Vox. UMz 25, NUMsaR 2S
21 DZCaMazR 1970
0. Bely, J. Phys. 8:
Proc. Phys. Soc., London 1, 718
(1968); O. Bely and P. Faucher, Astron. Astrophys. 1,
37 (1969).
G. W. F. Drake, to he published.
C. Bchwartz, private communication.
M. Lipeles, R. Novick, and
Tolk, Phys. Rev.
Lett. 15, 690 (1965).
9C. Artura,
Tolk, and R. Novick, Astrophys.
¹
J.
J. 157, L181 (1969).
B. C. Elton, I . J. Palumbo,
¹
and H. R. Griem, Phys.
Rev. Lett. 20, 783 (1968).
A. H. Gabriel and C. Jordan, Nature 221, 947 (1969);
J.
%. M.
Neupert and M. Swartz, Astrophys.
160, L189
(1970); A. B. C. Walker, Jr. , H. R. Ruggge, Astro
Astrophys. 5, 4 (1970).
R. W. Schmieder and R. Marrus, Phys. Rev. Lett.
25 1245 (1970).
A. S. Pearl- Phys. Rev. Lett. 24, 703 (1970); R. S.
Van Dyke, Jr., C. E. Johnson, and H. A. Shugart,
Phys. Rev. Lett. 25, 1403 (1970).
R. Marrus and B. %. Schmieder, Phys. Lett. 32A,
431 (1970).
R. Marrus and B. %. Schmieder, preceding Letter
[Phys. Rev. Lett. ~25 1689 (1970)j.
Calculation of the Indirect Spin-Spin
Coupling Constant in HD Molecule»'
Many-Body Perturbation
and T. P. Das
of Utah, Salt Luke City,
C. M. Dutta, N. C. Dutta,
DePa&ment of Physics,
University
Utah 84122
(Received 15 June 1970)
The linked-cluster many-body perturbation approach has been applied to the study of
indirect nuclear spin-spin coupling constant JHD in HD molecule. The complete set of
states used were the bound and continuum states of H2+ molecular ion with the internuclear separation appropriate to H2 molecule. Our calculated value of JHD through the
Fermi con'tact interaction mechanism ls +42. 57 Hz in good agreement with the most lecent experimental value of +42.7 + 0.7 Hz.
Although the indirect spin-spin interaction between nuclei in molecules has been utilized extensively for qualitative understanding of electronic structures in molecules, its quantitative
calculation even for the simple molecule HD has
proved to be a formidable task, with the current
situation far from conclusive. '
The various
theoretical approaches utilized so far can be divided broadly into two categories. In the first
category is the conventional second-order perturbation as first proposed by Ramsey and Purcell.
The difficulty with such an approach is
that one requires a knowledge of the complete
set of ground and excited states of the molecule,
which is not usually available.
To obviate the knowledge of excited states,
variation-perturbation
procedures have been
used by a number of authors. ' In one class of
such calculations, a diagonal-type perturbation
procedure" was used, the second-order energy
due to the hyperfine field of one nucleus being
minimized t:o obtain the first-order perturbed
wave function of the molecule. This function was
then utilized to calculate the cross term in the
second-order energy involving the other nucleus
to obtain PHD. The difficulty with this procedure
was that the second-order nuclear self-coupling
'
"
'
'
energy is infinite in nonrelativistic theory, no
real minimum thus being attainable in a variational approach. A second class of variation perturbation calculation has attempted to extremize
the cross terms in the second-order energy proportional to JHD using variational functions which
describe the first-order perturbation due to both
nuclei.
The difficulty with this procedure is
that the cross term due to two per turbations has
by itself no minimum and one does in fact get oscillatory behavior as the number of parameters
''
increased.
In this paper, we have revived the perturbation
approach in a form that: meets the major difficulty, namely, a knowledge of a complete set of
states for the molecule. This is accomplished
by using the linked-cluster many-body perturbation theory (LCMBPT), where a neighboring
Hamiltonian X„ for which the complete set of
states can be obtained exactly, is used as the
starting point for a perturbation treatment of 4X
=X-X . In our work here, a multiple perturbation approach is used in conjunction with the
LCMBPT, using the sum of 6+ and the two hyperfine interaction Hamiltonians XH' and XD' associated with the two nuclei. Fol the speed of convergence of the perturbation approach, it is nec-
"
Vox. UMz 25, NUMszR 25
21 DmcaMszR 1970
essary that the basis states chosen (and hence g)
describe the behavior of the charge densities
near the nuclei reasonably well. In particular, a
one-center choice for X might be expected to be
rather inadequate for the present problem since
this would require the inclusion of very high angular-momentum
states to describe properly the
charge densities near the nuclei. With this consideration in view, the H, molecular-ion Hamiltonian was chosen for X . This has the dual merit of providing a basis set that is both exactly de-
'
rlvRble RQd has the deslx'RMe cusp behavior Rt.
the nuclei.
The procedul e of cRlculRtlon follows b10adly
the same lines as that employed in handling atomsie systems subject to external pexturbations.
Only a very brief description will be presented
here with emphasis on some of the points that
have special importance for the present problem.
The starting Hamiltonian, corxesponding to two
noninteracting electrons in the H, molecule
framework, is given (in atomic units) by
"
'
~ f,
Xo= Z
i=
x
,
(
1
1
rHi
roi
leading to AX =1/r». The net perturbation Ham1ltonlRQ K' composed of 4+ Rnd the hyperflne 1Qteraction terms is then given by
X' = (1/ri2)
where the
+XH +XD,
X„' (A = H or
D)
are given by
representing the Bohx magneton Rnd
nuclear magnetic moment of nucleus A. and I„
and S; the nuclear and electron spin operators.
Following the usual linked-cluster perturbation
is givapproach, the total energy corxeetion
p, & Rnd p, ~
~
en by
AE =
.
Z &4. (-, 0)X.'(O)
I
.'(0, ")I+.&,
(4)
the suffix I. and other quantities in Eq. (4) having
their usual meanings in linked-cluster perturbaThe indirect nuclear spin-spin intion theory.
teraction Hamiltonian has the form
"
KHD
=h JHDIH- ID.
The spin-spin coupling constant PHD can be evaluated by equating the expectation values of both
sides of Eq. (5) over the nuclear spin states with
magnetic quantum numbers MH =IH and MD =ID.
The expectation value ~HO of the left-hand side
of Eq. (5) corresponds to the energy derived from
Eq. (4), keeping one order in XH' and one order
in XD' and all possible orders in 1/r». In diagrammatic representation, the corresponding
diagrams must contain the X„' and 3*D' vertices
once, whereas the 1/r» vertex can occur any
number of times. 81Dce the spin Hamiltonlan 18
isotropic, it i.s sufficient to woxk with only the
z-component term in Eq. (5).
For the one-electron basis set for the diagrammatic evaluation of
we have utilized the exact bound and continuum states of X which correspond to H2+ molecular-ion wave functions for the
internuclear distance R =1.4 a.u. These are expressed in the form
~»
~;(&, u, V) =A;(&)M;(V)e"',
where A,-(X) and M,-(p) are functions of the elliptic coordlnatesq A. = (r H+ r.n)/R~ p, = (riH
D)/
R, Rnd r,-H and r&D are the distances of the jth
electron from H and D nuclei. For the bound
states, we have to determine both the energy
eigenvalues as well as the eigenfunctions described by Eq. (6). This requires the solution
of the appropx iate second-ordex differential equations for A;(X) and M;(ii) which are coupled by the
energy and a separation constant A. ;. The eigenvalues and eigenfunctions for some of the lower
bound states axe already available in the literature.
For the higher bound states, only eigenvalues are available" and we had to solve the
necessary continued-fraction equations to obtain
A; and M;(p, ). The functions A;(i) were obtained
by solving the corresponding differential equaFox' the continuum statesq
tion numerically.
the eigenvalues are, of course, e; =k'/2. Again
A; and M;(p) are obtained by solving the requisite
continued-fraction equations and A,.(a) through
numerical integration of the coxxesponding differential equation. '
The diagrams involved in the calculation of
~HD Rx'e similar 1Q form to tho8e oDe eDcountex'8
in the perturbation of atomic systems in an external field.
The external fields in the present
molecule are the hyperfine fields of nuclei at the
sites of the electrons. For the hyperfine operators K„' and KD' we have utilized wiggly lines
terminating with dots in the diagrams. The lowest-order diagram for the present calculation is
of second order involving one order each in XH'
and XD'. The higher orders involve additional
vertices associated with 1/r». Figure 1 shows
the second-oxder and Rll of the third-order diagrams. The typical fourth-order diagrams involving two hyperfine vertices and two 1/r» ver~
~
"
"
"
~
-r
VOLUME
PHYSICAL REVIEW LETTERS
25, NUMBER 25
i
21 DECEMBER 1970
F
($)
FIG. 1. Second- and third-order diagrams for the
spin-spin interaction in HD molecule. The wiggly line
denotes the Fermi contact operators KH' and RD'
.
tices are
shown in Fig. 2. We have included all
the distinct diagrams that are obtained from
those in Fig. 2 by taking account of time reordering and permutations of the vertices among themselves and between hole and particle lines. The
most time-consuming aspect of the evaluation of
diagrams was the calculation of matrix elements
associated with the vertices. The hyperfine vertices only require the density of the wave funcHowtions at the nuclei and are straightforward.
ever, the matrix elements of 1/r», particularly
those involving continuum states, require special
attention, since elliptic coordinates are involved.
We have utilized the procedure developed by RMenberg" for variational molecular bound-state
calculations. Our computer program for this
purpose was checked by comparing our values
for a few two-center two-electron integrals using
Slater orbitals with currently available tables
for such integrals.
It should be pointed out that
because of the use of elliptic coordinates, the integration over the continuum states involves a
somewhat different multiplying factor than in
atomic work,
"
"
Q, - (2/vR)'J
dk.
The second-order diagram (1a) represents the
contribution to JHD from two noninteracting electrons in the ground state of a H, molecular ion
with internuclear separation of R =1.4 a.u. The
third-order diagrams (1b)-(1g) represent the
contribution to JHD from the first-order effect of
1/r» on diagram (1a). Of these, diagrams (lb)(1e) represent the role of the passive HartreeFock interaction between the electrons which
converts the electronic wave functions from
'
FIG. 2. Some typical fourth-order diagrams for the
spin-spin interaction in HD molecule.
'
those of H, molecular ion to the H, molecule.
Diagrams (1f) and (1g) represent the influence of
two-particle self-consistency effects in the perturbed state of the system when the nuclear hyperfine interaction is switched on.
The contributions to JHD from various diagrams are listed in Table I. For the second-order diagram (la) we have separately listed the
contributions from gerade and ungerade particle
states. A rather substantial cancelation is seen
to occur between the two types of excitations in
this order. For the diagrams (1b)-(lg), the contributions are listed after the gerade and ungerade cancelations have been carried out. , wherever they occur. For the sake of brevity, only the
net contribution from all the fourth-order diagrams is listed. The third-order result is seen
to be about one-half of the second order. This
somewhat slow decrease up to this order is a result of the rather severe cance1ation between
gerade and ungerade excitations in second order.
However, on comparing individual diagrams,
even the diagram with the largest contribution in
third order is found to be a factor of 4 smaller
than the smaller (gerade) of the second-order
diagrams. This situation is reminiscent of the
hyperfine interaction problems in atomic oxywhere the lowgen, nitrogen, and phosphorus,
"
"
1697
PHYSICAL REVIEW LETTERS
VOr. UME 25, NUMBER 25
Table
I.
Contributions
21 DECEMBER 1970
from various diagrams to JHD.
Contributions
Order of perturbation
Diagrams
(Hz}
Second order
(la) gerade
(la) ungerade
-385.65
458.67
73.02
Subtota1
Third order
-74.76
(lb)
-10.16
(1c}
(ld)
-47.60
(le)
47.41
24.90
24.90
-35.31
(1f}
(1g}
Fourth order
Subtotal
Subtota1
Grand total
Experiment
est order core-polarization diagrams have rather substantia1 eance1ations which make the higher orders of crucial importance.
In fourth order, there is strong cancelation
among the various diagrams due to the gerade
and ungerade symmetry and spins. Additionally,
there is also substantial cance1ation among different types of diagrams leading to the net relatively small contribution shown in Table I, namely, about 6.7% of the second-order result. We
have carried out rough estimation of important
fifth-order diagrams and find that this strong
cancelation persists and that the net contribution
in this order is expected to be no more than a
cycle. Our result compares very favorably with
the experimental value of +42.7+P.'7 Hz.
The
experimental value does include small contributions from second-order effects involving dipolar
and orbital hyperfine interactions.
Several estimates" of these contributions have been made
in the literature which are consistently rather
"
small.
It is our feeling" that these mechanisms
more than the quoted range of
experimental error. We would like to point out
that the present procedure can also be used to
evaluate these additional contributions to JHD.
It is worth mentioning that in contrast to
LCMBPT calculations on atomic systems, laddering procedures to sum over selected classes
of diagrams are not suitable in the present approach. This is because the diagrams (ld) and
(le) which are the progenies of the usual holehole and hole-particle ladders are comparable in
importance with the other diagrams in third order hand it is not appropriate to selectively apply
laddering procedure to any one class of diagrams.
do not contribute
4.86
42.57
42.7 +0.7
Instead, the correct procedure for the present
problem is to sum all diagrams for each order
and such a sum is seen from our work to show
satisfactory convergence. It should be pointed
out that in ea, rlier work, Matcha. and Brown" and
Goodisman" had used a zero-order H, + basis
set, as in the present work, and utilized variation-perturbation procedures to calculate the energies of the ground and excited states of H, molecule with rather satisfactory results. The convergence observed here using the same starting
Hamiltonian for a complex property such as the
spin-spin coupling constant demonstrates the general validity of the ionic basis set as a starting
point for perturbation calculations on diatomic
molecules.
It has been possible to carry out this calculation of the indirect spin-spin coupling constant
due first to the increased sophistication attained
in recent years in applying LCMBPT procedures
to atomic systems and secondly to the advance in
computing techniques which makes it practicable
to work with one-electron continuum states of
molecular systems. It is expected that the present procedure would be easily applicable to other
hyperfine and magnetic properties of hydrogen
molecule involving double perturbation by external sources, such as the rotational moment and
spin- rotation constant.
We are grateful to Professor D. R. Bates for
supplying us a table of eigenvalues prior to publication. Thanks a.re also due to Professor R. L.
Mateha for useful discussions.
*Work supported by National Science Foundation.
VQLUME
PHYSICAL REVIEW LETTERS
25, NUMBER 25
N. F. Ramsey and E. M. Po.rcell, Phys. Rev. 85,
143 (19M).
F. Ramsey, Phys. Rev. 91, 303 (1953).
3E. A. G. Armour,
Chem. Phys. ~49 5445 (1968).
E. Ishiguro, Phys. Rev. 111, 203 (1968).
T. P. Das and R. Bersohn, Phys. Rev. 115, 897
(1959) .
M.
Stephen, Proc. Roy. Soc., Ser. A 243, 274
¹
J.
J.
{1957}.
D. E. O'Reilly,
J. Chem. Phys. 36, 274 (1962).
J. Schaefer and R. Yaris, J. Chem. Phys. 46, 948
{1967).
S. Ray, M. Karplus, and T. P. Das, unpublished.
J. Goldstone, Proc. Boy. Soc., Ser. A 239, 267
{1967).
C. M. Dutta,
A 1, 661 (1970).
¹
21 DECEMSER 1970
demic, New York, 1968), Vol. 1V, p. 13.
' D. R. Bates, U. Opik, and G. Poots, Proc. Phys.
Soc., London, Sect. A 66, 113 {1963).
5K. Riidenberg,
Chem. Phys. 19, 1459 (1951).
8M. Kotani, A. Amemiya, E. Ishiguro, and T. Kimura„
Tables of Molecular Integrals {Maruzen Co. Ltd. ,
Tokyo, 1966).
H. P. Kelly, Phys. Rev. 131, 684 (1963).
H. P. Kelly, Phys. Rev. 173, 142 (1968}, aud 180, 66
(1969) .
N. C. Dutta, C. Matsubara, B. T. Pu, and T. P. Das,
Phys. Rev. 177, 33 (1968), and Phys. Rev. Lett. 21,
J.
1139 (1968).
T. F. Wimmett. Phys. Rev. 91, A476 (1963). Sign
determined hy I. Ozier, P. N. Yi, A. Khosha, and
F.
Ramsey, Bull. Amer. Phys. Soc. 12, 132 (1967).
S. D. Mahanti and T. P. Das, Phys. Bev. 170, 426
(1968) .
22R. I, . Matcha and W. B. Brown,
Chem. Phys. 48,
74 (1968).
J'. Goodisman, J'. Chem. Phys. 48, 2981 (1968).
¹
C. Dutta,
and
T. P. Das,
Phys. Rev.
D. R. Bates, K. Ledsham, and A. L. Stewart, Phil.
Trans. Roy. . Soc. London, Ser. A 246, 216 {1963).
D. R. Bates and B. H. G. Beid, in A.defences in Atomic and Molecular Physics, edited by D. R. Bates (Aca-
J.
Formation and Strncture of Electrostatic Collisionless Shocks*
D. %. Forslund and C. R. Shonk
Los Alamos Scientific Laboratory, University of California, Los Alamos,
(Received 4 November 1970)
Neav
Mexico 87544
A one-dimensional,
two-species numerical-simulation
code has been used to study the
formation and structure of coQisionless electrostatic shocks formed by two colliding
plasmas and by a plasma striking a perfectly reflecting piston. Collisionless shocks are
= vD/e
formed in hydrogen up to a maximum piston velocity of M —
s 3.5. Shocks with velocities up to M 4 have been produced which have a predominantly laminar structure
accompanied by strong collisionless dissipation.
-
Theoretical models for electrostatic shocks in
the absence of a magnetic field have been discussed by a number of authors. ' ' Lom-Machnumber [M=vo/cs&2, where cs=(T, /m;)'~']
shocks have been experimentally produced by
Alikhanov, Belon, and Sagdeev' and Taylor, Baker, and Ikezi, ' and numerically simulated by
Mason' and Sakanaka, Chu, and Marshall. ' In
this Letter me discuss formation of stable, highMach-number electrostatic shocks by means of
the particle-in-cell simulation technique treating both the ions and electrons exactly with the
mass ratio of hydrogen (1836). Since the electrons are treated exactly instead of isothermally,
the critical Mach number of
6 does not apply. The shocks described here are mell approximated by a rapid change from one equilibrium
state to another with little or no fluctuations in
the shock front (although perhaps having large
fluctuations behind the front), and thus correspond most closely to the "laminar"' ~ rather
I-1,
' model.
Initial conditions. —The shocks discussed here
are created in tmo mays. The fiI st method consists of a plasma of ions and electrons moving
to the right with Mach number M, as shown in
Fig. 1(a), with the plasma initially (T =0) just in
contact with the right-hand, perfectly reflecting
mall. The plasma is sustained at the left by continually injecting a Maxwellian plasma. The second method ls similar but with the above plasma
filling only the left half of the box and an equaldensity plasma with Mach numbeI of opposite
sign occupying the right half. This configuration
clearly should generate tmo shocks which dlffel'
only statistically.
For M &1 a shock will readily
form on the basis of the ion-ion instability'0 in
one dimension. Shocks formed with M &1 do so
by means of the nonlinear process described below.
Theory of formation and structure. —In the tmo
lllltlal conditions used the method of shock forthan "turbulent"