Вы находитесь на странице: 1из 20

Using The Hierarchy Equations of Motion to Simulate

Exciton Transfer in Photosynthesis

January 17, 2016

Contents
1 Introduction

2 Theoretical background

2.1

The antenna complex

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2

Hierarchy equations of motion . . . . . . . . . . . . . . . . . . . . . . . . .

2.3

Two dimensional spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . .

3 Method and Results

3.1

System bath model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.2

Writing the program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3.3

Simulating exciton transfer . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3.4

Creating electromagnetic spectra . . . . . . . . . . . . . . . . . . . . . . . 11

3.5

Two dimensional spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . 14

4 Discussion

15

4.1

Exciton transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

4.2

Electromagnetic spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4.3

Two dimensional spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

4.4

Improvements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

5 Conclusions

18

6 Acknowledgements

19
Abstract

A computer model was created to simulate pigment molecules in the photosynthesis system. This model is based on the hierarchy equations of motion, a set of
quantum mechanical equations. The model was used to show that the probability of
exciton transfer is larger if the excitation is being passed from a high energy molecule
to a lower energy molecule. It also revealed that the effectiveness of exciton transfer
is reduced if the molecules are tightly packed within the chloroplast. Investigations
of the electromagnetic spectra of the simulated pigment molecules showed that the
range of absorbance of the photosynthesis system can be increased by using pigments with different excitation energies. We saw that it can also be increased by
tighter molecular packing.

Introduction

In the past 10 years, advances in femtosecond pulse lasers and x-ray crystallography
have allowed us to view the shapes and positions of the molecules in the photosynthesis
system [8]. These discoveries, combined with improvements in computing power, enable
us to simulate the system at work.
The aim of my project is to simulate how energy is transferred during the early stages of
photosynthesis. I hope to use this simulation to investigate which factors affect energy
absorption and transfer. The Hierarchy Equations of Motion (HEOM) will form the
basis of my simulation. Simulations of the photosynthesis system have previously been
carried out by Dr. Alex Chin and others at the University of Cambridge [10], but outside
of the Tanimura group no simulations have been made using the HEOM.
The HEOM are a set of quantum mechanical equations which describe the time evolution
of a reduced density matrix in the canonical ensemble. Unlike the other reduced equations

of motion approaches, the HEOM are numerically exact and can be used for a finite
temperature bath [7]. See section 2.2 for an introduction to the HEOM.
Photosynthesis provides all of the food and most of the energy consumed by the human
race [8]. Discoveries in this field may lead to the production of genetically modified
plants, which have a greater capacity to produce food. These plants would enable us
to feed the growing world population. Ultra cheap solar panels could also be based on
knowledge gleaned from photosynthetic reactions.
The theoretical background of the experiment is summarised in the next section. Section 3 describes the method used and the results of each part of the experiment. In
Section 4, we discuss the results of the experiment, and potential sources of error. Section 5 states the conclusions we draw from the experiment.

2
2.1

Theoretical background
The antenna complex

The first stage of photosynthesis occurs in the antenna complex, here sunlight is absorbed
and energy is transported to the reaction centre (see figure 1). This energy drives a series
of redox reactions known as the Z scheme, which move protons across a membrane. The
proton gradient is then used to generate ATP, which can be used elsewhere in the plant
to build sugars.
The antenna complex consists of rings of pigment molecules, which surround the reaction
centre. The chlorophyll pigment molecules absorb light in the blue and red regions of the
visible spectrum, giving rise to the green colour of plants. When a photon is absorbed
by a pigment, it causes an electron to move from the orbital to the orbital within
the pigment. This exciton is passed along a chain of neighbouring pigments, eventually
reaching the reaction centre. Each pass occurs by a process known as exciton transfer.

2.2

Hierarchy equations of motion

The HEOM were developed by Y. Tanimura and R. Kubo in 1989 [9]. These equations
describe the time evolution of the reduced density matrix of a system in thermal contact
with a heat bath. The heat bath is characterised by the brownian oscillator model.
I.e., the bath radiates electromagnetic energy with a frequency distribution given by
J() = 2/( 2 + 2 ), where is the frequency of radiation, is the reorganisation
energy and 1/ represents the timescale of energy dissipation within the bath. The
HEOM are given by the following set of N equations;

Figure 1: A drawing of the antenna complex. The green disks represent chlorophyll molecules and the orange disks represent carotenoid
molecules.




i
n+1 n
n1 ,
n =
H + n n +
(1)
t
~ S
where 0 n N and N > 1/. Note that only 0 is equivalent to the reduced density
operator. The other levels of the hierarchy 1 , 2 , ... N are auxiliary operators which
take into account the fluctuation of energy levels within the system and the dissipation
of energy due to the heat bath. In all of my simulations, these matrices 1 , 2 , ... N
are set to zero for the initial condition.
The relaxation operators are given by
= iV ,

= i


2

V i V
,
~2
~

(2)

= AB
B
A,
and A B
=
where we have introduced the hyper-operator notations, A B
1

AB + B A. In equation (2), = kT , where k is Boltzmanns constant and T is the


temperature of the bath in Kelvin. ~ is Plancks constant.
4

Figure 2: Structure of the LH2 complex in purple photosynthetic bacteria. The chlorophyll pigments are shown in red, and the carotenoids
in orange. The yellow and grey molecules are structural proteins, which
hold the pigments in place. Figure taken from [8], p79.
S is the system Hamiltonian, it describes the energies of stationary states
The operator H
within the system. V represents the system bath coupling, and is determined by the
shapes of dipoles within the system.
The HEOM overcomes many limitations which the other reduced equations of motion
approaches suffer from. For example, the equations based on perturbation theory can
exhibit the positivity problem, whereby diagonal elements of the density matrix become
negative. This is unphysical, as the diagonal elements must express a probability between
zero and one. Another alternative is the Redfield equation, but this can only be used
for high temperatures.
The HEOM method can model the effect of a finite temperature bath in a non perturbative, numerically exact manner [1]. It has previously been applied in the areas of
quantum information [3] and multidimensional spectroscopy [2].

2.3

Two dimensional spectroscopy

Measurement of 2D spectra
Two dimensional spectroscopy allows us to observe the extent of coupling between different vibrational energy levels in a system, it can also be used to determine how excited
molecules behave over time. As mentioned in section 1, this measurement technique has
only become practical in the past 10 years, with the advancement of femtosecond laser
technology.
In order to make a 2D spectrum, the sample is injected with 3 pulses from a femtosecond
laser (see figure 3). Next, the light emitted by the relaxation of the molecules is observed.
The process is repeated several times, with different values of the excitation time t1 .
From these measurements, we can plot the response time t3 against t1 , giving a 2D
response function (figure 4). The 2D spectrum is found by taking a Fourier transform
of the 2D response function.

Figure 3: Measurement of 2D spectra. The three solid peaks represent


laser pulses incident on the sample, and the dashed line shows the
emitted light. The first two pulses are known as pump pulses, and
are used to excite the sample. The third pulse called the probe pulse,
it causes a transition back to the ground state. Figure courtesy of
Hironobu Ito, Kyoto University, with permission.

Interpretation of 2D spectra
It is instructional to think of 1 as the frequencies at which the system absorbs electromagnetic energy, and 3 as the frequencies at which this energy is emitted. Hence, a
peak on the line 1 = 3 corresponds to energy which is absorbed by a vibrational state,
and re-emitted by that same state. These are the bottom left and top right peaks in figure 5. On the other hand, the off diagonal peaks are produced when energy is absorbed
by a particular vibrational state, and re-emitted from a different vibrational state. I.e.,
the energy is transferred between states after absorption, and before emission. These

Figure 4: A simulation of the 2D response function. Blue and red


regions correspond to positive and negative fluctuations in the electric
field. The line t1 = 0 is equivalent to the 1D impulse response function
mentioned in section 3.4
states are said to be coupled. 2D spectra are a central part of my investigation, as the
size of the off diagonal peaks is directly related to the speed of excitation transfer. See
chapter 12 of [4] for a more complete introduction to 2D spectroscopy.

3
3.1

Method and Results


System bath model

I employed the system bath model in my simulations, whereby the state of the system
is calculated at every increment in time, and the bath is treated simply as a source and
sink of energy for the system. Let us assume that the time taken for the FORTRAN
MATMUL algorithm to carry out an n n matrix multiplication is O(n3 ) [11]. Adding
one additional molecule to the system would introduce at least one extra stationary state
to my system, therefore the size of the reduced density matrix would increase from n n
to at least (n + 1) (n + 1). Hence, the computation time for my simulation is no better
O(n3 ), where n is the number of molecules in the system. In other words, using more
molecules than necessary would greatly increase the computation time, forcing me to
increase the length of the time step, which would result in larger computational errors.

Figure 5: A 2D spectrum of two adjacent pigment molecules. E1 and


E2 are the excitation energies of pigment 1 and pigment 2. 1 and 3
are stated in arbitrary units.The purple regions correspond to peaks
in the spectrum. The bottom left peak is caused by the response of
pigment 1, and the top right peak is caused by the response from
pigment 2. The two other peaks are the result of coupling between
these states.
I chose to use two pigment molecules to make up my system. Two molecules is the minimum needed in order to simulate an exciton transfer, so it is the most computationally
efficient choice. Also, it is simple to quantify the extent of exciton transfer when there
are only two molecules involved.
As shown in figure 7, the system has three energy states;
|gi The HOMOs of pigments A and B are both fully occupied with two electrons. This
state is defined to have energy zero.
|e1 i Pigment A possesses one electron in its HOMO and one in its LUMO. Pigment B
has a fully occupied HOMO and an empty LUMO. This state has energy ~e1 .
|e2 i Pigment B possesses one electron in its HOMO and one in its LUMO. Pigment A
has a fully occupied HOMO and an empty LUMO. This state has energy ~e2 .

Figure 6: The system is made up of two pigment molecules, all of the


surroundings are modelled using a bath.

Figure 7: The three basis states of my system, |e1 i , |e2 i, and |gi. The
states |e1 i and |e2 i can interchange by a mechanism known as exciton
transfer (XT). Each of the diagrams shows the HOMO and LUMO of
pigment A and pigment B. The green dots represent electrons. Figure
courtesy of Hironobu Ito, Kyoto University, with permission.
The system Hamiltonian and reduced density matrix

~e1
J
0
e1 e1
= J
~e2 0 , = e2 e1
H
0
0
0
ge1
9

are as follows,

e1 e2 e1 g
e2 e2 e2 g ,
ge2 gg

where ~e1 and ~e2 are the excitation energies from ground state to states |e1 i and |e2 i
respectively, and J is the coupling energy between |e1 i and |e2 i.
The system bath coupling was set to,

1 0 0
V = 0 1 0 .
0 0 0

3.2

Writing the program

I wrote my code in the FORTRAN 90 programming language. I chose to use FORTRAN


as it is faster than other languages when working with arrays of data, and its long history
of use in science means there is a wealth of code that can be downloaded easily.
The Hierarchy Equations of Motion, (HEOM) were solved using a fourth order RungeKutta algorithm (RK4). This method is easy to use and understand. For a given
accuracy, it requires less computer time than other single step methods, such as the
Euler method. Also, unlike the multi step methods methods, it requires only one initial
condition.
I ensured that my calculation errors were small by reducing the time step by half and
re-running the simulation. If the final state of both the simulations was the same, then
I could be sure the size of the time step was not affecting my results.

3.3

Simulating exciton transfer

Once the program had been written, I put it to work in order to investigate which factors
affect the exciton transfer process. The density matrix was set to

1 0 0
= 0 0 0
0 0 0
for the initial condition. Which means that, at time t = 0, there is a 100% chance that
pigment A is found in the excited state and pigment B is found in the ground state.
This represents the quantum mechanical state of an antenna complex at the instant one
of its pigments is excited.
The time step, dt, was set to 1030 seconds and the constants in equation (1) were given
the following values,

1.00

0.50

0.25
10

J
-0.20

N
4

The results of two simulations are shown in figure 8. In simulation 1, pigment A has an
excitation energy of ~1 = 5 and pigment B has an excitation energy of ~2 = 2. In
simulation 2, both pigments have an excitation energy of ~1 = ~2 = 2.

Figure 8: These results show how the excitation energies of the two
pigments affects the probability of exciton transfer. The y axis shows
the probability of finding the system in a particular state. Time is
shown on the x axis, in arbitrary units. The blue line corresponds to
element e1 e1 of the density matrix, and the green line corresponds to
e2 e2 . Created using pyplot. See section 4.1 for a discussion of these
results.

3.4

Creating electromagnetic spectra

EM spectra are the primary method of observing organic molecules. Therefore it is


useful to calculate the spectra of my simulated pigments, allowing me to compare my
results with real experimental data.
The EM spectrum of a system is equivalent to the Fourier transform of its impulse
response function (IRF). The IRF can be measured by injecting the system with a short
pulse of laser light, and observing the subsequent changes in the electromagnetic field
(see figure 9).
The IRF of my system was calculated using the following formula,
R(t) =


i 
Tr
G(t)
eq ,
~

(3)

where eq is the equilibrium state of the density matrix, For non extreme temperatures, we can assume that the system is in equilibrium when in the ground state1 , i.e.,
1

The temperature T required for a 50% probability of excitation can be found by solving the equation

11

Figure 9: A method of obtaining the impulse response function (IRF).


The y axis shows electric field strength, and time is on the x axis. The
initial spike is caused by a femtosecond laser pulse, after which we see
the IRF of the system.
eq = |gi hg|. G(t) is the Greens function of equation (1), it represents the time propagation of any matrix to the right of it. For example,
G(t)
(0) = (t)
In the case of equation (3), The expression G(t)
eq is calculated by setting
eq as
the initial condition, and running the simulation for time t.
The transition dipole moment,
, represents the action of an electric field on the system,
it causes the transition between the ground state and an excited state;

0 0 1

= |e1 i hg| + |gi he1 | + |e2 i hg| + |gi he2 | = 0 0 1 .


1 1 0
I modified my code so that the value of R(t) is calculated at every time step, using
equation (3).
I used a Fast Fourier Transform (FFT) algorithm in order to create an EM spectrum
from the impulse response function. The source code fft.f90 was taken from the book
Numerical Recipes in Fortran 90 [6], which also contains an explanation of how fast
Fourier transform works. I decided a FORTRAN program was the easiest method for
making Fourier transforms of my data. As unlike Matlab FFT and other applications, it
can be run from the command line. Therefore I could create a batch file, which requires
only one command in order to compile and execute both my simulation and the FFT
code. Also, the code fft.f90 can be easily modified to carry out multi dimensional Fourier
transforms, which are used in section 4.3.
The step by step process for creating an EM spectrum is as follows;
1. Choose desired values for parameters , , , J , N , ~1 , ~2 , and dt.
exp(~) = 12 . For a 650nm transition, this gives T = 30 000 K.

12

2. Compile and run the simulation, starting with initial condition


eq . This outputs
values R(t) to a text file named R.txt.
3. Run the FFT algorithm, this reads R.txt and writes the Fourier transform to a file
called spec.txt.
4. Plot spec.txt using Pyplot.
Figure 10 shows a set of EM spectra, for various pigment excitation energies. Figure 11
shows how the dissipation , affects the EM spectrum of my system.

Figure 10: The calculated absorbance spectrum of a system of two


pigments. E1 is the energy required to excite pigment A, and E2 is
the energy required to excite pigment B. The wavelength is given in
arbitrary units. Notice how the width of the peak increases as the
difference between E1 and E2 increases.

13

Figure 11: The calculated absorbance spectrum of a system of two


pigments for various values of , the dissipation parameter. The wavelength is given in arbitrary units. Notice how the the peaks become
broader as increases.

3.5

Two dimensional spectroscopy

The following relationship was used to compute the nonlinear response function,
 3


i
R(t3 , t2 , t1 ) =
Tr
G(t3 )
G(t2 )
G(t1 )
eq .
~
In order to reduce the computation time of my simulation, t2 was set to zero, this gives
a response function which depends on two variables, t3 and t1 ,
 3


i
Tr
G(t3 )

G(t1 )
eq .
R(t3 , 0, t1 ) =
~

14

The computation of this function requires a nested do loop, here is the structure of my
program;
DO
propagate
eq to time t1 , giving matrix M (t1 )
DO
propagate

M (t1 ) to time t3 , giving matrix N (t3 , t1 )
calculate R(t3 , 0, t1 ) = ( ~i )3 Tr[
N (t3 , t1 )]
END DO
END DO.
Propagations are carried out in the same way as in sections 3.3 and 3.4, using the HEOM.
The FFT program was modified to carry out a 2D Fourier transform of the nonlinear
response function, giving the 2D Spectrum of the system. Figure (12) shows my results.

4
4.1

Discussion
Exciton transfer

Figure (8) shows my results for two exciton transfer simulations. A photon is absorbed
by pigment A (shown in blue) at time t = 0. In the first simulation, pigment A has a
greater excitation energy than pigment B. After 80 time units, there is an 80% chance
of exciton transfer. Whereas in the second simulation, both pigments have the same
excitation energy and the chance of exciton transfer is only 50% after 80 time units.
These results show that exciton transfer is more effective if the exciton is being passed
from a higher energy molecule to a lower energy molecule.
The excitation energy of a pigment is affected by its molecular structure. It can also be
increased by pigment protein interactions [8], p80.
It has been found that this principle is employed in nature, the pigments furthest from
the reaction centre have the highest energy excited states (see figure 13). This means the
system acts as a funnel; excitons are preferentially passed toward the reaction centre.
However, there is a trade off. When an exciton is passed to a lower energy molecule, the
extra energy is dissipated as heat. This results in a loss of efficiency. It is likely that
3000 million years of photosynthesis evolution has led to an optimum energy difference

15

Figure 12: The effect of on the 2D spectrum of the system. 1 and


3 are given in arbitrary units. Notice how the purple coloured peaks
become less well defined as increases.
between adjacent pigments.

4.2

Electromagnetic spectra

Figure 10 shows my calculated absorption spectrum of two pigment molecules. It can


be seen that a narrow peak is produced when the molecules have the same excitation
energy. The range over which absorption occurs is increased if the two molecules have
different excitation energies.
A broader absorption spectrum can allow the organism to absorb more light and photosynthesise more effectively. This is a second evolutionary reason for pigments to have
different excitation energies (see previous section). A broad absorption spectrum is
particularly useful in organisms which do not have access to direct sunlight, such as
16

Figure 13: The positioning of pigments around the reaction centre


(RC) in purple bacteria. The pigments are labelled according to the
frequencies of light that they absorb. For example, the B850 pigment absorbs 850nm light. Notice how the wavelength increases with
proximity to the reaction centre. Reproduced from Fleming and van
Grondelle (1997).
plants living on the forest floor. Another example is Prochlorococcus, a genus of deep
sea plankton which contains a unique form of chlorophyll not found in any other organism ([8], p20).
A similar spectral broadening effect is seen in figure 11, in this case it is due to an
increase in the dissipation parameter . The dissipation parameter describes the effect
of bath on the system. If was set to zero, the peaks would be unrealistically narrow.
This shows that it is essential to incorporate a bath in the simulations.
In a real organism, can be increased by moving the surrounding molecules closer to the
pigments, which could be achieved by using genetic modification to change the shape
of supporting proteins in the antenna complex (see figure 2). Hence the results shown
in figure 11 would be useful for a food scientist attempting to change the absorption
characteristics of a crop.

4.3

Two dimensional spectra

Figure 12 shows a series of 2D spectra for two pigments with various values for the dissipation parameter . It is a two dimensional version of the results shown in figure 11.
17

The size of the top left and bottom right purple regions corresponds to the amount of
energy transferred from one state to the other. (See section 2.3 for an explanation of this
process). These regions become smaller as the energy difference between the molecules
increases, suggesting the effectiveness of exciton transfer is reduced by a larger dissipation parameter. The fourth plot shows an extreme case of large dissipation, the peaks
are indistinguishable from each other and very little energy is absorbed.
It can be seen that increasing also causes the peaks to lose their diamond appearance
and become more circular. This is due to the fact that the dissipation term causes
the energy levels of the system to fluctuate. Therefore the energy of the absorbed and
emitted photons become more spread out, forming a less defined peak on the plot.

4.4

Improvements

In its current form, the model cannot make predictions of the values of excitation transfer
time, wavelength absorption, and broadness of the absorption peak. The result of these
quantities are given in arbitrary units.
In order to make quantitative predictions, the model needs to include real values for the
all of the parameters used. The values of E1 , E2 , and are well known. However, due
to the juvenility of the HEOM, the values of , , and J have not been measured for an
antenna system.
, , and J could be found by reverse engineering the model. I.e, it could be used to
make qualitative predictions about a simple system, and the parameters adjusted until
these predictions are correct.

Conclusions

A computer model of pigment molecules within the photosynthesis system was created.
This system behaves similarly to the real system, and was used to create realistic 1D
and 2D spectra.
It was found that exciton transfer is more probable when passing from a high energy
molecule to a lower energy molecule. Exciton transfer is also more probable if there
is a small dissipation parameter, which can be achieved with a loosely packed antenna
complex.
The spectra show that the frequency range of light absorption can be increased by
incorporating molecules with different excitation energies into the antenna complex, or
by increasing the dissipation parameter.
18

Further work needs to be done in this area in order to determine the values , , and
J. This would enable the model to make qualitative predictions, such as the time taken
for an excitation transfer, or the broadness of the absorption peak of the pigments.

Acknowledgements

I would like to thank Prof. Y. Tanimura, H. Ito, and J. Y. Jo for teaching me the
relevant areas of quantum mechanics and spectroscopy required for this investigation.
The project was made possible with funding from the Amgen scholars program.

Glossary
ATP adenosine triphosphate, a molecule used for energy storage in living organisms. 3
exciton a bound state of an electron and an electron hole, excitons are produced when
ground state electrons move to a higher state. 1, 3, 7, 8, 10, 11, 15, 17
multi step method a method of numerical integration which uses more than one previously calculated solution, xn , xn1 ... in order to calculate the next solution xn .
10
pigment molecule a molecule which absorbs in a region of the visible spectrum. Chlorophyll and carotenoids are the main pigment molecules involved in photosynthesis..
3, 6, 7, 16
reduced density matrix a density matrix which describes the state of the system,
but not the bath. 2, 3, 7, 8
single step method a method of numerical integration which uses only the most recently calculated solution xn in order to calculate the next step xn+1 . 10
stationary state a state of a quantum system which has only one possible energy,
particles in a stationary state have a probability density function which does not
vary in time. 4, 7
time step the time between successive calculations in a computer simulation. Using a
large time step requires fewer calculations, but produces larger errors in the results.
7, 10, 12

19

Z scheme a series of redox reactions occurring in photosynthesis. These reactions split


H2 O into O2 and H+ , as well as converting ADP into ATP. The Z scheme is a form
of electron transport chain. 3

References
[1] Xu R. X., Cui P., Li X. Q., Mo Y., Yan Y. J., 2005. Exact quantum master
equation via the calculus on path integrals J. Chem. Phys. 122, 041103
[2] Ishizaki A., Tanimura Y., 2006. Modeling vibrational dephasing and energy relaxation processes of intermolecular anharmonic modes for multidimensional infrared
spectroscopies J. Chem. Phys., 125, 084501
[3] Ma J., Sun Z., Wang X., Nori F., 2012. Entanglement dynamics of two qubits in
a common bath Phys. Rev. A, At., Mol., Opt. Phys. 85, 062323,
[4] Baiz C., Reppert M., Tokmakoff A., 2013. Ultrafast Infrared Vibrational Spectroscopy CRC Press, p361
[5] P. M. Mathews, K. Venkatesan Textbook Of Quantum Mechanics, 2nd Edition 2010.
Tata McGraw-Hill Education
[6] Press W., Teukolsky S., Vetterling W., Flannery B., 1996. Numerical Recipes in
Fortran 90. Cambridge University Press, England.
[7] L. Chen, Y. Zhao and Y. Tanimura, 2015 Dynamics of a one-dimensional Holstein
polaron with the hierarchical equations of motion approach, J. Phys. Chem, Let.
6, p3110-3115.
[8] Blakenship, R. E., 2002. Molecular Mechanisms of Photosynthesis Blackwell Science.
[9] Tanimura, Y., and Kubo, R., 1989. Time Evolution of a Quantum System in
Contact with a Nearly Gaussian-Markoffian Noise Bath Physical Society of Japan,
58, p101
[10] Caruso, F., Chin, A. W., Datta, A., Huelga, S. F., Plenio, M. B., 2009. Highly
efficient energy excitation transfer in light-harvesting complexes: The fundamental
role of noise-assisted transport J. Chem. Phys. 131, p105106
[11] Robinson, S., 2005. Toward an Optimal Algorithm for Matrix Multiplication
SIAM News 38 (9)

20

Вам также может понравиться