You are on page 1of 45

11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras

Página 1

REVISÕES DA FÍSICA MODERNA, VOLUME 77, OUTUBRO DE 2005

Supercondutividade ímpar trio e fenômenos relacionados


em estruturas de supercondutor-ferromagneto

FS Bergeret
Departamento de Física Teórica da Matéria Condensada, Universidad Autónoma de
Madri, E-28049 Madri, Espanha
AF Volkov
Theoretische Physik III, Ruhr-Universität Bochum, D-44780 Bochum, Alemanha
e Instituto de Radioengenharia e Eletrônica da Academia Russa de Ciências,
125009 Moscou, Rússia
KB Efetov
Theoretische Physik III, Ruhr-Universität Bochum, D-44780 Bochum, Alemanha
e LD Landau Institute for Theoretical Physics, 119334 Moscou, Rússia
Publicado 28 de novembro de 2005

Esta revisão considera efeitos incomuns em estruturas ferromagnéticas supercondutoras, em


componente tripleto do condensado gerado nesses sistemas. Este componente é estranho na frequência
e até mesmo no momento, o que o torna insensível a impurezas não magnéticas. Se o campo de troca é
não homogêneo no sistema, o componente tripleto não é destruído nem mesmo por uma troca forte
campo e pode penetrar o ferromagneto em longas distâncias. Alguns outros efeitos considerados aqui e
causados pelo efeito de proximidade são o aumento da corrente de Josephson devido à presença do
ferromagneto, indução de um momento magnético em supercondutores, resultando num rastreio do
momento magnético, e formação de estruturas magnéticas periódicas devido à influência do
supercondutor. Finalmente, as previsões teóricas são comparadas com experimentos existentes.

CONTEÚDO partícula ferromagnética em um supercondutor 1359


D. Interação spin-orbital e seu efeito na proximidade
efeito 1361
I. Introdução 1321
1. O deslocamento do cavaleiro em supercondutores
1361
II. O efeito de proximidade 1325
2. Influência da interação spin-órbita em
A. Supercondutor - estruturas metálicas normais 1325
o efeito de proximidade de longo alcance1362
B. Estruturas de ferromagneto supercondutor com 3. Interação spin-órbita e inversa
magnetização uniforme 1328 efeito de proximidade 1362
1. Densidade de estados 1329 VI. Discussão dos resultados e do Outlook 1362
2. temperatura de transição 1330 Agradecimentos 1365
3. O efeito Josephson nas junções S / F / S 1332 Apêndice A: Equações Básicas 1365
III Supercondutividade de tripla ímpar em estruturas1334
S / F Apêndice B: Direção Futura da Pesquisa Experimental1368
A. Supercondutividade convencional e não convencional
1334 Lista de símbolos e abreviações 1369
B. Componente Tripleto Estranho Homogêneo Referências 1369
magnetização 1336
C. Triplet supercondutividade ímpar inomogênea
I. INTRODUÇÃO
magnetização 1337
1. estrutura trilayer F / S / F 1337 Embora a supercondutividade tenha sido descoberta por H.
2. Mural de domínio na interface S / F 1340 Kammerlingh Onnes quase um século atrás de 1911, o
3. Interfaces spin-active 1343 O interesse em estudar esse fenômeno está longe de
D. Efeito de proximidade de longo alcance 1343 ing. Grande interesse em supercondutividade nos últimos 15 anos
IV. Efeito de Josephson em sistemas S / F não homogêneo anos é em parte devido à descoberta da alta
Magnetização 1347 supercondutores de temperatura Bednorz e Müller,
A. Josephson acoplamento entre camadas S através do trio1986, que promete importantes aplicações tecnológicas
componente 1348 ções. É claro que questões como a origem de
B. Aprimoramento da atual corrente Josephson 1350 supercondutividade de temperatura crítica, efeitos de
V. Redução da magnetização devido a campos nal e impurezas em Temperatura alta
1352 supercondutores, etc., permanecerão campos de interesse para
Supercondutividade: Efeito de Proximidade Inversa
A. Estado criptoferromagnético 1353 Anos por vir.
B. Ferromagnetismo induzido em um supercondutor 1356 Devido a grande atenção ao supercondens de alta temperatura
C. Rotação do momento magnético de um ductors, o interesse em c supercondutores de baixa T tradicionais

0034-6861 / 2005/774/132153 / US $ 50,00 1321 © 2005 a sociedade física americana

Página 2
1322 Bergeret, Volkov e Efetov: Supercondutividade do trigêmeo estranho e fenômenos relacionados…
https://translate.googleusercontent.com/translate_f 1/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras

não foi tão alto. No entanto, esse campo também já é bem sabido que os elétrons com diferentes rotações
sofreu um tremendo desenvolvimento. Tecnologicamente pertencem a diferentes bandas de energia. A mudança de energia do
Os supercondutores tradicionais costumam ser mais fáceis de duas
manipular.
bandas podem ser consideradas como uma troca eficaz
tarde do que de alto T c cuprates. Um dos principais campo atuando no spin dos elétrons. O condensado
mentos da última década é a produção de alta qualidade de supercondutores convencionais é fortemente influenciado
contatos entre supercondutores e metais normais por este campo de troca dos ferromagnetos e geralmente
S / N , supercondutores e ferromagnetos S / F , e su- isso reduz drasticamente as correlações supercondutoras.
percondutores e isoladores S / I. Essas heteroestruturas A supressão das correlações supercondutoras
podem ser muito pequenas com tamanhos característicos de é uma consequência do princípio de Pauli. Na maioria dos
micrômetros. condutores, a função de onda dos pares de Cooper é sin-
Isso abriu um novo campo de pesquisa. O pequeno glet para que os elétrons de um par tenham spins opostos.
O tamanho dessas estruturas fornece a coerência de super- Em outras palavras, ambos os elétrons não podem estar no mesmo
realização de correlações longo do comprimento total da re N estado, o que aconteceria se eles tivessem o mesmo giro. E se
gion. O comprimento da penetração de condensado no o campo de troca do ferromagneto é suficientemente
A região N é restrita por processos de decoerência forte, ele tenta alinhar os spins dos elétrons de um
espalhamento elástico ou spin-flip. Em baixas temperaturas Par de cooper paralelo entre si, destruindo assim o
comprimento característico ao longo do qual esses processos de supercondutividade.
decoerência Em relação às interfaces S / F e ao
Os cortes podem ocorrer em alguns microns. Super- penetração do condensado no ferromagneto,
conduzindo efeitos coerentes em nanoestruturas S / N , tais estes efeitos significam que o condensado supercondutor
como oscilações de condutância em um campo magnético externo, decaem rapidamente na região ferromagnética. Uma estimativa
foram estudados intensivamente durante a última década, ver, mate
por leva à conclusão de que a razão entre o
exemplo, os artigos de revisão de Beenakker 1997 e profundidade de penetração de densato em ferromagnetos ao de
Lambert e Raimondi 1998. metais não magnéticos com alta concentração de impurezas
A interação entre um supercondutor S e um é da ordem de T c / h , onde h é a energia de troca
metal normal N em tipos mais simples de estruturas S / N parae T c é a temperatura crítica do supercondutor
Por exemplo, as bicamadas S / N estão sendo estudadas ea transição. A energia de troca em ferro convencional
física principal deste chamado efeito de proximidade é bem ímãs, como Fe ou Co é várias ordens de magnitude
descrito por de Gennes 1964 e Deutscher e de superior a T c e, portanto, a profundidade de penetração
Gennes 1969. Nestes trabalhos notou-se que não os ferromagnetos são muito menores que os da nor-
só o supercondutor muda as propriedades do metais mal.
o metal normal, mas o metal normal também tem um forte O estudo do efeito de proximidade em estruturas S / F
efeito sobre o supercondutor. Foi demonstrado que perto do começou há não muito tempo, mas já evoluiu para um
Interface S / N a supercondutividade é suprimida campo muito ativo de pesquisa para uma revisão, ver Izyumov et
o comprimento de correlação S , o que significa que a ordem al. 2002; Golubov et al. 2004; Lyuksyutov e Pokémon
parâmetro é reduzido na interface em comparação rovsky 2004; Buzdin 2005a. O efeito do apoio
com seu valor em massa longe da interface. No pressão de supercondutividade por ferromagnetismo é
mesmo tempo, o condensado supercondutor penetra claramente visto experimentalmente e corresponde ao
o metal normal sobre o comprimento N , que a baixa temperatura imagem simples da destruição do superconjunto singlete
as peraturas podem ser muito maiores que S. Devido ao ductivity pelo campo de troca como discutido acima.
etração do condensado no metal normal ao longo À primeira vista, parece que, devido ao forte
grandes distâncias, o efeito de Josephson é possível em S / N /da
S supercondutividade o efeito de proximidade em S / F
junções com espessuras das N regiões do pedido estruturas é menos interessante do que nos sistemas S / N.
de algumas centenas de nanômetros. Os efeitos de Josephson em No entanto, isto não é assim porque a física da proximidade
As junções S / N / S foram estudadas em muitos trabalhos e O efeito iminente nas estruturas S / F não é exaurido
Uma boa visão geral, experimental e teórica, é supressão da supercondutividade e novos interesses muito
dada por Kulik e Yanson 1970, Likharev 1979, e efeitos de entrada entram em jogo. Além disso, sob alguns
Barone e Paterno 1982. a supercondutividade não é necessariamente supérflua
A situação descrita acima é bem diferente se um pressionado pelos ferromagnetos porque a presença do
camada isolante I é colocada entre dois superconductos este último pode levar a um emparelhamento supercondutor tripla
ors. A espessura do isolador em estruturas S / I / S Bergeret et al. , 2001a; Kadigrobov et al. , 2001. Em alguns
não pode ser tão grande quanto a dos metais normais porque casos não só o ferromagnetismo tendem a destruir
funções de onda de elétrons decair no isolador em atômica a supercondutividade, mas a supercondutividade pode
distâncias Como conseqüência, a corrente de Josephson é também suprimem o ferromagnetismo Buzdin e Bula-
extremamente pequeno em estruturas S / I / S com um evskii, 1988; Bergeret et al. , 2000. Isso pode afetar “real”
ing camada. ferromagnetos fortes como ferro ou níquel com um Curie
Mas o que dizer das heterojunções S / F / S , onde F temperatura muito maior que a temperatura de transição
observa um metal ferromagnético? Em princípio, o elétron ture do supercondutor.
função de onda pode estender para o ferromagneto ao longo de um Ao todo, está se tornando cada vez mais evidente
distância bastante grande sem uma decadência considerável. Como-
estudos experimentais e teóricos recentes de que os

Rev. Mod. Phys. Vol. 77, n ° 4, outubro de 2005

Página 3
1323
Bergeret, Volkov e Efetov: Supercondutividade do trigêmeo estranho e fenômenos relacionados…

diminuição inesperada da resistência de um ferromagnético


riedade de efeitos não triviais em estruturas S / F consideravelmente
excede o que se poderia esperar antes. Levando fio ligado a um supercondutor quando a temperatura
em conta possíveis aplicações tecnológicas, existe foi abaixada abaixo de T c . Em ambos os experimentos forte
não admira que os sistemas S / F hoje em dia atraiam muitos foram utilizados ferromagnetos Ni e Co, respectivamente. 1
atenção. esperaria que a mudança da resistência devesse ser
Este artigo de revisão é dedicado ao estudo de novas muito pequeno devido à destruição da supercondutividade
“fenómenos” nas heterojunções S / F. Por '' pelos ferromagnetos. No entanto, a queda observada
otic '' queremos dizer fenômenos que não poderiam ser esperados foi de cerca de 10% e só pode ser explicado por um
a partir da simples imagem de um supercondutor em contato efeito de proximidade de alcance.
com um ferromagneto homogêneo. De fato, a maioria Isto levanta uma questão natural: como pode tal
os efeitos de teste devem ocorrer quando o campo de troca é gama de efeitos supercondutores ocorrem em um ferromagneto

https://translate.googleusercontent.com/translate_f 2/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
não homogêneo.
intrínseco para o Essas não-homogeneidades
material ferromagnético, como podem
o ser com
seçõesumsecundárias
campo de troca que oforte?
campoVeremos
de trocano subseqüente
não é homogêneo
paredes principais, ou surgem como resultado de manipula- um componente de longo alcance da atmosfera.
como estruturas multicamadas com diferentes pode ser induzido no ferromagneto. este
direções de magnetização, que também pode ser falado componente está em um estado triplet e pode penetrar o F
como um alinhamento não homogêneo do região em distâncias comparáveis a N , como no caso de
mentos. um metal normal.
Claro, estamos longe de dizer que não há nada Nós agora delineamos a estrutura da presente revisão. Dentro
interessante ser visto quando o campo de troca é homo- Sec. II discutimos os efeitos de proximidade em estruturas S / N
Gene. Embora seja verdade que neste caso a penetra- e estruturas S / F com uma magnetização homogênea.
profundidade do condensado supercondutor no Os principais resultados ilustrados foram apresentados em
ferromagneto é curto, o decaimento exponencial do outros comentários e nós discutimos-los, a fim de dar o
função densate em ferromagnets é acompanhada por leitor uma introdução aos trabalhos anteriores. Seção II pode
oscilações no espaço. Essas oscilações levam, por também ajuda na compreensão dos métodos de cálculo
amplas, às oscilações da crítica supercondutora usado nas seções subseqüentes. Já se pode ver de
temperatura T c e a corrente Josephson atual I c em essa discussão de que os ferromagnetos homogêneos
Estruturas S / F em função da espessura d F. Pré- tato com supercondutores levar a novos e interessantes
ditado por Buzdin e Kupriyanov 1990 e Radovic et física.
al. 1991, a observação desse comportamento oscilatório No entanto, as não-homogeneidades o fazem ainda mais.
foi relatado pela primeira vez por Jiang et al. 1995, sobre as estruturas
Revimos Gd / Nbvários efeitos diferentes que surgem no
abaixo
turas. Indicações do comportamento não monotônico de T c como situação não homogênea. Acontece que um não
uma funo de dF foi tamb relatada por Wong et al. 1986, alinhamento homogêneo do campo de troca leva a um
Strunk et al. 1994, Mercaldo et al. 1996, Mühge et al. estrutura de rotação complicada do con- trapaça supercondutor
1996, Obiand et al. 1999 e Velez et al. 1999. densar. Como resultado, não só o compo
Entretanto, em outros experimentos, a dependência de T c existe um condensado do condensado, mas também um tripleto
em d F era monótono. Por exemplo, no trabalho de Bour- todas as projeções possíveis do spin total do Cooper
geois e Dynes 2002 a temperatura crítica do par S z = 0, ± 1. Em contraste com o componente singleto, o
bicamada Pb / Ni diminuiu aumentando a espessura da camadaosFgiros dos elétrons no tripleto um com S z = ± 1 são
ness d F de uma maneira monótona. Nos experimentos de paralelos um ao outro. O condensado Gor'kov funciona
Mühge et al. 1998 em estruturas Fe / Nb / Fe e por ção f tr do estado triplete é uma função ímpar do Mat-
Aarts et al. 1997 em sistemas V / Fe tanto um monotônico freqüência subara. 1 A singuleto parte f sng é, como de costume, uma
e comportamento não monotônico de T c foi observado. este mesmo função de mas muda de sinal quando intercambia
comportamento diferente foi atribuído às mudanças do os índices de spin. É por isso que a anticomutação
transmitância da interface S / F. Um abrangente relações para as funções iguais em tempo f tr t, t e f sng t, t
análise tendo em conta a qualidade da amostra foi permanecer válido; em particular, f tr t , t = 0 e f sng t , t 0
feita para diferentes materiais por Chien e Reich 1999. Portanto, a supercondutividade em estruturas S / F pode ser
Resultados mais convincentes foram encontrados medindo muito incomum. Juntamente com a parte usual de singles BCS,
Corrente crítica de Josephson em uma junção S / F / S. Devidotambém
a pode conter a parte tripla que é simétrica em
o comportamento oscilatório da condensação supercondutora espaço de momentum no caso difusivo e ímpar em fre-
Na região F , a corrente crítica de Josephson deveria quência. Ambos os componentes são insensíveis à dispersão
mudar seu sinal em um entroncamento S / junção.
F / S este por impurezas não magnéticas e, portanto, sobreviver em S / F
fenômeno, previsto há muito tempo por Bulaevskii et al. estruturas, mesmo que o caminho livre médio l seja curto. Quando
1977 só recentemente foi confirmado experimentalmente gerado, o componente tripleto não é destruído pelo
Kontos et al. 2001, 2002; Ryazanov et al. 2001; Blum et campo de troca e pode penetrar no ferromagneto sobre
al. 2002; Bauer et al. 2004; Sellier et al. 2004
Experimentos em propriedades de transporte de estruturas S / F
também foram realizados nos últimos anos. Por exemplo, 1 Supercondutividade causada pelo trio ímpar em condensado
oud et al. 1998 e Petrashov et al. 1999 observou uma é chamado aqui de supercondutividade ímpar.

Rev. Mod. Phys. Vol. 77, n ° 4, outubro de 2005

Página 4
1324 Bergeret, Volkov e Efetov: Supercondutividade do trigêmeo estranho e fenômenos relacionados…

longas distâncias da ordem de N = D F / 2 T. a supressão do primeiro pelo segundo, o oposto


Na sec. III analisamos propriedades deste novo tipo de também é possível e isso é discutido na Sec. V. Claro,
supercondutividade que pode surgir em estruturas S / F. Nós um ferromagnetismo fraco deve ser fortemente afetado por
enfatizar que esta supercondutividade tripla é generalizada a supercondutividade e esta situação é realizada em
pelo campo de câmbio e, na ausência do supercondutores magnéticos Bulaevskii et al. , 1985. Menos
campo, um teria o emparelhamento singlete convencional. trivial é que os ferromagnetos fortes convencionais
As multicamadas supercondutor-ferromagneto são Os sistemas S / F também podem ser consideravelmente afetados pela
objeto muito interessante e natural para observação de supercondutividade. Isso pode acontecer desde que o
Efeitos de Josephson. A espessura de ambos os supercondens A profundidade da camada ferromagnética é suficientemente pequena.
dutor e camadas ferromagnéticas, bem como a transparência pode
Entãoser
isso
energeticamente mais lucrativo para forçar a
da interface, pode ser variado experimentalmente. este momento nético para girar no espaço do que para destruir
torna possível um estudo detalhado de muitos percondutividade. Se o período de tais oscilações é
quantidades físicas. Como já mencionamos, um interesse menor que o tamanho dos pares de Cooper S , a influência
Manifestação do papel desempenhado pelo ferromagnetismo a força do magnetismo no supercondutor torna-se
é a possibilidade de um junção. muito pequena e a supercondutividade é preservada. Dentro
No entanto, este não é o único efeito interessante e camadas espessas como uma estrutura oscilante de criptoferro-
várias novas propostas foram recentemente propostas estado magnético custaria muita energia e a destruição
Cally. Eles ainda não foram confirmados inequivocamente A supercondutividade é mais favorável. Resultados
experimentalmente, mas não há dúvida de que a experiência de vários experimentos foram interpretados desta forma
em breve serão realizados. Na sec. IV nós discutimos novos Mühge et al. 1998; Garifullin et al. , 2002.
Os efeitos de Josephson em estruturas S / F de múltiplas camadasOutro fenômeno inesperado, a saber, o
em conta uma possível mudança da di- efeito de proximidade do verso, também é apresentado na Sec. V. It
rection em camadas ferromagnéticas. Discutimos uma situaçãoAcontece
simples que não só a condensação supercondutora
quando as direções dos momentos magnéticos em um sate penetre os ferromagnetos, mas também um

https://translate.googleusercontent.com/translate_f 3/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
A estrutura
aluguel flui SF / I / FS
através é colinear
de um isolantee aI, estrutura de Josephson
mas não através do mento
tato pode
com umser induzido emEste
ferromagneto. um efeito
supercondutor
tem umaque estámuito
forma em simples
ferromagnetos. Ingenuamente, seria de se esperar que a explicação. Existe uma probabilidade de que algumas das
A presença de ferromagnetos leva a uma redução do valor de tron de pares de Cooper entram no ferromagneto e sua rotação
a corrente crítica. No entanto, a situação é mais tende a ser paralelo ao momento magnético. No
esting. A corrente crítica é maior quando o magnético mesmo tempo, o spin do segundo elétron do Cooper
momentos das camadas F são antiparalelas do que quando par deve ser oposto ao primeiro, o par de singlete
são paralelos. Além disso, verifica-se que a crise crítica ou o trigêmeo com S z = 0 é assumido. Como um resultado,
aluguel para a configuração antiparalela é ainda maior do que um momento magnético com direção oposta à mag-
aquele na ausência de qualquer camada ferromagnética. Dentro momento nético do ferromagneto é induzido no
Em outras palavras, o ferromagnetismo pode melhorar a percondutor em distâncias da supercondutora co-
atual Bergeret et al. , 2001b. comprimento de herença S.
Outra configuração é sugerida para observar as viagens ímpares. Em princípio, o momento magnético total pode ser com-
deixe supercondutividade discutida na Sec. III Aqui o completamente rastreado pelo supercondutor. Formalmente, o
o aluguel deve fluir através das camadas ferromagnéticas. Usu- aparência do momento magnético no superconjunto
Por outro lado, pode-se pensar que a corrente crítica seria ductor é devido ao componente tripleto do condensado
decaimento muito rápido com o aumento da espessura do fer- que é induzido no ferromagnético F e penetra em
camada romagnética. No entanto, outro efeito é possível. o supercondutor S. É importante notar que isso
Alterando a direção mútua da ferida adicional efeito deve desaparecer se a supercondutividade estiver
camadas romagnéticas pode-se gerar o trio ímpar por exemplo, aquecimento, e isso dá a possibilidade de
componente do condensado supercondutor. Este componente uma observação do efeito. Além do Meissner
nente pode penetrar na camada ferromagnética como se fosse efeito, um este é mais um mecanismo de campo magnético
metal normal, levando a grandes valores da corrente crítica triagem por supercondutividade. Em contraste com o Meiss-
aluguel. efeito no qual o rastreio é devido ao orbital
Tais estruturas podem ser úteis para detectar e manipular movimento de elétrons, este é um tipo de triagem de spin.
nipulando o componente tripleto do condensado em Finalmente, na Sec. VI discutimos os resultados apresentados e
experimentos. Em particular, veremos que em alguns S / F tente antecipar as direções futuras da pesquisa. Ap-
estruturas do tipo de supercondutividade é diferente em O pêndulo A contém informações sobre a aplicação quasiclássica
direções diferentes. Na direção longitudinal na teoria da supercondutividade.
supercondutividade plana é causada principalmente pelo pecadoDevemos mencionar que vários artigos de revisão sobre
componente glet, enquanto que na direção transversal Tópicos relacionados a S / F foram publicados recentemente
componente tripleto contribui principalmente para a supercon-Izyumov et al. 2002; Golubov et al. 2004; Lyuksyutov
ductivity. Também discutimos as possibilidades de um experimento e Pokrovsky, 2004; Buzdin, 2005a. Nestes comentários
tal observação do componente tripleto. várias propriedades de estruturas S / F são discutidas para o
Embora o efeito mais pronunciado das interações caso de uma magnetização homogênea. Na revisão por
entre supercondutividade e ferromagnetismo é Lyuksyutov e Pokrovsky 2004 o foco principal está em

Rev. Mod. Phys. Vol. 77, n ° 4, outubro de 2005

Página 5
1325
Bergeret, Volkov e Efetov: Supercondutividade do trigêmeo estranho e fenômenos relacionados…

FIG. 1. S / N bicamada.

os efeitos causados por uma interação magnética entre o


ferromagneto e supercondutor, por exemplo, um FIG. 2. Densidade de tunelamento dos estados medidos a 60 mK no
criação espontânea de vórtices no supercondutor devido Superfície de Au de diferentes amostras de bicamadas Nb / Au com diferentes
à interação magnética entre o magnetismo magnético Espessura de Au d N. Adaptado de Gupta et al. 2004
vórtices ea magnetização no ferromagnetismo
líquido. Em contraste com essas revisões, nos concentramos na
cussão do componente tripleto com todos os possíveis com o aumento da espessura da camada N. Esse comportamento pode ser
interpretado como a quebra de alguns pares de Cooper devido
projeções do momento magnético S z = 0, surgindo ± 1
apenas no caso de uma magnetização não homogênea. Dentroà penetração de um dos elétrons dos pares
Além disso, discutimos o efeito de proximidade inversa, isto é,no metal normal, onde eles não estão mais em
tracionado pelos outros elétrons dos pares.
a influência da supercondutividade na magnetização
Ao mesmo tempo, ao penetrar no normal
M de estruturas S / F e outros efeitos. Desde a experiência
metal os pares Cooper induzem a correlação supercondutora
estudo mental dos efeitos de proximidade em estruturas S / F ainda
ções. Por exemplo, a influência da supercondutividade
permanece em sua infância, esperamos que esta revisão ajude
nas propriedades físicas do N metal manifesta-se
na compreensão das condições sob as quais se pode na supressão da densidade de estados. Experimentos
observar o novo tipo de supercondutividade e outras
determinar a densidade dos estados das bicamadas S / N
efeitos terestes e estimulará experimentalmente
a ajuda da espectroscopia de tunelamento foi realizada
atividade nesta área quente.
muitos anos atrás Adkins e Kington, 1969; Toplicar e
Finnemore, 1977, enquanto densidade espacialmente resolvida
II. O EFEITO DA PROXIMIDADE os estados foram posteriormente medidos por Guéron et al. 1996, An-
thore et al. 2003 e Gupta et al. 2004 ver Fig. 2.
Nesta seção, revisaremos os recursos básicos do A maneira mais simples de descrever o efeito de proximidade é
efeito de proximidade em diferentes heteroestruturas. O primeiro
use a equação de Ginzburg-Landau para os parâmetros de ordem
parte é dedicada a estruturas supercondutoras de metal normaléter Ginzburg e Landau, 1950. Essa equação é
enquanto na segunda parte os supercondutores em válido se a temperatura estiver próxima da temperatura crítica
https://translate.googleusercontent.com/translate_f 4/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
Tato com ferromagnetos homogêneos são considerados. da transição supercondutora T c . Neste caso, todos
quantidades podem ser expandidas no pequeno parâmetro / T c
A. Supercondutor - estruturas metálicas normais e variações lentas do parâmetro de ordem no espaço.
Usando a equação de Ginzburg-Landau escrita como
Se um supercondutor entrar em contato com um 2 r
material não supercondutor, as propriedades físicas de 2 2 = 0,
GL r 2 + r sgn T c , N , S - T - 2 r / 0 2,1
ambos os materiais podem mudar. Esse fenômeno, chamado
o efeito de proximidade , tem sido estudado por muitos anos. pode-se descrever a distribuição espacial da ordem
Tanto o experimento quanto a teoria mostram que as propriedades de
parâmetro em qualquer estrutura N / S. Aqui GL é a coerência
camadas supercondutoras em contato com o isolante I ma-
comprimento de trabalho nas regiões N e S a temperaturas próximas
os materiais permanecem quase inalterados. Por exemplo, para
às temperaturas críticas T cN , S. No limite difusivo
filmes percondutores evaporados em substratos de vidro, o
esse comprimento é igual a
temperatura crítica T c está muito próxima do valor em massa.
No entanto, as propriedades físicas de ambos os metais de um
N / S supercondutor de metal normal , veja a Fig. 1 GL = D N , S / 8 T - T cN , S , 2,2

junção com uma condutância de interface N / S alta pode onde D N , S é o coeficiente de difusão em N e S
mudar drasticamente. regiões. A quantidade 0 é o valor em massa do pedido
O estudo do efeito de proximidade remonta ao início parâmetro no supercondutor S. Desaparece quando T
ning dos anos 1960 e foi revisado em muitas publicações atinge a temperatura de transição T c .
ver, por exemplo, de Gennes 1964 e Deutscher and de Deve-se notar, no entanto, que a região aplicável
Gennes 1969. Verificou-se que a temperatura crítica da Eq. 2.1 para a descrição dos contatos S / N é
tura do supercondutor em um sistema S / N diminuiu bastante restrito. Claro, a temperatura deve ser

Rev. Mod. Phys. Vol. 77, n ° 4, outubro de 2005

Página 6
1326 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

perto da temperatura de transição T c, mas isso não é equação. O DOS é uma quantidade muito importante que pode
suficiente. A equação de Ginzburg-Landau descreve ser medido experimentalmente e, ao mesmo tempo, pode
variações dos parâmetros de ordem corretamente somente se ser calculado sem dificuldades.
são lentos nas escalas v F / T c para o caso limpo ou Consideramos a estrutura S / N mostrada na Fig. 1 e
D N , S / T c no caso “sujo” difusivo. Isso pode ser suponha que o sistema é difusivo, ou seja, a condição
alcançado se o metal normal for um mate supercondutor 1 é assumido como sendo cumprido, onde está a
rial tirado a uma temperatura superior à sua temperatura de tum tempo de relaxamento e é a energia e que o
perature T cN e as temperaturas de transição T cS e a transparência do S / N é baixa o suficiente. Neste caso, o
função de condensado verde f
T cN estão próximos uns dos outros. Se esta condição não for satisfatória = dt ft - t exp i t
Por exemplo, T cN = 0, deve-se usar mais complicado - t é pequeno na região de N e na equação de Usadel
equações, mesmo em temperaturas próximas a T cS , como mostramospode ser linearizado, consulte o Apêndice A.
abaixo. Assumindo que a fronteira entre o superconjunto
Segue da Eq. 2.1 que na região S , longe de O metal dutor e normal é plano e escolhe a coordenação
a interface N / S , o parâmetro order r é igual ao nate x perpendicular ao limite reduzimos o uso de
valor em massa 0 , enquanto na região N r decai equação adel na região N para a forma
ponencialmente a zero com comprimento N.
O parâmetro order r está relacionado ao condensado D N 2 f / x 2 + 2 i f = 0, 2,5
função ou função de Gor'kov onde D N = v F l / 3 é o coeficiente de difusão clássico.
A solução desta equação pode ser encontrada facilmente e
ft , t = T ↓t 2,3 escrito como

através da equação de auto-consistência f = f 0 exp - x - 2 i / D N , 2,6

2,4 onde f 0 é uma constante determinada a partir do limite


N , S t = N , S ft , t , condições.
onde N , S é a constante de acoplamento elétron-elétron Nós vemos que a solução para a função condensada f
levando à formação das condições supercondutoras decai exponencialmente na região N a distâncias
saciar. versely proporcional a . Em muitos casos, o principal
A equação 2.1 descreve realmente um contato entre tributação a quantidades físicas vem das energias
dois supercondutores com diferentes temperaturas críticas na ordem da temperatura, T. Isso significa que
T cN , S , quando a temperatura está entre T cS e T cN . Dentro o condensado supercondutor penetra na região N
o caso de um metal normal real a constante de acoplamento N sobre distâncias na ordem de N = D N / 2 T. Embaixo
é igual a zero e, portanto, N = 0. No entanto, isso temperatures this distance becomes very large, and if the
não implica que o metal normal não possua super- thickness of the normal-metal layer is smaller than the
conduzindo propriedades neste caso. O ponto é que inelastic relaxation length, the condensate spreads
muitas grandezas físicas importantes estão relacionadas throughout the entire N region.
o parâmetro de pedido mas para a função condensada f , In order to calculate the DOS it is necessary to know
Eq. 2.3. Por exemplo, o condensado não-secativo the normal Green's function g which is related to the
corrente j S é expressa em termos da função f mas não condensate function f via the normalization condition
do . Se o contato entre as regiões N e S for bom, see Appendix A
o condensado penetra no metal normal levando a um
valor finito de j S 0 nesta região. g 2 − f 2 = 1. 2,7
No caso geral de um N arbitrário é conveniente Equations 2.5 and 2.7 are written for the retarded
para descrever a penetração do condensado Cooper Green's function f = f R , see Appendix A. They are also
pares na região N no limite difusivo com o valid for the advanced Green's functions provided + i 0
Usadel equação Usadel, 1970, que é válido para todos is replaced by − i 0. The normalized density of states
temperaturas e para distâncias superiores à média livre we normalize the DOS to the DOS of noninteracting
caminho l . Esta equação determina o quasiclassical electrons is given by
As funções de Green incluem o Apêndice A, que pode ser

https://translate.googleusercontent.com/translate_f 5/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
veniently
que usado em de
o comprimento problemas envolvendo escalas de comprimento maiores
onda Fermi = Re g . 2,8
F e energias muito
menor que a energia de Fermi. Alternativamente, pode-se As the condensate function f is small, a correction para
tente encontrar uma solução exata do normal e anômala the DOS due to the proximity effect is also small. No
funções do elétron Green para as equações de Gor'kov, main approximation the DOS is very close to its value
mas isso é, na maioria dos casos, uma tarefa difícil. in the absence of the superconductor, 1. Corrections
A fim de ilustrar a conveniência de usar a to the DOS are determined by the condensate func-
método clássico agora calculamos a mudança do tion f . Da Eq. 2.7 one obtains
densidade de tunelamento de estados DOS no metal normal
devido ao efeito de proximidade com a ajuda do Usadel f 2 /2.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 7
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1327

Now we consider another case when the function f is equations near the interface and compare the
not small and the correction is on the order of unity. asymptotic behavior of this solution at large distances
Then the linearized Eq. 2.5 may no longer be used and with the solutions of the Usadel equation. This proce-
we should write a more general one. For a S / N system dure is equivalent to solving the quasiclassical equations
the general equation can be written as see Appendix A with some boundary conditions. These conditions were
derived by Zaitsev 1984 and Kuprianov and Lukichev
− iD S , N gg / x S , N / x + ˆ 3 , g S , N + ˆ 1988 see also Appendix A where these conditions are
S , g S , N = 0.
2,9 discutido em mais detalhes. For the present case they can
be written as
This nonlinear equation contains the quasiclassical ma-
trix Green's function g . Both normal g and anomalous 2 S , N gg / x S , N = g S , g N x =0 , 2,13
Green's functions f enter as elements of this matrix where S , N = R b S , N , R b , measured in units cm 2 , is the
through the following relation the phase in the super- S / N interface resistance per unit area in the normal
conductor is set to zero: state, and S , N are the conductivities of the S and N
g N = g N ˆ 3 + f N iˆ 2 , 2,10 metals in the normal state.
We assume that the thickness of the normal metal d N
where i , i =1,2,3 are Pauli matrices and A , B = AB is smaller than the characteristic penetration length
− BA is the commutator for any matrices A and B . N = D N / for a given energy , that is 2 D N / d2N
We consider a flat S / N interface normal to the x axis. = E Th . Then the functions g and f remain almost constant
The normal metal occupies the region 0 x d N . We as- over the thickness of the metal, and to find them, one
sume that in the normal metal N there is no electron- can average the Usadel equation over the thickness. Dentro
electron interaction other words, we assume that the thickness d N of the N
N =0, see Eq. 2.4 so that in this
region the superconducting order parameter vanishes, layer satisfies the inequality
ˆ
N =0. In the superconductor the matrix S has the dN DN/, bN 2.14
ˆ
estrutura S = iˆ 2 .
At large distances from the S / N interface the Green’s bN is a characteristic energy in the DOS of the N layer
and average Eq. 2.9 over the thickness d N considering
functions g S of the superconductor do not depend on
coordinates and the first term in Eq. 2.9 can be ne- g N as a constant in the second term of this equation.
glected. Then we obtain a simpler equation Using the boundary condition, Eq. 2.13, the first term
in Eq. 2.9 can be replaced after integration by the com-
ˆ 3 , g S + iˆ 2 , g S = 0. 2,11 mutator g S , g N x =0 . At x = d N the product gg / x N is
zero because the barrier resistance R b d N is infinite the
The solution for this equation satisfying the normaliza- current cannot flow into the vacuum. Finally, we obtain
tion condition 2.7 is Zaitsev, 1990
g BCS = / , f BCS = / , 2,12 + i b g S 0 ˆ 3 , g N + bN if S 0 iˆ 2 , g N = 0, 2,15
Onde = 2 − 2 . Equation 2.12 is just the BCS solu- where bN = D N /2 N d N is a new characteristic energy
tion for a bulk superconductor. that is determined by the S / N interface resistance R b .
In order to find the matrix gx in both the S and N This equation looks similar to Eq. 2.11 after making
regions, Eq. 2.9 should be complemented by boundary the replacement g S →g N . The solution is similar to the
conditions and this is a nontrivial problem. Começando de solution 2.12,
the initial Hamiltonian Hˆ tot , Eq. 2.22, one does not
need boundary conditions at the interface between the g N = ˜ /̃ , f N = ˜ bN / ˜ , 2,16
superconductor and the ferromagnet because the inter- ˜ = ˜ 2 − ˜ 2bN , ˜ bN = bN if S 0.
face can be described by introducing a proper potential Onde ˜ = + i bN g S 0,
in the Hamiltonian. In this case the self-consistent Therefore the Green's functions in the N layer g N and f N
Gor'kov equations can be derived. are determined by the Green's functions on the S side of
However, in deriving the Usadel equation, Eq. A18, the S / N interface g S 0 and f S 0. In order to find the
we have simplified the initial Gor'kov equations using values of g S 0 and f S 0, one has to solve Eq. 2.9 on
the quasiclassical approximation. Possible spatial varia- the superconducting side x 0. However, provided the
tion of the interface potential on a very small scale, due inequality
to the roughness of the interface, cannot be included in
the quasiclassical equations. Nevertheless, this problem N/S=N/S 1 2,17
is avoided by deriving the quasiclassical equations at dis- is fulfilled, one can easily show that in the main approxi-
tances from the interface exceeding the wavelength. Dentro mation the solution in the S region coincides with the
the diffusive case one should go away from the interface solution for bulk superconductors 2.12. If the transpar-
to distances larger than the mean free path l . A fim de
match the solutions in the superconducting and nonsu-
perconducting regions one should solve exactly the 2 The quantity E Th = D N / d2Nis the Thouless energy.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

https://translate.googleusercontent.com/translate_f 6/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras

Página 8
1328 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

ency of the S / N interface is not high, bN , the char- DOS has sharp peaks at = n for a recent review see
acteristic energies bN are much smaller than e Deutscher 2005 . If E Th is much larger than ,a
the functions g S 0 and f S 0 are equal to g S 0 DOS is zero at =0 and increases with increasing
g BCS 0 / i , f S 0 f BCS 0 1/ i . For these energies the energy no gap. However, this is true only for such
the functions g N and f N have the same form as the BCS a simple geometry. For samples of more complicated
functions g BCS and f BCS 2.12 with the replacement shapes the behavior of the DOS depends on
→ bN , whether the electron dynamics in the N region is chaotic
or integrable Melsen et al. , 1996; Beenakker, 1997; Lod-
gN= ,fN= bN , 2,18 der and Nazarov, 1998; Pilgram et al. 2000; Taras-
2
2 − bN 2
2 − bN Semchuk and Altland, 2001.
Finally, it was shown by Altland et al. 2000 and Os-
where bN = D N /2 R b N d N . The energy bN can be repre-
trovsky et al. 2001 that mesoscopic fluctuations smear
sented in another form,
out the singularity in the DOS at = bN and the DOS
2 RQ vF vF Tb in the diffusive limit is finite, although small, for
bN = 2 R b k2F d N = d N 4 , 2,19
bN . The minigap discussed above has been observed
on a Nb/Si bilayer system and on a Pb/Ag granular sys-
where R Q =/ e 2 is the resistance quantum, v F and k F are tem by Heslinga et al. 1994 and Kouh and Valles
the Fermi velocity and wave vector. When obtaining the 2003, respectively.
latter expression, we used a relation between the barrier From this analysis we see that the proximity effect
resistance R b and an effective coefficient of transmission changes the DOS of the normal metal which acquires
T b through the S / N interface Zaitsev, 1984; Kuprianov superconducting properties. In the next section we shall
and Lukichev, 1988: R b n =2/3 l / T b , where l = v F é focus our attention on the case in which the normal
the mean free path, T b = T cos /1− T , éo metal is a ferromagnet. We shall see that new interesting
angle between the momentum of an incoming electron physics will arise from the mutual interaction of super-
and the vector normal to the S / N interface, and T é conductivity and magnetism.
the angle-dependent transmission coefficient. The angle
brackets mean an averaging over .
An important result follows from Eq. 2,18. The DOS B. Superconductor-ferromagnet structures with a uniform
is zero at magnetization
bN , ie, bN is a minigap in the excitation
spectrum McMillan, 1968. Remarkably, in the limit In this section we consider the proximity effect be-
bN the value of bN does not depend on , but is tween a superconductor S and a ferromagnet F . We as-
determined by the interface transparency or, in other sume that the ferromagnet is a metal and has a conduc-
words, by the interface resistance R b . A aparência de tion band. In addition, there is an exchange field due to
the minigap is related to Andreev reflections Andreev, spins of electrons of other bands.
1964. As has already been mentioned, the effective ex-
Equation 2.19 for the minigap is valid if the inequali- change field acts on spins of the conduction electrons in
ties 2.14 and bN are fulfilled. Both inequalities can the ferromagnet, and an additional term Hˆ ex describing
be written as this action appears in the total Hamiltonian for more
D N // d b dN db, 2,20 details see Appendix A
where d b =2 R b N is a characteristic length. No caso de Hˆ tot = Hˆ + Hˆ ex , 2,22
a small interface resistance R b or a large thickness of the
N layer, that is, if the condition D N /, d b d N is ful-
filled, the value of the minigap in the N layer is given by Hˆ ex = − d 3 r + rhr rdr, 2,23
Golubov and Kupriyanov, 1996
where + are creation and destruction operators, h is
DN the exchange field, are Pauli matrices, and , are
bN = c 1d 2N , 2.21
spin indices. The Hamiltonian Hˆ stands for a nonmag-
where c 1 is a factor of the order 1. This result has been netic part of the Hamiltonian. It includes the kinetic en-
obtained from a numerical solution of the Usadel equa- ergy, impurities, external potentials, etc., and is sufficient
ção. The DOS for the case of arbitrary thickness d N and to describe all properties of the system in the absence of
interface transparency was calculated by Pilgram et al. the exchange field.
2000. The energy of the spin-up electrons differs from the
The situation changes in the clean limit. Let us con- energy of the spin-down electrons by the Zeeman en-
sider, for example, a normal slab of thickness d N in con- ergy 2 h . Due to the presence of the term Hˆ ex describing
tact with an infinite superconductor. If the Thouless en- the exchange interaction, all functions, including the
ergy E Th = v F / d N is less than , then discrete energy condensate Green's function f , are generally speaking
levels n appear Saint-James, 1964 in the N region due nontrivial matrices in the spin space with nonzero diag-
to Andreev reflections Andreev, 1964. As a result, the onal and off-diagonal elements.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 9
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1329

The situation is simpler if the direction of the ex- sufficient to consider some limiting cases. Isto será
change field does not depend on coordinates. Nisso done in the present section leaving the general equa-
case, choosing the z axis along the direction of h one can tions for Appendix A.
consider electrons with spin up and down separately. Dentro In the case of a weak proximity effect, the condensate
this section we concentrate on this case. This can help function f is small outside the S region. We consider
the reader understand several interesting effects and get again the diffusive limit. Then, the general Eq. A18
an intuition about what one can expect from the pres- can be linearized and one obtains an equation for the
ence of the exchange field. The results of this section will
matrix f similar to Eq. 2.5 but containing an extra term
https://translate.googleusercontent.com/translate_f 7/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
also help in understanding which
superconductor-ferromagnet effectscan
structures in be
theconsid- due to the exchange field h ,
ered as rather usual and what kind of behavior is exotic.
We shall see that exotic phenomena occur in cases when D F 2 f F / x 2 + 2 i ˆ 0 + hˆ 3 f F = 0. 2,26
the exchange field is not homogeneous and therefore The subscript F stands for the F region.
postpone their discussion until the next sections. In the absence of the exchange field h , Eq. 2.26 re-
If the exchange field h is homogeneous, the matrix f duces to Eq. 2.5. It is important to emphasize that Eq.
describing the condensate f is diagonal and can be rep- 2.26 is valid for a homogeneous h only. Any variation
resented in the form of h in space makes the equation much more compli-
cated.
f=f3ˆ3+f0ˆ0, 2,24 Equation 2.26 should be complemented by boundary
conditions which take the form see Appendix A
where f 3 is the amplitude of the singlet component and
f 0 is the amplitude of the triplet component with zero
projection of the magnetic moment of Cooper pairs on F fF/x=−fS, 2,27
the z axis S z =0. Note that in the case of a S / N struc- where F = R b F , R b is the boundary resistance per unit
ture, the condensate function has a singlet structure
area, F is the conductivity of the F region, and f F , S are
only, ie, it is proportional to ˆ 3 . A presença do the condensate matrix functions in the F and S regions.
exchange field leads to the appearance of the triplet
Since we assume a weak proximity effect, a deviation of
term proportional to ˆ 0 .
The amplitudes of the singlet and triplet components the f S from its BCS value f BCS = ˆ 3 f BCS is small. Lá-
are related to the correlation functions as follows fore on the right-hand side of Eq. 2.27 one can write
Leggett, 1975; Vollhardt and Wölfle, 1990: f S ˆ 3 f BCS , where f BCS is defined in Eq. 2,12. No
ferromagnet vacuum interface the boundary condition is
f3t ↑ t ↓ 0− ↓ t ↑ 0, given by the usual expression x f F =0, which follows
f0t 2,25 from the condition R b → .
↑ t ↓ 0+ ↓ t ↑ 0. Using Eq. 2.27, one can easily solve Eq. 2.26. Nós
One can see that a permutation of spins does not change assume, as in the previous section, that the normal metal
the function f 3 0, whereas such a permutation leads to a ferromagnet is in contact with the superconductor at
change of the sign of f 0 0. This means that the ampli- x =0 x is the coordinate perpendicular to the interface.
tude of the triplet component taken at equal times is The other boundary of the ferromagnet is located at x
zero in agreement with the Pauli exclusion principle. = d F and the space at xd F is empty.
Later we shall see that in the case of a nonhomogeneous The proper solution for the diagonal matrix elements
magnetization all triplet components including f ± f 1122 can be written as
↑ t ↑ 0 and ↓ t ↓ 0 differ from zero.
Once one determines the condensate function, Eq. f BCS cosh x − d
2.24, one is able to determine physical quantities such ± sinh d 0
± F
x dF,
f±x= ±F
± F
2.28
as DOS, the critical temperature T c , or the Josephson 0, x dF.
critical current through a S / F / S junction.
The next paragraphs are devoted to a discussion of Aqui ± = −2 i ± h / D F is a characteristic wave vector
these physical properties in F / S systems with homoge- that determines the inverse penetration depth of the
neous magnetization. condensate functions f 0,3 into the ferromagnet.
Usually, the exchange energy h is much larger than
the energy max, T . This means that the conden-
1. Density of states sate penetration depth F = D F / h is much shorter than
In this section we discuss the difference between the the penetration depth into a normal nonmagnetic
DOS in S / N and S / F structures. General equations for metal N . The strong suppression of the condensate in
the quasiclassical Green's functions describing the sys- the ferromagnet is caused by the exchange interaction
tem can be written but they are rather complicated see that tries to align the spins of electrons parallel to the
Appendix A. In order to simplify the problem and at magnetization. This effect destroys the Cooper pairs
the same time give the basic idea about the effects, it is with zero total magnetic moment.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 10
1330 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

N / S contacts but for finite energies . In the F / S con-


tacts the energy is shifted, → ± h time-reversal sym-
metry breaking and this leads to a nonmonotonic de-
pendence of F on the thickness d F even at zero energy.
On the other hand, nonoscillatory behavior of the DOS
has been found recently in experiments on
Nb/CoFe bilayers Reymond et al. , 2000. The discrep-
ancy between existing theory and experimental data
may be due to the small thicknesses of the ferromagnetic
layer 0.5 d F 2.5 nm which is comparable to the
Fermi wavelength F 0.3 nm. Strictly speaking, in this
case the Usadel equation cannot be applied.
The DOS in F / S structures was studied theoretically
FIG. 3. Calculated change of the local density of states for a in many papers. Halterman and Valls 2002b studied the
S / F bilayer at the outer F interface. The solid line corresponds
to a F thickness d F =0.5 0 , where 0 = D F /, while the dashed DOS variation numerically for ballistic F / S structures.
one corresponds to d F =0.8 0 . The latter curve is multiplied byThe DOS in quasiballistic F / S structures was investi-
a factor of 10. gated by Baladie and Buzdin 2001, Zareyan et al.
2001, and Bergeret, Volkov, and Efetov 2002 and for
dirty F / S structures by Fazio and Lucheroni 1999 and
It is worth mentioning that the condensate function f ± Buzdin 2000. The subgap in a dirty S / F / N structure
experiences oscillations in space. Indeed, for a thick F was investigated in a recent publication by Golubov et

https://translate.googleusercontent.com/translate_f 8/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
layer d F F we obtain from Eq. 2.28 al. 2005.
It is interesting to note that in the ballistic case h
f±=± exp− x / F cos x / F ± i sin x / F , 1, is the momentum relaxation time, the DOS in the
EF±F F layer is constant in the main approximation in the
2,29 parameter 1/ h while in the diffusive case h 1 it ex-
periences the damped oscillations. The reason for the
where E = 2 − 2 , F ± = ± at =0. The damped os- constant DOS in the ballistic case is that both parts of f ,
cillations of f ± lead to many interesting effects and, in the symmetric and antisymmetric in momentum space,
particular, to a nonmonotonic dependence of the critical contribute to the DOS. Each of them oscillates in space.
temperature on the thickness d F of a F / S bilayer, which However, while in the diffusive case the antisymmetric
will be discussed in the next section. part is small, in the ballistic case the contributions of
In order to calculate the DOS we have to use the both parts to the DOS are equal to each other, but op-
normalization condition, Eq. 2.7, which is also valid for posite in sign, thus compensating each other.
the matrix elements f ± and g ± . Thus for g ± we obtain Finally, we would like to emphasize that both the sin-
g ± = 1+ f 2± , which can be written for small f ± as g ± 1 glet and triplet components contribute to the DOS. Como
+ f2± /2. Then the correction to the normalized DOS in seen from Eq. 2.30, the changes of the DOS can be
the F region F = F −1 takes the form represented in the form 2 + f23/2, which explic-
F =Re f 0
2 + f 2−/4. itly demonstrates this fact.
F x = Re f + 2,30
Substituting Eq. 2.28 into Eq. 2.30, we obtain finally
the DOS variation the edge of the F film, 2. Transition temperature
As we have seen previously, the exchange field greatly
f BCS 2 affects singlet pairing in conventional superconductors.
−2
F d F = 1/4Re + sinh + d F Therefore the critical temperature of the superconduct-
F
ing transition T c is considerably reduced in S / F struc-
+ −2 . 2,31 tures with a high interface transparency.
− sinh − d F
The critical temperature for S / F bilayer and multilay-
In Fig. 3 we plot the function F for different thick- ered structures was calculated in many works. 3 Experi-
nesses d F and h /=20. It can be seen that at zero energy mental studies of the T c were also reported in many
=0 the correction to DOS F is positive for F films publications Jiang et al. 1995; Aarts et al. , 1997; Mühge
with d F =0.8 0 while it is negative for films with d F et al. 1998; Lazar et al. 2000; Gu et al. , 2002a. Boa
=0.5 0 , where 0 = D F /.
Such behavior of the DOS, which is typical for S / F 3 See, for example, Buzdin and Kupriyanov, 1991; Radovic et
systems, has been observed experimentally by Kontos et
al. 1991; Demler et al. , 1997; Khusainov and Proshin, 1997;
al. 2001 in a bilayer consisting of a thin PdNi film 5 Proshin and Khusainov, 1998, 1999; Tagirov, 1998; Baladie et
d F 7.5 nm on the top of a thick superconductor. o al. 2001; Proshin et al. 2001; Fominov et al. , 2002, 2003; Ba-
DOS was determined by tunneling spectroscopy. este grets et al. 2003; Baladie and Buzdin, 2003; You et al. , 2004;
type of dependence of N on d N can also be obtained in Tollis et al. , 2005.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 11
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1331

± h˜ ˜
g±= ,f±= , 2,32
E± E±

where E ± = ± h˜ 2 −˜ 2 , h˜ = r F h , ˜ = r S , r F =1− r S
= F d F / F d F + S d S . In this case the Green's functions
are uniform in space and have the same form as in a
magnetic superconductor or in a superconducting film in
a parallel magnetic field acting on the spins of electrons.
The difference between the S / F bilayer system and a
magnetic superconductor is that the effective exchange
energy h˜ depends on the thickness of the F layer and
may be significantly reduced in comparison with its
value in a bulk ferromagnet. A thin superconducting
film in a strong magnetic field H = h˜ / B B is an effec-
tive Bohr magneton is described by the same Green’s
funções. The behavior of these systems and, in particu-
lar, the critical temperature of the superconducting tran-
sition T c , was studied long ago by Sarma 1963, Larkin
FIG. 4. Dependence of superconducting transition tempera- and Ovchinnikov 1964, Fulde and Ferrell 1965, and
ture on the thickness of the Fe layer as determined by resistiv-Maki 1968. It was established that both first- and
ity measurements. The dashed line is a fit assuming a perfect
second-order phase transitions may occur in these sys-
interface transparency while the solid line corresponds to a
nonperfect interface. Adapted from Lazar et al. , 2000. tems if h˜ is less than or on the order of ˜ . If the effective
exchange field h˜ exceeds the value ˜ / 20.707 ˜ , a
system remains in the normal state the Clogston 1962
agreement between theory and experiment has been and Chandrasekhar 1962 limit. Independently from
achieved in some cases see Fig. 4. One should mention each other, Larkin and Ovchinnikov 1964 and Fulde
that despite many papers published on this subject, the and Ferrell 1965 found that in a clean system and in a
problem of the transition temperature T c in S / F struc-
tures is not completely clear. For example, Jiang et al. narrow interval of h˜ the homogeneous state is unstable
1995 and Ogrin et al. 2000 claimed that the nonmono- and an inhomogeneous state with the order parameter
varying in space is established in the system. This state,
tonic dependence of T c on the thickness of the ferro- denoted as the Fulde-Ferrel-Larkin-Ovchinnikov state,
magnet observed on Gd/Nb samples was due to the os- has not been observed yet in bulk superconductors. Dentro
cillatory behavior of the condensate function in F .
https://translate.googleusercontent.com/translate_f 9/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
However, Aarts et al. 1997 in another experiment on bilayeredofSa/ short
because F systems
meansuch
free apath.
state cannot be realized
V/FeV showed that the interface transparency plays a In the case of a first-order phase transition from the
crucial role in the interpretation of the experimental superconducting to the normal state, the order param-
data that showed both nonmonotonic and monotonic eter drops from a finite value to zero. The study of this
dependence of T c on d F . In other experiments Bour- transition requires the use of nonlinear equations for .
geois and Dynes, 2002 the critical temperature of the It was shown by Tollis 2004 that under some assump-
bilayer Pb/Ni decreases with increasing d F in a mono- tions both the first- and second-order phase transitions
tonic way. may occur in a S / F / S structure.
From the theoretical point of view, the T c problem in In the case of a second-order phase transition, one can
a general case cannot be solved exactly. In most papers it linearize the corresponding equations the Eilenberger
is assumed that the transition to the superconducting or Usadel equation for the order parameter and use the
state is of second order, ie, the order parameter var- Ginzburg-Landau expression for the free energy assum-
ies continuously from zero to a finite value with decreas- ing that the temperature T is close to the critical tem-
ing temperature T . However, this is generally not so. perature T c . Just this case was considered in most papers
Let us consider, for example, a thin S / F bilayer with on this topic. The critical temperature of an S / F struc-
thicknesses obeying the condition d F F , d S S , where ture can be found from an equation which is obtained
d F , S are the thicknesses of the FS layer. Neste caso, o from the self-consistency condition 2.4. In the Matsub-
Usadel equation can be averaged over the thickness ara representation it has the form
see, for instance, Bergeret et al. 2001b and reduced to
an equation describing a uniform magnetic supercon- Tc 1
ln T *c − if / , 2,33
ductor with an effective exchange energy h˜ and order T *c = n
parameter ˜ . where T c is the critical temperature in the absence of the
This problem can easily be solved. The Green's func- proximity effect and T *c is the critical temperature taking
tions g ± and f ± are given by into account the proximity effect.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 12
1332 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

The function f is the condensate Gor'kov function


in the superconductor; it is related to the function f S 3
as f S 3 i n = f , where n = 2 n +1 is the Matsubara fre-
quency. Strictly speaking, Eq. 2.33 is valid for a super-
conducting film with a thickness smaller than the coher-
ence length S because in this case f is almost constant
in space.
The quasiclassical Green's function f obeys the Us-
adel equation in the diffusive case or the more general FIG. 5. Measurement of the critical current I c as a function of
Eilenberger equation. One of these equations has to be temperature for a Nb/Cu 0.48 Ni 0.52 /Nb junction. A espessura
solved by using the boundary conditions at the S / F in- of the CuNi layer is d F =22 nm. Adapted from Ryazanov et al. ,
terface or S / F interfaces in case of multilayered struc- 2001.
turas. This problem was solved in different situations in
many works where an oscillation of T c as a function of
the F thickness was predicted see Fig. 4. In most of 2002; Kontos et al. 2002; Bauer et al. , 2004; Sellier et al. ,
these papers it was assumed that magnetization vectors 2004.
M in different F layers are collinear. Only Fominov et al. In the experiments of Ryazanov et al. 2001 and Blum
2003 considered the case of an arbitrary angle estar- et al. 2002, Nb was used as a superconductor and a
tween the M vectors in two F layers separated by a su- Cu x Ni 1− x alloy as a ferromagnet. Kontos et al. 2002
perconducting layer. used a more complicated S 1 / F / I / S structure, where S 1
As mentioned previously, in this case the triplet com- was a Nb/Al bilayer, S was Nb, I was the insulating
ponents with all projections of the spin S of the Cooper Al 2 O 3 layer, and F was a thin 40 d F 150 Å magnetic
pairs arise in the F / S / F structure. It was shown that T c layer of a PdNi alloy. All these structures exhibit oscil-
depends on decreasing from a maximum value T c max lations of the critical current I c . In Fig. 5 the tempera-
at =0 to a minimum value T c min at = . Nós não devemos ture dependence of I c measured by Ryazanov et al.
discuss the problem of T c for S / F structures in detail 2001 is shown. It can be seen that the critical current in
because this problem is discussed in other review articles the junction with d F =27 nm turns to zero at T 2 K,
Izyumov et al. 2002; Buzdin, 2005a. rises again with increasing temperature, and reaches a
maximum at T 5.5 K. If the temperature increases fur-
ther, I c decreases. In Fig. 6 we also show the dependence
3. The Josephson effect in S / F / S junctions
of I c on the thickness d F measured by Blum et al. 2002.
The oscillations of the condensate function in the fer- The measured oscillatory dependence is well fitted with
romagnet see Eq. 2.29 lead to interesting peculiarities the theoretical dependence calculated by Buzdin et al.
not only in the dependence T c d F but also in the Jo- 1982 and Bergeret et al. 2001c. o state in a Jo-
sephson effect in S / F / S junctions. Although as men- sephson junction leads to some observable phenomena.
tioned in the previous section that the experimental re- As was shown by Bulaevskii et al. 1977, a spontaneous
sults concerning the dependence T c d F are still supercurrent may arise in a superconducting loop with a
controversial, there is more evidence for these oscilla- ferromagnetic junction. This current has been mea-
tions in experiments on the Josephson current measure- sured by Bauer et al. 2004. Note also that the fractional
ments that we shall discuss here. Shapiro steps in a ferromagnetic junction were ob-
It turns out that under certain conditions the Joseph- served by Sellier et al. 2004 at temperatures at which
son critical current I c changes its sign and becomes nega-
the critical current I c reduces to zero.
tiva. In this case the energy of the Josephson coupling
E J = I c / e 1−cos has a minimum in the ground state
when the phase difference is equal not to 0, as in or-
dinary Josephson junctions, but to a junction.

https://translate.googleusercontent.com/translate_f 10/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
Thiseteffect
evskii was predicted
al. 1977. forconsidered
The authors the first time by Bula-
a Josephson
junction consisting of two superconductors separated by
a region containing magnetic impurities. The Josephson
current through a S / F / S junction was calculated for the
first time by Buzdin et al. 1982. Different aspects of the
Josephson effect in S / F / S structures were studied in
many subsequent papers Buzdin and Kupriyanov, 1991;
Fogelström, 2000; Heikkilä et al. 2000; Chtchelkatchev FIG. 6. Critical current of a Nb/Cu/Ni/Cu/Nb junction as a
et al. 2001; Barash et al. 2002; Golubov et al. , 2002a; function of the Ni layer thickness d . The squares are the mea-
Radovic et al. 2003; Zyuzin et al. , 2003. Recent experi- sured points. The theoretical fits are presented according to
ments confirmed the 0- transition of the critical current Buzdin et al. 1982 dashed line and Bergeret et al. 2001c
in S / F / S junctions Ryazanov et al. 2001; Blum et al. , solid line. Adapted from Blum et al. , 2002.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 13
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1333

Oscillations of the Josephson critical current I c are


related to the oscillatory behavior of the condensate
function f in space see Eq. 2.29 . The critical current I c
in a S / F / S junction can easily be obtained once the con-
densate function in the F region is known. Nós usamos o
following formula for the superconducting current I S in
the diffusive limit, which follows in the equilibrium case
from a general expression see Appendix A:

I S = L y L z F i T /4 e Tr ˆ 3 f + f + / x + f − f − / x ,

2,34
where L y L z is the area of the interface and F is the FIG. 7. Dependence of the critical current on T for h =0.06,
0 =0.03, d / l =, and different values of Ql . Aqui éo
conductivity of the F layer. momentum relaxation time.
In the considered case of a nonzero phase difference
the condensate functions f ± are matrices in the particle-
hole space. If in Eq. 2.34 instead of f ± we write a too large. If the exchange energy was not too large, the
44 matrix for f , then is given by ˆ effective penetration length F ,eff would be temperature
= iˆ 2 cos /2 iˆ 1 sin /2 ˆ 3 . We set the phase of the dependent. According to estimates presented by Ryaza-
right left superconductor equal to ± /2. For simplicity nov et al. , h 30 K, which means that the exchange en-
we assume that the overlap between the condensate ergy in this experiment was much smaller than in the
functions f ± induced in the F region by each supercon- one performed by Blum et al. and by Kontos et al. no
ductor is small. This assumption is correct in the case last reference h 35 meV.
d F F . Under this assumption the condensate function The conditions under which the state is realized in
may be written in the form of two independently in- S / F / S Josephson junctions of different types were stud-
duced f functions, ied theoretically in many papers Buzdin and Kupriy-
anov 1991; Buzdin, Vujicic, and Kupriyanov, 1992;
f ± x = 1/ ˆ Chtchelkatchev et al. 2001; Krivoruchko and Koshina
F ± F l exp− ± x + d F /2
2001a; Li et al. 2002; Buzdin and Baladie, 2003. In these
+ ˆ r exp− ± − x + d F /2 . 2,35 papers it was assumed that the ferromagnet consisted of
a single domain with a magnetization M fixed in space.
Aqui ˆ r is the order parameter in the right supercon-
The case of a S / F / S Josephson junction with a two-
ductor and ˆ l is the left. Substituting Eq. 2.35 into Eq. domain ferromagnet was analyzed by Blanter and Hek-
2.34, we get king 2004. The Josephson critical current I c was calcu-
IS I c sin = 4 TL y L z 2 exp− d F / F lated for parallel and antiparallel magnetization
F / FF orientations in both ballistic and diffusive limits. It turns
2 out that in such a junction the current I c is larger for the
cos d F / F 2,36
2 + 2 sin . antiparallel orientation.
A similar effect arises in a S / F / S junction with a ro-
When deriving Eq. 2.36, it was assumed that the ex- tating in-space magnetization, as was shown by Bergeret
change energy h is much larger than both T and . et al. 2001c. In this case not only the singlet and triplet
Calculating the sum in Eq. 2.36, we come to the final components with projection S z =0, but also the triplet
formula for the critical current, component with S z = ±1 arises in the ferromagnet. o
Ic= tanh/2 T F / FF 2 exp− d F / F cos d F / F . last component penetrates the ferromagnet over a large
length of the order of N and contributes to the Joseph-
2,37 son current.
As expected, according to Eq. 2.37 the critical current In Fig. 7 the temperature dependence of the critical
oscillates with varying the thickness of the ferromagnet current is presented for different values of Ql , where
d F . The period of these oscillations gives the value of F Q =2 / L m , L m is the period of the spatial rotation of the
and therefore the value of the exchange energy h . Para magnetization, and l is the mean free path. Pode ser
example, according to the experiments on Nb/Cu/Ni/Cu/ seen that at Q =0 homogeneous ferromagnet and low
Ni/Nb performed by Blum et al. 2002; see also Palevski, temperatures T the critical current I c is negative
2005, h 110 meV, which is a quite reasonable value state, whereas with increasing temperature, I c becomes
for CuNi. positive 0 state. If Q increases, the interval of negative
The nonmonotonic dependence of the critical current I c gets narrower and disappears completely at Ql 0.04,
on temperature observed by Ryazanov et al. 2001 can that is, the S / F / S structure with a nonhomogeneous M
be obtained only in the case of an exchange energy h is an ordinary Josephson junction with a positive critical
comparable with at least, the ratio h / should not be atual.

https://translate.googleusercontent.com/translate_f 11/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 14
1334 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

It is interesting to note that the -type Josephson cou- tional superconductors, are in a singlet state. Isto é, o
pling may also be realized in S / N / S junctions provided order parameter is represented in the form
the distribution function of quasiparticles n in the N =· i 3 , where 3 is the Pauli matrix in the spin
region deviates significantly from the equilibrium. este espaço. The difference between the s and d pairing is due
deviation may be achieved with the aid of a nonequilib- to a different dependence of the order parameter em
rium quasiparticle injection through an additional elec- the Fermi momentum p F = k F . In isotropic conventional
trode in a multiterminal S / N / S junction. The Josephson superconductors is a k almost independent quantity.
current in such a junction is again determined by Eq. In anisotropic conventional superconductors Depende
2.34 in which one has to set h =0, f + = f − , and replace on the k F direction but it does not change sign as a func-
tanh =1−2 n de 1/2 tanh + eV +tanh tion of the momentum k F orientation in space. In high-
− eV , where V is a voltage difference between N and T c superconductors where d -wave pairing occurs, the or-
S electrodes. der parameter k F changes sign at certain points at the
At a certain value of V the critical current changes Fermi surface.
placa. Thus there is some analogy between the sign- On the other hand, the Pauli principle requires the
reversal effect in a S / F / S junction and the one in a mul- função k F to be an even function of k F , which im-
titerminal S / N / S junction under nonequilibrium condi- poses certain restrictions for the dependence of the or-
ções. der parameter on the Fermi momentum. Por exemplo,
Indeed, when calculating I S in a multiterminal S / N / S for d pairing the order parameter is given by kF
junction one can shift the energy by eV or − eV . Então o =0 k x 2 − k2y, where k x , y are the components of the k F
function 1/2 tanh + eV +tanh − eV is trans- vector in the Cu-O plane. This means that the order
formed into tanh while in the other functions one parameter may have either positive or negative sign de-
performs the substitution → ± eV . So we see that eV is pending on the direction.
analogous to the exchange energy h that appears in the The change of sign of the order parameter leads to
case of a S / F / S junction. different physical effects. For example, if a Josephson
The sign-reversal effect in a multiterminal S / N / S junction consists of two high- T c superconductors with
junction under nonequilibrium conditions was observed properly chosen crystallographic orientations, the
by Baselmans et al. 1999 and studied theoretically by ground state of the system may correspond to the phase
Volkov 1995, Wilhelm et al. 1998, and Yip 1998. difference = junction. In some high- T c supercon-
Later Heikkilä et al. 2000 studied theoretically a com- ductors the order parameter may consist of a mixture of
bined effect of a nonequilibrium quasiparticle distribu- s - and d -wave components Tsuei and Kirtley, 2003.
tion on the current I c in a S / F / S Josephson junction. Another type of pairing, spin-triplet superconductiv-
Concluding this section we note that the experimental ity, has been discovered in materials with strong elec-
results by Strunk et al. 1994, Ryazanov et al. 2001, tronic correlations. The triplet superconductivity has
Blum et al. 2002, and Kontos et al. 2002 seem to con- been found in heavy-fermion intermetallic compounds
firm the theoretical prediction of an oscillating conden- and also in organic materials for a review, see Mineev
sate function in the ferromagnet and the possibility of and Samokhin 1999 . Recently a lot of work has been
switching between the 0 and the Estado. carried out studying the superconducting properties of
strontium ruthenate Sr 2 RuO 4 . Convincing experimental
data have been obtained in favor of triplet, p -wave su-
III ODD TRIPLET SUPERCONDUCTIVITY IN S / F perconductivity. For more details we refer the reader to
ESTRUTURAS the review articles by Maeno et al. 1994 and Eremin et
A. Conventional and unconventional superconductivity al. 2004.
Due to the fact that the condensate function
Since the development of the BCS theory of super- r , t r , t must be an odd function with respect
conductivity 1957, only one type of superconductivity to the permutations ↔ , r↔r for equal times, t = t ,
has been observed in experiments. This type is charac- the wave function of a triplet Cooper pair has to be an
terized by the s -wave pairing between the electrons with odd function of the orbital momentum, that is, the or-
opposite spin orientations due to the electron-phonon bital angular momentum L is an odd number: L =1 p
interação. It can be called conventional since it is ob- wave, 3, etc. Thus the superconducting condensate is
served in most superconductors with critical tempera- sensitive to the presence of impurities. Only the s -wave
ture below 20 K the low-temperature superconductors. L =0 singlet condensate is not sensitive to scattering by
Bednorz and Müller 1986 discovered that a nonmagnetic impurities Anderson theorem. In con-
La 2− x Sr x CuO 4 compound is a superconductor with a trast, the p -wave condensate in an impure material is
critical temperature of 30 K. This was the first known suppressed by impurities and therefore the order param-
high- T c copper-oxide cuprate superconductor. Nowa- eter =k kF k r , t r , t k is also sup-
days many cuprates have been discovered with critical pressed Larkin, 1965. That is why superconductivity in
temperatures above the temperature of liquid nitrogen. impure Sr 2 RuO 4 samples has not been observed. In or-
These superconductors the high- T c superconductors in der to observe triplet p -wave superconductivity or an-
general show a d -wave symmetry and similar to conven- other orbital order parameter with higher odd L , one

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 15
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1335

needs to use clean samples of appropriate materials. Abrahams 1992, Abrahams et al. 1993, and Balatsky
At first glance one cannot avoid this fact and there is et al. 1995 the case of a singlet odd pairing was consid-
no hope of seeing nonconventional superconductivity in ered, while in the other works a triplet odd pairing was
impure materials. However, another nontrivial possibil- estudou.
ity for the triplet pairing exists. The Pauli principle im- We would like to emphasize that while theories of un-

https://translate.googleusercontent.com/translate_f 12/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
poses r ,restrictions
t r , t k foronequal times.a In correlation
the Matsubarafunção conventional superconductivity are often based on the
presence of strong correlations where one has to use a
representação esta significa aquele a soma phenomenology, the triplet state induced in the S / F
r, r, k , must change sign under the per- structures can be studied within the framework of the
mutation r↔r for the triplet pairing the diagonal BCS theory, which is valid in the weak-coupling limit.
matrix elements = of these correlation functions This fact drastically simplifies the problem not only from
estamos not zero. Isso implica que o soma
the theoretical but also from experimental point of view
r, r, k , has to be either an odd function since well-known superconductors grown in a controlled
of k or just reduces to zero. The latter possibility does
way may be used in order to detect the triplet compo-
not mean that the pairing must vanish. It can remain
nent.
finite if the average r, r, k , is an odd func- We summarize the properties of this new type of su-
tion of the Matsubara frequency in this case it must
perconductivity which we speak of as triplet odd super-
be an even function of k . Then the sum over all fre-
conductivity :
quencies is zero and therefore the Pauli principle for the
equal-time correlation functions is not violated. • It contains the triplet component. In particular the
This type of pairing was first suggested by Berezinskii components with projection S z = ±1 on the direction
1975 as a possible mechanism of superfluidity in 3 He. of the field are insensitive to the presence of an ex-
He assumed that the order parameter change field and therefore long-range proximity ef-
,k r , r , k , is an odd function of : fects arise in S / F structures.
=−− . However, experiments on superfluid 3 He have
• In the dirty limit it has a s -wave symmetry. O con-
shown that the Berezinskii state is only a hypothetical
densate function is even in the momentum p and
state and the p pairing in 3 He has different symmetries.
therefore, contrary to other unconventional super-
As is known nowadays, the condensate in 3 He is anti-
symmetric in the momentum space and symmetric trip- conductors, is not destroyed by the presence of non-
let in the spin space. Thus the Berezinskii hypothetical magnetic impurities.
pairing mechanism remained unrealized for a few de- • The triplet condensate function is odd in frequency.
cades.
However, in recent theoretical works it was found that Before we turn to a quantitative analysis let us make a
a superconducting state similar to the one suggested by last remark. We assume that in ferromagnetic regions no
Berezinskii might be induced in conventional S / F sys- attractive electron-electron interaction exists, and there-
tems due to the proximity effect Bergeret et al. , 2001a, fore =0 in the F regions. The superconducting conden-
2003. In the next sections we shall analyze this new type sate arises in the ferromagnet due only to the proximity
of superconductivity with triplet pairing that is odd in efeito. This will become more clear later.
frequency and even in momentum. This pairing is pos- Another type of triplet superconductivity in the S / F
sible not only in the clean limit but also in samples with structures that differs from the one considered in this
a high impurity concentration. review was analyzed by Edelstein 2001. The author as-
It is important to note that, in spite of the similarity, sumed that spin-orbit interaction takes place at the S / F
there is a difference between this new superconducting interface due to a strong electric field which exists over
state in the S / F structures and that proposed by Berez- interatomic distances the so-called Rashba term in the
inskii. In the S / F structures both the singlet and triplet Hamiltonian Rashba, 1960 . It was also assumed that
types of the condensate f coexist. However, the order electron-electron interaction is not only not zero in the
parameter is not equal to zero only in the S region we s -wave singlet channel but also in the p -wave triplet
assume that the superconducting coupling in the F re- canal. The spin-orbit interaction mixes both the trip-
gion is zero and is determined there by the singlet part
let and singlet components. Then the triplet component
of the condensate only. This contrasts the Berezinskii
state where the order parameter should contain a trip- can penetrate into the F region over a large distance.
let component. However, in contrast to odd superconductivity, the
Note that attempts to find conditions for the existence triplet component analyzed by Edelstein is odd in mo-
of odd superconductivity were undertaken in several pa- mentum and therefore must be destroyed by scattering
pers in connection with the pairing mechanism in high- on ordinary nonmagnetic impurities. This type of triplet
T c superconductors Kirkpatrick and Belitz, 1991; Bal- component was also studied in two-dimensional systems
atsky and Abrahams, 1992; Belitz and Kirkpatrick, 1992; and in S / N structures in the presence of the Rashba-
Abrahams et al. , 1993; Coleman et al. , 1993a, 1993b, type spin-orbit interaction Edelstein, 1989, 2001;
1995; Balatsky et al. , 1995. In the works by Balatsky and Gor'kov and Rashba, 2001.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 16
1336 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

B. Odd triplet component homogeneous magnetization g BCS = g BCS ˆ 0 , f BCS = f BCS ˆ3, 3,6
As we have mentioned in Sec. II.B, even in the case of Onde
a homogeneous magnetization the triplet component
with the zero projection S z =0 of the total spin on the g BCS = / , f BCS =/i 3,7
direction of the magnetic field appears in the S / F struc-
e = 2+2.
ture. Unlike the singlet component it is an odd function
We now linearize the Usadel equation with respect to
of the Matsubara frequency . In order to see this, we
look for a solution of the Usadel equation in the Mat- a small deviation g S g S ˆ 3 + f S iˆ 2 = g S − g BCS and ob-
subara representation. In this representation the linear- tain for the condensate function f S in the supercon-
ized Usadel equation for the ferromagnet takes the form ductor the following equation:

D F 2 f F / x 2 − 2 ˆ 0 − ih ˆ 3 f F = 0, 3,1 2 / x 2 f S − S2 f S =2 i / D S g2BCSˆ 3 , 3,8


Onde = T 2 n +1 is the Matsubara frequency and where S 2 =2 2 + 2 / D S and x is a deviation of the
h =sgn h . superconducting order parameter from its BCS value in
The solution of Eq. 3.1 corresponding to Eq. 2,29 the bulk.
can be written as A solution for Eq. 3.8 determines the triplet compo-
nent f S 0 and a correction f S 3 to the singlet component.
https://translate.googleusercontent.com/translate_f 13/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
f± =± /i ± F exp− ± x, 3,2 To find the component f S 3 is a much more difficult task
Onde than to find f S 0 because x is a function of x and, in
its turn, is determined by the amplitude f S 3 . Assim sendo
± =2 ih / D F 3.3 the singlet component f S 3 obeys a nonlinear integro-
e = 2+2. differential equation; that is why the critical temperature
For the amplitudes of the triplet f 0 = f + + f − /2 and T c can be calculated only approximately Buzdin and
singlet f 3 = f + − f − /2 components we get in the ferro- Kupriyanov, 1990; Radovic et al. 1991; Demler et al. ,
magnet 1997; Tagirov, 1998; Izyumov et al. 2002; Bagrets et al. ,
2003; Baladie and Buzdin, 2003. Fominov et al. 2002
f 3,0 ,x= /2 i F proposed an analytical trick that reduces the T c problem
to a form allowing a simple numerical solution.
exp− + x exp− − x
± . 3,4 On the contrary, the triplet component f S 0 propor-
+ - tional to ˆ 0 can be found exactly in the linear approxi-
Equations 3.2 and 3.4 show that both the singlet and mation. The solution for f S 0 0 takes the form
the triplet component with S z =0 of the condensate func-
f S 0 x = f S 0 0exp− S x. 3,9
tions decay in the ferromagnet on the scale of Re +
having oscillations with Im + . Levando em conta que The constant f S 0 0 can be found from the boundary
+ = − − , we see that f 3 is an even function of condition see Appendix A
, whereas the amplitude of the triplet component f 0
is an odd function of . The mixing between the triplet f S 0 x / x x =0 = f F 0 0/ S . 3,10
and singlet components is due to the term proportional As follows from this equation, the triplet component in
to h ˆ 3 in Eq. 3.1. This term breaks the time-reversal the superconductor f S 0 has the same symmetry as the
symmetry. component f F 0 , that is, it is odd in frequency. Então o
Due to the proximity effect the triplet component f 0 triplet component of the condensate is inevitably gener-
also penetrates into the superconductor and the charac- ated by the exchange field in both the ferromagnet and
teristic length of the decay is the coherence length S . superconductor. Both the singlet component and the
The spatial dependence of this component inside the su- triplet component with S z =0 decay fast in the ferromag-
perconductor can be found provided the Usadel equa- net because the exchange field h is usually very large
tion is linearized with respect to a deviation of the f S see Eq. 3.3 . At the same time, the triplet component
matrix from its bulk BCS form f BCS . Na presença de decays much slower in the superconductor because the
inverse characteristic length of the decay k S is much
an exchange field the Green's functions g are 44 ma-
menor.
trices in the particle-hole and spin space. In the case of
To illustrate some consequences with the presence of
homogeneous magnetization they can be represented as
the triplet component in the superconductor, we use the
a sum of two terms the ˆ matrices operate in the
fact that the normalization condition g 2 =1 results in the
particle-hole space
relation
g = gˆ 3 + fiˆ 2 , 3,5 g0g3=f3f0. 3,11
where g and f are matrices in the spin space. The function g 0 entering Eq. 3.11 determines the
In a bulk superconductor these matrices are equal to change of the local density of states,

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 17
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1337

In order to determine the structure of the condensate


we shall use as before the method of quasiclassical
Green's functions. This allows us to investigate all inter-
esting phenomena except those that are related to quan-
tum interference effects.
Quasiclassical Green's functions can be used at spatial
scales much longer than the Fermi wavelength. 4 As we
have already mentioned, in order to describe S / F struc-
tures the Green's functions have to be 44 matrices in
the particle-hole and spin space. Such 44 matrix
FIG. 8. Trilayer geometry. The magnetization of the left F Green's functions not necessarily in the quasiclassical
layer makes an angle with the z axis and that of the right form were used by Vaks et al. 1962 and Maki 1969.
makes an angle −. Equations for the quasiclassical Green's functions in the
presence of the exchange field similar to the Eilenberger
and Usadel equations can be derived in the same way as
= Re g 0 , 3,12 the one used in the nonmagnetic case see Appendix A.
while the function g 3 determines the magnetic moment For example, a generalization of the Eilenberger equa-
M z of the itinerant electrons see Appendix A, tion was presented by Bergeret, Efetov, and Larkin
2000 and applied to the study of cryptoferromag-
Mz=BiT g3 . 3,13 netism.

We see that the appearance of the triplet component in 1. F / S / F trilayer structure


the superconductor leads to a finite magnetic moment in
We start the analysis of the nonhomogeneous case by
the S region, which can considered as an inverse prox-
considering the F / S / F system shown in Fig. 8. The struc-
imity effect. This problem will be discussed in more de-
ture consists of one S layer and two F layers with mag-
tail in Sec. VB
Thus even in the case of homogeneous magnetization, netizations inclined at the angle ± with respect to the z
the triplet component with S z =0 arises in the S / F struc- axis in the y - z plane.
ture. This fact was overlooked in many papers and was We demonstrate now that the triplet component with
noticed for the first time by Bergeret et al. 2003. This S z = ±1 inevitably arises due to the overlap of the triplet
component, as well as the singlet one, penetrates the components generated by the ferromagnetic layers in

https://translate.googleusercontent.com/translate_f 14/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
ferromagnet over a short length F because it consists of theso.
be S layer. It is not difficult to understand why it should
averages of two operators with opposite spins ↑ ↓ and
is strongly suppressed by the exchange field. The triplet As we have seen in the previous section, each of the
layers generates the triplet component with zero total
component with projections S z = ±1 on the direction of
the field results in more interesting properties of the sys- projection of the spin, S z =0, in the direction of the ex-
tem since it is not suppressed by the exchange interac- change field. If the magnetic moments of the layers are
ção. It can be generated by nonhomogeneous magneti- collinear to each other parallel or antiparallel, the total
projection remains zero. However, if the moments of the
zation as we shall discuss next.
ferromagnetic layers are not collinear, the superposition
of the triplet components coming from the different lay-
ers should have all possible projections of the total spin.
C. Triplet odd superconductivity inhomogeneous
magnetization From this argument we can expect the generation of
the triplet component with all projections of the total
According to the results of the last section the pres- spin provided the thickness of the S layer is not too
ence of an exchange field leads to the formation of the ampla. Since the only relevant length in the supercon-
triplet component of the condensate function. In a ho- ductors is S D S / T c , we assume that the thickness of
mogeneous exchange field, only the component with the the superconducting layer S does not exceed this length.
projection S z =0 is induced. Now we perform explicit calculations that support the
A natural question arises: Can the other components qualitative conclusion on the generation of the triplet
with S z = ±1 be induced? If they could, this would lead
to a long-range penetration of superconducting correla-
tions into the ferromagnet because these components 4 Note that as was shown by Shelankov and Ozana 2000 and
correspond to correlations of the type Galaktionov and Zaikin 2002, in the ballistic case and in the
↑ ↑ with paral- presence of several potential barriers some effects similar to
lel spins and are not as sensitive to the exchange field as
outras. the quantum interference effects may be important. We do not
consider purely ballistic systems assuming that the impurity
In what follows we analyze some examples of S / F scattering is important. In this case the quasiclassical approach
structures in which all projections of the triplet compo- is applicable. The applicability of the quasiclassical approxima-
nent are induced. The common feature of these struc- tion was discussed by Larkin and Ovchinnikov Larkin and
tures is that magnetization is nonhomogeneous. Ovchinnikov, 1968.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 18
1338 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

component with all projections of the total spin. Nós con- 3.22, we find the functions b i x and b¯ i x which are
sider the diffusive case when the Usadel equation is ap- decaying exponential functions and can be written as
plicable. This means that the condition
b k x = b k exp− x − d S ,
h 1 3,14
is assumed to be fulfilled is the elastic-scattering time. b¯ k x = b¯ k exp x + d S . 3,23
The linearized Usadel equation in the F region takes
the form see Appendix A Substituting Eq. 3.23 into Eqs. 3.15–3.18, we obtain
a set of linear equations for the coefficients b k . O con-
i h2 dition for the existence of nontrivial solutions yields an
2f−2f+ ˆ 0 ˆ 3 , f + cos ± ˆ 3 ˆ 2 , f sin = 0,
xx 2 equation for the eigenvalues . This equation reads
3,15 2−2 2 − 2 2 + h 4 = 0. 3,24
where f is a 44 matrix condensate function which is Equation 3.24 is of the sixth order and therefore has
assumed to be small and ˆ 3 , f + = ˆ 3 · f + f · ˆ 3 . The wave six solutions. Three of these solutions should be dis-
vectors and h entering Eq. 3.15 have the form carded because those corresponding to b k x grow when
going away from the interface. The remaining three so-
2 =2/ D F 3,16 lutions of Eq. 3.24 give three different physical values
e of .
If the exchange energy h is sufficiently large h
2 = 2 h sgn / D F . 3,17 T , , the eigenvalues are
h
The magnetization vector M lies in the y , z plane = , 3,25
and has components M = M 0 , ± sin ,cos . The sign +
corresponds to the right F film and − to the left. Nós
consider the simplest case of a highly transparent S / F ± 1 ± i h / 2. 3,26
interface and temperatures close to the critical tempera- We see that these solutions are completely different.
ture of the superconducting transition T c . Nesse caso The roots ± proportional to h cf. Eq. 3.17 are very
the function f , being small, obeys a linear equation simi- large and therefore the corresponding solutions b k s de-
lar to Eq. 3.8, cay very fast similar to the singlet component. Isto é
the solution that exists for a homogeneous magnetiza-
2 f / x 2 − S2 f = 2 i ˘ / D S g2BCS, 3,18 tion collinear magnetization vectors.
In contrast, the value for given by Eq. 3.25 is much
where S 2 =2 / D S . smaller see Eq. 3.16 and corresponds to a slow decay
The boundary conditions at the S / F interfaces are of superconducting correlations. Solutions correspond-
ing to the root given by Eq. 3.25 describe a long-range
fx=d S +0 =fx=d S −0 , 3,19 penetration of the triplet component into the ferromag-
netic region. For each root one can easily obtain rela-
f/xF=f/xS, 3,20 tions between the coefficients b k x . As a result, we ob-
tain
where = F / S and F is the conductivity in the ferro-
magnet and S is the conductivity of the superconductor. b 1 x = be − x − d − sin
S b 3+ e − + x − d S − b 3− e − − x − d, S

The first condition, Eq. 3.19, corresponds to the con- 3,27


tinuity of the condensate function at the S / F interface
with a high transparency, whereas Eq. 3.20 ensures the
https://translate.googleusercontent.com/translate_f 15/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
continuity of the current across the S / F interface b 0 x = − tan be − x − d − cos
S b 3+ e − + x − d S

Volkov et al. , 2003. − b 3− e − x − d ,


− S 3,28
A solution for Eqs. 3.15–3.18 with the boundary
conditions 3.19 and 3.20 can easily be found. O ma- e
trix f can be represented as b 3 x = b 3+ exp− + x − d S + b 3+ exp− − x − d S .
3,29
f = iˆ 2 f 2 + iˆ 1 f 1 , 3,21
The function b 1 x is the amplitude of the triplet com-
Onde
ponent penetrating into the F region over a long dis-
f1=b1xˆ1,f2=b3xˆ3+b0xˆ0. 3.22 tance on the order of −1 N . Its value as well as the
values of the other functions b k x , is found from the
In the left F layer the functions b k x are to be replaced boundary conditions 3.19 and 3.20 at the S / F inter-
by b¯ k x . For simplicity we assume that the thickness of rostos.
the F films d F exceeds F the case of an arbitrary d F was What remains to be done is to match the solutions for
analyzed by Bergeret et al. 2003. Using Eqs. 3.21 and the superconductor and the ferromagnets at the inter-

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 19
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1339

faces between them. The solution for the supercon- 1


ductor satisfies Eq. 3.18 and can be written as a3=− h , 3,36
iE S sinh2 S

f 3 x = / iE + a 3 cosh S x , 3,30 Onde = S d S .


As follows from the first of these equations, Eq. 3.35,
the correction to the bulk BCS solution for the singlet
f 0 x = a 0 cosh S x , 3,31
component is small in this approximation and this justi-
fies our approach.
f 1 x = a 1 sinh S x , 3,32 At the S / F interface the amplitude of the triplet com-
ponent b is small in comparison with the magnitude of
where E = 2+2. the singlet one b 3+ . However, the triplet component de-
Matching these solutions with Eqs. 3.27–3.29 at the cays over a long distance N while the singlet one van-
S / F interfaces we obtain the coefficients b k and b¯ k as ishes at distances exceeding the short length F . The am-
well as a k . Note that b 3± = b¯ 3± and b =− b¯ . Apesar de plitudes b and b 3± become comparable if the parameter
solution can be found for arbitrary parameters entering h / S is on the order of unity.
the equations, for brevity we present here the expres- It also follows from Eq. 3.34 that the amplitude of
sions for b 3± and b in some limiting cases only. the triplet component b is zero in the case of collinear
Let us consider first the case when the parameter vectors of magnetization, ie, at =0 or = /2. isto
reaches the maximum at the angle m for which sin m
h / S is small, that is, we assume the condition =1/ 3. Therefore the maximum angle-dependent factor
DFh in Eq. 3.34 is sin m cos 2 m =2/3 30.385.
F 1 3,33
h/S DSTc One can see from Eq. 3.34 that b becomes expo-
S nentially small if the thickness d S of the S films signifi-
to be fulfilled. cantly exceeds the coherence length S D S / T c . este
Here F , S is the density of states in the ferromagnet means that in order to have a considerable penetration
and superconductor, respectively in the quasiclassical of the superconducting condensate into the ferromagnet
approximation, the DOS for electrons with spin up and one should not make the superconducting layer too
spin down is nearly the same: h thick.
F . The condition, Eq. On the other hand, if the thickness d S is too small, the
3.33, can be fulfilled in the limit D F D S . Taking, for
example, F critical temperature T c is suppressed. A fim de evitar
S , l F 30 Å, and l S 300 Å, we find that
h should be smaller than 30 T c . this suppression one has to use, for instance, an F / S / F
In this limit the coefficients b 1,3± and a 1 can be written structure with a small width of the F films. Similar sys-
in a rather simple form, tems were considered by Beckmann et al. 2004 in
which nonlocal effects of Andreev reflections in a S / F
2 nanostructure were studied.
b - h sin cos 2 , 3,34
E Another limiting case that allows a comparatively
S sinh2 S simple solution is the limit of small angles Volkov et
al. , 2003 but an arbitrary parameter h / S , Eq. 3.33.
b 3+ b 3− , 3,35 At small angles the amplitudes of the triplet and sin-
2i glet components are given by the following formulas:

pecado h / S tanh S
b - , 3.37
E cosh 2 2 1+
S tanh S + h/S h / S tanh S
1
b 3± . 3,38
2 iE 1 + ± / S tanh S

One can see from Eqs. 3.37 and 3.38 that provided eral case of an arbitrary S / F interface transparency and
the parameter given by Eq. 3.33 is not small and a finite thickness of the F films.
,S 1, the amplitudes b and b 3± are again compa- In Fig. 9 we plot the spatial dependence of the triplet

https://translate.googleusercontent.com/translate_f 16/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
rable
Thetoamplitudes
each other.of the triplet and singlet components and singlet
seen components
from this in as
figure that, a Fexpected,
/ S / F structure. Pode
the triplet ser
com-
were calculated by Bergeret et al. 2003 in a more gen- ponent decays slowly, whereas the singlet component

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 20
1340 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

FIG. 10. S / F structure with a domain wall in the region 0 x


w . In this region = Qx , where Q is the wave vector which
describes the spiral structure of the domain wall. For xw it is
FIG. 9. Spatial dependence of Imsinglet component dashed assumed that magnetization is homogeneous, ie, w x
line and long-range part of Retriplet component. We have w = Qw .
escolhidoF / S =0.2, h / T C =50, F R b / F =0.05, d F T C / D S =2,
d S T C / D S =0.4, and =/4. The discontinuity of the triplet
component at the S / F interface occurs because the short-range limit it is still legitimate to speak about odd supercon-
part is not shown. Taken from Bergeret et al. , 2003. ductivity since the symmetric part is much larger than
the antisymmetric part of f .
decays fast over the short length h . For this reason, in a If the parameter h is not small, ie, the system is not
multilayered S / F structure with a varying direction of diffusive, the symmetric and antisymmetric parts are
the magnetization vector M and thick F layers h dF, comparable, and one cannot speak of odd superconduc-
a Josephson-like coupling between neighboring layers atividade. All this distinguishes the superconductivity in S / F
can be realized via the odd triplet component. Nisso structures from the odd superconductivity suggested by
case the in-plane superconductivity is caused by both Berezinskii 1975, who assumed that the order param-
triplet and singlet components. Properties of such S / F eter was an odd function of . In our discussion it
multilayered structures will be discussed in the next sec- is assumed that the order parameter é um
ção. -independent quantity and it is determined by the sin-
Let us mention an important fact. The quasiclassical glet component of the condensate function f 0 .
Green's function g in the diffusive case can be ex-
panded in spherical harmonics. In the present approach
only the first two terms of this expansion are taken into 2. Domain wall at the S / F interface
account such that In the previous section we have seen how the genera-
tion of the triplet component takes place. The appear-
g = g sym + g as cos , 3,39 ance of this component leads to long-range effects in a
structure in which the angle between the directions of
Onde is the angle between the momentum p and the magnetization in the different layers can be changed ex-
x axis, g as =− lg sym g sym / x is the antisymmetric part of perimentally. This is an example of a situation when the
g , and g sym is the isotropic part of g , which does long-range triplet component of the superconducting
not depend on . The antisymmetric part of g deter- condensate can be produced under artificial experimen-
mines the electric current in the system. tal conditions.
Higher-order terms in the expansion of g are small in In this section we show that the conditions under
the diffusive limit and can be neglected. In the case of a which the triplet long-range superconducting correla-
weak proximity effect the antisymmetric part of the con- tions occur are considerably more general. It is well
densate function in the F region can be written as known that the magnetization of any ferromagnet can
be quite inhomogeneous due to the presence of domain
paredes. They are especially probable near interfaces be-
f as cos − lˆ 3 ˆ 0 sgn f sym / x cos . 3,40 tween ferromagnets and other materials. Therefore in
making an interface between ferromagnets and super-
This expression follows from the fact that g 0 − ˆ 3 conductors one almost inevitably produces domain
ˆ 0 sgn corrections to g 0 are proportional to f 0 2. walls, and one should take special care to get rid of
Equation 3.40 holds for both the singlet and triplet eles.
componentes. In this section we consider a domain-wall-like struc-
ture and show that it will also lead to triplet long-range
As we have clarified previously, the symmetric part f 0 correlações. This structure is shown schematically in
is an odd function of . Thus according to Eq. 3.38 the Fig. 10. It consists of a S / F bilayer with a nonhomoge-
antisymmetric part is an even function of so that the neous magnetization in the F layer. Actually, the odd
total condensate function f = f 0 + f 1 cos is neither an odd triplet condensate has first been obtained in the dirty
nor an even function of . However, in the diffusive limit within this model Bergeret et al. , 2001b. Later, a

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 21
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1341

similar structure was considered in the clean limit by Performing this transformation we obtain instead of
Kadigrobov et al. 2001. We assume for simplicity that Eq. 3.15 a new equation,

https://translate.googleusercontent.com/translate_f 17/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
the magnetization
tates vector Mfrom
in the F film starting = Mthe
„0,sin x ,cos xx … ro-
S / F interface 2 − Q 2 /2 f n − 2 f n + i h 2 ˆ 3 , f n + −Q 2 ˆ 1 f n ˆ 1
=0 and the rotation angle has a simple piecewise x de- xx 2 2
pendence,
+ iQˆ 3 ˆ 1 , x f n + = 0. 3,45
Qx , 0 x w, Correspondingly, the boundary condition, Eq. 3.42,
x= 3,41
w = Qw , w x. takes the form
This form means that the M vector is aligned parallel to 3,46
the z axis at the S / F interface and rotates by the angle F Q /2 iˆ 3 ˆ 1, f n + + f n / x = − f s .
Equation 3.45 complemented by this boundary condi-
w = w over the length ww may be the width of a tion has to be solved in the region 0 x w . Lá-
domain wall. At xw the orientation of the vector M is
fixed. gion wx one needs to solve Eq. 3.15 with Q =0. Ambos
We calculate the condensate function in the F region solutions should be matched at x = w under the assump-
and show that it contains the long-range triplet compo- tion that there is no barrier at this point. Therefore the
nent LRTC. As in the preceding section, we assume matrix f n and its “generalized” derivative should be con-
that the condensate function in the F region is small. tinuous at x = w ,
The smallness of f in this case is either due to a mis-
match of the Fermi velocities in the superconductor and f nx = w −0 = f nx = w +0 , 3,47
ferromagnet or due to a possible potential barrier at the
S / F interface. In such cases the transparency of the in- Q
iˆ 3 ˆ 1, f n + + x f nx = w −0 = x f nx = w +0 . 3,48
terface is small and only a small portion of the supercon- 2
ducting electrons penetrates the ferromagnet.
In this case the solution has the same structure as Eq.
Due to the smallness of the transparency of the inter-
3.21 but small changes should be made. The eigenval-
face, the function f can experience a jump at the S / F ues obey the equation
interface, which contrasts with the preceding case. o
2 − Q 2 − 2 2 +4 Q 2 2 2 − 2 + h 4 2 − Q 2 − 2 = 0,
boundary condition for the 44 matrix f has the same
form as in Eq. 2.27, 3,49

3,42 Onde 2, hare determined in Eqs. 3.16 and 3.17. o


F x f =− f S . eigenvalue given by Eq. 3.25 changes. Now it is equal
The function f S on the right-hand side is the condensate para
matrix Green's function in the superconductor that, in 2=Q2+2, 3,50
the limit considered here, should be close to the bulk Q
solution while the eigenvalues ± , Eq. 3.26, remain unchanged
provided the condition
f S = f BCS iˆ 2 ˆ 3 . 3,43 Q, 3,51
h
We have to again solve Eq. 3.15 with boundary condi- is fulfilled.
tions 3.42. Therefore we assume that the domain-wall In the opposite limit of large Q
thickness w is larger than the mean free path l and the h , the eigenvalues
condition 3.14 is fulfilled dirty limit. This case was ± take the form
analyzed by Bergeret et al. 2001b. Another thin do- ± = ± iQ 1 i h2 / 2 Q 2 . 3,52
main wall case wl was considered by Kadigrobov et
al. 2001. Thus in this limit ± is imaginary in the main approxi-
The problem of finding the condensate functions in mation, which means that the function f n x oscillates
the case of the magnetization varying continuously in fast in space with the period 2 / Q . In this case the ei-
space is more difficult than the previous one because the genvalues 3.50 also change and have the form
angle now depends on x . However, one can use a trick 4
that helps to solve the problem, namely, we exclude the 2=2+ h 3,53
dependence x by introducing a new matrix f n related Q2.
to f via a unitary transformation a rotation in particle- Therefore the limit of a very fast rotating magnetization
hole and spin space, h / Q→ 0 is analogous to the case of a normal metal,
ie, when the condensate penetrates the ferromagnet
f = Uˇ · f n · Uˇ † , 3.44 over the length −1 D F /2 T , which is the characteris-
tic penetration length of the condensate in a S / N sys-
where Uˇ =exp iˆ 3 ˆ 1 x /2. tem.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 22
1342 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

More interesting and realistic is the opposite limit


when the condition Eq. 3.51 is fulfilled and the long-
range penetration of the triplet component into the fer-
romagnet becomes possible.
In the limit of large h , Eq. 3.51, the singlet compo-
nent penetrates the ferromagnet over a short length of
the order h =1/ h while the LRTC penetrates over the
length 1/ Q . As follows from Eq. 3.50, this penetra-
tion length is about 1/ Q provided w / w is smaller than
the length N .
Now let us find the amplitude of the LRTC. O então-
lution for Eq. 3.45 in the interval 0 x w is deter-
mined by Eqs. 3.21 and 3.22 with the functions b i x ,
i =0,1,3, given by the following formulas: FIG. 11. Dependence of the amplitude of the triplet compo-
nent on w = Qw . We have chosen w =0.01.

https://translate.googleusercontent.com/translate_f 18/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
b 1 x = b Q exp Q x + b¯ Q exp− Q x , 3,54
The maximum of c is achieved at
b 0 x = − b 3+ exp− + x + b 3− exp− − x , 3.55
e max = w 5 − 1/ 2 0.786 w . 3,61

b 3 x = b 3+ exp− + x + b 3− exp− − x . 3,56 At w = max the ratio in the first set of large parentheses
in Eq. 3.60 is equal to 0.68. This means that the am-
In the region wx the solution for the condensate func- plitude of the LRTC is of the order of the singlet com-
tion f n takes the form ponent at the S / F interface. The width w should not be
too small because in deriving the expression for c Q we
f n = iˆ 1 ˆ 1 c exp− x−w, 3.57 assumed the condition w h.
where c is a coefficient that has to be found by match- In Fig. 11 we present the dependence of c on w for
ing the solutions at x = w . a fixed w . The spatial dependence of the LRTC and the
Terms of the order of Q / h are small and they are singlet component is shown in Fig. 12. It can be seen that
omitted now. for the parameters chosen the LRTC is larger than the
singlet component and decays much slower with increas-
Then from the matching conditions at the S / F inter-
ing the distance x .
face, Eq. 3.48, we find the following relations for the
coefficients: If the magnetization vector M rotates by an angle uma
domain wall over a small length w so that Q /w
f BCS
b 3± = 3,58 w , then the ratio in the first set of large parentheses in
2F± Eq. 3.60 is equal to
e Q
Q / Q sinh 0.087,
Q sinh w + cosh w
b Q = − b¯ Q = Q / Q b 3+ − b 3− 3.59
3,62
the parameter F given by Eq. 3.42 .
One can see from the above equations that the con- which shows that the amplitude of the LRTC in this case
is smaller than the amplitude of the singlet component.
densate function f is small provided the parameter
F ± is large. It follows from Eq. 3.59 that the ampli-
tude of the LRTC, b Q , is not zero only if the magnetiza-
tion is nonhomogeneous, ie, Q 0.
Matching the solutions 3.54–3.57 at x = w , we find
for the amplitude of the LRTC
if BCS Q h sgn / D F
c=− ,
2F cosh w 2 Re +
Q sinh w + +
3.60
where w = Qw is the total angle of the magnetization
rotation. As has been mentioned, the amplitude of the
LRTC is an odd function of . FIG. 12. Spatial dependence of amplitudes of singlet dashed
As one can see from the last expression, the ampli- line and triplet solid line, components of the condensate
tude c increases from zero when increasing Q , reaches function in the F wire for different values of w . Here w
a maximum at Q max corresponding a certain angle max , = L /5, = E T , and h / E T =400. E T = D / L 2 is the Thouless en-
and then exponentially decreases at w max . ergy. From Bergeret et al. , 2001a.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 23
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1343

We can conclude from this analysis that in order to get mesoscopic structure. The authors considered a struc-
a large LRTC, a small total angle of rotation of the mag- ture with a thin normal-metal layer N and a ferromag-
netization vector is preferable. netic insulator F . The density-of-states variation in a
The amplitude of the condensate function calculated conventional superconductor which is in contact with a
here enters different physical quantities. In Sec. III.D we ferromagnetic insulator was analyzed by Tokuyasu et al.
discuss how the long-range penetration of the triplet 1988.
component into the ferromagnet affects transport prop- Eschrig et al. 2003 considered a clean S / F / S Joseph-
erties of F / S structures. son junction in which the ferromagnet F was a half metal
It is interesting to note that the type of magnetic so that the electrons with only one spin orientation say
structure discussed in this section differs drastically from the spin-up ↑ electrons existed in the ferromagnet. Dentro
the one in the case of an in-plane rotating magnetiza- this case only the triplet component corresponding to
ção. The latter was considered recently by Champel and the condensate function ↑ ↑ may penetrate the ferro-
Eschrig 2005a, 2005b. It was assumed that the magne- magnet. Assuming the p -wave triplet condensate func-
tization vector M F was parallel to the S / F interface tion, the authors have calculated the critical Josephson
e rotates; that is, it has the FormatoM F current I c . They showed that the state negative criti-
= M 0 0,sin Qy ,cos Qy the x axis is normal to the S / F cal current I c is possible in this junction. The dc Joseph-
interface plane. As shown by Champel and Eschrig son effect in a junction consisting of two superconduct-
2005b, the odd triplet component also arises in this ors and a spin-active interface between them was
case but it penetrates into the ferromagnetic region over analyzed by Fogelström 2000.
a short distance of the order of h . It would be of interest to analyze the influence of im-
purities on the critical current in such Josephson junc-
3. Spin-active interfaces tions because, as we noted, in the clean case the singlet
component can penetrate the ferromagnet not a half
In almost all papers discussing S / F structures it is as-
metal over a large distance.
sumed that the S / F interface is spin inactive, ie, the
spin of an electron does not change when the electron
goes through the interface. D. Long-range proximity effect
Although in many cases this is really so, one can imag-
In the last decade transport properties of mesoscopic
https://translate.googleusercontent.com/translate_f 19/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
ing another
ine through situation
the interface
whenchanges.
the spinOne
of an
can
electron
consider
pass-
a superconductor–normal-metal S / N structures were in-
region with a domain wall at the interface also as a tensively studied see, for example, the review articles
“spin-active interface” provided the width w of the do- by Beenakker 1997, Lambert and Raimondi 1998,
main wall is very small but the product Qw is of the and references therein. In the course of these studies
order unity. As we have seen in Sec. III.C.2, at this type many interesting phenomena have been discovered.
of interface the triplet condensate arises. Among them is a nonmonotonic voltage and tempera-
Boundary conditions at spin-active S / F interfaces for ture dependence of the conductance in S / N meso-
quasiclassical Green's functions were derived in a num- scopic structures, ie, structures whose dimensions are
ber of publications Millis et al. 1988; Kopu et al. 2004 less than the phase coherence length L and the
and were used in studying different problems. Kulic and inelastic-scattering length l . This means that the resis-
Endres 2000 employed these boundary conditions in tance R of a S / N structure changes nonmonotonically
the study of a system similar to the one shown in Fig. 8. when the temperature decreases below the critical tem-
Contrary to Bergeret et al. 2003, they assumed that the perature T c .
ferromagnets F are insulators so that the condensate This complicated behavior is due to the fact that there
does not penetrate them. Nevertheless, the calculated are two contributions to the resistance in such systems:
critical temperature T c of the superconducting transition one coming from the S / N interface resistance and the
depends on the mutual orientation of the magnetization resistance of the normal wire itself. The experimentally
M F in the ferromagnets. In accordance with the works of observed changes of the resistance can be both positive
Tagirov 1998, Fominov et al. 2002, and Baladie and R 0 and negative R 0 Shapira et al. 2000;
Buzdin 2003 in which metallic ferromagnets were con- Quirion et al. , 2002. The increase or decrease of the
sidered in a F / S / F structure, Kulic and Endres found resistance R depends, in particular, on the interface re-
that the critical temperature T c was maximal for the an- sistance R S / N . If the latter is very small, the resistance of
tiparallel magnetization orientation. If the directions of the S / N structure is determined mainly by the resistance
magnetization vector M F are perpendicular to each of the N wire R N . This resistance decreases with decreas-
other, a triplet component also arises in the supercon- ing temperature T , reaches a minimum at a temperature
ductor. The authors considered a clean case only, so that of the order of the Thouless energy D N / L N 2 , and in-
the influence of impurity scattering on the triplet com- creases again returning to the value in the normal state
ponent remained unclear. R N T c at low T , where D N is the diffusion coefficient
According to Huertas-Hernando et al. 2002, a spin- and L N is the length of the N wire. This is the so-called
active N / F interface plays an important role in the ab- reentrant behavior observed in many experiments
solute spin-valve effect which can take place in a S / N / F Gubankov and Margolin, 1979; Pothier et al. 1994; Di-

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 24
1344 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

moulas et al. , 1995; Petrashov et al. , 1995; Charlat et al. ,


1996; Chien and Chandrasekhar, 1999; Shapira et al. ,
2000.
Theoretical explanations for the nonmonotonic be-
havior of the resistance variation as a function of the
temperature T or voltage V in S / N structures were pre-
sented by Artemenko et al. 1979, Volkov et al. 1993,
1996, Nazarov and Stoof 1996, Golubov et al. 1997,
and Shapira et al. 2000. Such a variation of the resis-
tance of the normal-metal wire can be explained in
terms of the proximity effect that leads to the penetra- FIG. 13. Reduction of the resistance of a Ni wire attached to a
tion of the condensate into the N wire. Due to this pen- superconductor Al. Adapted from Petrashov et al. 1999
etration there are two types of contributions to the con-
ductance G N Volkov and Pavlovskii, 1996; Golubov et
al. , 1997. One of them reduces the density of states in by Petrashov et al. 1999. According to estimates of F
the N wire and therefore reduces the conductance G N . performed in this experiment, the observed R F is two
The other term, similar to the Maki-Thompson term orders of magnitude larger than expected from the con-
Volkov and Pavlovskii, 1996; Golubov et al. , 1997, ventional theory of S / F contacts. Therefore these results
leads to an increase of the conductance of the N wire. cannot be explained in terms of the penetration of the
In principle, the magnitude of the conductance varia- singlet component.
tion G N may be comparable with the conductance G N . In Fig. 14 we show similar data from the experiment
So there are no doubts that the proximity effect plays a on Co/Al structures performed by Giroud et al. 1998.
very important role in many experiments on S / N struc- In this experiment reentrance behavior of R was ob-
turas. served. In the limit of very low temperatures T→ 0 the
Recently, similar investigations have also been carried resistance was even larger than in the normal state.
out on mesoscopic F / S structures in which ferromagnets The final explanation of this effect remains unclear.
F were used instead of normal nonmagnetic metals. However, long-range proximity effects considered in the
According to our previous discussion, the depth of the previous sections may definitely contribute to the con-
condensate penetration into an impure ferromagnet ductance variation. In order to support this point of view
equals F = D / h . This length is extremely short 5–50 we analyze qualitatively the changes of the conductance
Å for strong ferromagnets such as Fe or Ni. Assim sendo due to the LRTC penetration into the ferromagnet and
one might expect that the influence of the proximity ef- demonstrate that the LRTC may lead to the conduc-
fect on the transport properties of such structures should tance variation comparable to that observed in the ex-
be negligibly small. perimentos.
It was a great surprise that experiments carried out However, before presenting these calculations it is
recently on F / S structures showed that the resistance reasonable to ask if one can explain the experiments in a
more simple way. Actually, the resistance of the S / F
variations R were quite visible varying from about 1%
structures was analyzed in many theoretical works. Para
to 10% when decreasing the temperature below T c
Lawrence and Giordano, 1996a, 1996b; Giroud et al. , example, de Jong and Beenakker 1994, Golubov
1999, and Belzig et al. 2000 analyzed a ballistic S / F
1998; Petrashov et al. , 1999; Aumentado and Chan-
drasekhar, 2001. For example, in the experiments by contato. It was shown that at zero exchange field h =0
https://translate.googleusercontent.com/translate_f 20/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
Lawrence and Giordano 1996a, 1996b in which a the contact conductance G F / S is twice as large as its con-
Sn/Ni structure was studied, the effective condensate ductance G F / N in the normal state above T c , as it
penetration length estimated from the measured resis-
tance was about 400 Å. This quantity exceeds F by an
order of magnitude. Similar results have been obtained
by Giroud et al. 1998 on Co/Al structures, by Pe-
trashov et al. 1999 on a Ni/Al structures, and by Au-
mentado and Chandrasekhar 2001 on Ni/Al structures.
It is worth mentioning that the change of the resis-
tance was both positive and negative. In some experi-
ments the variation R F was related to a change of the
interface resistance Aumentado and Chandrasekhar,
2001, whereas in others Lawrence and Giordano,
1996a, 1996b; Giroud et al. 1998; Petrashov et al. 1999
to the resistance variation of the ferromagnetic wire FIG. 14. Temperature dependence of the resistance of a Co
RF. wire attached to a superconductor Al measured by Giroud et
In Fig. 13 we show the temperature dependence of the al. 1998. Note that at low temperatures the authors observed
resistance of a Ni wire attached to an Al bank measured a reentrance behavior.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 25
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1345

should be. This agrees with the conductance in a N / S


ballistic contact according to theoretical predictions. No
the same time, it drops to zero at h = E F , where E F is the
Fermi energy.
The conductance of a diffusive point contact G F / S has
been calculated by Golubov 1999, who showed that
G F / S was always smaller than the conductance G F / N in
o estado normal. In the case of a mixed conductivity
mechanism partly diffusive and partly ballistic the con-
ductance G F / S was calculated by Belzig et al. 2000. Ac-
cording to their calculations it can be both larger or
FIG. 15. The cross geometry used to measured the changes of
smaller than the conductance in the normal state G F / N . the resistance of a F wire due to the proximity effect.
The resistance R F of a ferromagnetic wire attached to
a superconductor was calculated by Falko et al. 1999,
Jedema et al. 1999, and Bergeret, Pavlovskii, et al. and Mélin and Feinberg 2004 on the basis of the tunnel
2002. Let us briefly describe what happens in such a Hamiltonian method. The influence of superconducting
sistema. contacts on giant magnetoresistance in multilayered
The proximity effect was neglected in these works but structures was studied by Taddei et al. 2001. Tkachov et
a difference in the conductivities ↑↓ for spin-up and al. 2002 studied an enhancement of Andreev reflection
spin-down electrons was taken into account. O troco at the S / F interface due to inelastic magnon-assisted
of the conductance or resistance G F is caused by a scattering. In a recent publication, Marten et al. 2005
different form of the distribution functions below and studied the spin and charge transport in F/S/F structures
above T c because of Andreev reflections. on the basis of the Usadel equation.
The conductance G F T c of the F wire in the normal One can conclude from the works listed above that by
state TT c is given by the simple expression neglecting the penetration of the LRTC into the F wire,
an increase in the conductance G F cannot be explained.
GFTc=G↑+G↓, 3,63 Therefore let us discuss the consequences of the LRTC
where G ↑↓ = ↑↓ L F A , with L F , and A the length and penetration into the ferromagnetic wire. A fim de
cross-section area of the F wire. avoid the S / F interface contribution to the total resis-
This means that the total conductance is the sum of tance, we consider a cross geometry see Fig. 15 and
the conductances of the spin-up and spin-down chan- assume that the resistance of the interface between the F
nels. In this case not only the electric current but also wire and F reservoirs is negligible. Such a geometry was
the spin current is not zero. It turns out that below used, for example, in the experiments by Petrashov et al.
T c TT c the conductance decreases and at zero tem- 1995. The structure under consideration consists of two
perature it is equal to F wires attached to the F and S reservoirs. Nós presumimos
that there is a significant mismatch between parameters
GF0=4G↑G↓/G↑+G↓. 3,64 of the superconductor and ferromagnet so that the con-
Equation 3.64 shows that the zero-temperature con- densate amplitude induced in F is small and is deter-
ductance G F 0 for the system considered is smaller than mined by Eq. 3.37 or 3.60.
the normal-state conductance G F T c . According to our results obtained previously, the
It is possible to obtain the explicit formulas not only long-range proximity effect is possible provided there is
in the limiting cases, Eqs. 3.63 and 3.64, but also to a domain wall near the interface between the supercon-
describe the system at arbitrary temperatures. The gen- ductor and ferromagnet and we assume this for the
eral formula for the conductance of the F wire can be setup shown in Fig. 15. Another possibility for generat-
written as ing the triplet condensate would be to attach to the su-
perconductor an additional ferromagnet with a noncol-
G F T = G F 0tanh/2 T + G F T c 1 − tanh/2 T . linear magnetization.
3,65 The conductance can be found on the basis of a gen-
eral formula for the current see, for example, the book
Equations 3.63 and 3.64 can be obtained from Eq. by Kopnin 2001 and Appendix A,
3.65 by setting or T equal to zero. Equações
3.63–3.65 are valid provided the length L F satisfies I = 1/16 e L y L z
the condition F Tr ˆ 0

l ↑↓ LF L SO , L in , 3.66 ˆ 3 dg R x g K + g Kx g A , 3.67

https://translate.googleusercontent.com/translate_f 21/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras

where l ↑↓ is the mean free path of spin-up and spin-down


electrons, while L SO and L in are the spin-orbit and in- where F is the conductivity of the F wire in the normal
elastic relaxation length, respectively. Estado.
The resistance of multiterminal S / F structures was The matrix Green's function g K = g R Fˇ − Fˇ g A is the
calculated by Mélin 2001, Mélin and Peysson 2003, Keldysh function related to a matrix distribution func-

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 26
1346 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

tion Fˇ . The distribution function consists of two parts;


one of them is antisymmetric with respect to the energy
, the other one is symmetric in and determines the
dissipative current.
In the limit of a weak proximity effect the retarded
advanced Green's function has the form

g RA ± ˆ 3 ˆ 0 + f RA , 3,68
where f R A is given by Eq. 3.37 or 3.60.
We have to find the conductance of the vertical F wire
in Fig. 15. In the main approximation the distribution FIG. 16. The GT dependence. Here = h / R b , / E T 1,
function in this F wire is equal to and w / L =0.05. From Bergeret et al. , 2001a.
Fˇ = F 0 · ˆ 0 ˆ 0 + F 3 · ˆ 3 ˆ 0 , 3,69
where F 0,3 =tanh + V /2 T ±tanh − V /2 T . this leads to a difference in transport properties. Em um
The distribution function F 3 symmetric in deter- S / N system, the relation f R = f A =0 holds and this
mines the current I . The differential conductance G d equality is a consequence of the time-reversal symmetry.
= dI / dV can be represented as É por isso que G 0= GT c =0 in S / N structures,
Gd=G0+G, 3.70 whereas in a S / F structure f R fA =0 and why
ST T =0 0
where G 0 = F L F A is the conductance in the normal Although the LRTC may be the reason for the en-
state here for simplicity we neglect the difference be- hancement of the conductivity in the S / F structures this
tween ↑ and ↓ . possibility was also pointed out by Giroud et al. 2003 ,
The normalized correction to the conductance due to our understanding is based on the assumption that the
the proximity effect ST G / G 0 can be found using a magnetic moment is fixed and does not change with tem-
general formula Bergeret et al. , 2001a perature. Dubonos et al. 2002 suggested another
mechanism based on the assumption of a domain redis-
ST = − 32 T −1 Tr ˆ 0 df R − f A 2 F V , 3.71 tribution when the temperature drops below T c . The fer-
romagnetic wires or strips used in different experi-
Onde ments may consist of many domains. Their form and
number depend on the sample geometry and parameters
FV = cosh −2 + eV /2 T + cosh −2 − eV /2 T /2. do sistema. When the temperature decreases below
The angle brackets ¯ denote the average over the T c , stray magnetic fields excite the Meissner currents in
length of the ferromagnetic wire between the F or N the superconductor attached to the F wire. Assim sendo
the demagnetizing factors change, which may lead to a
reservatórios. The functions f R A are given by expressions new domain structure. At the same time, the total con-
similar to Eq. 3.60. This formula shows that if T ductance or resistance G F depends on the form and the
D F / L 2 , on the order of magnitude ST f tr 2 , where number of domains. So one might expect that the con-
L is the length of the ferromagnetic wire and f tr is the ductance G F T below T c would differ from G F in the
amplitude of the triplet component at the S / F interface normal state. This idea was supported by measurements
at a characteristic energy ch min T , D F / L . De acordo carried out by Dubonos et al. 2002. In this work a
to Eq. 3.60 the amplitude of the triplet component is of structure consisting of a two-dimensional electron gas
the order of c 1 h / R b , where is the resistivity of the and five Hall probes was used. An F / S system Ni+Al
ferromagnet and c 1 is determined by the factor in the disks was placed on top of this structure. By measuring
first term in large parentheses, that is, by the character- the Hall voltage, the authors were able to probe local
istics of the domain wall. In principle the amplitude f tr magnetic fields around the ferromagnetic disks. Eles
may be on the order of 1. found that these fields really change when the tempera-
Strictly speaking, both the singlet and triplet compo- ture dropped below T c .
nents contribute to the conductance. However, if the On the other hand, the Meissner currents and hence
length L F exceeds the short length F , only the contribu- the redistribution of the domain walls may be consider-
tion of the LRTC is essential. ably reduced in wires, as discussed previously. Mudando
In Fig. 16 we present the temperature dependence of the thickness of the superconducting wires in a con-
the correction to the conductance GT . Pode ser visto trolled way and measuring the conductance could help
that by increasing the temperature G F T decreases in to distinguish experimentally between the contribution
a monotonous way. This dependence differs from the of the triplet condensate to the conductivity and the re-
reentrant behavior discussed above that occurs in the distribution of domain walls.
S / N structures. The reason for this difference is that the An experiment in which the domain redistribution
time-reversal symmetry in S / F structures is broken and was excluded was performed by Nugent et al. 2004.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 27
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1347

https://translate.googleusercontent.com/translate_f 22/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras

The authors measured the resistance variation of a fer- junction observed in many experiments and discussed in
romagnetic wire Ni 1− x Cu x lowering the temperature T Sec. II.B.2. Another manifestation of the transition from
below the critical temperature T c of the superconductor the 0 state to the state is the sign reversal of the criti-
Al or Pb, which was attached to the middle part of the cal current observed in the experiment by Ryazanov et
ferromagnetic wire. A magnetic field, strong enough to al. 2001 on Nb/Cu x Ni 1− x /Nb Josephson junctions see
align all domains in the ferromagnet in one direction but Fig. 5. The proper choice of an alloy with a weak ferro-
not too strong to suppress the superconductivity, was ap- magnetic coupling was crucial for the observation of the
plied to the system. Under these conditions a small in- efeito.
crease in the resistance R / R 310 −3 was observed Subsequent experiments, Blum et al. 2002, Kontos et
when the temperature T drops below T c . A análise al. 2002, and Guichard et al. 2003, corroborated the
presented above shows that the triplet component leads observed sign change of the Josephson coupling when
to an increase of the conductance but not in the resis- varying the thickness of the intermediate F layer. Quali-
tance of the ferromagnetic wire. Therefore this particu- tatively, the experimental data on the Josephson effect
lar experiment can hardly be explained in terms of the in the S / F / S structures are in agreement with the theo-
long-range proximity effect. Perhaps the small increase retical works mentioned above. However, a more accu-
in the resistance of the ferromagnetic wire observed by rate control and understanding of the 0- transition de-
Nugent et al. 2004 was related to the “kinetic” mecha- mands knowledge on the magnetic structure of the
nism discussed above see Eq. 3.65 or to weak- ferromagnetic materials.
localization corrections caused by the triplet Cooperons In almost all theoretical papers very simplified models
McCann et al. , 2000. According to McCann et al. 2000 of the S / F / S junction were analyzed. Por exemplo,
the change of the resistance of the ferromagnetic wire is Blanter and Hekking 2004 assumed that the F layer
positive contrary to the contribution of the LRTC and consisted of either one domain or two domains with the
its order of magnitude is e 2 / R F , where R F is the resis- collinear orientation of the magnetization. Nesse caso
tance of the F wire in the normal state. A fim de and according to the discussion of Sec. III.C, the LRTC
clarify the role of the LRTC in the transport properties is absent in the system.
of S / F structures, further theoretical and experimental If the F layer is a single domain layer, the critical cur-
investigations are needed. Note that strong ferromag- rent I c is maximal at a nonzero external magnetic field
nets such as Fe are not suitable materials for observing H ext equal to −4 M F , where M F is the magnetization of
the contribution of the LRTC into the conductance the F layer. At the same time, in experiments Strunk et
variation because of the strong exchange field h . Nisso al. 1994; Kontos et al. , 2001, 2002; Ryazanov et al. 2001;
case, according to Eqs. 3.34 and 3.60, the amplitude Blum et al. 2002; Sellier et al. , 2004 a decrease of the
of the LRTC is small because it contains h in the de- current I c with increasing field H ext was observed and it
nominator. was maximal at H ext =0. This means, as assumed in these
experimental works, that the F layer in real junctions
IV. JOSEPHSON EFFECT IN S / F SYSTEMS contains many magnetic domains. In this case the Jo-
INHOMOGENEOUS MAGNETIZATION sephson critical current I c may change sign in S / F / S
junctions with a multidomain magnetic structure even if
As we have mentioned above, one of the most inter- the local Josephson current density j c is always positive.
esting issues in S / F structures is the possibility of switch- The reason for the sign reversal of I c in this case is a
ing between the so-called 0 and states in Josephson spatial modulation of the phase difference x due to
S / F / S junctions. o state denotes the state for which an alternating magnetization M x in the domains
the Josephson critical current I c becomes negative. este Volkov and Anishchanka, 2004. In order to determine
occurs for a certain thickness d F and temperature T . Dentro the mechanism that leads to the sign reversal of the criti-
this state the minimum of the Josephson coupling en- cal current further experiments are needed.
ergy E J = I c / e 1−cos corresponds to a phase differ- In this section we discuss a new phenomenon, namely,
ence of = but not to =0 as in conventional Joseph- how the Josephson coupling between the F layers in the
son junctions. S / F structures is affected by the LRTC.
The reason for the sign reversal of I c is the oscillatory First, we consider a planar S / F / S Josephson junction
dependence of the condensate functions f on the thick- with a ferromagnet magnetization M F rotating in the di-
ness d F see Eq. 2.37 . Since the critical current I c is rection normal to the junction plane. This model is an
sensitive to the phase of the condensate function at the idealization of a real multidomain structure with differ-
boundary, the state is a rather natural consequence of ent magnetization orientations. In this case, as discussed
the oscillations. in Sec. III.D, the LRTC arising in the structure strongly
The possibility of the state was predicted by Bula- affects the critical current I c .
evskii et al. 1977 and Buzdin et al. 1982, and studied Next, we shall analyze a multilayered S / F / S /¯ struc-
later in many other works eg, Radovic et al. 1991; ture in which the vector M F has a different direction in
Buzdin, Vujicic, and Kupriyanov, 1992. Experimentally, different F layers. Again, in this case the LRTC arises in
this phenomenon manifests in a nonmonotonic depen- the system. Interestingly, if the thickness of the F layers
dence of the critical temperature on the thickness of the d F is much larger than the penetration length F of the

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 28
1348 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

singlet component but less than or on the order of N , The presence of the second term in Eq. 4.3 is espe-
then Josephson coupling between the F layers is realized cially interesting in the case when the thickness d F of the
due to the LRTC odd triplet superconductivity in the ferromagnetic spacer between the superconductors
transverse direction. At the same time, the in-plane su- obeys F d F −1 eu
. Then the main contribution to the
perconductivity is due to the conventional singlet super- Josephson coupling comes from the long-range triplet
conducting pairing. component of the condensate. Another important fea-
Finally, we shall discuss the dc Josephson effect in a ture of this limit is that for sufficiently large values of Ql
SF / I / FS junction here SF is a superconductor- the critical current is always positive no possibility for
ferromagnet bilayer and I is a thin insulating layer. Dentro contato. This can be seen from Fig. 7.
this structure the exchange field may lead not only to a The fact that the superconductivity loses its “exotic
suppression of the Josephson coupling as one would na- properties” at large Q is quite understandable. The su-
ively expect but, under a certain condition, to its en- perconductivity is sensitive not to the local values of the

https://translate.googleusercontent.com/translate_f 23/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
hancement.
Let us consider first a planar S / F / S Josephson junc- exchange
order of thefield but to its average
superconducting on the scales
coherence length.ofSethe
o
ção. We assume the following spacial dependence of exchange field oscillates very fast such that the period of
the magnetization vector in the F layer: M F the oscillations is much smaller than the superconduct-
= M F „0,sin Qx ,cos Qx …, where the x axis is normal to ing coherence length, its average on this scale vanishes
the junction plane. and therefore all new properties of the superconductiv-
In this case, as seen in Sec. III.C.2, the LRTC arises. ity originating from the presence of the exchange field
Due to the long-range penetration into the ferromagnet become negligible.
the triplet component can give a contribution to the Jo- To conclude we summarize the results known for
sephson current. A general expression for the Josephson S / F / S Josephson junctions. When the magnetization in
current can be written in the form the ferromagnetic F is homogeneous, we have to distin-
guish between two different cases.
I J = L y L z /4 e F T i Tr ˆ 0 ˆ 3 · fxf. 4.1 In the dirty limit h 1 the change of sign of the
critical current occurs if the thickness of the F layer d F is
We assume that the impurity concentration is sufficiently on the order of F . The condensate function in the F
layer decays exponentially over this h and oscillates
high and therefore the condensate function f should be
with the same period.
determined from the Usadel equation. In the limit of a
In the opposite clean limit, h 1, the condensate
weak proximity effect the S / F interface transparency is
function oscillates in space with the period v F / h and de-
not too high this equation can be linearized and solved
cays exponentially over the mean free path l .
exatamente. The solution for the f matrix in the F region Finally, if the ferromagnetic region contains a domain
can be found in a similar way as done in Sec. III.C.2. wall described by a vector Q , a long-range component of
Due to rotation of the magnetization, the condensate the condensate appears. It decays in the F film over a
function contains the LRTC. We obtain for the Joseph-
considerably larger length on the order of N = D /2 T
son current Bergeret et al. , 2001c the following expres- that can greater exceed the characteristic length
sion:
D / h in a homogeneous F layer Q =0. Nesse caso
I J = I c sin , 4,2 the coupling between the superconductors survives even
if the thickness of F is larger than F .
where the critical current I c is equal to It is clear that the presence of a domain wall between
the superconductors is something that cannot be con-
Ql 2 e −
I c = L y L z F / l ˜ F2 Re f 2s e − d +
d
trolled very well experimentally. Therefore in the next
l F
+ F
,
0 +l 23 h 3/2 section we discuss a possible experiment on S / F multi-
4,3 layered structures that may help in detecting the LRTC
by measuring the Josephson critical current.
and l 2 =2 n /+ Q 2 . The parameter ˜ F =3/4 T é
an effective, averaged over angles, transmittance coeffi-
A. Josephson coupling between S layers via the triplet
cient which characterizes the S / F interface transparency componente
and + is defined in Eq. 3.52.
The first term in the brackets in Eq. 4.3 containing In this subsection we analyze another type of multi-
the parameter + corresponds to Eq. 2,36. It decays by layered S / F structure in which the LRTC arises. Isto é um
increasing the thickness d F over the short characteristic multilayered periodic ¯/ S / F n −1 / S / F n / S / F n +1 / S /¯
length F = D F / h and can change sign. The second term structure with alternating magnetization vector M F , n in
in Eq. 4.3 originates from the rotation of h along the x different F layers. We assume that the vector M F , n is
eixo. It decays with the thickness d F over another char- rotated with respect to the vector M F , n −1 by an angle 2 ,
acteristic length l −1 that can be much larger than F . such that the angle increases monotonously with increas-
Therefore this term results in a drastic change of the ing n . We call this type of magnetization the one with a
critical current. positive chirality.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 29
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1349

In an infinite system the magnetization vector M F av- d F is assumed to be much larger than F as usual, we
eraged over n is equal to zero it rotates when one assume that F N , the Josephson coupling between
moves from the n th to the n +1th, layer, etc.. Another the S layers is solely due to the LRTC. So in such sys-
type of chirality negative chirality is when the angle tems we come to a new type of superconductivity: an
between vectors M F , n and M F , n −1 is equal to 2 −1 n . Dentro odd triplet out-of-plane superconductivity and conven-
this case the averaged vector M F is not zero. tional singlet in-plane superconductivity the triplet
In Sec. III.C.1 we have seen that in a F / S / F structure component gives only a small contribution to the in-
with a noncollinear orientation of the magnetization plane superconductivity.
vectors in the F layers the LRTC arises. If one assumes Using Eqs. 3.21–3.29 one can perform explicit cal-
that the thickness of the F layers d F is larger than the culations for this case without considerable difficulties.
coherence length in the normal metal N , the overlap of As a result, the Josephson critical current I c can be writ-
the condensate functions created in a F layer by neigh- ten as follows Bergeret et al. , 2003:
boring S layers is weak, and the solutions given by Eqs.
3.21–3.29 remain valid for the multilayered S / F struc- eR F I c = ±2 T d F b2 1 1 + tan 2 e − d , 4,4 F

ture.
Using these solutions one can calculate the Josephson
current between neighboring S layers. As the thickness Onde

˜ 2S ˜ + − ˜ − sgn
b1 = − f BCS sin ,
cosh 2 S M + T − + M − T + g BCS + Ftanh F

https://translate.googleusercontent.com/translate_f 24/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras

S = s d S , F = d F , ˜ ± = ± / g BCS + F ± , ˜ = / g BCS A similar dependence of the Josephson current I c on


+F tanh F , ˜ S = S / g BCS , and chirality was predicted by Kulic and Kulic 2001 in a
Josephson junction S m IS m I is an insulator between
M ± = T ± ˜ S coth S + ˜ tanh F
two magnetic superconductors S m . For the magnetic su-
+ tan 2 C ± ˜ S tanh S + ˜ tanh F , perconductors considered in that work, the magnetiza-
tion vector M rotated with the angle of rotation equal to
T ± = ˜ S tanh S + ˜ ± , = x Q · n x , where Q is the wave vector of the x depen-
dence of the angle , and n x is the unit vector normal to
C ± = ˜ S coth S + ˜ ± . the insulating layer I . Therefore chirality or spiral he-
R F is defined as R F =2 d F / L y L z F . Equation 4.4 de- licity, in terms of Kulic and Kulic in this case is deter-
scribes the layered systems with both positive + sign mined by the sign of the product Q R · Q L , where Q L , R is
and negative − sign chirality. the wave vector in the left right magnetic supercon-
One can see from Eq. 4.4 that in the case of positive ductor.
chirality the critical current is positive, while if the However, there is an essential difference between the
chirality is negative the system is in the state negative multilayered S / F structure discussed here and magnetic
atual. This means that by changing the configuration
of the magnetization, one can switch between the 0 and
Estado.
It is important to emphasize that the nature of the
contact obtained here differs from that predicted by Bu-
laevskii et al. 1977 and observed by Ryazanov et al.
2001. In the latter cases the transition is due to the
change of the values of either the exchange field, the
temperature, or the thickness of the F film. No caso
considered in this section, negative Josephson coupling
originates from the presence of the triplet component
and can be realized in S / F structures with negative
chirality. Since for positive chirality the Josephson cur-
rent is positive, the result obtained gives a unique oppor-
tunity to switch experimentally between the 0 and FIG. 17. Dependence of the critical current on the angle .
states by changing the angles of the mutual magnetiza- The value of the current is given in arbitrary units. From Berg-
tion of the layers. eret et al. , 2003.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 30
1350 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

FIG. 19. The S / F / I / F / S system. I is an insulating thin layer.


FIG. 18. Schematic: Measurement of the change of the densityThe relative magnetization of F layers can be switched.
of states at the outer F interface by tunneling spectroscopy.
Kontos et al. 2001 performed such experiments on S / F struc-
turas. though the LRTC is not essential for the critical current
enhancement, the short-range triplet component arises
in this case and it plays a certain role in this effect. En-
superconductors. In magnetic superconductors with a hancement of the Josephson current in S / F / I / F / S tun-
spiral magnetization the triplet component also exists nel structures I stands for an insulating layer, see Fig.
but in contrast to S / F structures, the singlet and triplet 19 was predicted by Bergeret et al. 2001b and further
components cannot be separated. In particular, in the considered in a subsequent work by Golubov et al.
case of a collinear alignment of M , the Josephson cou- 2002b. As we shall see below, if the temperature is low
pling in S / F structures with thick F layers disappears, enough and the S / F interface transparency is good, one
whereas it remains finite in the S m IS m system. can expect an enhancement of the critical current with
Figure 17 shows the dependence of the Josephson cur- increasing the exchange field provided the magnetiza-
rent I c on the angle given by Eq. 4.4. If the mutual tions of the F layers are antiparallel to each other.
orientation of M is parallel =0 or antiparallel 2 This surprising result can be obtained in the limit
= , the amplitude of the triplet component is zero and when the thicknesses d S and d F of the S and F layers are
therefore there is no coupling between the neighboring smaller than the superconducting coherence length S
S layers, ie, I c =0. For any other angles between the D /2 T c and the penetration length of the conden-
magnetizations the amplitude of the triplet component is sate into the ferromagnet F D / h , respectively. Nisso
finite and this leads to a nonzero critical current. No case one can assume that the quasiclassical Green's func-
2 = /2 perpendicular orientation of M , the Joseph- tion does not depend on the space coordinates and, in
son current I c reaches its maximum value. particular, the superconducting order parameter é um
Another possible experiment for detecting the long- constant in space. Moreover, instead of considering the
range triplet component is the measurement of the den- dependence of the exchange field h on the coordinates
sity of states in the F / S / F system as shown in Fig. 18. one can replace it by a homogeneous effective exchange
Kontos et al. 2001 determined the spatial changes of field h eff with a reduced value. Therefore in our calcula-
the DOS in a PdNi/Al structure with the help of planar tions we use effective fields eff and h eff defined as
tunneling spectroscopy. This method could also be used
to detect the LRTC. It is clear that if the thickness of the
https://translate.googleusercontent.com/translate_f 25/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
F layer in Fig. 18 is larger than the penetration of the eff / = S d S S d S + F d F−1 , 4,5
short-range components, then any change of the DOS at
h eff / h = F d F −1 , 4,6
the outer boundary of the F layer may occur due only to SdS+FdF
the long-range penetration of the triplet component. E se
where S and F are the densities of states in the super-
the magnetizations of both F layers are collinear no ef-
conductor and ferromagnet, respectively.
fect is expected to be observed, while for a noncollinear
With this simplification, we can write the Gor'kov
magnetization a change of the DOS should be seen.
equations for the normal and anomalous Green's func-
tions in spin space as
B. Enhancement of the critical Josephson current
i + − h Gˆ + ˆ Fˆ + = 1, 4,7
Another interesting effect in S / F structures that we
will discuss is the enhancement of the Josephson critical
current by the exchange field. The common wisdom is − i + − h Fˆ + ˆ Gˆ = 0, 4,8
that any exchange field should reduce or destroy singlet
superconductivity. In the previous sections we have seen where = ˆ 1 , ˆ 2 , ˆ 3 are the Pauli matrices and = p
that this is not always so and superconductivity can sur- − F , F is the Fermi energy, p is the spectrum, and
vive in the presence of a strong exchange field. But still, =2 n +1 T are Matsubara frequencies. We omit the
it is not so simple to imagine that the superconducting subscript ''eff'' in Eqs. 4.7 and 4.8 and below.
properties can be enhanced by the exchange field. In order to calculate the Josephson current I J through
Surprisingly, this possibility exists and we shall dem- the tunnel junction represented in Fig. 19 we use the
onstrate now how this unusual phenomenon occurs. Al- well-known standard formula

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 31
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1351

I J = 2 T / eR Tr fh 1 fh 2 sin , 4,9
n
Onde
Eu
f= Fˆ d 4,10

is the quasiclassical anomalous Green's function, is the


phase difference between both the superconductors, R is
the barrier resistance, and h 1,2 are the exchange fields of FIG. 20. Dependence of the normalized critical current on h
the left and right F layers. for different temperatures in the case of a parallel orientation.
The only difference between Eqs. 4.9 and 4.10 and Here eV c = eRI c , h F is the effective exchange field, t = T / 0 , and
the corresponding equations in the absence of the ex- 0 is the superconducting order parameter at T =0 and h =0.
change field is the dependence of the condensate func-
tion f on h . This dependence can be found immediately exchange field for the antiparallel configuration at
from Eqs. 4.7 and 4.8, low temperatures, which is a new intriguing effect see
Fig. 21.
f= ˆ + i h 2 + 2 −1/2 . 4,11 This unexpected result can be understood from Eq.
4.14 rather easily without numerical calculation. No
What remains to be done is to insert the condensate limit T→ 0 the sum over the Matsubara frequencies can
function f into Eq. 4.9 for certain exchange fields h 1 be replaced by an integral and one can take for the su-
and h 2 and to calculate the sum over the Matsubara fre- perconducting order parameter the values = 0 if h
quencies . Although it is possible to carry out these
0 , and =0 if h 0 , where 0 is the BCS order pa-
calculations for arbitrary vectors h 1 and h 2 , we restrict rameter in the absence of an exchange field see, eg,
our consideration to the cases when the absolute values Larkin and Ovchinnikov 1964 .
of the magnetizations h 1 and h 2 are equal but the mag- Inserting this solution into Eq. 4.14 one can see that
netizations are either parallel or antiparallel to each the Josephson critical current I c uma increases with increas-
de outros. This simplifies the computation of the Josephson ing exchange field. Moreover, formally it diverges loga-
current but, at the same time, captures the essential rithmically when h→ 0 ,
physics of the phenomenon.
Using Eqs. 4.9–4.11 and assuming first that h 1 and Ic0
I ca h → 0 ln 0 / 0 , 4,15
h 2 are parallel to each other we write the expression for
the critical current as Bergeret et al. , 2001b
where I c 0 is the critical current in the absence of the
I J = I c sin , 4,12 magnetic moment at T =0, and 0 is a parameter needed
to diverge the logarithm at low energies.
2T4T 2+2T,h−h2 When deriving Eqs. 4.13 and 4.14 the conventional
I cp = singlet superconducting pairing was assumed. The elec-
eR 2 + 2 T , h − h 2 2 +4 2 h 2 . trons of a Cooper pair have opposite spins. This picture
4,13 of a superconducting pair with opposite spins of the
electrons helps in the understanding of the effect.
The corresponding equation for the antiparallel configu- If the magnetic moments in both the magnetic layers
ration is different from Eq. 4.13 and can be written as are parallel to each other, they serve as an obstacle for
the Cooper pair because the pairs located in the region
2T4T 1
Ic a = . of the ferromagnet demand more energy. This leads to a
eR 2 + 2 T , h − h 2 2 +4 2 h 2
4.14

One can easily check that the critical current I c p para o


parallel configuration, Eq. 4.13, is always smaller than

https://translate.googleusercontent.com/translate_f 26/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
the currentare
pressions I c equal
uma
for to
theeach
antiparallel case.
other only These
in the two ex-
absence of
any magnetization.
In Figs. 20 and 21 we present the dependence of the
critical current on the strength of the exchange field. Nós
see from Fig. 20 that for the parallel configuration the
exchange field reduces the value of the Josephson cur-
rent and this is exactly what one would expect. No FIG. 21. The same dependence as in Fig. 20 for the case of an
same time, the critical current increases with the antiparallel orientation.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Page 32
1352 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

reduction of the Josephson current. However, if the ex- junction. Similar results were obtained by Krivoruchko
change fields of the different layers are antiparallel, they and Koshina 2001a, 2001b. The case of an arbitrary
may favor the location of the Cooper pairs in the vicinity S / F transparency was also studied by Barash et al.
of the Josephson junction. A certain probability exists 2002, Chtchelkatchev et al. 2002, and Li et al. 2002.
that one of the electrons of the pair is located in one Barash et al. 2002 calculated the Josephson current as a
layer, whereas the other is in the second layer. Such a function of the angle between magnetizations in the F
possibility is energetically favorable because the spins of film.
the electrons of the pair can now have the same direc-
tion as the exchange fields of the layers. Then it is more
probable for the pairs to be near the junction even in
comparison with a junction without exchange fields and, V. REDUCTION OF THE MAGNETIZATION DUE TO
as a result, the critical current may increase. SUPERCONDUCTIVITY: INVERSE PROXIMITY EFFECT
The results presented above have been obtained for
the SF / I / FS structure by Bergeret et al. 2001b. Mais cedo Until now we have studied the superconducting prop-
a formula for the Josephson critical current in the S m IS m
erties of different S / F structures for a fixed magnetiza-
S m is the magnetic superconductor junction was pre- ção. This means that we have assumed a certain value
sented by Kulic and Kulic 2001. From that formula one
for this quantity and its dependence on coordinates. o
could, in principle, derive an enhancement of the critical
implied justification of this assumption was that ferro-
current for the antiparallel M orientation in magnetic
magnetism is a stronger phenomenon than superconduc-
superconductors S m . Unfortunately, the authors seem to
tivity and the magnetic moment of conventional ferro-
have missed this interesting effect.
Some remarks should be made at this point. magnets can hardly be affected by superconductivity.
This assumption is certainly correct in many cases but
Eu The results presented above are valid in the tun- not always. Often the presence of superconductivity can
neling regime, ie, when the transparency of the drastically change magnetic properties of ferromagnets
insulating barrier I is low enough. Golubov et al. even if they are strong.
2002b have shown that a smearing of the singu- Experiments performed by Mühge et al. 1998 and
larity of I cuma
is provided by a finite temperature or Garifullin et al. 2002 showed that the total magnetiza-
a not very low barrier transparency. The maxi- tion of certain S / F bilayers with strong ferromagnets de-
mum of the critical current for the antiparallel creased with lowering the temperature below the critical
configuration I cuma decreases with decreasing resis- superconducting transition temperature T c . As an expla-
tance of the I layer. The effect becomes weaker as nation, it was suggested that due to the proximity effect
the thickness of the F layer grows. domains with different magnetization appeared in the
ii We assumed that the S / F interface was perfect. Dentromagnetic materials and this could reduce the total mag-
a structure with a large S / F interface resistance netization. At the same time, quantitative estimates
R S / F the bulk properties of the S film are not con- based on an existing theory Buzdin and Bulaevskii,
siderably influenced by the proximity of the F film 1988 led to a conclusion that this mechanism was not
to be more precise, the condition R S / F very probable.
In this section we address the problem of the magnetic
F d F / S d SFF must be satisfied, where F is moment reduction by the presence of a superconductor
the specific resistance of the F film. Then, as one
can readily show see Sec. II.B, a minigap bF assuming that, in the absence of the ferromagnet, we
= D F /2 R S / F d F arises in the F layer. The Green’s have conventional singlet superconducting pairing. isto
functions in the F layers have the same form as turns out that two different and independent mecha-
nisms that lead to a decrease of the magnetization in
before with replaced by bF . The singularity in
S / F heterostructures due to the proximity effect exist
I c h first occurs at h equal to bF .
and we give a detailed account of them.
A physical explanation for the singular behavior of In order to study the magnetic properties we have to
the critical current I c uma was given by Golubov et al. choose a model. One can distinguish two different types
2002b. These authors noted that the density of states in of ferromagnetism: i itinerant ferromagnetism due to
the F layer has a singularity when h = bF . At this value the spin ordering of free electrons and ii ferromag-
of h the maximum of I cuma is achieved due to an overlap of netism caused by localized spins. Most ferromagnetic
two −1/2 singularities. This leads to the logarithmic di- metals show both types of ferromagnetism simulta-
vergency of the critical current in the limit T→ 0 in anal- neously, ie, their magnetization consists of both contri-
ogy with the well-known Riedel peak in S / I / S tunnel butions.
junctions for the voltage difference 2. In the latter case We consider a model in which the conducting elec-
the shift of the energy is due to the electric potential. trons interact with the localized moments via an effec-
Golubov et al. 2002b have also shown that, for the tive exchange interaction. The corresponding term in
parallel configuration, at h = bF the critical current the Hamiltonian is taken by the following form see Ap-
changes its signs, ie, there is a transition from 0 to a pendix A:

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

https://translate.googleusercontent.com/translate_f 27/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras

Página 33
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1353

ductor in contact with a ferromagnet. They showed that


- d3r†rJSr r. 5,1 the magnetic ordering in the magnet might take the
form of a structure consisting of small-size domains,
This term is suitable to describe s - d or s - f interaction such that superconductivity is not destroyed. Claro,
between the s and localized d and f electrons. Nós também as follows from Eq. 5.2, the formation of a domainlike
consider the ferromagnetic interaction between the lo- structure costs magnetic energy but this is compensated
calized moments. This interaction can be very compli- by the energy of the superconductor that would have
cated and to determine it, one should know the detailed been lost if the magnetic order remained ferromagnetic.
band structure of the metal as well as different param- This is only possible if the stiffness of the magnetic
eters. However, all these details are not important for us order parameter J is not too large. For instance, this
and we write the interaction between the localized spins nonhomogeneous magnetization occurs in magnetic su-
phenomenologically as perconductors such as those studied by Bulaevskii et al.
1985. But can one see it in heterostructures containing
- J ij S i S j . 5,2 strong ferromagnets such as Fe or Ni in contact with
eu j conventional superconductors?
It is assumed that J is positive. This interaction, Eq. At first glance it seems impossible since the Curie
5.2, is responsible for the ferromagnetic alignment of temperature of, for example, iron is a hundred times or
localized moments and is known as the Heisenberg more larger than the critical temperature of a conven-
Hamiltonian. tional superconductor. Therefore any change of the fer-
We consider a metallic ferromagnet in which the con- romagnetic order looks much less favorable energeti-
duction electrons interact with localized magnetic mo- cally than the destruction of the superconductivity in the
mentos. The ferromagnetic interaction 5.2 assures a fi- vicinity of the S / F interface.
nite magnetic moment of the background. O total This simple argument was, however, questioned in the
magnetization is the sum of the background magnetiza- experiments performed by Mühge et al. 1998 on Fe/Nb
tion localized moments and the magnetization of the bilayers and by Garifullin et al. 2002 on V/Pd 1− x Fe x
polarized free electrons. estruturas. Direct measurements of the ferromagnetic
In the next two sections we discuss the two different resonance have shown that in several samples with thin
mechanisms that lead to a decrease of the magnetization ferromagnetic layers the average magnetic moment
at low temperatures. In Sec. VA we consider the possi- started to decrease below the superconducting transition
bility of changing the magnetic order of the localized temperature T c .
magnetic spins in a F film deposited on top of a bulk Of course, one can reduce the influence of the ferro-
superconductor. The contribution from free electrons to magnet on the superconductor by diminishing the thick-
magnetization is first assumed to be small. We shall see ness of the ferromagnet. Using the formulas obtained by
that for not too strong ferromagnetic coupling J the Buzdin and Bulaevskii 1988, Mühge et al. 1998 esti-
proximity effect may lead to an inhomogeneous mag- mated the thickness of the ferromagnet for which super-
netic state. Conversely, we consider in Sec. VB an itin- conductivity was still possible and got a value on the
erant ferromagnet in which the main contribution to order of 1 Å, which created some doubt as to the expla-
magnetization is due to free electrons. We shall show nation of the experiment.
that the magnetization of free electrons may decrease at At the same time, the use of the formulas derived by
low temperatures due to spin screening. Thus both ef- Buzdin and Bulaevskii was not really justified because
fects may lead to the decrease in the magnetization ob- the calculations were done for thick but weak ferromag-
served in experiments Mühge et al. 1998; Garifullin et nets under the option of a strong anisotropy of the fer-
al. , 2002. romagnet which was necessary for formation of domain
walls with the magnetization vector changing its sign but
A. Cryptoferromagnetic state not axis.
Bergeret, Efetov, and Larkin 2000 theoretically in-
In 1959 Anderson and Suhl Anderson and Suhl, vestigated the possibility of a cryptoferromagneticlike
1959 suggested that superconductivity could coexist state in S / F bilayers with parameters corresponding to
with a nonhomogeneous magnetic order in some types the experiments by Mühge et al. and Garifullin et al. Dentro
de materiais. Anderson and Suhl called this state the that work a cryptoferromagnetic state with a magnetic
cryptoferromagnetic state. moment that rotates in space was considered. This cor-
The reason for this coexistence is that if the magneti- responds to a weak anisotropy of the ferromagnet,
zation direction varies over a scale smaller than the su- which was the case in the samples studied in Mühge et
perconducting coherence length, the superconductivity al. 1998. In particular, Bergeret et al. 2000 studied a
may survive despite the ferromagnetic background. este phase transition between the cryptoferromagnetic and
is due to the fact that the superconductivity is sensitive ferromagnetic phases. The calculations were carried out
to the ferromagnetic moment averaged on the scale of in the limit d F h = v 0 / h , T ch 0 , where v 0 and 0 are
the Cooper pairs rather than to its local values. the Fermi velocity and Fermi energy, respectively. este
In 1988 Buzdin and Bulaevskii discussed properties of limit is consistent with the parameters of the experi-
a bilayer system consisting of a conventional supercon- ments of Mühge et al. 1998 and Garifullin et al. 2002.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Page 34
1354 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

latter will be considered for the case with a rotating in


space magnetic moment = Qy . This should be energeti-
cally more favorable than the domainlike-structure one
provided the magnetic anisotropy of F is low. Tal
cryptoferromagnetic state corresponds to a Néel wall
see, for example, Aharoni, 1996.
Strictly speaking, one has to also take into account the
FIG. 22. A S / F bilayer consisting of a thin ferromagnet at- magnetostatic energy due to a purely magnetic interac-
tached to a bulk superconductor. The ferromagnet may be in tion of the magnetic moments. However, if the condition
either the a ferromagnetic or the b cryptoferromagnetic
https://translate.googleusercontent.com/translate_f 28/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
Estágio. J
d2, 5,6
M 2s
We present here the main ideas of this work.
The Hamiltonian describing the bilayer structure in
Fig. 22 can be written as where M s is the magnetic moment per volume, is ful-
filled one can neglect its contribution with respect to the
one of the exchange energy Aharoni, 1996.
H = H 0 + H BCS − d r + rhr r+HM,
Taking typical values of the parameters for Fe, M s
5,3 =800 emu/cm 3 and J =210 −6 erg/cm, one can see that
Eq. 5.6 requires that the thickness d of the ferromagnet
where the integration must be taken in the region − d be smaller than 10 nm, which corresponds to compara-
x 0. Here H 0 is the one-particle electron energy in- tively thick layers. Throughout this section this condi-
cluding an interaction with impurities, H BCS is the usual tion is assumed to be fulfilled.
term describing the conventional BCS superconductivity In this case the magnetic energy M per unit surface
in the superconductor S , and the third term describes area is given by
the interaction between localized moments and conduc-
tion electrons, where is a constant that will be set 5,7
M = JdQ 2 .
equal to 1 at the end see Appendix A.
The term H M describes the interaction between the In order to calculate the superconducting energy S
localized moments in the ferromagnet cf. Eq. 5.2 . Nós one has to take into account the fact that the order pa-
assume that the magnetization of the localized spins is rameter should be destroyed, at least partially, near con-
described by classical vectors and we take into account tact with the ferromagnet. This means that the order
the interaction between neighboring spins only. No parameter is a function of the coordinate x perpen-
limit of slow variations of the magnetic moment in space dicular to the interface. As we want to minimize the
and taking into account Eq. 5.2, the Hamiltonian H M energy we should look for a nonhomogeneous solution
can be written in the form para x of nonlinear equations describing the supercon-
ductivity. Near the critical temperature T c one can use
J Ginzburg-Landau equations. The proper solution of
HM= Sx2+ Sy2+ S z 2 dV , 5,4
these equations can be written in the form
where the magnetic stiffness J characterizes the strength
x
of the coupling between the localized moments in the F x = 0 tanh +C, 5,8
layer and the S i are the components of a unit vector that 2 GL T
are parallel to the local direction of the magnetization.
We assume that the magnetic moments are directed where 0 is the value of the order parameter in the bulk
parallel to the S / F interface and write the spin vector S and GL is the correlation length of the superconductor
as S =0,−sin ,cos . A perpendicular component of defined in Eq. 2.2. Near T c this length can be much
the magnetization would induce strong Meissner cur- larger than the length S . The parameter C in Eq. 5,8
rents in the superconductor, which would require greater is a number that has to be determined from boundary
additional energy. condições.
The condition for an extremum of the energy H M , Eq. The solution for x , Eq. 5.8, is applicable at dis-
5.4 can be written as tances exceeding the length S and therefore we cannot
use it near the interface.
=0 5.5
Having fixed the constant C one can compute the de-
Solutions of Eq. 5.5 can be written in the form = Qy , crease of the superconducting energy due to the sup-
where Q is the wave vector characterizing the rotation in pression of superconductivity in the S layer using the
space see Fig. 22. The value Q =0 corresponds to the Ginzburg-Landau free-energy functional eg, de
ferromagnetic state. Gennes, 1966. The decrease of the superconducting en-
What we want to do now is to compare the energies of ergy S per unit area at the F / S interface is a function of
the ferromagnetic and cryptoferromagnetic states. o C and can be written as

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 35
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1355

to the exchange energy h , while is the related to the


magnetic stiffness J .
The conclusion that the phase transition between fer-
romagnetic and cryptoferromagnetic states should be of
the second order was drawn neglecting the magneto-
static interaction. The direct magnetic interaction can
change this transition to a first-order one Buzdin,
2005b. However, in the limit of Eq. 5.6, this first-order
transition will inevitably be close to the second-order
1. Such a modification of this type of phase transition
FIG. 23. Phase diagrams , a for different values of = T c is out of the scope of this review.
− T / T c . The area above the curves corresponds to the ferro- Let us make estimates for the materials used in the
magnetic state and the area below to the cryptoferromagnetic experimentos. Performing ferromagnetic resonance mea-
Estado. surements, Mühge et al. 1998 observed a decrease of
the effective magnetization of a Nb/Fe bilayer. The stiff-
ness J for materials such as Fe and Ni is 60 K/Å. o
3/2 2 + K 1 − K 2 , 5,9 parameters characterizing Nb can be estimated as fol-
S=62 lows: T c =10 K, v F 10 8 cm/s, and l 100 Å. The thick-
ness of the magnetic layer is of order d =10 Å and the
where K =tanh C and = T − T c / T c . exchange field h 10 4 K, which is proper for iron.
It remains only to determine the contribution from Assuming that the Fermi velocities and energies of
the third term of the Hamiltonian 5.3. The correspond- the ferromagnet and superconductor are close to each
https://translate.googleusercontent.com/translate_f 29/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
ing free energy M / S is given by other, we obtain a 25 and 610 −3 . It is clear from
Tr 1 Fig. 23 that the cryptoferromagnetic state is hardly pos-
M / S =− iT 0 2 d d3rh g0, 5,10 sible in the Fe/Nb samples used in the experiment by
0 Mühge et al. 1998.
where 0 is the density of states and g 0 is the quasiclas- However, one can in principle explain the observed
sical Green's function averaged over all directions of the decrease of the magnetization by taking a closer look at
Fermi velocity. the structure of the S / F interface. In the samples ana-
Since the exchange field h in a strong ferromagnet lyzed by Mühge et al. the interface between the Nb and
may be much higher than the value of −1 here is the Fe layers is rather rough. So one can expect that in the
momentum relaxation time, one has to solve the Eilen- magnetic layers there were “islands” with smaller values
berger equation in the F region and the Usadel equation of J and/or h . A reduction of these parameters in the
in the S region. Solutions for these equations in both the Fe/Nb bilayers is not unrealistic because of the forma-
superconductor and ferromagnet were obtained by tion of nonmagnetic “dead” layers that can also affect
Bergeret et al. 2000. the parameters of the ferromagnetic layers. Se o
Thus the total energy is given by =M+S+M/S, cryptoferromagnetic state were realized only on the is-
Eqs 5.7, 5.9, and 5.10. As a result, one can express lands, the average magnetic moment would be reduced
the free energy as a function of two unknown param- but would remain finite. Such a conclusion correlates
eters, K and Q . One can find these parameters from the with what one observes experimentally. One can also
condition that the free energy must be minimal, which imagine islands very weakly connected to the rest of the
leads to the equations layer, which would lead to smaller energies of a nonho-
mogeneous state.
/K= / Q = 0. 5,11 Let us now consider the experiment by Garifullin et
al. 2002 on Pd 0.97 Fe 0.03 /V. Due to the low concentra-
One can show that the cryptoferromagnetic-
tion of iron, the magnetic stiffness and the exchange
ferromagnetic transition is of second order, which means
field of the F layers is much lower than the one in the
that near the transition the parameter Q is small. No
case of a pure iron. For this system, one estimates the
transition it vanishes and this gives an equation binding
the parameters. Solving the equation numerically we parameters as see Garifullin et al. 2002 J 60 K/nm,
find the phase diagram of Fig. 23 determining the h 100 K. Assuming again that the Fermi velocities of V
boundary between the ferromagnetic and cryptoferro- and Pd 1− x Fe x are close to each other, Garifullin et al.
magnetic states. The parameters a and used in Fig. 23 2002 obtained the following values of the parameters
are defined as a 1.2 and 1.310 −3 for the sample with d F
=1.2 nm.
2 h 2 2d f Jd F 73 Using these values for a and one can see from the
um 2 5,12 phase diagram in Fig. 23 that there can be a transition
DT c 2 , 22,
F2TcD3 from the ferromagnetic to the cryptoferromagnetic state
Onde is the ratio between the Fermi velocities v 0 F / vS.0 a 0.2, which corresponds to T 2.4 K. The de-
It is clear from Eqs. 5.12 that the parameter a is related crease of the effective magnetization M eff with decreas-

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 36
1356 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

Let us consider first a bulk ferromagnet and derive a


relation between the exchange field and the magnetiza-
tion of the free electrons. The exchange field h = JS in
the ferromagnet can be due to the localized moments
see Eq. 5.1 or due to the free electrons in the case of
an itinerant ferromagnet. 5 In some ferromagnets both
the localized and itinerant moments contribute to the
magnetization.
FIG. 24. S / F structure and schematic representation of the The magnetization of the free electrons is given by
inverse proximity effect. The dashed curves show the local
magnetization. Eu d d3p
M= 5,13
4 B 2 2 3 Tr ˆ 3 ˆ 3 Gˇ R − Gˇ A n p ,
ing temperature was not observed in samples with larger where B is an effective Bohr magneton and n P is the
F thickness d F . M eff was a temperature-independent con- Fermi distribution function of the free electrons. o
stant for the sample with d F =4.4 nm and d S =37.2 nm. Dentroexpression in front of n P in Eq. 5.13 determines the
the sample with d F =1.2 nm and d S =40 nm, the effective DOS that depends on the exchange field h . Nós presumimos
magnetization M eff decreased by 50% with cooling that the magnetization is oriented along the z axis.
from T 4 to 1.5 K. This fact is again in accordance with Using Eq. 5.13 one can easily compute the contribu-
the predictions of Bergeret et al. 2000. tion of the free electrons to the magnetization in a bulk
The results of this section demonstrate that not only ferromagnet. In the simplest case of a normal metal with
do ferromagnets change superconducting properties but a quadratic energy spectrum we have
also that superconductivity can affect ferromagnetism.
This result is valid, in particular, for strong ferromag-
nets, although the thickness of the ferromagnetic layers MF= B p 2 dp n p − h − n p + h , 5,14
22
must be small in this case.
The exchange interaction between the superconduct- where p = p 2 /2 m − F . At T =0 the magnetization is given
ing condensate and the magnetic order parameter re- de
duces the energy of the system if the direction of the
magnetization vector M F is not constant in space but M F 0 = B p3+ − p 3− , 5,15
oscillates. Provided the energy of the anisotropy is small, 23 2
this interaction leads to the formation of a spiral mag-
netic structure in the F film. where p ± = 2 m F ± h are the Fermi momenta for
spin-up and spin-down electrons. In the quasiclassical
As we shall see in the next section, the appearance of
limit it is assumed that h F , and therefore
the cryptoferromagnetic state is not the only effect that
leads to a reduction of the effective magnetization in
https://translate.googleusercontent.com/translate_f 30/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
S / F structures. We shall show that the proximity effect MF0 Bh, 5.16
may also lead to a change of the absolute value of the where = p F 0 m / 2 is the density of states at the Fermi
magnetic moment M F in the ferromagnet and to an in- level and p F 0 = 2 m F is the Fermi momentum in the ab-
duced magnetization M S in the superconductor. sence of the exchange field. 6 For the temperature range
we are interested in, T h , one can assume that the
B. Ferromagnetism induced in a superconductor magnetization of the ferromagnet does not depend on T
and is given by Eq. 5.16.
In the previous section we have seen that supercon- Now let us consider a S / F system with a thin F layer
ductivity can affect the magnetic ordering changing the see Fig. 24 and ask a question: Is the magnetization of
orientation of magnetic moments in a ferromagnetic the itinerant electrons modified by the proximity effect?
film. In this section we demonstrate that another mecha- We assume that the exchange field of the ferromagnet F
nism for a change of the total magnetization of a S / F is homogeneous and aligned in the z direction, which is
system exists. In contrast to the phenomenon discussed the simplest situation.
in the previous section, the orientation of the magnetic At first glance, it is difficult to expect anything inter-
moments in the F film does not change but the magni- esting in this situation and, to the best of our knowledge,
tude of the magnetization in both the F and S films does. such a system has not been discussed until recently.
This change is related to the contribution of free elec-
trons both in the ferromagnet M F and in the super-
5 In many papers the exchange field h is defined in another
conductor M S to the total magnetization. Por um lado,
way h =− JS so that the energy minimum corresponds to ori-
the DOS in the F film is reduced due to the proximity entation of the vector antiparallel to the vector h . Nisso
effect and therefore M F is reduced. Por outro lado, case the magnetic moment m =− B is parallel to h . Ambos
the Cooper pairs in the S film are polarized in the direc- definitions lead to the same results.
tion opposite to M F , where M F is the magnetization of 6 Actually Eq. 5.16 is valid not only in the case of a quadratic
free electrons in the ferromagnet. spectrum but also in a more general case.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 37
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1357

However, the physics of this heterostructure is actu- h E Th = D F / 2d F 5.17


ally very interesting and is general for any shape of the S
and F regions. It turns out that the proximity effect re- is fulfilled. If the thickness of the F film or radius of the
duces the total magnetization of the system and this ef- F grain is not small in comparison to the correlation
fect can be seen as a certain kind of “spin screening.” length S , the situation changes. The induced magnetic
Before doing explicit calculations we explain the phe- moment M¯ S is much smaller than M¯ F 0 but variation of
nomenon in simple words. If the temperature is above the magnetic moment of the ferromagnetic film or
T c , the total magnetization of the system M tot equals grain M F becomes comparable to M¯ F 0 . Este último é
M 0 F d F , where d F is the thickness of the F layer. Quando caused by a change in the density of states of the ferro-
the temperature is lowered below T c , the S layer be- magnet due to the proximity effect. However, the case of
comes superconducting and Cooper pairs with the size a large ferromagnet size is less interesting because the
of the order of S D S /2 T c arise in the supercon- exchange field h should be smaller than the full
ductor. Due to the proximity effect the Cooper pairs screening of M¯ F 0 occurs only if the second condition in
penetrate the ferromagnet. In the case of a homoge- Eq. 5.17 is fulfilled.
neous magnetization the Cooper pairs consist, as usual, Using similar arguments we come to a related effect.
of electrons with opposite spins, such that the total mag- The magnetic moment in the ferromagnet should be re-
netic moment of a pair is equal to zero. The exchange duced in the presence of superconductivity because
field is assumed to be not too strong, otherwise the pairs some of the electrons located in the ferromagnet con-
would break down. densate into Cooper pairs and do not contribute to the
It is clear from this simple picture that pairs located magnetization.
entirely in the superconductor cannot contribute to the From this qualitative and somewhat oversimplified
magnetic moment of the superconductor because their picture one expects that the total magnetization of the
magnetic moment is simply zero, which is what one S / F system will be reduced for temperatures below T c .
would expect. Nevertheless, some pairs are located in Both the mechanism studied here and that of the last
space in a more complicated manner. One of the elec- section lead to a negative change of the total magnetiza-
trons of the pair is in the superconductor, while the ção. Thus independently of the origin of ferromag-
other moves in the ferromagnet. These are the pairs that netism, they can explain, at least qualitatively, the ex-
create the magnetic moment in the superconductor. este perimental data of Mühge et al. 1998 and Garifullin et
follows from the simple fact that the direction along the al. 2002.
magnetic moment M in the ferromagnet is preferred for The ideas presented above can be confirmed by calcu-
the electron located in the ferromagnet we assume a lations based on the Usadel equation. In order to deter-
ferromagnetic type of exchange field and this makes the mine the change of the magnetization it is sufficient to
spin of the other electron of the pair be antiparallel to compute the quasiclassical Green's functions g R A
M . So all such pairs with one electron in the ferromagnet =i/ d Gˇ R A and, in particular, the component pro-
and one in the superconductor equally contribute to the portional to ˆ 3 ˆ 3 .
magnetic moment in the bulk of the superconductor. Como The matrix Green's function has the form we write g
a result, a ferromagnetic order is created in the super- in Matsubara representation g = g R i for positive
conductor, the direction of the magnetic moment in this
region being opposite to the direction of the magnetic g = ˆ 3 g + iˆ 2 f . 5,18
moment M in the ferromagnet. Moreover, the induced
magnetic moment penetrates the superconductor over In the ferromagnet we represent, for convenience, the
the size of the Cooper pairs S that can be much larger matrix f in the spin space as
than d F .
This means that although the magnetization M S in- f+ 0
. 5,19
duced in the superconductor is less than the magnetiza- 0f−
tion in the ferromagnet M F 0 , the total magnetic moment
The diagonal form of the matrix is a consequence of the
in the superconductor M¯ S = S d 3 rM S r may be compa-
https://translate.googleusercontent.com/translate_f 31/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
rable to the magnetic moment of the ferromagnet in the uniformity
same form.of the exchange field h . The matrix g has the
normal state M¯ F 0 = M F 0 V F , where V F = d F in the case of a In order to find the function g 3 that determines the
flat geometry M¯ F 0 is the magnetic moment per unit magnetization, we have to solve the Usadel equation
square and V F =4 a F 3 /3 is the volume of the spherical A18 in the F and S regions and to match the corre-
ferromagnetic grain. It turns out that the total magnetic sponding solutions with the help of the boundary condi-
moment of the ferromagnetic region film or grain tions A21.
The simplest case when the Usadel equation can be
M¯ F 0 = BF hV F due to free electrons is compensated at solved analytically is the case of a thin F layer. We sup-
zero temperature by the total magnetic moment M¯ S in- pose that the thickness d F of the F layer is small com-
duced in the superconductor. This statement is valid if pared with the characteristic length F of the condensate
the condition penetration into the ferromagnet this condition is ful-

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 38
1358 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

0
M¯ S = dx M S x .
−d s

Assuming aquele h bF = D F /2 F d F ou h
D F /2 d2 F F d F / R b , we can easily compute the ratio
M¯ S D SS 2 T 1
- 5,23
MF0dF d F SF bF 2 + 2 3/2 = − 1,

where F is the resistivity of the F region.


We see that in the case of a thin ferromagnet at low
temperatures and a not too strong exchange field the
FIG. 25. Change of the magnetization of a F / S bilayer as a
magnetization induced in the superconductor compen-
função da temperatura.
sates completely the magnetization in the ferromagnet.
This result follows from the fact that the magnetization
filled in the experiments by Garifullin et al. 2002 . Dentro induced in the superconductor it is proportional to g S 3
this case we can average the exact Usadel equation spreads over distances of the order of S . Em vista disso
A18 over x in the F layer assuming that the Green’s result one can expect that the magnetic moment of a
functions are almost constant in space. In addition, pro- small ferromagnetic particle embedded in a supercon-
vided the ratio F / S is small enough, the Green's func- ductor should be completely screened by the Cooper
tions in the superconductor are close to the bulk values pairs. We discuss the screening of a ferromagnet particle
f BCS and g BCS . This allows us to linearize the Usadel by the Cooper pairs in the next subsection.
equation in the superconductor. The component of the It is worth mentioning that the problem of finding the
Green's function in S that enters the expression for the magnetization in a S / F structure consisting of thin
magnetization can be obtained from the boundary con- SdS S and F d F F layers is equivalent to the
dition A21 and is given by problem of magnetic superconductors where ferromag-
netic exchange interaction and superconducting corre-
1 lations coexist. If we assume a strong coupling between
gS3x=− − g BCS f F 0 + f BCS g F 3 e x ,
s 5,20
SS the thin S and F layers, we can again average the equa-
tions over the thickness of the structure and arrive at the
where S 2 =2 2 + 2 / D S , f F 0 = f + + f − /2, g F 3 = g + − g − /2,Usadel equation for the averaged Green's function with
and g ± and f ± are the components of the matrices g and
an effective exchange field h˜ = hd F / d and an effective
f . They are defined as order parameter ˜ = d S / d , where d = d S + d F . Nesse caso
g F ± = ˜ ± / ± , f F ± = ± bF f BCS / ± , 5.21 the magnetization is given by M = g B h˜ 2 −̃ 2 h˜ −˜ ,
Onde x is the step function. Isso significa que o
where ˜ ± = + bF g BCS ih , 2 − bF f BCS2 , ˜ . Esse resultado
±=˜± bF total magnetization M is zero for h˜
= D F /2 F d F . The magnetization variation is determined
agrees with that obtained by Karchev et al. 2001 and
de
Shen et al. 2003, who studied the problem of the coex-
istence of superconductivity and itinerant ferromag-
M=−i T Tr g · ˆ 3 . 5.22 netism in magnetic superconductors.
=− One of the assumptions made for obtaining the previ-
ous results is the quasiclassical condition h / F 1. For
Using Eqs. 5.20–5.22 for Tr g · ˆ 3 /2 g 3 = g + − g − /2, some materials the latter is not fulfilled and one has to
one can easily calculate M . In Fig. 25 we show the go beyond the quasiclassical approach. Halterman and
change of the magnetization M induced in the super- Valls 2002a studied the imbalance of spin-up and spin-
conductor as a function of temperature. We see that for down electrons in pure S / F structures ie, without im-
low enough temperatures the decrease of the magneti- purities in the case of strong exchange fields h / F 1
zation can be very large. At the same time, the change of In that case superconductivity is strongly suppressed at
the magnetization in the ferromagnet is small Bergeret the S / F interface. Solving the Bogoliubov–de Gennes
et al. , 2004a. equations numerically the authors showed that there
It is interesting to calculate the total magnetic mo- was magnetic “leakage” from the ferromagnet into the
ment M¯ S induced in the superconducting film and com- superconductor, which led to a polarization of the elec-
pare it with the total magnetization of the ferromagnet trons in S over the short length scale F . The direction of
M F 0 d F as we have mentioned, the magnetization varia- the induced magnetic moment in the superconductor
tion M F in the ferromagnet is small and can be ne- was parallel to that in the ferromagnet, which contrasts
glected. with our finding.
The total magnetization of the superconductor is At the same time, the limit of a very strong exchange
dado por field considered by Halterman and Valls 2002a differs

https://translate.googleusercontent.com/translate_f 32/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 39
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1359

completely from ours. It is clear that due to the strong


suppression of the superconductivity at the S / F inter-
face, the magnetic moment cannot be influenced by the
superconductivity and therefore thick ferromagnetic lay-
ers with exchange energies on the order of the Fermi
energy are not suitable for observing the reduction of
the magnetization described above.
The DOS for states with spin-up and spin-down elec-
trons in a S / F structure has been calculated on the basis
of the Usadel equation by Fazio and Lucheroni 1999.
The authors have found that the DOS of these states
was different in the superconductor over the length of
the order S . However, the change of the magnetization
has not been calculated in this work.
This was done later by Krivoruchko and Koshina
2002 for a S / F structure. Using the Usadel equation, FIG. 26. Ferromagnetic grain embedded in a superconductor.
the authors numerically calculated the magnetization in- Due to the inverse proximity effect the magnetic moment of
duced in the superconductor. They found that the mag- the grain is screened by the electrons of the superconductor.
netic moment leaked from the F layer into the S layer
and changed sign at some distance on the order of S , A recent work by Stahn et al. 2005 studied a multilay-
thus becoming negative at sufficiently large distances ered S / F / S / F /¯ structure. This structure consists of the
só. In our opinion, leakage of the magnetic moment high- T c superconductor YBa 2 Cu 3 O 7 S layer and of the
M S obtained in that paper is a consequence of the use of ferromagnet La 2/3 Ca 1/3 MnO 3 F layer. Two samples
the wrong expression for the magnetic moment. Eles with the S and F layers of the same thickness were used.
did not add to the formula, obtained in the quasiclassical Layers of sample 1 are 98 Å and those of sample 2 160 Å
approximation, a contribution from the energies levels thick. The Curie temperature of the ferromagnet and
located far from the Fermi energy. The latter contribu- temperature of the superconducting transition are equal
tion is not captured by the quasiclassical approach and to 165 K and 75 K, respectively. By using neutron reflec-
should be written additionally. tometry the authors obtained information on the spatial
We have seen that under certain conditions a finite distribution of the magnetic moment in the structure.
magnetic moment is induced inside the superconductor. Analyzing the temperature dependence of the Bragg
Does this magnetic moment affect the superconductiv- peaks intensity they came to the conclusion that the
ity? The magnetic field B S in the superconductor equals most probable scenario for explaining important fea-
the magnetization 4 M s . The induced magnetization in tures of this observed dependence was the assumption
the superconductor M S is smaller than the magnetiza- that an induced magnetization arises in the S layers. E se
tion in the ferromagnet: M S = M F max d F / S , d S . o this explanation was correct, the sign of the induced
critical field for superconducting thin films is given by magnetization had to be opposite to the sign of the mag-
Hc L / d S H bulk , where L is the London penetration netization in the F layers. It is quite reasonable to think
depth and H bulk is the critical field of the bulk material. that the mechanism discussed above for conventional su-
Superconductivity is not affected by the induced field B S perconductors should also be present in high- T c super-
if the field B S 4 M F d F / S we set d S S is smaller conductors and then the theoretic scenario analyzed in
than H c . Therefore the condition 4 M F L / d F H bulk this section can serve as an explanation of the experi-
should be satisfied. If we take L 1 m and d F 50 Å, mento.
we arrive at the condition 4 M F 200 H bulk . This condi-
tion is easily fulfilled for the case of not too strong fer- C. Spin screening of the magnetic moment of a
romagnets. Due to the presence of magnetization in the ferromagnetic particle in a superconductor
ferromagnet and superconductor spontaneous currents
arise in the system. The spontaneous Meissner currents Let us now consider a ferromagnetic particle grain
induced by the magnetization in S / F structures were embedded in a superconductor see Fig. 26. As in the
studied by Bergeret et al. 2001c and Krawiec et al. previous subsection, we analyze the magnetic moment
2004. induced in the superconductor around the particle and
The phenomenon discussed in this section can be con- compare it with the magnetic moment of the F particle
sidered as an alternative mechanism for the decrease of 4 a 3 /3 M F 0 we assume that the particle has a spherical
the total magnetic moment observed by Garifullin et al. form and radius a .
2002. In order to clarify which of these two effects is It is well known that the superconducting currents
more important for the experimental observations one Meissner currents in a superconductor screen a mag-
needs more information. netic field that decays from the surface over the London
The most direct check for the cryptoferromagnetic penetration length L and vanishes in the bulk of the
phase would be measurements with polarized neutrons. superconductor. The same length characterizes the de-

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 40
1360 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

cay of the magnetic field created by a ferromagnetic F We assume again that the transmission coefficient
grain embedded in a superconductor if the radius of the through the S / F interface is not small and the condition
grain a is larger than L . However, if the radius a is h D F / a 2 is fulfilled. In this case the expression for
small, the Meissner effect can be neglected and a stray g S 3 drastically simplifies. Indeed, in this limit g F 3
magnetic field around the grain should decay, as in a = f F 0 f BCS / g BCS and f F 0 ihf BCS g BCS / bF . Therefore Eq.
normal metal, over a length of the order a . We now 5.25 acquires the form

https://translate.googleusercontent.com/translate_f 33/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
consider just this case. f 2BCS umih2
Above the critical temperature T c the stray magnetic gS3= e− r−a.
S 5,26
field polarizes the spins of free electrons and induces a S r bF
magnetic moment. This magnetic moment is very small
2 10 −6 . This solution can be obtained from Eq. 5.24 if one
because the Pauli paramagnetism is weak B writes down the term 4 A r on the right-hand side of
In addition, the total magnetic moment induced by the 2 a 2 ih / S bF . Isso significa que
this equation with A = f BCS
stray magnetic field is zero. The penetration depth L
can be on the order of hundreds of interatomic distances the ferromagnetic grain acts on Cooper pairs as a mag-
netic impurity embedded in a dirty superconductor. isto
or larger, so that if a is smaller or on the order of 10 nm,
induces a ferromagnetic cloud of the size of the order S
the Meissner effect can be neglected.
The screening of the magnetic moment is a phenom- with a magnetic moment − B hV F .
enon specific to superconductors. It is usually believed In order to justify the assumptions made above we
that in the situation when screening due to the orbital estimate the energy D F / a 2 assuming that the mean free
electron motion can be neglected small grains and thin path is on the order of a . For a =30 Å and v F
films, the total magnetic moment is just the magnetic =10 8 cm/sec, we get D F / a 2 1000 K. This condition is
moment of the ferromagnetic particle and no additional fulfilled for ferromagnets with exchange energy of the
magnetization is induced by the electrons of the super- order of several hundred K.
condutor. In the limit of low temperatures the calculation of the
This common wisdom is quite natural because in con- magnetic moment becomes very easy and we obtain for
ventional superconductors the total spin of a Cooper the magnetic moment M¯ S induced in the superconductor
pair is equal to zero and the polarization of the conduc- the following expression:
tion electrons is even smaller than in the normal metal.
Spin-orbit interactions may lead to a finite magnetic sus- M¯ S
ceptibility of the superconductor but it is positive and = − 1. 5,27
M F 0 4 a 3 /3
smaller anyway than the one in the normal state Abri-
kosov and Gor'kov, 1962; Abrikosov, 1988. This is a remarkable result which shows that the in-
Let us now take a closer look at the results of the last duced magnetic moment is opposite in sign to the mo-
subsection. We have seen that the proximity effect in- ment of the ferromagnetic particle and their absolute
duces in the superconductor a magnetic moment with values are equal to each other. In other words, the mag-
sign opposite to the one in the ferromagnet. In view of netic moment of the ferromagnet is completely screened
this result it is quite natural to expect that the magnetic by the superconductor Bergeret et al. , 2004b. The char-
moment of a small ferromagnetic particle embedded in a acteristic radius of the screening is the coherence length
superconductor may be screened by the Cooper pairs as S , which contrasts to the orbital screening due to the
is sketched in Fig. 26. So let us consider this situation in Meissner effect characterized by the London penetra-
more detail. tion depth L .
We consider a ferromagnetic grain of radius a embed- To avoid misunderstanding we emphasize once again
ded in a bulk superconductor. If the size of the particle is that full screening occurs only if the magnetization per
smaller than the length F , we can again assume that the unit volume of the ferromagnetic grain M F 0 is given by
quasiclassical Green's functions in the F region are al- Eq. 5.16, which means that the ferromagnetic grain is
most constant and given by Eq. 5.21, where now bF an itinerant ferromagnet. If the magnetization of the fer-
=3 D F /2 F a . In the superconductor we have to solve the romagnet is caused by both localized moments M loc
linearized Usadel equation for the component g S 3 deter- and itinerant electrons M itin , full screening is not
mining the magnetization alcançado. Moreover, magnetization M loc may be larger
than M itin and have opposite direction. Neste caso nós
2 g S 3 −2Sg S 3 = 0, 5.24 would have antiscreening Bergeret and García, 2004.
Actually, we have discussed the diffusive case only.
where 2 = rr +2/ r r is the Laplace operator in spheri-
However, it turns out that spin screening also occurs in
cal coordinates.
the clean case provided the exchange field is not too
Using the boundary conditions A21 we write the so-
high: hv F / d F , where v F and d F are the Fermi velocity
lution of this equation as
and the thickness radius of the ferromagnetic film or
f BCS um 2 e − r − a
S
grain Bergeret et al. 2005; Kharitonov et al. , 2005.
gS3= g BCS f F 0 − f BCS g F 3 , 5,25 The energy spectrum of a superconductor with a
S 1 + S a r
pointlike classical magnetic moment was studied many
where f F 0 = f F + + f F − /2 and g F 3 = g F + − g F − /2. years ago by Shiba 1968, Rusinov 1969, and Sakurai

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 41
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1361

1970 and more recently by Salkola et al. 1997. The ˇ , g + hSˇ , g


− iD r g r g + i ˆ 3 t g + t gˆ 3 +
magnetic impurity leads to a bound state 0 inside the
superconducting energy gap. There is some critical Eu ˇ , 3gS= 0.
strength h c + S ˇ ˆ 3 gˆ 5,28
F of the exchange coupling h that sepa- assim
rates two different ground states of the system denoted
de if hh c and if hh c . The bound state 0 corre- All symbols are defined in Appendix A. The spin-orbit
sponds to a localized quasiparticle with spin up. 7 Since relaxation time so takes very different values depending
the total electronic spin in the state is zero, one says on the material used in the experiments. Some estimates
that the continuum localizes a spin up. A energia for the values of 1/ h so can be found in Oh et al. 2000.
needed to create a quasiparticle excitation decreases For example, for transition metals such as Fe one ob-
when increasing h . At h = h c the state becomes un- tains 1/ h so 10 −2 , while for a magnetic rare earth the
stable against a spontaneous creation of an excitation value 1/ h so 0.3 is more typical. In the latter case the
with spin up and the transition to the state ocorre. DentroSO interaction should clearly affect the penetration of
this state the electronic spin at the impurity site is now the condensate into the ferromagnet.
equal to −1/2. All previous work considering this prob- In order to study the influence of the SO interaction
lem focused the attention on the subgap structure of the on both the long-range and the inverse proximity effect
spectrum and did not address the problem of the screen- we shall use Eq. 5.28. We consider first the well-known
ing of the magnetic moment by the continuum spectrum. problem of the Knight shift. This example will show the
This is no surprise because a sufficiently large magnetic convenience of using the quasiclassical approach.

https://translate.googleusercontent.com/translate_f 34/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
moment of the impurity S 1 cannot be screened by
the quasiparticles.
1. The Knight shift in superconductors
Since the pioneering work of Abrikosov and Gor'kov
1962 it has been well established that the magnetic sus-
D. Spin-orbit interaction and its effect on the proximity ceptibility of small superconducting samples is not zero
efeito due to the spin-orbit interaction. This explains the ex-
periments performed for the first time by Androes and
In this section we discuss the influence of the spin- Knight 1961 who used the nuclear-magnetic-resonance
orbit SO interaction on the proximity effect. Apesar técnica.
in general its characteristic energy scale is much smaller Let us consider a superconductor in an external mag-
than the exchange energy h , it can be comparable to the netic field H . In the Usadel equation, Eq. 5.28, the field
superconducting gap and therefore this effect can be H plays the role of the exchange energy h . We are inter-
muito importante. Since SO scattering leads to a mixing of ested in the linear response to this field, ie, in the mag-
the spin channels, we expect that it will affect not only netic susceptibility S of the superconductor. Nós presumimos
the singlet component of the condensate but also the that the superconductor is homogeneous and therefore
triplet one in the ferromagnet. we drop the gradient term in Eq. 5.28:
In conventional superconductors the SO interaction
does not affect thermodynamic properties. No entanto, um - ˆ 3 , g + i ˇ , g + i B H n , g − 1/ so S ˇ ˆ 3 gˆ 3 Š, g
nonvanishing magnetic susceptibility at zero tempera-
ture Knight shift observed in small superconducting = 0, 5,29
samples and films can be explained only if the SO inter-
action is taken into account Abrikosov and Gor'kov, g 2 = 1. 5,30
1962. In the F / S structures considered here the ex-
The solution of Eq. 5.29 has the form
change field h breaks the time-reversal symmetry in
analogy to the external magnetic field in the Knight-shift g = g BCS + g 3 ˆ 3 ˆ 3 + f BCS ˆ 3 + f 0 iˆ 2 , 5.31
problema. Therefore the SO interaction in the supercon-
ductor is expected to influence the inverse proximity ef- where the functions g 3 and f 0 are corrections to the nor-
fect studied in this section. mal g BCS and anomalous f BCS Green's functions. No
In this section we shall generalize the analysis of the particle-hole space the matrix g has the usual form, ie,
long-range proximity effect and the inverse proximity it is expanded in matrices ˆ 3 and iˆ 2 . In spin space the
effect presented above taking the SO interaction into triplet component the g 3 and f 0 terms appears due to
conta. The quasiclassical equations in the presence of the magnetic field acting on the spins. Using Eqs.
the SO interaction were derived by Alexander et al. 5.29–5.31 one can readily obtain
1985 and used for the first time for the F / S systems by
Demler et al. 1997. 2
g3=−i BH , 5.32
The derivation of these equations is presented in Ap- E 2 E + 4/ so
pendix A. The resulting Usadel equation takes the form
where E = 2+2.
Substituting Eq. 5.32 into Eq. 5.22 we can write the
7 One assumes that the magnetic impurity has spin up. magnetization M as follows:

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 42
1362 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

1 2 = 2 +2 4
M=M0−B 2T2 H. 5,33 0 . 5.36
E 2 E + 4/ so so D F
The first term in Eq. 5.33 cannot be calculated in the We see from these equations that both the singlet and
framework of the quasiclassical theory and one should triplet components are affected by the spin-orbit inter-
use exact Green's functions. It corresponds to the Pauli action making the decay of the condensate in the ferro-
paramagnetic term given by M 0 = B H . In the quasiclas- magnet faster. In the limiting case, when 4/ so h,Tc,
sical approach this term is absent. This term does not both components penetrate over the same distance so
depend on temperature on the energy scale on the order = so D F and therefore the long-range effect is sup-
of T c and originates from a contribution of short dis- pressed. In this case the characteristic oscillations of the
tances where the quasiclassical approximation fails. singlet component are destroyed Demler et al. , 1997. In
This situation is rather typical for the quasiclassical the more interesting case 4/ so T c h , the singlet com-
approach and one usually adds by hand to formulas ob- ponent does not change and penetrates over the short
tained with this approach contributions coming from distance F . At the same time, the triplet component is
short distances or times see, for example, Artemenko more sensitive to the spin-orbit interaction and the pen-
and Volkov 1980, Rammer and Smith 1986, and etration length equals min so , T F.
Kopnin 2001 . Equation 5.33 was first obtained by Therefore if the spin-orbit interaction is not very
Abrikosov and Gor'kov 1962. strong, the penetration of the triplet condensate over the
In the absence of the spin-orbit interaction the mag- long distances discussed in the preceding sections is still
netization at T =0 is, as expected, equal to zero. Como- possible, although the penetration length is reduced.
ever, if the SO interaction is finite, the spin susceptibility
S does not vanish at T =0. It is interesting that, as fol- 3. Spin-orbit interaction and the inverse proximity
lows from Eq. 5.29, the singlet component of the con- efeito
densate is not affected by the SO interaction. The origin
In studying a S / F bilayer we have seen that the in-
of the finite susceptibility is the existence of the triplet
duced magnetic moment in the superconductor S is re-
component f 0 of the condensate.
lated to the appearance of the triplet component f 0 .
In S / F structures there is no exchange field in the
Moreover, we have shown that this component is af-
superconductor and therefore the situation is in prin-
fected by the SO interaction, while the singlet one f 3 is
ciple different. However, we have seen that due to the
não. So one should expect that the SO interaction may
proximity effect the triplet component f 0 is induced in
change the scale over which the magnetic moment is
the superconductor.
induced in the superconductor and one can easily esti-
From the above analysis one expects that the SO in-
https://translate.googleusercontent.com/translate_f 35/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
teraction may affect the penetration length of the mate
Assuming
this length.
that the Green's functions in the supercon-
triplet-component into the superconductor. In the next
ductor take values close to the bulk values we linearize
sections we consider the influence of the SO on the su-
the Usadel equation 5.28 in the superconductor. o
perconducting condensate in both the ferromagnet and
solution has the same form as before, Eq. 5.20, but S
the superconductor.
should be replaced by
2 → S2 + so2 , 5,37
2. Influence of the spin-orbit interaction on the long- S
range proximity effect where so 2 =8 D S / so . Therefore the length of the penetra-
Now we again consider the S / F / S / F / S structure of tion of g S 3 and, in turn, of M S into the S region decreases
Sec. IV.A and assume that the long-range triplet compo- if S 2 −2S
2.
assim
nent is created, which is possible provided the angle In principle, one can measure the spatial distribution
between the magnetizations differs from 0 and . Em or- of the magnetic moment in the S region as done by
der to understand how the SO interaction affects the Luetkens et al. 2003 by means of muon spin rotation
triplet component it is convenient to linearize Eq. 5,28 and get information on the SO interaction in supercon-
in the F layer assuming, for example, that the proximity ductors. As Eq. 5.37 shows, this would be an alterna-
effect is weak. One can easily obtain a linearized equa- tive method for measuring the strength of the SO inter-
action in superconductors, complementary to the
tion similar to Eq. 3.15 for the condensate function f .
measurement of the Knight shift Androes and Knight,
The solution of this equation is represented again in the
1961
Formato
fx = iˆ 2 f 0 x ˆ 0 + f 3 x ˆ 3 + iˆ 1 f 1 x ˆ 1 . 5.34 VI. DISCUSSION OF THE RESULTS AND OUTLOOK
The functions f i x are given as before by f i x In this review we have discussed new unusual proper-
= j b j exp j x but now the new eigenvalues j are writ- ties of structures consisting of conventional supercon-
ten as ductors in contact with ferromagnets. It has been known
2i 4 2 4 that such systems might exhibit very interesting proper-
2 =± h2− + , 5,35 ties such as a nonmonotonic reduction of the supercon-
± DF assim so D F
Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 43
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1363

ducting temperature as a function of the thickness of the Moreover, it was not even easy to see from the models
superconductor, the possibility of a contact in Joseph- studied in these papers that the odd frequency supercon-
son junctions with ferromagnetic layers, etc. ductivity had really to exist. The prediction that the odd
However, as we have seen, everything is even more triplet condensate must be generated in a real system
interesting and some spectacular phenomena are pos- has been made for the first time in the work by Bergeret
sible that might even at first glance look to be paradoxi- et al. 2001b.
cal. The common feature of the effects discussed in this The triplet component with the projection of the total
review is that almost all of them originate in situations spin S z = ±1 penetrates the ferromagnet over a long dis-
when the exchange field is not homogeneous. As a con- tance on the order of N D F /2 T , which shows that
sequence of the inhomogeneity, the spin structure of the the exchange field does not affect the triplet part of the
superconducting condensate function becomes very non- condensate. At the same time, the exchange field sup-
trivial and, in particular, the triplet components are gen- presses the amplitude of the singlet component at the
erated. In the presence of the inhomogeneous exchange S / F interface that determines the amplitude of the trip-
field, the total spin of a Cooper pair is not necessarily let component. The long-range triplet component arises
equal to zero and the total spin equal to unity with all only in the case of a nonhomogeneous magnetization.
projections onto the direction of the exchange field is The triplet component also appears in a system with a
possible. homogeneous magnetization but in this case it corre-
We have discussed the main properties of odd triplet sponds to the projection S z =0 and penetrates the ferro-
superconductivity in S / F structures. This superconduc- magnet over a short length F = D F / h N.
tivity differs from the well-known types of superconduc- The triplet component also exists in magnetic super-
tivity: i singlet superconductivity with s -wave conven- conductors Bulaevskii et al. 1985; Kulic and Kulic,
tional T c superconductors and d -wave high- T c 2001 with a spiral magnetic structure. However, it al-
superconductors types of pairing; ii odd in momentum ways coexists with the singlet component and cannot be
p and even in frequency triplet superconductivity ob- separated from it. In contrast, in multilayered S / F struc-
served, eg, in Sr 2 RuO 4 . tures with nonhomogeneous magnetization and with the
The odd triplet superconductivity discussed in this re- thickness of F layers d F exceeding F , Josephson cou-
view has a condensate Gor'kov function that is an odd pling between S layers is realized only through the long-
function of the Matsubara frequency and an even range triplet component and this separates the singlet
function in the main approximation of momentum p in and triplet components from each other. Como um resultado,
the diffusive limit. It is insensitive to scattering on non- “real” odd triplet superconductivity may be realized in
magnetic impurities and therefore may be realized in the transverse direction in such structures.
thin-film S / F structures where the mean free path is very Another interesting peculiarity of S / F structures is the
short. inverse proximity effect, namely, the penetration of the
A condensate function of this type was first suggested magnetic order parameter spontaneous magnetic mo-
by Berezinskii 1975 as a possible candidate to describe ment M into the superconductor and spatial variation
superfluidity in 3 He. Later, it was established that the of the magnetization direction in the ferromagnet under
superfluid condensate in 3 He had a different the influence of superconductivity. It turns out that both
structure—it was odd in p and even in . Em princípio, effects are possible. A homogeneous distribution of the
there is an important difference between the triplet su- magnetization M F in the S / F bilayer structures may not
perconductivity discussed here and that predicted by Be- be energetically favored in F even in a one-domain case
rezinskii, who assumed that the order parameter estava resulting in a nonhomogeneous distribution of M F in the
also an odd function of . In our case the order param- ferromagnet.
eter is determined by the singlet, s -wave condensate Moreover, the magnetic moment penetrates the super-
function and has ordinary BCS structure ie, it does not conductor induced ferromagnetism changing sign at

https://translate.googleusercontent.com/translate_f 36/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
depend on momentum p and frequency . No outro thethe
of S / system
F interface. Therefore
is reduced. thesome
Under total conditions
magnetic moment
full
hand, the structure of the triplet condensate function f in spin screening of M F occurs. For example, at zero tem-
the diffusive case considered here is similar to that sug- perature the itinerant magnetic moment of a ferromag-
gested by Berezinskii: it is an odd function of the Mat- netic grain embedded in a superconductor is completely
subara frequency and, in the main approximation, is screened by spins of the Cooper pairs in S . The radius of
constant in momentum space. The antisymmetric part of the screening cloud is on the order of the superconduct-
f is small compared with the symmetric part, being odd ing coherence length S . If the magnetization vector M F
in p and even in . The possibility of an odd frequency is oriented in the opposite direction to the ferromagnetic
superconductivity in solids was investigated by Kirk- exchange field h , antiscreening is possible.
patrick and Belitz 1991, Balatsky and Abrahams As to the experimental situation, there are indications
1992, Belitz and Kirkpatrick 1992, Abrahams et al. in favor of the long-range triplet component, although
1993, Coleman et al. 1993a, 1993b, 1995, and Balatsky so far unambiguous evidence does not exist. Por exemplo
et al. 1995. However, to the best of our knowledge, ample, the resistance of ferromagnetic films or wires in
none of these suggestions is realized experimentally. S / F structures changes on distances that exceed the

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 44
1364 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

length of the decay of the singlet component h Giroud range proximity effects already observed experimentally
et al. 1998; Petrashov et al. , 1999; Aumentado and Chan- are due to triplet pairing or to a simple redistribution of
drasekhar, 2001. A possible reason for this long-range the domain walls by the Meissner currents. Nós acreditamos
proximity effect in S / F systems is the long-range pen- that measurements on thin ferromagnetic wires in which
etration of the triplet component. However, a simpler the Meissner currents are reduced may clarify the situa-
effect might also be the reason for this long-range prox- ção.
imity effect. It is related to the rearrangement of the It is very interesting to distinguish experimentally be-
domain structure in the ferromagnet when the tempera- tween two possible inverse proximity effects. Apesar
ture lowers below T c . The Meissner currents induced in both the formation of the cryptoferromagnetic state and
the superconductor by a stray magnetic field affect the the induction of magnetic moments in superconductors
domain structure and the resistance of the ferromagnet are very interesting effects, it is not clear yet which of
may change Dubonos et al. , 2002. At the same time, these effects causes the magnetization reduction ob-
the Meissner currents should be considerably reduced in served by Mühge et al. 1998 and Garifullin et al. 2002.
a one-dimensional geometry for the ferromagnet such as The enhancement of the Josephson current by the
that used by Giroud et al. 1998 and the explanation in presence of the ferromagnet near the junction is one
terms of the long-range penetration of the triplet com- more theoretical prediction that has not been observed
ponent are more probable here. yet but still deserves attention. An overview for experi-
Sefrioui et al. 2003 and Peña et al. 2004 also ob- mentalists interested in these subjects is presented in
tained indications on the existence of a triplet compo- Appendix B, where we briefly discuss different experi-
nent in a multilayered S / F / S / F /¯ structure. o ments on S / F structures, focusing our attention on the
samples used by Sefrioui et al. contained the high- T c ma- materials for which we expect the main effects discussed
terial YBa 2 Cu 3 O 7 as a superconductor and the half- in this review may be observed.
metallic ferromagnet La 0.7 Ca 0.3 MnO 3 as a ferromag- In addition, further theoretical investigations are
net. They found that superconductivity persisted even necessário. The odd triplet component has been studied
in the case when the thickness of F layers d F essentially mainly in the diffusive limit h 1. It would be inter-
exceeded F d F 10 nm and F 5 nm. In a half-metal esting to investigate the properties of the triplet compo-
ferromagnet with spins of free electrons aligned in one nent for an arbitrary impurity concentration h 1. No
direction the singlet Cooper pairs cannot exist. Lá- theoretical work on the dynamics of magnetic moments
fore it is reasonable to assume that superconducting in S / F structures has been performed yet, although the
coupling between neighboring S layers is realized via the triplet component may play a very important role in the
triplet component Eschrig et al. 2003; Volkov et al. , dynamics of these structures. Transport properties of
2003. S / F structures also require also further theoretical con-
A reduction of the magnetic moment of S / F struc- siderações. It would be useful to study the influence of
tures due to superconducting correlations has already domain structures on properties of S / F structures, etc.
been observed Garifullin et al. , 2002. This reduction In other words, the physics of the proximity effects in
may be caused both by the spin screening of the mag- superconductor-ferromagnet structures is evolving into a
netic moment M F and by the rotation of M F in space very popular field of research, both experimentally and
Bergeret et al. , 2000, 2004a. Perhaps the spin screening theoretically.
can be observed directly by probing the spatial distribu- The study of the proximity effect in S / F structures
tion of the magnetic field or magnetic moment M with may be extended to include ferromagnets in contact
the aid of the muon spin rotation technique Luetkens et with high-temperature superconductors. Apesar de alguns
al. , 2003. The variation of the magnetic moment M oc- experiments have been done already Sefrioui et al. ,
curs on a macroscopic length S and therefore can be 2003; Stahn et al. , 2005, one can expect much more
detectou. broad experimental investigations in the future. Modern
Evidence in favor of the inverse proximity effect has techniques allow the preparation of multilayered
also been obtained in another experimental work Stahn S / F / S / F /¯ structures consisting of thin ferromagnetic
et al. , 2005. Analyzing data of neutron reflectometry on layers eg, La 2/3 Ca 1/3 MnO 3 and thin layers of high- T c
a multilayered YBa 2 Cu 3 O 7 /La 2/3 Ca 1/3 MnO 3 structure, superconductor eg, YBa 2 Cu 3 O 7 with variable thick-
the authors concluded that a magnetic moment was in- . It would be very interesting to study, both experi-
duced in superconducting YBa 2 Cu 3 O 7 layers. O sinal mentally and theoretically, such a system with noncol-
of this induced moment was opposite to the sign of the linear magnetization orientations. In this case d -wave
magnetic moment in ferromagnetic La 2/3 Ca 1/3 MnO 3 lay- singlet and odd triplet superconductivity should coexist
ers, which correlates with our prediction. in the system. It is well known that many properties of
In spite of these experimental results that may be con- ordinary BCS superconductivity remain unchanged in
sidered preliminary confirmation of the existence of the high- T c superconductors. This means that many effects
triplet component in S / F structures, there is a need for considered in this review can also occur in S / F struc-
additional experimental studies of the unconventional tures containing high- T c materials, but there will cer-

https://translate.googleusercontent.com/translate_f 37/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
superconductivity
important issues isdiscussed in this
to understand review.the
whether Um dos
long- tainly be differences
conductors that have with respect
s -wave to conventional super-
pairing.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 45
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1365

We hope that this review will encourage experimen- conducting and magnetic h order parameter. o
talists and theorists to make further investigations in this exchange field h is parallel to the magnetization M F in
fascinating field of research. F . 8 In strong ferromagnets the magnitude of h is much
higher than and corresponds to an effective magnetic
AGRADECIMENTOS field H exc = h / B on the order of 10 6 Oe where B
= g Bohr , g is the g factor and Bohr is the Bohr magne-
We appreciate fruitful discussions with AI Buzdin, ton.
YV Fominov, IA Garifullin, A. Gerber, AA Gol- In order to describe the ferromagnetic region we use a
ubov, A. Palevski, LR Tagirov, K. Westerholt, and H. simplified model that contains all the physics we are in-
Zabel. We would like to thank SFB 491 for financial terested in. Ferromagnetism in metals is caused by the
Apoio, suporte. FSB would like to thank the EU network electron-electron interaction between electrons belong-
DIENOW for financial support. ing to different bands that can correspond to localized
and conducting states. Only the latter participate in the
proximity effect. If the contribution of free electrons
APPENDIX A: BASIC EQUATIONS strongly dominates an itinerant ferromagnet, one has
Throughout this review we mainly use the well- M F M e and the exchange energy is caused mainly by
established method of quasiclassical Green's functions. free electrons.
Within this method the Gor'kov equations can be dras- If the polarization of the conduction electrons is due
tically simplified by integrating the Green's function to the interaction with localized magnetic moments, the
over momentum. This method was first introduced by Hamiltonian Hˆ F can be written in the form
Eilenberger 1968 and Larkin and Ovchinnikov 1968
and then extended by Usadel 1970 for a dirty case and Hˆ F = − h 1 a+sp S ss a s p , A2
by Eliashberg 1971 for a nonequilibrium case. o p,s
method of the quasiclassical Green's functions is dis- where S = a S a r − r a and S a is the spin of a particular
cussed in many reviews Serene and Reiner 1983; Lar- ion. A constant h 1 is related to h via the equation h
kin and Ovchinnikov 1984; Rammer and Smith 1986;
= h 1 n M S 0 , where n M is the concentration of magnetic
Belzig et al. 1999 and in the book by Kopnin 2001. In
ions and S 0 is a maximum value of S a we consider these
this appendix we present a brief derivation of equations
spins as classical vectors; see Gor'kov and Rusinov
for the quasiclassical Green's functions and write formu-
las for the main observable quantities in terms of these 1963 . In this case the magnetization is a sum, M
funções. Special attention will be paid to the depen- = M loc + M e , and the magnetization M e can be aligned
dence of these functions on the spin variables that play a parallel h 1 0, the ferromagnetic type of the exchange
crucial role in S / F structures. In particular, we take into field to M or antiparallel h 1 0, the antiferromagnetic
account the spin-orbit interaction along with the ex- type of the exchange field. In the following we shall
change interaction in the ferromagnet. assume a ferromagnetic exchange interaction M e and M
We start with a general Hamiltonian describing a con- are oriented in the same direction. In principle, one can
ventional BCS–superconductor-ferromagnet structure: add to Eq. A2 the term a , b S a S b describing a di-
rect interaction between localized magnetic moments
Hˆ = „ a+sp p pp + eV + U imp ss + U so but in most of the review this term is not important
p,s except in Sec. VA, where the cryptoferromagnetic state
−h· a s p − a†sp a†s p + cc…. A1 is discussed.
Starting from the Hamiltonian A1 and using a stan-
The summation is carried out over all momenta p , p dard approach Larkin and Ovchinnikov, 1984, one can
and spins s , s the notation s¯ , p¯ means inversion of derive the Eilenberger and Usadel equations. Initially
both spin and momentum, p = p 2 /2 m − F is the kinetic these equations have been derived for 22 matrix
energy counted from the Fermi energy F , and V is a Green's functions g n , n , where indices n , n relate to nor-
smoothly varying electric potential. The superconduct- mal g 11 , g 22 and anomalous or condensate f 12 , f 21
ing order parameter must be determined self- Green's functions. These functions describe the singlet
consistently. It vanishes in the ferromagnetic regions. componente. In the case of nonhomogenous magnetiza-
The potential U imp = U p − p describes the interaction tion considered in this review, one has to introduce ad-
of the electrons with nonmagnetic impurities, and U so ditional Green's functions depending on spins and de-
describes a possible spin-orbit interaction Abrikosov scribe not only the singlet but also the triplet
and Gor'kov, 1962: componente. These matrices depend not only on n , n in-
dices but also on the spin indices s , s , and are 44 ma-
u soEu trices in the spin and Gor'kov space sometimes the n , n
U so = p .
2p space is called the Nambu or Nambu-Gor'kov space.
Eup F
Here the summation is performed over all impurities.
The representation of the Hamiltonian in A1 implies 8 We remind the reader that the exchange field h is measured
that we use the mean-field approximation for the super- in energy units; see also footnote 5.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 46
1366 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

In order to define the Green's functions in a custom- states, the Josephson current, etc.. The matrix
ary way it is convenient to write the Hamiltonian A1 in Gˇ t , t K = Gˇ t 1 , t 2 + Gˇ t 2 , t is
1 related to the distribu-
terms of new operators c nsp† and c nsp that are related to tion function and has a nontrivial structure only in
https://translate.googleusercontent.com/translate_f 38/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
the creation and annhilation operators a s + and a s by the the nonequilibrium case. In the equilibrium case it is
relation we drop the index p related to the momentum equal to Gˇ K = dt − t Gˇ t − t K exp i t − t = Gˇ R
a s , n = 1, − Gˇ A tanh /2 T .
c ns = A3
a†s¯ , n = 2. In order to obtain the equations for the quasiclassical
Green's functions, we follow the procedure introduced
These operators for s =1 were introduced by Nambu by Larkin and Ovchinnikov 1984. The equation of mo-
1960. The new operators allow one to express the tion for the Green's functions is
anomalous averages a ↑ · a ↑ introduced by Gor'kov as
conventional averages c 1 · c +
2 and therefore the theory i t − Hˇ − ˇ imp −ˇ so ˇG = 1̌, A8
of superconductivity can be constructed by analogy to a Onde
theory of normal systems. Thus the index n operates in
particle-hole Nambu-Gor'kov space, while the index s 2
Hˇ = − ˆ 3 r − F − h ˆ 3 ˇS+ ˆ ˜ˆ 3
operates in spin space. In terms of the c ns operators the 2m
Hamiltonian can be written in the form
e ˇ ˇ
imp and so are the self-energies given in the Born
H= c+ns H nn ss c n s , A4 approximation by
p,n,s
where the summation is performed over all momenta, ˇ 2 3 G ˇ ˆ 3 , Ǧ = d
imp = N imp u ˆimp dp
4
Ǧ,
particle-hole, and spin indices. The matrix Hˇ is given by
ˇ 2 Gˇ
Hˇ =1 ˆ ˜ˆ 3 − h ˆ 3 S ˇ so = N imp u so so , A9
2 p pp + eV + U imp ˆ 3 ˆ 0 +
d
u soEu Ǧ so = dp n n Sˇ ǦŠ n n.
+ 2p p Sˇ . A5 4
Eup F Here N imp is the impurity concentration, is the density
The matrices ˆ i and ˆ i are the Pauli matrices in particle- of states at the Fermi level, and n is a unit vector parallel
hole and spin space, respectively; i =0,1,2,3 where ˆ 0 and to the momentum.
The next step is to subtract from Eq. A8, multiplied
0 are the corresponding unit matrices. The matrix vec-
tor Š is defined as by ˆ 3 from the left, its conjugate equation multiplied by
ˆ 3 from the right. Then one has to go from the variables
Š = ˆ 1 , ˆ 2 , ˆ 3 ˆ 3 , r , r to „ r + r / 2 , r − r … and to perform a Fourier trans-
formation with respect to the relative coordinate. De
and the matrix order parameter equals ˆ=
˜ ˆ 1 Re making use of the fact that the Green's functions are
− ˆ 2 Im . Now we can define the matrix Green's func- peaked at the Fermi surface, one can integrate the re-
tions in particle-hole spin space in the Keldysh rep- sulting equation over p , and finally one obtains
resentation in a standard way, 1
ˆ 3 t g + t gˆ 3 + v F g − i hŠ, g − i ˇ , g + g,g
1 2
Gˇ t i , t k = T C c ns t i c†n s t k , A6
Eu 1
+ ˆ 3 g so ˆ 3 , g = 0, A10
where the temporal indices take the values 1 and 2, 2 so
which correspond to the upper and lower branches of
the contour C , running from − to + and back to −. Onde ˇ = ˆ 3 ˆ 3ˆ ˜and the quasiclassical Green's functions
One can introduce a matrix in the Keldysh space of gt i , t k are defined as
the form
Eu
gpF,r= ˆ3ˆ0 d p Ǧ t i , t k ; p , r , A11
Gˇ t , t R Gˇ t , t K
Ǧ t , t = , A7
0 Gˇ t , t A and v F is the Fermi velocity. The scattering times appear-
where the retarded advanced Green's functions ing in Eq. A10 are defined as
Gˇ t , t RA estamosrelacionadopara a matrices −1 = 2 N imp u2imp , A12
Gˇ t i , t k : Gˇ t , t RA = Gˇ t 1 , t 1 − Gˇ t 12 , t21 . All these
elements are 44 matrices. These functions determine −1 = 1 N imp
d 2 sin 2 .
u so A13
thermodynamic properties of the system density of assim3 4

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Página 47
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1367

Equation A10 is a generalization of an equation de- If we take the elements A11 or A22 of the super-
rived by Eilenberger 1968 and Larkin and Ovchinni- matrix g , we obtain the Usadel equation for the retarded
kov 1968 for the general nonequilibrium case. este and advanced Green's functions g R A t , t generalized
generalization in the absence of spin-dependent inter- for the case of the exchange field acting on the spins of
actions was done by Eliashberg 1971 and Larkin and electrons. In this review we are mainly interested in sta-
Ovchinnikov 1984. A solution for Eq. A10 is not tionary processes, when the matrices g R A t , t depend
único. The proper solutions must obey the normaliza- only on the time difference t − t . Performing the Fou-
tion condition rier transformation g R A = dt − tg RA t − t exp i t
− t , we obtain for g R A the following equation we
d 1 /2 g p F , r ; , 1 · g p F , r ; 1 , =1 A14 drop the indices RA :

Generalization for the case of exchange and spin-orbit D x g x g + i ˆ 3 ˆ 0 , g + ih ˆ 3 ˆ 3 , g cos x


interaction was presented by Bergeret et al. 2000,
2001c. The solution to Eq. A10 can be obtained in + ˆ 0 ˆ 2 , g sin x + i ˇ , g + EuS ˇ ˆ 3 gˆ
ˇ , 3gS= 0.
some limiting cases, for example, in the homogeneous assim
caso. However, finding its solution for nonhomogeneous A19
https://translate.googleusercontent.com/translate_f 39/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
structures with an arbitrary impurity concentration may
be a quite difficult task. Further simplifications can be It is assumed here that h has the components
made in the case of a dirty superconductor when the h 0,sin ,cos . This equation was first obtained by Us-
energy −1 related to the elastic scattering by nonmag- adel 1970 and is known as the Usadel equation. Incluindo
netic impurities is larger than all other energies involved sion of the exchange and spin-orbit interaction was
in the problem, and the mean free path l is smaller than made by Alexander et al. 1985 and Demler et al. 1997.
all characteristic lengths except the Fermi wavelength Equation A18 can be solved analytically in many
that is set in the quasiclassical theory to zero. Nisso cases and it is used in most of the previous sections in
case one can expand the solution of Eq. A10 in terms order to describe different S / F structures. Soluções para
of spherical harmonics and retain only the first two of the Usadel equation must obey the normalization condi-
them, ie, ção

gpF,r;=gsr+pF/pFgˇar, A15 gpF,r;·gpF,r;=1 A20


where g s r is a matrix that depends only on coordinates. The Usadel equation is complemented by the boundary
The second term is the antisymmetric part the first Leg- conditions presented by Kuprianov and Lukichev 1988
endre polynomial that determines the current. It is as- on the basis of Zaitsev's boundary conditions Zaitsev,
sumed that the second term is smaller than the first one. 1984. Various aspects of the boundary conditions have
The parameter l / x 0 determines its smallness, where l is been discussed by Lambert et al. 1997, Nazarov 1999,
the mean free path and x 0 is a characteristic length of Xia et al. 2002, and Kopu et al. 2004. In the absence
the problem. In S / F structures x 0 D F / h is the shortest of spin-flip processes at the interface they take the form
length because usually h . In the limit l / x 0 1, that
is, if the product h is small, one can express g ˇ a r from 1
g1xg1= g1,g2, A21
Eq. A10 in terms of g s r , 2a
g ˇ a r ; , = − lg s r ; , 1 gsr;1, . A16 where 1 = R b 1 , 1 is the conductivity of the conductor
When obtaining Eq. A16, we used the relations 1, and R b is the interface resistance per unit area. The x
coordinate is assumed to be normal to the plane of the
gsr;,1 gsr;1, = 1, interface.
A17 The boundary condition A21 implies that we accept
g ˇ as r ; , 1 g s r ; 1 , +gsr;,1 gˇar;1, =0 the simplest model of the S / F interface which is used in
most papers on S / F structures. We assume that the in-
The symbolically written products in Eqs. A16 and terface separates two dirty regions: a singlet supercon-
A17 imply an integration over the internal energy 1 as ductor and a ferromagnet. The superconductor and the
mostrado na Eq. A14. ferromagnet are described in the mean-field approxima-
The equation for the isotropic component of the tion with different order parameters: the off-diagonal or-
Green's function after averaging over the direction of der parameter in the superconductor in the weak-
the Fermi velocity v F reads coupling limit and the exchange field h in the
ˇ , g + hSˇ , g ferromagnet acting on the spins of free electrons. Não
− iD gg + i ˆ 3 t g + t gˆ 3 + spin-flip scattering processes are assumed at the S / F in-
Eu terface. A generalization of the boundary conditions was
+ ˇ , 3gS= 0,
S ˇ ˆ 3 gˆ A18 used for a spin-active S / F interface as carried out in the
assim papers by Millis et al. 1988, Eschrig 2000, Fogelström
where D is the diffusion coefficient. 2000, and Kopu et al. 2004.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Page 48
1368 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

Equations A18 and A21 together with the self- Mx,y=B 1/2 i T Tr ˆ 0 ˆ 1,2 g , A29
consistency equation that determines the superconduct-
ing order parameter constitute a complete set of equa-
tions from which one can obtain the Green's functions. where is the density of states at the Fermi level in the
The Usadel equation can be solved in some particular normal state and B = g Bohr is an effective Bohr magne-
casos. We often use the linearized Usadel equation ob- ton.
tained by representing the Green's functions g of the Finally, it is important to remark on the notations
superconductor in the form used in this review. In most works where S / F structures
with homogeneous magnetization are studied, the
Green's function g is a 22 matrix with the usual nor-
g p F , r ; = g BCS +gS+fS, A22
mal and Gor'kov components. Of course, this simplifica-
tion can be made provided magnetizations of F layers
where g BCS = ˆ 3 g BCS + iˆ 2 f BCS , g BCS = i / f BCS ,
involved in the problem are aligned in one direction.
and f BCS =/ i 2 + 2 . We have written the matrix g in
However, this simple form leads to erroneous results if
the Matsubara representation. This means that a substi- magnetizations are arbitrarily oriented with respect to
tution ⇒ i = T 2 n +1, n =0,±1,±2,… is done entre si. The 44 form of the Green's function is
and g coincides with g R for positive and with unavoidable if one studies structures with nonhomoge-
g A for negative . The linearized Usadel equation has neous magnetization. Of course, the c operators in Eq.
the form A3 can be defined in different ways. Por exemplo,
Maki 1969 introduced a spinor representation of the
2 f S − S 2 f S = 2 i ˘ / D S g2BCS A23 field operators, which is equivalent to letting the spin
xx
index of the operator a in Eq. A3 be unchanged when
in the S region and n =2. This notation was used in later works eg, Alex-
ander et al. 1985; Demler et al. , 1997 in which the
2 f − 2 f + i h 2 ˆ 3 , f + cos ± ˆ 3 ˆ 2 , f − sin =0 Green's functions have a 22 block matrix form. o
xx diagonal blocks represent the normal Green's functions,
A24 while the off-diagonal blocks represent the anomalous
1. With this notation the matrix, Eq. A5, changes its
in the F region. Aqui 2 =2 E / D S , 2 =2 / D F , 2
S h Formato. For example, the term containing is propor-

https://translate.googleusercontent.com/translate_f 40/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
= h sgn
signs / DEq.
± in F , and A , B ± = AB ± BA ,
A24 correspond to the right˘ =and
iˆ 2left
ˆ3 .o tional
pends tooniˆthe
2 and not to ˆ 3 . The choice of notation de-
problem to solve. In order to study the
layers, respectively. triplet superconductivity induced in S / F systems and to
The boundary conditions for f S and f F f in zero- see explicitly the three projections S z =0,±1 of the con-
order approximation f F =0 are obtained from Eq. densate function, it is more convenient to use the opera-
A21. They have the form tors defined in Eq. A3 see, for example, Bergeret et al.
2001c and Fominov et al. 2003.
2
x f S = 1/ S g BCS f − g BCS f BCS ˆ 3 g F 3 , A25
APPENDIX B: FUTURE DIRECTION OF EXPERIMENTAL
A26 PESQUISA
x f F = 1/ F g BCS f − f S ,
As we have seen throughout the paper there are a
where F , S = R b F , S . great number of experiments on S / F structures. The va-
If the Green's functions are known, one can calculate riety of superconducting and ferromagnetic materials is
macroscopic quantities such as the current, magnetic very large. In this section we briefly review some of
moment, etc. For example, the current is given by Lar- these experiments. We shall not dwell on specific fabri-
kin and Ovchinnikov 1984, cation techniques but rather focus on which pairs of ma-
terial S and F are more appropriate for the observation
I S = L y L z /16 F Tr ˆ 3 ˆ 0 dg s g s / x 12 , A27 of the effects studied in this review.
First experiments on S / F structures used strong ferro-
magnets large exchange fields like Fe, Ni, Co, or Gd
where L y , z are the widths of the films in the y and z and conventional superconductors like Nb, Pb, V, etc.
directions the current flows in the transverse x direc- Hauser et al. , 1963. In these experiments the depen-
tion and the subscript 12 shows that one has to take the dence of the superconducting transition temperature on
Keldysh component of the supermatrix g s g s / x . Variação the thicknesses of S and F layers was measured. Em outro
tion of the magnetic moment due to the proximity effect words, the suppression of the superconductivity due to
is determined by the formulas the strong exchange field of the ferromagnet was ana-
lyzed. It is clear that for such strong ferromagnets spin
Mz=B 1/2 i T Tr ˆ 3 ˆ 3 g , A28 splitting is large and therefore a mismatch in electronic
parameters of S and F regions is large. This leads to a

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Page 49
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1369

low interface transparency and a weak proximity effect. though the materials employed in this experiment can-
This was confirmed by Aarts et al. 1997 in experiments not be quantitatively described with the methods pre-
on V/V 1− x Fe x multilayers. By varying the concentration sented in this review the ferromagnet they used is a half
of Fe in VFe alloys they could change the values of the metal with an exchange field comparable to the Fermi
exchange field and indirectly the transparency of the in- energy and the superconductor is unconventional, the
terface. Such systems consisting of a conventional super- experimental technique may be used in other experi-
conductor and a ferromagnetic alloy, both with similar ments in order to detect induced magnetization pre-
band structure in the above experiment the mismatch dicted in Secs. VB and VC
was 5%, are good candidates for observing the effects
discussed in Secs. IV.A, VB, and VC LIST OF SYMBOLS AND ABBREVIATIONS
Weak ferromagnets have been used in recent years in S superconductor
many experiments on S / F structures. Before we turn our N nonmagnetic normal metal
attention to ferromagnets with small exchange fields it is F ferromagnetic metal
worth mentioning the experiment by Rusanov et al. Eu insulator
2004. They analyzed the so-called spin-switch effect. Dentro LRTC long-range triplet component
particular, they studied the transport properties of Per- ˆ i , i =1,2,3 Pauli matrices in particle-hole space
malloy Py/Nb bilayers. They observed an enhance- ˆ i , i =1,2,3 Pauli matrices in spin space
ment of superconductivity in the resistive transition in ˆ0,ˆ0 unit matrices
the field range where the magnetization of the Py D diffusion coefficient
switches and many domains were present. Interesting density of states
for us is that Py shows a well-defined magnetization = T 2 n +1 Matsubara frequency
switching at low fields and therefore could be used to real frequency energy
detect the long-range triplet component that appears g BCS quasiclassical normal Green's func-
when magnetization of the ferromagnet is not homoge- tion for a bulk superconductor
neous see Sec. III.C. Finally, a magnetic configuration f BCS quasiclassical anomalous Green’s
analysis of strong-ferromagnetic structures used in trans- function for a bulk superconductor
port experiments, such as those performed by Giroud et Tc superconducting critical temperature
al. 1998 and Petrashov et al. 1999, may also serve to Ic Josephson critical current
confirm the predictions of Sec. III.C. As discussed be-
Rb interface resistance per unit area
fore, increase in the conductance of the ferromagnet for
temperatures below the superconducting T c may be ex- bN = D N /2 R b N d N minigap induced in a normal metal
S,F conductivity in the normal state
plained assuming a long-range proximity effect.
S,F R bS , F
The proximity effect in S / F is stronger if one uses
ratio F / S
dilute ferromagnetic alloys. Thus such materials are the J magnetic coupling between localized
best candidates in order to observe most of the effects
magnetic moments
discussed in this review. The idea of using ferromagnetic
h exchange field acting on the spin of
alloys with small exchange fields was used by Ryazanov
conducting electrons
et al. 2001. They were the first in observing the sign
N = D N /2 T characteristic penetration length of
reversal of the critical current in a S / F / S Josephson
the condensate into a dirty normal
junction. Nb was used as superconductor while
metal
Cu 0.48 Ni 0.52 alloy was used as a ferromagnet exchange
characteristic penetration length of
https://translate.googleusercontent.com/translate_f 41/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
field 25 et
Kontos K.al.
Later
2002ononsimilar results
Nb/Al/Al 2 Owere obtained
3 /PdNi/Nb by
struc- F=DF/h the condensate into a dirty ferro-
magnet
turas. The CuNi alloy was also used in the experiment
S = D S /2 T c superconducting coherence length
by Gu et al. 2002b on F / S / F structures. In this experi-
for a dirty superconductor
ment the authors determined the dependence of the su-
perconducting transition temperature on the relative
magnetization orientation of two F layers. A fim de REFERÊNCIAS
get different alignments between the two CuNi layers an
exchange-biased spin-valve stack of CuNi/Nb/ Aarts, J., JME Geers, E. Brück, AA Golubov, and R.
CuNi/Fe 50 Mn 50 was employed. With a small magnetic Coehoorn, 1997, Phys. Rev. B 56 , 2779.
Abrahams, E., AV Balatsky, JR Schrieffer, and PB Allen,
field the authors could switch the magnetization direc-
1993, Phys. Rev. B 47 , 513.
tion of the free NiCu layer. This technique could be very
Abrikosov, AA, 1988, Fundamentals of the Theory of Metals
useful for observing Josephson coupling via the triplet North-Holland, Amsterdam.
component as described in Sec. IV.A. Abrikosov, AA, and LP Gor'kov, 1962, Zh. Eksp. Teor. Fiz.
Finally, it is worth mentioning the experiment by 42 , 1088 Sov. Phys. JETP 15 , 752 1962 .
Stahn et al. 2005 on YBa 2 Cu 3 O 7 /La 2/3 Ca 1/3 MnO 3 . Nos- Adkins, CJ, and BW Kington, 1969, Phys. Rev. 177 , 777.
ing the neutron reflectometry technique they observed Aharoni, A., 1996, Introduction to the Theory of Ferromag-
an induced magnetic moment in the superconductor. Al- netism Clarendon Press, Oxford.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Page 50
1370 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

Alexander, JAX, TP Orlando, D. Rainer, and PM Ted- Bergeret, FS, AF Volkov, and KB Efetov, 2004b, Euro-
row, 1985, Phys. Rev. B 31 , 5811. phys. Lett. 66 , 111.
Altland, A., D. Taras-Semchuk, and BD Simons, 2000, Adv. Bergeret, FS, AL Yeyati, and A. Martín-Rodero, 2005,
Phys. 49 , 321. Phys. Rev. B 72 , 064524.
Anderson, P., and H. Suhl, 1959, Phys. Rev. 116 , 898. Blanter, YM, and FWJ Hekking, 2004, Phys. Rev. B 69 ,
Andreev, A., 1964, Sov. Phys. JETP 19 , 1228. 024525.
Androes, GM, and WD Knight, 1961, Phys. Rev. 121 , 779. Blum, Y., MKA Tsukernik, and A. Palevski, 2002, Phys.
Anthore, A., H. Pothier, and D. Esteve, 2003, Phys. Rev. Lett. Rev. Lett. 89 , 187004.
90 , 127001. Bourgeois, O., and RC Dynes, 2002, Phys. Rev. B 65 , 144503.
Artemenko, SN, and AF Volkov, 1980, Sov. Phys. Usp. 22 , Bulaevskii, LN, ML Buzdin, AI Kulic, and SV
295 Panyukov, 1985, Adv. Phys. 34 , 175.
Artemenko, SN, A. Volkov, and AV Zaitsev, 1979, Solid Bulaevskii, LN, VV Kuzii, and AA Sobyanin, 1977, Pis'ma
State Commun. 30 , 771. Zh. Eksp. Teor. Fiz. 25 , 314 JETP Lett. 25 , 290 1977 .
Aumentado, J., and V. Chandrasekhar, 2001, Phys. Rev. B 64 Buzdin,
, A., 2000, Phys. Rev. B 62 , 11377.
054505. Buzdin, A., 2005a, Rev. Mod. Phys. 77 , 935.
Bagrets, A., C. Lacroix, and A. Vedyayev, 2003, Phys. Rev. BBuzdin, A., 2005b, private communication.
68 , 054532. Buzdin, A., and I. Baladie, 2003, Phys. Rev. B 67 , 184519.
Baladie, I., and A. Buzdin, 2001, Phys. Rev. B 64 , 224514. Buzdin, AI, B. Bujicic, and M. Yu. Kupriyanov, 1992, Zh.
Baladie, I., and A. Buzdin, 2003, Phys. Rev. B 67 , 014523. Eksp. Teor. Fiz. 101 , 231 Sov. Phys. JETP 74 , 124 1992 .
Baladie, I., A. Buzdin, N. Ryzhanova, and A. Vedyayev, 2001,Buzdin, AI, and LN Bulaevskii, 1988, Sov. Phys. JETP 67 ,
Phys. Rev. B 63 , 054518. 576.
Balatsky, A., and E. Abrahams, 1992, Phys. Rev. B 45 , 13125. Buzdin, AI, LN Bulaevskii, and SV Panyukov, 1982,
Balatsky, A., E. Abrahams, DJ Scalapino, and JR Schrief- Pis'ma Zh. Eksp. Teor. Fiz. 35 , 147 JETP Lett. 35 , 178
fer, 1995, Phys. Rev. B 52 , 1271. 1982 .
Barash, YS, IV Bobkova, and T. Kopp, 2002, Phys. Rev. B Buzdin, AI, and MY Kupriyanov, 1990, Pis'ma Zh. Eksp.
66 , 140503. Teor. Fiz. 52 , 1089 JETP Lett. 52 , 487 1990 .
Bardeen, J., LN Cooper, and JR Schrieffer, 1957, Phys. Buzdin, AI, and MY Kupriyanov, 1991, Pis'ma Zh. Eksp.
Rev. 106 , 162. Teor. Fiz. 53 , 308 JETP Lett. 53 , 321 1991 .
Barone, A., and G. Paterno, 1982, Physics and Applications ofChampel, T., and M. Eschrig, 2005a, e-print cond-mat/0504198.
the Josephson Effect Wiley, New York. Champel, T., and M. Eschrig, 2005b, Phys. Rev. B 71 , 220506.
Baselmans, JJA, A. Morpurgo, BJ van Wees, and TM Chandrasekhar, BS, 1962, Appl. Phys. Lett. 1 , 7.
Klapwijk, 1999, Nature London 397 , 43. Charlat, P., H. Courtois, P. Gandit, D. Mailly, AF Volkov, and
Bauer, A., J. Bentner, M. Aprili, MLD Rocca, M. Reinwald, B. Pannetier, 1996, Phys. Rev. Lett. 77 , 4950.
W. Wegscheider, and C. Strunk, 2004, Phys. Rev. Lett. 92 , Chien, CJ, and V. Chandrasekhar, 1999, Phys. Rev. B 60 ,
217001. 3655.
Beckmann, D., HB Weber, and H. v. Löhneysen, 2004, Phys. Chien, CL, and DH Reich, 1999, J. Magn. Magn. Mater.
Rev. Lett. 93 , 197003. 200 , 83.
Bednorz, JG, and KA Müller, 1986, Z. Phys. B: Condens. Chtchelkatchev, NM, W. Belzig, and C. Bruder, 2002, JETP
Matter 64 , 189. Lett. 75 , 646.
Beenakker, CWJ, 1997, Rev. Mod. Phys. 69 , 731. Chtchelkatchev, NM, W. Belzig, Y. Nazarov, and C. Bruder,
Belitz, D., and TR Kirkpatrick, 1992, Phys. Rev. B 46 , 8393. 2001, JETP Lett. 74 , 323.
Belzig, W., A. Brataas, YV Nazarov, and GEW Bauer, Clogston, AM, 1962, Phys. Rev. Lett. 9 , 266.
2000, Phys. Rev. B 62 , 9726. Coleman, P., E. Miranda, and A. Tsvelik, 1993a, Phys. Rev.
Belzig, W., F. Wilhelm, C. Bruder, G. Schön, and A. Zaikin, Lett. 70 , 2960.
1999, Superlattices Microstruct. 25 , 1251. Coleman, P., E. Miranda, and A. Tsvelik, 1993b, Phys. Rev. B
Berezinskii, VL, 1975, JETP Lett. 20 , 287. 49 , 8955.
Bergeret, FS, KB Efetov, and AI Larkin, 2000, Phys. Rev. Coleman, P., E. Miranda, and A. Tsvelik, 1995, Phys. Rev. Lett.
B 62 , 11872. 74 , 1653.
Bergeret, FS, and N. García, 2004, Phys. Rev. B 70 , 052507. de Gennes, PG, 1964, Rev. Mod. Phys. 36 , 225.
Bergeret, FS, VV Pavlovskii, AF Volkov, and KB Efetov, de Gennes, PG, 1966, Superconductivity of Metals and Alloys
2002, Int. J. Mod. Phys. B 16 , 1459. Benjamin, New York.
Bergeret, FS, AF Volkov, and KB Efetov, 2001a, Phys. de Jong, MJM, and CWJ Beenakker, 1994, Phys. Rev.
Rev. Lett. 86 , 4096. Lett. 74 , 1657.
Bergeret, FS, AF Volkov, and KB Efetov, 2001b, Phys. Demler, EA, GB Arnold, and MR Beasley, 1997, Phys.
Rev. Lett. 86 , 3140. Rev. B 55 , 15174.
Bergeret, FS, AF Volkov, and KB Efetov, 2001c, Phys. Deutscher, G., 2005, Rev. Mod. Phys. 77 , 109.
Rev. B 64 , 134506. Deutscher, G., and PG de Gennes, 1969, Superconductivity
Bergeret, FS, AF Volkov, and KB Efetov, 2002, Phys. Dekker, New York, Vol. 2
https://translate.googleusercontent.com/translate_f 42/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
Rev. B 65 , 134505. Dimoulas, A., JP Heida, BJ van Wees, and TM Klapwijk,
Bergeret, FS, AF Volkov, and KB Efetov, 2003, Phys. 1995, Phys. Rev. Lett. 74 , 602.
Rev. B 68 , 064513. Dubonos, SV, AK Geim, KS Novoselov, and IV Grig-
Bergeret, FS, AF Volkov, and KB Efetov, 2004a, Phys. orieval, 2002, Phys. Rev. B 65 , 220513.
Rev. B 69 , 174504. Edelstein, VM, 1989, Sov. Phys. JETP 68 , 1244.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Page 51
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1371

Edelstein, VM, 2001, Phys. Rev. B 67 , 020505. Heslinga, DR, SE Shafranjuk, H. van Kempen, and TM
Eilenberger, G., 1968, Z. Phys. 214 , 195. Klapwijk, 1994, Phys. Rev. B 49 , 10484.
Eliashberg, GM, 1971, Sov. Phys. JETP 34 , 668. Huertas-Hernando, D., YV Nazarov, and W. Belzig, 2002,
Eremin, I., D. Manske, SG Ovchinnikov, and JF Annett, Phys. Rev. Lett. 88 , 047003.
2004, Ann. Phys. 13 , 149. Izyumov, YA, YN Proshin, and MG Khusainov, 2002,
Eschrig, M., 2000, Phys. Rev. B 61 , 9061. Phys. Usp. 45 , 109.
Eschrig, M., J. Kopu, JC Cuevas, and G. Schön, 2003, Phys. Jedema, FJ, BJ van Wees, BH Hoving, AT Filip, and T.
Rev. Lett. 90 , 137003. M. Klapwijk, 1999, Phys. Rev. B 60 , 16549.
Falko, VI, AF Volkov, and CJ Lambert, 1999, Phys. Rev. Jiang, JS, D. Davidovic, D. Reich, and CL Chien, 1995,
B 60 , 15394. Phys. Rev. Lett. 74 , 314.
Fazio, R., and C. Lucheroni, 1999, Europhys. Lett. 45 , 707. Kadigrobov, A., RI Skehter, and M. Jonson, 2001, Europhys.
Fogelström, M., 2000, Phys. Rev. B 62 , 11812. Lett. 54 , 394.
Fominov, YV, NM Chtchelkatchev, and AA Golubov, Karchev, NI, KB Blagoev, KS Bedell, and PB Little-
2002, Phys. Rev. B 66 , 014507. wood, 2001, Phys. Rev. Lett. 86 , 846.
Fominov, YV, AA Golubov, and MY Kupriyanov, 2003, Kharitonov, M., AF Volkov, and KB Efetov, 2005, unpub-
JETP Lett. 77 , 510. lished.
Fulde, P., and RA Ferrell, 1965, Phys. Rev. 135 , 550. Khusainov, MG, and YN Proshin, 1997, Phys. Rev. B 56 ,
Galaktionov, AV, and AD Zaikin, 2002, Phys. Rev. B 65 , R14283.
184507. Kirkpatrick, TR, and D. Belitz, 1991, Phys. Rev. Lett. 66 ,
Garifullin, IA, DA Tikhonov, NN Garif'yanov, MZ 1533.
Fattakhov, K. Theis-Broehl, K. Westerholt, and H. Zabel, Kontos, T., M. Aprili, J. Lesueur, F. Genet, B. Stephanidis, and
2002, Appl. Magn. Reson. 22 , 439. R. Boursier, 2002, Phys. Rev. Lett. 89 , 137007.
Ginzburg, VL, and LD Landau, 1950, Zh. Eksp. Teor. Fiz. Kontos, T., M. Aprili, J. Lesueur, and X. Grison, 2001, Phys.
20 , 1064. Rev. Lett. 86 , 304.
Giroud, M., H. Courtois, K. Hasselbach, D. Mailly, and B. Kopnin, NB, 2001, Theory of Nonequilibrium Superconduc-
Pannetier, 1998, Phys. Rev. B 58 , R11872. tivity Clarendon Press, Oxford.
Giroud, M., K. Hasselbach, H. Courtois, D. Mailly, and B. Kopu, J., M. Eschrig, JC Cuevas, and M. Fogelström, 2004,
Pannetier, 2003, Eur. Phys. J. B 31 , 103. Phys. Rev. B 69 , 094501.
Golubov, AA, 1999, Physica C 326-327 , 46. Kouh, T., and J. Valles, 2003, Phys. Rev. B 67 , 140506.
Golubov, AA, and MY Kupriyanov, 1996, Physica C 259 , Krawiec, M., BL Györffy, and JF Annett, 2004, Phys. Rev.
27 B 70 , 134519.
Golubov, AA, MY Kupriyanov, and YV Fominov, 2002a, Krivoruchko, VN, and EA Koshina, 2001a, Phys. Rev. B 63 ,
JETP Lett. 75 , 588. 224515.
Golubov, AA, MY Kupriyanov, and YV Fominov, 2002b, Krivoruchko, VN, and EA Koshina, 2001b, Phys. Rev. B 64 ,
JETP Lett. 75 , 190. 172511.
Golubov, AA, MY Kupriyanov, and E. Il'ichev, 2004, Rev. Krivoruchko, VN, and EA Koshina, 2002, Phys. Rev. B 66 ,
Mod. Phys. 76 , 411. 014521.
Golubov, AA, MY Kupriyanov, and M. Siegel, 2005, JETP Kulic, ML, and M. Endres, 2000, Phys. Rev. B 62 , 11846.
Lett. 81 , 180. Kulic, ML, and IM Kulic, 2001, Phys. Rev. B 63 , 104503.
Golubov, AA, FK Wilhelm, and AD Zaikin, 1997, Phys. Kulik, IO, and IK Yanson, 1970, The Josephson Effect in
Rev. B 55 , 1123. Superconducting Tunneling Structures Nauka, Moscow.
Gor'kov, LP, and EI Rashba, 2001, Phys. Rev. Lett. 87 , Kuprianov, MY, and VF Lukichev, 1988, Sov. Phys. JETP
037004. 67 , 1163.
Gor'kov, LP, and AI Rusinov, 1963, Zh. Eksp. Teor. Fiz. 46 , Lambert, C., and R. Raimondi, 1998, J. Phys.: Condens. Matter
1363 Sov. Phys. JETP 19 , 922 1964 . 10 , 901.
Gu, JY, C.-Y. You, JS Jiang, J. Pearson, YB Bazaliy, and Lambert, C., R. Raimondi, V. Sweeney, and AF Volkov,
SD Bader, 2002a, Phys. Rev. Lett. 89 , 267001. 1997, Phys. Rev. B 55 , 6015.
Gu, JY, C.-Y. You, JS Jiang, J. Pearson, YB Bazaliy, and Larkin, AI, 1965, JETP Lett. 2 , 130.
SD Bader, 2002b, Phys. Rev. Lett. 89 , 267001. Larkin, AI, and YN Ovchinnikov, 1964, Zh. Eksp. Teor.
Gubankov, VN, and NM Margolin, 1979, JETP Lett. 29 , Fiz. 47 , 1136 Sov. Phys. JETP 20 , 762 1965 .
673. Larkin, AI, and YN Ovchinnikov, 1968, Zh. Eksp. Teor.
Guéron, S., H. Pothier, NO Birge, D. Esteve, and MH Fiz. 55 , 2262 Sov. Phys. JETP 28 , 1200 1969 .
Devoret, 1996, Phys. Rev. Lett. 77 , 3025. Larkin, AI, and YN Ovchinnikov, 1984, Nonequilibrium
Guichard, W., MAO Bourgeois, T. Kontos, J. Lesueur, and Superconductivity Elsevier, Amsterdam, p. 530
P. Gandit, 2003, Phys. Rev. Lett. 90 , 167001. Lawrence, MD, and N. Giordano, 1996a, J. Phys.: Condens.
Gupta, AK, L. Crtinon, N. Moussy, B. Pannetier, and H. Matter 8 , 563.
Courtois, 2004, Phys. Rev. B 69 , 104514. Lawrence, MD, and N. Giordano, 1996b, J. Phys.: Condens.
Halterman, K., and OT Valls, 2002a, Phys. Rev. B 66 , 224516.Matter 11 , 1089.
Halterman, K., and OT Valls, 2002b, Phys. Rev. B 65 , 014509.Lazar, L., K. Westerholt, H. Zabel, LR Tagirov, YV
Hauser, JJ, HC Theurer, and NR Werthamer, 1963, Phys. Goryunov, NN Garif'yanov, and IA Garifullin, 2000,
Rev. 142 , 118. Phys. Rev. B 61 , 3711.
Heikkilä, TT, FK Wilhelm, and G. Schön, 2000, Europhys. Leggett, AJ, 1975, Rev. Mod. Phys. 47 , 331.
Lett. 51 , 434. Li, X., Z. Zheng, DY Xing, G. Sun, and Z. Dong, 2002, Phys.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Page 52
1372 Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena …

https://translate.googleusercontent.com/translate_f 43/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras
Rev. B 65KK,
Likharev, , 134507.
1979, Rev. Mod. Phys. 51 , 101. Quirion, D., B
Phys. Rev. C.65
Hoffmann,
, 100508.F. Lefloch, and M. Sanquer, 2002,
Lodder, A., and Yu. V. Nazarov, 1998, Phys. Rev. B 59 , 5783.Radovic, Z., L. Dobrosavljevic-Grujic, AI Buzdin, and JR
Luetkens, H., et al. , 2003, Phys. Rev. Lett. 91 , 017204. Clem, 1991, Phys. Rev. B 44 , 759.
Lyuksyutov, IF, and V. Pokrovsky, 2004, e-print cond-mat/ Radovic, Z., N. Lazarides, and N. Flytzanis, 2003, Phys. Rev. B
0409137. 68 , 014501.
Maeno, Y., H. Hashimoto, K. Yoshida, S. Nishizaki, T. Fujita,Rammer, J., and H. Smith, 1986, Rev. Mod. Phys. 58 , 323.
JG Bednorz, and F. Lichtenberg, 1994, Nature London Rashba, E., 1960, Sov. Phys. Solid State 2 , 1109.
372 , 532. Reymond, S., P. SanGiorgio, M. Beasley, J. Kim, T. Kim, and
Maki, K., 1968, Prog. Theor. Phys. 39 , 897. K. Char, 2000, e-print cond-mat/0504739.
Maki, K., 1969, Superconductivity Dekker, New York. Rusanov, AY, M. Hesselberth, and J. Aarts, 2004, Phys. Rev.
Marten, JP, A. Brataas, and W. Belzig, 2005, Phys. Rev. B 72 , Lett. 93 , 057002.
014510. Rusinov, AI, 1969, JETP Lett. 9 , 85.
McCann, E., VI Falko, AF Volkov, and CJ Lambert, 2000, Ryazanov, VV, VA Oboznov, AY Rusanov, AV Vereten-
Phys. Rev. B 62 , 6015. nikov, AA Golubov, and J. Aarts, 2001, Phys. Rev. Lett. 86 ,
McMillan, WL, 1968, Phys. Rev. 175 , 537. 2427.
Mélin, R., 2001, J. Phys.: Condens. Matter 13 , 6445. Saint-James, D., 1964, J. Phys. Paris 25 , 899.
Mélin, R., and D. Feinberg, 2004, Phys. Rev. B 70 , 174509. Sakurai, A., 1970, Prog. Theor. Phys. 44 , 1472.
Mélin, R., and S. Peysson, 2003, Phys. Rev. B 68 , 174515. Salkola, MI, AV Balatsky, and JR Schrieffer, 1997, Phys.
Melsen, JA, PW Brouwer, KM Frahm, and C. Beenakker, Rev. B 55 , 12648.
1996, Europhys. Lett. 35 , 7. Sarma, G., 1963, J. Phys. Chem. Solids 24 , 1029.
Mercaldo, V., C. Affanasio, C. Coccorese, L. Maritato, SL Sefrioui, Z., D. Arias, V. Peña, JE Villegas, M. Varela, P.
Prischepa, and M. Salvato, 1996, Phys. Rev. B 53 , 14040. Prieto, C. León, JL Martinez, and J. Santamaria, 2003, Phys.
Millis, A., D. Rainer, and JA Sauls, 1988, Phys. Rev. B 38 , Rev. B 67 , 214511.
4504. Sellier, H., C. Baraduc, F. Lefloch, and R. Calemczuk, 2004,
Mineev, VP, and KV Samokhin, 1999, Introdction to Uncon- Phys. Rev. Lett. 92 , 257005.
ventional Superconductivity Gordon and Breach, Amster- Serene, JW, and D. Reiner, 1983, Phys. Rep. 101 , 222.
dam. Shapira, S., EH Linfield, CJ Lambert, R. Serviour, AF
Mühge, T., N. Garif'yanov, YV Goryunov, GG Khaliullin, Volkov, and AV Zaitsev, 2000, Phys. Rev. Lett. 84 , 159.
LR Tagirov, K. Westerholt, IA Garifullin, and H. Zabel, Shelankov, A., and M. Ozana, 2000, Phys. Rev. B 61 , 7077.
1996, Phys. Rev. Lett. 77 , 1857. Shen, R., ZM Zheng, S. Liu, and DY Xing, 2003, Phys. Rev.
Mühge, T., N. Garif'yanov, YV Goryunov, K. Theis-Bröhl, K. B 67 , 024514.
Westerholt, IA Garifullin, and H. Zabel, 1998, Physica C Shiba, H., 1968, Prog. Theor. Phys. 40 , 435.
296 , 325. Stahn, J., J. Chakhalian, C. Niedermayer, J. Hoppler, T. Gut-
Nambu, Y., 1960, Phys. Rev. 117 , 648. berlet, J. Voigt, F. Treubel, H.-U. Habermeier, G. Cristiani, B.
Nazarov, YV, 1999, Superlattices Microstruct. 25 , 1221. Keimer, and C. Bernhard, 2005, Phys. Rev. B 71 , 140509.
Nazarov, YV, and TH Stoof, 1996, Phys. Rev. Lett. 76 , 823. Strunk, C., C. Sürgers, U. Paschen, and H. v. Löhneysen, 1994,
Nugent, P., I. Sosnin, and VT Petrashov, 2004, J. Phys.: Con- Phys. Rev. B 49 , 4053.
dens. Matter 16 , L509. Taddei, F., S. Sanvito, and CJ Lambert, 2001, Phys. Rev. B 63 ,
Obiand, Y., M. Ikebe, T. Kubo, and H. Fujimori, 1999, Physica012404.
C 317-318 , 149. Tagirov, LR, 1998, Physica C 307 , 145.
Ogrin, FY, SL Lee, AD Hillier, A. Mitchell, and T.-H. Taras-Semchuk, D., and A. Altland, 2001, Phys. Rev. B 64 ,
Shen, 2000, Phys. Rev. B 62 , 6021. 014512.
Oh, S., Y.-H. Kim, D. Youm, and MR Beasley, 2000, Phys. Tkachov, G., E. McCann, and VI Falko, 2002, Phys. Rev. B
Rev. B 63 , 052501. 65 , 024519.
Ostrovsky, PM, MA Skvortsov, and MV Feigel'man, 2001, Tokuyasu, T., JA Sauls, and D. Rainer, 1988, Phys. Rev. B 38 ,
Phys. Rev. Lett. 87 , 027002. 8823.
Palevski, A., 2005, private communication. Tollis, S., 2004, Phys. Rev. B 69 , 104532.
Peña, V., Z. Sefriouri, D. Arias, C. Leon, J. Santamaria, M. Tollis, S., M. Daumens, and A. Buzdin, 2005, Phys. Rev. B 71 ,
Varela, SJ Pennycook, and JL Martinez, 2004, Phys. Rev. 024510.
B 69 , 224502. Toplicar, JR, and DK Finnemore, 1977, Phys. Rev. B 16 ,
Petrashov, VT, VN Antonov, P. Delsing, and T. Claeson, 2072.
1995, Phys. Rev. Lett. 74 , 5268. Tsuei, CC, and JR Kirtley, 2003, Rev. Mod. Phys. 72 , 969.
Petrashov, VT, IA Sosnin, I. Cox, A. Parsons, and C. Troa- Usadel, KL, 1970, Phys. Rev. Lett. 25 , 507.
dec, 1999, Phys. Rev. Lett. 83 , 3281. Vaks, VG, VM Galitskii, and AI Larkin, 1962, Sov. Phys.
Pilgram, S., W. Belzig, and C. Bruder, 2000, Phys. Rev. B 62 , JETP 14 , 1177.
12462. Velez, M., MC Cyrille, S. Kim, JL Vicent, and IK
Pothier, H., S. Gueron, D. Esteve, and MM Devoret, 1994, Schuller, 1999, Phys. Rev. B 59 , 14659.
Phys. Rev. Lett. 73 , 2488. Volkov, AF, 1995, Phys. Rev. Lett. 74 , 4730.
Proshin, YN, YA Izyumov, and MG Khusainov, 2001, Volkov, AF, N. Allsopp, and CJ Lambert, 1996, J. Phys.:
Phys. Rev. B 64 , 064522. Condens. Matter 8 , 45.
Proshin, YN, and MG Khusainov, 1998, JETP 86 , 930. Volkov, AF, and A. Anishchanka, 2004, Phys. Rev. B 71 ,
Proshin, YN, and MG Khusainov, 1999, JETP 89 , 1021. 024501.

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

Page 53
Bergeret, Volkov, and Efetov: Odd triplet superconductivity and related phenomena … 1373

Volkov, AF, FS Bergeret, and KB Efetov, 2003, Phys. Xia, K., PJ Kelly, GEW Bauer, and I. Turek, 2002, Phys.
Rev. Lett. 90 , 117006. Rev. Lett. 89 , 166603.
Volkov, AF, and VV Pavlovskii, 1996, in Proceedings of the Yip, SK, 1998, Phys. Rev. B 58 , 5803.
XXXI Rencontres de Moriond , edited by T. Martin, G. Mon-You, C.-Y., YB Bazaliy, JY Gu, S.-J. Oh, LM Litvak, and
tambaux, and J. Tran Thanh Van Frontiers, France. SD Bader, 2004, Phys. Rev. B 70 , 014505.
Volkov, AF, AV Zaitsev, and TM Klapwijk, 1993, Physica Zaitsev, AV, 1984, Zh. Eksp. Teor. Fiz. 86 , 1742 Sov. Phys.
C 210 , 21.
JETP 59 , 1015 1984 .
Vollhardt, D., and P. Wölfle, 1990, The Superfluid Phases of He
3 Taylor and Francis, New York. Zaitsev, AV, 1990, JETP Lett. 51 , 35.
Wilhelm, FK, G. Schön, and AD Zaikin, 1998, Phys. Rev. Zareyan, M., W. Belzig, and YV Nazarov, 2001, Phys. Rev.
Lett. 81 , 1682. Lett. 86 , 308.
Wong, HK, B. Jin, HQ Yang, JB Ketterson, and JE Zyuzin, AY, B. Spivak, and M. Hruska, 2003, Europhys. Lett.
Hillard, 1986, J. Low Temp. Phys. 63 , 307. 62 , 97.

https://translate.googleusercontent.com/translate_f 44/45
11/05/2018 Supercondutividade do tripleto ímpar e fenômenos relacionados em estruturas ferro-elétricas supercondutoras

Rev. Mod. Phys., Vol. 77, No. 4, October 2005

https://translate.googleusercontent.com/translate_f 45/45