1: % INJ
2:
3: %\documentclass[aps,pra,superscriptaddress,showpacs,preprint]{revtex4}
4: \documentclass[aps,prb,superscriptaddress,twocolumn]{revtex4}
5: %\documentclass[aps,prb,superscriptaddress,showpacs,preprint]{revtex4}
6: %\usepackage[draft]{graphicx}
7: \usepackage{graphicx}
8: \usepackage{amssymb}
9: \usepackage{amsmath}
10:
11: \newcommand{\bk}{{\bf k}}
12: \newcommand{\bq}{{\bf q}}
13: \newcommand{\bQ}{{\bf Q}}
14: \newcommand{\bp}{{\bf p}}
15: \newcommand{\bx}{{\bf x}}
16: \newcommand{\br}{{\bf r}}
17: \newcommand{\Li}{{\mathop{\rm{Li}}\nolimits}}
18: \renewcommand{\Im}{{\mathop{\rm{Im}}\nolimits\,}}
19: \renewcommand{\Re}{{\mathop{\rm{Re}}\nolimits\,}}
20: \newcommand{\sgn}{{\mathop{\rm{sgn}}\nolimits\,}}
21: \newcommand{\Tr}{{\mathop{\rm{Tr}}\nolimits\,}}
22: \newcommand{\EF}{E_{\mathrm{F}}}
23: \newcommand{\kB}{k_{\mathrm{B}}}
24: \newcommand{\kF}{k_{\mathrm{F}}}
25: \newcommand{\nF}{n_{\mathrm{F}}}
26: \newcommand{\Green}{{G}}
27: \newcommand{\dSC}{{\mathrm{dSC}}}
28: \newcommand{\dPG}{{\mathrm{dPG}}}
29: \newcommand{\dDW}{{\mathrm{dDW}}}
30: \newcommand{\Ret}{{\mathrm{R}}}
31: \newcommand{\Tau}{T_\tau}
32: \newcommand{\HF}{{\mathrm{HF}}}
33: \newcommand{\TF}{{\mathrm{TF}}}
34: \newcommand{\vW}{{\mathrm{vW}}}
35:
36: \begin{document}
37:
38: \title{Equilibrium geometries of low-lying isomers of some Li clusters,
39: within Hartree-Fock theory plus
40: bond order or MP2 correlation corrections}
41:
42: \author{A. Grassi}
43: \affiliation{Dipartimento di Scienze Chimiche, Facolt\`a di Farmacia,
44: Universit\`a di Catania,\\
45: Viale A. Doria, 6, I-95126 Catania, Italy}
46: \author{G. M. Lombardo}
47: \affiliation{Dipartimento di Scienze Chimiche, Facolt\`a di Farmacia,
48: Universit\`a di Catania,\\
49: Viale A. Doria, 6, I-95126 Catania, Italy}
50: \author{G. G. N. Angilella}
51: \affiliation{Dipartimento di Fisica e Astronomia, Universit\`a di
52: Catania,\\ and Istituto Nazionale per la Fisica della Materia,
53: UdR di Catania,\\ Via S. Sofia, 64, I-95123 Catania, Italy}
54: \author{N. H. March}
55: \affiliation{Department of Physics, University of Antwerp (RUCA),\\
56: Groenenborgerlaan 171, B-2020 Antwerp, Belgium}
57: \affiliation{Oxford University, Oxford, England}
58: \author{R. Pucci}
59: \affiliation{Dipartimento di Fisica e Astronomia, Universit\`a di
60: Catania,\\ and Istituto Nazionale per la Fisica della Materia,
61: UdR di Catania,\\ Via S. Sofia, 64, I-95123 Catania, Italy}
62:
63: \date{\today}
64:
65:
66: \begin{abstract}
67: \medskip
68: In a recent study by Kornath \emph{et al.} [J. Chem. Phys. {\bf 118},
69: 6957 (2003)], the Li$_n$ clusters with $n=2,4$ and $8$ have been
70: isolated in argon matrices at 15~K and characterized by Raman
71: spectroscopy.
72: This has prompted us to carry out a theoretical study on such clusters
73: up to $n=10$, using Hartree-Fock theory, plus low-order
74: M\o{}ller-Plesset perturbation corrections.
75: To check against the above study of Kornath \emph{et al.,} as a
76: by-product we have made the same approximations for $n=6$ and $8$
77: as we have for $n=10$.
78: This has led us to emphasize trends with $n$ through the Li$_n$
79: clusters for (i) ground-state energy, (ii) HOMO-LUMO energy gap,
80: (iii) dissociation energy, and (iv) Hartree-Fock eigenvalue sum.
81: The role of electron correlation in distinguishing between low-lying
82: isomers is plainly crucial, and will need a combination of
83: experiment and theory to obtain decisive results such as that of
84: Kornath \emph{et al.} for Li$_8$.
85: In particular, it is shown that Hartree-Fock theory plus bond order
86: correlations does account for the experimentally observed symmetry
87: T$_d$ symmetry for Li$_8$.
88: \end{abstract}
89:
90: \maketitle
91:
92: \section{Background and outline}
93:
94: Lithium clusters by now have been studied by a variety of experimental
95: techniques which include electron spin resonance, laser induced
96: fluorescence, depletion spectroscopy, photoionization, and Raman
97: spectroscopy \cite{Kornath:03}.
98: This Ref.~\onlinecite{Kornath:03} has investigated, in particular, the
99: lithium clusters Li$_n$, for $n=2,4$ and $8$.
100: This has been done by isolating these clusters in argon matrices at
101: 15~K.
102: Whereas Kornath \emph{et al.} \cite{Kornath:03} point out that most of
103: the techniques listed above are difficult to correlate with the
104: cluster geometries, their experimental work was able to determine
105: the geometry of Li$_4$ as a rhombic structure (D$_{2h}$) and for
106: Li$_8$ a hypertetrahedral structure (T$_d$) was shown to be in
107: agreement with their Raman studies.
108:
109: This experimental work has motivated the present theoretical study of
110: Li$_n$ clusters.
111: Most attention is focussed here on $n=10$, but we have also
112: included $n=6$ and $n=8$.
113: In the latter case, we must mention the previous theoretical work of
114: Bonacic-Koutecky, Fantucci and Koutecky \cite{Bonacic-Koutecky:88}.
115:
116: The outline of the paper is then as follows.
117: In Section~\ref{sec:1-10} some general trends of the electronic
118: structure of Li$_n$ clusters are depicted with $n$ ranging from $1$
119: to $10$.
120: We note that Kornath \emph{et al.} \cite{Kornath:03} studied $n=2$,
121: $4$, and $8$ extensively using high quality techniques.
122: For Li$_8$ our more modest techniques essentially confirmed
123: the findings of Kornath \emph{et
124: al.} \cite{Kornath:03} for the equilibrium geometry and the
125: vibrational frequencies.
126: So we focus in Section~\ref{sec:6-8} dominantly on Li$_6$.
127: Section~\ref{sec:10} presents results for Li$_{10}$.
128: Finally, discussion of some additional trends, plus some suggestions
129: for future work are covered in Section~\ref{sec:summary}.
130:
131: \section{Quantum chemically predicted trends in L\lowercase{i$_n$}
132: clusters for \lowercase{$n=1$} to $10$}
133: \label{sec:1-10}
134:
135: In this Section we present our results of the calculations made on
136: Li$_n$ clusters ($n=2,4,6,8,10$), by using the Gaussian package
137: (G03 Linux version) \cite{Frisch:03}.
138: All the calculations were performed using the standard self-consistent
139: field Hartree-Fock (HF) theory with the 6-311G$^\ast$ basis set.
140: The correlation energy was obtained using low-order M\o{}ller-Plesset (MP2)
141: corrections \cite{Moeller:34}, considering all (valence and core)
142: electrons.
143: For both HF and HF+MP2 level, full geometry optimization and
144: vibrational analysis were performed for all Li clusters.
145:
146: While Sections~\ref{sec:6-8} and \ref{sec:10} will develop further the
147: systematic studies we have made especially for $n=6$ and $10$, with
148: the approximate quantum chemical approach set out above, this
149: Section will present the general trends predicted by HF+MP2 theory
150: for $n=1$ to $10$.
151:
152: \begin{figure}[t]
153: \centering
154: \includegraphics[width=\columnwidth]{energybw.eps}
155: \caption{Ground-state energy per atom of Li$_n$ clusters, with
156: $n=1-10$.
157: \underline{\sl Inset:} accepted D$_{2h}$ planar structure of Li$_4$.
158: }
159: \label{fig:GS}
160: \end{figure}
161:
162:
163: To this end, Fig.~\ref{fig:GS} depicts the ground-state energy per
164: atom for increasing $n$ from $1$ to $10$.
165: No commentary is needed on the results for $n=1-4$, except to show as
166: an inset the accepted planar structure of Li$_4$.
167: The HF energy of this cluster is $-29.759$ and the lowering by
168: second-order M\o{}ller-Plesset perturbation theory yields $-29.859$,
169: both in Hartrees (as all energies below, unless stated otherwise).
170:
171: As seen in Fig.~\ref{fig:GS}, two structures were found to be
172: low-lying isomers of Li$_6$, with symmetries D$_{4h}$ and
173: D$_{3h}$, and these will be discussed in detail in
174: Section~\ref{sec:6-8} below.
175: Likewise, for Li$_8$, C$_s$ and T$_d$ structures emerge and again
176: Section~\ref{sec:6-8} presents details.
177: Li$_{10}$ has the lowest MP2 energy per atom of the clusters
178: considered and is the subject of Section~\ref{sec:10}.
179:
180:
181: \begin{figure}[b]
182: \centering
183: \includegraphics[width=\columnwidth]{gap.eps}
184: \caption{LUMO-HOMO gap for Li$_n$ clusters as a function of $n$.}
185: \label{fig:HL}
186: \end{figure}
187:
188:
189: Again concerning general trends, Fig.~\ref{fig:HL} shows the variation
190: of the HOMO-LUMO gap for Li$_n$ clusters with the number of atoms
191: $n$ in the lithium cluster.
192: It is of interest to note that though the variation is irregular for
193: $n=4-8$, the smallest gap is at $n=10$.
194: It is relevant to add here that solid lithium is metallic, with
195: therefore zero energy gap.
196: Of course the gap at $n=10$ is still $\sim 5$~eV!
197: It is the trend, not the absolute numbers, which should be the focus
198: here.
199:
200: With this relatively brief discussion of trends we turn to consider
201: our detailed results for Li$_6$ and, quite briefly, also Li$_8$ in
202: Section~\ref{sec:6-8} immediately below.
203:
204:
205: \section{Geometry, energy, and vibrational frequencies of
206: L\lowercase{i}$_6$ and L\lowercase{i}$_8$ using HF+MP2 theory}
207: \label{sec:6-8}
208:
209: In Ref.~\onlinecite{Kornath:03}, a comment was made in their experimental
210: considerations of a result which `may indirectly support the
211: absence of a Li$_6$ species.'
212: Motivated by this, we have, by geometry optimization, `converged' on
213: the geometries of two low-lying isomers which are depicted in
214: Fig.~\ref{fig:Li6}.
215: The first of these is the planar structure (Fig.~\ref{fig:Li6}, left),
216: with HF energy $-44.658$ and with addition of MP2 is $-44.810$.
217: The present HF+MP2 approach, however, predicts as the isomer we
218: identify as the ground state, the planar square with a `diatomic'
219: Li$_2$-like species perpendicular to the plane, through the center
220: of gravity.
221: While this has a (very) slightly higher HF energy than the planar
222: structure, it lies below that structure energetically when
223: `correlation' treated at the MP2 level is added, the results being
224: $E_\HF = -44.655$ and with the MP2 correction $-44.824$.
225:
226: \begin{figure}[t]
227: \centering
228: \includegraphics[width=0.49\columnwidth]{Li6_D3Hbw.eps}
229: \includegraphics[width=0.49\columnwidth]{Li6bw.eps}
230: \caption{Low-lying isomers of Li$_6$.
231: \underline{\sl Left:} Planar structure, with D$_{3h}$ symmetry.
232: \underline{\sl Right:} Most stable isomer predicted by HF+MP2, with
233: D$_{4h}$ symmetry, and
234: a dimer-like Li$_2$
235: structure perpendicular to the plane.
236: }
237: \label{fig:Li6}
238: \end{figure}
239:
240:
241:
242: Turning quite briefly to Li$_8$, the geometry was settled as T$_d$
243: from the Raman studies of Ref.~\onlinecite{Kornath:03}, whereas as can be
244: seen from Fig.~\ref{fig:GS}, the C$_s$ symmetry is predicted as
245: very slightly lower by our present approximation.
246: This is enough to illustrate the stringent test of many-electron
247: approximate theories that is afforded by attempts to predict
248: decisively the ground-state energies of isomers of Li$_n$ clusters.
249: However, the vibrational frequencies of Li$_8$ (T$_d$) are more in
250: line with the observed values than those of the C$_s$ structure,
251: which is somewhat encouraging in the light of the known geometry
252: (T$_d$) of the lowest lying isomer.
253:
254: We turn next to discuss, with fuller details, the Li$_{10}$ cluster.
255:
256:
257: \begin{figure}[t]
258: \centering
259: \includegraphics[width=0.49\columnwidth]{Li10bw.eps}
260: \includegraphics[width=0.49\columnwidth]{Li10abw_mod.eps}
261: \caption{Low-lying isomers of Li$_{10}$.
262: \underline{\sl Left:} Asymmetric isomer.
263: \underline{\sl Right:} Symmetric (D$_{2d}$) isomer.
264: The D$_{2d}$ isomer of Li$_{10}$ is characterized by
265: reflection symmetry with respect to the plane indicated in the
266: figure, and by 180$^\circ$ rotation symmetry with respect to the
267: C$_2$ axis.
268: See also \protect\cite{epaps:Li10} for calculated properties of both
269: isomers.}
270: \label{fig:Li10}
271: \end{figure}
272:
273:
274: \section{Geometry, energy, and vibrational frequencies of two low-lying
275: isomers of L\lowercase{i}$_{10}$}
276: \label{sec:10}
277:
278: We have carried out similar calculations which have led us to two
279: geometries of Li$_{10}$ which appear to be serious candidates for
280: low-lying isomers.
281: The first of these is the very asymmetrical structure shown in
282: Fig.~\ref{fig:Li10} (left).
283: Here, with an atom, roughly speaking, at the `center' of the cluster,
284: there can, of course, be no symmetrical arrangement around it.
285:
286:
287: We have evaluated the total energies, the
288: distance matrix (in \AA) and the calculated frequencies (in
289: cm$^{-1}$) of the geometrical structure shown in
290: Fig.~\ref{fig:Li10} (left) \cite{epaps:Li10}.
291: The second structure we found as a low-lying isomer is shown in
292: Fig.~\ref{fig:Li10} (right) \cite{epaps:Li10}.
293: This second isomer is characterized by D$_{2d}$ symmetry, and may be
294: thought as obtained from the Li$_8$ cluster, by `adding' two more
295: Li atoms (with, of course, different bond lengths and angles),
296: shown in Fig.~\ref{fig:Li10} (right) in the top-right-hand and
297: bottom-left-hand positions.
298:
299:
300: \begin{figure}[t]
301: \centering
302: \includegraphics[width=\columnwidth]{mp.eps}
303: \caption{Total energies $E_n$ versus eigenvalue sum $E_{sn}$ of Li$_n$
304: ($n=2-10$).
305: The straight line is Eq.~(\protect\ref{eq:MP})
306: \protect\cite{March:56}.
307: }
308: \label{fig:En}
309: \end{figure}
310:
311:
312: \section{Discussion and some future directions}
313: \label{sec:summary}
314:
315: Let us begin this discussion of the results presented above by dealing
316: first with further trends, to add to those reported in
317: Section~\ref{sec:1-10}.
318:
319: \subsection{Sum of eigenvalues ($E_{sn}$) related to total energy
320: $E_n$, for $n$ atom cluster}
321:
322:
323: In Fig.~\ref{fig:En} we have plotted for $n=2$ to $10$ the total
324: energies $E_n$ versus the eigenvalue sum $E_{sn}$ from the HF
325: occupied eigenvalues $\epsilon_{in}$, defined by
326: \begin{equation}
327: E_{sn} = \sum_{\mathrm{occupied}~i} \epsilon_{in} .
328: \end{equation}
329:
330: We find an approximately linear relation, which is compared with a
331: theoretical prediction from initially Thomas-Fermi statistical
332: theory, given by March and Plaskett \cite{March:56} for neutral
333: atoms:
334: \begin{equation}
335: E = \frac{3}{2} E_s .
336: \label{eq:MP}
337: \end{equation}
338: The result (\ref{eq:MP}) was found, independently, by Ruedenberg
339: \cite{Ruedenberg:77} from self-consistent field results for
340: molecules, and its foundation in density functional theory was
341: subsequently discussed by one of us \cite{March:77}.
342:
343: \subsection{Trends in dissociation energy $D_n$ with number of atoms
344: $n$ in Li$_n$ clusters defined as $|E_n - n E_1 |$}
345:
346: As a further trend through the clusters Li$_n$ with $n$ going from $2$
347: to $10$, we have collected in Table~\ref{tab:dissoc} results from
348: our HF+MP2 calculations for the dissociation energy $D_n$.
349:
350:
351: \begin{table}[b]
352: \centering
353: \begin{tabular}{ccccc}
354: \hline
355: $n$ & Symmetry & $D_n$ & $N$ & $D_n /N^2$ \\
356: \hline
357: 2 & & 0.02530 & 6 & 0.000702778 \\
358: 4 & & 0.07896 & 12 & 0.000548333 \\
359: 6 & D$_{3h}$ & 0.14086 & 18 & 0.000434753 \\
360: 6 & D$_{4h}$ & 0.15411 & 18 & 0.000475648 \\
361: 8 & T$_d$ & 0.23765 & 24 & 0.000412587 \\
362: 8 & C$_s$ & 0.24471 & 24 & 0.000424844 \\
363: 10 & D$_{2d}$ & 0.31460 & 30 & 0.000349555 \\
364: 10 & C$_1$ & 0.31843 & 30 & 0.000353811 \\
365: \hline
366: \end{tabular}
367: \caption{Dissociation energy $D_n = |E_n - nE_1 |$, where $E_n$ is
368: energy of Li$_n$ and $E_1$ that of isolated Li atom.
369: The second column refers to symmetry, when two low-lying isomers have
370: been treated, while the fourth column refers to the total number of
371: electrons, $N$.
372: All energies are in Hartree.
373: }
374: \label{tab:dissoc}
375: \end{table}
376:
377:
378: Mucci and March \cite{Mucci:83}, in early work, stressed the merit of
379: Teller's theorem \cite{Teller:62}, which states that
380: molecules/clusters do not bind
381: in any wholly local density approximation (LDA).
382: To avoid confusion with current terminology, `wholly LDA' in Teller's
383: theorem refers to also treating the kinetic energy $T$ by the
384: Thomas-Fermi (TF) result
385: \begin{equation}
386: T_\TF = c_k \int [\rho(\br)]^{5/3} \, d\br , \quad
387: c_k = \frac{3h^2}{10m} \left( \frac{3}{8\pi} \right)^{2/3} ,
388: \end{equation}
389: where $\rho(\br)$ denotes the electron density of the
390: molecule/cluster.
391: If $T$ denotes the correct (single-particle) kinetic energy at the
392: equilibrium geometry, Mucci and March \cite{Mucci:83} pointed out that
393: the difference between $T$ and $T_\TF$ was entirely due to electron
394: density gradients, \emph{e.g.} $\nabla\rho$, $\nabla^2 \rho$, etc.
395:
396: This idea was followed up by Allan \emph{et al.} \cite{Allan:85}, who
397: showed that $D/N^2$, with $N$ the total number of electrons in the
398: molecule, indeed correlated with the simplest energy
399: constructed from $\nabla\rho$, namely the von~Weizs\"acker (vW)
400: inhomogeneity kinetic energy defined by
401: \begin{equation}
402: T_\vW = \frac{1}{8} \int \frac{(\nabla\rho)^2}{\rho} \, d\br
403: .
404: \end{equation}
405: This is the first-order correction due to density gradients in the
406: difference $T-T_\TF$.
407: Therefore, in Fig.~\ref{fig:dissoc} we have plotted $D_n /N^2$,
408: already recorded in Table~\ref{tab:dissoc}, versus $T_\HF^{(n)}
409: -T_\TF^{(n)}$
410: (where $T_\HF^{(n)} =-E_\HF^{(n)}$ at equilibrium from the virial
411: theorem) for the series Li$_n$, with $n$
412: again running from $2$ to $10$.
413:
414:
415: \begin{figure}[t]
416: \centering
417: \includegraphics[width=\columnwidth]{dissenergy.eps}
418: \caption{Dissociation energy $D_n /N^2$, where $N$ is the total number
419: of electrons in cluster for Li$_n$, versus HF kinetic energy
420: $T_\HF^{(n)}$ minus Thomas-Fermi kinetic energy $T_\TF^{(n)}$.}
421: \label{fig:dissoc}
422: \end{figure}
423:
424:
425: Subsequently, one of us \cite{March:91a} has given the inhomogeneous
426: electron gas theory of molecular dissociation energies.
427: This reveals that $T_\vW^{1/6}$ is the crucial variable to
428: use in characterizing $D_n /N^2$.
429: Therefore, in Fig.~\ref{fig:dissoc} we have made a further plot of
430: $D_n /N^2$ already given in Table~\ref{tab:dissoc}, versus
431: $|T_\HF^{(n)} - T_\TF^{(n)} |$, the latter quantity being expected
432: to reflect $T_\vW^{(n)}$, albeit approximately.
433: There is a quite clear correlation from Fig.~\ref{fig:dissoc} between
434: $D/N^2$ and the `gradient difference' $T_\HF - T_\TF$,
435: substantiating the work reported in
436: Refs.~\onlinecite{Mucci:83,Allan:85,March:91a}.
437:
438: \subsection{Comments on individual clusters and especially role of
439: electron correlation}
440:
441: In cases when $n=6$ and $8$, two low-lying isomers have emerged from
442: our studies using the HF+MP2 approximation.
443: A strong pointer to the importance of electron correlation in any
444: quite decisive prediction of the true ground-state geometry is
445: afforded by the two structures, considered for Li$_8$, where
446: Kornath \emph{et al.} \cite{Kornath:03} have experimentally
447: verified from their Raman studies that the correct structure is
448: T$_d$.
449: This is indeed predicted to be lower in energy than C$_s$ in our HF
450: studies, but the situation is (wrongly!) changed over by treating
451: electron correlation at MP2 level.
452: However, as noted, HF+MP2 vibrational frequencies are considerably
453: higher for T$_d$ than for C$_s$, the latter clearly being in marked
454: disagreement with experiment!
455: On the other hand, a comparison of HF+MP2 energies and HF energy plus
456: bond-order correlation energy (Sec.~\ref{ssec:correlation})
457: shows that indeed inclusion of bond-order correlation allows to
458: confirm the experimentally observed T$_d$ structure for Li$_8$.
459:
460: Turning more briefly to Li$_6$, we find two low-lying isomers, both
461: with lower energies than either 6 isolated Li atoms or 3 isolated
462: Li$_2$ molecules.
463: Also stability with respect to isolated Li$_4$ and Li$_2$ components
464: is clear.
465: Again, however, while D$_{4h}$ lies higher in energy than D$_{3h}$ at
466: the HF level of approximation, including electron correlation at
467: order MP2 reverses the ordering.
468: Our conclusion here then which seems to us firm is that two low lying
469: isomers of Li$_6$ are found.
470: However, the energy ordering is not decisive, though our prejudice
471: here is in favor of the MP2 addition, namely D$_{4h}$ symmetry, but
472: that must remain conjecture until electron correlation is treated
473: by more refined approaches such as the coupled cluster
474: approximation.
475: Finally, the largest cluster studied here seems fairly `strongly'
476: bound, being stable again with respect to 10 Li atoms, 5 Li$_2$
477: dimers, and also the isolated `fragments' Li$_6$~$+$~Li$_4$ and
478: Li$_8$~$+$~Li$_2$.
479: None of this must be taken to mean that our structure for Li$_{10}$ is
480: the lowest-lying isomer of this cluster, though that is our
481: prediction at the HF+MP2 level.
482:
483: In the light of the importance of electron correlation, we shall now
484: present some `heuristic' ideas on this subject, in which appeal
485: will be made to bond-order versus bond-length relations which we
486: have used earlier for polyatomic molecules \cite{Grassi:96}.
487: We note here that the study of Kornath \emph{et al.} \cite{Kornath:03}
488: already includes extensive and careful \emph{ab initio} MO
489: calculations with high level of treatment of electronic correlations.
490:
491:
492: \subsection{Some approximate considerations on the magnitude of electron
493: correlation energy in the L\lowercase{i$_n$} clusters with
494: \lowercase{$n$} from $2$ to $10$}
495: \label{ssec:correlation}
496:
497:
498: Using the L\"owdin definition
499: \cite{Loewdin:55a,Loewdin:55b,Loewdin:55c} of electron correlation,
500: say $E_c^{\mathrm{L}}$, as
501: \begin{equation}
502: E_c^{\mathrm{L}(n)}= E_{\mathrm{exact}}^{(n)} - E_\HF^{(n)} ,
503: \end{equation}
504: we might use, as a first approximation to $E_c^{\mathrm{L}(n)}$ the
505: correlation energy in the K shells, namely
506: \begin{equation}
507: E_c^{\mathrm{L}(n)} \approx n E_c^{\mathrm{L}(1)} .
508: \label{eq:L2}
509: \end{equation}
510: One can utilize, for example, the recent study of Alonso \emph{et al.}
511: \cite{Alonso:03} to estimate $E_c^{\mathrm{L}(1)}$ for the Li atom
512: ground state.
513: Their pairing energy $E_{ss}$ is, for neutral atoms, given in their
514: Table~2 as $-1.28$~eV, and hence for the Li$_{10}$ cluster we
515: estimate (essentially from 10 separate K shells) the correlation
516: energy in magnitude to be $0.47$~Hartree.
517: The upper curve in Fig.~\ref{fig:linear} shows this K-shell--like
518: magnitude of $E_c$ for the
519: Li$_n$ clusters under consideration.
520: Of course, the original $2s$ electrons will form molecular orbitals
521: where again there is pairing of electrons with antiparallel spins,
522: and we therefore suppose that the uppermost curve in
523: Fig.~\ref{fig:linear} will be below the `true' magnitude of the
524: correlation energy curve.
525:
526: For comparison with this K shell estimate, we have plotted the
527: `correlation energy' given by HF+MP2 perturbation theory.
528: The circles in Fig.~\ref{fig:linear} show our results for Li$_n$
529: ($n=2,4,6,8$ and $10$).
530: Two low-lying isomers were considered in the present study of both
531: Li$_6$ and Li$_8$.
532: The small-dashed curve is proportional to $n$ as in Eq.~(\ref{eq:L2}),
533: whereas the long-dashed curve is linear in $n$ but does not pass
534: through the origin, in contrast to Eq.~(\ref{eq:L2}).
535:
536: Evidently, while MP2 corrections obviously improve the HF energies,
537: they yield only $\sim 50$ to $60$~\% the Li correlation energies,
538: the poorest result being for the dimer.
539:
540:
541: \begin{figure}[t]
542: \centering
543: \includegraphics[height=\columnwidth,angle=-90]{linear.ps}
544: \caption{Different estimates of the magnitude of the electron
545: correlation energy in Li$_n$ clusters versus $n$
546: (Sec.~\protect\ref{ssec:correlation}).
547: Solid line is based on `estimate' of $n$ `independent' K shell
548: contributions and here correlation energy is directly proportional
549: to $n$.
550: Open circles show results due to MP2 perturbation theory.
551: Two different `linear' approximations are represented by the dashed
552: lines.
553: Small-dashed line passes through origin at $n=0$.
554: Long-dashed line is a (slightly) better fit to the MP2 results shown
555: as circles.
556: Open triangles show results due to bond-order correlation, for which
557: the dash-dotted line is a guide for the eye.
558: For Li$_6$, Li$_8$ and Li$_{10}$, values for two low-lying isomers
559: studied in each case are shown.
560: }
561: \label{fig:linear}
562: \end{figure}
563:
564:
565: \begin{table}[b]
566: \centering
567: \begin{tabular}{cccccc}
568: \hline
569: $n$ & & $E_c^{\mathrm{BO}}$ & $E_c^{\mathrm{MP2}}$ &
570: $E_c^{\mathrm{BO}} - E_c^{\mathrm{MP2}}$ & $\epsilon_c
571: = E_c^{\mathrm{BO}} /Z$ \\
572: \hline
573: 4 & & $-0.27126$ & $-0.09887$ & $-0.17239$ & $-0.06782$ \\
574: 6 & D$_{3h}$ & $-0.41049$ & $-0.15191$ & $-0.25858$ & $-0.06842$ \\
575: 6 & D$_{4h}$ & $-0.42696$ & $-0.16724$ & $-0.25972$ & $-0.07116$ \\
576: 8 & T$_d$ & $-0.57063$ & $-0.23078$ & $-0.33985$ & $-0.07133$ \\
577: 8 & C$_s$ & $-0.57561$ & $-0.24506$ & $-0.33055$ & $-0.07195$ \\
578: 10 & D$_{2d}$ & $-0.72000$ & $-0.30339$ & $-0.41661$ & $-0.07200$ \\
579: 10 & C$_1$ & $-0.72041$ & $-0.31088$ & $-0.40953$ & $-0.07204$ \\
580: \hline
581: \end{tabular}
582: \caption{Correlation energies from both bond order
583: (BO) and MP2 calculations.
584: The second column lists the symmetry of the cluster.
585: All energies are in Hartree.
586: }
587: \label{tab:correlation}
588: \end{table}
589:
590:
591:
592: Following the technique reported in Ref.~\onlinecite{Grassi:96}, we have
593: calculated the correlation energies for lithium clusters using the
594: HF bond orders.
595: Fig.~\ref{fig:linear} compares our results with the values obtained
596: from MP2 calculations.
597: In particular, the experimental correlation energy
598: $E_c^{\mathrm{exp}}$ is defined as:
599: \begin{equation}
600: E_c^{\mathrm{exp}} = E_S - E_\HF ,
601: \end{equation}
602: with $E_S$ the exact Schr\"odinger energy and $E_\HF$ the Hartree-Fock
603: energy.
604: Following Ref.~\onlinecite{Grassi:96}, we calculate the theoretical
605: correlation energy $E_c^{\mathrm{BO}}$ as the sum of atomic and
606: bond contributions:
607: \begin{equation}
608: E_c^{\mathrm{BO}} = \sum_{i=1}^N E_S^A n_{\mathrm{eff}} +
609: \sum_{i=1}^{N-1} \sum_{j=i+1}^N a_{ij} B_{ij} ,
610: \label{eq:corrBO}
611: \end{equation}
612: with $N$ the number of atoms in the cluster.
613: The $E_S^A$ term is the atomic Schr\"odinger correlation energy that
614: can be obtained from experimental data, as explained in
615: Ref.~\onlinecite{Grassi:96}, $n_{\mathrm{eff}}$ denotes the effective
616: atomic electron number, which takes into account the electrons not
617: involved in the bonds, \emph{i.e.}
618: \begin{equation}
619: n_{\mathrm{eff}} = \frac{Z-n_{\mathrm{bond}}}{Z} ,
620: \end{equation}
621: with $Z$ the atomic number and $n_{\mathrm{bond}}$ the total number of
622: electrons involved in molecular bonds for each atom.
623: $B_{ij}$ is the bond order (BO) between atoms $i$ and $j$, as obtained
624: from HF calculations, and $a_{ij}$ is a parameter depending on the
625: particular bond $i-j$.
626: For Li clusters, this parameter was obtained from the experimental
627: correlation energy of Li$_2$ molecule and its value is
628: $0.06157$~a.u.
629: While up to now all binding and correlation energies have
630: been calculated within HF+MP2, with full geometry optimization, the
631: bond order correlation energies discussed here refer to the
632: optimized geometries within HF.
633:
634:
635: \begin{table*}[t]
636: \begin{tabular}{ccccccccc}
637: \hline
638: $n$ & & $E_\HF$ & $E_{\mathrm{BO}}$ &
639: $E_{\mathrm{MP2}}$ & & $\Delta E_\HF$
640: &
641: $\Delta E_{\mathrm{BO}}$ & $\Delta
642: E_{\mathrm{MP2}}$ \\
643: & & & & & & ($\times 1000$) & ($\times 1000$) & ($\times 1000$)\\
644: \hline
645: 4 & & $-29.75973$ & $-30.03099$ & $-29.85860$ & &
646: & & \\
647: 6 & D$_{3h}$ & $-44.65841$ & $-45.06890$ & $-44.81032$ & &
648: & & \\
649: 6 & D$_{4h}$ & $-44.65633$ & $-45.08329$ & $-44.82357$ &
650: D$_{4h}$~$-$~D$_{3h}$& $2.0800$ &
651: $-14.390$ & $-13.252$ \\
652: 8 & T$_d$ & $-59.56615$ & $-60.13678$ & $-59.79693$ & &
653: & & \\
654: 8 & C$_s$ & $-59.55893$ & $-60.13454$ & $-59.80399$ &
655: C$_{s}$~$-$~T$_{d}$ & $7.2191$
656: & $2.239$ & $-7.058$ \\
657: 10 & D$_{2d}$ & $-74.46031$ & $-75.18031$ & $-74.76370$ & &
658: & & \\
659: 10 & C$_1$ & $-74.45665$ & $-75.17706$ & $-74.76753$ &
660: C$_{1}$~$-$~D$_{2d}$ & $3.6587$ &
661: $3.249$ & $-3.827$ \\
662: \hline
663: \end{tabular}
664: \caption{Hartree-Fock and total energies from both bond order
665: (BO) and MP2 calculations, defined as $E_{\mathrm{BO}} = E_\HF +
666: E_c^{\mathrm{BO}}$ and $E_{\mathrm{MP2}} = E_\HF +
667: E_c^{\mathrm{MP2}}$.
668: The second column lists the symmetry of the cluster.
669: Last three columns list differences of these energies between clusters
670: with same $n$ but different symmetry (D$_{4h}$~$-$~D$_{3h}$, for
671: $n=6$; C$_{s}$~$-$~T$_{d}$, for
672: $n=8$; C$_{1}$~$-$~D$_{2d}$, for
673: $n=10$).
674: All energies are in Hartree.}
675: \label{tab:HFcorr}
676: \end{table*}
677:
678: Table~\ref{tab:correlation} reports the calculated correlation
679: energies for Li clusters obtained from MP2 calculations
680: ($E_c^{\mathrm{MP2}}$) and by using Eq.~(\ref{eq:corrBO})
681: ($E_c^{\mathrm{BO}}$), along with the HF energies.
682: As shown in Table~\ref{tab:correlation} and in Fig.~\ref{fig:linear},
683: the calculated correlation energies from Eq.~(\ref{eq:corrBO}) are
684: higher than the MP2 energies, giving for each Li cluster a total
685: molecular energy less than that calculated at HF+MP2 level.
686: Moreover, it is noteworthy that the difference $E_c^{\mathrm{MP2}} -
687: E_c^{\mathrm{BO}}$ increases with increasing number of atoms in the
688: cluster, and that the correlation energy per electron, $\epsilon_c
689: = E_c^{\mathrm{BO}} /Z$, is about constant for all clusters,
690: indicating that the total correlation energy is proportional to the
691: total number of electrons in the molecule.
692: In addition, a comparison of the total HF+MP2 and the HF+BO energies in
693: Table~\ref{tab:HFcorr} shows
694: that, while for $n=6$ both models predict the cluster with $D_{4h}$
695: symmetry to be stabler than the cluster with $D_{3h}$ symmetry, at
696: variance with the HF result,
697: both for $n=8$ and $n=10$ only the HF+BO energies do predict the
698: $T_d$ and $D_{2d}$ to be stabler than the $C_s$ and $C_1$,
699: respectively, in agreement with the purely HF results and, in the
700: case of Li$_8$, in agreement with the experimental result
701: \cite{Kornath:03}.
702:
703: Fig.~\ref{fig:linear} shows that the MP2 correlation energy estimated
704: with the 6-311G$^\ast$ basis set is systematically much smaller
705: than the sum of the $(1s)^2$ atomic pair correlation.
706: The referee has pointed out to us that this probably results from the
707: basic set deficiency for the inner shell correlation.
708:
709:
710: \begin{acknowledgments}
711: N.H.M. brought his contribution to the present study to fruition
712: during a visit to Catania in 2003.
713: He thanks the Department of Physics and Astronomy for the stimulating
714: atmosphere and for generous hospitality.
715: \end{acknowledgments}
716:
717: \medskip
718:
719: \begin{small}
720: \bibliographystyle{apsrev}
721: \bibliography{a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u,v,w,x,y,z,zzproceedings,Angilella,notes}
722: \end{small}
723:
724:
725: \end{document}
726: