JOURNAL OF THE OPTICAL SOCIETY OF AMERICA VOLUME 51, NUMBER 8 AUGUST, 1961

Emission Spectra of the Doubly and Triply Ionized Rare Earths*

G. H. DIEKE, H. M. CROSSWHITE, AND B. DUNNt
The Jo/hns Hopkins University, Baltimore, Maryland
(Received February 6, 1961)

The spectral of the rare earths have been photographed under controlled excitation so that either the
spectra of the doubly or triply ionized elements are brought out with maximum intensity. A mild excitation
is used in addition to give the first and second spectra for comparison. Some regularities are immediately
apparent and vary very gradually through the rare-earth group. The general features of the spectra are
discussed in this paper.

1. INTRODUCTION spectively. The two-electron spectrum Ce in (Russell,

THE spectra of the rare earths and the correspond-
ing actinides are the most complex spectra in
existence and, accordingly, the least well known. Most
of the investigations in the past have been concentrated
on the first and second spectra, while very little work
has been done on the higher stages of ionization. On
the other hand, the fourth spectra, those of the triply
ionized atoms, are of very great importance because of
their connection with the crystal spectra of the rare-
earth salts. As is well known, the salts of the trivalent
rare earths show elaborate sharp-lined absorption and
fluorescence spectra which are due to transitions
between the various levels of the 4fn configuration. For
an understanding of the structure of these crystal
spectra and the crystal forces which modify the levels
of the ion inside the crystal, it is very desirable to have
the levels of the free ion for comparison. This would
make it possible to find exactly the modifications caused
by the crystal field. No such comparison has been pos-
sible in the past because the free ion spectra have been
completely unknown except for Cei+, where so far the
crystal spectrum is not known.
In order to remove these deficiencies in our knowl-
edge, an experimental study of the free-ion emission
spectra has been started in this laboratory. It soon
became evident, for reasons elaborated below, that this
study should also encompass the spectra of the divalent
ions.
Results are now available on several elements in
various stages of completion. They will be dealt with
in detail in future publications. The present paper is
intended to give the common experimental background
and the general results which can be derived without
going into the complex details of the structure of the
energy level systems.
While very little is known about the third and fourth
spectra of the rare earths, the few cases where a reasona-
bly complete analysis exists have been of great help.
The one-electron spectra Ce IV and La III have been
investigated by Lang' and Russell and Meggers2 , re-
* Work carried out with the support of the Air Force Office of
Scientific Research, Air Research and Development Command.
t Now with Braddock, Dunn, and McDonald Inc., El Paso,
Texas.
1 R. J. Lang, Can. J. Research A14, 127 (1936).
2 H. N. Russell and W. F. Meggers, Bur. Standards J. Research
9, 625 (1932).
820
King, and Lang'), and fragments of the spectrum of
Lu iii (Meggers and Scribner 4) are also known.
The rare-earth spectra have the deserved reputation
of being extremely complex and very rich in lines. The
methods for attacking the problem had to be carefully
scrutinized in order to achieve some success in a
reasonable time. The problem divides itself into the
following natural stages: (a) production and photog-
raphy of the spectrum with separation of the different
ionization stages; (b) measurement of the wavelengths
and the establishment of reasonably complete tables
of the various spectra; (c) empirical analysis, i.e.,
establishment of the energy levels from the observa-
tions; (d) identification and interpretation of the levels,
and comparison with the results derived from theoreti-
cal considerations; and (e) comparison of the free-ion
levels with the crystal levels.
While the accomplishment of stage (e) was the prime
moving factor for the investigation, we shall leave it
out of consideration in the present paper. Stage (a)
represents the chief experimental effort and its general
features will be described in Sec. 2. (b) represents
mainly routine work but is very time consuming
because of the large number of lines involved.
Stage (c) can also be largely reduced to routine work
with the help of computing machines if a sufficiently
accurate and complete wavelength table is available,
i.e., if the work of stages (a) and (b) has been done well.
Finally the real theoretical analysis, stage (d),
requires detailed consideration of each individual case
and its discussion must be left for the further individual
papers.
Although, originally, the identification of the levels
of the 4f" configuration which give rise to the crystal
spectrum was the principal aim, it was neither feasible
nor desirable to restrict the investigation to this con-
figuration. In the first place, the free-ion analogs of the
crystal lines, the lines themselves being forbidden
transitions, do not appear in the emission spectrum. It
is possible, therefore, to obtain the 4fn configuration
only through allowed transitions, among which the
most prominent ones are 4fn 4fn-1 5d. In turn, the
3H. N. Russell, R. B. King, and R. J. Lang, Phys. Rev. 52,
456 (1937).
4 W. F. Meggers and B. F. Scribner, Bur. Standards J. Research
5, 73 (1930); 19, 651 (1937).
August 1961 SPECTRA OF
4fa-1 5d levels are connected by allowed transitions to
the 4fn-1 6p configuration which, in turn, is connected
to the 4Jn-1 6s levels. It is therefore advisable to study
at least these four basic configurations and the transi-
tions between them which should, in general, furnish
the strongest lines of the spectrum.
Since for some of the elements each configuration may
have hundreds of levels, it is evident that even the
spectrum resulting from only these four configurations
will generally be extremely complicated. For the analy-
sis of such spectra it is always advantageous to have
other spectra of similar structure with which they can
be compared. Those of the isoelectric ion, e.g., Pr iii,
which is isoelectric with Nd iv, will be particularly
useful. Pr iII is also connected to Pr iv, which serves as
its parent ion. This indicates that the third spectra
should be studied together with the fourth. This is all
the more desirable, as accurate empirical data can be
obtained with greater ease for the lower stages of
ionization.
The inclusion of the first and second spectra in this
study, which at first sight seems a logical extension, did
not appear to have any great advantages. In the first
place, work on the first two spectra is being undertaken
at present in several other laboratories. Moreover, the
relative positions of the energy levels are so different
from those of the fourth spectrum that the analogies
unquestionably present would often be so hidden as to
be of no particular aid in the analysis.
2. EXPERIMENTAL PROCEDURE
It is not possible to obtain the spectrum of one stage
of ionization free of that of the neighboring stages, but
in each case it is possible to find conditions which will
bring out a particular stage at its maximum intensity.
A number of different discharges were tried, and finally
three were chosen which give the desired results in the
simplest possible way. The essential features of each
are given in the following table:
Peak Pulse Induct-
current length ance Capacity Pressure
Source (A) (usec) Voltage (ph) (paf) (mm He)
A. dc arc 3 *.. 250 *.. ..
140
B. mild spark
C. hot spark
50
900
2000
90
300
1000
2500
30 140
Many modifications have been tried and further infor-
mation about them may be found in the detailed papers.
Since the ionization potentials of the rare earths are
relatively low, the neutral and singly ionized spectra
are both well developed in the dc arc and even the third
spectrum appears weakly. In the mild spark discharge,
the first and second spectra are greatly weakened and
the third enhanced, while the fourth also now appears
weakly. The hot spark greatly enhances the fourth
spectrum while leaving the third strong, but the first
and second spectra have become nearly extinct. More
IONIZED RARE EARTHS 821
energetic discharges are not used because of the de-
terioration in line sharpness.
In the spark discharges B and C, an ignitron type
GL-7171 is used as a switching tube. The ignitron is
triggered by a FG-105 mercury thyratron, which in
turn is controlled by the voltage on the capacitor bank.
The discharges are critically damped by the addition
of a few ohms resistance to the discharge circuit. Dis-
charge efficiency is improved and electromagnetic
radiation virtually eliminated.
Experience with these nearly critically damped dis-
charges has shown them to be superior to oscillatory or
square pulses for the separation of the ionization stages.
The current-wave-form signal as developed across a
low inductance 0.0136-ohm resistor is displayed on an
oscillograph. It rises immediately to the peak value
given in the table and decays almost linearly to zero.
The pulse length is sharply defined, as there is no over-
shoot. This is important for the suppression of the un-
desirable ionization stages.
Helium carrier gas is used to minimize corrosion of the
metal electrodes which generally are very reactive. The
gas pressure is chosen for best containment and radiance
of the discharge. Wandering and diffusion were also
restricted by a C-shaped Lavite disk placed between
the electrodes. This was made by drilling a -!--in. hole
in the center, sawing half-way through, and then hard-
ening by baking. When it is necessary to obtain rare-
earth spectra free of helium lines, other rare gases have
been used.
The data were obtained with three spectrographs.
The wavelength region from 1900 to 9900 A was photo-
graphed with a 21-ft concave grating in a Paschen
mounting with a 7-in. grating having 30 000 lines/in.
and a linear first-order dispersion of 1.2 A/mm. In
general the lines above 4900 A were obtained in the
first order, those between 3200 and 4900 A in the
second, and those below 3200 A in the third order,
which gave maximum resolution and intensity because
the blaze of the grating was at rather high angles. The
wavelengths beyond 9900 A to about 1.2 u were ob-
tained in the first order with a 21-ft concave grating in
a Wadsworth mounting with a dispersion of 5 A/mm.
500 Many of the strong lines of the third and, particularly,
20
20 the fourth spectra fall in the vacuum ultraviolet and
this region, from about 500 to 2000 A, was covered by
a 2-m vacuum spectrograph with grazing incidence and
an average dispersion of 5 A/mm.
The separation of the orders presents no difficulty in
the first order throughout, and none for the longer
wavelengths of the second and third orders. However,
this cannot be achieved with existing filters for wave-
lengths shorter than about 2700 A, and the following
device, which depends on selective total internal re-
flection, was developed to separate the orders.
5
A pair of quartz parallelepipeds is adjusted so that
H. M. Crosswhite (to be published).
822 I)IEKE, CROSSWHITE, AND DUN N Vol. 51

0II*A

2730 2740 2750 2760 i770 2'0

A
: s : : . -~~~~~~~
B

2780 2790 2810 2820

--- ;" I, IEE'I: I

1t1
1 1 1 X I W 1
FIG. 1. The praseodymium spec-
c~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

trum with different degrees of

3320 :.,
excitation. A, hot spark; B, mild
3330 3340 . ' - 3350 ; 3360
spark; C, arc. Lines marked by
four or three dots above the spec-
tra belong to Pr iv or Pr in, re-
spectively; those with two dots
below, to Pr II or Pr I.
I11 A141
.~~~~~~~~~~~

3360 `0> 3370 3380 3390 V 3400

I :

~~B

6010 6020 ,"I ' 6030 6040 6050 6060 6070 6080 6090 6100 6110

. . . ....... . .0 - .............................-
- A_
6e O 6120 630 6140 6150 6160 6170 6180 6190 6200

for some convenient wavelength, say 2600 A, a parallel the same intensity in B and C but weaker in A, while
beam passes through at just the critical angle for this Pr ii lines are almost entirely confined to spectrum C.
wavelength. Longer wavelengths will be only partially It is generally, therefore, not difficult to make a decision
reflected internally, but, because of their higher index regarding the stage of ionization to which a particular
of refraction, shorter wavelengths will undergo total line belongs. Figure 2 gives similar information for the
internal reflection. For eight such reflections, the trans- ytterbium spectrum in the vacuum region. Here a de-
mittance for wavelengths greater than 3000 A is less cision need be made only between Yb in and Yb iv as
than 1%. Quartz of very high purity must be used, the spectra of Yb and Yb ii do not exist in this wave-
since the optical path of the internal beam is over 5 in. length region. The transitions of Yb iv occurring in this
However, the transmittance of this system is greater region are marked above the spectrum.
than 60% at 1900 A, and, of course, better for longer Figure 1 shows many broad lines (e.g., near 3380 A)
wavelengths. which on the original plates have resolved hyperfine
Figure 1 shows the situation for the praseodymium structure. As large hyperfine structure is found only
spectrum; A represents the hot spark; B, the mild when at least one of the outer electrons is in an s state,
spark, and C, the arc. A number of positively identified the hyperfine structure is a very helpful criterion for the
lines are marked by dots above or below the spectrum classification of the lines.
(two dots for Pr or II, three and four for Pr iII and iv,
respectively). It is seen that Pr iv lines are much 3. RESULTS
stronger in spectrum A than in B, and very faint or A detailed energy-level diagram of the doubly and
entirely absent in C. Pr iII lines are of approximately triply ionized rare-earth atoms can only be provided
August 1961 SPECTRA OF IONIZED RARE EARTHS 823

after the completed analysis of the spectra. This will

take some time, as even the wavelength measurements
alone in these spectra, which are all extremely rich in
lines, represent a major undertaking. Several general
features, however, become apparent even with a pre-
liminary inspection of the spectra. These qualitative
features are brought out in Fig. 3, which gives a re-
production of low-dispersion spectra from 800-3500 A
6
of all the rare earths except promethium.
The figure shows that the lines occur in clusters and
that these change systematically from one element to
the next. This suggests that the analogous groups have
a similar origin in each element. Extrapolation to the
one-electron spectra at each end, where the structure is
simple and the analysis known, gives the clue to the
interpretation. FiG. 2. The ytterbium spectrum between 900 and 1900 A;
We see that the two groups found at the longest A. hot spark, B. mild spark. The principal line groups of Yb IV
reference 8).
wavelength in the third spectra must be the 4f8-I 6s- are marked (B. W. Bryant,
4fn-l 6p transitions. In La iii this is a doublet with a
separation of 3096 cm-l at 3172 and 3517 A. This detailed analysis, but the results in Yb iii and Gd iII
6s-6p transition shifts regularly to shorter wave- support this. The relative shift of the 5d and 6s levels
lengths with increasing Z and is at about 2200 and seems to be due to the fact that for low Z the 5d levels
2600 A for LuIII. The' doublet separation becomes are still partially penetrating. Because of stronger con-
gradually larger, as the increase in the nuclear charge traction of the core with increasing Z, this penetration
is not entirely compensated for by the screening of the becomes less, with the result that the energy of a 5d
added 4f electrons. In the intermediate elements, the level rises above that of 6s where penetration remains
6s and 6p energies are superimposed on the elaborate important.
system of the 4Jni-core configuration, which can extend In the fourth spectra the 5d-6p transition in Ce iv
50 000 cm-l or more. Nevertheless the reason that we is near 1300 A and, if the qualitative evidence of Fig. 3
see the lines of the 4fn- 6s-4f-' 6p transition clus- is trustworthy, shifts only very little so that it is near
tered in relatively narrow wavelength regions is that 1400 in Lu iv.
for the strongest lines the 4fn-1 core remains unchanged For our immediate purpose, the location of the levels
when the outermost electron goes from 6p to 6s, because of the 4fn configuration is most important, because the
the coupling is close to Jj coupling. transitions between the levels of this configuration in
In the fourth spectra the analogous 6s-6p transitions the fourth spectra give us the crystal spectra of these
can also be picked out easily. For Ce iv the doublet is at ions. The strongest transitions involving the 4fn levels
2457 and 2778 A, with a doublet separation of 4706.8 are the 4fn -4fn-1 5d transitions which in Ce iv lie
cm-l. It also shifts gradually to shorter wavelengths just below 2000 A and appear to move to shorter wave-
and is near 1800 and 2100 A for Yb iv. lengths with increasing Z. Beyond europium these
The 6s-6p transitions are of prime importance for transitions can no longer be followed on our survey
the analysis of the third and fourth spectra; of the spectra but appear near 1000 A for Yb Iv.
easily recognizable fundamental transitions they alone For the third spectra, the 4fn configuration is com-
are in the region of the spectrum where large spectro- pletely unknown even 2for the simplest cases. For La 2
ii

graphs can be used and therefore very accurate wave- the lowest level of 4f lies 55 000 cm'l above 5d ; in
lengths can be obtained. Ba i this is about 44 000, while in the isoelectric Ce iv,
2
The 4fn-1 5d-4fn-1 6p transitions center in La iII 4f is 45 000 below 5d and in Pr iv 4f is 50 000 below
around 2300 A with the lowest Sd level about 13 600 4f 5d. If we interpolate the third spectra between the
cm-' below the 6s level. This transition appears to move second and fourth spectra, we see that we must expect
very gradually to the longer wavelengths remaining the 4fn and 4f-I Sd configurations to have approxi-
nearly stationary in the beginning. If the analysis of mately the same energy for the beginning of the rare-
Lu iII is correct it centers around 2800 A in that spec- earth group. The exact location cannot, of course, be
trum, which means that somewhere near Dy iII the 6s predicted in this way with any accuracy, but it is evi-
and 5d levels must cross. The appearance of this group dent that the strong 4f-5d transitions should be7 in
is less conspicuous than the previous one and is more the infrared and the visible. They have been found in
difficult to trace in the intermediate elements without Pr iII where the lowest 4f3 level 4Ig/2, the ground state
6
We first saw the regular sequence of agglomeration of lines of the ion, lies about 11 000 cm-l below the lowest level
between 2300 and 3000 A similar to part of Fig. 3 in a collection 7 J. Sugar (to be published).
of rare-earth spectra prepared by M. Fred and F. Tomkins.
824 DIEKE, CROSSWHITE, AND DUNN Vol. 51

-000
> S : E t > r000;0t0t 02

~0
ot*
0

t- | g s1
cli~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~c

to~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~c

('4~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.

C~~~~~~~~~~~~
0 l. ~~ 0 ~~ ;~~~~~~~~~a
f r
~~~~~~~~~~~~~~~~~~~~~~~~

0 ~~~~~~~~~~~0
- oO.
>
U

,~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~c e- d;X;

a a, Tn E : -0 > 0
J U CL z co uJ CD C) T. Lw
August 1961 SPECTRA OF IONIZED RARE EARTHS 825

of the 4f' 5d configuration (see Fig. 4). It has also been

observed 8 in Yb iii, where the 4f14 configuration consists 80 d
of only one 'So level about 35 000 cm-' below the 4f" 5d
configuration. Here, some of the 4f-5d transitions lie 60
in the ultraviolet below 2000 A. 40
The location, even approximately, of the 5d levels in
the fourth spectra is of some importance for the study
of the crystal spectra. The 4f-5d transitions should be -20
Dx00t'f = 6dX
5d
about 106 times as strong as the forbidden transitions
within the 4fn configuration which give rise to the
sharp-line crystal-absorption spectrum. This absorption
spectrum is well known in many cases in the visible and 100. x1OO cm- _

near ultraviolet, but very little is known below 3000 A.

80 rsd
Often continuous absorption has been observed in this
region. This might be attributed to absorption to the 60 6p

5d levels. The free-ion results show that no absorption 40

to 4fn-1 5d can be expected above 2000 A even if the 20
broadening of these levels should be considerable. The
continuous absorption between 2000 and 3000 A should 0
probably be attributed to impurities. This makes it - 20
fiuan ine rd arndS furt Tb
spetr ofh
5d
are earths.
probable that the levels of the 4fn configuration of the Celogn
lecrno Pr the Eu GdelTb idictd ErThe we partL
-40-
crystal spectra can be observed to at least 50 000 cm-l.
-60-

-80-
140
Even
x I000 cm-,
-120 f

20k_ PrIII _f16

FIG. 5. Relative position of the centers of the principal con-
figurations in the third and fourth spectra of the rare earths.
Circles are data obtained from actual analyses. Only the outer
electron not belonging to the 4f shell is indicated. The lower part
marked RI+ represents the levels of the triply ionized atoms; the
Odd upper part, those of R+±.

The extension of the observed data to this limit should

90t- yield many theoretically interesting levels.

80k I_ Figure 4 shows the extent of the low configurations

in Pr iii. The calculations were made with the empirical
parameters obtained from a partial analysis of the
70k Pr iii spectrum by Sugar7 and are sufficiently accurate
to show the range on this scale. At least six of the con-
60- figurations shown in the figure are now empirically
known.
Figure 5 gives an indication of the relative positions
50k
of the centers of the lowest configurations. The circles
are empirical results from spectra which have been
40k analyzed in detail. By linear interpolation it is possible
to obtain the relative positions of the levels of the
30k intermediate elements at least with sufficient accuracy
to locate the position of the chief line groups in the
20k spectrum. This gives a more quantitative formulation
of the qualitative results obtained by inspection of
I0 Fig. 3. It confirms the crossing of the Sd and 6s levels
near Dy III. The parameters obtained from the analyses
of a few spectra can be used to obtain approximate
values of these parameters for the intermediate energy
FIG. 4. The lowest configurations of Pr III.
levels. This can be used to calculate the structure of the
I B. W. Bryant (to be published). energy levels in each configuration. Thus it is possible
826 DIEIKE, CROSSWVHITE, AND DUNN Vol. 51

Lx
36 _ 1

10 cm-,
34 I
..tso
- 6p3 H

32 2
G

0
P
M
30 F
0 0

N
_ P.
E N

28 o D _ M
L
N C
2
L K0
26 1

II
-K

K
_ 2

2 M~~~~~~~ 2
K H g
2ow
24 2D.5 I
J

H 4
- I F 3
3 ZH
p
22 -~~~
G
-
I
- ~~lq-
H
I2
F 5

II

_
F
E
_2 F H
4
- 'G4
-0 A 5I NW G F 7
_- 4 _ - I
20 F T
F H 1
8 - _ 1
E E 4s 3 FIG. 6. The lower energy
18 . .-l- A 4F I
levels of the 4fn configuration
-0 for the triply ionized rare
196iI earths as obtained from crystal
16 D D 4F
absorption and fluorescence
0,Fg_ -2
spectra.
-3I
14 A S
E OF _

_8 _ 3F
12 ";i D
-
L
I

I
G o° F II OF
a A Li
10 A 7-
2
2
____

P 9 A
i -

8 Q 7 w 9
2 -

- 4 F 215 3
_ 32
6 - 2 U 1 4
2 _
7 -g
- 2 V 6H 13 F 6

4 36 2 *3
Y -
2 -4 -2
-4
A 9
2 7
-2
5 Y
-~~~~~~~~~~~ in23
-5

Y 7
i 2 2
_

2
F%
3
4 19,2 14
6
H5 1 2 7 F0 aS
7
F6 6
H,2,2 I. I,%/2 3 H6 F7 ,2
Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb

to obtain a semiquantitative idea of the structure of the one to two electrons is quite remarkable. For n= 3
entire spectrum, without which the analysis and inter- (Nd iv and Pr iii) the number of levels has increased to
pretation of the more complicated cases would be vir- 107 with 5780 possible lines and for n = 5 (Sm iv, Pm iii)
tually hopeless. That some of these spectra are indeed we have 977 levels, but we have made no attempt to
very complicated can be seen from the following calculate the number of possible lines, and this is by no
remarks. means the most complicated case. Also there are many
If we consider only the four basic configurations other configurations besides the four basic ones con-
4ft, 4f"-1 6s, 4fn-I 6p, 4fn-1 5d, we have a total of 7 sidered here which contribute to the observable line
levels for n= 1 (Ce iv and La iii) with 8 allowed transi- strengths.
tions among them. For it= 2 (Pr iv and Ce iii) we have Even with the incomplete statistics just given it is
20 levels with 296 lines allowed by the selection rules of evident that the spectra of the rare earths in any stage
which some are perhaps so weak as not to be observable, of ionization are extremely rich in lines and of very
but the increase in the number of lines when we go from complex structure. Their analysis in each case requires
August 1961 SPECTRA OF IONIZED RARE EARTHS 827

voluminous experimental data and all the help which studied in order to enable us to obtain the modification
the theory is able to give. It is also clear that without of the levels by the crystal forces, the results obtained
automatic computing facilities the burden of handling for the crystal levels directly are very useful for the
the data would be intolerable. We use such facilities for unravelling of the free-ion spectra.
obtaining the wave numbers from the photographs of This is illustrated in Fig. 6 which gives the lower
the spectra and, once a wavelength table is made up, part of the 4fn level system obtained from the absorp-
for making the innumerable subtractions which are tion and fluorescence spectra of the trivalent rare-earth
necessary to establish the constant differences from the ions in crystalline salts (in most cases the anhydrous
observed line wave numbers, which are the foundation chlorides). With the exception of Pm3 +, for which no
for building an energy-level system with the help of the detailed data are as yet available, and Ho3+, for which
combination principle. one low-lying level is missing and the identification of
In finding constant-frequency differences with the some of the levels is uncertain, the levels are essentially
help of automatic computers or by any other systematic complete for the indicated energy range. The thickness
method, a number of accidental spurious coincidences of the lines represents the total splitting by the electric
occur which may be so numerous that they make the crystal field. The center of gravity is close to the position
task much more laborious; if they outnumber the of the free-ion levels and the knowledge of the approxi-
genuine constant differences they make the method mate position is of great help in the analysis of the
futile. The larger the possible error Av in the wave fourth spectra of the free ions.
number, the more numerous are the spurious coinci-
dences which may easily be seen to be proportional to ACKNOWLEDGMENTS
(Lv)2 . This emphasizes the great importance of accurate We thank Dr. J. Sugar, Mr. B. W. Bryant, and
wavelength measurements. Mr. W. R. Callahan for the use of material on the
While the free-ion spectra of the rare earths are being spectra of Pr iii, Yb III, Yb Iv, and Gd III.

JOURNAL OF THE OPTICAL SOCIETY OF AMERICA VOLUME 51, NUMBER 8 AUG.JST, 1961

Kr86 and Atomic-Beam-Emitted Hg'9 8 Wavelengths

R. L. BARGER
National Bureau of Standards, Boulder, Colorado
AND

K. G. KESSLER
National Bureau of Standards, Washington 25, D. C.
(Received January 10, 1961)

Vacuum wavelengths are given for the Hg 9 8 2537- and 3132-A lines and the Kr' 6 6013- and 5651 A lines,
referred to the Kr86 primary standard line 6057 A. The light sources were an Hg' 8 atomic-beam and a
Kr" hot-cathode lamp. A vacuum Fabry-Perot interferometer was employed for the measurements. The
wavelength of the Hg'98 3132 A line was also determined by using the Hg"'1 2537-A line as the reference
standard. The accuracy of measurement of the Hg'98 lines is about a factor of 5 higher than that of the
Kr6 lines.

INTRODUCTION to nonorthogonality between the atomic-beam axis and

AS part of the investigation into the problem of
redefining the standard of length in terms of a
wavelength of light, an emission atomic-beam light
source has been developed' which utilizes the isotope
Hg195 . It has been shown that the Hg 98 resonance line
2537 A emitted by this source is quite suitable for a
standard; the line is of great sharpness, with a half-
width of approximately 0.0016 cm-', the atoms in the
atomic beam source experience a very low level of
perturbation, and any systematic wavelength shifts due
R. L. Barger and K. G. Kessler, J. Opt. Soc. Am. 50, 651
(1960).
the direction of observation can easily be reduced to a
negligible quantity.2 Thus, this line emitted by the Hg' 9 '
atomic beam approaches the ideal line which would be
emitted by unperturbed atoms at rest with respect to
the observer.
The present investigation was undertaken in order to
calibrate the Hg19s resonance line in terms of the wave-
length of the 6057-A line of Kr". The Eleventh General
Conference on Weights and Measures, meeting in Paris
on October 14, 1960,3 adopted the value 1 650 763.73
2 R. L. Barger and K. G. Kessler, J. Opt. Soc. Am. 50, 352
(1960).
3Comite intern. poids et mesures, proces verbaux, 20 sr. (to
be published).