ANNALS
01’ PHYSICS
76, 137-160 (1973)
The Quantum Higher Order
Theory
of Nucleon
Correlation
Effects and Additive L.
Department
Correlation
oj’ Physics,
in Finite Nuclei.
II.
Pair Approximation
SZASZ
Fordham
University,
New
York,
10458
AND
J. SCHROEDER Science
Department,
State
University New
of New York, York, 10465
Maritime
College,
Fort
Schuyler,
Received February 17,1972
In the first part of this article it was shown that the variational solution of the Schroedinger equation of a finite Fermion system can be written as a finite sum of A terms (for A particles) the first of which is the Hartree-Fock energy, while the rest represent the correlation effects. In the first part explicit formulas for the 2-particle correlation energy were given. In this paper explicit formulas are given for the higher order correlation energies. It is shown that two different models can be developed depending on the orthogonality condition used. Beginning with the 4th order effects the “linked” and “unlinked” correlation terms are separated. An exact formula is given for the case in which only the 2-particle effects, linked and unlinked are taken into account. The “additive pair approximation” in which the correlation energy is given as the sum of 2-particle
energies
is investigated
and it is shown
formula by a clearly defined set of approximations. model are discussed.
I.
to be related
to the exact
Various possible applications of the
INTRODUCTION
In the first part [I] of this paper (hereafter referred to as I) we have considered Fermion system with the Hamiltonian
a3nite
H =
i
ti + 3 ;
uij ,
i,i=l
i=l
where fi is a one-particle operator and Uij is a general, nonlocal interaction without hard core. The system described by H may be a nucleus in which case vij is the Copyright Q 1973 by Academic Press, Inc. All rights of reproduction in any form resewed.
137
138
SZASZ
AND
SCHROEDER
nucleon-nucleon model potential developed by Tabakin [2] (or some other similar nucleon-nucleon potential); or H may represent the electrons of an atom in which case vij is the Coulomb potential. We have shown in Part I that the variationalsolution of the Schroedinger equation with (1.1) can be written as follows. The trial function is 9 = #F + ; f’“W
+ 2 f’“‘W>
+ ... + .vA’,
(1.2)
which is a finite sum of A terms; the first term is the Hartree-Fock (HF) determinant, the second term represents 2-particle correlations, etc. The symbols are defined by (I, 2.6) and (I, 2.7) respective1y.l The variational energy is given as
E’A’ = EF + 2 j@‘.
(1.3)
n=2
In this finite sum of A terms l?cn) (2 < n < A) is the correlation “n-particle approximation” defined as g(n)
=
E(n)
_
E'n-l,,
(2
< n < A),
where the (total) energy of the n-particle approximation
energy of the (1.4)
is given by
(1.5) with zp
=
*F
+
i
g’k’
(2 < n < A),
(1.6)
(2 < k < A).
(1.7)
k=2
and
g’“) =
c f’“)(ij
... Z)
ij...l
The total effects up which are In I we
energy of the “n-particle approximation” (1.5) contains all correlation to and including n-particle effects. For n = 1, $(l) = #F and ,V) = EF the HF function and energy respectively. have investigated the HF model and the 2-particle approximation. We
1 It is assumed that the reader is familiar with Part I. References which were used in the development of the model presented in these articles were listed in Part I, and will not be repeated here. All notations are taken over from that work without changes. We shall refer to an equation in Part I by writing, e.g., (I, 2.6), meaning Eq. (2.6) in Part I.
THE
QUANTUM
THEORY
have seen that the l-particle HF equations
OF
NUCLEON
functions
CORRELATION
IN
FINITE
NUCLEI
139
& (i = 1, 2,..., A), are the solutions of the
HFA
= 4
3
(1.8)
where HF = t + U,
(1.9)
with u=
i: us,
(1.10)
S=l
where U, is the HF potential given by (I, 2.21). For EF we obtained EF =
C Ei -
+ C
l ,j ,
(1.11)
ij=l
i=l
with Eij
=
<$i
I uj
I $i>*
(1.12)
As was shown in I, EC2)can be computed from the “independent pair model”.2 In this model we computed the correlation functions & one by one, i.e., we considered first #t2) with all but one of thef(ij) zero in the gc2).Varying the total energy with respect to the CJ&(with fixed l-particle functions) we obtained the equation for the best +ij . We have seen also that two different models can be developed depending on the orthogonality condition used. The 2-particle functions are always “strong-orthogonal” to the l-particle functions: (1.13) If in this equation S = 1 *.. A but S # ij we talk about “partial” orthogonality whereas if S = 1 ... A with (ij) included we have “full” orthogonality. Now the equation for the best +ij with partial orthogonality is (Hij + P12Oij) #tj = Gj$ij + (Eij - p120,j) pij y
(1.14)
where Hzj is given by (I, 4.16), Oij by (I, 4.14) PI2 = PIP2 where P is given by (I, 2.18) and pij = det(+,+j). The Cij is the minimum of the correlation energy for the pair (ij). The equation for the best $ij with full orthogonality is
a For the: development of this model, see references given in Part I.
140
SZASZ
AND
SCHROEDER
where the symbols have the same meaning as in (1.14) except now the projection operator PI2 removes all $i’S including & and +j . For ,!?c2)we derived in I the formula (1.16) where Cii is the solution of (1.14), Nij is given by (I, 4.23) N by (I, 4.21) and X is the sum of the off-diagonal matrix components of the Hamiltonian. With full orthogonality we obtained jp)
= c Eii !$ + A$,
(1.17)
ij
where Cij is the solution of (1.15), Nij , fi and T? have similar meanings as in (1.16) but their explicit form is different because of the different orthogonality condition; e.g., 8 is a much simpler expression than X. Having computed the &‘s, one at a time from (1.14) or (1.15) one can compute the .!?c2) from (1.16) or (1.17) by building up these expressions from the pair energies Zij and using the 2-particle functions which are the solutions of (1.14) or (1.15) for the calculation of the matrix components Nij , N and X. In this model the 2-particle functions are computed from independent equations but there are no terms neglected in the energy expressions; consequently (1.16) and (1.17) are upper limits to the exact solution of the Schroedinger equation. Denoting the latter by E we have E’2’ = EF + Et2) > E. In the present paper we pursue the following
(1.18)
goals:
(1) We want to derive the explicit formulas for Et3), ,??c4)etc., i.e., we want to see the full variational solution (1.3). (2) Assuming that the 2-particle correlation effects are much more important than the higher order effects we shall investigate the approximation in which all 2-particle effects-“individual pair” as well as simultaneous “more than one pair” correlations-are taken into account. An exact formula for the correlation energy in this approximation will be derived. In connection with this exact formula, the “additive pair approximation ” in which the correlation energy is given as a sum of the pair energies will be discussed and it will be shown that the “additive pair” formula can be obtained from the exact one by a set of clearly defined approximations. Sections II-V contain the discussion of these topics. In Section VI we review our results and present some thoughts on their applications.
THE
QUANTUM
THEORY
OF
NUCLEON
CORRELATION
IN
FINITE
NUCLEI
141
II. THE ~-PARTICLE APPROXIMATION (A) Partial Orthogonality In developing the 3-particle approximation we adopted the same ideas and methods as in the case of the 2-particle approximation. First we consider a wavefunction in which all but one of theft3) functions are zero:
An approximation based on (2.1) and (2.2) can be called the “independent triplet” model. We want to obtain the equation for the best z,&~ or &jl . We first compute the energy with (2.1). Let f,jL be the correlation energy of the triplet (ijr) with an arbitrary I,!Q~. Then n l iiz = g - EF = <# I H - EF I $Y<#
I #i,
(2.3~
where l? is the expectation value of H with (2.1). We want to vary (2.3) with respect to #ijl. The subsidiary conditions which have to be taken into account are as follows: (a) The l-particle orbitals are the solutions of the HF equations and they are orthonormal: (4, I +t> = sst ; (b)
The 3-particle function is subjected to partial orthogonality: (4s I #iid = O,
(c)
(2.4)
(s = 1 *a-A; s # ijl);
(2.5)
#ijl is normalized: <#ijZ I #ijZi = l*
(2.6)
2,j~ = <#QZI HOZ+ Oijl I #ijt?>
(2.7)
Using (I, 2.20) we get for (2.3) where Hijz = H,(l) 3 In all formulas
the symbol
+ H,(2) + H,(3) - ei - cj - l t,
A means
always
an antisymmetrizer
in this paper.
(2.8)
142
SZASZ
AND
[(Ida)
+
SCHROEDER
and Oijz
=
2’12 +
013 +
u23 -
2
u,(a)
+
UZ,(N)]
f
Eij f
QZ +
E~Z 7
(2.9)
which are the generalizations of (I, 4.16) and (I, 4.14). The variation of Enijlis carried out exactly in the same way as the variation of the corresponding expression in the 2-particle approximation. Using the same methods as in Section IV-B of I, we get IHiiZ
+
p123°ijZl
#iYZ =
cijl#ijL
(2.10)
9
where Zijl is the minimum of the correlation energy of the triplet (ijl) and Pl23 = P,P,P, , where PI is a projection operator (operating on functions of sl) defined as follows: (2.11) From any function on which it operates, P123removes the “basic set” $1 **. +A with the exception of & , $j , & . Multiplying (2.10) from the left by $& we get Zijl in the form of (2.7) which shows that &r becomes identical with Eijl if we substitute the solution of (2.10) into (2.7). We get the equation for the best &jl by putting (2.12)
+ijZ = N(/-%Z + dijZ)> where N is a normalization constant. Putting (2.12) into (2.10) we get LHijZ
+
p123°ijLl
+ijL
=
;ijl$ijl
+
LcijL -
p123°ijLl
PijZ
*
(2.13)
This equation is exactly analogous to (1.14), i.e., the equation for $ij, has the same form as the equation for $ij . We proceed now to the calculation of Et3). First we define and AljZ
= 2(/-kjl
I OijZ
I +ijZI)
+
<$tjZ
i HijZ
+
OijZ
I +ijZ>.
(2.15)
From (2.13) we get Cijl = A
(2.16)
Let us consider now the wavefunction of the full 3-particle approximation. Using the notation introduced in (1.7) we can write $(3)
=
p2,
+
&$3'*
(2.17)
THE
QUANTUM
THEORY
The 3-particle correlation
OF
NUCLEON
CORRELATION
IN
FINITE
NUCLEI
143
energy is by definition Jp’ =
E’3’
_
(2.18)
E’2’.
Using (2..17) to build Ec3) we get ~(3) = 2(c+U2 [ H - E12’ / g’3’) + ( g’3’ 1H - E”’ / g’3’) (zp”’ j tp”‘) + 2
(2.19)
.
It is easy to introduce (2.16) into this formula. First we note that E(2) = EF + ,??(‘J). Substituting this into (2.19) we get matrix components of H-E, which we compute from (I, 2.20). Naturally, in doing that we write $c2) = $F + g(2) and for g(2) and gt3) which appear in (2.19) we substitute their form from (1.7). We obtain by simple calculation ~(3) = c -+$-E’2,(N3;3N2). SijlNijt ijl N3 In this formula Cijl is the solution and X3 are defined as follows:
(2.20)
of (2.13), Nij, is given by (2.14) and N, , N3
N, = 3 ! ( #‘2’ 1 t,V”‘),
(2.21)
N3 = 3! (c+V3) 1 #‘“‘),
(2.22)
and X,=3!‘2(g’2)IH-EFIg’3)j+ 1
cc (f(ij)jH-EFlf(sfu)). ii1 stu (ijl# stu)
(2.23) I
As we see the formula for E”c3) is very similar to the formula for ,!?c2)given in (1.16). Apart from the last term which is proportional to e(z) and very small because of the structure of (N3 - N,), E(3) consist of a term additive in triplets plus the term containing the off-diagonal matrix components X3 . We can compute eC3) as follows. First the 3-particle correlation functions diil and the correlation energies CiiE can be computed from (2.13), one at a time. Besides the &‘s we need the matrix components N, , N3 and X3. These can be computed using the $ifl’s obtained from (2.13). The Et2) appearing in (2.20) assumed to be known from preceding calculations. Using the same arguments as in the case of the 2-particle approximation we can state that if we add Et3), computed as described, to the previously computed Ec2) we get an upper limit to the exact energy: E’3’ = EF + ,??2) + Z?‘3) > E. 595/76/1-10
(2.24)
144
SZASZ
AND
SCHROEDER
(B) Full Orthogonality We have seen in developing the 2-particle approximation that we can develop two different modeIs depending on the orthogonality condition used. The use of partial orthogonality, which does not restrict the generality of the wave function, led to the Eq. (1.14) for the $ij and to the Eq. (1.16) for Et3). The full orthogonality, which restricts the generality of the wavefunction slightly (as we discussed in Section II-B of I) leads to Eq. (1.15) and (1.17), respectively. Comparing these two sets of equations we see that while the two equations for & are not much different the two energy expressions differ considerably. Although (1.16) has the same form as (1.17) the two expressions are different because 2 is a much simpler expression than X. Since the two orthogonality conditions led to two different models in the 2-particle approximation, we can expect the same to be the case in the 3-particle approximation, and since one of the models was much simpler than the other it is worthwhile to explore whether this will be the case in-the 3-particle approximation. Let us start again with the “independent triplet” wavefunction (2.2) but les us now replace the orthogonality condition (2.5) with the following: (4s I 6,t> = 0,
(s = I,..., A; (ijl> included).
(2.25)
We do not have a normalization condition like (2.6) in this case. We get for iij, , the correlation energy of the triplet with an arbitrary $ij2 : (2.26) The variation of iij, with respect to & is carried out exactly in the same way as in the 2-particle approximation (Sec. IV-D of I). We get for dijl the equation
iHijL + p123°ijll Ail = zijl+ijl 9
(2.27)
where PlZ3 is a projection operator which eliminates the “basic set,” including & , r#~, & from the function on which it operates. cijl is the minimum of gijl . As we see this equation is quite different from (1.15). The term -P120ijoij present in (1.15) is missing from (2.27). The reason is that the full orthogonality eliminates the matrix component from which this term would come. With full orthogonality we have <$ijL I Oijl I Pijl) = O. Also, for both orthogonality
(2.28)
conditions, -hjt I Oijl I PLijL) = O*
(2.29)
THE
QUANTUM
Let us multiply We get
THEORY
OF NUCLEON
CORRELATION
(2.27) from the left by (&,
IN
1, and subtract
FINITE
NUCLEI
145
the result from (2.26).
It is clear that iijl should become equal to Cijl if the solution of (2.27) is substituted into (2.26). But (2.30) shows that Eni,l- CijCwill be zero only if +rjl
s 0.
(2.31)
The result is that (2.27) is not compatible with (2.26); therefore (2.27) has only the trivial solution (2.31). Since (2.28) is valid for any higher order correlation function, the same result would be obtained for higher order approximations. In the 2-particle approximation we do not have this problem because the corresponding matrix component is not zero. It is clear therefore that, in the case of full orthogonality, instead of (2.2) we have to start with the following function: * = #F + g’2’ + f($)
= $/J(2)+ f(i)).
(2.32)
Using (1, 2.20) we get by simple calculation
where all symbols are familiar
except the following
<+ijt I Bijz) =
matrix component:
H i gt2’),
(2.34)
which means that we have compressed all terms depending on the 2-particle functions into c&,, , i.e., this function does not contain the 3-particle functions &.r . Variation of (2.33) with the full orthogonality as subsidiary condition gives
where Ci,il is the minimum
of gijl and
$ij, = &Jij, .
(2.36)
The equation for the best 3-particle function now contains all the 2-particle functions in $,, , which is, of course, the consequence of the presence of gt2) in (2.32). Next let us calculate Cijc from (2.35) and compare it with (2.33). The iij, of
146
SZASZ
AND
SCHROEDER
(2.33) must be equal to gija if the &, occurring in (2.33) is the solution of (2.35). From the condition II = ) (2.37) Ejl
Eijl
we get by simple calculation (2.38) where fiz = ( I,P / #(“J>. W e note that (2.38) is exactly analogous (I, 4.33) obtained for cii in I: Eij = &($ij 1Oij j /Jdij)* To show the analogy we use (I, 2.20) to transform
to the result
(2.39) (2.39) back into the form
T&<~G I Oij I ~1) =
I H I #F>-
(2.40)
Now let us recall that in our notation I&
(2.41)
EC l/(l),
so we can write Rl = (Ip
j p)
= (#,
I #F) = 1)
(2.42)
and using (2.40) and (2.42) we can write (2.39) in the form
Gj = Wij)l H I VW% .
(2.43)
The analogy between (2.43), for the 2-particle case and (2.38) for the 3-particle case is complete, thereby showing that our choice for the starting trial function of the 3-particle approximation, Eq. (2.32), is the right one. Let us consider now the complete trial function of the 3-particle approximation: #(3)
=
#‘2’
+
g’3’.
Using our standard formula (I, 2.20) for the matrix components simple calculation, for the 3-particle correlation energy the formula
(2.44) we get by
where Iv3 = (Ip
I $h’“‘),
(2.46)
THE
QUANTUM
THEORY
OF
NUCLEON
CORRELATION
IN
FINITE
NUCLEI
147
and
23= 1231cstu IIf If(stuD.
(2.47)
(ijZ #stu)
Next we introduce Simple calculation
the correlation gives
energies Cijl , the solutions of (2.35), into (2.45).
where Nijl = (i,P) + f(ijZ) / t,V2) + f(i$)).
(2.49)
Our result is that we can compute the correlation functions &it from (2.35), one at a time, if the l-particle functions and the 2-particle functions are known from a -previous calculation. Using the correlation energies Cijr , obtained from the “independent triplet” equations we can compute the total correlation energy ,!?c3)from (2.48). This formula is completely analogous to (1.17), the formula for Ec2). Applying the same arguments which led to (2.24), we obtain again E’3’
= EF +
E(3)
+
E(3)
>, E,
(2.50)
where ,??t2)is obtained from (1.17), E(3) from (2.48) and E is the exact energy.
III.
THE ~-PARTICLE
APPROXIMATION
In the treatment of the higher order effects beginning with the 4-particle approximation, we have to consider now the problem of interacting and non-interacting groups. The quadruple excitations which represent the 4-particle effects may describe the collision of 4 particles (or rather the property of the wave function making such collisions very unlikely in case of a repulsive interaction) or they may describe the simultaneous collision of two pairs of particles. In the former case we may talk about an interacting group which was called a “linked cluster” in the earlier literature; in the latter case we talk about a noninteracting group (“unlinked cluster”). Let us write the full 4-particle correlation function in the form +ijkl = 4&t + +F.L 7
(3.1)
where the superscripts L and U refer to “linked” and “unlinked,” respectively. It is plausible to assumethat the unlinked terms are built from the 2-particle functions +ij , already known from the 2-particle approximation. Therefore our
148
SZASZ
AND
SCHROEDER
task here is to determine the linked term. We will proceed in such a way that first we shall put c$” = 0 and determine 4”. Then with the 4” known, we shall enIarge the model to include 4”. It is obvious from the preceding discussion that we can again develop two different models according to the orthogonality condition. In connection with this we note that the energy expression (2.20) is much more complex than the energy expression (2.48). It is clear that as we proceed to higher approximations we should try to keep the formulas as simple as possible. When we consider a trial function like (3.1) which is rather complex first because it is a 4-particle term, second because it consists of two parts, we must try to work with that energy expression which is the simpler; from the S-particle approximation we see that the full orthogonality leads to a simpler energy expression. For this reason, we shall develop the model with partial orthogonality with a trial function containing only c#+; the model containing 4” as well as 4” will be developed only with full orthogonality. (A) Partial
Orthogonality
For the starting point we choose, in analogy with (2.2), the trial function # = #F + f”fijkl>,
(3.2)
where fL contains only the “linked” correlation function. Since this function is completely analogous to the 3-particle case, we do not need to repeat the details of the calculations here. For the best 4-particle linked correlation functions we get the equation
in complete analogy with (2.13). For the 4-particle correlation energy we get ~(4)
=
c
GjkfijkL 4
+
+
-
(e(2)
+
E’(3))
( N4
N,
N3),
(3.4)
ilk2
where Zijkl is the solution of (3.3) and
NiikL = , N4 = 4! ( i,b4)j a,+b’“‘),
(3.5) (3.6)
and X4 is the collection of off-diagonal matrix components between 4-particle functions of difficult indices, as well as between 4-particle and lower order functions.
THE
QUANTUM
THEORY
OF
NUCLEON
CORRELATION
IN
FINITE
NUCLEI
149
(B) Full Orthogonality Let us start now with the following
trial function:
5) = $bF+ g(2) + g(3) + fyijkz)
= lp3) + fyijkz).
(3.7)
In this trial function we have only the “linked” term. The mathematical procedure is therefore completely analogous to the 3-particle case. We get first the correlation energy of the quartet ([jkl):
where pi = E(2) + Et31 and iijk, contains all lower order correlation functions. Variation of (3.8) with respect to the 4-particle function yields the equation
lHift2 +
- al +i”jkL= Eijkl$kkl- P1234Jijkl 3
p1234°ijk,
where Jij,, is the antisymmetrized form of &,, (3.9) and (3.8) gives the relationship hkl
= <<+EkL I = ((f%‘jkZ)
. The “compatibility
BijkC>im3>
(3.9)
test” between
(3.10)
I H I P’
+ gc3’)/fi3),
where fi3 = (ip
/ zp).
(3.11)
The 4-particle linked terms can be computed from (3.9), one at a time, similarly as in the 3-particle case. As we see, this equation contains all the lower order functions through the presence Of Jijl;, . Let us proceed now to the full 4-particle approximation which contains all linked terms for all quartets:
$A’“’= 3p”’+ c fyijkr). ijkl
(3.12)
With (3.12) we can build up the expression for the 4-particle correlation energy. Using the standard formula for the matrix components, Eq. (I, 2.20), we get
where iv4 = (fp
I tp”‘),
(3.14)
and 24
=
C ijkl
C sptu
(f’-(ijkl)
I H
IfYvW
KijkO
#
(wu>>.
(3.15)
150
SZASZ
AND
Using (3.9) we get for the correlation j34)
=
SCHROEDER
energy the formula ^ ~m$-hlcz I 2 ,
c
ijkl
4
(3.16)
4
where N
/ I+ht3)+ f=(ijk/)).
(3.17)
The philosophy behind the procedure described here is of course the same as the general philosophy of the independent pair model. Writing (3.7) we assumed that the linked correlation function 4” can be accurately computed in the absence of other linked as well as unlinked correlation functions. Having determined the linked terms, one at a time, from (3.9) we must now include the unlinked terms in the total 4-particle correlation energy. In order to do that let us consider now the complete 4-particle approximation including all linked and unlinked terms: (3.18)
p4) = p3) + i;z (fL(ijkZ) + f U(ijkz)).
We get the energy with (3.18) easily from (3.13). Let us introduce the following notations: (3.19) Riikz E Hiikz + Oijkz a; Ah
E
$
<~&cz
1 &kz
1 #hkz)
A%Lz
=
$
<$ikz
1 &kz
1 +zkz);
A&Z
=
&
<$zkZ
I &cl I 4&z)
24L = 2;”
c
c
ijkZ (ijkZ#
sntu sptu)
c
c
=
(fL(ijkZ)
+
2(&k~
(3.20)
1 ‘$iiikz);
(3.21)
+ ~<#‘L%c I $ijkz);
(3.22)
I H If”(sptu));
(3.23)
I H If”(sptu));
(3.24)
(f u(ijkZ) I H I f “(sptu)).
(3.25)
2(fL(ijkZ)
ijkl sptu (ijkl #s&u)
z4”
=
c
c
ilk1 sptu (ijkl #sptu)
Using these notations we get, for the correlation energy in the 4-particle approximation, E(4)
=
c
A&z
iikl
where &a is the normalization
+
Ai’iz
+
Tc4L
A&z
+
xy
+
T4”
(3.26)
+ fi4
integral of (3.18): N4 = (ip 1 zp).
iv4
’
(3.27)
THE QUANTUM
THEORY OF NUCLEON CORRELATION
IN FINITE NUCLEI
151
The Z?L4)can be computed as follows. The linked terms $kkl are the solutions of (3.9), and the unlinked terms #,, can be built up from the 2-particle functions & and (fikl . (See th e discussionbelow.) The expression (3.26) is arranged in such a way that it is easy to extract from it the energies related to the linked or the unlinked terms separately. For example, let us assumethat r#~”< 4” and let us put 4” == 0 in (3.18). Then (3.26) becomes E’(4)
=
c
Afkl
ijkl
j
$L.
4
(3.28) 4
From (3.8) and (3.9) we seethat
where N;,, = (#(3) + f”(ijkl)
1yY3) + fL(i$l)).
(3.30)
On putting (3.29) into (3.28) we get back (3.16). On the other hand if we assume that 4”
=
c ijkl
Z4”
&& N+ 4
(3.31)
K’
Regardlesswhether we use (3.26) or (3.28) or (3.31) we have again the relationship E’4’
=
,5’3’
+
E(4)
3
E,
(3.32)
where E is the exact energy.
IV. THE HIGHER ORDER EFFECTS: FULL VARIATIONAL ENERGYOF THE SYSTEM It is clear from the preceding two sections that we are able to write down the correlation energy of the n-particle approximation, for n = 5, 6,..., A, simply by inspecting the formulas for the 3-, and 4-particle cases.The sameis true for the equations for the n-particle functions. Also, a general procedure can be developed for the unlinked correlation functions as we have demonstrated in the case of 4-particle unlinked functions. The procedure is first to write down the energy with an unspecified general n-th order correlation function, and then write this as the sum of linked and unlinked terms. The result is an expression like (3.26) in which we have “linked, ” “unlinked,” and “mixed” matrix elements. We ar’e now in the position to compare the two different orthogonality conditions. Comparing first the equations for the correlation functions, we see that
152
SZASZ
AND
SCHROEDER
the equations are much simpler with partial orthogonality. To demonstrate this let us look at (2.13) and (2.35) which are the equations for &, . Writing #ii1 = PC1+ $221we have from (2.13) &l#ijC
=
cijl#zjl
(4.1)
3
where Sijl = Hzj, + f’l2,Oijl *
(4.2)
On the other hand, we get from (2.35) szjl&il
=
Gijl
+
ec2))
+ifl
-
p123Jijt
-
(4.3)
Equation (4.1) is much simpler than (4.3) mainly because&, contains all the lower order correlation functions and this complicated expression is missing from (4.1). A comparison of the energy expressions, however, shows that the formulas there are much simpler with full orthogonality than with partial orthogonality; the reason for this being that the off-diagonal matrices are much simpler with full orthogonality. Faced with the dilemma of a choice between the two models we note that one possibleprocedure would be to determine the correlation functions with partial orthogonality first, then orthogonalize them to the HF functions &. , $,i , & ,..., etc., thereby making them fully orthogonal, and then use them in the energy expressionswith full orthogonality. The energy obtained this way is still an upper limit to the exact energy. How can the equations for $ij , & , etc., be solved? No direct (exact) method of solution exist at the present time. Approximate solutions can be obtained by using the CT (configuration interaction) method or by using correlated functions. (We refer to the method developed by Hylleraas [3] and by Clark and Feenberg [4]). We are able now to write down the variational solution of the Schroedinger equation. On the basisof the preceding discussionwe get, using full orthogonality:
+ E(5) +
... +
E(A),
(4.4)
THE QUANTUM
THEORY OF NUCLEON CORRELATION
IN FINITE NUCLEI
153
where the higher order terms can be constructed as outlined. We note that the energies related to the linked correlation functions consist always of a term additive in pairs, triplets, etc. plus a nonadditive term. In order to achieve a certain desired accuracy, it may not be necessary to go beyond the first few terms of the series (4.4). There is evidence in the atomic calculations [5] as well as in the calculations for nuclear matter [6] which seem to show that the 2-particle correlation effects are much more important than the higher order effects. (Actually, for atoms the HF calculations generally give 99 % of the total energy (except for He) which showsthat even the 2-particle correlations are rather small. This is not true, however, in nuclei where the HF calculations gave only about 30-50 % of the total energy per particle [7]). It is plausible therefore to consider an approximation in which only the 2-particle effects are considered. This will be done in the next section.
V. THE “ADDITIVE
PAIR" APPROXIMATION
In the reaction-matrix theory of the nuclear matter developed by Brueckner, Bethe and Goldstone [8] the basic assumption is to consider the interaction of one particle pair in detail while neglecting the interaction of the other particles. This is particularly clear in the formulation given by Gomes, Walecka and Weisskopf (GWW) [9], which they called the independent pair model. In addition to computing the interaction energy of individual pairs independently from the other particles, GWW have also assumed that the total energy of the system is equal to the sum of the energies of the individual (independent) pairs. We shall call this assumption the “additive pair approximation.” We note that a generalization of the Brueckner theory to finite systemssuch as developed by Rodberg [lo] also leads to an additive pair approximation. The importance of pair correlations was early recognized by Macke [l 11. Investigating the properties of an electron gas Macke observed that the second order perturbation energy diverges: a convergent expression can be obtained, however. if one summed up the energies of 2-particle unlinked clusters of all orders. Investigating finite atomic systems Sinanoglu has shown [12] that the second order perturbation theory leads to an “additive pair” formula. In a theory developed for the treatment of electron correlation in atoms [13] Sinanoglu assumedthat the 3-, 4-, and higher order linked terms are negligibly small in atoms becausethe Pauli principle keeps the electrons spatially separated; important are only the 2-particle correlations plus the 2-particle unlinked clusters.
154 According to Sinanoglu
SZASZ
AND
SCHROEDER
[13] for a atom one can write in good approximation (5.1)
where the gij’s are the energies of the independent pairs. Equation (5.1) represents the sum of ,!V?@) plus the energies of all 2-particle unlinked clusters. In the present work we shall adopt the basic idea of Sinanoglu which is that the summation of the unlinked 2-particle terms results in an additive pair approximation. Adopting this idea we shall present here an “additive pair approximation” which will be a generalized and improved version of Sinanoglu’s theory. The generalization will consist of our theory being applicable to finite Fermion systems with a Hamiltonian given by (1.1) and not only to Coulomb interactions. The improvement will consist of some of the approximations being investigated in more detail here than in Sinanoglu’s work. Actually our goal will not be to show that (5.1) is a good approximation. We shall try to answer the question: What exactly are the approximations which lead to (5.1)? While we shall answer this question clearly we will not be able to prove that (5.1) is a good approximation. Let us assume that we omit from the general variational energy expression (4.4) all linked correlation terms beginning with the 3rd order. What is left is the sum of all 2-particle terms linked plus unlinked:
where &?, Et), etc., are the contributions of 2-particle unlinked terms to the higher order correlation energies. We shall assume that these are small and concentrate on the 4th order term. The wavefunction underlying this approximation is (3.18) with the linked terms omitted:
z)(4)= t,P) + ; C C f “(ijkl), 13 kl
where we have written the (ij) and (kl) summations as independent (hence the 4). For the 4-particle unlinked correlation function we put [13]
We evaluate Agkl by putting (5.4) into (3.22). Trying to separate the important
THE
QUANTUM
THEORY
OF
NUCLEON
CORRELATION
terms from the smaller ones we shall proceed as follows. in (3.22):
IN
FINITE
Consider
NUCLEI
155
the first term
(5.5) where we have used the fact that R is symmetric. Now we select from A($ij+ki) that term in which the position of the coordinates is the same as in the function on the left of (5.5) assuming that that term is the leading term. We put therefore
where the word “Exchange” indicates those terms which result from the antisymmetrization of 4&l . The argument underlying this step is the assumption that if we would use a nonantisymmetrized correlation function in (3.22)-although this would violate the Pauli principle-it would not lead to a too large numerical error. A nonantisymmetrized function would yield the first “diagonal” term of (5.6) and in the final analysis we shall keep only this term neglecting the “Exchange” terms. An argument in favor of this method is presented by the HF calculations for atoms. In that case the “diagonal” terms (the Coulomb integrals) are always much larger than the “exchange” terms (the “exchange” integrals). We assume that this argument can be carried over to the case of general nonlocal potentials. We separate similarly the “direct” and “exchange” terms in the second term of (3.32) as well as in fld . In order to compare the 4-particle unlinked correlation energy to the 2-particle correlation, we introduce the Eij’s into the former. This is accomplished by rewriting (3.19) as follows:
where Hij and Oij are the 2-particle operators defined by (I, 4.16) and (I, 4.14) and Bijkl is the remainder of Oiilcl after Oij and O,, are extracted. Putting (5.7) into (3.22) it is easy to seethat we shall get terms of the form of (I, 4.20) which lead directly to cij through (I, 4.22). The lengthy but straightforward calculation gives finally E = EF + ,f@ + j3~4,= & + c %Lii i
l
R L f L’ ’
(5.8)
i.e., the total energy of the 2-particle correlations plus the 2-particle unlinked term is given by EC=+%+&. i
L
i-
L’
156
SZASZ
AND
SCHROEDER
Here Eii is the solution of (1.15) and (5.10) (5.11) (5.12)
i
k
+
2(4i,
/ $h’)
1 u 1 $k,)
+
2(4jk
-
2<$jL
I $ij)
I v 1 +k,>
+
(%k
+
4&(i2)
- 2
i %3 i 4di2)
4kt(34)
j +ij)+il +
%Z +
‘%k
+
%)
,%j~kl
$k1(34))
hcz(34)l Ui(3) + Uj(3) + U,(l) &(34))
/ 2’ I ‘$k,)
+ Wl)
+ -& + -%a” + exchange terms.
(5.13)
In the above expressions the prime on the summations means that (kl) # (ij); the 2-particle functions are the solutions of (1.15); l ilz is defined by (1.12) and Ui is the HF potential (I, 2.21), xz and xdu are the off-diagonal terms in (5.2). Finally L’ is the “exchange” part of the normalization integral fld . It is easy to seefrom (5.9) that if we putfU = 0 in (5.3) then (5.9) becomesequal to EQ). We are now in the position to answer the question: What approximations lead to (5.1)? The expression (5.9) is exact with (5.3). We get the additive pair formula from (5.9) by assumingthat (1) R and L’ are negligibly small, and (2) L,JL can be replaced by 1. As we have indicated before we cannot prove that these approximations are permissible. We can present a semiquantitative argument, however, showing that the above approximations are not unreasonable. Consider first the L,,/L. Replacing pij by its average /3 we get from (5.10) and (5.12) for a system with A particles Lij _L -
1 + gp + ;/?(A - 2)(A - 3) + gl”(A - 2)(A - 3) 1 + &&4
- 1) + &@A(A
- l)(A - 2)(/t - 3) .
(5.14)
In I we have presented arguments showing that /I must be a small number. We have evaluated (5.14) for the following cases:6 = k (constant) or p = kA-” for n = 1, 2, 3. Setting k = 0.01 we get the curves shown in Diagrams 1 and 2.
THE
QUANTUM
THEORY
OF
NUCLEON
CORRELATION
1.0 ‘T-.‘-Lij I ‘Ice ..‘. .-\ i 0.8 L ‘,‘.\\‘\ I\r .’ ‘.\\.\ i \ o.b!-
-,
\
0.4
c
\
JO.94
1.
J \
\\
\
\ 50
100
I57
0.96
\
\\
------l.\\i.
NUCLEI
IO.98 1 \
0.2 ;o.,L.
FINITE
\\
\
/
IN
I 150
I 200
0.92 '\ '\\I! *
o.go
FIG. I. The quantity L,JL from Eq. (5.14) as a function of A. The full curve, plotted on the left hand scale, is for the case ,!?= 0.01. The dotted curve, plotted on the right hand scale, is for the case /3 = 0.01/A.
FIG. 2. The quantity L,/L from Eq. (5.14) as a function of A. The full curve is for the case p = 0.01/,42. The dotted curve is for the case p = 0.01/M.
As we seefrom (5.14) the limit of L,j/L is 1 for A = 2 asit should be. The diagrams show that for ,8 = 0.01 the curve goes to zero at about A - 90; for k = 0.01/A the curve goes to zero for A + a3 but it is still about 0.9 for A - 230. For /3 = 0.01/A2 and p = 0.01/A3 the curves go to 1 for A + co and the deviation
158
SZASZ
AND
SCHROEDER
from 1 is only about 1%. Therefore if /3 = 0.01 and constant the Lij/L cannot be replaced by 1; if /3 = 0.01/A the error introduced would be about 10 “/,; for ,/?I= 0.01/A2 and jI = 0.01/A3 the error introduced would be less than 1%. A different choice of k leads to similar curves. It is clear that in order to get a meaningful approximation in (5.1) or indeed in (5.9) /3 must be proportional to A-” with n > 1. If this condition is satisfied then L,JL can be replaced by 1 in good approximation. Looking at L’ and R we can neglect L’ as an exchange term and we can neglect the similar terms in R. In R we have the off-diagonal terms xZ and $“. We have seen in I that &, can be considered small relative to the diagonal term Cij Aij [I, Eq. (4.19)]. Also, calculations for Coulomb interactions [5] have shown that 8, is small compared to the diagonal term. While the numerical values will be very different for nuclear calculations, in a semiquantitative discussion we can assume that the relative magnitude of these terms will be the same for a nonlocal nucleon-nucleon interaction as it is for Coulomb interaction. The same argument applies to the relative magnitude of X4” and A& (See (5.2)). In the first four terms of R the matrix (pFLjl/ v I $,,) has the same magnitude as Zij [I, 4.331 but they are multiplied by a term of the magnitude of /3; also there will be considerable cancelation between these four terms. The fifth term is proportional to j3”. Finally, no argument could be formulated for the 6th and 7th terms of R. Summarizing, we may say that the approximations leading from (5.9) to (5.1) are plausible, although we cannot ascertain their accuracy at this stage. Our attitude will be that (5.1) is certainly a useful starting point because of its simplicity; should it prove to be inaccurate, one can always fall back on the exact formuIa (5.9).
VI.
DISCUSSION
We consider Eq. (4.4) the main result of this paper. Looking at both parts of the article our main result is that the variational solution of the many-particle Schroedinger equation is a finite sum of A terms as given by (1.3); the explicit form of that expression is given by (4.4). We have also shown that the terms of (1.3) can be calculated in successive approximations. At each step a system of independent equations must be solved for the correlation functions. The energy is an upper limit to the exact energy at each stage of the approximations. We have also seen that the unlinked correlation terms can be incorporated into the energy as in (5.9); plausible approximations led us to the “additive pair” formula given in (5.1). We can discuss our results from several different points of view. From the point of view of general mathematical physics, our results provide useful insight into
THE
QUANTUM
THEORY
OF
NUCLEON
CORRELATION
IN
FINITE
NUCLEI
159
the structure of the many-particle Schroedinger equation with fairly generally defined particle-particle interactions. For the nuclear physics our method provides a scheme on the basis of which highly accurate calculations can be made. It is clear that calculations with, e.g., Tabakin potentials or with other similar phenomenological nucleon-nucleon potentials can now be made utilizing our scheme; such calculations can be based, e.g., on the two-particle approximations or on the additive pair approximation. The accuracy of such calculations will certainly surpass the accuracy of the HF calculations. We note that while our method is developed for finite systems, the generalization of the method to arbitrarily large systems will be straightforward, opening the way for the development of a Thomas-Fermi type model and for nuclear-matter calculations. Work in this direction is in progress. Finalty the question might be asked: Is it possible to test the applicability of phenomenological nucleon-nucleon potentials to many-body problems with this method? What we have in mind is whether on the basis of a calculation made with our method one could decide if a given phenomenological potential accurately describes the nucleon-nucleon interaction in the nucleus? A clear answer could only be given if the exact solution of the Schroedinger equation would be available. Since that is not the case, what can be said at this stage is that e.g. using the 2-partitle approximation we can get a more reliable answer to the question above than what would be obtained from a HF or perturbation theory calculation.
ACKNOWLEDGMENT The
authors
are indebted
to Dr.
J. Shapiro
for useful
discussions.
REFERENCES 1. 2. 3. 4. 5.
L. SZASZ AND J. SCHROEDER, Ann. Whys. (N.Y.) 68 (1971), 1. F. TAIBAKIN, Anu. Phys. (N.Y.) 30 (1964), 51. E. A. HYLLERAAS, Z. Phys. 54 (1929), 347; 60 (1930), 624; 63 (1930), 291. J. W. CLARK AND E. FEENBERG, Phys. Rev. 113 (1959), 388. R. E. WATSON, Php. Rev. 119 (1960), 170; D. F. TUAN AND 0. SINANOGLU, .I. Chem. Phys. 41 (1964), 2677; R. K. NESBETT, Phys. Rev. 175 (1968), 2; L. SZASZ AND J. BYRNE, Phys. Rev. 158 (1967), 34; H. F. SCHAEFER AND F. E. HARRIS, Phys. Rev. 167 (1968), 67. 6. H. A. BETHE, Phys. Rev. B 138 (1965), 804. 7. A. K. KERMAN, J. P. SVENNE, AND F. M. VILLARS, Phys. Rev. 147 (1966), 710; W. H. BASSICHIS, A. K. KERMAN AND J. P. SVENNE, Phys. Rev. 160 (1967), 746; A. K. KERMAN AND M. K. PAL, Phys. Rev. 162 (1967), 970. 8. K. A. BRIJECKNER AND C. A. LEVINSON, Phys. Rev. 97 (1955), 1344; H. A. BETHE, Phys. Rev. 103 (1956), 1353; H. A. BETHE AND J. GOLDSTONE, Proc. Roy Sot. A 238 (1957), 551. 595/76/1-
1I
160 9. 10. 11. 12. 13.
SZASZ L. L. W. 0. 0.
C. GOMES, J. D. WALECKA S. RODBERG, Ann. Phys. (iV. MACKE, Z. Naturforschung SINANOGLU, Proc. Roy. Sot. SINANOGLU, J. Chem. Phys.
AND
SCHROEDER
AND V. F. WEISSKOPF, Ann. Y.) 2 (1957), 199. A 5 (1950), 192. A 260 (1960), 379. 36 (1962), 706, 3198.
Phy.s. (N.Y.)
3 (1958),
241