1: %%\documentclass{jpsj2}
2: %%\documentclass[letter]{jpsj2} %% for letters
3: %%\documentclass[shortnote]{jpsj2} %% for short notes
4: %%\documentclass[comment]{jpsj2} %% for comments
5: %%\documentclass[addenda]{jpsj2} %% for addenda
6: %%\documentclass[errata]{jpsj2} %% for errata
7: %%\documentclass[twocolumn]{jpsj2} %% two-column layout
8: %%\documentclass[seceq]{jpsj2} %% It makes equation numbers included within the section number (for regular paper only).
9: %%% The following is the list of packages loaded automatically into this class file.
10: %% amsmath.sty
11: %% amssymb.sty
12: %% graphicx.sty
13: %% overcite.sty
14: %
15: \documentclass[letter,twocolumn]{jpsj2} %% for letters
16: %\documentclass[letter]{jpsj2} %% for letters
17:
18: \usepackage{graphicx}
19:
20: \title{Interaction and Localization of
21: One-electron Orbitals in an Organic Molecule:\\
22: Fictitious Parameter Analysis for Multi-physics Simulations}
23:
24: \author{
25: Toshiya \textsc{Takami}\thanks{E-mail address: takami@cc.kyushu-u.ac.jp},
26: Jun \textsc{Maki}\thanks{E-mail address: maki-j@cc.kyushu-u.ac.jp},
27: Jun-ichi \textsc{Ooba},
28: Taizo \textsc{Kobayashi},
29: Rie \textsc{Nogita},
30: and Mutsumi \textsc{Aoyagi}}
31:
32: \inst{Computing and Communications Center, Kyushu University,
33: Hakozaki 6--10--1, Higashi-ku, Fukuoka 812-8581, Japan}
34:
35: \abst{We present a new methodology to analyze complicated
36: multi-physics simulations by introducing a fictitious parameter.
37: Using the method,
38: we study quantum mechanical aspects of an organic molecule in water.
39: The simulation is variationally constructed from
40: the {\it ab initio\/} molecular orbital method
41: and the classical statistical mechanics
42: with the fictitious parameter
43: representing the coupling strength between solute and solvent.
44: We obtain a number of one-electron orbital energies
45: of the solute molecule derived from the Hartree-Fock approximation,
46: and eigenvalue-statistical analysis developed in the study
47: of nonintegrable systems is applied to them.
48: Based on the results, we analyze localization properties
49: of the electronic wavefunctions under the influence of the solvent.
50: }
51:
52: \kword{eigenvalue statistics, avoided crossings, quantum chaos,
53: multi-physics simulation, variational construction of coupled simulation,
54: electronic states of bio-molecules, one-electron orbital energy,
55: solute-solvent system, RISM/SCF}
56:
57: \begin{document}
58: \maketitle
59:
60: The multi-physics simulation is one of
61: %the powerful methods to construct complicated simulations
62: the powerful methods to construct complex simulations
63: representing realistic systems.
64: Such a calculation is often constructed
65: by combining multiple theories with different scales of description
66: based on different approximations,
67: e.g., climate simulations by fluid dynamics of the atmosphere
68: surrounded by various external heat and humidity sources\cite{AGCM},
69: quantum materials bound
70: to classical large degrees of freedom\cite{Ogata,Sato}, etc.
71: Since reality and accuracy are required,
72: those simulations have been larger and more complicated year by year.
73: Then, guaranteeing their convergence and reliability
74: becomes a more difficult problem.
75:
76: In this letter, we study electronic states of a peptide in water
77: obtained by a multi-physics simulation which consists of
78: {\it ab initio\/} molecular orbital (MO) methods
79: and classical statistical mechanics.
80: The quantum chemical nature of large molecules in a solvent
81: is also one of the most popular topics in physics and chemistry.
82: In the present calculation,
83: the simulation is constructed from a combination of
84: the self-consistent field (SCF) calculation
85: under the Hartree-Fock (HF) approximation\cite{SCF}
86: and the statistical mechanics calculation by three dimensional version
87: of the reference interaction site model (3D-RISM)\cite{rism}.
88: Although the coupled simulation of RISM and SCF itself
89: was developed in 1990's\cite{Ten-no}
90: and has been applied to several molecular systems,
91: it attracts public attention again due to the recent developments
92: of the distributed computing environments\cite{grid}.
93: Since the 3D-RISM/SCF coupled simulation\cite{Sato-Kovalenko}
94: heterogeneously couples
95: the different theories both of which require large computational resources,
96: the execution of the simulation for large solute molecules
97: is difficult not only in an arrangement of the computer resources
98: but also in theoretical aspects of computation
99: such as stability of the convergence and reliability of the simulation.
100:
101: The aim of this letter is two fold:
102: one is to establish the methodology to ensure a proper execution
103: of the complicated simulation,
104: and the other is to introduce the new analysis
105: for a electronic state of molecules using eigenvalue statistics.\cite{Haake}
106: The target of our analysis is one-electron orbital
107: energies $\varepsilon_j$ and its wavefunction $|\phi_j\rangle$
108: obtained as spatially independent wavefunctions
109: under HF approximation.
110: Although $\varepsilon_j$ and $|\phi_j\rangle$ are
111: introduced through the variational minimization of the total energy,
112: they still retain physical reality without Koopman's theorem
113: if we consider energies and wavefunctions of
114: quasiparticles and quasiholes\cite{HF-statistics}.
115: Our key to accomplish these aims
116: is an introduction of a fictitious parameter
117: which can be used not only to investigate the stability
118: and continuity of the results
119: but also to perform sophisticated analyses
120: \cite{GRMN1990,TTcurvature,TTcharacter,ZD93c,ZD93a,MS2005}
121: based on the eigenvalue statistics.
122:
123: % Variational Construction of Coupled Simulation
124: At first, we construct 3D-RISM/SCF coupled simulation
125: according to the standard variational way used in 1D-RISM/MCSCF
126: \cite{sato-hirata-kato}.
127: In the present calculation,
128: the electronic states of the peptide are obtained
129: by SCF\cite{SCF} under the restricted HF approximation (RHF),
130: i.e., all electrons are assumed to configure closed orbitals.
131: The variational functional is chosen as
132: a total Helmholtz free energy\cite{rism,sato-hirata-kato}
133: \begin{equation}
134: \label{eqn:free-energy}
135: A(\lambda)=E_{\rm solute}(\{|\phi_j\rangle\})
136: +\Delta\mu(\{|\phi_j\rangle\}; \lambda),
137: \end{equation}
138: where $E_{\rm solute}$ is an energy of the solute molecule
139: and $\Delta\mu$ is the excess chemical potential
140: of the solute obtained in 3D-RISM with the coupling strength $\lambda$.
141: Statistical correlation functions of the solvent
142: are calculated under the charge distribution of the solute
143: represented by a set of the orbitals $\{|\phi_j\rangle\}$.
144: The variation of the first term in {\it r.h.s\/} of (\ref{eqn:free-energy})
145: with respect to $|\phi_j\rangle$ gives usual SCF equations for the MO theory
146: while the solvation effect is introduced through the second term.
147: %The standard strategy using Lagrange multipliers
148: %for variation with constraints gives conditions to be satisfied.
149: Variation with constraints using Lagrange multipliers
150: gives conditions to be satisfied.
151: Thus, the total free energy (\ref{eqn:free-energy})
152: of the coupled system is stabilized by solving these conditions.
153:
154: One of the merits of the variational construction of complicated simulations
155: is that a set of equations to be solved is given by a simple calculus
156: once we define a functional of the whole system.
157: In the present case, we only have to define
158: the solute-solvent interaction potential
159: \begin{equation}
160: u_\gamma({\bf r};\lambda)=\lambda\sum_a\left[
161: \frac{q_aq_\gamma}{r_a}
162: +4\epsilon_{a\gamma}\left\{
163: \left(\frac{\sigma_{a\gamma}}{r_a}\right)^{12}
164: -\left(\frac{\sigma_{a\gamma}}{r_a}\right)^{6}
165: \right\}
166: \right]
167: \end{equation}
168: where $q_a$ is a point charge on a solute atom $a$,
169: $q_\gamma$ is a partial charge on a solvent site $\gamma$,
170: $r_a\equiv|{\bf R}_a-{\bf r}|$ is a distance between
171: the solute atom $a$ and the solvent site $\gamma$,
172: $\epsilon_{a\gamma}$ and $\sigma_{a\gamma}$ is given
173: by the standard parameter sets of Lennard-Jones potentials.
174: We adopt the method of Mulliken population
175: among several schemes to define $q_a$,
176: \begin{equation}
177: q_a=Z_a-\sum_{j=1}^{n/2}\sum_{\nu\in a}\sum_{\nu'}\left[
178: C^*_{\nu j}C_{\nu'j}\langle\chi_\nu|\chi_{\nu'}\rangle+\hbox{c.c.}
179: \right]
180: \end{equation}
181: where $C_{\nu j}$ are coefficients in a basis-set expansion of $|\phi_j\rangle$
182: \begin{equation}
183: \label{eqn:basis}
184: |\phi_j\rangle=\sum_\nu|\chi_\nu\rangle C_{\nu j}
185: \end{equation}
186: by the atomic basis function $|\chi_\nu\rangle$.
187: This representation reduces computational costs
188: of the whole simulation,
189: and can be used even in large-scale simulations for proteins.
190: Although the reduction of the cost
191: is often a result of compensation in accuracy,
192: it can be retained by combining other schemes of
193: the potential representation \cite{calc-3drism}.
194:
195: After the variation with respect to $C_{\nu j}$,
196: the solvated Fock matrix element $F_{\mu\nu}$ between $\mu$ and $\nu$
197: is given by
198: \begin{equation}
199: \label{eqn:fock}
200: % F_{\mu\nu}(\lambda)=F^{(0)}_{\mu\nu}-\lambda V_{\mu\nu}.
201: F_{\mu\nu}(\lambda)=F^{(0)}_{\mu\nu}-\lambda V_{\mu\nu}S_{\mu\nu},
202: \end{equation}
203: where the overlap integral $S_{\mu\nu}$ is defined by
204: $\langle\chi_\mu|\chi_\nu\rangle$.
205: $V_{\mu\nu}$ represents the environmental potential term
206: induced by the partial charge distribution of water molecules,
207: \begin{equation}
208: V_{\mu\nu}
209: =\frac{1}{2}\int\left[
210: \frac{1}{|{\bf R}_\mu-{\bf r}|}+\frac{1}{|{\bf R}_\nu-{\bf r}|}
211: \right]Q({\bf r};\lambda)\ d{\bf r},
212: \end{equation}
213: where ${\bf R}_\mu$ and ${\bf R}_\nu$ represent atomic centers
214: of the basis functions $|\chi_\mu\rangle$ and $|\chi_\nu\rangle$,
215: respectively, and
216: \begin{equation}
217: Q({\bf r};\lambda)
218: \equiv\sum_\gamma\rho_\gamma q_\gamma g_\gamma({\bf r};\lambda)
219: \end{equation}
220: is the partial charge distribution
221: calculated by summing partial charges given by the distribution
222: function $g_\gamma({\bf r};\lambda)$ over all the solvent sites $\gamma$.
223: We can numerically solve the Roothaan equation\cite{SCF}
224: \begin{equation}
225: \label{eqn:roothaan-hall}
226: {\bf F}(\lambda){\bf C}={\bf S}{\bf C}\varepsilon
227: \end{equation}
228: to obtain coefficients $C_{\mu\nu}$ of the basis-set expansion
229: as elements of the matrix ${\bf C}$ and the orbital
230: energies $\varepsilon_j$ as diagonal elements of $\varepsilon$.
231: %where the element $S_{\mu\nu}$ of ${\bf S}$ is an overlap integral
232: %$\langle\chi_\mu|\chi_\nu\rangle$.
233:
234: \begin{figure}[tb]
235: \begin{center}
236: \includegraphics[scale=0.5]{density.eps}
237: \caption{The integrated density of one-electron orbitals
238: of a met-enkephalin. Two curves (vacuum and solvated) are plotted
239: around the area near the orbitals of HOMO and LUMO.}
240: \label{fig:staircase}
241: \end{center}
242: \end{figure}
243:
244: The actual procedure to execute the coupled simulation is the following.
245: The SCF calculation is performed from an appropriate
246: initial distribution of the solvent.
247: For the given distribution of partial charges of the solvent,
248: the electron density is obtained after a convergence of the SCF.
249: Then, 3D-RISM calculation to obtain $g_\gamma({\bf r})$
250: is carried out using the charge distribution of the solute molecule.
251: Converged $g_\gamma({\bf r})$ determines
252: the environmental term $\lambda V_{\mu\nu}$ in the solvated-Fock matrix
253: by which the SCF calculation is performed again.
254: This procedure repeats until the whole distribution is converged.
255: From the converged simulation,
256: we obtain the solute electronic states
257: and the solvent distribution functions.
258:
259: Before entering the analysis using $\lambda$,
260: we show results of a methionine-enkephalin in two specific cases.
261: This peptide consists of a sequence of 5 residues,
262: Tyrosine, Glycine, Glycine, Phenylalanine, and Methionine,
263: with 75 atoms in total.
264: From the standard SCF calculation, which corresponds to $\lambda=0$,
265: the orbital energies of the methionine-enkephalin are obtained in vacuum,
266: while the fully-solvated case $\lambda=1$ gives
267: the orbital energies in aqueous solution.
268: For the electronic ground state, we assume that
269: 304 electrons occupy 152 independent spatial orbitals under RHF.
270: Among several band-like structures in the orbital energies,
271: we concentrate the analysis of the band just below the
272: highest occupied molecular orbital (HOMO) and the lowest
273: unoccupied molecular orbital (LUMO),
274: which contains 108 orbitals from 45th to 152th ones.
275: Orbitals strongly localized on 1s atomic orbital of C, N, O,
276: and 1s, 2s, 2p orbitals of S (known as core orbitals)
277: can be found in the region far from the band we analyze.
278: Since they exhibit almost no interaction with other orbitals,
279: we do not consider these orbitals in the present analyses.
280: The integrated density of states, $D(E)=\sum\Theta(E-E_j)$,
281: is shown for the band around HOMO and LUMO in Fig.~\ref{fig:staircase}.
282: In this case, the 152nd orbital is HOMO and the 153rd is LUMO.
283:
284: \begin{figure*}[tb]
285: \begin{center}
286: \includegraphics[scale=0.5]{integ-spc-vacuum.eps}
287: \includegraphics[scale=0.5]{integ-spc-solvated.eps}
288: \caption{The integrated distribution of the nearest-neighbor spacing of
289: the orbital energies of a met-enkephalin
290: (a) in vacuum and (b) in water.
291: We also show histograms for the nearest-neighbor spacing in the insets,
292: where curves show the Brody distribution (see text)
293: with (a) $q=0.52$ and (b) $q=0.33$.}
294: \label{fig:spacing}
295: \end{center}
296: \end{figure*}
297:
298: % Eigenvalue Statistics of One-electron Orbital Energies
299: Despite that these orbital energies calculated from
300: the generalized eigen-equation (\ref{eqn:roothaan-hall})
301: are not usual eigenvalues defined in an autonomous system,
302: they have almost the same property as the usual eigenvalues and eigenstates,
303: i.e., the orbital energies are real, and the orbital states are
304: orthogonal to the state with a different orbital energy.
305: Moreover, repulsive interactions between them
306: have the same origin as those of the eigenvalues\cite{HF-statistics}.
307: Thus, we can perform the eigenvalue-statistical analysis
308: of these values after a certain unfolding procedure.
309: In Figs. \ref{fig:spacing} (a) and (b),
310: we show the integrated nearest-neighbor spacing distribution
311: obtained after the standard procedure of unfolding the spectrum\cite{Haake}.
312: It is known that the fully chaotic system
313: shows the Wigner distribution while the regular system shows
314: the Poisson distribution.
315: For an intermediate case, we can compare the distribution
316: to a Brody distribution.
317: In figs. \ref{fig:spacing} (a) and (b),
318: we also plot curves for the Brody distribution
319: which is obtained by fitting the distribution
320: to the integrated Brody distribution,
321: \begin{equation}
322: P_B(s)=1-\exp(-\alpha s^{q+1})
323: \end{equation}
324: where $q$ represents the Brody parameter, and
325: \begin{equation}
326: \alpha\equiv\left[\Gamma\left(\frac{q+2}{q+1}\right)\right]^{q+1}
327: \end{equation}
328: is obtained by normalization condition.
329: The values of the Brody parameters obtained are
330: $q=0.52$ in vacuum and $q=0.33$ in water.
331: The smaller value of $q$ reflects the smaller interaction
332: between orbitals.
333:
334: In order to study further quantum aspects of the solvation effect,
335: we enter the detailed analysis of energies and wavefunctions
336: to the fictitious parameter $\lambda$.
337: For chaotic quantum systems, the Brody parameter $q$ is
338: usually related to the area of stochastic region in phase space\cite{Brody}.
339: However, we do not intend to study any stochastic dynamics
340: behind the electronic states.
341: Instead, we concentrate on the interaction between the orbitals
342: when we slightly vary the parameter $\lambda$,
343: i.e., avoided crossings\cite{TTavoided} between the orbital energies.
344:
345: % One-electron Energies Parameterized by Coupling Strength
346: The Fock matrix (\ref{eqn:fock}) depends on the fictitious
347: parameter $\lambda$ nonlinearly
348: since $\lambda$ is also included in $V_{\mu\nu}$,
349: which is different from linearly perturbed systems.
350: Even then, we can plot the orbital energies to the parameter as usual
351: if we perform SCF calculations from $\lambda=0$ (vacuum)
352: to $\lambda=1$ (fully solvated).
353: Figure \ref{fig:avoided} is a plot of the orbital energies
354: with respect to the parameter $\lambda$,
355: where we chose $\lambda^2$ instead of $\lambda$ as the abscissa
356: because of the nonlinear dependence on $\lambda$.
357: Various sizes of avoided crossing are found in the figure,
358: and there seems to be a small number of special orbitals
359: which have almost no interaction with others.
360: Such a separation of orbital states into groups without interaction
361: each other can lead to smaller values of the Brody parameter.
362:
363: % Localization
364: The separation of eigenfunctions is often seen in the intermediate
365: semiclassical limit of quantum chaotic systems,
366: i.e., statistical properties in a range around a finite energy.
367: A stadium billiard system has a series of eigenfunctions
368: corresponding to a bouncing-ball orbit
369: by which the nearest-neighbor spacing deviates
370: from the ideal case of random matrix systems.
371: The eigenstates corresponding to the bouncing-ball orbit
372: is extremely localized in momentum space\cite{Localize1,Localize2}.
373: In the present results,
374: such localization can be observed in the coordinate space.
375: Actually, the core orbitals, which we have removed from the Brody analysis,
376: are strongly localized on a certain atomic basis function.
377: Although the orbitals in the region we study here do not localize so strong,
378: it is reasonable to analyze the property of orbital states
379: with respect to the localization on the atomic basis functions.
380:
381: \begin{figure}[b]
382: \begin{center}
383: \includegraphics[scale=0.57]{avoided.eps}
384: \caption{One-electron orbital energies near the highest occupied
385: molecular orbital. The abscissa is $\lambda^2$, the square of
386: the coupling strength between the solute and solvent.}
387: \label{fig:avoided}
388: \end{center}
389: \end{figure}
390:
391: \begin{figure}[tb]
392: \begin{center}
393: \includegraphics[scale=0.5]{localize.eps}
394: \caption{Distribution $P(W_j)$ of the delocalization index $W_j$
395: from 45th to 152nd orbital states.}
396: \label{fig:delocalize}
397: \end{center}
398: \end{figure}
399:
400: In order to analyze the extent of localization,
401: we define the degree of delocalization of $j$-th orbital state by
402: \begin{equation}
403: \label{eqn:delocalize}
404: W_j=\left|\displaystyle\sum_\nu\left|C_{\nu j}\right|\right|^2
405: \left/\displaystyle\sum_\nu\left|C_{\nu j}\right|^2\right.
406: \end{equation}
407: through the basis-set expansion coefficient $C_{\nu j}$ in (\ref{eqn:basis}).
408: This is a representation-dependent quantity, and
409: we define $W_j$ on the standard atom-centered basis
410: set $\{|\chi_\nu\rangle\}$ in the present work.
411: Although $\{|\chi_\nu\rangle\}$ is
412: not orthogonal each other and we cannot relate the coefficients
413: $C_{\nu j}$ directly to the extent of the spatial distribution,
414: the value of (\ref{eqn:delocalize}),
415: the approximate number of basis functions contained,
416: still represents the relative strength of the delocalization.
417: In Fig. \ref{fig:delocalize},
418: we show the distribution of (\ref{eqn:delocalize}).
419: It can be seen that the distribution has two peaks
420: over the whole values of $\lambda$
421: while the places of the peaks slightly change.
422: That is, the orbitals are divided into two groups:
423: relatively localized orbitals ($0<W_j<60$) and
424: delocalized orbitals ($60<W_j<130$).
425: This separation of orbitals is considered to be the reason why
426: the Brody parameter exhibits the intermediate values.
427:
428: % Summary and Discussion
429: Before summarizing this work,
430: we give some comments for future studies.
431: Since our analysis is based on the parameter variation\cite{TTslope},
432: more sophisticated analysis such as the level-curvature analysis
433: \cite{GRMN1990,TTcurvature,ZD93c} and
434: distribution of gap sizes of avoided crossings\cite{ZD93a,MS2005}
435: can be performed.
436: Since these properties are known to be sensitive for intermediate
437: dynamical systems,
438: it will be more appropriate for analyzing molecular systems.
439: By the use of these analysis,
440: it is desirable to investigate quantum aspects of large molecules in water.
441: When we investigate large systems such as proteins or nucleic acids,
442: more effective methods must be needed.
443: One of such methods is the fragment MO calculation\cite{FMO}
444: by which the orbital energies can be obtained with small errors\cite{FMO-MO}.
445:
446: Another interesting problem is whether the regularity of the
447: quantum many-body system is influenced by solvation effects or not,
448: while in the present work
449: we have concentrated on the localization of molecular orbitals.
450: Although we could not give a clear answer within our Brody analysis,
451: such analysis on the regularity will be performed
452: through the detailed analysis of nodal patterns, Husimi representation,
453: etc. of the orbital wavefunctions with respect to the fictitious parameter.
454: Introducing the parameter
455: enables us to assign each solvated orbital
456: in terms of the orbitals in vacuum.
457: This procedure will successfully be done by
458: connecting characteristic features of the orbital over avoided crossings
459: since it is possible even in strongly chaotic systems\cite{TTcharacter}
460: if many pairwise avoided crossings are found.
461: From a technical point of view,
462: the continuous parameter is also important as a computational technique.
463: In our case, the stability of the result is numerically supported
464: by varying the parameter slightly,
465: since physically meaningful quantities can only be obtained
466: as stable results under small perturbations.
467: In order to obtain stable and significant results
468: from complicated realistic simulations,
469: the fictitious parameter approach will be one of the powerful methods.
470:
471: % Summary and Conclusion
472: In summary, we have studied statistics of the orbital energies
473: and localization of orbital states
474: through the variation of the fictitious parameter.
475: This is our first report on the eigenvalue-statistical analysis
476: of the orbitals as a new approach to quantum many-body systems.
477: We hope that these new methods will be used to study
478: quantum aspects of large bio-molecules
479: as well as to complete large-scale realistic simulations.
480:
481: \section*{Acknowledgments}
482:
483: We would like to thank Prof. F. Hirata
484: for allowing us to use the 3D-RISM program developed in his group.
485: We also thank Dr. S. Ho, Mr. S. Kubo and Mr. K. Fukuda for
486: their collaborative development on the 3D-RISM/SCF coupled simulation.
487: One of the authors (T.T.) is grateful to Prof. S. Sawada, Prof. M. Toda,
488: Prof. A. Shudo, Dr. M. Sano and Dr. A. Tanaka
489: for extensive discussion on chaos and dynamical systems.
490: This work is supported by the Ministry of Education, Sports, Culture,
491: Science and Technology (MEXT) through the Science-grid NAREGI Program
492: under the Development and Application of Advanced High-performance
493: Supercomputer Project.
494:
495: \begin{thebibliography}{99} %% The number "99" means that this list has more than nine items.
496: \bibitem{AGCM}
497: M.-I. Lee, I.-S. Kang, B. E. Mapes: {\it J. Meteor. Soc. Japan\/} {\bf 81}
498: (2003) 963.
499: \bibitem{Ogata}
500: S. Ogata: {\it Phys. Rev. B\/} {\bf 72} (2005) 045348.
501: \bibitem{Sato}
502: T. Sato: {\it J. Phys: Conference Ser.\/} {\bf 16} (2005) 310.
503: \bibitem{SCF}
504: C. C. J. Roothaan: {\it Rev. Mod. Phys.\/} {\bf 23} (1951) 69.
505: \bibitem{rism}
506: F. Hirata, ed.: {\it Molecular Theory of Solvation\/}
507: (Kluwer Academic Pub. 2003).
508: \bibitem{Ten-no}
509: S. Ten-no, F. Hirata and S. Kato:
510: {\it Chem. Phys. Lett.\/} {\bf214} (1993) 391;
511: S. Ten-no, F. Hirata and S. Kato:
512: {\it J. Chem. Phys.\/} {\bf100} (1994) 7443.
513: \bibitem{grid}
514: S. Matsuoka, S. Shimojo, M. Aoyagi, S. Sekiguchi, H. Usami, and K. Miura:
515: {\it the Proceedings of the IEEE\/} {\textbf 93} (2005) 522.
516: \bibitem{Sato-Kovalenko}
517: H. Sato, A. Kovalenko and F. Hirata:
518: {\it J. Chem. Phys.\/} {\bf112} (2000) 9463.
519: \bibitem{Haake}
520: F. Haake: {\it Quantum Signatures of Chaos\/} (2nd ed.) (Springer, 2001).
521: \bibitem{HF-statistics}
522: S. Levit and D. Orgad: {\it Phys. Rev. B\/} {\bf 60} (1999) 5549.
523: \bibitem{GRMN1990}
524: P. Gaspard, S.A. Rice, H.J. Mikeska, and K. Nakamura:
525: {\it Phys. Rev. E\/} {\bf 42} (1990) 4015.
526: \bibitem{TTcurvature}
527: T. Takami and H. Hasegawa: {\it Phys. Rev. Lett.\/} {\bf 68} (1992) 419.
528: \bibitem{TTcharacter}
529: T. Takami: {\it Phys. Rev. Lett.\/} {\bf 68} (1992) 3371.
530: \bibitem{ZD93c}
531: J. Zakrzewski and D. Delande: {\it Phys. Rev. E\/} {\bf 47} (1993) 1650.
532: \bibitem{ZD93a}
533: J. Zakrzewski and D. Delande: {\it Phys. Rev. E\/} {\bf 47} (1993) 1665.
534: \bibitem{MS2005}
535: M. Machida and K. Saito: {\it Phys. Rev. E\/} {\bf 72} (2005) 056206.
536: \bibitem{sato-hirata-kato}
537: H. Sato, F. Hirata and S. Kato: {\it J. Chem. Phys.\/} {\bf 105} (1996) 1546.
538: \bibitem{calc-3drism}
539: N. Yoshida and F. Hirata: {\it J. Comp. Chem.\/} {\bf 27} (2006) 453.
540: \bibitem{Brody}
541: T. A. Brody, J. Flores, J. B. French, P. A. Mello, A. Pandey,
542: and S. S. M. Wong: {\it Rev. Mod. Phys.\/} {\bf 53} (1981) 385.
543: \bibitem{TTavoided}
544: T. Takami: {\it Phys. Rev. E\/} {\bf 52} (1995) 2434.
545: \bibitem{Localize1}
546: W.~E.~Bies, L.~Kaplan, M.~R.~Haggerty and E.~J.~Heller:
547: {\it Phys. Rev. E\/} {\bf 63} (2001) 066214.
548: \bibitem{Localize2}
549: T.~Timberlake, F.~Petruzielo and L.~E.~Reichl:
550: {\it Phys. Rev. E\/} {\bf 72} (2005) 016208.
551: \bibitem{TTslope}
552: T. Takami: {\it Prog. Theor. Phys. Suppl.} {\bf 116} (1994) 303.
553: \bibitem{FMO}
554: K. Kitaura, E. Ikeo, T. Asada, T. Nakano and M. Uebayasi:
555: {\it Chem. Phys. Lett.\/} {\bf 313} (1999) 701.
556: \bibitem{FMO-MO}
557: Y. Inadomi, T. Nakano, K. Kitaura and U. Nagashima:
558: {\it Chem. Phys. Lett.\/} {\bf 364} (2002) 139.
559: \end{thebibliography}
560:
561: \end{document}
562: