HARTREE —FOCK METHOD

beta-decay theory are not fu&&&~ed in this decay process. Be' used in these experiments. They are also grateful
Certainly all other experimental results support the for a very valuable discussion of the experiment with
former conclusion. Dr. Raymond Davis of the 8rookhaven National
The authors wish to express their appreciation to Laboratory, to a great deal of excellent advice on
the crews of the University of Illinois and the Univer- surface preparation given by Prof. C. W. Sherwin of
sity of Chicago cyclotrons, who provided most of the the Department of Physics, University of Illinois.

PH YSICAL REVIEW VOLUME 81, NUMBER 3 FEBRUARY 1 ~ 19S1

A S™p»&cationof the Hartree-Fock Method

J. C. Sz,ATER
Massachusetts Inset'tgte of Technology, ~ Cambrntge, Massachgsetts
(Received September 28, 1950)

It is shown that the Hartree-Fock equations can be regarded as ordinary Schrodinger equations for the
motion ot electrons, each electron moving in a slightly diferent potential field, which is computed by
electrostatics from all the charges of the system, positive and negative, corrected by the removal of an
exchange charge, equal in magnitude to one electron, surrounding the electron whose motion is being in-
vestigated. By forming a weighted mean of the exchange charges, weighted and averaged over the various
electronic wave functions at a given point of space, we set up an average potential Geld in which we can
consider all of the electrons to move, thus leading to a great simplification of the Hartree-Fock method, and
bringing it into agreement with the usual band picture of solids, in which all electrons are assumed to move
in the same Geld. We can further replace the average exchange charge by the corresponding value which we
should have in a free-electron gas whose local density is equal to the density of actual charge at the position
in question; this results in a very simple expression for the average potential field, which still behaves quali-
tatively like that of the Hartree-Fock method. This simplified Geld is being applied to problems in atomic
structure, with satisfactory results, and is adapted as well to problems of molecules and solids.

I. INTRODUCTION u2(x), u„(x), in such a way as to make the energy

HE Hartree-Fock equations' furnish the best set tui*(xi) u (x)Bui(xi) . .u (x)dxi. . dx an ex-
of one-electron wave functions for use in a self- treme, where H is the energy operator of a problem in-
consistent approximation to the problem of the motion volving n electrons in the field of certain nuclei, and
of electrons in the field of atomic nuclei. However, they where the functions u, are required to be normalized.
are so complicated to use that they have not been em- Similarly the Hartree-Fock equations, as modified by
ployed except in relatively simple cases. It is the pur- Dirac, ' are obtained by varying the s so as to make I
pose of the present paper to examine their meaning the energy
suKciently closely so that we can see physically how to
1 t' u, *(x,) u, *(x„) u, (x,) u, (x )
set up a simplification, which still preserves their main ~ s ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ O
g ~ . . 61 dX„
features. This simplified method yields a single poten- n!" u„'(x,) u„*(x„) u. (x'). u„(x.)
tial in which we can assume that the electrons move,
and we shaB show the properties of this field for prob- an extreme, where in this latter expression the u's are
lems not only of single atoms but of molecules and assumed to be functions depending on coordinates
solids, showing that it leads to a simplified self- and spin, and where the integrations over the dx's are
consistent method for handling atomic wave functions, interpreted to include summing over the spins. The
easy enough to apply so that we can look forward to Hartree-Fock equations can then be written in the form
using it even for heavy atoms.
II. THE HARTREE-POCK
THEIR MEANING
EQUATIONS AND Hyu;(xy)+ g
k=1
us*(x2)ua(x2)(e/kreoty2)dx2 u, (xg)'

It is well known that the Hartree equations are ob- n

tained by varying one-electron wave functions u'(x), "u.*(x2)u;(*2)(~/4 &o.u)dx2 uk(xl)
~
a-1
The work described in this paper was supported in part by
the Signal Corps, the Air Materiel Command, and the ONR,
through the Research Laboratory of Electronics of M.I.T.
=E,u;(xg). (l)
' J. C. Slater, Phys. Rev. 35, 210 (1930); V. Fock, Z. Physik
61, 126 (1930};L. Brillouin, Les GhemPs Self-Goesistnds de Har- Here H1 is the kinetic energy operator for the electron
tree et de I;oak, Actuahths Scientifiques et Industrielles No. 159 of coordinate x1, plus its potential energy in the field of
(Hermann et Cie. , 1934); D. R. Hartree and %'. Hartree, Proc.
Roy. Soc. A156, 9 (1935); and many other references. ~ P. A. M. Dirac, Proc. Cambridge Phil. Soc. 26, 376 (1930).
386 J. C. SLATE R
all nuclei; e'/krcor~m is the Coulomb potential energy falling off as vre go avray from that point. We can get
of interaction betvreen electrons 1 and 2, expressed a rough idea of the distance in vrhich it has fallen to a
in mks units; to get the corresponding formula in small value by replacing it by a constant density p+
Gaussian units vre omit the factor 4n-fo, and to get it in inside a sphere of radius ro, zero outside the sphere.
atomic units we replace e'/4s ea by 2. The I s as before We have ~~sro' p+ = e, or
l l

are assumed to depend on spin as well as coordinates,

and the integrations over dx~ include summation over
ro= (3s/4 I p+ I)'. (2)
spin, so that the exchange terms, the last ones on the The situation is then much as if the corrected charge
left side of Eq. (I), automatically vanish unless the density equaled the actual total electronic charge
functions u, and ul, correspond to spins in the same density outside this sphere, but was only p within
direction. the sphere; there is a sort of hole, sometimes called the
The Hartree-Fock equations in the form given pre- Fermi' or exchange hole, surrounding the electron in
sent an appearance which seems to differ from the question, consisting of a deficiency of charge of the
ordinary one-electron type of Schrodinger equation, same spin as the electron in question. Actually, of
and for this reason it is ordinarily thought that they course, this exchange hole does not have a sharp bound-
cannot be given a simple physical interpretation. This ary, but the charge density of the same spin as the elec-
assumption arises partly from the paper of Dirac', in tron in question gradually builds up as vre go away
which they are interpreted in a rather involved way. from this electron. Similar statements hold for the
The second term on the left of (I) is simple: it is clearly field acting on an electron of minus spin.
the Coulomb potential energy, acting on the electron The exchange hole clearly is different for wave func-
at position x~, of all the electronic charge, including that tions of the two spins, provided p+ and p are different;
of the ith wave function vrhose vrave equation we are examination proves further that it is different for each
vrriting. The last term on the left, the exchange term, different wave function I;.
It is this di6'erence which
however, is peculiar, in that is is multiplied by uz(x&) leads to the complicated form of the Hartree-Fock
rather than by I;(x&). It must somehow correct for the equation; and the simpli6cation which we shall intro-
fact that the electron does not act on itself, vrhich it duce in a later section is that of using sort of an averaged
would be doing if this term vrere omitted. In the Hartree, exchange hole for all the electrons. The difference be-
as opposed to the Hartree-Fock, equations, this is tvreen the exchange charge for two wave functions u;
obvious. There the last term differs from that in the corresponding to the same spin is not great, however.
Hartree-Fock equations only in that all terms in the W'e have already seen that the radius ro vrhich we ob-
summation are omitted except the ith; the exchange tain by assuming a hole of constant density depends
term in that case then merely cancels the term in 0=i only on p+ (for a plus spin), and hence is the same for
from the Coulomb interaction found in the second term. all I s of that spin. Thus the exchange holes for dif-
The main point of our discussion is to show that an ferent u s of the same spin vrill only show small dif-
equally simple interpretation of this term can be given ferences. W'e shall later examine these differences for
in the Hartree-Fock equations. the case of a free electron gas, and show that they are
Let us 6rst state in words vrhat the interpretation really not large. It is this small dependence on vrhich I;
proves to be; then vre can more easily describe the way wi11 make it reasonable to use an averaged exchange
in vrhich the equations lead to it. We can subdivide the charge in the simplified method which vre shall suggest
total density of all electrons into two parts, p+ from later.
those with plus spins, p from those with minus spins; To agree vrith the qualitative description which we
the two together add to the quantity — eg(k= l. e) have just given, vre then expect the exchange charge
Xgq*(x)mq(x), where e is the magnitude of the elec- density at point x2, producing a 6eld acting on the
tronic charge. Then vre can show that the Hartree- electron at xq whose wave function N, (x~) we are deter-
Fock Kq. (I) for a wave function e; which happens to mining by the Hartree-Fock Kq. (I), to integrate over
correspond to an electron with a plus spin is an ordinary dx2 to — e (a single electronic charge), and to be equal
Schrodinger equation for an electron moving in a per- when x2 approaches x& to the quantity
fectly conventional potential 6eld. This 6eld is calcu-
lated by electrostatics from all the nuclei, and from a —e Ng (xg)Qk(xl) ~ (3)
distribution of electronic charge consisting of the whole k=|
spin k =spin s
of p, but of p+, corrected by removing from the im-
We shall novr show that this is the case.
mediate vicinity of the electron, whose vrave function
we are 6nding, a correction or exchange charge density 3 E. Wigner and F. Seitz,
Phys. Rev. 43, 804 (1933); ibid. 46,
whose total amount is just enough to equal a single 509 (1934). The discussion of Wigner and Seitz was one of the
6rst to show a proper understanding of the main points taken up
electronic charge. That is, this corrected charge dis- in the present paper, which must be understood to represent a
tribution equals the charge of e —1 electrons, as it generalization and extension of previously suggested ideas, rather
should. The exchange charge density equals just p+ at than an entirely new approach. See also L. Brillouin, J. de Phys.
et le Radium, 5, 413 (1934) for a discussion somewhat similar to
the position of the electron in question, gradually the present one.
 HARTREE —FOCK METHOD
To show it; we rewrite (1) in the equivalent form' an electronic charge on each of the X atoms of the
crystal. In the Hartree scheme, the potential acting on
Hgu;(xg)+ P up*(xg)ug(x2)(e'/4s eorgg)dxg u;(xg)
the electron in the wave function I;
is that of all elec-
k 1 aj trons, minus this charge corresponding to 1/E of an
electron on each atom. This correction charge is so
spread out that its effect on the potential field is com-
u' (xl)uk (x2)uk(xl)u'(x2)(e'/+rearm)dxm pletely negligible, and each electron acts as if it were in
the 6eld of all electrons, thus 6nding itself in the 6eld
u;(xg)
u;*(xg) u, (xg) of a neutral atom when near any of the nuclei of the
metal. On the other hand, with the Hartree-Fock equa-
tions, the exchange charge is located near the position
x& of the electron in question, moving around with it,
The exchange term now appears as the product of a so that when this electron is on a given atom, the ex-
function of x~, times the function u, (x~); thus it has change charge is removed largely from that atom,
the standard form of a potential energy term in a one- leaving it in the form of a positive ion, which, as our
electron Schrodinger equation. This exchange potential physical. intuition tells us, is the correct situation.
energy is the potential energy, at the position of the
6rst electron, of the exchange charge density, GI. AVERAGED EXCHANGE CHARGE
We have seen that the exchange charges for diBerent
u, *(xg)up*(x2) up(xg) u, (x2)
wave functions u; corresponding to the same spin are
u;*(xg)u, (xg) not very diferent from each other, since in every case
they reduce to the same value when x2 ——x&, and in-
located at the position x2 of the second electron. We tegrate to the same value over all space. Furthermore,
note as we expect that the exchange charge density de- in a system containing equal or approximately equal
pends on the position of the 6rst electron, as well as numbers of electrons with both spins, p+ and p will
the second, and also on the quantum state i in which be at least approximately the same, so that exchange
this first electron is located. We note, furthermore, charges for diGerent Is even of opposite spins will be
that the total charge is that of a single electron. To nearly the same. It then seems clear that we shall make
show this, we integrate the exchange charge density (5) no very great error if we use a weighted mean of the ex-
over dx2, and find at once, on account of the orthogo- change charge density, weighting over i, for each value
nality of the u s (which follows from the Hartree-Fock of x~. The result of this will be that we shall have a
equations) that the integral over all space is — e. Fur- single potential 6eld to use for the Schrodinger equation
thermore, as x2 approaches x~, we see at once that the for each e;, simplifying greatly the application of the
exchange charge density approaches the correct value Hartree-Fock method. Let us 6rst set up this average
(3), where the restriction that the spins of i and k must exchange charge and the consequent averaged exchange
be equal arises from (1), where an exchange term
potential, then give some discussion of their properties
uz (x2)u;(x&) is automatically zero unless this condition
is satisfied. Thus we have shown that the exchange and uses.
charge density (5) satisfies all the conditions necessary The probability that an electron at x& should be in
to justify our qualitative discussion of its behavior. In the state i is evidently u; (x~) u;(x~)/[P, ~,*(x~)u;(x~) j.
a later section, where we work out detailed values for We can then use this quantity as a weighting factor to
the free-electron case, we can examine its properties weight the exchange charge density (5). When we do
more in detail. this, we find as the average exchange charge density
The great di8erence between the Hartree and the the quantity'
Hartree-Fock methods is the fact that in the Hartree-
Fock method the exchange hole or correction charge n n

appropriate for an electron at x~ moves around to follow Q u, *(xg)up*(xm) ug(xg) u, (xm)

that electron; in the Hartree method it does not, the —ej=i %=i (6)
correction charge depending only on the index i of the
wave function e;. If our problem is a single atom, this u, *(xg)u, (xg)
j=l
is not very important, but in a crystal, for instance a
metal, the difference is profound. Thus consider a
Using this average exchange charge density, we come
periodic lattice, in which the one-electron functions I; to the following Schrodinger equations for the e; s, as
are modulated plane waves, corresponding to 1/X of
I J. C. Slater, Rev. Mod. Phys. 6, 2Q9 (1934), particularly
' J. C. Slater and H. M. Krutter, Phys. Rev. 47, 559 (1935); p. 267, where this same expression is used for similar purposes,
particularly p. 564, where this same method is used in discussing but without pointing out that it is the weighted mean of the ex-
the Thomas-Fermi method. change charge density found in the Hartree-Fock equations.
 J. C. SLATE R
substitutes for the Hartree-Fock equations:

+1Ni(x1)+ Q '
22k (x'2)lk(x2)(+/4Ir00r12)dx2

n n
tgj (xl)sk (x2)20k(xl)Qj(x2)(8 /42I 009'12)dx2
j~l k~1 ~
u, (x1) = E,22;(x1). (7)
e
Q I;~(x1)u;(x1)

The wave functions e;, and energy values E;, as deter- quantum numbers, and then carry out perturbations.
mined from these equations, will not be quite so ac- If we take the Hartree-Fock scheme literally, we shall
curate as those determined from the Hartree-Fock use diBerent potentials for finding the I
s of each of
equations; but they will at least be much better than these various unperturbed functions. It is highly de-
those found from the Hartree equations, particularly sirable in such cases, in the interests of simplicity, to
for the case of the crystal, and they have the great modify the method so as to use the same potential
advantage that they are all solutions of the same po- function for the calculation of each wave function. This
tential problem. This automatically brings one good may involve even more averaging than is contemplated
feature, which the solutions possess in common with in setting up Eqs. (6) and (7). As one illustration,
solutions of the Hartree-Fock equations, but which Hartree's use of a spherical potential for discussing
solutions of the Hartree equations do not have: the atomic structure is an example of this procedure; this
functions u, are all orthogonal to each other. involves averaging over all orientations of the various
There is one aspect of Eqs. (7) which is very im- orbital angular momenta of electrons which are not in
portant. In the 1ast few years there has been a great closed shells. Whether we are using the Hartree scheme
development of the energy-band theory of semicon- or the present simpli6cation of the Hartree-Pock
ductors. This is al1. based on the hypothesis that we can scheme, such averaging over orientations seems cer-
buiM up a model of a solid in which each electron moves tainly desirable. Again, in studying ferromagnetism,
independently in a potential 6eld which is made up the potentials to use, according to the scheme of the
from the nuclei, and all other electrons except the one present paper, will depend on the net magnetization,
in question. The electric 6eld derived from this poten- or on the number of electrons of each spin. It is much
tial is sometimes called the motive field acting on the simpler to handle such a problem, however, by using a
electron. Each wave function corresponds to a definite single potential function, and that will usually be
energy level, and the Fermi statistics are applied to the chosen to be that representing the unmagnetized state,
distribution of the electrons in these levels. The soundest with equal. numbers of plus and minus spins.
way to set up this potential acting on each electron is In all these cases which we have just been discussing,
by the Hartree-Fock method, but we see by our present we use one-electron wave functions which are slightly
discussion that this implies a diGerent potential energy less accurate than those found by the Hartree-Fock
or motive for each electron, or each I;.
If we wish to scheme. When we apply perturbation methods, we
have a single motive 6eld appropriate for all electrons, must remember this, computing the matrix components
the best thing we can do is to use the weighted mean of the exact energy operator with respect to these
suggested in the present section. Thus Eqs. (7) may somewhat incorrect wave functions, remembering the
well be taken to be the basis of the ordinary form of the wave equation (for instance (7)) which they actually
energy-band theory of solids. satisfy. Nondiagonal matrix components of energy
In many problems, we are interested in cases of de- between these somewhat inaccurate wave functions
generacy, not merely in evaluating the wave function will be somewhat larger than those between exact
of a single nondegenerate stationary state. Thus we Hartree-Fock functions. Nevertheless they wil1. still
may be solving a problem of multiplet structure in an not be very large, for the wave functions are still quite
atom or molecule, or discussing ferromagnetism in a accurate; the slight decrease in exactness is much more
solid. In such a case we start with a number of de- than made up by the simplicity of the method. The
generate or approximately degenerate energy levels, energy values computed by averaging the exact energy
corresponding to diBerent orientations of orbit or spin, operator over the wave function will be very nearly the
or in some cases (as in the hybridization of atomic same as for Hartree-Fock functions, on account of the
orbitals) corresponding to different total or azimuthal theorem that the mean value of energy over an in-
 HARTREE —FOCK METHOD
correct wave function has errors only of the second order
of small quantities.
IV. THE EXCHANGE CHARGE FOR THE 05—
FREE-ELECTRON CASE
The calculations of exchange charge and exchange
potential which we have been describing in general
2.0
language can be carried out exactly for the case of a r/rO

free-electron gas, as is well known. In this section we FIG. 2. Exchange charge density (divided by p/2) plotted as a
shall give the results, as an illustration of the general function of r/ro, from Eq. {12),where ro is given by Eq. (9).
case. Then we shall point out in the next section that
by using a free-electron approximation we can get an with uniform charge density p+~ =me/2V, the poten-
~

exchange potential much simpler than that of Eq. (7), tial energy of an electronic charge at the center of the
which still is accurate enough for many purposes. sphere would be s3(e'/4NEorp), while the value from Eq.
Let us have a free-electron gas with n electrons in (10) is 1.54(e'/47r@ro) at the bottom of the Fermi
the volume V, half of them of each spin; the volume is band, half this value at the top. Thus this simple model
assumed to be 6lled with a uniform distribution of of an exchange hole of constant charge density gives a
positive charge, just enough to make it electrically qualitatively correct value for the exchange potential,
neutral. The electrons are assumed to obey the Fermi and rather accurate quantitative value; and the ex-
statistics. Then by elementary methods we 6nd that treme difference between top and bottom of the band
they occupy energy levels with uniform density in corresponds only to a factor of 2 in the exchange po-
momentum space, out to a level whose energy is tential.
P'/2m=(h'/2m)(3e/8 Vs)&, corresponding to a maxi- If now we average over-aB wave functions, we 6nd
mum momentum P = h(3n/Ss. V) &. The de Broglie that the properly weighted average of F(q) is 3~. Thus
wavelength the exchange potentia1. energy of the averaged exchange
d = h/I'= (Ss.V/3n) ~ charge' is (x3)(6/s)&(e'/4s coro). This can also be found
from the averaged exchange charge density. This
F ('g)
l.0
charge density is'
FIG. 1. J (g} as function of p 3 sin(r/d) —(r/d)cos(r/d) '
g {from Eq. (11}}. (12)
2 (r/d)'
2.0
where p is the total charge density of electrons, d is
associated with this maximum momentum is clearly given by Eqs. (8) and (9), and r is the distance from
related to the radius ro of the exchange hole, which we point xi, where the electron whose wave function we
introduced in Eq. (2). When we notice that p+ which ~ ~
are computing is located, to x2, where we are finding
appeared there equals ne/2V, we see that the exchange charge density. This function (12) is
shown in Fig. 2, plotted as a function of r/ro, and we
(9) see that it does in fact represent a density which equals
We can now state some of the principal results of p/2 when r=o, and falls to small values at approxi-
the application of this model to the exchange charge mately r=ro. The potential energy of an electron at
density and exchange energy. The exchange potential the center of this averaged exchange charge distribution
energy can be conveniently stated in terms of the ratio is just the value (34)(6/s)&(e'/4seoro) previously given.
g = p/I' of the magnitude of the momentum of the elec-
tron to the maximum momentum corresponding to the V. USE OF THE FREE-ELECTRON APPROXIMATION
FOR THE EXCHANGE POTENTIAL
top of the Fermi distribution. It is'
From the argument of Sec. III, it is clear that the
exchange potential energy = (e'/4s eo) (4E/h) F (g)
exchange charge density (6), and the corresponding
= (6/s) &(e'/4s 60rp) F(g) (10) potential appearing in (7), must depend on the density
where of electronic charge, but not greatly on anything else.
Thus in no case will we expect it to be very di8erent
from what we should have in a free-electron gas of the
same charge density. We may then make a further
The function F(p) is shown in Fig. 1. It goes from unity approximation and simplification, beyond that of Sec.
when g=o, for an electron of zero energy, to —, ' when III; we may approximate the averaged exchange po-
q=1, at the top of the Fermi distribution. We see that tential by what we should have ih a free-electron gas
this exchange potential energy is of the form which we e F. Bloch, Z. Physik 57, 545 {1929}
gave the 6rst derivation of
should expect. If we had a sphere of radius ro, 61led this value.
 J. C. SLATE R
of the same density, as given in Sec. IV.' Thus, corn- structure. Let the electrostatic potential of the nucleus,
bining (10) and (2), we have and of all electrons, at distance r from the nucleus of a
spherical atom, be Z~(r) e/kroor. Then the charge
exchange potential energy density is given by Poisson s equation as p = — ooV'(Z~e/
= — (6/s ) &(e'/kr ooro)
a4
4xoor). When we express the Laplacian in spherical co-
= —3(e'/4oroo)(3n/8xV) (13) ordinates, this gives at once p= —
&,
(e/4s)(1/r)d'Z~/dr'.
where we are now to interpret n/V as the local density This is the quantity which is expressed as — eP(k)
of electrons, a function of position. If we recall that No*(x)li(x). Thus the exchange potential energy be-
P
this is (k)N&*(x)ez(x), we finally have as our simplified comes — 3(e/4xoo)$(oox )(1/r) d'Z~/ dr o]&, and, finally,
Schrodinger equation for the one-electron functions u, , the total potential energy, for use in the Schrodinger
to replace (7), equation for u;(xi), is

s
Z +31
( ) '(
3 tPZy'11
Hindi'(xi)+ Q(k) ~+i (x2)No(x2)(e/4ir&or12)dxo
i32+i ( d. i
To carry out a self-consistent solution for an atom,
—3(e'/4xoo)
3
— g(k)go*(xi)No(xi) N, (xi) using this simplified method, we then find a Z„such
8 that the wave functions u;, determined from a single
Schrodinger equation using the potential energy (15),
= E,e;(x,). (14) determined from Z„, add to give a charge density which
would lead by Poisson's equation to a potential energy
This equation is in practice a very simple one to e'Z„—
/4xoor
apply. %e solve it for each of the wave functions u;, In order to check the applicability of the method
then find the total charge density arising from all Mr. George %. Pratt is carrying out a self-consistent
these wave functions, and can at once calculate the solution of the ion Cu+ by this method. This ion was
potential energy, including the exchange term, to go chosen, as being the heaviest one for which solutions
into (14), so as to check the self-consistency of the
by both the Hartree and the Hartree-Fock methods are
solution. Here, as before, we change to Gaussian units available for comparison. The solution has gone far
by omitting 4moo, and atomic units by changing e'/kroo enough to show that the wave functions and energy
to 2. parameters E; determined from it are not far from those
One result of this formulation of the self-consistent found by the Hartree and the Hartree-Fock methods.
problem is of immediate interest. In a periodic potential The discrepancies come principally from large values of
problem such as a crystal, it is obvious that the total r, where the charge density is small, and our free-
charge density will have the same periodicity as the electron approximation for exchange is not very good.
potential. Thus the corrected potential of Eq. (14) will Over most of the range of r, however, the approxima-
also be periodic, and hence the functions u; will be tion seems very satisfactory. Detailed results will be
modulated according to Block's theorem. In other reported later, The great advantages of this method for
words, such modulated functions are the only type numerical calculation are clear from this example which
which can follow from a proper application of our has been worked out. Actual calculation is simpler than
simplification of the Hartree-Fock method to a periodic for the original Hartree scheme, since only one poten-
potential problem. tial function need be computed, and can be used for
Our general method is applicable to any problem of all wave functions. The wave functions have the ad-
atoms, molecules, or solids. It is easy to give it a very vantage of being orthogonal; and they possess a con-
explicit formulation for the case of atoms, which can siderable part of the accuracy which the solutions of
then be used for the self-consistent treatment of atomic the Hartree-Fock equations possess, in contrast to the
Hartree equations. It is to be hoped that they will
7 This method of treating the exchange potential as if the elec-
make enough simplification so that it will be possible
trons formed part of a free-electron gas is similar to what is done
in the Thomas-Fermi method with exchange (see Dirac (reference to carry out calculations for more complicated atoms
2}, Slater and Krutter {reference 4), and L. Brillouin, L'Atone de
" than have yet been attempted by the Hartree-Fock
Thomas-Ferosi et lu Nethode de Chaesp "Self-Consistent, Ac-
tualiths Scienti6ques et Industrielles No. 160 (Hermann et Cie. ,
method. At the same time the method should prove
1934)). valuable in setting up solutions for molecules and solids.