1: %$Id: bf.tex,v 1.18 2003/04/14 15:59:25 nuclear Exp $
2: \documentclass[superscriptaddress,twocolumn,showpacs,floatfix,prc]{revtex4}
3: \usepackage{graphicx}
4: \renewcommand{\vec}[1]{\mbox{\boldmath ${#1}$}}
5:
6: \begin{document}
7:
8: \title{
9: Spin-orbit induced backflow in neutron matter with
10: auxiliary field diffusion Monte Carlo
11: }
12:
13: \author{L.~Brualla}
14: \email{lbrualla@sissa.it}
15: \affiliation{International School for Advanced Studies, SISSA,
16: and \\ INFM {\sl DEMOCRITOS} National Simulation Center \\ Via Beirut
17: I-34014 Trieste, Italy}
18: \author{S. Fantoni}
19: \email{fantoni@sissa.it}
20: \affiliation{International School for Advanced Studies, SISSA,
21: and \\ INFM {\sl DEMOCRITOS} National Simulation Center \\ Via Beirut
22: I-34014 Trieste, Italy}
23: \author{A. Sarsa}
24: \email{sarsa@ugr.es}
25: \altaffiliation{Present address:
26: Departamento de F\'{\i}sica Moderna, Universidad de Granada, E-18071 Granada,
27: Spain}
28: \affiliation{International School for Advanced Studies, SISSA,
29: and \\ INFM {\sl DEMOCRITOS} National Simulation Center \\ Via Beirut
30: I-34014 Trieste, Italy}
31: \author{K. E. Schmidt}
32: \email{kevin.schmidt@asu.edu}
33: \altaffiliation{Permanent address: Department of Physics and Astronomy,
34: Arizona State University, Tempe, AZ, 85287}
35: \affiliation{International School for Advanced Studies, SISSA,
36: and \\ INFM {\sl DEMOCRITOS} National Simulation Center \\ Via Beirut
37: I-34014 Trieste, Italy}
38: \author{S.~A.~Vitiello}
39: \email{vitiello@unicamp.br}
40: \affiliation{Instituto de F{\'\i}sica {\sl Gleb Wataghin}\\
41: Universidade Estadual de Campinas, Unicamp}
42:
43: \date{\today}
44:
45: \begin{abstract}
46: The energy per particle of zero-temperature neutron matter is investigated,
47: with particular emphasis on the role of the
48: $\vec L\cdot\vec S$ interaction. An analysis
49: of the importance of explicit spin--orbit correlations in the description
50: of the system is carried out by the auxiliary field diffusion Monte
51: Carlo method. The improved nodal structure of the
52: guiding function, constructed by explicitly considering these correlations,
53: lowers the energy. The proposed spin--backflow orbitals can conveniently
54: be used also in Green's Function Monte Carlo calculations of light nuclei.
55: \end{abstract}
56:
57: \pacs{26.60.+c,21.65.+f,21.30.Fe,05.10.Ln}
58: %26.60.+c Nuclear matter aspects of neutron stars
59: %21.65.+f Nuclear matter ...
60: %21.30.Fe Forces in hadronic systems and effective interactions
61: %05.10.Ln Monte Carlo methods
62: \maketitle
63:
64: \section{Introduction}
65: In recent investigations \cite{sarsa03,fantoni01b} of the ground state and
66: the magnetic properties of neutron matter with
67: modern nuclear interactions of the
68: Urbana-Argonne type \cite{wiringa95,pudliner97}
69: good agreement was observed between results
70: obtained with the Auxiliary Field Diffusion Monte Carlo
71: (AFDMC) \cite{schmidt99}, a calculation by
72: \textcite{morales02},
73: performed with the Variational Chain Summation (VCS)
74: method \cite{akpa97,akmal98}, and Brueckner Hartree Fock
75: estimates \cite{baldo00,vidana02}.
76:
77: However, in spite of the overall agreement of the equation of state and
78: the spin susceptibility,
79: marked differences exist between the AFDMC and the VCS calculations, concerning
80: the contribution to the energy due to the spin--orbit component of the
81: two--body interaction. VCS calculations, performed with the
82: Argonne $v_{18}$ \cite{wiringa95} and the Urbana-IX three--body potential
83: (AU18 Hamiltonian),
84: find large and negative contributions from the cluster terms $C_{LS}$
85: with either spin-orbit correlations in the trial function and/or the
86: spin--orbit potential. For instance, in correspondence
87: with optimal trial functions,
88: $C_{LS}$ amounts to -5.8 MeV at a density $\rho=\rho_0=0.16$ fm$^{-3}$ and
89: -12.1 MeV at twice the same density. The AFDMC calculations of
90: Ref.~\cite{sarsa03}, performed with the simplified,
91: but still realistic, version of the Argonne two--body
92: potential, $v_{8'}$ \cite{pudliner97},
93: plus the Urbana IX three--body interaction ($AU8'$ Hamiltonian)
94: get energy differences $\Delta E_{LS}$
95: between the $AU8'$ Hamiltonians with and without
96: the spin--orbit potential which are small and positive. At $\rho=\rho_0$,
97: $\Delta E_{LS}= 0.2$ MeV and at $\rho=2 \rho_0$,
98: $\Delta E_{LS}= 1.12$ MeV.
99:
100: Such a discrepancy cannot be ascribed to differences between
101: the two potentials $v_{18}$ and $v_{8'}$. It is well known that they
102: provide very close results for the energy per
103: particle \cite{pudliner97}. In addition, this
104: discrepancy is confirmed by other FHNC/SOC calculations \cite{sarsa03},
105: performed by using the approximations of
106: Ref.~\cite{wiringa88} and the $AU8'$
107: Hamiltonian. They give $\Delta E_{LS}=-3.7$ MeV at $\rho_0$ and
108: $\Delta E_{LS}=-10.1$ MeV at $2\rho_0$, with $C_{LS}$ being
109: $-3.9$ MeV and $-10.7$ MeV respectively.
110:
111: A possible source for this
112: disagreement might be the use of a less than satisfactory guiding
113: function in the AFDMC method \footnote{
114: In the AFDMC method a path constraint is used to deal with the fermion
115: sign or phase problem. In all the calculations reported here the
116: real part of the guiding function, evaluated at the walker position
117: and spin, is constrained to be positive in analogy with the fixed
118: node approximation for central potentials. For convenience we refer
119: to the effect of the constraint as the nodal structure of the guiding
120: function.
121: }.
122: The nodes of the plane wave Slater
123: determinant might be too poor, particularly when the interaction
124: includes a spin--orbit potential, like in $AU8'$. The results of the
125: FHNC/SOC calculations of Ref.~\cite{sarsa03} with the $AU8'$
126: Hamiltonian and a trial function of the type $F_6$ (not
127: containing spin--orbit correlations), give values
128: for $\Delta E_{LS}$ quite close to the AFDMC ones, which may confirm
129: this hypothesis.
130:
131: To clarify this issue we modify the guiding functions in our quantum
132: Monte Carlo calculations to contain explicit spin--orbit
133: correlations. This is efficiently done by considering
134: orbitals of the spin--backflow form in the
135: Slater determinant, as explained below.
136:
137: The structure of this paper is as follows. In Sec. \ref{sec.2} we show
138: the guiding function used. The computational details are given in
139: Sec. \ref{sec.3}. The results are shown and discussed in Sec.
140: \ref{sec.4}. The conclusions and perspectives of the present work
141: can be found in Sec. \ref{sec.5}.
142:
143: \section{Spin-Orbit Induced Backflow}
144: \label{sec.2}
145:
146: The $\vec L\cdot\vec S$ correlation in FHNC/SOC and VCS calculations takes
147: the form
148:
149: \begin{eqnarray}
150: F_b(1,2) = \frac{1}{4i}f_b(r_{12})\left[ \vec r_{12}\times
151: \left(\vec\nabla_1-\vec\nabla_2\right) \right ] \cdot\left(\vec\sigma_1+
152: \vec\sigma_2\right)
153: \label{lscorr}
154: \end{eqnarray}
155: where $r_{12}=|\vec r_{12}|$ and $\vec\sigma_j$ are the Pauli matrices for
156: the $j$th particle.
157:
158: By inspection of the cluster terms of $\langle AU8'\rangle$
159: at the two--body order, one
160: observes that the leading of the spin--orbit terms are those usually denoted
161: as {\sl bbc} and {\sl cbb} in FHNC/SOC
162: theory \cite{pandharipande79}. In the calculations
163: of Ref.~\cite{sarsa03} these terms are responsible for
164: $\sim 80\%$ of the total
165: contribution from all the $C_{LS}^{(2)}$ terms in the density range
166: $(3/4)\rho_0 \leq \rho \leq (5/2)\rho_0$.
167: One can easily prove that exactly the same expressions
168: of the {\sl cbb} and {\sl bbc} terms are obtained by the following simplified
169: $\vec L\cdot\vec S$ correlation.
170:
171: \begin{eqnarray}
172: \tilde{F}_b(1,2) = \frac{1}{4i}f_b(r_{12})\left[ \vec r_{12} \cdot
173: \left(\vec\nabla_1\times\vec\sigma_1
174: -\vec\nabla_2\times\vec\sigma_2\right)\right] \ .
175: \label{lscorrback}
176: \end{eqnarray}
177: It is found that, at the two--body level of the FHNC/SOC theory,
178: $\tilde{F}_b$ leads to an energy which is only $\sim 10\%$ different from
179: that obtained with $F_b$.
180:
181: The important feature of the $\vec L\cdot\vec S$
182: correlation of Eq. (\ref{lscorrback})
183: for AFDMC calculations is that, similarly to the case of standard
184: backflow \cite{schmidt81}, it can be implemented in quantum Monte
185: Carlo simulations by substituting the plane wave orbitals of the Slater
186: function with the following spin--backflow ones
187: \begin{eqnarray}
188: \label{eq:bkf}
189: & &\exp(i\vec k \cdot \vec r_j)
190: \rightarrow \, \\
191: & & \phi_{\vec k}(j) = \exp\left(i\vec k \cdot \vec r_j
192: +{\beta\over 2} \sum_{k\ne j} f_b(r_{jk})
193: (\vec r_{jk} \times \vec k) \cdot \vec\sigma_j\right) \ , \nonumber
194: \end{eqnarray}
195: where $\beta$ is a spin--orbit strength parameter. For $\beta=1$ there
196: is a direct correspondence of $\phi_{\vec k}(j)$ with
197: $\tilde{F}_b$. This spin--backflow ansatz can also be used
198: in the GFMC simulations of small nuclear
199: systems \cite{pudliner97,pieper02,carlson03},
200: to include $\vec L\cdot\vec S$ correlations.
201:
202: We present and discuss, in the following, the results obtained in
203: AFDMC simulations of neutron matter energy with the $AU8'$ Hamiltonian
204: and the nodal surface of the spin--backflow Slater function. We will show
205: that these nodes serve to lower the AFDMC energy per particle of
206: neutron matter by a sizable amount, which
207: however is too small to solve the CVS and AFDMC discrepancy,
208: particularly at higher densities.
209:
210: \section{Calculation Details}
211: \label{sec.3}
212:
213: We describe the neutron system by the non relativistic Hamiltonian
214:
215: \begin{eqnarray}
216: H & = & T+V_2+V_3 \nonumber \\
217: & = &-\frac{\hbar^2}{2m}\sum_{j=1,N}\nabla^2_j +
218: \sum_{j<k}v_{jk} + \sum_{j<k<l} V_{jkl} \ ,
219: \label{ham_ham}
220: \end{eqnarray}
221:
222: \noindent
223: where $m$ is assumed to be the average of the neutron and proton masses
224: and $\hbar^2/m=41.47108$ MeV fm$^2$;
225: the two-body and three-body potentials, $v_{ij}$ and $V_{jkl}$,
226: are the Argonne $v'_8$ and the the Urbana IX potentials \cite{pudliner97}.
227: The three-body contributions to the
228: energy are large, particularly at high densities, and cannot be neglected
229: even in a survey calculation like ours.
230:
231: In neutron matter this interaction can be written in terms of four components:
232:
233: \begin{eqnarray}
234: v_{ij} &=& \sum_{p=1}^4 v_p(r_{ij}) {\cal O}^p (ij) \\
235: {\cal O}^p (ij) &=& 1, \vec\sigma_i\cdot\vec\sigma_j, S_{ij},
236: \vec L\cdot\vec S \ ,
237: \end{eqnarray}
238: where $S_{ij}$ and $\vec L\cdot \vec S$ are the usual tensor and spin--orbit
239: operators. The functions $v_p(r_{ij})$ can be found
240: on Ref.~\cite{pudliner97} and also on Ref.~\cite{sarsa03}.
241:
242: The Urbana-IX three-body interaction is given by the sum
243:
244: \begin{eqnarray}
245: V_{jkl} = V_{jkl}^{SI} + V_{jkl}^{SD},
246: \label{v3}
247: \end{eqnarray}
248: where $V_{jkl}^{SI}$ is a spin independent three--body short range part,
249: and the spin--dependent part, $V_{jkl}^{SD}$, in neutron matter,
250: reduces to a sum of terms containing only two--body spin operators,
251: with a form and strength that depends on the positions of three particles.
252: Their explicit expressions can be found in Ref.~\cite{sarsa03}.
253:
254: We have also considered an interaction obtained from $AU8'$ by dropping
255: the spin--orbit term, and, as in Ref.~\cite{sarsa03}, it is denoted as
256: $AU6'$.
257:
258: \begin{figure}[tb]
259: \includegraphics[height=\columnwidth,angle=-90]{backfig}
260: \caption[]{Neutron matter energy per particle in MeV as a function of
261: the strength $\beta$ of the spin--orbit correlation function $f_b$ at
262: $\rho=2\rho_0$ for 14 neutrons.
263: Solid line stands for the $AU6'$ interaction, dotted line for the
264: $AU8'$ potential.
265: }
266: \label{fig1}
267: \end{figure}
268:
269: The AFDMC \cite{schmidt99} method used in these calculations has been
270: previously described in detail \cite{sarsa03}.
271: It allows Monte Carlo simulations to be performed on a relatively large nuclear
272: system at the required accuracy, thanks to the introduction of
273: auxiliary fields, which
274: uncouple the spin--dependent interaction between particles by means of a
275: Hubbard--Stratonovich transformation. While the propagation of the particle
276: coordinates is done as in diffusion Monte Carlo, that of the
277: spin variables, after
278: having sampled the auxiliary fields, results in a rotation of each
279: particle's spinor.
280:
281: The guiding function used in this work is given
282: by a Jastrow product correlating an $N\times N$ Slater determinant
283:
284: \begin{eqnarray}
285: \langle R,S |\Psi_{JSB} \rangle = \prod_{j<k} f(r_{jk}) \Phi_B(R,S) ,
286: \label{jastrow}
287: \end{eqnarray}
288:
289: \noindent
290: where $R$ and $S$ denote the set of particle coordinates and of spinors
291: respectively. The space and spin orbitals in $\Phi_B(R,S)$ are given
292: in Eq. (\ref{eq:bkf}), and they operate on the spin states in the
293: following way
294:
295: \begin{eqnarray}
296: \phi_{\vec k}(i) |\uparrow\rangle &=& e^{i\vec k\cdot \vec r_i} \nonumber \\
297: &\times& \left[\left(\text{cosh}(A_{\vec k}(i))+\hat{A}_{\vec k}^z(i)
298: \text{sinh}(A_{\vec k}(i))\right) |\uparrow\rangle\right. \nonumber \\
299: &+&\left. \left(\hat{A}_{\vec k}^x(i)+i \hat{A}_{\vec k}^y(i)\right)
300: \text{sinh}(A_{\vec k}(i)) |\downarrow\rangle\right] \ ,
301: \end{eqnarray}
302:
303: \noindent
304: where
305:
306: \begin{eqnarray}
307: \vec A_{\vec k}(i) = \frac{\beta}{2}\sum_{j\ne i} f_b(r_{ij}) \vec r_{ij}
308: \times \vec k
309: \end{eqnarray}
310: and $\hat A_{\vec k}^{\alpha}(i)$, $\{\alpha=x,y,z\}$ are the components
311: of an unit vector in the direction of $\vec A_{\vec k}(i)$.
312: A similar expression can be obtained for the action over the spin down
313: single particle state.
314: We denote this function by JSB as opposed to JS which stands for a function
315: without explicit backflow correlations, {\sl i.~e.},
316: with simple plane waves in the spatial part of the orbitals in the
317: Slater determinant ($f_b(r)=0$). The AFDMC method described in
318: Ref.~\cite{sarsa03} needs only slight modifications to deal with
319: $\langle R,S | \Psi_{JSB}\rangle$ as guiding function. They refer mainly to
320: calculating the kinetic energy part of the local energy, and the
321: gradient of the guiding function in the drift of the walker.
322:
323: The Jastrow and spin--orbit correlation functions $f(r)$ and $f_b(r)$
324: have been taken as the first and fourth components respectively
325: of the FHNC/SOC correlation operator that minimizes
326: the energy per particle of neutron matter at the desired
327: density \cite{wiringa88}.
328:
329: \section{Results}
330: \label{sec.4}
331:
332: We have made our calculations within the full simulation box, and we
333: have taken into account the 26 neighboring boxes in the tabulation of
334: the correlations $f(r)$ and $f_b(r)$, and of the various
335: components $v_p(r)$ of the two--body potential, as described in
336: Ref.~\cite{sarsa03}. Our results are therefore already tail corrected
337: for a Hamiltonian with two--body force only.
338: Tail corrections for the three-body potential were not included.
339: However previous analyses \cite{sarsa03} have shown
340: that they are small for systems with 66 neutrons.
341:
342: \begin{table}
343: \caption{\label{tab:bf}
344: AFDMC energies per particle in MeV for the $AU6'$ and $AU8'$
345: interactions obtained for a system of 14 or 66 neutrons in a periodical box
346: with the guiding functions, $\Psi$, JS and JSB (with $\beta=1$), at $\rho_0$.
347: Error bars for the last digit are shown in parentheses.
348: }
349: \begin{ruledtabular}
350: \begin{tabular}{ccc}
351: $\Psi$ & $N=14$ & $N=66$ \\
352: \hline
353: JS(AU6') & $48.27(9)$ & $ 53.11(9)$ \\
354: JS(AU8') & $48.4(1) $ & $ 54.4(6) $ \\
355: JSB(AU8') & $46.8(1) $ & $ 52.9(2) $
356: \end{tabular}
357: \end{ruledtabular}
358: \end{table}
359:
360: Fig.~\ref{fig1} shows the AFDMC energies per particle of 14 neutrons
361: in a periodic simulation cell at $\rho=2\rho_0$ for the $AU6'$ and
362: $AU8'$ interactions, as a function of the strength parameter $\beta$.
363: One can see that, for the $AU8'$ interaction the energy minimum is around
364: $\beta=1$, consistent with the conjecture that $f_b(r)$ moves the nodes
365: in the optimal way when the spin--orbit potential is included
366: in the Hamiltonian. We have obtained a lowering of the energy of
367: 1.6 MeV
368: with respect to the case of the JS nodal surface, which corresponds to
369: $\beta=0$ in the figure. If we switch off the spin--orbit potential
370: by considering the $AU6'$ interaction, the minimum is found at
371: $\beta=0$, confirming that the spin--backflow nodes are energetically
372: advantageous only in the presence of the spin--orbit component of $v_{8'}$.
373:
374: AFDMC simulations for the $AU8'$ Hamiltonian with the JSB guiding function
375: have been also carried out for 66 neutrons. As in the case of the 14
376: neutron system, we have found a minimum of the energy at $\beta\sim 1$.
377: The result obtained is compared in Table~\ref{tab:bf}
378: with those at $\beta=0$ for both the $AU8'$ and the $AU6'$ interactions.
379: In spite of sizable
380: differences between the energies per particle of the 14 neutron
381: and the 66 neutron systems, the gain in energy $(E(JSB)-E(JS))/N$
382: is roughly independent on the number of particles in the box.
383: The large differences between
384: E(14) and E(66) are mainly due to the effect of three--body
385: interaction. It has been shown \cite{sarsa03} that the finite size effects on
386: E(66) are rather small.
387:
388: The dependence of the JSB energy on the density is reported in
389: Table~\ref{tab:density} for the 14 neutron system. There is a very
390: weak dependence of $\Delta E_{LS}$ on the density, in contrast with
391: VCS results.
392:
393: In order to make comparison with recent quantum Monte Carlo calculations
394: performed for 14 neutrons interacting via the $v_8'$
395: two--body potential by \textcite{carlson03},
396: we report in Table~\ref{tab:carlson} the corresponding AFDMC results.
397: The table displays the energy difference $\Delta E_{LS}$ between the energy
398: obtained with $v_8'$ and $v_6'$ two--body potential.
399:
400: \begin{table}
401: \caption{\label{tab:density}
402: AFDMC energies per particle in MeV for the $AU6'$ and $AU8'$
403: interactions obtained for a system of 14 neutrons in a periodical box
404: with the guiding functions, $\Psi$, JS and JSB (with $\beta=1$),
405: as a function of the density $\rho$.
406: Error bars for the last digit are shown in parentheses.
407: }
408: \begin{ruledtabular}
409: \begin{tabular}{cccc}
410: $\rho$ & $AU6'$ & $JS$ & $JSB$ \\
411: \hline
412: $\rho_0$ & $19.73(5)$ & $19.76(6)$ & $18.76(5)$ \\
413: $2\rho_0$ & $48.27(9) $ & $48.4(1) $ & $46.8(1)$
414: \end{tabular}
415: \end{ruledtabular}
416: \end{table}
417:
418: The calculations of Ref.~\cite{carlson03}, to which the results reported
419: in Table~\ref{tab:carlson} refer, have been performed with the $v_{8'}$
420: potential cut off at the edge of the box. Therefore, the values of
421: $\Delta E_{LS}$ extracted from there might not be completely comparable
422: with ours, since ours are already tail corrected and have been
423: obtained without introducing any discontinuity in the potential.
424: AFDMC seems to agree reasonably well with GFMC in the constrained
425: path approximation (GFMC--CP). The VCS estimate seems to be too large.
426:
427: \section{Conclusions}
428: \label{sec.5}
429:
430: \begin{table}
431: \caption{\label{tab:carlson}
432: Spin--orbit contribution to the energy per particle in MeV of neutron matter
433: at density $\rho_0$.
434: The constrained (CP) and unconstrained (UC) GFMC results, as well as the
435: VCS ones are taken from Ref.~\cite{carlson03}. The AFDMC results
436: obtained with the JS guiding function are taken from Ref.~\cite{sarsa03}.
437: Error bars for the last digit are shown in parentheses.
438: }
439: \begin{ruledtabular}
440: \begin{tabular}{cc}
441: method & $\Delta E_{LS}$ \\
442: \hline
443: GFMC--CP & -1.26(4) \\
444: GFMC--UC & -2.9(3) \\
445: AFDMC--JS & -0.14(6) \\
446: AFDMC--JSB & -1.2(1) \\
447: VCS & -3.8
448: \end{tabular}
449: \end{ruledtabular}
450: \end{table}
451:
452: In this paper we have proposed a new kind of space--spin
453: orbitals with a spin--backflow form,
454: which is particularly useful for taking into account
455: the spin--orbit interaction of nuclear systems in quantum Monte
456: Carlo simulations. An efficient parameterization of the
457: spin--backflow is obtained from the spin--orbit correlation of
458: FHNC/SOC theory. These spin-backflow orbitals can be conveniently used
459: also in other quantum Monte Carlo calculations, for instance
460: the Green's Function Monte Carlo simulations of small nucleon systems.
461: The nodal surface provided by this new guiding function
462: is able to decrease the energy about 5 per cent. This amount is not
463: sufficient to solve the spin--orbit discrepancy between the variational
464: chain summation and the AFDMC results, however the AFDMC results
465: are in good agreement
466: with constrained GFMC simulations.
467:
468: \begin{acknowledgments}
469: SAV wants to thanks the kind hospitality of SF and SISSA where most of
470: this work has been accomplished.
471: Portions of this work were supported by the Italian MIUR-National Research
472: Project 2001025498.
473: AS acknowledges the Spanish Ministerio
474: de Ciencia y Tecnologia for partial support under contract BMF2002-00200
475: \end{acknowledgments}
476:
477: \bibliography{bf}
478: \bibliographystyle{apsrev}
479:
480: \end{document}
481: