CORRELATIONS IQ~INFINITE SYSTEMS .S Roeati Istituto di Fisica dell'UniversitY, Piss, Italy Istituto Nazionale di Fisica Nucleate, Sezione di Piss, Italy ÷ .S Fantoni Institut for Theoretische Physik Universit~t zu KOln, Cologne, W.-Germany Abstract. An extensive analysis of strongly-interacting infinite Fermi systems is presented within the framework of the variational theory ba- sed on trial correlated wave ?unctions. The Fermi hypernetted chain (FHNC) procedures for calculating the energy expectation value and other quantities of physical interest are derived in correspondence to the following two cases: )a{ polarized Fermi systems described by Jastrow- Slater wave ?unctions with state-independent correlation faotors,(b)un- polarized Fermi systems described by Jastrow-type wave ?unctions, with correlation factors depending on the spin-isoepin state of the partic- les. Numerical results are presented for a simplified model of nuclear matter, for liquid 3He and for the hard-spheres Fermi system. The cal- culation of the surface energy coefficient for nuclear matter is also discussed. ÷ Permanent address: Istituto di Fisica deIl'Universit~, Piss, Italy .I Introduction In the present paper some problems related to the variational approach to the study of strongly interacting Fermi systems are discussed. Basi- cally, the analysis is limited to the case of infinite, homogeneous and translationally invariant systems, but some results for finite systems, obtainable in rather Simple ways, are also presented. The approach is non relativistic, and the particles which constitute the system are sup- posed to interact by means of two-body forces. The Hamiltonian is then written in the form A A H=T+V= t(i) + ~ V(i,j) , (1.1) i=I i<j:1 where t(i) is the kinetic energy operator and A is the number of par- ticles. In the calculations, A and the volume 0 of the cubic box contai- ning the system are taken to be finite but large enough to finally con- eider the thermodynamic limit: A and £ are left to go to infinity, but the density ~ is kept fixed. The wave function @ for the system is written as @:F¢=F(1 ..... A)@(1 ..... A} . (1.2) The model function ¢ must adequately describe the ground state of the system in absence of strong interparticle correlations. ¢ is therefore chosen as the Slater determinant of the Fermi sea ~(1 ..... A)=detlCBz(i)=. (8)s61... 8A¢61(I)...¢6 A(A) , i.3) where (6) stands for 18 ..... 6 A and s8 .8 is the permutation symbol, The free-particle wave function ~6 (i)li'n" wh~chthe argument refers to the particle . and the subscrzpt • to the si tate, has the form BS ~.iB~i(pxe2/I-~)i()is(pq=)i( )i , )4.1( l where the spin (-isospin) functions qp(i), p=1 ..... v, constitute a com- plete orthonormalized set. The number v is called the multiplicity of the system. The allowed momenta ~. satisfy the usual periodicity con- 1 ditions on the walls of the box, ki,x=ni,x~/a, ki,y = ni,y~/a , ki,z=ki,z~/a (ni, x,ni,y ,ni,z ~0) , I( • )5 where a is the side of the box. The correlation factor F, in eq. (1.2), is an operator which takes care of the correlations induced by the strong interparticle interactions. It is require d to satisfy the cluster property, i.e. when a subset i I ..... i r of particles is far yaw'a from all the remaining ones, the following relation holds: F(I ..... )A ÷Fr(i I ..... ir)FA,r(ir+ 1 ..... i )A . (I.6) Interesting applications are possible if the system is allowed to pos- ses different populations in the v single-particle spin states. The problem of evaluating the mean value of the total energy for those sys- tems has been treated by ROSATI and FANTONI )I( in the case of a state -independent Jastrow choice )2( for the correlation factor F, namely A =F ~ fir..) . (1.7) i>j= I jI In section 2 the problem is discussed in details, and the cluster ex- pansion for the two-body distribution function is given and summed up in the frame of the Fermi hypernetted chain (FHNC) theory. Section 3 presents some numerical results obtained for three specific polarized Fermi systems, namely a semirealistic model of non-symmetrical nuclear matter, the polarized 3He and the hard-spheres system. A state-independent correlation factor, as specified by eq. (1.7), ap- pears to be appropriate for systems of particles interacting through central forces at not to high densities (see, for example, the review of KRALC (3)). However, to obtain very accurate results, some improve- ments are necessary. First of all, at high densities, effective three- body correlations become important: the generalization of eq. (1.7) to include effective three-body correlation factors is obvious and the corresponding calculations, which do not present particular adjunctive difficulties, can be found in ref. (4). On the other side, serious pro- blems are encountered if the correlation factors are taken to be state- dependent, i.e. they contain some dependence on the spin operators, the velocities, the angular momenta, and so on, of the particles (see, for example, ref. (5)). The correlation factors considered in section 4 in- blude a rather simple spin dependence which does not give troubles for the calculation of all the cluster terms contributing to the two-body distribution functionj the corresponding wave function reveals to be of some interest for a satisfactory description of certain configurations of a Fermi system. Section 5 is devoted to some problems of interest in the variational theory for Fermy systems. First of all, the importance of a two-body correlation ~actor with a structure, ~or both small and medium inter- particle separations, is stressed. The calculation o~ the momentum dis tribution is then briefly discussed. Finally the results ~or the sur- face energy in e Fermi system are given. 5 2. FHNC approach to polarized Fermi s~sflems 2.1. Cluster expansion. Let us indicate with x (K), K=I,...,v, the per- centage of particles in the single-particle spin state K. The density of particles in the state K is related to the total density P=A/Q by the relation p(K)=x(K)p. The allowed values for the momenta of the par- ticles in the state K are contained in the Fermi sphere of radius k~K)=(6~2p(K)) 1/3 , (K=I ..... v) . (2.1) The two-body distribution function g(I'J)(1,2}, with particles I and 2 in the spin states I and ,3 is defined as (I,3) A(A-1) ~- g (1,2) = @#~(I)(I) #(3)(2)~dx3...d~A, (2.2) ~p2 (~T where x i represents both position ~. and spin ~ariables of the particle 1 i, so that summation over ~i' implies both integral ever all the space and sum over the spin varxables.~r=~ . ~ ~dXl...dXA÷ ÷ , and ~(K) (i) is the projection operator on the state qK for the particle i. The particles (indices) I and 2 will be called as "external" particles (indices), the remaining ones as "internal". In the limit of an infinite and translationally invariant system, the distribution function depends on- ly on the distance of the particles 1 and 2. With the choice of W given by eqs. (1.2) and (1.7), we can write (I,J) r12) ~#-~ (1) 2)¢i>j=I~ (r.lj)dx3...dXA .2.3) The procedure adopted here to derive the cluster expansion for the di- stribution function, differs in some respects from the one of ref. )I( and is quite similar to that one utilized for unpolarized systems in ref. }6( (see also refs. )7( and (8)}. First of all, we have to obtain the expansion of the squared Slater determinant AA=¢ ~ ¢. The model func- tion ~ specified in eq. (1.3) can be written by performing the antlsymme trization with respect to the particles or to the states, so that $~(1 ..... A)= £ .~A~ ~ * (~A) = (~) ~1'' (~I)'''~A (2.4) =E61...6A {~)e 1'''~A¢B1(~1)'''~BA (aA) ' If the final expression is used together with eq. (1.3), we obtain AA(1 ..... A) = ~ s ~)¢;Ie1)'''~6ACmA)¢61(1)'''¢BA A) ' (2.5 ~) ~I"''~A Due to the antisymmetrization with respect to (~), the summation on 18 ..... B A can be extended to include also equal values, so that we £et AA(1 ..... A)= detlP(i,3)l , (2.6 where A 2.7 P(i,J) = !l¢8(i)¢B(J) • 8 The p-particle subdeterminants Ap defined as A (1 ..... pl=detlp(i,j) l, i,jg1,p , (2.8) P satisfy the relations (see refs. (6,8)) fAp+ld~p+i=A-p)&p 2.9 & ~0, p>A (2.10) P Let us examine the structure of the terms which contribute to 6p defi- ned by eq. (2.8). Any permutation of p particles can be seen as pro- duct of independent cyclic permutations of separate subsets of these particles. The parity of a cyclic permutation involving q particles is (_)q-1. We have, therefore, the following rules: The terms of Ap are products of closed loops of factors p(i,j) involving the p particles in separate loops; to every loop of q p-factors (or par- ticles) is associated a factor (_)q-1. In particular, if a particle i is not exchanged with other particles, a factor p(i;i),which reduces to the density p when the spin traces are calculated, is associated with that particle. For the in?inite polarized system, we are interested ,ni one £ets A v KA p(i,j)= ¢~(i)¢~(j) = n~i)nKj)a -1 exp( ikBK. + ÷ rj-ri)= + a=1 K=I 6K=I (2.11) v oCK) 1 ~K) = ~ nKi)nK J) (k rij), K=I where l(k~K)y)= 1 dR expi~.~) 2.12) (2~)~p '}K f k ~k p The function defined by the latter equation is called the statistical correlation factor and it satisfies the two following relations d;1Ck~K)y):;d;12k~ K)y)=I/p (K) , (2,13) ;dglCkF{K)y)lCk }K I; 1 (2.14 The explicit analytical expression of the statistical factor is iCz)=3z-3(slnz -zcos z } . C2.15) Let us now consider the quantity OcJ c2)A X C2.16 ~ P where the spin sums are evaluated for all particles, except I and 2. The various terms which contribute to thls quantity can be easily evaluated using the expression obtained for & . The operator ~(I)c1) ~(J)C2) is P associated to each term of A and then the spin sums are calculated. P The following statistical loop rules are satisfied: If aloop of p(i,j) factors does not include particles I and 2, after - spin summations,v different loops are obtained. The first is construc- ted with the statistical factor x(1)l(k~l)rij),-~ the second with x(2)itk(2)r " F ij), and so on. To every loop is associated a factor (_)q-1 where q is the number of statistical correlations. If particle I is involved with other q-1 particles, but not particle - 2, in a loop of p(i,j) factors, after spin summation, a loop of fac- tors x(I)l(k(I)r ) involving the same q particles, with an adjunc- F i' tive factor (_)q-l, is obtained. If both the particles I and 2 are involved in the same loop of p(i,j) - factors, after spin summation, two chains of statistical factors, having the particles I and 2 as extremitles, result. One chain is con- structed with x(I)l(k~I)rij)- and the other chain with x(J) l(k~J)rij).- Let us sxpllcitly calculate the two-body distribution function in the case of a non interacting polarized Fermi gas. This is obtained 9rom eq. (2.3 with the positions ~I)= ~(3)=I and f(r..)=1, JI g2FCr I 2 ) A(A-I) A ÷ d= ÷ A" c2.17) where the spin summations are mow extended to ell the particles. By using eq. (2.9) and ~=A!, we get = AA-I) & =1 ! g2F(r12) E o~ r (A-2)I 2 2p T£(I'1)p(2'2)-p(1"2)p(2'1) v ~K) 2 = 1- x(K)l(k r12) o 2.18) K=I For an unpolarized system x(K)=l/v and kF(K)=k F for K=I ..... v, and the distribution function reduces to the well known expression I12 (2.19) g2F(r12)=l"~ (kFr12). The calculation of the two-body distribution function for a correlated system is a far more difficult problem. First of all, let us consider the quantity F 2= A f~rij)J the function f(r) which will be considered i>j=1 is such that fCr)~l as r÷® and, moreover, it does not differ apprecia- bly from unity for most of the values of r. As a consequence, the func- tion h(r)=f2(r)_1 (2,20) is appreciably different from zero only in a small region and can be used as an expansion parameter in evaluating the distribution function or other interesting quantities. Let us now substitute by 1+hrij) all the factors f2{r..) which constitute F 2 and then expand the products in JI terms of the function hit..). If all the terms related to a given num- JI bet p of particles particles I and 2 being always included) are grou- ped together, we get the expression A A F2=X2 (1,21 + ~ X31,2,i 3) + Z X411,2,i ,i 4} (2 21) i3=3 i3>i4= 3 3 . . . . One has X211,2)=f2r12) and all the remaining Xp functions contain f2(r12) as a factorj the explicit expressions of the various Xp func- tions can be derived without difficulty. If expression 2.21) is sub- stituted in eq. (2.3) for g(I'O{m12), since the quantity in eq. (2.16) is symmetric with respect to the interchange of the particle coordina- tes ~3 ..... A+X , all the terms Xp which differ only in the labels of their arguments may be relabelled and summed to~ether to give gCI'Jlcr12 =A(A-! ~' ~ (11 21& + ~Pp~ T 2 (A-plIp-2IlXp(1 .... P)d~3"'drA" {2.22) Integratlon and spin summation over the particles p+1 .... ,A not appea- ring in Xp, can be readily performed using the identity (2.9) to give a factor A-pIA : P g(I'Jlr121Np p 2 T~'TT • 1 2)A Xp(1 .... p)dr3..drp.2.23) The summation over p has been extended to ~ in virtue of identiy (2.10). The calculation can be carried on in the same way as that one employed for unpolarized systems dna discussed in ref. C6). eW od not repeat here all the details of the procedure but limit ourselves to exhibit the most important feature of the calculation. ehT quantity enclosed in square brackets in eq. [2.23) nac eb expanded sa stated before, sa mus of statistical terms Tstat, each characterized by a product of non-overlapping closed loops of statistical factors. A generic term T appearing sa integrand in eq. (2.23) is the product of a statistical term Tstatwithadynamical eno ydT n, contributing to Xp, os that T=TstatTdy n is a function of the dynamical correlation factors h(rij) dna the statistical correlation factors. In general T nac eb written sa product of unlinked terms,i.e, functions without nommoc in- dices. Let su consider all the terms which enter in eq. (2.23) dna have a same llnked part involving a given number of particles together with particles I dna 2. All the unlinked parts of these terms nac eb integra- ted with respect to their arguments ,and the results demmus pu to give the normalization constant~ . In this yaw ew ~et C(I.') 1 r F n ~ [1,2 ..... n) id~3...d~.n, gCI,J)Crl2)=~Zn_L2j (n-2)l (2.24) where C[I'3)(1,.2. ... n) indicates the mus of all the linked terms con- n structed sa allowed products of statistical ant dynamical factors inte- resting the particles 1,2 ..... n. All the ways f labelling the n-2 in- ternal particles in a term lead to the same coqtribution in eq. [2.24): the number of different labelling ways is (n-2)l! ~ divided by the symmetry I )9((rebmun ') ,S of the particular term considered. I Moreover, it nac eb nwohs that all the constributions from the Ireducible [") terms can- ceil out so that, finally, ew set ii I I ! ® CI,J) ~ r CI'J] ,2]=~ ~ CI'J)C1,2, i.,p)d~3 .d; (2.25) g Cri2)= (I . . . . 2=p 2=p i P " I I (') ehT symmetry number ,S associated with a given term containing the indices 1,2 ..... n , is defined sa the number of permutations of the indices 3,4 ..... n which leave the term unchanged. (") A term is called reducible when its integral, sa specified by eq. (2.24), nac eb factorized into a product of 2 or more integrals. 01 In the latter equation ~(I'3)(I,2 .... )p indicates the sum Of ail irredu- cible, topologically distinct terms corresponding to the external par- ticles I and 2 in the states I and ,J respectively, and with p-2 inter- nal particles. Each term is divided by its symmetry number S and is an irreducible product of statistical and dynamical correlation factors. The following rules hold: The statistical correlation factors constitute closed, separate loops - and satisfy the statistical loop rules discussed before. The external particles I and 2 are associated with the function - f2(r12). With each internal particle i is associated at least one function - h(rik). Each integration on particle coordinates implies a factor .p - 2.2. FHNC equations. Once we have specified all the terms which con- tribute to the two-body distribution function, we must device a technique to sum all these contributions. In the case of an unpolarized system this sum can be expressed in terms of four functions which satisfy a set of integral equations )4( known as FHNC equations. These functions will be denoted here as Ndd(r12), Nde(r12), Nee(r12) and Ncc(r12), where the subcripts ,d e and o stand for dynamical, exchange and cyclic, respec- tively. Each function N is the sum of an infinite number of terms with structures symilar to those involved in eq. (2.25). The terms corres- ponding to Ndd(r12) include the particles I and 2 only with dynamical correlation factors, those corresponding to Nde(r12) include the par- ticle I only with dynamical correlation factors and the particle 2 with two statistical factors. Those corresponding to Nee(r12) (Ncc(r12)) in- clude both the external particles with two (one) statistical correlation factors and an arbitrary number of dynamical ones. It is to be noted that for unpolarized systems, there is only one sta- tistical function l(kFrij)J while for polarized systems there are, in general, v different statistical functions and, therefore, v different chains of statistical factors. As a consequence, to sum up the terms contributing to the distribution function of a polarized system, it is convenient to introduce a larger number of functions, which are denoted N(I'J)( ) N(I'J)(r12 ) N(I'Jl(r12 ) and N(I'J)(r12 ) However. when as -dd r12 ' de ' ee cc ' the external particle 1 (or 2 is connected only by dynamical correla- tion £actors, no dependence on the state I (or J) occurs, so that we have: