1: \documentclass[aps,groupedaddress,showpacs,twocolumn,floatfix]{revtex4}
2: \usepackage{amsmath}
3: \usepackage{amssymb}
4: \usepackage{graphicx}
5: \usepackage{epsfig}
6:
7: \begin{document}
8:
9: \title{Frozen local hole approximation}
10: \author{Elke Pahl}
11: \affiliation{Max-Planck-Institute for the Physics
12: of Complex Systems, N{\"o}thnitzer Stra{\ss}e 38, 01187 Dresden, Germany}
13: \author{Uwe Birkenheuer}
14: \affiliation{Max-Planck-Institute for the Physics
15: of Complex Systems, N{\"o}thnitzer Stra{\ss}e 38, 01187 Dresden, Germany}
16: %\author{Peter Fulde}
17: %\affiliation{Max Planck Institute for the Physics
18: %of Complex Systems, N{\"o}thnitzer Stra{\ss}e 38, 01187 Dresden, Germany}
19:
20: \date{\today}
21:
22: \begin{abstract}
23: The frozen local hole approximation (FLHA) is an adiabatic approximation which
24: is aimed to simplify the correlation calculations of valence and conduction
25: bands of solids and polymers or, more generally, of the ionization potentials
26: and electron affinities of any large systems. Within this approximation
27: correlated {\it local} hole states (CLHSs) are explicitely generated by
28: correlating local Hartree-Fock (HF) hole states, i.~e.~($N$-1)-particle
29: determinants in which the electron has been removed from a {\it local}
30: occupied orbital. The hole orbital and its occupancy is kept frozen during
31: these correlation calculations, implying a rather stringent configuration
32: selection. Effective Hamilton matrix elements are then evaluated with the
33: above CLHSs; diagonalization finally yields the desired correlation
34: corrections for the cationic hole states. We compare and analyze the results
35: of the FLHA with the results of a full MRCI(SD) (multi-reference configuration
36: interaction with single and double excitations) calculation for two prototype
37: model systems, (H$_2$)$_n$ ladders and H--(Be)$_n$--H chains. Excellent
38: numerical agreement between the two approaches is found. Comparing the FLHA
39: with a full correlation treatment in the framework of quasi-degenerate
40: variational perturbation theory reveals that the leading contributions in the
41: two approaches are identical. In the same way it could be shown that replacing
42: a correlation calculation around a frozen {\it local} hole is in fact
43: equivalent, up to first order, to perform a much less demanding SCF
44: (self-consistent field) calculation around the hole to relax all other
45: occupied orbitals. Thus, both, the FLHA and the above SCF approximation, are
46: well-justified and provide a very promising and efficient alternative to fully
47: correlated wavefunction-based treatments of the valence and conduction bands
48: in extended systems.
49:
50:
51: \end{abstract}
52:
53: \pacs{71.10.-w,71.15.-m,71.20.-b}
54: % 71.10.-w Theories and models of many-electron systems
55: % 71.15.-m Methods of electronic structure calculations
56: % 71.20.-b Electron density of states and band structure of crystalline solids
57: \maketitle
58:
59: \section{Introduction}
60:
61: In the last decade, increasing interest in {\it wavefunction}-based
62: correlation methods for excited states in extended systems can be observed
63: \cite{GSF93,GSF97,GSB99,AFS00,AI01,A02,HVOB02,HPGB03,BB04,BBAF05}. This is
64: because one can rely on the very sophisticated numerical methods for molecules
65: which are well-established in quantum chemistry. These methods are
66: conceptually very clear and allow for systematic improvement: based on
67: self-consistent field calculations, electron correlation can be included
68: successively by considering determinants of increasing excitation order in
69: configuration interaction (CI) or related procedures. They yield approximate
70: correlated wavefunctions from which all quantities of interest can be derived.
71:
72: Of course, wavefunction-based methods are computationally very demanding. In
73: order to arrive at manageable schemes for the description of solids or other
74: extended systems, local orbitals have to be introduced and the local character
75: of electron correlation must rigorously be exploited. Among these local
76: approaches \cite{P02} one can distinguish local MP2 (M{\o}ller-Plesset
77: perturbation theory up to 2$^{\rm nd}$ order) and coupled electron pair or
78: coupled cluster schemes \cite{SP87,FS95,HW96,SHW99,SB96,AKS01,PBCC05}, methods
79: which are based on a Green's function formalism \cite{A02,BBAF05} and
80: techniques which use local Hamiltonian matrix elements
81: \cite{GSF93,GSF97,AFS00,BB04} or other partitioning concepts
82: \cite{S92-1,S92-2,S92-3,PFS95,F02,WB04,ACCE03,KIANU99,FK04,MKAS05,BME05}.
83: Whereas the first-mentioned methods work in infinite systems and aim to
84: truncate the excitation space by an appropriate configuration selection
85: scheme, the effective Hamiltonian approach as well as the other partitioning
86: schemes assemble the correlation effects from finite subsystems (embedded
87: fragments) which are finally transfered to the infinite system. Most of these
88: methods focus on ground state properties, only the Green's function and local
89: Hamiltonian approaches are explicitely designed for excited states. In
90: particular, the local Hamiltonian approach has been used in the past quite
91: successfully to describe the correlation effects on the band structure of
92: covalent solids \cite{GSF93,GSF97,AFS00} and polymers \cite{BB04}. Yet, this
93: approach is quite involved and a more simpler, approximative way of computing
94: the local Hamiltonian has been proposed recently by our lab \cite{BFSpp}.
95:
96: In this work we want to analyze the guiding approximations introduced in that
97: recent approach, and therefore address the question to which extent it is
98: possible to simplify the correlation calculations by focusing on ``frozen''
99: local hole configurations although, in reality, the electron hole is usually
100: delocalized over the entire system. Intuitively, this frozen local hole
101: approximation, is guided by the idea that the shape of the correlation
102: hole which is carried along by a traveling electron is essentially invariant
103: and is thus following the electron hole adiabatically.
104:
105: A related approach is the frozen core hole approximation \cite{RS03} for the
106: calculation of core-hole states where the core hole orbital is frozen and all
107: other orbitals are allowed to relax during an Hartree-Fock calculation. Such a
108: local view on core holes in bulk materials is also adopted in other
109: wavefunction-based investigations on excited core hole states
110: \cite{GBNB97,VHBB02}. While it seems quite natural to describe core holes in
111: that way, although in solids even core orbitals are Bloch waves, it is much
112: less obvious that a frozen local hole treatment should also be possible for
113: delocalized valence hole states.
114:
115: In the following, we will outline the theoretical background of the frozen
116: local hole approximation (for valence holes) and then present and analyze the
117: numerical results obtained for two model systems, namely (H$_2$)$_n$ ladders
118: and H--(Be)$_n$--H chains. These two examples have been chosen to represent
119: the two cases of prevailing van der Waals and predominantly covalent
120: binding. Finally, an analytical, perturbative analysis of the frozen local
121: hole approximation is given.
122:
123: \section{Theoretical background}
124:
125: Starting point for our correlation calculations is the Hartree-Fock (HF)
126: ground state Slater determinant $|\Phi_0^N \rangle $ and energy $E_0^N$ of the
127: $N$-particle system. The canonical HF orbitals $|\nu \rangle $ with orbital
128: energies $\epsilon _{\nu} $ are divided into the $N/2$ energetically
129: lowest-lying orbitals which are occupied in $|\Phi_0^N \rangle $ and the
130: unoccupied, virtual orbitals. Furthermore, the occupied orbitals are
131: subdivided into core and valence orbitals depending on whether they are kept
132: frozen or are active during the subsequent correlation calculation. Within
133: these orbital groups we will switch between delocalized canonical orbitals and
134: localized orbitals, $\{ |\nu \rangle \}$ and $\{ |a\rangle \}$, using unitary
135: transformations, such as
136: \begin{equation}
137: |a\rangle = \sum_\nu | \nu \rangle U_{\nu a}
138: \end{equation}
139: for the occupied orbitals. In order to find the correlated cationic
140: ($N$-1)-particle states of the system in mind, one first constructs the
141: so-called reference states $|\Phi_i^{N-1} \rangle$ of the system by removing
142: one electron out of one of the valence orbitals $|i\rangle $: $|\Phi_i^{N-1}
143: \rangle = \hat{c}_i |\Phi_0^N \rangle$. The so generated {\em canonical hole}
144: states $|\Phi_{\nu}^{N-1} \rangle$ and {\em local hole} states $|\Phi_a^{N-1}
145: \rangle$ are connected via
146: \begin{equation}
147: \label{trafo} |\Phi_{\nu}^{N-1} \rangle = \sum_a |\Phi_a^{N-1} \rangle
148: (U^{-1})_{a\nu }^\ast =\sum_a |\Phi_a^{N-1} \rangle U_{\nu a} \quad
149: \end{equation}
150: Since we are only dealing with cationic states here we will omit the upper
151: index ($N$-1) in the following. The {\it canonical} hole states are the eigenstates
152: of the CI matrix in the reference space. They describe the hole states of the
153: system according to Koopmans' theorem \cite{K33}.
154:
155: Including correlation effects one arrives at correlated wave functions
156: $|\Psi_{\nu} \rangle $ with energies $E_{\nu}$. In analogy to
157: Eq.~(\ref{trafo}) we then define so-called {\em correlated local hole states}
158: $|\Psi_a \rangle $ (see Fig.~\ref{figholestates})
159: \begin{equation}
160: \label{corrlocstates}
161: |\Psi_a\rangle := \sum_{\nu} |\Psi_{\nu}\rangle U_{\nu a}^\ast
162: \end{equation}
163: and an effective Hamilton matrix $\underline{\underline{H}}^{\rm eff} =
164: (H_{ab}^{\rm eff})$ with
165: \begin{equation}
166: \label{effham} H_{ab}^{\rm eff} :=
167: \langle \Psi_a | \hat{H} | \Psi_b \rangle =
168: \sum_{\nu} U_{\nu a} E_{\nu} U_{\nu b}^\ast
169: \quad.
170: \end{equation}
171: The effective Hamiltonian $\hat{H}^{\rm eff}$ is constructed such that
172: diagonalizing its local matrix representation $\underline{\underline{H}}^{\rm
173: eff}$ precisely yields the desired correlated energies $E_{\nu}$ of the
174: cationic states. Thus, provided one knows the correlated local hole states
175: $|\Psi_a\rangle$ defined in Eq.~(\ref{corrlocstates}), one can find the
176: cationic energies by simply solving a small eigenvalue problem.
177:
178: \setlength{\unitlength}{1cm}
179: \begin{figure}
180: \begin{picture}(6,3.5)
181: \put(0.5,3){$\{ |\Phi_{\nu} \rangle \} \quad \stackrel{U}{\longleftrightarrow}
182: \quad \{ | \Phi_{a} \rangle \} \quad \quad $ HF states}
183: \put(0.5,1){$\{ |\Psi_{\nu} \rangle \} \quad \stackrel{U}{\longleftrightarrow}
184: \quad \{ | \Psi_{a} \rangle \} \quad \quad $ correlated states}
185: \put(0.9,2.5){\vector(0,-1){0.8}}
186: \put(1.1,2.1){{\scriptsize MRCI}}
187: \multiput(3.3,2.1)(0,0.2){3}{\line(0,-1){0.15}}
188: \put(3.3,1.9){\vector(0,-1){0.15}}
189: \put(3.5,2.1){{\scriptsize FLHA}}
190: \end{picture}
191: \caption{Relation between the delocalized hole state on the Hartree-Fock
192: level $|\Phi_{\nu}\rangle$ and the corresponding correlated states
193: $|\Psi_{\nu}\rangle$, and their localized counterparts, $|\Phi_{a}\rangle$ and
194: $|\Psi_{a}\rangle$, respectively. Within the frozen local hole approximation
195: (FLHA) approximate correlated local hole states $|\tilde{\Psi}_a\rangle$ are
196: constructed without explicit reference to the true correlated states
197: $|\Psi_{\nu}\rangle$.}
198: \label{figholestates}
199: \end{figure}
200:
201: Within our frozen local hole approach we do not follow the three-step process
202: depicted in Fig.~\ref{figholestates} but directly generate approximate
203: correlated local hole states (CLHSs) $|\tilde{\Psi}_a\rangle$ by performing
204: separate correlation calculations for each reference state $|\Phi_a\rangle$
205: during which the hole in the localized orbital $|a\rangle$ is kept
206: frozen. This implies a configuration selection where only those configurations
207: are used in which the hole in $|a\rangle$ is maintained. In the case of an
208: CI(SD) (CI with single and double excitations) calculation the approximate
209: localized hole states $|\tilde{\Psi}_a\rangle$ take on the following form:
210: \begin{equation}
211: \label{flhstates}
212: |\tilde{\Psi}_a\rangle =\alpha_a |\Phi_a\rangle + \sum_{x,v} \alpha_{a,x}^v
213: |\Phi_{a,x}^v\rangle + \sum_{\substack{x,x'\\v,v'}} \alpha_{a,x,x'}^{v,v'}
214: |\Phi_{a,x,x'}^{v,v'}\rangle
215: \;.
216: \end{equation}
217: The indices $x,x'\in \{\bar{a},b,\bar{b},...\}$ run over all the remaining
218: valence electrons, with $\bar{a}$ being the electron with opposite spin to the
219: removed electron $a$. $v,v'$ denote the electrons in the virtual
220: orbitals. Thus, e.~g., $|\Phi_{a,\bar{a}}^v\rangle$ is the 2h1p configuration
221: with no electrons remaining in the spatial orbital associated with $a$ whereas
222: $|\Phi_{a,b}^v\rangle$ contains holes in two different spatial
223: orbitals. Obviously, the configuration spaces for the different hole states
224: $|\tilde{\Psi}_a\rangle$ are overlapping. Excited configurations with holes in
225: two or more valence orbitals $|a\rangle $, $|b\rangle $, ... are, in fact,
226: present in the correlation calculations of each reference state
227: $|\Phi_a\rangle$, $|\Phi_b\rangle$, $\ldots$ (see Fig.~\ref{figconfig}). As we
228: will see in detail later, this fact does not harm the procedure. On the
229: contrary, it is important to build up the correlation effects in the final,
230: delocalized hole states properly through a suitable mixing of the
231: configurations in the final diagonalization step of the effective Hamiltonian
232: (\ref{effham}),
233: \begin{eqnarray}
234: \label{def:Psi_tilde}
235: |\tilde{\Psi}_\nu \rangle = \sum_a \lambda_a(\nu) |\tilde{\Psi}_a \rangle
236: \quad\quad\mbox{where} \\
237: \label{diagonalization}
238: \sum_b H_{ab}^{\rm eff} \lambda_b(\nu) =
239: \tilde{E}_{\nu} \sum_b S_{ab} \lambda_b(\nu) \quad .
240: \end{eqnarray}
241: Note, that the individual approximated CLHSs $|\tilde{\Psi}_a\rangle$ are not
242: orthonormal with respect to each other. Thus, one has to solve a generalized
243: eigenvalue problem here (with $S_{ab} = \langle \tilde{\Psi}_a |\tilde{\Psi}_b
244: \rangle$ being the overlap matrix) in order to arrive at the approximate
245: canonical hole states $|\tilde{\Psi}_{\nu}\rangle$ and the approximate hole
246: state energies $\tilde{E}_{\nu}$.
247: \vspace{-0.3cm}
248:
249: \section{Applications}
250: \vspace*{-0.2cm}
251: As first applications of the frozen local hole approximation (FLHA) we choose
252: two simple model systems, (H$_2$)$_n$ ladders (see Fig.~\ref{figh2ladder}) and
253: linear H--(Be)$_n$--H chains, for which we compare and analyze the results of
254: the approximation with the results of the corresponding complete
255: multi-reference CI(SD) (MRCI(SD)) calculation. All results are obtained by
256: using the {\sc MOLPRO} program package \cite{MOLPRO}, in particular the
257: MRCI(SD) option \cite{WK88,KW88,KW92}. For the calculations {\it sp} cc-pVDZ
258: basis sets by Dunning \cite{D89} are used for H and Be except for the
259: terminating H atoms in the Be chains which are described by a reduced {\it s}
260: cc-pVDZ basis. The \nobreak{Be--H} distance has been set to $1.37$~\AA\ which
261: is the equilibrium distance found in all neutral chains of length
262: $n$=4--10. The computation of the Hartree-Fock ground state of the neutral
263: system $|\Phi^{N}_0\rangle$ yields the canonical Hartree-Fock orbitals. The
264: set of valence orbitals is subsequently localized by means of the Foster-Boys
265: procedure \cite{FB60} and is used for the construction of the localized hole
266: states $\{|\Phi_{a}\rangle\}$. Each $|\Phi_{a}\rangle$ is then separately
267: correlated in a CI(SD) calculation with the above-described configuration
268: selection, where only those single and double excitations are allowed with a
269: hole in the localized orbital $|a\rangle$. In order to achieve this
270: configuration restriction within the {\sc MOLPRO} program, one has to declare
271: all valence and unoccupied orbitals as active orbitals. The maximum number of
272: active orbitals is fixed to 32 in the {\sc MOLPRO} code, which only allowed us
273: to handle relatively small chains and basis sets chosen here. Nevertheless,
274: the effects of the approximation can be studied very well in these model
275: systems.
276:
277: \begin{figure}
278: \includegraphics[width=7.5cm,clip=true]{Fig2.eps}
279: \caption{Geometry of the (H$_2$)$_n$ ladders used in this paper. The H$_2$
280: bond length is fixed at its molecular value of $R=0.7417$~\AA; the distance
281: between the H$_2$ units is varied during the calculations.}
282: \label{figh2ladder}
283: \end{figure}
284:
285: \begin{figure}
286: \includegraphics*[width=7.5cm,clip=true]{Fig3.eps}
287: \caption{Total Hartree-Fock, MRCI(SD) and FLHA energies of (H$_2$)$_3^+$
288: as a function of the H$_2$--H$_2$ distance.}
289: \label{figh2chainsa}
290: \end{figure}
291:
292: \begin{figure}
293: \includegraphics*[width=7.5cm,clip=true]{Fig4.eps}
294: \caption{Total Hartree-Fock, MRCI(SD) and FLHA energies of (H$_2$)$_5^+$
295: as a function of the H$_2$--H$_2$ distance.}
296: \label{figh2chainsb}
297: \end{figure}
298:
299: The results for the (H$_2$)$_3^+$ and (H$_2$)$_5^+$ ladders are shown in
300: Figs.~\ref{figh2chainsa} and \ref{figh2chainsb}. Here, the total energies of
301: the hole states on the HF, the MRCI(SD) and the FLHA level are depicted in
302: dependence of the distance between the H$_2$ units. In accordance with the
303: number of valence orbitals, and thus possible hole states, we find three and
304: five low-lying cationic states of (H$_2$)$_3^+$ and (H$_2$)$_5^+$,
305: respectively. In their lowest states the cationic (H$_2$)$_n$ ladders are
306: stable. Compared to the HF data, very pronounced correlation effects are
307: observed: In the case of the (H$_2$)$_3^+$ chain the lowest-lying state is
308: lowered by about 2.9~eV at its equilibrium distance, which is slightly shifted
309: from $R_{\rm eq}=1.7$ to $1.6$~\AA, and the potential well is deepened from
310: 1.6 to 1.9~eV. Also for the high-lying states of (H$_2$)$_3^+$ and the states
311: of (H$_2$)$_5^+$ states we find an overall energetical lowering of about 3~eV
312: through correlation. The most important finding in the present context,
313: however, is that all the described effects are very well accounted for by the
314: frozen local hole approximation. The FLHA curves follow the MRCI reference
315: data very closely with the only noticeable slight deviations at the outer left
316: edge of the potential curves.
317:
318: In order to make sure that this excellent agreement is not only an artefact of
319: the relatively weak van der Waals binding between the H$_2$ units, we choose
320: the predominantly covalently bound H--(Be)$_n$--H chains as second
321: application. For all systems considered, H--(Be)$_3$--H through
322: H--(Be)$_5$--H, the FLHA and the ``exact'' MRCI(SD) data agree very well. In
323: the following, we only want to discuss in detail the results for the smallest
324: system Be$_3$H$_2^+$ shown in Fig.~\ref{figbe3h2}. The calculated correlated
325: potential energy curves for the two lowest cationic states exhibit shallow
326: minima at Be--Be distances of $R_{\rm eq}=2.3$ and $2.4$~\AA. For the most
327: stable cationic state of Be$_3$H$_2$ the correlation energies vary from about
328: 1.1~eV at the minima to about 1.5~eV for small and large Be--Be distances.
329:
330: \begin{figure}
331: \includegraphics[width=7.5cm,clip=true]{Fig5.eps}
332: \caption{Total Hartree-Fock, MRCI(SD) and FLHA energies for
333: Be$_3$H$_2^+$ in dependence of the Be--Be distance.}
334: \label{figbe3h2}
335: \end{figure}
336:
337: The dominating configurations of the two lowest cationic states are the two
338: reference configurations $|\Phi_a\rangle$ and $|\Phi_b\rangle$ which result
339: from taking one electron out of one of the highest occupied molecular orbitals
340: (HOMOs) of the HF determinant. The corresponding localized orbitals are
341: centered at the Be--Be bonds (see top row of Fig.~\ref{figvirtualorb}). Within
342: the FLHA two configuration spaces I and II are constructed, built upon the two
343: reference configurations $|\Phi_a\rangle$ and $|\Phi_b\rangle$. They are shown
344: schematically in Fig.~\ref{figconfig}. Clearly, the two configuration spaces
345: strongly overlap. For this simple example with two reference configurations
346: only, all mixed (single-excited) 2h1p configurations $|\Phi_{a,b}^v\rangle$,
347: $|\Phi_{\bar{a},b}^v\rangle$ and $|\Phi_{a,\bar{b}}^v\rangle$ and all
348: (doubly-excited) 3h2p configurations $|\Phi_{a,x,x'}^{v,v'}\rangle$ and
349: $|\Phi_{b,x,x'}^{v,v'}\rangle$ fall into the overlap region. Nevertheless,
350: very important configurations, especially the reference configurations
351: themselves, lie in only one of the spaces. That is why this very simple
352: example is already suited to show the main features of the
353: approximation. Strictly speaking, the single determinants
354: $|\Phi_{a,\bar{b}}^v\rangle$ only belong to space I (as the
355: $|\Phi_{\bar{a},b}^v\rangle$ only belong to space II). But since {\sc MOLPRO}
356: uses spin-adapted excitations these two types of single determinants always
357: come in singlet and triplet-like linear combinations
358: $(|\Phi_{a,\bar{b}}^v\rangle \pm |\Phi_{\bar{a},b}^v\rangle)/\sqrt{2}$ and are
359: thus both attributed to the overlap region I $\cap$ II.
360:
361: In the present example, no substantial saving in the computational cost can be
362: gained from using the FLHA, since we substitute the single MRCI(SD)
363: calculation by two CI calculations with configuration spaces of almost the
364: same size as the original one. But this will change dramatically with bigger
365: systems because a given configuration can be found in at most three different
366: configuration spaces. E.~g., the double excited configuration
367: $|\Phi_{a,b,c}^{v,v'}\rangle$ would only be part of the configuration spaces
368: generated from the reference configurations $|\Phi_a\rangle$, $|\Phi_b\rangle$
369: and $|\Phi_c\rangle$. Once suitable cut-off criteria are introduced, the
370: configuration spaces of each reference configuration remain finite no matter
371: how large the systems becomes; and since the number of these spaces only grow
372: linearly with system size the FLHA directly leads to an ${\cal O}(N)$ method.
373:
374: \begin{figure}
375: \includegraphics[width=5.5cm,clip=true]{Fig6.eps}
376: \caption{Sketch of the overlapping configuration spaces I and II for a system
377: with two valence orbitals and two hole states. The nomenclature is the same as
378: in Eq.~(\ref{flhstates}).}
379: \label{figconfig}
380: \end{figure}
381:
382: Coming back to the discussion of the Be$_3$H$_2$ results, only a few
383: configurations with noticeable weights ($>0.05$) are found in the approximated
384: correlated local hole states
385: $|\tilde{\Psi}_a\rangle$ and $|\tilde{\Psi}_b\rangle$
386: (Eq.~(\ref{flhstates})). The same
387: holds true for the ``full'' canonical MRCI wave functions,
388: $|\Psi^{\rm MRCI}_1\rangle$ and $|\Psi^{\rm MRCI}_2\rangle$. In order to enable an
389: analysis of the individual correlation contributions, we also localize the
390: virtual orbitals, although this is by no means a prerequisite of our FLHA. All
391: relevant configurations contain excitations into one of the four pair-wise
392: degenerated virtual orbitals shown in Fig.~\ref{figvirtualorb}. The first
393: pair, named $c_x^\ast$ and $c_y^\ast$, is localized at the central Be atom,
394: the other pair, $a^\ast$ and $b^\ast$, represents anti-bonding orbitals on the
395: two Be--Be bonds.
396:
397: \begin{figure}
398: \includegraphics[width=6.5cm,clip=true]{Fig7.eps}
399: \caption{Localized highest occupied orbitals (HOMOs) of Be$_3$H$_2^+$ together
400: with those localized virtual orbitals which are involved in the most relevant
401: configurations of the approximate correlated local hole states
402: $|\tilde{\Psi}_a\rangle$ and $|\tilde{\Psi}_b\rangle$. The Hartree-Fock
403: orbital energies $\epsilon$ of each energetically degenerate pair of orbitals
404: are given as well.}
405: \label{figvirtualorb}
406: \end{figure}
407:
408: In Table \ref{tabconfig} we summarize the CI coefficients $\alpha_I$ and
409: $\beta_I$ of the most relevant contributions to the approximate CLHSs
410: $|\tilde{\Psi}_a\rangle$ and $|\tilde{\Psi}_b\rangle$, as well as the CI
411: coefficients $(\alpha_I \pm \beta_I) / \sqrt{2}$ of the approximated wave
412: functions $\tilde{\Psi}_\nu$ of the hole states 1 and 2 as resulting from the
413: diagonalization of the effective Hamiltonian, a ($2 \times 2$) matrix,
414: here. The latter CI coefficients are compared with those of the ``exact'' MRCI
415: wave functions, $\alpha_{1,I}^{\rm MRCI}$ and $\alpha_{2,I}^{\rm MRCI}$. For
416: all spin-adapted single determinants considered here: the reference
417: configurations, the single and double excitations into the anti-bonding
418: orbitals $a^\ast$ and $b^\ast$ and the 3h2p configurations involving the
419: orbitals $c_x^\ast$ and $c_y^\ast$, an excellent agreement is found between
420: the approximate and the ``exact'' CI coefficients of the holes states with
421: deviations well below 10\% in most of the cases. In contrast to our initial
422: expectation, no cancellation of artificial, orbital-relaxation-based
423: contributions in the FLHA wave functions occur. All configurations which are
424: important in the CLHSs also contribute significantly to (at least) one of the
425: final hole states $|\tilde{\Psi}_1\rangle$ and
426: $|\tilde{\Psi}_2\rangle$. Apparently, the ultimate diagonalization step just
427: mixes the individual contributions in such a way that the delocalized nature
428: of the hole states is recovered, precisely as one would expect for an
429: adiabatic approximation.
430: \vspace{-0.3cm}
431:
432: \section{Perturbative analysis}
433: \vspace*{-0.2cm}
434: To understand, why the frozen local hole approximation (FLHA) works so well,
435: we switch from the MRCI level of theory to quasi-degenerate variational
436: perturbation theory (QDVPT) \cite{CD88} and try to find the leading
437: contributions in both, the ``exact'' hole state and the approximate hole
438: states according to the FLHA. Within QDVPT the correlated wave functions
439: $|\Psi_\nu\rangle$ of a system are written as
440: \begin{equation}
441: |\Psi_\nu\rangle = \hat{\Omega} |\Psi^{\cal M}_\nu \rangle
442: \end{equation}
443: where $|\Psi^{\cal M}_\nu \rangle = \hat{P} |\Psi_\nu \rangle$ are the
444: projections of the correlated wave functions $|\Psi_\nu \rangle$ onto the
445: space ${\cal M}$ spanned by the local Hartree-Fock hole configurations
446: $\{|\Phi_a \rangle\}$, and $\hat{\Omega}$ is an operator which acts on the
447: ``model space'' ${\cal M}$ and provides the full, correlated wave functions
448: $|\Psi_\nu \rangle$. It is assumed that the $|\Psi_\nu \rangle$ are those
449: eigenstates of the hole system which are dominated by the single determinants
450: forming the model space and, in particular, that the projections $|\Psi^{\cal
451: M}_\nu \rangle$ remain linear independent. In fact, \begin{equation}
452: \hat{\Omega} = \hat{P} + \hat{Q} \hat{\Omega} \quad\mbox{with}\quad
453: \hat{Q}\,\hat{\Omega} = \sum_{I,a} |\Phi_I\rangle \Omega_{Ia} \langle\Phi_a|
454: \end{equation}
455: where $\hat{Q} = \sum_I |\Phi_I \rangle \langle \Phi_I|$ is the projector onto
456: the orthogonal complement of the model space and the configurations $| \Phi_I
457: \rangle$ run over all single, double and higher excitations one can form from
458: the local model space configurations $\{|\Phi_a \rangle\}$.
459:
460: In first order perturbation theory the wave operator $\hat{\Omega}$ is given
461: by \cite{CD88}
462: \begin{eqnarray}
463: \label{def:Omega}
464: \hat{\Omega} &=& \sum_c | \Phi_c \rangle \langle \Phi_c |
465: \\ \nonumber &+& \sum_c \sum_{a<b,v} |\Phi_{ab}^v \rangle \,
466: \frac{ \langle v c || a b \rangle }
467: { \varepsilon_v - \varepsilon_a - \varepsilon_b + \varepsilon_c } \langle
468: \, \Phi_c | \quad.
469: \end{eqnarray}
470: We take the zeroth-order Hamiltonian $H^0$ to consist of the diagonal terms of
471: the Fock operator only, i.~e., $H^0_{ij} = \delta_{ij} \varepsilon_i$ with
472: $\varepsilon_i = F_{ii}$. The perturbation then contains both, the
473: off-diagonal terms $F_{ij} = \langle \varphi_i | \hat{F} | \varphi_j \rangle
474: \;,\; i\ne j$ of the Fock operator $\hat{F}$ and the usual two-electron
475: contributions $\langle i j || k l \rangle - \sum^{\rm occ}_a \langle i a || j
476: a \rangle$. Here and in the following, indices $a$, $b$, $\ldots$ are
477: understood to run over occupied localized orbitals while indices $v$, $w$,
478: $\ldots$ run over virtual orbitals. For arbitrary orbitals indices $i$, $j$,
479: $\ldots$ are used. The 12,12 notation
480: \begin{eqnarray}
481: \langle i j || k l \rangle =
482: \langle i j | \hat{g} | k l \rangle - \langle i j | \hat{g} | l k \rangle
483: \quad\quad\mbox{with} \\ \nonumber
484: \langle i j | \hat{g} | k l \rangle =
485: \int \! \! \! \int \varphi^\ast_i(1) \varphi^\ast_j(2) \:r_{12}^{-1}\:
486: \varphi_k(1) \varphi_l(2) \: d1\,d2
487: \end{eqnarray}
488: is employed throughout. The $|\varphi_i \rangle$ are the (localized)
489: Hartree-Fock orbitals of the neutral $N$-electron system. Using that form of
490: the wave operator as {\it variational} ansatz for the eigenstates of the hole
491: system, i.~e.
492: \begin{equation}
493: |\Psi_\nu \rangle
494: = \hat{\Omega} \big[ \: \sum_a \lambda_a(\nu) |\Phi_a \rangle \: \big]
495: = \sum_a \lambda_a(\nu) \: \hat\Omega |\Phi_a \rangle
496: \quad,
497: \end{equation}
498: leads to the following effective secular matrix
499: \begin{eqnarray}
500: H^{\rm eff}_{ef} &:=&
501: \langle \, \hat{\Omega}\!\big[ \Phi_e \big] |\hat{H}| \,
502: \hat{\Omega}\!\big[ \Phi_f \big] \rangle \\[2mm]
503: &=& - F_{fe} \\ \nonumber
504: %%&& + \sum_{c<d,w} W_{e|cd,f}^{\;\;w} \: \langle w f || c d \rangle
505: && + \sum_{a<b,v} W_{e|ab,f}^{\;\;v} \: \langle v f || a b \rangle
506: \\ \nonumber
507: && + \sum_{a<b,v} W_{e,ab|f}^{\;\;v} \: \langle a b || v e \rangle
508: \\ \nonumber
509: && + \sum_{a<b,v} \sum_{c<d,w} W_{e,ab|cd,f}^{\;\;v\;\;\,w} \:
510: \langle a b || v e \rangle \: \langle w f || c d \rangle
511: \\ \nonumber
512: \mbox{with} \\
513: %%W_{e|cd,f}^{\;\;w} &=& \frac{ \langle \Phi_e | \hat{H} | \Phi_{cd}^w \rangle }
514: %%{ ( \varepsilon_w - \varepsilon_c - \varepsilon_d + \varepsilon_f ) }
515: W_{e|ab,f}^{\;\;v} &=& \frac{ \langle \Phi_e | \hat{H} | \Phi_{ab}^v \rangle }
516: { ( \varepsilon_v - \varepsilon_a - \varepsilon_b + \varepsilon_f ) }
517: \\ \nonumber
518: W_{e,ab|f}^{\;\;v} &=& \frac{ \langle \Phi_{ab}^v | \hat{H} | \Phi_f \rangle }
519: { ( \varepsilon_v - \varepsilon_a - \varepsilon_b + \varepsilon_e ) }
520: \;=\; ( W_{f|ab,e}^{\;\;\,v} )^\ast
521: \\ \nonumber
522: W_{e,ab|cd,f}^{\;\;v\;\;\,w} &=&
523: \frac{ \langle \Phi_{ab}^v | \hat{H} | \Phi_{cd}^w \rangle }
524: { ( \varepsilon_v - \varepsilon_a - \varepsilon_b + \varepsilon_e ) \:
525: ( \varepsilon_w - \varepsilon_c - \varepsilon_d + \varepsilon_f ) }
526: \;\;.
527: \end{eqnarray}
528: This is not what is usually done in QDVPT, but to preceed this way allows to
529: find the direct link to the frozen local hole approximation we are looking
530: for.
531:
532: In the first step of the FLHA, the model space configurations and the excited
533: configurations are restricted to those determinants which already contain a
534: hole in a particular local occupied orbital $|\varphi_h \rangle$, such that
535: the wave operator reduces to
536: \begin{equation}
537: \hat{\Omega}^{\rm FLH}(h) = |\Phi_h \rangle \langle \Phi_h |
538: \: + \sum_{a \ne h,v} | \Phi_{ah}^v \rangle \,
539: \frac{ \langle v h || a h \rangle }
540: { \varepsilon_v - \varepsilon_a } \langle \Phi_h |
541: \quad.
542: \label{eqomegaflh}
543: \end{equation}
544: Only single excitations $|\Phi_{ah}^v\rangle$ with respect to the given local
545: hole configuration $|\Phi_h\rangle$ show up here. Inspection of
546: Eq.~(\ref{def:Omega}) reveals that these restricted wave operators are
547: precisely the leading terms in the full wave operator $\hat{\Omega}$, in the
548: sense that
549: \begin{equation}
550: \hat{\Omega} = \sum_h \hat{\Omega}^{\rm FLH}(h) + \hat{\Pi}
551: \end{equation}
552: where the remainder
553: \begin{equation}
554: \hat{\Pi} = \sum_{a<b,v} \sum_{c \not\in \{a,b\}}
555: |\Phi_{ab}^v\rangle \,
556: \frac{\langle v c || a b \rangle }
557: {\varepsilon_v - \varepsilon_a - \varepsilon_b + \varepsilon_c }
558: \, \langle\Phi_c|
559: \end{equation}
560: contains all terms in which the three occupied spin orbitals $|a\rangle$,
561: $|b\rangle$, and $|c\rangle$ are all {\it different}, while the leading terms
562: are made up by all those contributions in which two of the hole indices
563: coincide.
564:
565: The final secular matrix of the frozen local hole approach,
566: \begin{equation}
567: H^{\rm FLH}_{gh} :=
568: \langle\,\hat{\Omega}^{\rm FLH}(g)\!\big[ \Phi_g \big] | \hat{H} |\,
569: \hat{\Omega}^{\rm FLH}(h)\!\big[ \Phi_h \big] \rangle \quad,
570: \end{equation}
571: is then given by
572: \begin{eqnarray}
573: H^{\rm FLH}_{gh} &=& - F_{hg} \\ \nonumber
574: && + \sum_{c \ne h,w} W_{g|ch,h}^{\;\;w} \langle w h || c h \rangle
575: \\ \nonumber
576: && + \sum_{a \ne g,v} W_{g,ag|h}^{\;\;v} \langle a g || v g \rangle
577: \\ \nonumber
578: && +\sum_{a \ne g,v} \sum_{c \ne h,w} W_{g,ag|ch,h}^{\;\;v\;\;\;w}
579: \langle a g || v g \rangle \langle w h || c h \rangle
580: \end{eqnarray}
581: which is again precisely the leading coinciding-hole-indices part of the full
582: effective Hamiltonian $H^{\rm eff}_{gh}$ (despite the always present
583: zeroth-order contribution $-F_{hg}$). Thus, up to first order perturbation
584: theory the FLHA can be understood as a simple neglection of all
585: three-distinct-spin-orbital contributions to the wave operator $\hat{\Omega}$
586: and the resulting effective Hamiltonian.
587:
588: A further simplification of the FLHA can be achieved by treating the
589: ($N$-1)-particle system with a frozen local hole on the Hartree-Fock level
590: only. This has already been done with some success in our ``simplified
591: method'' to correlation effects in band structures \cite{BFSpp}. Orbital
592: relaxation around the localized hole is the only effect which can be accounted
593: for in this approach and we want to analyze here, to which extent the
594: correlation effects in the ($N$-1)-particle system can be mimicked this way.
595:
596: Removal of an electron from a fixed localized occupied spin orbital
597: $|\varphi_h\rangle$ leads to the following modified Fock operator
598: $\hat{F}^{\rm FLH}(h)$,
599: \begin{eqnarray}\nonumber
600: \langle i | \hat{F}^{\rm FLH}(h) | j \rangle
601: &=& F_{ij} - \langle i h || j h \rangle
602: \:=\: \delta_{ij} \, \varepsilon_i + \langle i | V | j \rangle
603: \\
604: \mbox{with} \quad \langle i | V | j \rangle &=&
605: (1 - \delta_{ij} ) \, F_{ij} - \langle i h || j h \rangle
606: \end{eqnarray}
607:
608: Up to first order in the perturbation potential $V$ the relaxed Hartree-Fock
609: orbitals $|\tilde{\varphi}_i(h)\rangle$ which result from diagonalizing
610: $\hat{F}^{\rm FLH}(h)$ under the constraint that $|\tilde{\varphi}_h\rangle$
611: remains unchanged read
612: \begin{equation}
613: |\tilde{\varphi}_i(h)\rangle = \left\{ \begin{array}{ll} \displaystyle{
614: |\varphi_i\rangle - \sum_{j \not\in\{i,h\}} |\varphi_j \rangle \,
615: \frac{\langle j | V | i \rangle }{\varepsilon_j - \varepsilon_i} }
616: & \quad\mbox{for } i \ne h \\
617: |\varphi_h\rangle & \quad\mbox{else}
618: \end{array} \right.
619: \end{equation}
620: and the corresponding Slater determinant
621: \begin{eqnarray}
622: |\Phi_h^{\rm SCF}\rangle &:=& \frac{(-1)^{N-h}}{\sqrt{(N-1)!}} \times \\ \nonumber
623: && \mbox{det} (\tilde{\varphi}_1(h),\ldots,\tilde{\varphi}_{h-1}(h),
624: \tilde{\varphi}_{h+1}(h),\ldots,\tilde{\varphi}_N(h))
625: \end{eqnarray}
626: becomes
627: \begin{equation}
628: |\Phi_h^{\rm SCF}\rangle =
629: |\Phi_h\rangle - \sum_{a \ne h} \sum_{j \not\in\{a,h\}} |\Phi_{ah}^j\rangle \,
630: \frac{\langle j | V | a \rangle }{\varepsilon_j - \varepsilon_a } +
631: {\cal O}(V^2) \quad .
632: \end{equation}
633: Excitations into occupied spin orbitals $|\varphi_j\rangle $ other than
634: $|\varphi_a\rangle $ and $|\varphi_h\rangle $ are not possible in the 2h1p
635: configurations $|\Phi_{ah}^j\rangle $. Thus, $j$ only runs over the virtual
636: orbitals and one has
637: \begin{eqnarray}
638: |\Phi_h^{\rm SCF}\rangle &=& |\Phi_h\rangle +
639: \sum_{a \ne h,v} |\Phi_{ah}^v\rangle \,
640: \frac{\langle v h || a h \rangle }{\varepsilon_v - \varepsilon_a} + {\cal O}(V^2)
641: \\ &=& \hat{\Omega}^{\rm FLH}(h) |\Phi_h\rangle + {\cal O}(V^2)
642: \end{eqnarray}
643: Obviously, up to first order, the local frozen hole wave function obtained on
644: the Hartree-Fock level is identical to the fully correlated local hole state
645: $|\tilde{\Psi}_h\rangle = \hat{\Omega}^{\rm FLH}(h) |\Phi_h\rangle$ of the
646: system (see Eq.~(\ref{eqomegaflh})). This result is somewhat surprising, at
647: first glance, because it tells, that what formally looks like pure orbital
648: relaxation around a frozen local defect, is actually bare electron
649: correlation. It also tells, that performing a Hartree-Fock calculation around
650: a frozen local hole rather than a much more demanding wavefunction-based
651: correlation calculations is a rather promising approximation.
652:
653: \section{Conclusions}
654:
655: The so-called frozen local hole approximation (FLHA) has been analyzed in this
656: study. It presents a method that allows to perform correlation calculations
657: for cationic and anionic ($N$$\pm$1)-electron systems in a much more efficient
658: way than running a full quantum chemical correlation calculation such as
659: MRCI(SD). In this work we were focusing on cationic holes states, but, in a
660: totally analogue way, the approximation can also be applied to anionic
661: electron attachment states (as done in Ref.~\cite{BFSpp}).
662:
663: The FLHA is a two step procedure: In the first step, the electron to be
664: removed is assumed to reside in a given, fixed local occupied orbital and
665: standard wavefunction-based correlation calculations are performed in order to
666: find the correlation hole around each of these local holes. Because of the
667: frozen character of the given hole orbital the configuration space is
668: substantially reduced in these correlation calculations. Therefore, the
669: approximation is a perfect, linear scaling divide-and-conquer approach. In the
670: second step, the many-body Hamiltonian matrix elements with the correlated
671: local hole states generated in the first step are evaluated and the resulting
672: effective Hamiltonian matrix is diagonalized. In this way, the hole can
673: delocalize over the whole system to form proper Bloch states. Hence, the FLHA
674: can be understood as an adiabatic treatment, in which the shape of the
675: correlation hole around each electron hole is kept fixed during the final
676: hybridization of the correlated local hole states.
677:
678: It has been shown numerically, for two sample systems, $({\rm H}_2)_n$ ladder
679: chains and hydrogen terminated linear H--(Be)$_n$--H chains, that the FLHA
680: performs astonishingly well. Potential energy curves have been calculated on
681: the MRCI(SD) level of theory, and deviations of at most 0.1 eV between a full
682: correlation calculations and the frozen local hole approach were found, both,
683: for the van der Waals bound H$_2$ chains and for the covalent Be chains.
684:
685: Using quasi-degenerate variational perturbation theory, it was possible to
686: demonstrate that the FLHA indeed assembles all leading terms of a full
687: correlation calculation. Actually, the approximation consists in a neglection
688: of all three-distinct-hole-site contributions over those where (at least) two
689: hole orbitals coincide. Furthermore, it could be shown that the correlation
690: calculation around a frozen {\it local} hole can very well be replaced by a
691: numerically much less demanding SCF calculation. In first order perturbation
692: theory the resulting orbital relaxations are totally equivalent to the
693: correlation effects around a frozen local hole.
694:
695: \begin{thebibliography}{41}
696: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
697: \expandafter\ifx\csname bibnamefont\endcsname\relax
698: \def\bibnamefont#1{#1}\fi
699: \expandafter\ifx\csname bibfnamefont\endcsname\relax
700: \def\bibfnamefont#1{#1}\fi
701: \expandafter\ifx\csname citenamefont\endcsname\relax
702: \def\citenamefont#1{#1}\fi
703: \expandafter\ifx\csname url\endcsname\relax
704: \def\url#1{\texttt{#1}}\fi
705: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
706: \providecommand{\bibinfo}[2]{#2}
707: \providecommand{\eprint}[2][]{\url{#2}}
708:
709: \bibitem[{\citenamefont{{Gr\"{a}fenstein}
710: et~al.}(1993)\citenamefont{{Gr\"{a}fenstein}, Stoll, and Fulde}}]{GSF93}
711: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{{Gr\"{a}fenstein}}},
712: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Stoll}}, \bibnamefont{and}
713: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Fulde}},
714: \bibinfo{journal}{Chem.~Phys.~Lett.} \textbf{\bibinfo{volume}{215}},
715: \bibinfo{pages}{611} (\bibinfo{year}{1993}).
716:
717: \bibitem[{\citenamefont{{Gr\"{a}fenstein}
718: et~al.}(1997)\citenamefont{{Gr\"{a}fenstein}, Stoll, and Fulde}}]{GSF97}
719: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{{Gr\"{a}fenstein}}},
720: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Stoll}}, \bibnamefont{and}
721: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Fulde}},
722: \bibinfo{journal}{Phys.~Rev.~B} \textbf{\bibinfo{volume}{55}},
723: \bibinfo{pages}{13588} (\bibinfo{year}{1997}).
724:
725: \bibitem[{\citenamefont{Albrecht et~al.}(2000)\citenamefont{Albrecht, Fulde,
726: and Stoll}}]{AFS00}
727: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Albrecht}},
728: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Fulde}}, \bibnamefont{and}
729: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Stoll}},
730: \bibinfo{journal}{Chem.~Phys.~Lett.} \textbf{\bibinfo{volume}{319}},
731: \bibinfo{pages}{355} (\bibinfo{year}{2000}).
732:
733: \bibitem[{\citenamefont{Albrecht and Igarashi}(2001)}]{AI01}
734: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Albrecht}} \bibnamefont{and}
735: \bibinfo{author}{\bibfnamefont{J.-I.} \bibnamefont{Igarashi}},
736: \bibinfo{journal}{J.~Phys.~Soc.~Jap.} \textbf{\bibinfo{volume}{70}},
737: \bibinfo{pages}{1035} (\bibinfo{year}{2001}).
738:
739: \bibitem[{\citenamefont{Albrecht}(2002)}]{A02}
740: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Albrecht}},
741: \bibinfo{journal}{Theo.~Chem.~Acc.} \textbf{\bibinfo{volume}{107}},
742: \bibinfo{pages}{71} (\bibinfo{year}{2002}).
743:
744: \bibitem[{\citenamefont{V.Bezugly and Birkenheuer}(2004)}]{BB04}
745: \bibinfo{author}{\bibnamefont{V.Bezugly}} \bibnamefont{and}
746: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Birkenheuer}},
747: \bibinfo{journal}{Chem.~Phys.~Lett.} \textbf{\bibinfo{volume}{399}},
748: \bibinfo{pages}{57} (\bibinfo{year}{2004}).
749:
750: \bibitem[{\citenamefont{de~Graaf et~al.}(1999)\citenamefont{de~Graaf, Sousa,
751: and Broer}}]{GSB99}
752: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{de~Graaf}},
753: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Sousa}}, \bibnamefont{and}
754: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Broer}},
755: \bibinfo{journal}{J.~Mol.~Struc.\ (TheoChem)} \textbf{\bibinfo{volume}{458}},
756: \bibinfo{pages}{53} (\bibinfo{year}{1999}).
757:
758: \bibitem[{\citenamefont{Hozoi et~al.}(2002)\citenamefont{Hozoi, de~Vries, van
759: Oosten, Broer, Cabrero, and de~Graaf}}]{HVOB02}
760: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Hozoi}},
761: \bibinfo{author}{\bibfnamefont{A.~H.} \bibnamefont{de~Vries}},
762: \bibinfo{author}{\bibfnamefont{A.~B.} \bibnamefont{van Oosten}},
763: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Broer}},
764: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Cabrero}}, \bibnamefont{and}
765: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{de~Graaf}},
766: \bibinfo{journal}{Phys.~Rev.~Lett.} \textbf{\bibinfo{volume}{89}},
767: \bibinfo{pages}{076407} (\bibinfo{year}{2002}).
768:
769: \bibitem[{\citenamefont{Hozoi et~al.}(2003)\citenamefont{Hozoi, Presura,
770: de~Graaf, and Broer}}]{HPGB03}
771: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Hozoi}},
772: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Presura}},
773: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{de~Graaf}}, \bibnamefont{and}
774: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Broer}},
775: \bibinfo{journal}{Phys.~Rev.~B} \textbf{\bibinfo{volume}{67}},
776: \bibinfo{pages}{035117} (\bibinfo{year}{2003}).
777:
778: \bibitem[{\citenamefont{Buth et~al.}(2005)\citenamefont{Buth, Birkenheuer,
779: Albrecht, and Fulde}}]{BBAF05}
780: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Buth}},
781: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Birkenheuer}},
782: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Albrecht}}, \bibnamefont{and}
783: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Fulde}},
784: \bibinfo{journal}{Phys.~Rev.~B} \textbf{\bibinfo{volume}{72}},
785: \bibinfo{pages}{195107} (\bibinfo{year}{2005}).
786:
787: \bibitem[{\citenamefont{Pisani}(2002)}]{P02}
788: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Pisani}},
789: \bibinfo{journal}{J.~Mol.~Struct.~(TheoChem)} \textbf{\bibinfo{volume}{621}},
790: \bibinfo{pages}{141} (\bibinfo{year}{2002}).
791:
792: \bibitem[{\citenamefont{Saeb{\o} and Pulay}(1987)}]{SP87}
793: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Saeb{\o}}} \bibnamefont{and}
794: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Pulay}},
795: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{86}},
796: \bibinfo{pages}{914} (\bibinfo{year}{1987}).
797:
798: \bibitem[{\citenamefont{Hampel and Werner}(1996)}]{HW96}
799: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Hampel}} \bibnamefont{and}
800: \bibinfo{author}{\bibfnamefont{H.-J.} \bibnamefont{Werner}},
801: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{104}},
802: \bibinfo{pages}{6286} (\bibinfo{year}{1996}).
803:
804: \bibitem[{\citenamefont{{Sch\"{u}tz} et~al.}(1999)\citenamefont{{Sch\"{u}tz},
805: Hetzer, and Werner}}]{SHW99}
806: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Sch\"{u}tz}}},
807: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Hetzer}}, \bibnamefont{and}
808: \bibinfo{author}{\bibfnamefont{H.-J.} \bibnamefont{Werner}},
809: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{111}},
810: \bibinfo{pages}{5691} (\bibinfo{year}{1999}).
811:
812: \bibitem[{\citenamefont{Sun and Bartlett}(1996)}]{SB96}
813: \bibinfo{author}{\bibfnamefont{J.-Q.} \bibnamefont{Sun}} \bibnamefont{and}
814: \bibinfo{author}{\bibfnamefont{R.~J.} \bibnamefont{Bartlett}},
815: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{104}},
816: \bibinfo{pages}{8553} (\bibinfo{year}{1996}).
817:
818: \bibitem[{\citenamefont{Ayala et~al.}(2001)\citenamefont{Ayala, Kudin, and
819: Scuseria}}]{AKS01}
820: \bibinfo{author}{\bibfnamefont{P.~Y.} \bibnamefont{Ayala}},
821: \bibinfo{author}{\bibfnamefont{K.~N.} \bibnamefont{Kudin}}, \bibnamefont{and}
822: \bibinfo{author}{\bibfnamefont{G.~E.} \bibnamefont{Scuseria}},
823: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{115}},
824: \bibinfo{pages}{9698} (\bibinfo{year}{2001}).
825:
826: \bibitem[{\citenamefont{Pisani et~al.}(2005)\citenamefont{Pisani, Busse,
827: Capecchi, Casassa, Dovesi, Maschio, Zicovich-Wilson, and
828: {Sch\"{u}tz}}}]{PBCC05}
829: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Pisani}},
830: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Busse}},
831: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Capecchi}},
832: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Casassa}},
833: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Dovesi}},
834: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Maschio}},
835: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Zicovich-Wilson}},
836: \bibnamefont{and}
837: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Sch\"{u}tz}}},
838: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{122}},
839: \bibinfo{pages}{094113} (\bibinfo{year}{2005}).
840:
841: \bibitem[{\citenamefont{Fink and Staemmler}(1995)}]{FS95}
842: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Fink}} \bibnamefont{and}
843: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Staemmler}},
844: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{103}},
845: \bibinfo{pages}{2603} (\bibinfo{year}{1995}).
846:
847: \bibitem[{\citenamefont{Kitaura et~al.}(1999)\citenamefont{Kitaura, Ikeo,
848: Asada, Nakano, and Uebayasi}}]{KIANU99}
849: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Kitaura}},
850: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Ikeo}},
851: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Asada}},
852: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Nakano}}, \bibnamefont{and}
853: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Uebayasi}},
854: \bibinfo{journal}{Chem.~Phys.~Lett.} \textbf{\bibinfo{volume}{313}},
855: \bibinfo{pages}{701} (\bibinfo{year}{1999}).
856:
857: \bibitem[{\citenamefont{Fedorov and Kitaura}(2004)}]{FK04}
858: \bibinfo{author}{\bibfnamefont{D.~G.} \bibnamefont{Fedorov}} \bibnamefont{and}
859: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Kitaura}},
860: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{121}},
861: \bibinfo{pages}{2483} (\bibinfo{year}{2004}).
862:
863: \bibitem[{\citenamefont{Mochizuki et~al.}(2005)\citenamefont{Mochizuki,
864: Koikegami, Amari, Segawa, Kitaura, and Nakano}}]{MKAS05}
865: \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Mochizuki}},
866: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Koikegami}},
867: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Amari}},
868: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Segawa}},
869: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Kitaura}}, \bibnamefont{and}
870: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Nakano}},
871: \bibinfo{journal}{Chem.~Phys.~Lett.)} \textbf{\bibinfo{volume}{406}},
872: \bibinfo{pages}{283} (\bibinfo{year}{2005}).
873:
874: \bibitem[{\citenamefont{Stoll}(1992{\natexlab{a}})}]{S92-1}
875: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Stoll}},
876: \bibinfo{journal}{Chem.~Phys.~Lett.} \textbf{\bibinfo{volume}{191}},
877: \bibinfo{pages}{548} (\bibinfo{year}{1992}{\natexlab{a}}).
878:
879: \bibitem[{\citenamefont{Stoll}(1992{\natexlab{b}})}]{S92-2}
880: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Stoll}},
881: \bibinfo{journal}{Phys.~Rev.~B)} \textbf{\bibinfo{volume}{B46}},
882: \bibinfo{pages}{6700} (\bibinfo{year}{1992}{\natexlab{b}}).
883:
884: \bibitem[{\citenamefont{Stoll}(1992{\natexlab{c}})}]{S92-3}
885: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Stoll}},
886: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{97}},
887: \bibinfo{pages}{8449} (\bibinfo{year}{1992}{\natexlab{c}}).
888:
889: \bibitem[{\citenamefont{Paulus et~al.}(1995)\citenamefont{Paulus, Fulde, and
890: Stoll}}]{PFS95}
891: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Paulus}},
892: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Fulde}}, \bibnamefont{and}
893: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Stoll}},
894: \bibinfo{journal}{Phys.~Rev.~B} \textbf{\bibinfo{volume}{51}},
895: \bibinfo{pages}{10572} (\bibinfo{year}{1995}).
896:
897: \bibitem[{\citenamefont{Fulde}(2002)}]{F02}
898: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Fulde}},
899: \bibinfo{journal}{Adv.~Phys.} \textbf{\bibinfo{volume}{51}},
900: \bibinfo{pages}{909} (\bibinfo{year}{2002}).
901:
902: \bibitem[{\citenamefont{Willnauer and Birkenheuer}(2004)}]{WB04}
903: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Willnauer}} \bibnamefont{and}
904: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Birkenheuer}},
905: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{120}},
906: \bibinfo{pages}{11910} (\bibinfo{year}{2004}).
907:
908: \bibitem[{\citenamefont{Angeli et~al.}(2003)\citenamefont{Angeli, Calzado,
909: Cimiraglia, Evangelisti, Guih{\'e}ry, Leininger, Malrieu, Maynau, Ruiz, and
910: Sparta}}]{ACCE03}
911: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Angeli}},
912: \bibinfo{author}{\bibfnamefont{C.~J.} \bibnamefont{Calzado}},
913: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Cimiraglia}},
914: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Evangelisti}},
915: \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{Guih{\'e}ry}},
916: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Leininger}},
917: \bibinfo{author}{\bibfnamefont{J.-P.} \bibnamefont{Malrieu}},
918: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Maynau}},
919: \bibinfo{author}{\bibfnamefont{J.~V.~P.} \bibnamefont{Ruiz}},
920: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Sparta}},
921: \bibinfo{journal}{Mol.~Phys.} \textbf{\bibinfo{volume}{101}},
922: \bibinfo{pages}{1389} (\bibinfo{year}{2003}).
923:
924: \bibitem[{\citenamefont{Borini et~al.}(2005)\citenamefont{Borini, Maynau, and
925: Evangelisti}}]{BME05}
926: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Borini}},
927: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Maynau}}, \bibnamefont{and}
928: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Evangelisti}},
929: \bibinfo{journal}{J.~Comput.~Chem.} \textbf{\bibinfo{volume}{26}},
930: \bibinfo{pages}{1042} (\bibinfo{year}{2005}).
931:
932: \bibitem[{\citenamefont{Birkenheuer et~al.}(2005)\citenamefont{Birkenheuer,
933: Fulde, and Stoll}}]{BFSpp}
934: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Birkenheuer}},
935: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Fulde}}, \bibnamefont{and}
936: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Stoll}},
937: \bibinfo{journal}{Theo.~Chem.~Acc.} (\bibinfo{year}{2005}),
938: \bibinfo{note}{submitted, {\tt arXiv:cond-mat/0511626}}.
939:
940: \bibitem[{\citenamefont{{R\"{o}ssler} and Staemmler}(2003)}]{RS03}
941: \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{{R\"{o}ssler}}}
942: \bibnamefont{and}
943: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Staemmler}},
944: \bibinfo{journal}{Phys.~Chem.~Chem.~Phys.} \textbf{\bibinfo{volume}{5}},
945: \bibinfo{pages}{3580} (\bibinfo{year}{2003}).
946:
947: \bibitem[{\citenamefont{de~Graaf et~al.}(1997)\citenamefont{de~Graaf, Broer,
948: Nieuwpoort, and Bagus}}]{GBNB97}
949: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{de~Graaf}},
950: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Broer}},
951: \bibinfo{author}{\bibfnamefont{W.~C.} \bibnamefont{Nieuwpoort}},
952: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{P.~S.} \bibnamefont{Bagus}},
953: \bibinfo{journal}{Chem.~Phys.~Lett.} \textbf{\bibinfo{volume}{272}},
954: \bibinfo{pages}{341} (\bibinfo{year}{1997}).
955:
956: \bibitem[{\citenamefont{de~Vries et~al.}(2002)\citenamefont{de~Vries, Hozoi,
957: Broer, and Bagus}}]{VHBB02}
958: \bibinfo{author}{\bibfnamefont{A.~H.} \bibnamefont{de~Vries}},
959: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Hozoi}},
960: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Broer}}, \bibnamefont{and}
961: \bibinfo{author}{\bibfnamefont{P.~S.} \bibnamefont{Bagus}},
962: \bibinfo{journal}{Phys.~Rev.~B} \textbf{\bibinfo{volume}{66}},
963: \bibinfo{pages}{035108} (\bibinfo{year}{2002}).
964:
965: \bibitem[{\citenamefont{Koopmans}(1933)}]{K33}
966: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Koopmans}},
967: \bibinfo{journal}{Physica} \textbf{\bibinfo{volume}{1}}, \bibinfo{pages}{104}
968: (\bibinfo{year}{1933}).
969:
970: \bibitem[{\citenamefont{Werner et~al.}(2003)\citenamefont{Werner, Knowles,
971: Lindh, {Sch\"{u}tz}, Celani, Korona, Manby, Rauhut, Amos, Bernhardsson
972: et~al.}}]{MOLPRO}
973: \bibinfo{author}{\bibfnamefont{H.-J.} \bibnamefont{Werner}},
974: \bibinfo{author}{\bibfnamefont{P.~J.} \bibnamefont{Knowles}},
975: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Lindh}},
976: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Sch\"{u}tz}}},
977: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Celani}},
978: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Korona}},
979: \bibinfo{author}{\bibfnamefont{F.~R.} \bibnamefont{Manby}},
980: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Rauhut}},
981: \bibinfo{author}{\bibfnamefont{R.~D.} \bibnamefont{Amos}},
982: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Bernhardsson}},
983: \bibnamefont{et~al.}, \emph{\bibinfo{title}{Molpro, version 2002.6, a package
984: of ab initio programs}} (\bibinfo{year}{2003}), \bibinfo{note}{see
985: http://www.molpro.net}.
986:
987: \bibitem[{\citenamefont{Werner and Knowles}(1988)}]{WK88}
988: \bibinfo{author}{\bibfnamefont{H.-J.} \bibnamefont{Werner}} \bibnamefont{and}
989: \bibinfo{author}{\bibfnamefont{P.~J.} \bibnamefont{Knowles}},
990: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{89}},
991: \bibinfo{pages}{5803} (\bibinfo{year}{1988}).
992:
993: \bibitem[{\citenamefont{Knowles and Werner}(1988)}]{KW88}
994: \bibinfo{author}{\bibfnamefont{P.~J.} \bibnamefont{Knowles}} \bibnamefont{and}
995: \bibinfo{author}{\bibfnamefont{H.-J.} \bibnamefont{Werner}},
996: \bibinfo{journal}{Chem.~Phys.~Letters} \textbf{\bibinfo{volume}{145}},
997: \bibinfo{pages}{514} (\bibinfo{year}{1988}).
998:
999: \bibitem[{\citenamefont{Knowles and Werner}(1992)}]{KW92}
1000: \bibinfo{author}{\bibfnamefont{P.~J.} \bibnamefont{Knowles}} \bibnamefont{and}
1001: \bibinfo{author}{\bibfnamefont{H.-J.} \bibnamefont{Werner}},
1002: \bibinfo{journal}{Theor.~Chim.~Acta} \textbf{\bibinfo{volume}{84}},
1003: \bibinfo{pages}{95} (\bibinfo{year}{1992}).
1004:
1005: \bibitem[{\citenamefont{T.~H.~Dunning}(1989)}]{D89}
1006: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{T.~H.~Dunning}},
1007: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{90}},
1008: \bibinfo{pages}{1007} (\bibinfo{year}{1989}).
1009:
1010: \bibitem[{\citenamefont{Foster and Boys}(1960)}]{FB60}
1011: \bibinfo{author}{\bibfnamefont{J.~M.} \bibnamefont{Foster}} \bibnamefont{and}
1012: \bibinfo{author}{\bibfnamefont{S.~F.} \bibnamefont{Boys}},
1013: \bibinfo{journal}{Rev.~Mod.~Phys.} \textbf{\bibinfo{volume}{32}},
1014: \bibinfo{pages}{300} (\bibinfo{year}{1960}).
1015:
1016: \bibitem[{\citenamefont{Cave and Davidson}(1988)}]{CD88}
1017: \bibinfo{author}{\bibfnamefont{R.~J.} \bibnamefont{Cave}} \bibnamefont{and}
1018: \bibinfo{author}{\bibfnamefont{E.~R.} \bibnamefont{Davidson}},
1019: \bibinfo{journal}{J.~Chem.~Phys.} \textbf{\bibinfo{volume}{89}},
1020: \bibinfo{pages}{6798} (\bibinfo{year}{1988}).
1021:
1022: \end{thebibliography}
1023:
1024:
1025: \begin{table*}
1026: \begin{ruledtabular}
1027: \caption{Analysis of the most relevant configurations in the two lowest
1028: correlated cationic hole states $|\Psi_1\rangle$ and $|\Psi_2\rangle$ of
1029: Be$_3$H$_2$. The first columns give the occupation numbers of the involved
1030: spatial orbitals (Fig.~\ref{figvirtualorb}). The indices $S$ and $T$ refer to
1031: spin-adapted configurations with a singlet and triplet-like linear combination
1032: of the remaining electrons in a and b, respectively. The next two columns show
1033: the CI coefficients $\alpha_I$ and $\beta_I$ of the approximate correlated
1034: hole states $|\tilde{\Psi}_a\rangle$ and $|\tilde{\Psi}_b\rangle$,
1035: respectively, as defined in Eq.~(\ref{flhstates}), followed by the CI
1036: coefficients of the hole states $|\tilde{\Psi}_{\nu=1}\rangle$ and
1037: $|\tilde{\Psi}_{\nu=2}\rangle$ after the diagonalization step
1038: (\ref{diagonalization}). They are to be compared to the CI coefficients
1039: $\alpha_{1,I}^{{\rm MRCI}}$ and $\alpha_{2,I}^{{\rm MRCI}}$ of the ``exact''
1040: MRCI wave functions $|\Psi^{\rm MRCI}_{\nu=1}\rangle$ and $|\Psi^{\rm
1041: MRCI}_{\nu=2}\rangle$. \\}
1042: \begin{tabular}{llllll|cccccc|c}
1043: \multicolumn{6}{c|}{configuration $I$} &
1044: $|\tilde{\Psi}_a\rangle$ & $|\tilde{\Psi}_b\rangle$ &
1045: $|\tilde{\Psi}_1\rangle$ & $|\Psi^{\rm MRCI}_1\rangle$ &
1046: $|\tilde{\Psi}_2\rangle$ & $|\Psi^{\rm MRCI}_2\rangle$ & type of configuration
1047: \\[0.8ex]
1048: a & b & $|$ & a$^\ast$ & b$^\ast$ & c$^\ast$ &
1049: {\small $\alpha_I$} &
1050: {\small $\beta_I$} &
1051: {\small $\frac{1}{\sqrt{2}}(\alpha_I + \beta_I)$ } &
1052: {\small $\alpha^{\rm MRCI}_{1,I}$} &
1053: {\small $\frac{1}{\sqrt{2}}(\alpha_I - \beta_I) $} &
1054: {\small $\alpha^{\rm MRCI}_{2,I}$} &
1055: \\[0.8ex]
1056: \hline
1057: 1&2&$|$&0&0&0 & \phantom{-}0.9528 & -- & \phantom{-}0.6737 &
1058: \phantom{-}0.6767 & \phantom{-}0.6737 & \phantom{-}0.6682 & ref. configuration
1059: $| \Phi_a \rangle$\\
1060: 2&1&$|$&0&0&0 & -- & -0.9528 & -0.6737 & -0.6767
1061: & \phantom{-}0.6737 & \phantom{-}0.6682 & ref. configuration $| \Phi_b
1062: \rangle$
1063: \\[0.8ex]
1064: 1&1&$|$&1&0&0$^S$ & \phantom{-}0.0824 & -0.0901 & -0.0054 & -0.0052 & \phantom{-}0.1220 & \phantom{-}0.1450 \\
1065: 1&1&$|$&1&0&0$^T$ & \phantom{-}0.0595 & -0.1028 & -0.1148 & -0.1026 & \phantom{-}0.0306 & \phantom{-}0.0307 \\
1066: 1&1&$|$&0&1&0$^S$ & \phantom{-}0.0901 & -0.0824 & \phantom{-}0.0054 &
1067: \phantom{-}0.0052 & \phantom{-}0.1220 & \phantom{-}0.1450 & \raisebox{1.5ex}[-1.5ex]{mixed 2h1p configurations}\\
1068: 1&1&$|$&0&1&0$^T$ & -0.1028 & -0.0595 & -0.1148 & -0.1026 &
1069: -0.0306 & -0.0307 &
1070: \\[0.8ex]
1071: 0&2&$|$&0&1&0 & \phantom{-}0.0507 & -- & \phantom{-}0.0359 &
1072: \phantom{-}0.0302 & \phantom{-}0.0359 & \phantom{-}0.0451 & type I 2h1p configurations \\
1073: 2&0&$|$&1&0&0 & -- & -0.0507 & -0.0359 & -0.0302
1074: & \phantom{-}0.0359 & \phantom{-}0.0451 & type II 2h1p configurations
1075: \\[0.8ex]
1076: 1&0&$|$&0&0&2 & -0.0651 & -0.0298 & -0.0671 & -0.0617 & -0.0250 & -0.0297 \\
1077: 0&1&$|$&0&0&2 & \phantom{-}0.0298 & \phantom{-}0.0651 & \phantom{-}0.0671 & \phantom{-}0.0617 & -0.0250 & -0.0297& \raisebox{1.5ex}[-1.5ex]{3h2p configurations} \\
1078: \end{tabular}
1079: \label{tabconfig}
1080: \end{ruledtabular}
1081: \end{table*}
1082:
1083:
1084:
1085:
1086: \end{document}
1087: