1: \documentclass[a4paper, 12pt]{article}
2:
3: % eps inserting and fiddling
4: \usepackage{epsfig, psfrag}
5:
6: % labels
7: %\usepackage{showkeys}
8:
9: \usepackage{afterpage}
10:
11: % ams stuff
12: \usepackage{amsfonts, amsmath, amssymb, amsthm}
13:
14: % improves handling of maths in tables
15: \usepackage{tabls}
16: % allows entries over multiple rows in tables
17: \usepackage{multirow}
18: % allows use of rotated tables
19: \usepackage{rotating}
20:
21: % blackboard bold 1
22: \usepackage{bbm}
23:
24: \newcounter{thmNumber}[section]
25: \newcounter{lemmaNumber}[section]
26: \newcounter{corollaryNumber}[section]
27: \newcounter{definitionNumber}[section]
28: \numberwithin{thmNumber}{section}
29: \numberwithin{lemmaNumber}{section}
30: \numberwithin{corollaryNumber}{section}
31: \numberwithin{definitionNumber}{section}
32: \newtheorem{theorem}[thmNumber]{Theorem}
33: \newtheorem{definition}[thmNumber]{Definition}
34: \newtheorem{corollary}[corollaryNumber]{Corollary}
35: \newtheorem{lemma}[lemmaNumber]{Lemma}
36:
37: \newcommand{\R}{{\mathbb R}}
38: \newcommand{\C}{{\mathbb C}}
39: \newcommand{\Z}{{\mathbb Z}}
40: \newcommand{\N}{{\mathbb N}}
41:
42: \newcommand{\be}{\begin{equation}}
43: \newcommand{\ee}{\end{equation}}
44:
45: \newcommand{\mc}[1]{\multicolumn{2}{|c|}{#1}}
46: % place the entry over two columns of a table
47: % used in Eigenspaces
48:
49: \newcommand{\HNZ}{H^{N,Z}}
50: \newcommand{\ENZ}[1]{E^{N,Z}_{#1}}
51: \newcommand{\EkZ}[2]{E^{#1,Z}_{#2}}
52: \newcommand{\PsiNZ}[1]{\Psi^{N,Z}_{#1}}
53:
54: \newcommand{\HoNZ}{H^{N,Z}_0}
55: \newcommand{\HpNZ}{H'^{N,Z}}
56: \newcommand{\UNZ}{U_{N,Z}}
57: \newcommand{\HokZ}[1]{H^{#1,Z}_0}
58:
59: \newcommand{\PsiPT}{\Psi^{N,Z}_{PT,n}}
60: \newcommand{\EPT}{E^{PT}}
61: %\newcommand{\EPT}{E^{N,Z}_{PT,n}}
62: \newcommand{\PPT}{P_{N,Z}^{PT}}
63:
64: \newcommand{\XNpq}{X_{N, \mathbf{p}, \mathbf{q}}}
65: \newcommand{\SigmaNpq}{\Sigma_{N, \mathbf{p}, \mathbf{q}}}
66: \newcommand{\PNpq}{P_{N,\mathbf{p},\mathbf{q}}}
67:
68: \newcommand{\PsiCI}{\Psi^{N,Z}_{CI,n}}
69: \newcommand{\ECI}{E^{N,Z}_{CI,n}}
70: \newcommand{\PCI}{P_{N,Z}^{CI}}
71: \newcommand{\XN}{X_N}
72: \newcommand{\Xk}[1]{X_{#1}}
73: \newcommand{\SigmaN}{\Sigma_N}
74: \newcommand{\PCIN}{P_{N}^{CI}}
75:
76: \newcommand{\Rhat}{\hat{R}}
77: \newcommand{\LL}{{\bf\underline{L}}}
78: \newcommand{\SSS}{{\bf\underline{S}}}
79: \newcommand{\Etilde}{\tilde{E}}
80: \newcommand{\Span}{\,\mbox{Span}\,}
81: \newcommand{\Conf}{\mathcal C}
82:
83: \textheight=225mm
84: \voffset=-10mm
85:
86: \begin{document}
87:
88: \title{Explicit large nuclear charge limit of electronic ground states for Li, Be, B, C, N, O, F, Ne
89: and basic aspects of the periodic table}
90:
91: \author{Gero Friesecke
92: and Benjamin D. Goddard}
93:
94: \date{January 16, 2009}
95:
96: \maketitle
97:
98: \begin{abstract} This paper is concerned with the Schr\"odinger equation for atoms and ions with
99: $N=1$ to 10 electrons. In the asymptotic limit of large nuclear charge $Z$,
100: we determine explicitly the low-lying energy levels and eigenstates.
101:
102: The asymptotic energies and wavefunctions are in good quantitative agreement with experimental data for positive ions,
103: and in excellent qualitative agreement even for neutral atoms ($Z=N$). In particular, the predicted ground state spin and angular momentum
104: quantum numbers ($^1S$ for He, Be, Ne, $^2S$ for H and Li, $^4S$
105: for N, $^2P$ for B and F, and $^3P$ for C and O) agree with experiment in every case.
106:
107: The asymptotic Schr\"odinger ground states agree, up to small corrections, with the semi-empirical
108: hydrogen orbital configurations developed by Bohr, Hund and Slater to explain the periodic table.
109: In rare cases where our results deviate from this picture, such as the ordering of
110: the lowest ${}^1D^o$ and ${}^3S^o$ states of the Carbon isoelectronic
111: sequence, experiment confirms our, not Hund's, predictions.
112:
113: \end{abstract}
114:
115:
116: \section{Introduction}
117: How do the striking chemical differences between some elements, and the similarities between others,
118: emerge from the universal laws of quantum mechanics? In the physics and chemistry literature,
119: this fundamental question is discussed on a semi-empirical level, via the ``hydrogen orbital configurations''
120: developed by Bohr, Hund and Slater (see e.g. \cite{Bohr22, Hund25, Slater30, CondonShortley35, LanLif, Schwabl01, Atkins01}), or via numerical simulation
121: of simplified quantum mechanical models (see e.g. \cite{Hartree28, Hartree57, FroeseFischer77, Tatewaki94, NIST05, Tayloretal1}).
122: In this article, we address this question from a mathematical perspective.
123:
124: The theoretical possibility of making chemically specific predictions
125: was realized almost immediately after
126: the Schr\"odinger equation had been introduced (see e.g. \cite{Hartree28, Dirac29}).
127: But we are not aware of previous rigorous results, as mathematical research
128: on the basic quantum mechanical equations has hitherto focused overwhelmingly
129: on universal, qualitative properties.
130:
131: The perhaps most basic non-universal properties of atoms relevant to
132: chemical behaviour are the total spin and angular momentum quantum
133: numbers $S$ and $L$ of the ground state, which describe the amount
134: of symmetry under spin and spatial rotation. These two numbers not
135: only determine the ground state dimension $d$, but, as argued below,
136: they also allow to predict, up to at most two possibilities, the
137: group of the atom in the periodic table.
138:
139: Other quantities of interest include the energy levels $E_n$ and, more importantly,
140: energy differences such as spectral gaps
141: $E_n-E_m$ (which govern the photon frequencies which the atom can emit or absorb)
142: and the ionization energy (whose striking empirical periodicities
143: lay at the origin of the design of the periodic table).
144:
145: Our principal result is that such quantitites can be extracted analytically
146: from the many-electron Schr\"odinger equation in a natural scaling limit.
147: More precisely, we show that for ions with $N=1$ to 10 electrons,
148: %
149: as the nuclear charge $Z$ gets large the low-lying energy levels
150: and eigenstates converge to well defined limits,
151: which can be determined explicitly.
152: %
153: In particular, this yields rigorous values of $L$, $S$, $d$ for the ground state for all sufficiently large $Z$.
154: See Theorems \ref{GStheorem}, \ref{T:isolimit}, \ref{mainresult} and Tables \ref{Tab:GSqnumbers}, \ref{Tab:PTGS}, \ref{Tab:PTEigenspaces1},
155: \ref{Tab:PTEigenspaces2}. We call the above fixed-$N$, large-$Z$ limit {\it iso-electronic limit}, because
156: it is realized physically by an iso-electronic sequence such as Li, Be$^+$, B$^{++}$, ... Note that this limit
157: is different from the Thomas-Fermi limit $N=Z\to\infty$, which is
158: of interest in other contexts but does not retain any chemical specificity.
159:
160: The asymptotic levels and eigenstates are in good quantitative agreement with experimental data for positive ions,
161: and in excellent qualitative agreement even for neutral atoms ($Z=N$). In particular
162: the predicted values of $L$, $S$ and $d$ (see Table \ref{Tab:GSqnumbers}) agree with the experimental atomic values
163: \cite{NIST} in all cases.
164:
165: \begin{table}[ht]
166: \begin{center}
167: \resizebox{0.8\textwidth}{!}{
168: \begin{tabular}{|c|c|c|c|c|c|c|c|c|c|c|}
169: \hline
170: {Iso-electronic sequence} & H & He& Li&Be & B & C & N & O & F & Ne \\
171: \hline
172: $\sharp$ electrons & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 \\
173: \hline\hline
174: $L$ & 0 & 0 & 0 & 0 & 1 & 1 & 0 & 1 & 1 & 0 \\
175: \hline
176: $S$ & $\frac{1}{2}$ & 0 & $\frac{1}{2}$ & 0 & $\frac{1}{2}$ & 1 & $\frac{3}{2}$ & 1 & $\frac{1}{2}$ & 0 \\
177: \hline
178: {\tiny Chemist's notation} & {\tiny $^2S$} & {\tiny $^1S$} & \tiny{$^2S$} & {\tiny $^1S$} & {\tiny $^2P$} &
179: {\tiny $^3P$} & {\tiny $^4S$} & {\tiny $^3P$} & {\tiny $^2P$} & {\tiny $^1S$} \\
180: \hline
181: dim & 2 & 1 & 2 & 1 & 6 & 9 & 4 & 9 & 6 & 1 \\
182: \hline
183: \end{tabular}
184: } % end resizebox
185: \end{center}
186: \caption{Angular momentum and spin quantum numbers and dimension of the Schr\"odinger ground state for large $Z$,
187: as calculated in this paper.
188: All numbers agree with the experimental values for neutral atoms \cite{NIST}.}
189: \label{Tab:GSqnumbers}
190: \end{table}
191:
192: The asymptotic ground states we calculate (see Theorem \ref{GStheorem}) provide for the first time a mathematical
193: justification of the celebrated semi-empirical ``hydrogen orbital configurations'' developed
194: notably by Bohr, Hund and Slater to explain the periodic table. In our approach,
195: none of the underlying nontrivial postulates (electrons filling hydrogen
196: orbitals, shell and sub-shell formation, sub-shell ordering rules
197: such as 2s$<$2p, Hund's rules) need to be invoked, but are seen to emerge in a natural way.
198: The only corrections are as follows (see Sections \ref{Sec:Aufbau} and \ref{Sec:Comparison} for a detailed discussion):
199:
200: (1) Alongside each Slater determinant built from
201: admissible hydrogen orbitals, the ground state must contain its orbit under the symmetry group $SO(3)\times SU(2)\times \Z_2$
202: of the many-electron Schr\"odinger equation.
203:
204: (2) For the three elements Be, B, C, a ten to twenty percent admixture of a particular ``higher sub-shell''
205: configuration is also present, an effect we term $2s^2$--$2p^2$ resonance.
206:
207: (3) In rare cases, such as that of the
208: lowest ${}^1D^o$ and ${}^5S^o$ states of Carbon, the ordering of excited states disagrees with
209: the semi-empirical Hund's rules (with experiment confirming our orderings).
210:
211: %
212: %
213: %
214: We now outline our mathematical strategy to obtain explicit asymptotic energy levels and eigenstates, focusing for
215: simplicity on the ground state.
216:
217: The first step is the derivation of a simplified model governing the asymptotics. This can be done via
218: a scaling argument plus standard perturbation theory, as follows.
219: For fixed $N$ and large $Z$, attraction of an electron by the nucleus dominates
220: its interaction with the other electrons, so one expects
221: the true ground state to be close to the ground state of the corresponding system with electron interaction turned off
222: (which is known explicitly via hydrogen atom theory). After a little more thought, one realizes that this cannot be quite correct.
223: The non-interacting ground state eigenspaces of atoms
224: happen to be highly degenerate (see Table \ref{Tab:H0Dim}), but the underlying symmetry is broken by
225: the interaction, so the true ground state eigenspaces should
226: converge only to particular subspaces of them.
227: (Experimentally, this phenomenon is well known, from observed energy splittings.)
228: Mathematically, we prove that the difference between the Schr\"odinger ground states and the ground states
229: of the problem $PHP\Psi=E\Psi$, where $P$ is the projector onto the non-interacting ground state but
230: $H$ is the full Hamiltonian (eq. (\ref{ham}) below), tend to zero. We call this simplified problem {\it PT model},
231: because it corresponds to (i) re-scaling the problem so as to make the ground state of the reduced problem
232: $Z$-independent, (ii) applying degenerate first order perturbation theory, (iii) undoing the re-scaling.
233: Physically it corresponds to resolving, within the non-interacting ground state eigenspace,
234: the fully interacting problem.
235:
236: The second step is to determine the lowest PT eigenspace. This requires a careful analysis of the interplay
237: between hydrogen orbital formation (promoted by the Laplacian and electron-nuclei interaction), antisymmetry,
238: spin, and electron interaction. More technically,
239: the following difficulties arise. \\
240: (i) The non-interacting ground state, i.e. the state space of the PT model, is of somewhat daunting looking
241: dimension, e.g. 70 in case of Carbon (see Table \ref{Tab:H0Dim}). \\
242: (ii) The PT Hamiltonian $PHP$ is easy to write down abstractly (as we have just done), but unknown; one needs to
243: devise a method to determine it explicitly. \\
244: (iii) The PT model is a strongly interacting many-body model. \\
245: Difficulties (i) and (iii) are overcome via careful use of the
246: symmetry group of the original equation and its representation theory in terms of many-body spin and
247: angular momentum operators, which allows one to split the Hamiltonian $PHP$ into small invariant blocks. (ii) is addressed by combining
248: ideas from quantum chemistry which have not hitherto played a role in mathematical studies, such as Slater's rules \cite{SzaboOstlund96}
249: (which allow to express the components of the Hamiltonian via six-dimensional integrals of a product of four hydrogen eigenstates
250: and a Coulomb repulsion term), Fourier analysis (while the Fourier transform of individual hydrogen eigenstates is well known,
251: here one requires the Fourier transform of pointwise products of these), and residue calculus. In principle, our methods apply to
252: arbitrary atoms, except that the relevant $PHP$ matrices can become significantly higher dimensional.
253:
254: One curious mathematical phenomenon we observe is that the Hamiltonian
255: $PHP$ arising in the $Z\to\infty$ limit of quantum mechanics is always a rational matrix,
256: despite $H$ being a somewhat complicated partial differential operator and
257: $P$ a ``transcendental'' projector (onto tensor products of scaled
258: hydrogen eigenfunctions such as $\pi^{-1/2}e^{-|x|}$).
259:
260: The structure of this paper is as follows. In Section 2 we describe the basic quantum mechanical equations and their
261: symmetry group. In Section 3 we state the asymptotic limit of the Schr\"odinger ground states for
262: Li to Ne (see Theorem \ref{GStheorem}). In Sections 4 and 5, we justify the reduction to the
263: PT model and determine explicitly its state space.
264: Sections 6--7 contain more technical material: the explicit determination of the PT Hamiltonian and
265: the derivation of Theorem \ref{GStheorem}, as well as of the excited states and levels of the PT model. Finally,
266: sections \ref{Sec:Aufbau} and \ref{Sec:Comparison} compare
267: our results to experimental data and to methods in the physics and chemistry literature.
268:
269: %
270: %
271: %
272: \section{Schr\"odinger equation and mathematical definition of basic quantitites of chemical physics} \label{Sec:Basic}
273: %
274: %
275: The exact (nonrelativistic, Born-Oppenheimer) time-independent Schr\"odinger equation
276: for atoms and ions is
277: \be \label{SE}
278: H \Psi = E \Psi,
279: \ee
280: where, for nuclear charge $Z>0$ and $N$ electrons and in atomic units,
281: \be \label{ham}
282: H = \sum_{i=1}^N \Big( -\frac{1}{2} \Delta_{x_i} - \frac{Z}{|x_i|}
283: \Big) + \sum_{1 \leq i < j \leq N} \frac{1}{|x_i-x_j|},
284: \ee
285: $E\in\R$, and
286: \be \label{space}
287: \Psi \in L^2_a\big((\R^3 \times \Z_2)^N\big).
288: \ee
289: Here and below the $x_i\in\R^3$ are electronic coordinates, $s_i\in\Z_2=\{\pm\frac12\}$ are spin coordinates,
290: and $L^2_a$ is the usual Hilbert space of $N$-electron
291: functions $\Psi\, : \, (\R^3\times\Z_2)^N\to\C$ which are square-integrable,
292: \be \label{norm}
293: \int_{\R^{3N}}\sum_{(\Z_2)^N} |\Psi(x_1,s_1,\dots,x_N,s_N)|^2 = ||\Psi||^2 <\infty,
294: \ee
295: and satisfy the antisymmetry principle that, for all $i$ and $j$,
296: \be \label{anti}
297: \Psi(\dots, x_i,s_i, \dots, x_j,s_j, \dots)= -\Psi(\dots, x_j,s_j, \dots, x_i,s_i, \dots).
298: \ee
299: Mathematically, $H$ is a bounded below, self-adjoint operator with domain $L^2_a\cap H^2$, where
300: $H^2$ is the usual Sobolev space of $L^2$ functions with second weak derivatives belonging to $L^2$ \cite{Kat51}.
301:
302: We are interested in the mathematical derivation of a number of quantities of basic physical and chemical interest,
303: and begin by recalling how these are defined in terms of the Schr\"odinger equation (\ref{SE}).
304: \\[2mm]
305: %
306: %
307: %
308: %
309: {\bf Definitions, 1} An {\it energy level of an atom or ion} is an eigenvalue of the corresponding operator $H$.
310: An {\it eigenstate of the atom or ion} is an eigenstate of $H$ (i.e. a nonzero solution $\Psi$ to (\ref{SE})
311: belonging to the domain of $H$). By Zhislin's theorem (\cite{Zhislin60}, see \cite{Friesecke03} for a short proof),
312: for atoms ($N=Z$) and positive ions ($N<Z$) there exist countably many energy levels $E_1<E_2<...$ below the bottom of
313: the essential spectrum of $H$, the corresponding eigenspaces being finite-dimensional. $E_1$ is called the {\it ground state
314: energy} and the corresponding eigenspace is known as the {\it ground state}. Eigenspaces corresponding to the higher energy levels
315: are known as {\it excited states}. The {\it excitation energy} or {\it spectral gap} of an excited state with energy $E_m$ is
316: defined to be $E_m-E_1$. Physically it corresponds to the energy required to promote the electrons from the ground state to the
317: excited state.
318: \\[2mm]
319: %
320: %
321: %
322: %
323: Besides the quantized energy levels $E_n$, there exist important additional discrete quantum numbers associated with
324: the atomic Schr\"odinger equation
325: which arise from its symmetries. Their precise definition, albeit very natural, takes
326: a little more work.
327: \\[2mm]
328: The model (\ref{SE}), (\ref{ham}), (\ref{norm}), (\ref{anti}) is invariant under
329: \begin{itemize}
330: \item [(i)] simultaneous rotation of all electron positions about the origin, \\
331: $\Psi(x_1,s_1,..,x_N,s_N)\mapsto\Psi(R^Tx_1,s_1,..,R^Tx_N,s_N)$, $R\in SO(3)$
332: \item [(ii)] simultaneous rotation of all electron spins \\
333: (by a unitary matrix $U\in SU(2)$)
334: \item [(iii)] simultaneous inversion of all electron positions at the origin, \\
335: $\Psi(x_1,s_1,..,x_N,s_N)\mapsto\Psi(-x_1,s_1,..,-x_N,s_N)=:\Rhat
336: \Psi$.
337: \end{itemize}
338: (In group theory language, the symmetry group is $SO(3)\times SU(2) \times \Z_2$, the third factor being the inversion group
339: consisting of $\Rhat$ and the identity. When $N=1$, there exists an additional symmetry, which gives rise to conservation of
340: the quantized Runge-Lenz vector; but it is broken by the interaction term in (\ref{ham}) when $N\ge 2$. Note also that even though
341: the Hamiltonian (\ref{ham}) is invariant under the larger group of non-simultaneous rotation of spins,
342: the antisymmetry condition (\ref{anti}) is not.)
343: %
344: %
345: %
346:
347: The conserved quantities, i.e., operators which commute with the Hamiltonian, which arise from the above symmetries are
348: \begin{itemize}
349: \item[(i)]
350: $\LL=\sum_{j=1}^N\LL(j)$ (many-electron angular momentum operator)
351: \item[(ii)]
352: $\SSS=\sum_{j=1}^N\SSS(j)$ (many-electron spin operator)
353: \item[(iii)]
354: $\Rhat$ (parity operator),
355: \end{itemize}
356: where
357: $$
358: \LL(j)=\begin{pmatrix} L_1(j) \\ L_2(j) \\ L_3(j) \end{pmatrix},
359: \quad
360: \SSS(j)=\begin{pmatrix} S_1(j) \\ S_2(j) \\ S_3(j) \end{pmatrix},
361: $$
362: and $L_\alpha(j)$, $S_\alpha(j)$ ($\alpha=1,2,3$) denote the usual angular momentum respectively spin operators acting on the $j^{th}$
363: coordinate. Explicitly, on $N$-electron states $\Psi(x_1,s_1,..,x_N,s_N)$, $x_j\in\R^3$, $s_j\in\{\pm \frac12\}$, and denoting
364: $x_j=(y^{(1)}, y^{(2)}, y^{(3)})$, $L_\alpha(j)$ is the partial differential operator
365: \be \label{Lj}
366: L_\alpha(j) = \frac1i \Bigl( y^{(\alpha+1)}\frac{\partial}{\partial y^{(\alpha-1)}} - y^{(\alpha-1)}\frac{\partial}{\partial y^{(\alpha+1)}}
367: \Bigr),
368: \ee
369: and $S_\alpha(j)$ is multiplication by a Pauli matrix,
370: $$
371: \left(\begin{array}{c} (S_\alpha(j)\Psi)(x_1,s_1,\ldots,
372: x_j,\mbox{$\frac12$},\ldots,x_N,s_N) \\ (S_\alpha(j)\Psi)
373: (x_1,s_1,\ldots,
374: x_j,-\mbox{$\frac12$},\ldots,x_N,s_N)
375: \end{array}\right)
376: = \sigma_\alpha
377: \left(\begin{array}{c} \Psi(x_1,s_1,\ldots,
378: x_j,\mbox{$\frac12$},\ldots,x_N,s_N) \\ \Psi(x_1,s_1,\ldots,
379: x_j,-\mbox{$\frac12$},\ldots,x_N,s_N)
380: \end{array}\right),
381: $$
382: where the $\sigma_\alpha$ are the Pauli matrices
383: $$
384: \sigma_1:=\frac12\begin{pmatrix} 0 & 1 \\ 1 & 0 \end{pmatrix}, \quad
385: \sigma_2:=\frac12\begin{pmatrix} 0 & -i \\ i & 0 \end{pmatrix}, \quad
386: \sigma_3:=\frac12\begin{pmatrix} 1 & 0 \\ 0 & -1 \end{pmatrix}.
387: $$
388: The fact that the operators (i), (ii), (iii) commute with the Hamiltonian (\ref{ham})
389: can be checked by direct inspection using the above formulae, without reference to the underlying symmetry group.
390:
391: The components of total angular momentum and total spin,
392: $L_\alpha=\sum_{j=1}^NL_\alpha(j)$ and $S_\alpha=\sum_{j=1}^NS_\alpha(j)$,
393: obey the usual commutator relations
394: $$
395: [L_\alpha,L_\beta]=iL_\gamma, \;\;\; [S_\alpha,S_\beta]=iS_\gamma \;\;\;\; (\alpha,\beta,\gamma \mbox{ cyclic}).
396: $$
397: Angular momentum representation theory, together with
398: simple considerations concerning the above specific action of the operators on $N$-electron states,
399: yields the following well known facts (see e.g. \cite{Friesecke}).
400: %
401: %
402: %
403: \begin{lemma} \label{symlem} (a) For arbitrary $N$ and $Z$, a set of operators which commutes with the Hamiltonian $H$
404: and with each other is given by
405: \be \label{ops}
406: \LL^2, \, L_3, \, \SSS^2, \, S_3, \, \Rhat.
407: \ee
408: (b) The eigenvalues of $\LL^2$, $\SSS^2$, and $\Rhat$ (acting on $L^2_a((\R^{3}\times\Z_2)^N)$) are, respectively,
409: \begin{align}
410: & L(L+1), \;\; L=0,1,2,..., \\
411: & S(S+1), \;\; S=\left\{\begin{array}{ll} \mbox{$\frac12$},\,\mbox{$\frac32$},\,\mbox{$\frac52$},...,\mbox{$\frac{N}{2}$}, & N \mbox{ odd}, \\
412: 0, \, 1, \, 2,...,\mbox{$\frac{N}{2}$}, & N \mbox{ even} \end{array}\right. \\
413: & p=\pm 1.
414: \end{align}
415: (c) For fixed $L$, $S$ and $p$, on any joint eigenspace of $H$, $\LL^2$, $\SSS^2$ and $\Rhat$, $L_3$ has eigenvalues
416: $
417: M=-L,-L+1,...,L,
418: $
419: and $S_3$ has eigenvalues
420: $
421: M_S = -S, -S+1,...,S.
422: $
423: In particular, the eigenspace has dimension greater
424: or equal to $(2L+1)\cdot (2S+1)$, with equality in the case when the joint eigenspaces of $H$ and the operators (\ref{ops})
425: are non-degenerate (i.e., one-dimensional).
426: \end{lemma}
427: %
428: %
429: %
430: \noindent
431: Here we have employed the usual notation $\LL^2=L_1^2+L_2^2+L_3^2$ (and analogously for $\SSS^2$).
432:
433: From the above we see that the main differences between the symmetries of many-electron atoms and those of hydrogen are the absence of an analogon of the
434: quantized Runge-Lenz vector and the nontrivial action of the spin operator
435: $\SSS^2$ (for $N=1$ it is equal to the trivial operator $\frac34 I$).
436: %
437: %
438: %
439: \\[2mm]
440: {\bf Definitions, 2.} The values of $L$, $S$ and $p$ for eigenstates of $\LL^2$, $\SSS^2$, $\Rhat$ are called the {\it total angular momentum quantum
441: number}, the {\it total spin quantum number}, and the {\it parity} of the state.
442: \\[2mm]
443: From the above lemma we see that for each energy level of $H$ there
444: exist unique quantum numbers $L$, $S$, $p$ characterizing the
445: symmetry of the eigenspace (except in ``non-generic'' cases where
446: the joint eigenspaces of $H$ and the operators (\ref{ops}) are
447: degenerate, in which case there exists a unique finite set of such
448: quantum numbers). In the chemistry literature these numbers for an
449: energy level are usually given in the form $^{2S+1}X^{\nu}$, where
450: $L$ corresponds to $X$ via $0\to S$, $1\to P$, $2\to D$, $3\to F$,
451: and where no superscript $\nu$ means $p=1$, and $\nu=o$ (for odd)
452: stands for $p=-1$. For example, the Carbon values $L=1$, $S=1$,
453: $p=1$ from Table \ref{Tab:GSqnumbers} would be denoted $^3P$, and
454: the Nitrogen values $L=0$, $S=3/2$, $p=-1$ by $^4S^o$.
455:
456: Of particular physical and chemical interest is the energetic ordering in which different combinations of $L$ and $S$ appear in the spectrum of
457: the Hamiltonian (\ref{ham}); see Section \ref{Sec:AtSpec}.
458: %
459: %
460: %
461: %
462: %
463: \section{Ground states of the first ten atoms in the large $Z$ limit}
464: We are now in a position to state
465: a principal result of this paper.
466: \\[2mm]
467: {\bf Notation}
468: $|\eta_1\dots\eta_N\rangle$ denotes the Slater determinant (or antisymmetrized tensor product) of the orbitals
469: $\eta_j\!\in\! L^2(\R^3\times\Z_2)$,
470: $|\eta_1...\eta_N\rangle(x_1,s_1,..,x_N,s_N)={(N!)^{-1/2}}\det(\eta_i(x_j,s_j)_{i,j=1}^N)$.
471: $\phi\!\uparrow$, $\phi\!\downarrow$ stands for the spin-up and spin-down orbitals
472: $\phi(x)\delta_{\pm 1/2}(s)\in L^2(\R^3\times\Z_2)$. For a linear operator on the $N$-electron Hilbert space
473: (\ref{space}), $|||A|||$ denotes the usual operator norm $\sup\{||A\Psi|| \, : \, \Psi\in
474: L^2_a((\R^3\times\Z_2)^N), \, ||\Psi||=1\}$.
475: %
476: %
477: %
478: \begin{theorem} \label{GStheorem} Consider the atomic Schr\"odinger equation (\ref{SE}),
479: (\ref{ham}), (\ref{space}) for the atom/ion with $N$=1 to 10 electrons and nuclear charge $Z$. \\[1mm]
480: (i) For sufficiently large $Z$, the ground state has the spin and angular momentum quantum number $S$, $L$
481: and the dimension given in Table \ref{Tab:GSqnumbers}. \\[1mm]
482: %
483: (ii) In the limit $Z\to\infty$, the ground state is asymptotic to the explicit vector space given in Table
484: \ref{Tab:PTGS}, in the sense that the projection operators $P_0$, $\tilde{P}_0$ onto these spaces satisfy
485: $\lim_{Z\to\infty}|||P_0-\tilde{P}_0|||=0$.
486: Here $1$, $2$, $3$, $4$, $5$ stand for the scaled hydrogen orbitals (mathematically: hypergeometric functions)
487: $
488: \phi_{1s}\!\!\uparrow, \;\phi_{2s}\!\!\uparrow, \;
489: \phi_{2p_3}\!\!\uparrow, \; \phi_{2p_1}\!\!\uparrow, \; \phi_{2p_2}\!\!\uparrow
490: $
491: from (\ref{PTorbitals}), (\ref{porbitals}), and $\overline{1}$, $\overline{2}$, $\overline{3}$, $\overline{4}$, $\overline{5}$ denote the
492: corresponding spin-down orbitals.
493: \end{theorem}
494:
495: \begin{table}[h!]
496: \resizebox{\textwidth}{!}{
497: \begin{tabular}{|c|c|c|c|}
498: \hline
499: $\mbox{Iso-electronic}\atop{\mbox{sequence}}$ & Symmetry & Ground state & Dimension \\
500: \hline
501: H & $^2S$ & $|1\rangle$, $|\overline{1}\rangle$ & 2 \\
502:
503: \hline He & $^1S$ & $|1\overline{1}\rangle$ & 1 \\
504:
505: \hline Li & $^2S$ & $|1\overline{1}2\rangle$,
506: $|1\overline{1}\overline{2}\rangle$ & 2 \\
507:
508: \hline Be & $^1S$ & $\frac{1}{\sqrt{1+c^2}}\Big(
509: |1\overline{1}2\overline{2}\rangle + c \frac{1}{\sqrt{3}}\big(
510: |1\overline{1}3\overline{3}\rangle +
511: |1\overline{1}4\overline{4}\rangle +
512: |1\overline{1}5\overline{5}\rangle\big) \Big)$ & 1 \\
513: & & $c = - \frac{\sqrt{3}}{59049}(2\sqrt{1509308377} - 69821)
514: = -0.2310995\dots$ & \\
515:
516: \hline B & $^2P^o$ & $\frac{1}{\sqrt{1+c^2}}\Big(
517: |1\overline{1}2\overline{2}i\rangle + c \frac{1}{\sqrt{2}} \big(
518: |1\overline{1}ij\overline{j}\rangle +
519: |1\overline{1}ik\overline{k}\rangle \big) \Big)$ & 6 \\
520: & & $\frac{1}{\sqrt{1+c^2}}\Big(
521: |1\overline{1}2\overline{2}\overline{i}\rangle + c
522: \frac{1}{\sqrt{2}} \big(
523: |1\overline{1}\overline{i}j\overline{j}\rangle +
524: |1\overline{1}\overline{i}k\overline{k}\rangle \big) \Big)$ & \\
525: & & $(i,j,k)=(3,4,5),(4,5,3),(5,3,4)$ & \\
526: & & $c = -\frac{\sqrt{2}}{393660}(\sqrt{733174301809}-809747)
527: =-0.1670823\dots$ & \\
528:
529: \hline C & $^3P$ & $\frac{1}{\sqrt{1+c^2}}\big(
530: |1\overline{1}2\overline{2}ij\rangle + c
531: |1\overline{1}k\overline{k}ij\rangle \big)$ & 9 \\
532: & & $\frac{1}{\sqrt{1+c^2}}\Big( \frac{1}{\sqrt{2}}\big(
533: |1\overline{1}2\overline{2}i\overline{j}\rangle +
534: |1\overline{1}2\overline{2}\overline{i}j\rangle\big) +
535: c\frac{1}{\sqrt{2}}\big(
536: |1\overline{1}k\overline{k}i\overline{j}\rangle +
537: |1\overline{1}k\overline{k}\overline{i}j\rangle\big) \Big)$ & \\
538: & & $\frac{1}{\sqrt{1+c^2}}\big(
539: |1\overline{1}2\overline{2}\overline{i}\overline{j}\rangle + c
540: |1\overline{1}k\overline{k}\overline{i}\overline{j}\rangle \big)$ & \\
541: & & $(i,j,k)=(3,4,5),(4,5,3),(5,3,4)$ & \\
542: & & $c= - \frac{1}{98415}(\sqrt{221876564389}-460642)
543: =-0.1056317\dots$ & \\
544:
545: \hline N & $^4S^o$ & $|1\overline{1}2\overline{2}345\rangle$ & 4 \\
546: & &
547: $\frac{1}{\sqrt{3}}(|1\overline{1}2\overline{2}\overline{3}45\rangle
548: + |1\overline{1}2\overline{2}3\overline{4}5\rangle +
549: |1\overline{1}2\overline{2}34\overline{5}\rangle)$ & \\
550: & &
551: $\frac{1}{\sqrt{3}}(|1\overline{1}2\overline{2}\overline{3}\overline{4}5\rangle
552: + |1\overline{1}2\overline{2}\overline{3}4\overline{5}\rangle +
553: |1\overline{1}2\overline{2}3\overline{4}\overline{5}\rangle)$ & \\
554:
555: & & $|1\overline{1}2\overline{2}\overline{3}\overline{4}\overline{5}\rangle$ & \\
556:
557: \hline O & $^3P$ &
558: $|1\overline{1}2\overline{2}i\overline{i}jk\rangle$ & 9 \\
559: & &
560: $\frac{1}{\sqrt{2}}(|1\overline{1}2\overline{2}i\overline{i}j\overline{k}\rangle
561: + |1\overline{1}2\overline{2}i\overline{i}\overline{j}k\rangle)$ & \\
562: & & $|1\overline{1}2\overline{2}i\overline{i}\overline{j}\overline{k}\rangle$ & \\
563: & & $(i,j,k)=(3,4,5),(4,5,3),(5,3,4)$ & \\
564:
565: \hline F & $^2P^o$ &
566: $|1\overline{1}2\overline{2}i\overline{i}j\overline{j}k\rangle$ & 6 \\
567: & & $|1\overline{1}2\overline{2}i\overline{i}j\overline{j}\overline{k}\rangle$ & \\
568: & & $(i,j,k)=(3,4,5),(4,5,3),(5,3,4)$ & \\
569:
570: \hline Ne & $^1S$ &
571: $|1\overline{1}2\overline{2}3\overline{3}4\overline{4}5\overline{5}\rangle$ & 1 \\
572: \hline
573: \end{tabular}
574: } % end resizebox
575: \caption{Ground states of the atomic Schr\"odinger equation in the limit $Z\to\infty$.
576: The indicated wave functions are an orthonormal basis of the ground
577: state. See Theorem \ref{GStheorem} for notation. The symmetry agrees with experiment
578: for each sequence and all $Z$.}\label{Tab:PTGS}
579: \end{table}
580:
581: \noindent
582: In fact, all low-lying energy levels and eigenstates can be determined exactly in the above limit.
583: See Theorem \ref{mainresult}. Note also that, as we will see below,
584: the asymptotic ground states in the table are the exact ground states of the limit model (\ref{PT}),
585: (\ref{H0H'}), and become $Z$-independent after the re-scaling (\ref{scaling}).
586:
587: The derivation of these results requires five steps: \\[1mm]
588: 1. Reduction to the finite-dimensional problem $PHP\Psi=E\Psi$, $\Psi\in V_0(N)$ described in (\ref{PT}), (\ref{H0H'}). \\[1mm]
589: 2. Explicit determination of $V_0(N)$. \\[1mm]
590: 3. Choice of a suitable basis of $V_0(N)$ making $PHP$ particularly simple. \\[1mm]
591: 4. Explicit determination of the $d\times d$ (i.e., in case of Carbon, $70\times 70$) matrix
592: representing the Hamiltonian $PHP$ from eq. (\ref{PT}) in this basis. \\[1mm]
593: 5. Spectral analysis of this matrix.
594: \\[1mm]
595: These steps are carried out in Sections \ref{Sec:Pert}, \ref{Sec:GS}, \ref{S:Eigenspaces},
596: \ref{Sec:SymbolicVee}--\ref{Sec:ExplicitPHP}, and \ref{Sec:AtSpec}. Steps 1. and 2. follow from standard perturbation
597: theory respectively hydrogen atom theory.
598: 3. is achieved by a basis adapted to the symmetries of $PHP$ (see Lemma \ref{symlem}) leading to block diagonal structure.
599: 4. exploits, in addition, the fact that the Hamiltonian contains only one-body and two-body terms, allowing
600: to reduce evaluation of the required $N$-electron matrix elements $\langle\Psi|H|\tilde{\Psi}\rangle$, which are integrals
601: over $\R^{3N}$, to $1$- and $2$-electron matrix elements.
602: %
603: %
604: %
605:
606: \subsection{Comparison with the semi-empirical Bohr-Hund-Slater picture of the periodic table}
607: \label{Sec:Aufbau} The result of Theorem \ref{GStheorem} provides a mathematical justification of the semi-empirical
608: ``Aufbau principle'' (from the German word for building up)
609: developed notably by Bohr, Hund, and Slater to explain the periodic
610: table \cite{Bohr22, Hund25, LanLif, Schwabl01, Atkins01}. The aufbau principle is based on three semi-empirical postulates: \\[1mm]
611: (a) Each electron in an atom occupies a hydrogenic orbital.\footnote{In fact, in
612: Bohr's and Hund's original works \cite{Bohr22, Hund25}, which narrowly predate the Schr\"odinger equation,
613: the electrons were supposed to occupy hydrogenic Bohr orbits.}
614: \\[1mm]
615: (b) {\it Sub-shell ordering} The orbitals in each hydrogen energy level, or shell, form sub-shells which
616: are occupied in the order
617: $1s \;\;\; 2s \;\;\; 2p \;\;\; 3s \;\;\; 3p \;\;\; 4s \;\;\; 3d\;\;\; \dots$
618: \\[1mm]
619: (c) {\it Hund's rule} Within any partially filled sub-shell, the electrons adopt a configuration with the
620: greatest possible number of aligned spins.
621: \\[1mm]
622: Thus, for example, in Carbon the six electrons
623: would occupy the orbitals $1s\!\uparrow 1s\!\downarrow 2s\!\uparrow 2s\!\downarrow 2p_1\!\uparrow 2p_2\!\uparrow$ (note
624: that the alternative choices $2p_1\!\downarrow$ or $2p_2\!\downarrow$ for the last orbital would be consistent with (b)
625: but not (c)).
626:
627: This beautiful heuristic picture is seen to emerge in Theorem \ref{GStheorem} in a natural way,
628: without reliance on the above nontrivial postulates or numerical simulations (up to small but interesting corrections).
629:
630: (i) For seven out of ten elements (H, He, Li, N, O, F, Ne), the Aufbau principle configuration (when interpreted not as a collection of
631: individual electronic states, but as a Slater-determinantal many-electron wavefunction) is an element of the asymptotic
632: Schr\"odinger ground state.
633:
634: (ii) For the remaining three elements (Be,
635: B, C), the Aufbau principle configuration is the dominant part of an element of the asymptotic Schr\"odinger ground state.
636:
637: The corrections to the semi-empirical rules emerging in the large $Z$ limit of quantum mechanics are:
638: \\[1mm]
639: (1)
640: Alongside any Slater determinant, the asymptotic ground state contains
641: its orbit under the symmetry group of the many-electron Schr\"odin\-ger equation (see Lemma \ref{symlem}).
642: \\[1mm]
643: (2)
644: The corrections to the Aufbau principle configurations in (ii) come
645: from different sub-shells, indicating that rule (b)
646: is not strictly obeyed. We term this effect, which does not seem to have received
647: mathematical attention, {\bf 2s$^2$--2p$^2$ resonances}: besides the aufbau principle configurations
648: $1s^22s^2$, $1s^22s^22p$, and $1s^22s^22p^2$, a significant percentage is also present of,
649: respectively, $1s^22p^2$, $1s^22p^3$, and $1s^22p^4$. This could be
650: described as a resonance of the standard configuration with a
651: configuration in which the two 2s orbitals have been substituted by
652: two 2p orbitals. Why this phenonenon occurs only for Be, B, C has a simple group-theoretic reason:
653: Tables \ref{Tab:VeeMatrix1}, \ref{Tab:VeeMatrix2} show that such a substitution
654: which preserves the total quantum numbers $L$ and $S$ is only possible in these three cases.
655: \\[1mm]
656: (3)
657: For excited states, Hund's rules are in rare cases found to disagree with the
658: experimental and mathematical results; see Section \ref{Sec:Comparison}.
659: %
660: %
661: \subsection{Comparison with other approaches} First we comment on the asymptotic regime in which the above picture emerges,
662: namely $N$ fixed, $Z\to\infty$. These parameters are the only ones contained in
663: the electronic Schr\"odinger equation that can be varied in ground states of a physical system. To capture chemical
664: specificity, $N$ must be kept fixed.
665: A priori, $Z$ could be made either small or large; but making it small leads to non-existence of bound electrons \cite{Lieb84}.
666: This leaves
667: $Z\to\infty$, i.e. the limit studied above, as the only option.
668:
669: This limit has attracted considerable previous attention in the physics and chemistry literature in connection with
670: asymptotic expansions of energies and numerical evaluation of the coefficients. See Section 4 and the references given
671: there. But we do not know of any effort to derive a numerical analogue of Table \ref{Tab:PTGS} in this way (which
672: would correspond to truncating the expansion at first order in $1/Z$, see Theorem \ref{T:isolimit}), even though, in
673: principle, the tools to do so were available. In fact, in the only case we are aware of where first order wavefunctions
674: are given \cite[Table 1, p. 288]{Layzer59}, those for Be, B, C are incorrect
675: (the standard aufbau principle configurations are given, instead of the correlated states in Table \ref{Tab:PTGS}).
676:
677: Another interesting strategy would be to vary mathematical parameters in the Schr\"odinger equation which cannot be varied physically, such as
678: $\hbar\to 0$ (semiclassical limit), or $D\to\infty$ where $D$ is the dimensionality of single-particle state space.
679: Despite interesting results (see e.g. the semiclassical picture of highly excited states of Helium in \cite{Tanner00}, and
680: the total atomic energies in \cite{Loeser87} via large-D asymptotics \cite{Witten80, Herschbach86}), these
681: ideas have so far not been aimed at, or led to, explanations of the main features of the periodic table.
682:
683: The large majority of the literature on atomic systems is
684: computational, and does not take the Schr\"odinger equation, but
685: various simplified models as a starting point. Explanations of
686: the filling order (postulate (b) above) in terms of numerical solutions of the Hartree and Hartree-Fock equations began with the
687: pionieering work of Hartree on Rubidium \cite{Hartree28}; for treatments of large classes of atoms see e.g. \cite{Hartree57,
688: FroeseFischer77, Tatewaki94}. Note that these models assume a significant part of postulate (a) from the outset,
689: namely that electrons occupy individual orbitals and that these have $1s$, $2s$, $2p$, ... symmetry. The asymptotic Schr\"odinger
690: ground states which we determined above beautifully illustrate both the power of the Slater determinant ansatz in
691: Hartree-Fock theory and its limitations:
692: in several cases other than noble gases the ground state contains a determinantal state, but for some of the atoms it
693: does not.
694:
695: Finally we remark that the refined, multi-configurational methods developed in quantum chemistry (see e.g. \cite{FroeseFischer77,
696: Tayloretal1, SzaboOstlund96}) assume, on a somewhat ad hoc basis, exactly the structure of the wavefunctions
697: which emerges naturally in Table \ref{Tab:PTGS}, namely finite linear combinations of Slater determinants.
698: The only reason preventing these methods from being asymptotically exact within numerical error as $Z\to\infty$
699: is the use of Gaussians to represent the orbitals (see our companion paper \cite{FG09}).
700: %
701: %
702: %
703: %
704:
705: \section{Reduction to Perturbation Theory Model}\label{Sec:Pert}
706:
707: The first step in establishing the above result is to show that (\ref{SE}) simplifies
708: to a finite-dimensional model in the limit of fixed electron number $N$ and large nuclear charge $Z$.
709:
710: If $\Psi$ solves the original Schr\"odinger equation (\ref{SE}),
711: then an elementary calculation shows that its rescaling
712: \be \label{scaling}
713: \tilde{\Psi}(x_1,s_1,\dots,x_N,s_N) = Z^{-3N/2} \Psi(Z^{-1}x_1,s_1,\dots,Z^{-1}x_N,s_N)
714: \ee
715: solves the equation
716: \be \label{scaledSE}
717: \Bigl(\tilde{H}_0 + \mbox{$\frac{1}{Z}$}V_{ee}\Bigr) \tilde{\Psi} = \Etilde \tilde{\Psi},
718: \ee
719: where $\tilde{H}_0$ is the $Z$-independent Hamiltonian
720: \be \label{ham0}
721: \tilde{H}_0 = \sum_{i=1}^N\Bigl(-\frac12\Delta_{x_i}-\frac{1}{|x_i|}\Bigr)
722: \ee
723: (acting on the $N$-electron Hilbert space (\ref{space}))
724: and $\Etilde=\frac{1}{Z^2}E$. The elementary but important observation now is that the interaction term
725: $\frac{1}{Z}V_{ee}$ in (\ref{scaledSE}) becomes small if $Z$ is large, allowiong to treat the
726: interaction by perturbation theory.
727:
728: Let us first derive the ensuing perturbation-theoretic model informally, then formulate a theorem.
729: By first order Rayleigh-Schr\"odinger perturbation theory, (\ref{scaledSE}) is expected to
730: be well approximated by
731: \begin{eqnarray} \label{scaledPTone}
732: & & \tilde{P}(\tilde{H}_0 + \mbox{$\frac{1}{Z}$}V_{ee})\tilde{P}\tilde{\Psi} = \tilde{E}\tilde{\Psi}, \;\;
733: \tilde{\Psi}\in \tilde{V}_0, \;\;\tilde{P}=\mbox{orth.projector onto }\tilde{V}_0 \\
734: & & \tilde{V}_0 = \mbox{ground state eigenspace of }\tilde{H}_0, \label{scaledPTtwo}
735: \end{eqnarray}
736: with $\tilde{H}_0$ as in (\ref{ham0}). Now we undo the rescaling (\ref{scaling}). This yields the model
737: \begin{eqnarray} \label{PT}
738: & & PHP\Psi = E\Psi, \;\;\;\Psi\in V_0, \;\;\; P=\mbox{orth.$\,$projector onto }V_0, \\
739: & &
740: V_0 = \mbox{ground state eigenspace of }H_0, \;\; H_0=\sum_{i=1}^N \Big( -\frac{1}{2} \Delta_{x_i} - \frac{Z}{|x_i|} \Big),
741: \label{H0H'}
742: \end{eqnarray}
743: where $H$ is the original Hamiltonian (\ref{ham}).
744:
745: We call eqns. (\ref{PT}), (\ref{H0H'}) the {\it PT model}. While it is still a
746: fully interacting quantum many-body model,
747: the key simplification is that the space $V_0$ is finite-dimensional. Its dimension for different atoms is easily read off
748: from Lemma \ref{L:meEfns} below:
749: \begin{table}[h!]
750: \begin{center}
751: \begin{tabular}{|r|c|c|c|c|c|c|c|c|c|}
752: \hline
753: Atom & He& Li&Be & B & C & N & O & F & Ne \\
754: \hline
755: $N$ & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 \\
756: dim $V_0$ & 1 & 8 & 28 & 56 & 70 & 56 & 28 & 8 & 1 \\
757: \hline
758: \end{tabular}
759: \end{center}
760: \caption
761: {Dimensions of degenerate $H_0$ ground
762: states, as given by Lemma \ref{L:meEfns}.}
763: \label{Tab:H0Dim}
764: \end{table}
765:
766: \noindent
767: An important feature of the PT model is that it
768: retains the full symmetries of the atomic Schr\"odinger equation.
769: %
770: %
771: %
772: \begin{lemma} \label{PTsymm} For arbitrary $N$ and $Z$, with $P$ as defined above and with $H$ denoting the
773: Hamiltonian (\ref{ham}), the operators (\ref{ops}) \\
774: (i) leave the ground state $V_0$ of $H_0$ invariant \\
775: (ii) commute with the PT Hamiltonian $PHP \, : \, V_0 \to V_0$.
776: \end{lemma}
777: %
778: %
779: %
780: \noindent
781: {\bf Proof} By direct inspection the operators (\ref{ops}) commute with $H_0$. Since $V_0$ is an eigenspace of
782: $H_0$, they must therefore leave $V_0$ invariant, and commute with the projector $P$ onto $V_0$.
783: As already shown (see Lemma \ref{symlem}), the operators (\ref{ops}) also commute with $H$,
784: and hence with the composition $PHP$.
785: %
786: %
787: %
788: %
789: %
790: %
791: \\[2mm]
792: We now come to the rigorous justification of the PT model (\ref{PT}), (\ref{H0H'}).
793: %
794: %
795: %
796: \begin{theorem} \label{T:isolimit} Let $N=1,\dots,10$, $Z>0$, and let $n(N)$ be the
797: number of energy levels of the PT model (\ref{PT}), (\ref{H0H'}). Then: \\[1mm]
798: (a) For all sufficiently large $Z$, the lowest $n(N)$ energy levels $E_1(N,Z)<\dots<E_{n(N)}(N,Z)$ of the
799: full Hamiltonian (\ref{ham}) have exactly the same dimension, total spin quantum number, total angular momentum
800: quantum number, and parity as the corresponding $PT$ energy levels $E_1^{PT}(N,Z)<\dots<E_{n(N)}^{PT}(N,Z)$. \\[1mm]
801: (b) The lowest $n(N)$ energy levels of the full Hamiltonian have the asymptotic expansion
802: $$
803: \frac{E_j(N,Z)}{Z^2} = \frac{E_j^{PT}(N,Z)}{Z^2} + O(\mbox{$\frac{1}{Z^2}$}) =
804: \Etilde^{(0)} + \frac{1}{Z}\Etilde^{(1)}_j + O(\mbox{$\frac{1}{Z^2}$}) \mbox{ as }Z\to\infty,
805: $$
806: where $\Etilde^{(0)}$ is the lowest eigenvalue of $\tilde{H}_0$
807: and the $\Etilde^{(1)}_j$ are the energy levels of $\tilde{P}V_{ee}\tilde{P}$ on $\tilde{V}_0$. \\[1mm]
808: (c) The projectors $P_1,\dots,P_{n(N)}$ onto the lowest $n(N)$ eigenspaces of the full Hamiltonian satisfy
809: $$
810: |||P_j - P_j^{PT}|||\to 0 \mbox{ as }Z\to\infty,
811: $$
812: where the $P_j^{PT}$ are the corresponding projectors for the PT model.
813: \end{theorem}
814: %
815: %
816: %
817: \noindent
818: The idea that for large $Z$ the inter-electron term $\Etilde^{(1)}_j$ provides the first order
819: correction to the non-interacting energy is well known in the physics literature (see e.g.
820: \cite{Hylleraas30, BetheSalpeter57, SharmaCoulson62, SeungWilson67, RileyDalgarno71, Wilson84} who treat
821: non-degenerate eigenvalues, and see \cite{Layzer59} who gives an expansion similar to that in (b), not accounting
822: for multiplicities, and numerical tables of $\Etilde^{(1)}_j$ in the degenerate case). The main new insight here
823: is the absence of further splittings at higher orders of perturbation theory (see statement (a) in the theorem).
824: This is remarkable, considering that it fails in the simple $3\times 3$ matrix example
825: $$
826: H(\epsilon) = \begin{pmatrix} 0 & 1 & 0 \\ 1 & 0 & 0 \\ 0 & 0 & 1\end{pmatrix}
827: + \epsilon \begin{pmatrix} 1 & 0 & 0 \\ 0 & -1 & 0 \\ 0 & 0 & 0\end{pmatrix}.
828: $$
829: The eigenvalues are $\sqrt{1+\epsilon^2}$, $1$, $-\sqrt{1+\epsilon^2}$, and hence nondegenerate for $\epsilon\neq 0$,
830: but the leading eigenvalue is degenerate in first order perturbation theory.
831: \\[2mm]
832: {\bf Proof} Let $\tilde{E}_1^{PT}<...<\tilde{E}_{n(N)}^{PT}$ be the energy levels of the scaled model
833: (\ref{scaledPTone}), (\ref{scaledPTtwo}), and let $d_j$, $\tilde{P}_j^{PT}$ be the corresponding
834: eigenspace dimensions and eigenspace projectors. By perturbation theory for relatively bounded
835: perturbations of self-adjoint operators (see e.g. \cite{Kat95, Friesecke}), exactly $d_j$ eigenvalues of the scaled Schr\"odinger equation (\ref{scaledSE}) including multiplicity
836: are asymptotic to first order in $1/Z$ to the $j^{th}$ eigenvalue of (\ref{PT}), (\ref{H0H'}).
837: More precisely: The lowest $d_1+...+d_{n(N)}$ eigenvalues of
838: (\ref{scaledSE}) including multiplicity, labelled $\tilde{E}_{j,k}$, $j=1,..,n(N)$, $k=1,..,d_j$,
839: $\tilde{E}_{1,1}\le ... \le \tilde{E}_{1,d_1}\le \tilde{E}_{2,1}\le ... \le \tilde{E}_{2,d_2}\le ...$, satisfy
840: $$
841: \tilde{E}_{j,k} = \tilde{E}^{(0)} + \frac{1}{Z}\tilde{E}_j^{(1)} + O\big(\frac{1}{Z^2}\big) \mbox{ as }Z\to\infty,
842: \;\; k=1,..,d_j.
843: $$
844: Moreover the projector $\tilde{P}_j$ onto the span of these $d_j$ eigenstates satisfies
845: \be \label{threestar}
846: ||| \tilde{P}_j - \tilde{P}_j^{PT} ||| \to 0 \mbox{ as }Z\to\infty.
847: \ee
848: Next, we investigate the Schr\"odinger eigenspace dimensions. By Lemma \ref{PTsymm}, each PT eigenspace possesses well defined spin, angular momentum and parity
849: quantum numbers $L$, $S$, and $p$, and by inspection of the explicit formulae in Theorem \ref{mainresult} below,
850: the space has minimal dimension subject to these numbers.
851: On the other hand, by (\ref{threestar}), for sufficiently large $Z$ these numbers must agree with those of the
852: eigenspaces of (\ref{scaledSE}); hence by Lemma \ref{symlem} (c), $\tilde{E}_{j,1}=...=\tilde{E}_{j,d_j}$.
853: Note that without the information on minimality of the PT dimensions, we would not be able to exclude
854: the possibility of further splittings of the Schr\"odinger eigenvalues beyond the PT splittings,
855: at higher orders of perturbation theory; this is the only reason why the restriction $N\le 10$ is needed.
856:
857: The theorem now follows by applying the isometric scaling transformation (\ref{scaling}).
858: %
859: %
860: %
861: %
862: %
863:
864: \section{State space of the PT model}\label{Sec:GS}
865: %
866: %
867: %
868: The important starting point for solving the PT model is the fact
869: that its state space, the GS of $H_0$, can be determined explicitly. This will follow
870: from the exact solubility of the Schr\"odinger equation of hydrogen and basic many-body arguments.
871: To explain these matters, we start from the hydrogen atom Hamiltonian
872: \be \label{hydrogen}
873: h=-\tfrac{1}{2}\Delta - \frac{Z}{|x|},
874: \ee
875: $x\in\R^3$, acting on $L^2(\R^3\times\Z_2)$. For hydrogen, $Z=1$, but the parameter $Z>0$ will be useful later. Its
876: eigenvalues are given by (see e.g. \cite{Griffiths95})
877: \be \label{hydevals}
878: e_n = - \frac{Z^2}{2n^2}, \;\;\; n\in\N,
879: \ee
880: and have corresponding $2n^2$-dimensional eigenspaces with orthonormal basis
881: \be \label{hydfctns}
882: {\cal B}_n =\{ \phi_{n\ell m s}(x,s) \; | \; \ell=0, \dots , n-1, \; m=-\ell, -\ell+1, \dots,
883: \ell, \; s=-\mbox{$\frac12$}, \mbox{$\frac12$} \},
884: \ee
885: where $\phi_{n \ell m s}\in L^2(\R^3\times\Z_2)$ is the, up to normalization, unique
886: eigenfunction of $h$, $\underline{L}^2$, $L_3$ and $S_3$ with
887: eigenvalues $-Z^2/(2n^2)$, $\ell(\ell+1)$, $m$ and $s$ respectively.
888:
889: For later it will be useful to have an explicit form for these
890: so-called hydrogen orbitals, which in polar coordinates with spin coordinate $s\in\Z_2$ are
891: given by
892: \be
893: \phi_{n,l,m,\sigma}(r,\theta,\phi,s) = \phi_{n\ell m}(r,\theta,\phi)\delta_\sigma(s) =
894: Z^{3/2}R_{n,\ell}(Z r)Y_{\ell,m}(\theta,\phi)\delta_{\sigma}(s)
895: \label{psinlm}
896: \ee
897: where
898: \be
899: R_{n,\ell}(r):= \biggl( \Big(\frac{2}{n}\Big)^3
900: \frac{(n-\ell-1)!}{2n[(n+\ell)!]} \biggr)^{1/2} e^{-r/n} \Big(
901: \frac{2r}{n} \Big)^\ell
902: L_{n-\ell-1}^{2\ell+1}\Big(\frac{2r}{n}\Big). \label{HRnl}
903: \ee
904: Here $L_n^k(x)$ is a generalized Laguerre polynomial and
905: $Y_{\ell,m}(\theta,\phi)$ is a spherical harmonic
906: \cite{AbramowitzStegun72}. In cartesian coordinates, the $n=1$ and $n=2$ orbitals are
907: \begin{align}
908: \phi_{1,0,0}(x)&=\frac{Z^{3/2}}{\sqrt{\pi}} e^{-Z|x|} =:\phi_{1s}(x), \notag \\
909: \phi_{2,0,0}(x)&=\frac{Z^{3/2}}{\sqrt{8 \pi}}
910: \left(1-\frac{Z|x|}{2}\right)e^{-Z|x|/2} =:\phi_{2s}(x), \notag \\
911: \phi_{2,1,0}(x)&=\frac{Z^{5/2}}{\sqrt{32 \pi}} x_3
912: e^{-Z\frac{|x|}{2}} =: \phi_{2p_3},
913: \label{PTorbitals}\\
914: \phi_{2,1,\pm 1}(x)&=\frac{Z^{5/2}}{\sqrt{32 \pi}} \frac{x_1\pm ix_2}{\sqrt{2}} e^{-Z|x|/2} =:\phi_{2p_\pm}. \notag
915: \end{align}
916: Often, it is convenient to work -- instead of the last two functions -- with their real linear
917: combinations
918: \be \label{porbitals}
919: \frac{Z^{5/2}}{\sqrt{32 \pi}} x_j
920: e^{-Z\frac{|x|}{2}} =: \phi_{2p_j}(x), \;\;\; j=1,2.
921: \ee
922:
923: The following lemma describes how the eigenfunctions for the
924: non-interacting many-electron system are formed from these
925: one-electron eigenfunctions.
926: \begin{lemma} \label{L:meEfns} (Standard ``folklore'', see \cite{Friesecke} for a rigorous proof)
927: \text{} (a)
928: The lowest eigenvalue of the operator
929: $$
930: H_0:=-\frac{1}{2}\sum_{i=1}^N \Delta_i - \sum_{i=1}^N\frac{Z}{|x_{i}|}
931: $$
932: on the space $L^2_a\big((\R^3\times\Z_2)^N\big)$ of square-integrable functions $\Psi : (\R^3 \times
933: \Z_2)^N \rightarrow \C$ satisfying the antisymmetry condition (\ref{anti})
934: is $E\!=\!\sum^N_{n=1}\tilde{e}_n$, where $\tilde{e}_1\!\le\!\tilde{e_2}\!\le\! ...$ is an ordered list including
935: multiplicity of the hydrogen eigenvalues (\ref{hydevals}).
936: \\[2mm]
937: (b)
938: The corresponding eigenspace is
939: \begin{align} V_0 = & \Span \Big\{|\chi_1\dots \chi_{d_*}\psi_{i_1} \dots
940: \psi_{i_{N-d_*}} \rangle \, \Big| \nonumber \\
941: & \qquad\qquad 1\le i_1 < \dots <i_{N-d_*}\le 2(n_*+1)^2\}\Big\}, \label{manybodybasis}
942: \end{align}
943: where the functions $\chi_i$ and $\psi_i$ (``core orbitals'' and ``valence orbitals'') and the
944: integers $d_*$ and $n_*$ (``number of core orbitals'' and ``number of closed shells'') are defined
945: as follows:
946: $d_*(N)$ is the largest number of form
947: $\sum_{j=1}^n2j^2$ which is less or equal to $N$,
948: $n_*(N)$ is the corresponding value of $n$,
949: $$
950: \{\chi_1,\dots,\chi_{d_*}\}={\cal B}_1\cup\cdots\cup{\cal B}_{n_*}
951: $$
952: (union of the ON bases (\ref{hydfctns}) of the first $n_*$ hydrogen eigenspaces), and
953: $$
954: \{\psi_1,\dots,\psi_{2(n_*+1)^2}\}={\cal B}_{n_*+1}
955: $$
956: (ON basis of the $(n_*+1)^{st}$ hydrogen eigenspace).
957: \end{lemma}
958: Thus the ground state of the non-interacting
959: Hamiltonian is spanned by Slater determinants (alias antisymmetrized tensor products) formed
960: from scaled hydrogen orbitals, ``filled'' in order of increasing
961: one-electron energy.
962:
963: Due to the freedom of choosing any $N-d_*$ eigenfunctions $\psi_i$ (``valence
964: orbitals'') from the basis of the highest relevant hydrogen eigenspace, whose dimension is $2(n_*+1)^2$, the noninteracting GS
965: has typically a large degeneracy:
966: \be
967: d_0 := \mbox{dim GS of $H_0$} = \binom{2(n_*(N)+1)^2}{N-d_*(N)}. \label{dimGS}
968: \ee
969: %
970: %
971: %
972: {\bf Specialization to the second row atoms and their isoelectronic ions (N=3,...,10)} In this case, the number $d_*$ of
973: core orbitals equals $2$, the number $n_*$ of closed shells equals $1$,
974: and the dimension $2(n_*+1)^2$ of the hydrogen eigenspace from which the valence orbitals are selected
975: equals $8$. Thus by (\ref{dimGS}), the dimension of the ground state equals
976: $$
977: d_0 = \binom{8}{N-2}.
978: $$
979: These numbers are given in Table \ref{Tab:H0Dim}. The set of core
980: respectively valence orbitals is (using the real orbitals
981: $\phi_{2p_1}$, $\phi_{2p_2}$ instead of $\phi_{2p_{\pm}}$)
982: \begin{align}
983: & \{\chi_1,\chi_2\} = \{\phi_{1s}\!\uparrow , \, \phi_{1s}\!\downarrow \}, \label{Corbitals} \\
984: & \{\psi_1,\dots,\psi_8\} = \{\phi_{2s}\!\uparrow, \, \phi_{2s}\!\downarrow, \,
985: \phi_{2p_1}\!\uparrow, \, \phi_{2p_1} \!\downarrow, \,
986: \phi_{2p_2}\!\uparrow, \, \phi_{2p_2}\!\downarrow, \,
987: \phi_{2p_3}\!\uparrow, \, \phi_{2p_3} \!\downarrow \} \label{Vorbitals}.
988: \end{align}
989: Here we have employed the standard notation $\phi\!\uparrow$, $\phi\!\downarrow$ for the two spin orbitals
990: $\phi(x)\delta_{\pm 1/2}(\sigma)$.
991:
992: Finally, the ground state of $H_0$ is
993: \be \label{GS2ndperiod}
994: V_0(N) = \Span \Bigl\{ |\chi_1\chi_2\psi_{i_1}\dots\psi_{i_{N-2}}\rangle \, \Bigl|\Bigr. \, 1\le i_1<\dots<i_{N-2}\le 8\Bigr\}.
995: \ee
996: %
997: %
998: %
999: %
1000: %
1001: %
1002: %
1003: %
1004:
1005: \section{Determining the matrix PHP} \label{Sec:Matrix}
1006: In this section we determine explicitly the Hamiltonian matrices $PHP$, for all second period atoms.
1007:
1008: Most of our arguments do not rely on the special radial form of the hydrogen orbitals
1009: (\ref{Corbitals}), (\ref{Vorbitals}) appearing in the definition of the subspace $V_0(N)$. Hence in this section,
1010: unless stated otherwise, $V_0(N)$ denotes the space
1011: (\ref{GS2ndperiod}), (\ref{Corbitals}), (\ref{Vorbitals}) with the more general orbitals
1012: \be
1013: \varphi_{1s}(x)=R_1(|x|), \;\; \varphi_{2s}(x)=R_2(|x|), \;\; \varphi_{2p_i}(x)=R_3(|x|)x_i \;(i=1,2,3), \label{GeneralOrbitals}
1014: \ee
1015: where the $\varphi$'s are in $L^2(\R^3)$ with norm one, $R_i: \R \to
1016: \R$, and $\int_0^\infty R_1(r)R_2(r) r^2 dr=0$.
1017: %
1018: %
1019: %
1020:
1021: \subsection{Spin and angular momentum calculus on Slater determinants}\label{S:7.1}
1022:
1023: The action of the spin and angular momentum operators on $V_0(N)$ can be calculated
1024: from their action on the orbitals (\ref{Corbitals}) and (\ref{Vorbitals}) together
1025: with the following simple identitites for the action of linear operators of form
1026: $$
1027: B = \sum_{i=1}^N b(i), \;\;\; B^2 = \sum_{i,j=1}^N b(i) b(j)
1028: $$
1029: on Slater determinants, where $b$ is a linear operator on $L^2(\R^3\times\Z_2)$:
1030: \begin{align}
1031: B \, | \chi_1,\dots,\chi_N\rangle &= \sum_{i=1}^N |\chi_1, \dots, b\chi_i, \dots, \chi_N\rangle,
1032: \label{L:OneParticleOnSD} \\
1033: B^2 \, | \chi_1,\dots,\chi_N\rangle &= \sum_{i=1}^N |\chi_1, \dots, b^2\chi_i, \dots, \chi_N\rangle \nonumber \\
1034: &\quad + 2 \sum_{1\le i<j\le N} |\chi_1,\dots, b\chi_i, \dots, b\chi_j, \dots, \chi_N\rangle.
1035: \label{L:TwoParticleOnSD}
1036: \end{align}
1037: %
1038:
1039: Direct calculations show that, for any two spatial orbitals $\psi,
1040: \phi \in L^2(\R^3)$, and orthogonal spin states $\alpha,
1041: \beta:\Z_2\to\C$,
1042: \begin{align*}
1043: \underline{S}\cdot \underline{S} \psi \alpha &= \tfrac{3}{4}
1044: \psi\alpha, \\
1045: (\underline{S}(1) \cdot \underline{S}(2)) \psi \alpha \otimes
1046: \phi \beta &= \tfrac{1}{2} \psi\beta \otimes \phi\alpha -
1047: \tfrac{1}{4} \psi\alpha \otimes \phi\beta, \\
1048: (\underline{S}(1) \cdot \underline{S}(2)) \psi\alpha \otimes \phi\alpha
1049: &= \tfrac{1}{4} \psi\alpha \otimes \phi\alpha.
1050: \end{align*}
1051:
1052: In particular, by (\ref{L:OneParticleOnSD}) and
1053: (\ref{L:TwoParticleOnSD}), $S_3|\psi\alpha \psi\beta\rangle=0$ and
1054: $\SSS^2|\psi\alpha \psi\beta\rangle = (
1055: \frac{3}{4}+\frac{3}{4})|\psi\alpha \psi\beta\rangle - 2 \cdot
1056: \frac{3}{4}|\psi\alpha \psi\beta\rangle=0$.
1057:
1058: The angular momentum operators (\ref{Lj}) act on the orbitals (\ref{GeneralOrbitals}) as follows,
1059: independently of the choice of spin $\alpha:\Z_2\to\C$:
1060: \begin{align*}
1061: & L_j \varphi_{1s} \alpha = L_j \varphi_{2s} \alpha = L_j \varphi_{2p_j}\alpha =0, \\
1062: L_{j+1}\varphi_{2p_j}\alpha=&-i\varphi_{2p_{j-1}}\alpha,
1063: \quad L_{j-1}\varphi_{2p_j}\alpha=i\varphi_{2p_{j+1}}\alpha, \; j=1,2,3,
1064: \end{align*}
1065: where the indices are understood modulo three. Hence we need only
1066: consider the action of $\LL^2$ on $\varphi_{2p_i}$, giving, for any
1067: two spin states $\alpha$ and $\beta$, and $i \neq j$,
1068: \begin{align*}
1069: \underline{L} \cdot \underline{L} \varphi_{2p_i} \alpha &= 2\varphi_{2p_i} \alpha \\
1070: (\underline{L}(1) \cdot \underline{L}(2)) \varphi_{2p_i}\alpha \otimes
1071: \varphi_{2p_i} \beta &= -(\varphi_{2p_{i-1}}\alpha \otimes \varphi_{2p_{i-1}}\beta + \varphi_{2p_{i+1}}\alpha
1072: \otimes \varphi_{2p_{i+1}} \beta) \\
1073: (\underline{L}(1) \cdot \underline{L}(2)) \varphi_{2p_i}\alpha \otimes
1074: \varphi_{2p_j} \beta &= \varphi_{2p_j}\alpha \otimes \varphi_{2p_i}\beta.
1075: \end{align*}
1076:
1077: Finally we see that, for any spin state $\alpha$,
1078: $\hat{R}\varphi_{1s}\alpha=\varphi_{1s}\alpha$,
1079: $\hat{R}\varphi_{2s}\alpha=\varphi_{2s}\alpha$ and
1080: $\hat{R}\varphi_{2p_i}\alpha=-\varphi_{2p_i}\alpha$, $i=1,2,3$.
1081:
1082: A useful and well known consequence of the above is that the pair of 1s orbitals makes no contribution
1083: to spin, angular momentum, and parity on the space (\ref{GS2ndperiod}). More precisely:
1084: %
1085: %
1086: %
1087: \begin{lemma} \label{eliminate1s} (See [Fri0X])
1088: The matrix of any of the operators (\ref{ops}) on $V_0(N)$ with respect
1089: to the basis (\ref{GS2ndperiod}) is the same as that on the corresponding fewer-particle
1090: space obtained by deleting the orbitals $\chi_1$, $\chi_2$,
1091: with respect to the corresponding basis
1092: $\{ | \psi_{i_1}\dots\psi_{i_{N-2}}\rangle \, | \, 1\le i_1<\dots<i_{N-2}\le
1093: 8\}$.
1094: \end{lemma}
1095:
1096: \subsection{Particle-hole duality}\label{S:Duality}
1097:
1098: A further observation that simplifies the calculation of the eigenfunctions is a particle-hole duality result.
1099: We introduce a
1100: dual operator, in the spirit of the Hodge star operator from differential geometry (see e.g.
1101: \cite{Jost02}), by
1102: \\[2mm]
1103: {\bf Definition} \label{D:star}
1104: {\it The dual of $\alpha|\Psi\rangle$, with
1105: $$
1106: |\Psi\rangle = |\varphi_{1s}\!\uparrow \varphi_{1s}\!\downarrow \psi_{i_1}
1107: \dots \psi_{i_{N-2}}\rangle,
1108: $$
1109: being any element of the basis (\ref{GS2ndperiod}) and
1110: $\alpha \in \C$, denoted by $\ast(\alpha|\Psi\rangle)$, is given by
1111: \be
1112: \ast \big(\alpha|\Psi\rangle \big) := \alpha^* a(\psi_{i_{N-2}}) \dots
1113: a(\psi_{i_1})|\mathbbm{1}\rangle,
1114: \label{*2ndQ}
1115: \ee
1116: where
1117: $$
1118: |\mathbbm{1}\rangle:=|\varphi_{1s}\! \uparrow \varphi_{1s}\!\downarrow
1119: \varphi_{2s}\!\uparrow \varphi_{2s}\!\downarrow \varphi_{2p_1}\! \uparrow
1120: \varphi_{2p_1}\!\downarrow \varphi_{2p_2} \!\uparrow \varphi_{2p_2}\!\downarrow
1121: \varphi_{2p_3}\! \uparrow \varphi_{2p_3} \!\downarrow \rangle
1122: $$
1123: and $a(\psi)$ is the usual annihilation operator which maps $|\psi \, \psi_{i_1}\dots\psi_{i_k}\rangle$
1124: to $|\psi_{i_1}\dots\psi_{i_k}\rangle$.
1125: }
1126: %end of definition
1127: %
1128: %
1129: \\[2mm]
1130: We extend $\ast$ linearly to real linear combinations of the
1131: $\alpha|\Psi\rangle$, thereby obtaining an antilinear map from $V_0(N)$ to
1132: $V_0(10-(N-2))$. We then have the following result:
1133: \begin{lemma} \label{L:Duality}
1134: Suppose $\Psi \in V_0(N)$ satisfies $\LL^2\Psi=\mathcal{L}\Psi$ and
1135: $\SSS^2\Psi=\mathcal{S}\Psi$. Then
1136: $\LL^2(\ast\Psi)=\mathcal{L}(\ast\Psi)$ and
1137: $\SSS^2(\ast\Psi)=\mathcal{S}(\ast\Psi)$. Furthermore, if $L_3\Psi=M
1138: \Psi$, $S_3 \Psi = M_s \Psi$ and $\hat{R}\Psi=p\Psi$, then
1139: $L_3(\ast\Psi)=-M (\ast\Psi)$, $S_3 (\ast\Psi) = -M_s (\ast\Psi)$
1140: and $\hat{R}(\ast\Psi)=p(\ast\Psi)$.
1141: \end{lemma}
1142: \noindent \textbf{Proof} Direct calculations using the second
1143: quantized forms of $\LL$ and $\SSS$ show that both operators
1144: anticommute with $\ast$ on $V_0(N)$. The results for angular momentum
1145: and spin are then trivial. The result for the inversion operator
1146: follows from the fact that the parity of a wavefunction is
1147: equivalent to the parity of the number of $p$-orbitals present in
1148: each Slater determinant (since $\hat{R}\varphi_{ns} \alpha=
1149: \varphi_{ns}\alpha$ for both $n=1$ and $2$, and
1150: $\hat{R}\varphi_{2p_i}\alpha= -\varphi_{2p_i}\alpha$ for $i=1,2,3$)
1151: and the number of $p$-orbitals in the dual of a Slater determinant
1152: with $k$ $p$-orbitals is $6-k$, preserving the parity. \qed
1153: %
1154: %
1155: \\[2mm]
1156: The above result is a modest generalization of insights by spectroscopists
1157: (who did not know they were speaking what mathematicians would call supersymmetry).
1158: They introduced the dual of a configuration with respect to a single open shell
1159: (termed ``conjugate configuration'' in \cite{Condon80}), and noticed that it gives rise to the same
1160: set of $L_3$ and $S_3$ eigenvalues \cite{Condon80} and, what is more, the same
1161: $\LL^2$, $\SSS^2$, and $\LL\cdot\SSS$ matrices \cite{CondonShortley35}.
1162: %
1163: %
1164: \subsection{Simultaneous $\LL^2$-$\SSS^2$ Eigenspaces} \label{S:Eigenspaces}
1165:
1166: We now form the joint angular momentum and spin eigenspaces within $V_0(N)$.
1167: Lemma \ref{L:Duality} shows that we only need to do this for
1168: Lithium-Carbon, the remaining cases follow by duality. Under the restriction of
1169: a fixed number of $1s$, $2s$, $2p$ orbitals, the results were no doubt known to early
1170: spectroscopists, who realized that the multiplet structure of observed spectra can only
1171: be captured via superposition of aufbau principle Slater determinants into ``terms'' (in our
1172: language, joint $\LL^2$--$\SSS^2$ eigenspaces); see e.g. \cite{CondonShortley35, Condon80}. We do not
1173: however know of a complete tabulation.
1174:
1175: \begin{theorem} For the Lithium-Neon sequences ($N=3, \dots, 10$, $Z>0$),
1176: orthonormal basis for the $\LL^2$-$\SSS^2$ simultaneous eigenspaces
1177: within $V_0(N)$ are as given in
1178: Tables \ref{Tab:LiFullTable}-\ref{Tab:NeFullTable}. See below for
1179: the notation used in the tables.
1180: \end{theorem}
1181: %
1182: %
1183: \noindent
1184: \textbf{Proof} We only give the proof for the highest dimensional
1185: case, Carbon, the other cases being analogous but easier.
1186:
1187: By Lemma \ref{eliminate1s}, it suffices to find the joint $\LL^2$-$\SSS^2$-eigenstates
1188: in the four-electron vector space spanned by
1189: ${\cal B}=\{|\psi_{i_1}\psi_{i_2}\psi_{i_3}\psi_{i_4}\rangle\, | \, 1\le i_1<i_2<i_3<i_4\le 8\}$,
1190: with the $\psi_i$ as in (\ref{Vorbitals}).
1191:
1192: We note first that each Slater determinant in the above basis
1193: is already an eigenfunction of
1194: $\SSS_3$, and that the space with $S_3$-eigenvalue $M$
1195: is isomorphic (by flipping all spins) to that
1196: with eigenvalue $-M$. Since both $\LL^2$ and $\SSS^2$ commute with $S_3$,
1197: it suffices therefore to consider their action on the
1198: eigenspaces of $S_3$ with eigenvalue $M\ge 0$.
1199:
1200: Next we observe that within each such $S_3$-eigenspace, the span of
1201: those Slater determinants which share the same number of different
1202: spatial orbitals (4, 3, or 2) is also invariant under $\LL^2$ and
1203: $\SSS^2$.
1204:
1205: We now calculate the matrices of $\LL^2$ and $\SSS^2$ with respect to the
1206: so-obtained subsets of the basis ${\cal B}$, using (\ref{L:OneParticleOnSD}), (\ref{L:TwoParticleOnSD}), and
1207: the formulae from subsection \ref{S:7.1}.
1208: To shorten the notation, we will write
1209: $s\!\uparrow$, $s\!\downarrow$, $p_i\!\uparrow$, $p_i\!\downarrow$ instead of
1210: $\varphi_{2s}\!\uparrow$, $\varphi_{2s}\!\downarrow$, $\varphi_{2p_i}\!\uparrow$, $\varphi_{2p_i}\!\downarrow$.
1211: \\[1mm]
1212: {\bf Four different spatial orbitals, M=2} On $|s\!\uparrow\, p_1\!\uparrow \, p_2\!\uparrow \,
1213: p_3\!\uparrow\rangle$,
1214: $$
1215: \LL^2 = 0, \;\;\; \SSS^2 = 6.
1216: $$
1217: {\bf Four different spatial orbitals, M=1} With respect to the basis
1218: $\{|s\!\uparrow\, p_1\!\uparrow\, p_2\!\uparrow\, p_3\!\downarrow\rangle, \;
1219: |s\!\uparrow\, p_1\!\uparrow\, p_2\!\downarrow\, p_3\!\uparrow\rangle, \;
1220: |s\!\uparrow\, p_1\!\downarrow\, p_2\!\uparrow\, p_3\!\uparrow\rangle, \;
1221: |s\!\downarrow\, p_1\!\uparrow\, p_2\!\uparrow\, p_3\!\uparrow\rangle\}$,
1222: $$
1223: \LL^2 = \left(\begin{array}{rrrr} 4 & -2 & -2 & 0 \\ -2 & 4 & -2 & 0 \\
1224: -2 & -2 & 4 & 0 \\
1225: 0 & 0 & 0 & 0 \end{array}\right), \;\;\;
1226: \SSS^2 = \left(\begin{array}{rrrr} 3 & 1 & 1 & 1 \\ 1 & 3 & 1 & 1 \\
1227: 1 & 1 & 3 & 1 \\
1228: 1 & 1 & 1 & 3 \end{array}\right).
1229: $$
1230: {\bf Four different spatial orbitals, M=0} With respect to the basis
1231: $\{|s\!\uparrow\, p_1\!\uparrow\, p_2\!\downarrow\, p_3\!\downarrow\rangle, \;
1232: |s\!\uparrow\, p_1\!\downarrow\, p_2\!\uparrow\, p_3\!\downarrow\rangle, \;
1233: |s\!\uparrow\, p_1\!\downarrow\, p_2\!\downarrow\, p_3\!\uparrow\rangle, \;
1234: |s\!\downarrow\, p_1\!\downarrow\, p_2\!\uparrow\, p_3\!\uparrow\rangle, \;
1235: |s\!\downarrow\, p_1\!\uparrow\, p_2\!\downarrow\, p_3\!\uparrow\rangle, \;
1236: |s\!\downarrow\, p_1\!\uparrow\, p_2\!\uparrow\, p_3\!\downarrow\rangle\}$,
1237: $$
1238: \LL^2 = \left(\begin{array}{rrrrrr} 4 & -2 & -2 & 0 & 0 & 0 \\
1239: -2 & 4 & -2 & 0 & 0 & 0 \\
1240: -2 & -2 & 4 & 0 & 0 & 0 \\
1241: 0 & 0 & 0 & 4 & -2 & -2 \\
1242: 0 & 0 & 0 & -2 & 4 & -2 \\
1243: 0 & 0 & 0 & -2 & -2 & 4 \end{array}\right), \;\;\;
1244: \SSS^2 = \left(\begin{array}{rrrrrr} 2 & 1 & 1 & 0 & 1 & 1 \\
1245: 1 & 2 & 1 & 1 & 0 & 1 \\
1246: 1 & 1 & 2 & 1 & 1 & 0 \\
1247: 0 & 1 & 1 & 2 & 1 & 1 \\
1248: 1 & 0 & 1 & 1 & 2 & 1 \\
1249: 1 & 1 & 0 & 1 & 1 & 2 \end{array}\right). \;\;\;
1250: $$
1251: {\bf Three different spatial oritals, M=1} On each Slater determinant
1252: $|s\!\uparrow\, s\!\downarrow\, p_i\!\uparrow\, p_j\!\uparrow\rangle$, and each Slater determinant
1253: $|p_k\!\uparrow\, p_k\!\downarrow\, p_i\!\uparrow\, p_j\!\uparrow\rangle$,
1254: $$
1255: \LL^2 = 2, \;\;\; \SSS^2=2.
1256: $$
1257: (In total, these span a 6-dimensional subspace.)
1258: With respect to each basis
1259: $\{|p_i\!\uparrow\, p_i\!\downarrow\, s\!\uparrow\, p_j\!\uparrow\rangle, \;
1260: |p_k\!\uparrow\, p_k\!\downarrow\, s\!\uparrow\, p_j\!\uparrow\rangle\}$,
1261: $$
1262: \LL^2 = \left(\begin{array}{rr} 4 & -2 \\
1263: -2 & 4 \end{array}\right), \;\;\;
1264: \SSS^2 = 2.
1265: $$
1266: (In total, these span a 6-dimensional subspace.)
1267: \\[1mm]
1268: {\bf Three different spatial orbitals, M=0} With respect to each of the bases
1269: $\{|s\!\uparrow\, s\!\downarrow\, p_i\!\uparrow\, p_j\!\downarrow\rangle, \;
1270: |s\!\uparrow\, s\!\downarrow\, p_i\!\downarrow\, p_j\!\uparrow\rangle\}$
1271: and
1272: $\{|p_k\!\uparrow\, p_k\!\downarrow\, p_i\!\uparrow\, p_j\!\downarrow\rangle, \;
1273: |p_k\!\uparrow\, p_k\!\downarrow\, p_i\!\downarrow\, p_j\!\uparrow\rangle\}$,
1274: $$
1275: \LL^2 = \left(\begin{array}{rr} 4 & -2 \\
1276: -2 & 4 \end{array}\right), \;\;\;
1277: \SSS^2 = \left(\begin{array}{rr} 4 & -2 \\
1278: -2 & 4 \end{array}\right).
1279: $$
1280: (In total, these span a 12-dimensional subspace.)
1281: With respect to each of the bases
1282: $\{|p_i\!\uparrow\, p_i\!\downarrow\, s\!\uparrow\, p_j\!\downarrow\rangle, \;
1283: |p_k\!\uparrow\, p_k\!\downarrow\, s\!\uparrow\, p_j\!\downarrow\rangle, \;
1284: |p_i\!\uparrow\, p_i\!\downarrow\, s\!\downarrow\, p_j\!\uparrow\rangle, \;
1285: |p_k\!\uparrow\, p_k\!\downarrow\, s\!\downarrow\, p_j\!\uparrow\rangle$,
1286: $$
1287: \LL^2 = \left(\begin{array}{rrrr} 4 & -2 & 0 & 0 \\
1288: -2 & 4 & 0 & 0 \\
1289: 0 & 0 & 4 & -2 \\
1290: 0 & 0 & -2 & 4 \end{array}\right), \;\;\;
1291: \SSS^2 = \left(\begin{array}{rrrr} 1 & 0 & 1 & 0 \\ 0 & 1 & 0 & 1 \\
1292: 1 & 0 & 1 & 0 \\
1293: 0 & 1 & 0 & 1 \end{array}\right).
1294: $$
1295: (In total, these span a 12-dimensional subspace.)
1296: \\[1mm]
1297: {\bf Two different spatial orbitals, M=0} With respect to the bases
1298: $\{|s\!\uparrow\, s\!\downarrow\, p_1\!\uparrow\, p_1\!\downarrow\rangle, \;
1299: |s\!\uparrow\, s\!\downarrow\, p_2\!\uparrow\, p_2\!\downarrow\rangle, \;
1300: |s\!\uparrow\, s\!\downarrow\, p_3\!\uparrow\, p_3\!\downarrow\rangle\}$
1301: and
1302: $\{|p_1\!\uparrow\, p_1\!\downarrow\, p_2\!\uparrow\, p_2\!\downarrow\rangle, \;
1303: |p_2\!\uparrow\, p_2\!\downarrow\, p_3\!\uparrow\, p_3\!\downarrow\rangle, \;
1304: |p_3\!\uparrow\, p_3\!\downarrow\, p_1\!\uparrow\, p_1\!\downarrow\rangle\}$,
1305: $$
1306: \LL^2 = \left(\begin{array}{rrr} 4 & -2 & -2 \\
1307: -2 & 4 & -2 \\
1308: -2 & -2 & 4 \end{array}\right), \;\;\;
1309: \SSS^2 = 0.
1310: $$
1311: This completes the explicit description of the action of $\LL^2$ and $\SSS^2$.
1312:
1313: The eigenfunctions and eigenvalues are now found by explicit diagonalization of
1314: the above matrices. \qed\\
1315:
1316: Tables \ref{Tab:LiFullTable}-\ref{Tab:NeFullTable} use the following
1317: conventions: \\[1mm]
1318: 1) The two $1s$ orbitals present in every Slater
1319: determinant are not shown. \\[1mm]
1320: 2) The eigenfunctions are not normalized. \\[1mm]
1321: 3) In all cases, it is assumed that $i = 1,2,3$,
1322: $(i,j)=(1,2),(2,3),(3,1)$ and $(i,j,k)$ is any cyclic permutation of
1323: $(1,2,3)$. In particular, any eigenfunction containing a variable corresponds to a 3D subspace. \\[1mm]
1324: 4) Eigenfunctions of the form $a\Psi_1 + b\Psi_2 +
1325: c\Psi_3$ are such that $a+b+c=0$ and stand for two linearly
1326: independent orthogonal choices of $(a,b,c)$, and hence correspond to a
1327: two-dimensional subspace. \\[1mm]
1328: 5) Within each $\LL^2$-$\SSS^2$-eigenspace, the different
1329: $S_3$ eigenspaces are separated by a line, in the order $M=S$, $-S$, $S-1$, $-(S-1)$, $\dots$, $0$. \\[1mm]
1330: 6) The spin orbitals
1331: $\varphi_{2s}\!\uparrow$, $\varphi_{2s}\!\downarrow$, $\varphi_{2p_i}\!\uparrow$,
1332: $\varphi_{2p_i}\!\downarrow$ are abbreviated $s$, $\overline{s}$, $p_i$,
1333: $\overline{p_i}$. \\
1334:
1335: Note that the parity of the eigenfunctions in the tables, although not shown explicitly,
1336: can be read off by counting the number of $p$ orbitals (see the previous section).
1337: %
1338: %
1339: %
1340: %
1341:
1342: % needs the following for entries covering multiple rows in tables
1343: % \usepackage{multirow}
1344: % and the following for entries covering multiple columns
1345: % \newcommand{\mc}[1]{\multicolumn{2}{|c|}{#1}}
1346:
1347: % Lithium
1348:
1349: \begin{table}[h!]
1350: \begin{minipage}[h]{0.25\textwidth}
1351:
1352: \begin{center}
1353: \scalebox{0.75}{
1354: \begin{tabular}{|r|c|}
1355: \hline & $\SSS^2=\tfrac{3}{4}$ \\
1356: \hline
1357:
1358: $\LL^2=0$ &
1359: \begin{tabular}{c}
1360: $|s\rangle$ \\
1361: \hline
1362: $|\overline{s}\rangle$ \\
1363: \end{tabular} \\
1364: \hline
1365:
1366: $\LL^2=2$ &
1367: \begin{tabular}{c}
1368: $|p_i\rangle$ \\
1369: \hline
1370: $|\overline{p_i}\rangle$ \\
1371: \end{tabular} \\
1372:
1373: \hline
1374: \end{tabular}
1375: } %end scalebox
1376: \end{center}
1377: \caption{Lithium sequence $\protect\LL^2$-$\protect\SSS^2$
1378: eigenspaces.} \label{Tab:LiFullTable}
1379: \end{minipage}
1380: \begin{minipage}[h]{0.66\textwidth}
1381:
1382: \begin{center}
1383:
1384: \scalebox{0.75}{
1385: \begin{tabular}{|r|c|c|}
1386: \hline & $\SSS^2=0$ & $\SSS^2=2$ \\
1387:
1388: \hline $\LL^2=0$ &
1389:
1390: \begin{tabular}{c}
1391: $|s\overline{s}\rangle$ \\
1392: $|p_1\overline{p_1}\rangle+|p_2\overline{p_2}\rangle+|p_3\overline{p_3}\rangle$\\
1393: \end{tabular}
1394: & \\
1395:
1396: \hline
1397:
1398: $\LL^2=2$
1399:
1400: &
1401: \begin{tabular}{c}
1402: $|s\overline{p_i}\rangle-|\overline{s}p_i\rangle$
1403: \end{tabular}
1404:
1405: &
1406: \begin{tabular}{cc}
1407:
1408: $|sp_i\rangle$
1409:
1410: & $|p_ip_j\rangle$\\
1411: \hline
1412:
1413: $|\overline{s}\overline{p_i}\rangle$
1414:
1415: & $|\overline{p_i}\overline{p_j}\rangle$ \\
1416: \hline
1417:
1418: $|s\overline{p_i}\rangle+|\overline{s}p_i\rangle$
1419:
1420: & $|p_i\overline{p_j}\rangle+|\overline{p_i}p_j\rangle$
1421: \vspace{1mm}
1422: \end{tabular}
1423:
1424: \\
1425:
1426:
1427: \hline
1428:
1429: $\LL^2=6$ &
1430:
1431: \begin{tabular}{c}
1432: $|p_i\overline{p_j}\rangle-|\overline{p_i}p_j\rangle$\\
1433: $a|p_1\overline{p_1}\rangle+b|p_2\overline{p_2}\rangle+c|p_3\overline{p_3}\rangle$
1434: \\
1435: \end{tabular}
1436:
1437: &\\
1438:
1439: \hline
1440:
1441: \end{tabular}
1442:
1443: }% end scalebox
1444:
1445: \end{center}
1446: \caption{Beryllium sequence
1447: $\protect\LL^2$-$\protect\SSS^2$ eigenspaces.}
1448: \label{Tab:BeTableFull}
1449: \end{minipage}
1450:
1451: \end{table}
1452:
1453:
1454: % Boron
1455: \begin{table}[h!tp]
1456: \begin{center}
1457: \scalebox{0.75}{
1458: \begin{tabular}{|r|c|c|}
1459: \hline & $\SSS^2=\tfrac{3}{4}$
1460: & $\SSS^2=\tfrac{15}{4}$ \\
1461:
1462: \hline $\LL^2=0$ &
1463:
1464: \begin{tabular}{c}
1465: $|sp_1\overline{p_1}\rangle +|sp_2\overline{p_2}\rangle
1466: +|sp_3\overline{p_3}\rangle$ \\[3mm]
1467: \hline
1468: \rule{0pt}{6mm}
1469:
1470: $|\overline{s}p_1\overline{p_1}\rangle
1471: +|\overline{s}p_2\overline{p_2}\rangle
1472: + |\overline{s}p_3\overline{p_3}\rangle$\\
1473: \end{tabular}
1474:
1475: &
1476:
1477: \begin{tabular}{c}
1478:
1479: $|p_1p_2p_3\rangle$ \\
1480: \hline
1481:
1482: $|\overline{p_1}\overline{p_2}\overline{p_3}\rangle$ \\
1483: \hline
1484:
1485: $|p_1p_2\overline{p_3}\rangle+|p_1\overline{p_2}p_3\rangle+|\overline{p_1}p_2p_3\rangle$
1486: \\
1487: \hline
1488:
1489: $|\overline{p_1}\overline{p_2}p_3\rangle
1490: +|\overline{p_1}p_2\overline{p_3}\rangle
1491: +|p_1\overline{p_2}\overline{p_3}\rangle$
1492: \\
1493:
1494: \end{tabular}
1495:
1496: \\
1497:
1498: \hline $\LL^2=2$ &
1499:
1500: \begin{tabular}{c}
1501:
1502: $|s\overline{s}p_i\rangle$ \\
1503:
1504: $|p_ip_j\overline{p_j}\rangle
1505: +|p_ip_k\overline{p_k}\rangle$ \\
1506:
1507: $2|\overline{s}p_ip_j\rangle-|s\overline{p_i}p_j\rangle-|sp_i\overline{p_j}\rangle$\\
1508: \hline
1509:
1510: $|s\overline{s}\overline{p_i}\rangle$ \\
1511:
1512: $|\overline{p_i}p_j\overline{p_j}\rangle
1513: +|\overline{p_i}p_k\overline{p_k}\rangle$ \\
1514:
1515: $2|s\overline{p_i}\overline{p_j}\rangle-|\overline{s}p_i\overline{p_j}\rangle-|\overline{s}\overline{p_i}p_j\rangle$
1516: \\
1517:
1518: \end{tabular}
1519:
1520: &
1521:
1522: \begin{tabular}{c}
1523:
1524: $|sp_ip_j\rangle$ \\
1525: \hline
1526:
1527: $|\overline{s}\overline{p_i}\overline{p_j}\rangle$ \\
1528: \hline
1529:
1530: $|sp_i\overline{p_j}\rangle+|s\overline{p_i}p_j\rangle+|\overline{s}p_ip_j\rangle$\\
1531: \hline
1532:
1533: $|s\overline{p_i}\overline{p_j}\rangle+|\overline{s}p_i\overline{p_j}\rangle+|\overline{s}\overline{p_i}p_j\rangle$\\
1534:
1535: \end{tabular}
1536:
1537: \\
1538:
1539: \hline $\LL^2=6$ &
1540:
1541: \begin{tabular}{c}
1542:
1543: $|sp_i\overline{p_j}\rangle-|s\overline{p_i}p_j\rangle$ \\
1544:
1545: $a|sp_1\overline{p_1}\rangle
1546: +b|sp_2\overline{p_2}\rangle
1547: +c|sp_3\overline{p_3}\rangle$ \\
1548:
1549: $|p_ip_j\overline{p_j}\rangle-|p_ip_k\overline{p_k}\rangle$ \\
1550:
1551: $a|\overline{p_3}p_1p_2\rangle
1552: +b|p_3p_1\overline{p_2}\rangle
1553: +c|p_3\overline{p_1}p_2\rangle$ \\
1554:
1555: \hline
1556:
1557: $|\overline{s}p_i\overline{p_j}\rangle-|\overline{s}\overline{p_i}p_j\rangle$\\
1558:
1559: $a|\overline{s}p_1\overline{p_1}\rangle
1560: +b|\overline{s}p_2\overline{p_2}\rangle
1561: +c|\overline{s}p_3\overline{p_3}\rangle$
1562: \\
1563:
1564: $|\overline{p_i}p_j\overline{p_j}\rangle
1565: -|\overline{p_i}p_k\overline{p_k}\rangle$ \\
1566:
1567: $a|p_3\overline{p_1}\overline{p_2}\rangle
1568: +b|\overline{p_3}p_1\overline{p_2}\rangle
1569: +c|\overline{p_3}\overline{p_1}p_2\rangle$ \\
1570:
1571: \end{tabular}
1572:
1573: &
1574: \\
1575: \hline
1576:
1577: \end{tabular}
1578: } % end scalebox
1579: \caption{Boron sequence $\protect\LL^2$-$\protect\SSS^2$
1580: eigenspaces.}
1581: \end{center}
1582: \end{table}
1583:
1584:
1585: % Carbon
1586:
1587: \begin{table}[h!tp]
1588: \hspace{23mm}
1589: \rotatebox{90}{ %
1590: \resizebox{0.93\textheight}{!}{
1591: \begin{tabular}{|r|c|c|c|}
1592: \hline & $\SSS^2=0$ & $\SSS^2=2$ & $\SSS^2=6$ \\
1593: \hline
1594: $\LL^2=0$ &
1595: %
1596: \begin{tabular}{c}
1597: $|s\overline{s}p_1\overline{p_1}\rangle
1598: +|s\overline{s}p_2\overline{p_2}\rangle
1599: +|s\overline{s}p_3\overline{p_3}\rangle$ \\
1600: %
1601: $|p_1\overline{p_1}p_2\overline{p_2}\rangle
1602: +|p_1\overline{p_1}p_3\overline{p_3}\rangle
1603: +|p_2\overline{p_2}p_3\overline{p_3}\rangle$\\
1604: \end{tabular}
1605: %
1606: &
1607: %
1608: \begin{tabular}{c}
1609: $3|\overline{s}p_1p_2p_3\rangle -|sp_1p_2\overline{p_3}\rangle
1610: -|s\overline{p_1}p_2p_3\rangle -|sp_1\overline{p_2}p_3\rangle$\\[3mm]
1611: \hline
1612: %
1613: $3|s\overline{p_1}\overline{p_2}\overline{p_3}\rangle
1614: -|\overline{s}\overline{p_1}\overline{p_2}p_3\rangle
1615: -|\overline{s}p_1\overline{p_2}\overline{p_3}\rangle
1616: -|\overline{s}\overline{p_1}p_2\overline{p_3}\rangle$
1617: \rule{0pt}{6mm}\\[3mm]
1618: \hline
1619: %
1620: $|sp_1\overline{p_2}\overline{p_3}\rangle +
1621: |s\overline{p_1}p_2\overline{p_3}\rangle +
1622: |s\overline{p_1}\overline{p_2}p_3\rangle$ \phantom{xxxxxx}\\
1623: \phantom{xxxxxx} $-|\overline{s}\overline{p_1}p_2p_3\rangle
1624: -|\overline{s}p_1\overline{p_2}p_3\rangle -
1625: |\overline{s}p_1p_2\overline{p_3}\rangle$ \rule{0pt}{6mm}\\
1626: \end{tabular}
1627: %
1628: &
1629: %
1630: \begin{tabular}{c}
1631: $|sp_1p_2p_3\rangle$\\
1632: \hline
1633: %
1634: $|\overline{s}\overline{p_1}\overline{p_2}\overline{p_3}\rangle$\\
1635: \hline
1636: %
1637: $|\overline{s}p_1p_2p_3\rangle +|sp_1p_2\overline{p_3}\rangle
1638: +|sp_1\overline{p_2}p_3\rangle+|s\overline{p_1}p_2p_3\rangle$\\
1639: \hline
1640: %
1641: $|s\overline{p_1}\overline{p_2}\overline{p_3}\rangle +
1642: |\overline{s}\overline{p_1}\overline{p_2}p_3\rangle+
1643: |\overline{s}\overline{p_1}p_2\overline{p_3}\rangle+
1644: |\overline{s}p_1\overline{p_2}\overline{p_3}\rangle $\\
1645: \hline
1646: %
1647: $|sp_1\overline{p_2}\overline{p_3}\rangle +
1648: |s\overline{p_1}p_2\overline{p_3}\rangle +
1649: |s\overline{p_1}\overline{p_2}p_3\rangle$ \phantom{xxxx}\\
1650: \phantom{xxxx}$+|\overline{s}\overline{p_1}p_2p_3\rangle
1651: +|\overline{s}p_1\overline{p_2}p_3\rangle +
1652: |\overline{s}p_1p_2\overline{p_3}\rangle$ \\
1653: \end{tabular}
1654: \\
1655: \hline
1656: %
1657: %
1658: $\LL^2=2$ &
1659: %
1660: \begin{tabular}{c}
1661: $|s\overline{p_i}p_j\overline{p_j}\rangle -
1662: |\overline{s}p_ip_j\overline{p_j}\rangle +
1663: |s\overline{p_i}p_k\overline{p_k}\rangle -
1664: |\overline{s}p_ip_k\overline{p_k}\rangle$\\
1665: \end{tabular}
1666: %
1667: &
1668: %
1669: \begin{tabular}{c}
1670: $|s\overline{s}p_ip_j\rangle$\\
1671: %
1672: $|p_1p_2p_3\overline{p_i}\rangle$ \\
1673: %
1674: $|sp_ip_j\overline{p_j}\rangle + |sp_ip_k\overline{p_k}\rangle$
1675: \\
1676: \hline
1677: %
1678: $|s\overline{s}\overline{p_i}\overline{p_j}\rangle$ \\
1679: %
1680: $|\overline{p_1}\overline{p_2}\overline{p_3}p_i\rangle$\\
1681: %
1682: $|\overline{s}\overline{p_i}p_j\overline{p_j}\rangle +
1683: |\overline{s}\overline{p_i}p_k\overline{p_k}\rangle$
1684: \\
1685: \hline
1686: %
1687: $|s\overline{s}p_i\overline{p_j}\rangle +
1688: |s\overline{s}\overline{p_i}p_j\rangle$\\
1689: %
1690: $|p_i\overline{p_i}p_j\overline{p_k}\rangle +
1691: |p_i\overline{p_i}\overline{p_j}p_k\rangle$ \\
1692: %
1693: $|s\overline{p_i}p_j\overline{p_j}\rangle
1694: +|\overline{s}p_ip_j\overline{p_j}\rangle
1695: +|s\overline{p_i}p_k\overline{p_k}\rangle
1696: +|\overline{s}p_ip_k\overline{p_k}\rangle$\\
1697: \end{tabular}
1698: &
1699: \\\hline
1700: %
1701: %
1702: $\LL^2=6$ &
1703: %
1704: \begin{tabular}{c}
1705: $|s\overline{s}p_i\overline{p_j}\rangle -
1706: |s\overline{s}\overline{p_i}p_j\rangle$\\
1707: %
1708: $a|s\overline{s}p_1\overline{p_1}\rangle +
1709: b|s\overline{s}p_2\overline{p_2}\rangle +
1710: c|s\overline{s}p_3\overline{p_3}\rangle$\\
1711: %
1712: $|p_i\overline{p_i}p_j\overline{p_k}\rangle -
1713: |p_i\overline{p_i}\overline{p_j}p_k\rangle$\\
1714: %
1715: $a|p_1\overline{p_1}p_2\overline{p_2}\rangle +
1716: b|p_1\overline{p_1}p_3\overline{p_3}\rangle +
1717: c|p_2\overline{p_2}p_3\overline{p_3}\rangle$ \\
1718: %
1719: $|s\overline{p_i}p_j\overline{p_j}\rangle-
1720: |\overline{s}p_ip_j\overline{p_j}\rangle
1721: -|s\overline{p_i}p_k\overline{p_k}\rangle +
1722: |\overline{s}p_ip_k\overline{p_k}\rangle$\\
1723: %
1724: $a(|s\overline{p_1}\overline{p_2}p_3\rangle
1725: +|\overline{s}p_1p_2\overline{p_3}\rangle)
1726: +b(|sp_1\overline{p_2}\overline{p_3}\rangle +
1727: |\overline{s}\overline{p_1}p_2p_3\rangle)$\\
1728: $+c(|s\overline{p_1}p_2\overline{p_3}\rangle
1729: +|\overline{s}p_1\overline{p_2}p_3\rangle)$\\
1730: \end{tabular}
1731: %
1732: &
1733: %
1734: \begin{tabular}{c}
1735: $|sp_ip_j\overline{p_j}\rangle-|sp_ip_k\overline{p_k}\rangle$\\
1736: %
1737: $a|sp_1p_2\overline{p_3}\rangle +b|sp_1\overline{p_2}p_3\rangle
1738: +c|s\overline{p_1}p_2p_3\rangle$\\
1739: \hline
1740: %
1741: $|\overline{s}\overline{p_i}p_j\overline{p_j}\rangle
1742: -|\overline{s}\overline{p_i}p_k\overline{p_k}\rangle$\\
1743: %
1744: $a|\overline{s}\overline{p_1}\overline{p_2}p_3\rangle
1745: +b|\overline{s}\overline{p_1}p_2\overline{p_3}\rangle
1746: +c|\overline{s}p_1\overline{p_2}\overline{p_3}\rangle$\\
1747: \hline
1748: %
1749: $|s\overline{p_i}p_j\overline{p_j}\rangle+|\overline{s}p_ip_j\overline{p_j}\rangle
1750: -|s\overline{p_i}p_k\overline{p_k}\rangle-|\overline{s}p_ip_k\overline{p_k}\rangle$\\
1751: %
1752: $a(|s\overline{p_1}\overline{p_2}p_3\rangle
1753: -|\overline{s}p_1p_2\overline{p_3}\rangle)
1754: +b(|sp_1\overline{p_2}\overline{p_3}\rangle -
1755: |\overline{s}\overline{p_1}p_2p_3\rangle)$\\
1756: $+c(|s\overline{p_1}p_2\overline{p_3}\rangle
1757: -|\overline{s}p_1\overline{p_2}p_3\rangle )$
1758: \end{tabular}
1759: & \\
1760: \hline
1761: \end{tabular}
1762: } % end resizebox
1763: } % end rotatebox
1764: \caption{Carbon sequence $\protect\LL^2$-$\protect\SSS^2$
1765: eigenspaces.}
1766: \end{table}
1767:
1768:
1769: % Nitrogen
1770: \begin{table}[h!tp]
1771: \begin{center}
1772: \resizebox{\textwidth}{!}{
1773: \begin{tabular}{|r|c|c|}
1774: \hline & $\SSS^2=\tfrac{3}{4}$ & $\SSS^2=\tfrac{15}{4}$ \\
1775:
1776: \hline $\LL^2=0$ &
1777:
1778: \begin{tabular}{c}
1779: $|sp_1\overline{p_1}p_2\overline{p_2}\rangle
1780: +|sp_1\overline{p_1}p_3\overline{p_3}\rangle
1781: +|sp_2\overline{p_2}p_3\overline{p_3}\rangle$\\[3mm]
1782:
1783: \hline \rule{0pt}{6mm}
1784:
1785: $|\overline{s}p_1\overline{p_1}p_2\overline{p_2}\rangle
1786: +|\overline{s}p_1\overline{p_1}p_3\overline{p_3}\rangle
1787: +|\overline{s}p_2\overline{p_2}p_3\overline{p_3}\rangle$\\
1788: \end{tabular}
1789:
1790: &
1791:
1792: \begin{tabular}{c}
1793: $|s\overline{s}p_1p_2p_3\rangle$\\
1794: \hline
1795:
1796: $|s\overline{s}\overline{p_1}\overline{p_2}\overline{p_3}\rangle$\\
1797: \hline
1798:
1799: $|s\overline{s}p_1\overline{p_2}p_3\rangle+|s\overline{s}\overline{p_1}p_2p_3\rangle+|s\overline{s}p_1p_2\overline{p_3}\rangle$
1800: \\ \hline
1801:
1802: $|s\overline{s}p_1\overline{p_2}\overline{p_3}\rangle+|s\overline{s}\overline{p_1}p_2\overline{p_3}\rangle+|s\overline{s}\overline{p_1}\overline{p_2}p_3\rangle$
1803: \\
1804: \end{tabular}
1805:
1806: \\
1807:
1808: \hline $\LL^2=2$ &
1809:
1810: \begin{tabular}{c}
1811: $|s\overline{s}p_ip_j\overline{p_j}\rangle
1812: +|s\overline{s}p_ip_k\overline{p_k}\rangle$\\
1813:
1814: $|p_1p_2p_3\overline{p_j}\overline{p_k}\rangle$ \\
1815:
1816: $2|\overline{s}p_i\overline{p_i}p_jp_k\rangle-|sp_i\overline{p_i}\overline{p_j}p_k\rangle
1817: -|sp_i\overline{p_i}p_j\overline{p_k}\rangle$\\
1818:
1819: \hline
1820:
1821: $|s\overline{s}\overline{p_i}p_j\overline{p_j}\rangle
1822: +|s\overline{s}\overline{p_i}p_k\overline{p_k}\rangle$\\
1823:
1824: $|\overline{p_1}\overline{p_2}\overline{p_3}p_jp_k\rangle$\\
1825:
1826: $2|sp_i\overline{p_i}\overline{p_j}\overline{p_k}\rangle
1827: -|\overline{s}p_i\overline{p_i}p_j\overline{p_k}\rangle
1828: -|\overline{s}p_i\overline{p_i}\overline{p_j}p_k\rangle$\\
1829:
1830: \end{tabular}
1831:
1832: &
1833:
1834: \begin{tabular}{c}
1835: $|sp_1p_2p_3\overline{p_i}\rangle$ \\ \hline
1836:
1837: $|\overline{s}\overline{p_1}\overline{p_2}\overline{p_3}p_i\rangle$
1838: \\ \hline
1839:
1840: $|sp_i\overline{p_i}p_j\overline{p_k}\rangle+
1841: |sp_i\overline{p_i}\overline{p_j}p_k\rangle
1842: +|\overline{s}p_i\overline{p_i}p_jp_k\rangle$ \\ \hline
1843:
1844: $|sp_i\overline{p_i}\overline{p_j}\overline{p_k}\rangle
1845: +|\overline{s}p_i\overline{p_i}p_j\overline{p_k}\rangle
1846: +|\overline{s}p_i\overline{p_i}\overline{p_j}p_k\rangle$\\
1847: \end{tabular}
1848:
1849: \\
1850:
1851: \hline $\LL^2=6$ &
1852:
1853: \begin{tabular}{c}
1854:
1855: $|s\overline{s}p_ip_j\overline{p_j}\rangle
1856: -|s\overline{s}p_ip_k\overline{p_k}\rangle$\\
1857:
1858: $a|s\overline{s}p_1p_2\overline{p_3}\rangle
1859: +b|s\overline{s}p_1\overline{p_2}p_3\rangle
1860: +c|s\overline{s}\overline{p_1}p_2p_3\rangle$\\
1861:
1862: $|sp_i\overline{p_i}p_j\overline{p_k}\rangle
1863: -|sp_i\overline{p_i}\overline{p_j}p_k\rangle$\\
1864:
1865: $a|sp_1\overline{p_1}p_2\overline{p_2}\rangle
1866: +b|sp_1\overline{p_1}p_3\overline{p_3}\rangle
1867: +c|sp_2\overline{p_2}p_3\overline{p_3}\rangle$\\
1868:
1869: \hline
1870:
1871: $|s\overline{s}\overline{p_i}p_j\overline{p_j}\rangle
1872: -|s\overline{s}\overline{p_i}p_k\overline{p_k}\rangle$\\
1873:
1874: $a|s\overline{s}\overline{p_1}\overline{p_2}p_3\rangle
1875: +b|s\overline{s}p_1\overline{p_2}\overline{p_3}\rangle
1876: +c|s\overline{s}\overline{p_1}p_2\overline{p_3}\rangle$\\
1877:
1878: $|\overline{s}p_i\overline{p_i}p_j\overline{p_k}\rangle
1879: -|\overline{s}p_i\overline{p_i}\overline{p_j}p_k\rangle$\\
1880:
1881: $a|\overline{s}p_1\overline{p_1}p_2\overline{p_2}\rangle
1882: +b|\overline{s}p_1\overline{p_1}p_3\overline{p_3}\rangle
1883: +c|\overline{s}p_2\overline{p_2}p_3\overline{p_3}\rangle$\\
1884:
1885: \end{tabular}
1886:
1887: &
1888: \\
1889: \hline
1890:
1891: \end{tabular}
1892:
1893: } % end resizebox
1894:
1895: \end{center}
1896: \caption{Nitrogen sequence
1897: $\protect\LL^2$-$\protect\SSS^2$ eigenspaces.}
1898: \label{Tab:NFullTable}
1899: \end{table}
1900:
1901: % Oxygen
1902: \begin{table}[h!tp]
1903: \begin{center}
1904: \resizebox{\textwidth}{!}{
1905: \begin{tabular}{|r|c|c|}
1906: \hline & $\SSS^2=0$ & $\SSS^2=2$ \\
1907:
1908: \hline $\LL^2=0$ &
1909:
1910: \begin{tabular}{c}
1911: $|s\overline{s}p_1\overline{p_1}p_2\overline{p_2}\rangle
1912: +|s\overline{s}p_1\overline{p_1}p_3\overline{p_3}\rangle
1913: +|s\overline{s}p_2\overline{p_2}p_3\overline{p_3}\rangle$\\
1914:
1915: $|p_1\overline{p_1}p_2\overline{p_2}p_3\overline{p_3}\rangle$\\
1916: \end{tabular}
1917: & \\
1918:
1919: \hline
1920:
1921: $\LL^2=2$ &
1922:
1923: \begin{tabular}{c}
1924: $|s\overline{p_1}\overline{p_2}\overline{p_3}p_jp_k\rangle
1925: -|\overline{s}p_1p_2p_3\overline{p_j}\overline{p_k}\rangle$\\
1926: \end{tabular}
1927:
1928: &
1929:
1930: \begin{tabular}{c}
1931: $|s\overline{s}p_1p_2p_3\overline{p_i}\rangle$\\
1932:
1933: $|sp_1p_2p_3\overline{p_j}\overline{p_k}\rangle$\\
1934:
1935: \hline
1936:
1937: $|s\overline{s}\overline{p_1}\overline{p_2}\overline{p_3}p_i\rangle$\\
1938:
1939: $|\overline{s}\overline{p_1}\overline{p_2}\overline{p_3}p_jp_k\rangle$\\
1940:
1941: \hline
1942:
1943: $|s\overline{s}p_i\overline{p_i}p_j\overline{p_k}\rangle
1944: +|s\overline{s}p_i\overline{p_i}\overline{p_j}p_k\rangle$\\
1945:
1946: $|s\overline{p_1}\overline{p_2}\overline{p_3}p_jp_k\rangle
1947: +|\overline{s}p_1p_2p_3\overline{p_j}\overline{p_k}\rangle$\\
1948:
1949: \end{tabular}
1950:
1951: \\
1952: \hline
1953:
1954: $\LL^2=6$ &
1955:
1956: \begin{tabular}{c}
1957: $|s\overline{s}p_i\overline{p_i}p_j\overline{p_k}\rangle
1958: -|s\overline{s}p_i\overline{p_i}\overline{p_j}p_j\rangle$ \\
1959:
1960: $a|s\overline{s}p_1\overline{p_1}p_2\overline{p_2}\rangle
1961: +b|s\overline{s}p_1\overline{p_1}p_3\overline{p_3}\rangle
1962: +c|s\overline{s}p_2\overline{p_2}p_3\overline{p_3}\rangle$ \\
1963: \end{tabular}
1964: & \\
1965:
1966: \hline
1967:
1968: \end{tabular}
1969: } % end scalebox
1970: \end{center}
1971:
1972: \caption{Oxygen sequence $\protect\LL^2$-$\protect\SSS^2$
1973: eigenspaces.}
1974: \end{table}
1975:
1976: % Fluorine
1977: \begin{table}[h!tp]
1978: \begin{minipage}{0.5\textwidth}
1979: \begin{center}
1980: \begin{tabular}{|r|c|}
1981: \hline & $\SSS^2=\tfrac{3}{4}$ \\
1982:
1983: \hline
1984:
1985: $\LL^2=0$ &
1986:
1987: \begin{tabular}{c}
1988: $|sp_1\overline{p_1}p_2\overline{p_2}p_3\overline{p_3}\rangle$
1989: \\
1990: \hline
1991:
1992: $|\overline{s}p_1\overline{p_1}p_2\overline{p_2}p_3\overline{p_3}\rangle$\\
1993: \end{tabular}
1994:
1995: \\
1996: \hline
1997:
1998: $\LL^2=2$ &
1999:
2000: \begin{tabular}{c}
2001: $|s\overline{s}p_1p_2p_3\overline{p_i}\overline{p_j}\rangle$\\
2002: \hline
2003:
2004: $|s\overline{s}\overline{p_1}\overline{p_2}\overline{p_3}p_ip_j\rangle$\\
2005: \end{tabular}
2006: \\
2007: \hline
2008:
2009: \end{tabular}
2010: \end{center}
2011: \caption{Fluorine sequence
2012: $\protect\LL^2$-$\protect\SSS^2$ eigenspaces.}
2013: \label{Tab:FFullTable}
2014: \end{minipage}
2015: % Neon
2016: \begin{minipage}{0.5\textwidth}
2017: \begin{center}
2018: \begin{tabular}{|r|c|}
2019: \hline & $\SSS^2=0$ \\
2020: \hline
2021:
2022: $\LL^2=0$ & $|s\overline{s}p_1\overline{p_1}p_2\overline{p_2}p_3\overline{p_3}\rangle$ \\
2023: \hline
2024:
2025: \end{tabular}
2026: \end{center}
2027: \caption{Neon sequence $\protect\LL^2$-$\protect\SSS^2$
2028: eigenspaces.} \label{Tab:NeFullTable}
2029: \end{minipage}
2030: \end{table}
2031:
2032: Inspecting these eigenspaces reveals a number of interesting properties.
2033: %
2034: %
2035: %
2036: \begin{corollary} For any $N=3,\dots,10$, the maximum dimension of any simultaneous eigenspace within $V_0(N)$ of
2037: the operators (\ref{ops}) is two.
2038: \end{corollary}
2039: %
2040: %
2041: %
2042: As regards diagonalization of the Hamiltonian,
2043: this is clearly much more promising
2044: than the 70-dimensional space of Carbon.
2045: %
2046: %
2047: %
2048: \begin{corollary}
2049: \label{C:SDDifferByEven} For any $N=3,\dots,10$, and any simultaneous
2050: $\LL^2$-$\SSS^2$-$L_3$-$S_3$-$\hat{R}$ eigenspace within $V_0(N)$ with $L_3$ eigenvalue equal to zero, all Slater
2051: determinants occuring within the space differ by an even
2052: number of orbitals.
2053: \end{corollary}
2054: %
2055: %
2056: %
2057: This is remarkable, and will greatly simplify the structure of the
2058: Hamiltonian matrix in the basis (\ref{GS2ndperiod}), due to
2059: the simpler structure of Slater's rules (see below). Also, it
2060: implies that even for the correlated eigenstates of the Hamiltonian,
2061: the orbitals (\ref{Vorbitals}) are natural orbitals in the sense of
2062: L\"owdin. A more abstract proof of
2063: Corollary \ref{C:SDDifferByEven} will be given elsewhere.
2064: %
2065: %
2066: %
2067: \subsection{Symbolic interaction matrix}\label{Sec:SymbolicVee}
2068: %
2069: %
2070: %
2071: In order to calculate the Hamiltonian matrix $PHP$ on each $\LL^2$-$\SSS^2$-$\hat{R}$ eigenspace, note first
2072: that $PHP$ commutes with (\ref{ops}) (see Lemma \ref{PTsymm}, which remains valied for the more general orbitals (\ref{GeneralOrbitals}), cf.
2073: the calculus in Section \ref{S:7.1}). Hence it suffices to pick arbitrary components of $\LL$ and $\SSS$, say $L_3$ and $S_3$, and
2074: calculate this matrix on the $\LL^2$-$L_3$-$\SSS^2$-$S_3$-$\hat{R}$ eigenspace with maximal $S_3$ and $L_3=0$.
2075: These spaces are shown in Tables \ref{Tab:VeeMatrix1}--\ref{Tab:VeeMatrix3}.
2076: Here, because of their importance for the interaction energy, the
2077: $1s$ orbitals are shown and the eigenfunctions are normalized. We
2078: find it convenient to abbreviate the spin orbitals
2079: \be \label{not1}
2080: \varphi_{1s}\!\uparrow,\;\varphi_{1s}\!\downarrow,\;\varphi_{2s}\!\uparrow, \;
2081: \varphi_{2s}\!\downarrow,\;\varphi_{2p_3}\!\uparrow,\;\varphi_{2p_3}\!\downarrow, \;
2082: \varphi_{2p_1}\!\uparrow,\;\varphi_{2p_1}\!\downarrow, \;
2083: \varphi_{2p_2}\!\uparrow,\;\varphi_{2p_2}\!\downarrow
2084: \ee
2085: (even more drastically than in Tables
2086: \ref{Tab:LiFullTable}--\ref{Tab:NeFullTable}) by
2087: \be \label{not2}
2088: 1,\;\overline{1},\;2,\;\overline{2},\;3,\;\overline{3},\;4,\;\overline{4},\;5,\;\overline{5}.
2089: \ee
2090: Thus, for example, the top Carbon state of Table \ref{Tab:VeeMatrix2},
2091: $$
2092: \frac{1}{\sqrt{3}}( |1\overline{1}2\overline{2}3\overline{3}\rangle
2093: +|1\overline{1}2\overline{2}4\overline{4}\rangle +
2094: |1\overline{1}2\overline{2}5\overline{5}\rangle),
2095: $$
2096: stands for
2097: \begin{align*}
2098: \frac{1}{\sqrt{3}}\Bigl( |\varphi_{1s}\!\!\uparrow \varphi_{1s}\!\!\downarrow
2099: \varphi_{2s}\!\!\uparrow \varphi_{2s}\!\!\downarrow \varphi_{2p_3}\!\!\uparrow
2100: \varphi_{2p_3}\!\!\downarrow \rangle & + |\varphi_{1s}\!\!\uparrow
2101: \varphi_{1s}\!\!\downarrow \varphi_{2s}\!\!\uparrow \varphi_{2s}\!\!\downarrow
2102: \varphi_{2p_1}\!\!\uparrow \varphi_{2p_1}\!\!\downarrow \rangle \\
2103: & +
2104: |\varphi_{1s}\!\!\uparrow \varphi_{1s}\!\!\downarrow \varphi_{2s}\!\!\uparrow
2105: \varphi_{2s}\!\!\downarrow \varphi_{2p_2}\!\!\uparrow \varphi_{2p_2}\!\!\downarrow
2106: \rangle\Bigr).
2107: \end{align*}
2108:
2109: We begin by analyzing the $V_{ee}$ matrix elements between
2110: the eigenfunctions of Tables
2111: \ref{Tab:VeeMatrix1}--\ref{Tab:VeeMatrix3}. Using Slater's rules
2112: \cite[Section 2.3]{SzaboOstlund96}, these are straightforward to
2113: express in terms of Coulomb and exchange integrals $(aa|bb)$ and
2114: $(ab|ba)$ of the spatial orbitals (\ref{GeneralOrbitals}), where -- in common notation --
2115: \be
2116: (ab | cd)
2117: = \int_{\R^6} dx_1 dx_2 a^*(x_1)b(x_1)
2118: \frac{1}{|x_1-x_2|} c^*(x_2)d(x_2).
2119: \label{ijklIntegral}
2120: \ee
2121:
2122: \begin{lemma} \label{L:7.3}
2123: Let $N\in\{3,\dots,10\}$.
2124: Orthonormal bases of the simultaneous $\LL^2$-$\SSS^2$-$L_3$-$S_3$-$\hat{R}$
2125: eigenspaces within $V_0(N)$ with $S_3$ maximal and $L_3=0$, and the corresponding
2126: $V_{ee}$ matrix elements $\langle\Psi|V_{ee}|\tilde{\Psi}\rangle$ in terms of
2127: Coulomb and exchange integrals of the one-electron
2128: orbitals (\ref{GeneralOrbitals}), are as given in Tables
2129: \ref{Tab:VeeMatrix1}-\ref{Tab:VeeMatrix3}. Here the orbitals are abbreviated as in (\ref{not1})--(\ref{not2}),
2130: and the off-diagonal matrix elements ($\Psi\neq\tilde{\Psi}$) in the
2131: two-dimensional eigenspaces are denoted by ``cross''.
2132: \end{lemma}
2133:
2134: The shortness of the expressions for the $\langle\Psi|V_{ee}|\tilde{\Psi}\rangle$, and the absence of
2135: Coulomb and exchange integrals involving the last orbital, comes from
2136: the absence of single excitations (Cor.~\ref{C:SDDifferByEven}) and the equivalence of the $p$ orbitals
2137: in (\ref{GeneralOrbitals}) up to rotation. The latter would be destroyed by changing to a basis of $L_3$ eigenfunctions
2138: (which is why we have not done so even though this would have been more convenient
2139: for the diagonalization of $\LL^2$ in the previous subsection).
2140:
2141:
2142: \begin{table}[htbp]
2143: \begin{center}
2144:
2145: \resizebox{\textwidth}{!}{
2146: \begin{tabular}{|c|c|c|c|c|l|}
2147: \hline & $\LL^2$&
2148: $\SSS^2$ & $\hat{R}$& $\Psi$
2149: & \multicolumn{1}{c|}{$\langle V_{ee}\rangle$} \\
2150: \hline
2151:
2152: Li & $0$ & $\tfrac{3}{4}$ & $1$ & $|1\overline{1}2\rangle$ & $(11|11) + 2(11|22) - (12|21)$\\
2153:
2154: \cline{2-6} & $2$ & $\tfrac{3}{4}$ & $-1$ & $|1\overline{1}3\rangle$ &
2155: $(11|11) + 2(11|33)-(13|31)$
2156: \\
2157: \hline
2158:
2159: Be & $0$ & $0$ & $1$ & $|1\overline{1}2\overline{2}\rangle$ &$(11|11) +
2160: 4(11|22) - 2(12|21) + (22|22)$\\
2161:
2162: \cline{5-6} & & & &
2163: $\tfrac{1}{\sqrt{3}}\left(|1\overline{1}3\overline{3}\rangle +
2164: |1\overline{1}4\overline{4}\rangle +
2165: |1\overline{1}5\overline{5}\rangle \right)$ & $(11|11) + 4(11|33)
2166: - 2(13|31) + (33|33) + 2(34|43)$ \\
2167:
2168: \cline{5-6} & & & & cross &
2169: $\sqrt{3}(23|32)$ \\
2170:
2171: \cline{2-6} & $2$ & $0$ & $-1$ &
2172: $\tfrac{1}{\sqrt{2}}\left(|1\overline{1}2\overline{3}\rangle
2173: -|1\overline{1}\overline{2}3\rangle\right)$ & $(11|11) + 2(11|22)
2174: - (12|21) +2(11|33)- (13|31)$ \\
2175: & & & & & $ + (22|33) + (23|32)$ \\
2176:
2177: \cline{2-6} & $2$ & $2$ & $-1$ & $|1\overline{1}23\rangle$ & $(11|11) +
2178: 2(11|22) - (12|21) + 2(11|33)- (13|31)$ \\
2179: & & & & & $ + (22|33) - (23|32)$\\
2180:
2181: \cline{4-6} & & & $1$ & $|1\overline{1}45\rangle$ &$(11|11) + 4(11|33)
2182: -2(13|31) + (33|44) - (34|43)$ \\
2183:
2184: \cline{2-6} & $6$ & $0$ & $1$ &
2185: $\tfrac{1}{\sqrt{6}}\left(2|1\overline{1}3\overline{3}\rangle-
2186: |1\overline{1}4\overline{4}\rangle
2187: -|1\overline{1}5\overline{5}\rangle \right)$
2188: &$(11|11) + 4(11|33)
2189: -2(13|31) + (33|33) -(34|43)$\\
2190:
2191: \hline B & $0$ & $\tfrac{3}{4}$ & $1$ &
2192: $\tfrac{1}{\sqrt{3}}\left(|1\overline{1}23\overline{3}\rangle +
2193: |1\overline{1}24\overline{4}\rangle +
2194: |1\overline{1}25\overline{5}\rangle\right)$&
2195: $(11|11) + 2(11|22) - (12|21) + 4(11|33) -2(13|31)$ \\
2196: & & & & & $+ 2(22|33) - (23|32) + (33|33) + 2(34|43)$\\
2197:
2198: \cline{2-6} & $0$ & $\tfrac{15}{4}$ & $-1$ & $|1\overline{1}345\rangle$ &
2199: $(11|11) + 6(11|33) -3(13|31) + 3(33|44) -3(34|43)$ \\
2200:
2201: \cline{2-6} & $2$ & $\tfrac{3}{4}$ & $-1$ &
2202: $|1\overline{1}2\overline{2}{3}\rangle$ &
2203: $(11|11) + 4(11|22) -2(12|21) + 2(11|33) - (13|31) $\\
2204: & & & & & $ + (22|22) + 2(22|33) - (23|32)$ \\
2205:
2206: \cline{5-6} & & & &
2207: $\tfrac{1}{\sqrt{2}}\left(|1\overline{1}34\overline{4}\rangle +
2208: |1\overline{1}35\overline{5}\rangle \right)$ & $ (11|11) +
2209: 6(11|33) -3(13|31) + (33|33) + 2(33|44)$ \\
2210:
2211: \cline{5-6} & & & & cross & $\sqrt{2}(23|32)$\\
2212:
2213: \cline{4-6} & & & $1$ &
2214: $\tfrac{1}{\sqrt{6}}\left(2|1\overline{1}\overline{2}45\rangle
2215: -|1\overline{1}2\overline{4}5\rangle
2216: -|1\overline{1}24\overline{5}\rangle \right)$
2217: & $(11|11) + 2(11|22) -(12|21) + 4(11|33) -2(13|31) $ \\
2218: & & & & & $+ 2(22|33) + (23|32) + (33|44) -(34|43) $ \\
2219:
2220: \cline{2-6} & $2$ & $\tfrac{15}{4}$ & $1$ & $|1\overline{1}245\rangle$ &
2221: $(11|11) + 2(11|22) - (12|21) + 4(11|33) -2(13|31)$ \\
2222: & & & & & $+ 2(22|33) -2(23|32) + (33|44) - (34|43)$\\
2223:
2224: \cline{2-6} & $6$ & $\tfrac{3}{4}$ & $1$ &
2225: $\tfrac{1}{\sqrt{6}}\left(2|1\overline{1}23\overline{3}\rangle
2226: -|1\overline{1}24\overline{4}\rangle
2227: -|1\overline{1}25\overline{5}\rangle \right)$
2228: & $(11|11) + 2(11|22) -(12|21) + 4(11|33) -2(13|31)$ \\
2229: & & & & & $ + 2(22|33) -(23|32) + (33|33) -(34|43)$ \\
2230:
2231: \cline{4-6} & & & $-1$ &
2232: $\tfrac{1}{\sqrt{6}}\left(2|1\overline{1}\overline{3}45\rangle
2233: -|1\overline{1}3\overline{4}5\rangle
2234: -|1\overline{1}34\overline{5}\rangle \right)$ & $(11|11) +
2235: 6(11|33) -3(13|31) + 3(33|44)$ \\
2236: \hline
2237:
2238: \end{tabular}
2239: } % end resizebox
2240: \end{center}
2241: \caption[$V_{ee}$ matrix element expressions for Li-B]
2242: {$V_{ee}$ matrix element expressions for the
2243: Li-B sequences, `cross' denotes the
2244: off-diagonal term in the $2\times2$ matrix. See Lemma \ref{L:7.3} for notation.}
2245: \label{Tab:VeeMatrix1}
2246: \end{table}
2247:
2248: \begin{table}[htbp]
2249: \begin{center}
2250:
2251: \resizebox{\textwidth}{!}{
2252: \begin{tabular}{|c|c|c|c|c|l|}
2253: \hline & $\LL^2$&
2254: $\SSS^2$ & $\hat{R}$& $\Psi$
2255: & \multicolumn{1}{c|}{$\langle V_{ee} \rangle$} \\
2256: \hline
2257:
2258: C & $0$ & $0$ & $1$ &
2259: $\tfrac{1}{\sqrt{3}}\left(|1\overline{1}2\overline{2}3\overline{3}\rangle
2260: + |1\overline{1}2\overline{2}4\overline{4}\rangle +
2261: |1\overline{1}2\overline{2}5\overline{5}\rangle \right) $
2262: & $(11|11) + 4(11|22) -2(12|21) + 4(11|33) -2(13|31)$ \\
2263: & & & & & $+ (22|22) + 4(22|33) -2(23|32) + (33|33) + 2(34|43)$\\
2264:
2265: \cline{5-6} & & & &
2266: $\tfrac{1}{\sqrt{3}}\left(|1\overline{1}3\overline{3}4\overline{4}\rangle
2267: + |1\overline{1}3\overline{3}5\overline{5}\rangle +
2268: |1\overline{1}4\overline{4}5\overline{5}\rangle \right) $
2269: & $(11|11) + 8(11|33) -4(13|31) + 2(33|33) + 4(33|44)$ \\
2270:
2271: \cline{5-6} & & & &
2272: cross
2273: &$2(23|32)$ \\
2274:
2275: \cline{2-6} & $0$ & $2$ & $-1$ & $\hspace{-15mm} \tfrac{1}{\sqrt{12}}\big(
2276: 3|1\overline{1}\overline{2}345\rangle -
2277: |1\overline{1}2\overline{3}45\rangle $
2278: & $(11|11) + 2(11|22) - (12|21) + 6(11|33) -3(13|31) $ \\
2279: & & & & $\hspace{15mm} -|1\overline{1}23\overline{4}5\rangle-
2280: |1\overline{1}234\overline{5}\rangle \big)$ &
2281: $+ 3(22|33) + (23|32) + 3(33|44) -3(34|43) $ \\
2282:
2283: \cline{2-6} & $0$ & $6$ & $-1$ & $|1\overline{1}2345\rangle$ & $(11|11) +
2284: 2(11|22) - (12|21) + 6(11|33) -3(13|31)$\\
2285: & & & & & $+ 3(22|33) -3(23|32) + 3(33|44) -3(34|43)$ \\
2286:
2287: \cline{2-6} & $2$ & $0$ & $-1$ &
2288: $\hspace{-15mm} \tfrac{1}{2}\big(|1\overline{1}2\overline{3}4\overline{4}\rangle -
2289: |1\overline{1}\overline{2}34\overline{4}\rangle$
2290: & $(11|11) + 2(11|22) - (12|21) + 6(11|33) -3(13|31)$ \\
2291: & & & & $\hspace{15mm} +|1\overline{1}2\overline{3}5\overline{5}\rangle
2292: - |1\overline{1}\overline{2}35\overline{5}\rangle \big)$
2293: & $+ 3(22|33) + (33|33) + 2(33|44)$ \\
2294:
2295: \cline{2-6} & $2$ & $2$ & $1$ & $|1\overline{1}2\overline{2}45\rangle$
2296: & $(11|11) + 4(11|22) -2(12|21) + 4(11|33) -2(13|31)$ \\
2297: & & & & & $ + (22|22) + 4(22|33) -2(23|32) + (33|44) - (34|43)$ \\
2298:
2299: \cline{5-6} & & & & $|1\overline{1}3\overline{3}45\rangle$ &
2300: $(11|11) + 8(11|33) -4(13|31) + (33|33) + 5(33|44)$ \\
2301: & & & & & $-3(34|43)$\\
2302:
2303: \cline{5-6} & & & & cross & $(23|32)$ \\
2304:
2305: \cline{4-6} & & & $-1$ &
2306: $\tfrac{1}{\sqrt{2}}\left(|1\overline{1}234\overline{4}\rangle +
2307: |1\overline{1}235\overline{5}\rangle \right)$
2308: & $(11|11) + 2(11|22) -(12|21) + 6(11|33) -3(13|31)$ \\
2309: & & & & & $+ 3(22|33) -2(23|32) + (33|33) + 2(33|44)$ \\
2310:
2311: \cline{2-6} & $6$ & $0$ & $1$ & $\tfrac{1}{\sqrt{6}} \left(
2312: 2|1\overline{1}2\overline{2}3\overline{3}\rangle -
2313: |1\overline{1}2\overline{2}4\overline{4}\rangle -
2314: |1\overline{1}2\overline{2}5\overline{5}\rangle\right)$
2315: & $(11|11) + 4(11|22) -2(12|21) + 4(11|33) -2(13|31)$ \\
2316: & & & & & $+ (22|22)+ 4(22|33) -2(23|32) + (33|33) -(34|43)$ \\
2317:
2318: \cline{5-6} & & & & $\tfrac{1}{\sqrt{6}} \left(
2319: 2|1\overline{1}4\overline{4}5\overline{5}\rangle -
2320: |1\overline{1}3\overline{3}4\overline{4}\rangle -
2321: |1\overline{1}3\overline{3}5\overline{5}\rangle \right)$ &
2322: $(11|11) + 8(11|33) -4(13|31) + 2(33|33) + 4(33|44)$ \\
2323: & & & & & $-3(34|43)$ \\
2324:
2325: \cline{5-6} & & & & cross &
2326: $-(23|32)$\\
2327:
2328: \cline{4-6} & & & $-1$ &$\tfrac{1}{\sqrt{12}} \big(
2329: 2|1\overline{1}23\overline{4}\overline{5}\rangle
2330: -|1\overline{1}2\overline{3}4\overline{5}\rangle
2331: -|1\overline{1}2\overline{3}\overline{4}5\rangle$ &
2332: $(11|11) + 2(11|22) - (12|21) + 6(11|33) -3(13|31)$ \\
2333: & & & & $+2|1\overline{1}\overline{2}\overline{3}45\rangle
2334: -|1\overline{1}\overline{2}34\overline{5}\rangle
2335: -|1\overline{1}\overline{2}3\overline{4}5\rangle \big)$
2336: & $+ 3(22|33) + 3(33|44)$\\
2337:
2338: \cline{2-6} & $6$ & $2$ & $-1$ & $\tfrac{1}{\sqrt{6}} \big(
2339: 2|1\overline{1}2\overline{3}45\rangle
2340: -|1\overline{1}234\overline{5}\rangle -
2341: |1\overline{1}23\overline{4}5\rangle\big)$
2342: & $(11|11) + 2(11|22) - (12|21) + 6(11|33) -3(13|31)$\\
2343:
2344: & & & & &
2345: $+ 3(22|33) -2(23|32) + 3(33|44)$\\
2346: \hline
2347:
2348: \end{tabular}
2349:
2350: } % end resizebox
2351:
2352: \end{center}
2353:
2354: \caption[$V_{ee}$ matrix element expressions for C]
2355: {$V_{ee}$ matrix element expressions for the
2356: C sequence, `cross' denotes the
2357: off-diagonal term in the $2\times2$ matrix. See Lemma \ref{L:7.3} for notation.}
2358: \label{Tab:VeeMatrix2}
2359: \end{table}
2360:
2361:
2362: \begin{table}[htbp]
2363: \begin{center}
2364:
2365: \resizebox{\textwidth}{!}{
2366: \begin{tabular}{|c|c|c|c|c|l|}
2367: \hline & $\LL^2$&
2368: $\SSS^2$ & $\hat{R}$& $\Psi$
2369: & \multicolumn{1}{c|}{$\langle V_{ee} \rangle$} \\
2370: \hline
2371:
2372: N & $0$ & $\tfrac{3}{4}$ & $1$ & $\tfrac{1}{\sqrt{3}} \big(
2373: |1\overline{1}23\overline{3}4\overline{4}\rangle +
2374: |1\overline{1}23\overline{3}5\overline{5}\rangle $
2375: & $(11|11) + 2(11|22) - (12|21) + 8(11|33) -4(13|31)$ \\
2376: & & & & $+|1\overline{1}24\overline{4}5\overline{5}\rangle \big)$
2377: & $+ 4(22|33) -2(23|32) + 2(33|33) + 4(33|44)$ \\
2378:
2379: \cline{2-6} & $0$ & $\tfrac{15}{4}$ & $-1$ &
2380: $|1\overline{1}2\overline{2}345\rangle$
2381: & $(11|11) + 4(11|22) -2(12|21) + 6(11|33) -3(13|31)$ \\
2382: & & & & & $+ (22|22)+ 6(22|33) -3(23|32) + 3(33|44) -3(34|43)$ \\
2383:
2384: \cline{2-6} & $2$ & $\tfrac{3}{4}$ & $-1$ &
2385: $|1\overline{1}34\overline{4}5\overline{5}\rangle$ & $(11|11) +
2386: 10(11|33) -5(13|31) + 2(33|33) + 8(33|44)$ \\
2387: & & & & & $ -4(34|43)$ \\
2388:
2389: \cline{5-6} & & & & $\tfrac{1}{\sqrt{2}}\left(
2390: |1\overline{1}2\overline{2}34\overline{4}\rangle +
2391: |1\overline{1}2\overline{2}35\overline{5}\rangle \right)$ &
2392: $(11|11) + 4(11|22) -2(12|21) + 6(11|33) -3(13|31)$\\
2393: & & & & & $+ (22|22)+ 6(22|33) -3(23|32) + (33|33) + 2(33|44)$ \\
2394:
2395: \cline{5-6} & & & & cross & $\sqrt{2}(23|32)$\\
2396:
2397: \cline{4-6} & & & $1$ & $\tfrac{1}{\sqrt{6}} \big(
2398: 2|1\overline{1}\overline{2}3\overline{3}45\rangle -
2399: |1\overline{1}23\overline{3}\overline{4}5\rangle $
2400: & $(11|11) + 2(11|22) -(12|21) + 8(11|33) -4(13|31) $ \\
2401: & & & & $- |1\overline{1}23\overline{3}4\overline{5}\rangle \big)$
2402: & $+ 4(22|33) + (33|33) + 5(33|44) -3(34|43)$ \\
2403:
2404: \cline{2-6} & $2$ & $\tfrac{15}{4}$ & $1$ &
2405: $|1\overline{1}23\overline{3}45\rangle$
2406: & $(11|11) + 2(11|22) - (12|21) + 8(11|33) -4(13|31)$ \\
2407: & & & & & $+ 4(22|33) -3(23|32) + (33|33) + 5(33|44) -3(34|43)$ \\
2408:
2409: \cline{2-6} & $6$ & $\tfrac{3}{4}$ & $-1$ &
2410: $\tfrac{1}{\sqrt{6}}\big(2|1\overline{1}2\overline{2}\overline{3}45\rangle
2411: - |1\overline{1}2\overline{2}34\overline{5}\rangle$
2412: & $(11|11) + 4(11|22) -2(12|21) + 6(11|33) -3(13|31)$ \\
2413: & & & & $- |1\overline{1}2\overline{2}3\overline{4}5\rangle \big)$
2414: & $+ (22|22)+ 6(22|33) -3(23|32) + 3(33|44)$ \\
2415:
2416: \cline{4-6} & & & $1$ &
2417: $\tfrac{1}{\sqrt{6}}\big(2|1\overline{1}24\overline{4}5\overline{5}\rangle
2418: - |1\overline{1}23\overline{3}4\overline{4}\rangle$
2419: & $(11|11) + 2(11|22) -(12|21) + 8(11|33) -4(13|31)$ \\
2420: & & & & $-|1\overline{1}23\overline{3}5\overline{5}\rangle\big)$
2421: & $+ 4(22|33) -2(23|32) + 2(33|33) + 4(33|44) -3(34|43)$ \\
2422:
2423: \hline
2424:
2425: O & $0$ & $0$ & $1$ & $\tfrac{1}{\sqrt{3}}\big(
2426: |1\overline{1}2\overline{2}3\overline{3}4\overline{4}\rangle +
2427: |1\overline{1}2\overline{2}3\overline{3}5\overline{5}\rangle$ &
2428: $(11|11) +
2429: 4(11|22) -2(12|21) + 8(11|33) -4(13|31)$\\
2430: & & & & $+
2431: |1\overline{1}2\overline{2}4\overline{4}5\overline{5}\rangle\big)$
2432: & $+
2433: (22|22) + 8(22|33) -4(23|32) + 2(33|33) + 4(33|44)$\\
2434:
2435: \cline{5-6} & & & & $|1\overline{1}3\overline{3}4\overline{4}5\overline{5}\rangle$ &
2436: $(11|11) + 12(11|33) -6(13|31) + 3(33|33)+ 12(33|44)$\\
2437: & & & & & $ -6(34|43)$ \\
2438:
2439: \cline{5-6} & & & & cross &
2440: $\sqrt{3}(23|32)$ \\
2441:
2442: \cline{2-6} & $2$ & $0$ & $-1$ & $\tfrac{1}{\sqrt{2}}\left(
2443: |1\overline{1}2\overline{3}4\overline{4}5\overline{5}\rangle-
2444: |1\overline{1}\overline{2}34\overline{4}5\overline{5}\rangle\right)$
2445: & $(11|11) + 2(11|22) -(12|21) + 10(11|33) -5(13|31)$ \\
2446: & & & & & $+ 5(22|33) -(23|32) + 2(33|33) + 8(33|44) -4(34|43)$ \\
2447:
2448: \cline{2-6} & $2$ & $2$ & $-1$ &
2449: $|1\overline{1}234\overline{4}5\overline{5}\rangle$
2450: & $(11|11) + 2(11|22) - (12|21) + 10(11|33) -5(13|31)$ \\
2451: & & & & & $ + 5(22|33)-3(23|32) + 2(33|33) + 8(33|44) -4(34|43)$ \\
2452:
2453: \cline{4-6} & & & $1$ &
2454: $|1\overline{1}2\overline{2}3\overline{3}45\rangle$
2455: & $(11|11) + 4(11|22) -2(12|21) + 8(11|33) -4(13|31)$ \\
2456: & & & & & $+ (22|22) + 8(22|33) -4(23|32) + (33|33) + 5(33|44)$ \\
2457: & & & & & $ -3(34|43)$ \\
2458:
2459: \cline{2-6} & $6$ & $0$ & $1$ & $\tfrac{1}{\sqrt{6}}\big(
2460: 2|1\overline{1}2\overline{2}4\overline{4}5\overline{5}\rangle-
2461: |1\overline{1}2\overline{2}3\overline{3}4\overline{4}\rangle$
2462: & $(11|11) + 4(11|22) -2(12|21) + 8(11|33) -4(13|31)$ \\
2463: & & & &
2464: $-|1\overline{1}2\overline{2}3\overline{3}5\overline{5}\rangle
2465: \big)$
2466: & $+ (22|22) + 8(22|33) -4(23|32) + 2(33|33) + 4(33|44)$\\
2467: & & & & & $-3(34|43)$\\
2468: \hline
2469:
2470: F & $0$ & $\tfrac{3}{4}$ & $1$ &
2471: $|1\overline{1}23\overline{3}4\overline{4}5\overline{5}\rangle$
2472: & $(11|11) + 2(11|22) - (12|21) + 12(11|33) -6(13|31)$ \\
2473: & & & & & $ + 6(22|33) -3(23|32) + 3(33|33) + 12(33|44) -6(34|43)$ \\
2474:
2475: \cline{2-6} & $2$ & $\tfrac{3}{4}$ & $-1$ &
2476: $|1\overline{1}2\overline{2}34\overline{4}5\overline{5}\rangle$
2477: & $(11|11) + 4(11|22) -2(12|21) + 10(11|33) -5(13|31)$ \\
2478: & & & & & $+ (22|22) + 10(22|33) -5(23|32) + 2(33|33) + 8(33|44)$\\
2479: & & & & & $ -4(34|43)$ \\
2480: \hline
2481:
2482: Ne & $0$ & $0$ & $1$ &
2483: $|1\overline{1}2\overline{2}3\overline{3}4\overline{4}5\overline{5}\rangle$
2484: & $(11|11) + 4(11|22) -2(12|21)+12(11|33)-6(13|31)$\\
2485: & & & & & $+(22|22)+12(22|33)-6(23|32)+3(33|33) + 12(33|44)$ \\
2486: & & & & & $-6(34|43)$ \\
2487: \hline
2488:
2489: \end{tabular}
2490: } % end scalebox
2491:
2492: \end{center}
2493:
2494: \caption[$V_{ee}$ matrix element expressions for N-Ne]
2495: {$V_{ee}$ matrix element expressions for the
2496: N-Ne sequences, `cross' denotes the
2497: off-diagonal term in the $2\times2$ matrix. See Lemma \ref{L:7.3} for notation.}
2498: \label{Tab:VeeMatrix3}
2499: \end{table}
2500:
2501: %\newpage\noindent
2502:
2503: \subsection{Explicit interaction matrix}
2504: %
2505: In order to obtain explicit values, we
2506: finally need to substitute the explicit PT orbitals (\ref{PTorbitals}), (\ref{porbitals}), and
2507: evaluate the ensuing Coulomb and exchange integrals.
2508: We do this via a four-step procedure: reduce the original integrals over $\R^6$ to
2509: integrals over $\R^3$ via Fourier transform calculus; explicitly determine the Fourier transforms
2510: of pointwise products of the above orbitals; reduce to 1D integrals with the help of spherical
2511: polar coordinates in Fourier space; evaluate the remaining 1D integrals, which turn out to have
2512: rational integrands.
2513: %
2514: %
2515: %
2516:
2517: The Fourier transform of a function $f\in L^1(\R^n)$ will be
2518: denoted $\hat{f}$; we find it convenient to use the definition
2519: \be \label{Fourier}
2520: \hat{f}(k):= \int_{\R^n} f(x) e^{-ik\cdot x} dx
2521: \ee
2522: which does not contain any normalization constants.
2523:
2524: \begin{lemma}\label{L:CoulombIntegrals}
2525: For one-electron orbitals $\psi_\alpha \in L^2(\R^3)\cap
2526: L^\infty(\R^3)$, with $\hat\psi_\alpha \in L^2(\R^3) \cap
2527: L^\infty(\R^3)$ and $\alpha \in \{i,j,k,\ell\}$, let
2528: $f(x):=\psi_i(x)\psi_j^*(x)$ and $g(x):=\psi_k^*(x)\psi_\ell(x)$.
2529: Then
2530: \begin{align}
2531: (\psi_i \psi_j | \psi_k \psi_\ell)
2532: &= \int_{\R^6} dx_1 dx_2 \psi_i^*(x_1)\psi_j(x_1)
2533: \frac{1}{|x_1-x_2|} \psi_k^*(x_2)\psi_\ell(x_2) \notag \\
2534: &= \frac{1}{2\pi^2}\int_{\R^3}dk \frac{1}{|k|^2}
2535: (\widehat{f})^*(k)\widehat{g}(k). \label{ijklFT}
2536: \end{align}
2537: \end{lemma}
2538: %
2539: %
2540: %
2541: \noindent
2542: Note that this shows that exchange integrals $(\psi_i\psi_j|\psi_j\psi_i)$ are positive.
2543: \\[2mm]
2544: \noindent\textbf{Proof} Since the $\psi_\alpha$ are in
2545: $L^2(\R^3)\cap L^\infty(\R^3)$, their products $f$ and $g$ are in
2546: $L^1(\R^3)\cap L^\infty(\R^3)$ and hence their Fourier transforms
2547: are well defined. Considering the integral
2548: $I(\lambda)=\int_{\R^3}\int_{\R^3} dx dy
2549: \frac{e^{-\lambda|x-y|}}{|x-y|} f^*(x)g(y)$, $\lambda>0$. It is
2550: easy to show that $\widehat{\frac{e^{-\lambda|x|}}{|x|}}
2551: =\frac{4\pi}{\lambda^2+|k|^2}$. Since $f,g \in L^1(\R^3)\cap
2552: L^\infty(\R^3)$ it follows that $|\frac{e^{-\lambda|x-y|}}{|x-y|}
2553: f^*(x)g(y)| \in L^1(\R^6)$ and so by dominated convergence
2554: \be
2555: I(\lambda) \to \int_{\R^3}\int_{\R^3} dx dy \frac{1}{|x-y|}
2556: f^*(x)g(y) \quad (\lambda \to 0). \label{DCT1}
2557: \ee
2558:
2559: Setting $h=e^{-\lambda|x|}/|x|$ we have $I(\lambda)= \int_{\R^3}dy
2560: \big(f \ast h\big)(y)g(y)dy$ and $\widehat{f\ast h}=\hat{f}\hat{h}
2561: \in L^1(\R^3)$ since $\hat{f}\in L^1(\R^3)$ and $\hat{h}\in
2562: L^\infty(\R^3)$. By Plancherel's theorem we have
2563: $I(\lambda)=\frac{1}{(2\pi)^3}\int_{\R^3} dk \frac{4\pi}{\lambda^2 +
2564: |k|^2}(\hat{f})^*(k)\hat{g}(k)$, and again by dominated convergence and using $f,g \in
2565: L^1(\R^3) \cap L^\infty(\R^3)$,
2566: \be
2567: I(\lambda) \to \frac{1}{2\pi^2}\int_{\R^3}dk \frac{1}{|k|^2}
2568: (\widehat{f})^*(k)\widehat{g}(k) \quad (\lambda \to 0).
2569: \label{DCT2}
2570: \ee
2571: Combining (\ref{DCT1}) and (\ref{DCT2}) gives the result.
2572: \qed \\
2573:
2574: Next we calculate the Fourier transforms of pointwise products of
2575: the hydrogen orbitals (\ref{PTorbitals}), (\ref{porbitals}). Here and below, by expressions such as
2576: $\widehat{|x|e^{-\lambda |x|}}(k)$ we mean the Fourier transform
2577: $\widehat{f}(k)$ of the function $f(x)=|x|e^{-\lambda|x|}$.
2578: %
2579: %
2580: %
2581: \begin{lemma} \label{L:FTs}
2582: With the Fourier transform as defined in (\ref{Fourier}), and $\lambda>0$,
2583: \begin{center}
2584: \begin{tabular}{|c|c|}
2585: \hline Function & Fourier transform \\
2586: \hline
2587:
2588: $e^{-\lambda |x|}$ & $\frac{8 \lambda \pi}{(\lambda^2 +|k|^2)^2}$ \\
2589:
2590: $|x|e^{-\lambda |x|}$ & $\frac{32 \lambda^2 \pi}{(\lambda^2 + |k|^2)^3}
2591: -\frac{8\pi}{(\lambda^2 + |k|^2)^2}$ \\
2592:
2593: $|x|^2e^{-\lambda |x|}$ & $\frac{192\lambda^3 \pi}{(\lambda^2+|k|^2)^4} - \frac{96\lambda
2594: \pi}{(\lambda^2+|k|^2)^3}$ \\
2595:
2596: $x_j e^{-\lambda |x|}$ & $-\frac{32 i \lambda \pi
2597: k_j}{(\lambda^2+|k|^2)^3}$\\
2598:
2599: $x_j^2 e^{-\lambda |x|}$ & $\frac{32\lambda \pi}{(\lambda^2+|k|^2)^3}
2600: - \frac{192 \lambda \pi k_j^2}{(\lambda^2+|k|^2)^4}$\\
2601:
2602: $x_{\ell}x_j e^{-\lambda |x|}$ \scriptsize{$(j \neq \ell)$} & $ - \frac{192 \lambda \pi
2603: k_j k_{\ell}}{(\lambda^2+|k|^2)^4}$\\
2604:
2605: $|x|x_je^{-\lambda |x|}$ & $\frac{32 i \pi k_j}{(\lambda^2+|k|^2)^3} - \frac{192 i
2606: \lambda^2 \pi k_j}{(\lambda^2+|k|^2)^4}$\\
2607: \hline
2608: \end{tabular}
2609: \end{center}
2610:
2611: \end{lemma}
2612: \noindent \textbf{Proof} Let $f(x):=e^{-\lambda |x|}$. We have
2613: $\widehat{f}(k)=8\pi\lambda/(\lambda^2+|k|^2)^2$,
2614: which is easy to verify by
2615: direct calculation (convert to spherical polar coordinates and
2616: integrate). All remaining Fourier transforms can be expressed in terms of derivatives of $\widehat{f}$, as follows.
2617: Using $|x|e^{-\lambda |x|}=-\frac{d}{d\lambda}e^{-\lambda |x|}$, $|x|^2e^{-\lambda |x|}=\frac{d^2}{d\lambda^2}e^{-\lambda |x|}$,
2618: and noting that differentiation with respect to $\lambda$ commutes
2619: with the Fourier transform gives
2620: $$
2621: \widehat{|\cdot|e^{-\lambda |\, \cdot \, |}}(k)=-\frac{\partial}{\partial \lambda}\widehat{f}(k), \;\;
2622: \widehat{|\cdot|^2e^{-\lambda |\cdot|}}(k) =\frac{d^2}{d\lambda^2}\widehat{f}(k).
2623: $$
2624: For the next three Fourier transforms, we recall the well known
2625: differentiation identities for Fourier transforms:
2626: $$
2627: \widehat{x_j f}(k) =i \frac{\partial}{\partial k_j}\widehat{f}(k)
2628: \text{ and } \widehat{x_{\ell}x_j f}(k) =-
2629: \frac{\partial^2}{\partial k_{\ell} \partial k_j}\widehat{f}(k).
2630: $$
2631: Consequently
2632: $$
2633: \widehat{(x_j e^{-\lambda|x|})}(k) \!=\! i\frac{\partial}{\partial k_j}\widehat{f}(k), \;
2634: \widehat{x_j x_\ell e^{-\lambda|x|}}(k) \!=\! - \frac{\partial^2}{\partial k_j \partial k_\ell}\widehat{f}(k).
2635: $$
2636: The final case needed is
2637: $$
2638: \widehat{(|x|x_j e^{-\lambda|x|})}(k) = - \frac{d}{d\lambda} \Bigl(i\frac{\partial}{\partial k_j}\widehat{f}(k)\Bigr).
2639: $$
2640: Working out the above derivatives of $\widehat{f}$ explicitly is straightforward, yielding the formulae given in the lemma. \qed
2641: %
2642: %
2643: %
2644: %
2645: \begin{lemma} \label{L:OrbitalFTs}
2646: The Fourier transforms of pointwise products of the hydrogen orbitals
2647: (\ref{PTorbitals}), (\ref{porbitals}) are as given in the following table. In
2648: all cases $j, \ell =1,2,3$, $j \neq \ell$.
2649: \begin{center}
2650: \begin{tabular}{|c|c|}
2651: \hline Function & Fourier Transform \\
2652: \hline $\phi_{1s}\phi_{1s}$ & $\frac{16Z^4}{(4Z^2+|k|^2)^2}$ \\
2653: $\phi_{2s}\phi_{2s}$ & $\frac{2Z^4}{(Z^2+|k|^2)^2}-
2654: \frac{7Z^6}{(Z^2+|k|^2)^3} +
2655: \frac{6Z^8}{(Z^2+|k|^2)^4}$\\
2656: $\phi_{1s}\phi_{2s}$ & $\frac{4\sqrt{2}Z^4}{((\frac{3}{2}Z)^2+|k|^2)^2}
2657: -\frac{9\sqrt{2}Z^6}{((\frac{3}{2}Z)^2+|k|^2)^3}$ \\
2658: $\phi_{2p_j}\phi_{2p_j}$ &
2659: $\frac{Z^6}{(Z^2+|k|^2)^3}-\frac{6Z^6k_j^2}{(Z^2+|k|^2)^4}$ \\
2660: $\phi_{1s}\phi_{2p_j}$& $-\frac{6\sqrt{2}iZ^5 k_j}{((\frac{3}{2}Z)^2 +|k|^2)^3}$
2661: \\
2662: $\phi_{2s}\phi_{2p_j}$& $\frac{6Z^7 i
2663: k_j}{(Z^2+|k|^2)^4}-\frac{3Z^5ik_j}{(Z^2+|k|^2)^3}$ \\
2664: $\phi_{2p_j}\phi_{2p_\ell}$& $-\frac{6k_jk_{\ell}Z^6}{(Z^2+|k|^2)^4}$ \\
2665: \hline
2666: \end{tabular}
2667: \end{center}
2668: \end{lemma}
2669: \noindent \textbf{Proof} This is simply an application of the
2670: results of Lemma \ref{L:FTs}. \qed
2671: \\[2mm]
2672: Finally we use these Fourier transforms, along with the
2673: reformulation of the Coulomb and exchange integrals from Lemma
2674: \ref{L:CoulombIntegrals}, to derive the explicit values of these
2675: integrals.
2676:
2677: \begin{lemma}\label{L:VeeElements}
2678: Using the abbreviated notation
2679: $1=\phi_{1s}$, $2=\phi_{2s}$, $3=\phi_{2p_3}$, $4=\phi_{2p_1}$, $5=\phi_{2p_1}$, the Coulomb and
2680: exchange integrals (\ref{ijklIntegral}) occuring in Tables \ref{Tab:VeeMatrix1}--\ref{Tab:VeeMatrix3}
2681: with the PT orbitals (\ref{PTorbitals}), (\ref{porbitals}) are given by
2682: \vspace*{-3mm}
2683: \begin{center}
2684: \resizebox{\textwidth}{!}{
2685: \begin{tabular}{|c|c|c|c|c|c|c|c|c|c|c|}
2686: \hline
2687: $(11|11)$ & $(11|22)$ & $(12|21)$ & $(22|22)$
2688: & $(11|33)$ & $(13|31)$ & $(22|33)$ & $(23|32)$
2689: & $(33|33)$ & $(33|44)$ & $(34|43)$ \\
2690: \hline
2691: $\frac{5}{8}Z$ & $\frac{17}{81}Z$ & $\frac{16}{729}Z$ & $\frac{77}{512}Z$
2692: & $\frac{59}{243}Z$ & $\frac{112}{6561}Z$ & $\frac{83}{512}Z$ & $\frac{15}{512}Z$
2693: & $\frac{501}{2560} Z$ & $\frac{447}{2560} Z$ & $\frac{27}{2560} Z$ \\
2694: \hline
2695: \end{tabular}
2696: } %end resizebox
2697: \end{center}
2698: \end{lemma}
2699: \noindent \textbf{Proof} We insert the Fourier transforms from Lemma
2700: \ref{L:OrbitalFTs} into (\ref{ijklFT}), change to spherical polar coordinates and
2701: integrate. The angular integrals are elementary and the final radial
2702: integrals, which on account of Lemmas \ref{L:OrbitalFTs} and \ref{L:CoulombIntegrals} have
2703: rational integrands, may be evaluated with Maple; for a truly pen and paper
2704: method, one can use complex contour integration. \qed
2705:
2706: Note the insteresting multiscale effect that the exchange integrals are much smaller then the Coulomb integrals,
2707: by about one order of magnitude. Nevertheless, as we will see later, the exchange terms play an important role
2708: in energy level splitting.
2709:
2710: The table in the above lemma together with Tables \ref{Tab:VeeMatrix1}--\ref{Tab:VeeMatrix3} completes the task of
2711: evaluating the matrix $PV_{ee}P$ on $V_0(N)$.
2712: %
2713: %
2714: %
2715: %
2716: %
2717:
2718: \subsection{The matrix $PHP$} \label{Sec:ExplicitPHP}
2719:
2720: The remaining part $PH_0P$ of the Hamiltonian $PHP$ is trivial
2721: to determine, because the space $V_0(N)$ is an eigenspace
2722: of $H_0$, with eigenvalue given in Lemma \ref{L:meEfns}:
2723: \be \label{H0PT}
2724: H_0 = PH_0P = Z^2\Bigl(-1-\mbox{$\frac{N-2}{8}$}\Bigr)I \;\;\mbox{on }V_0(N).
2725: \ee
2726: %
2727: %
2728: %
2729: By inspection of (\ref{H0PT}), Lemma \ref{L:VeeElements}, and Tables \ref{Tab:VeeMatrix1}--\ref{Tab:VeeMatrix3}
2730: we obtain an interesting corollary.
2731: %
2732: %
2733: %
2734: \begin{corollary} The matrix of the PT Hamiltonian $PHP$ with respect to the basis in Tables
2735: \ref{Tab:VeeMatrix1}--\ref{Tab:VeeMatrix3}
2736: with the orbitals (\ref{PTorbitals}), (\ref{porbitals}) is a rational matrix.
2737: \end{corollary}
2738: %
2739: %
2740: %
2741: %
2742: \section{Atomic energy levels and eigenstates} \label{Sec:AtSpec}
2743: The spectral decomposition of the PT Hamiltonians $PHP$ is almost immediate from the block form
2744: derived in the previous section , the only remaining task being the
2745: diagonalization of the $2\times 2$ blocks, which may be done explicitly:
2746: For (orthonormal) wavefunctions $\Psi_1$, $\Psi_2$
2747: and $E_i:=\langle \Psi_i | H | \Psi_i\rangle$, the eigenvalues are given by
2748: \be \label{degevals}
2749: \lambda_{\pm} = \frac{E_1+E_2}{2} \pm
2750: \sqrt{\big(\tfrac{E_1-E_2}{2}\big)^2 +
2751: |\langle \Psi_1 |V_{ee} |\Psi_2\rangle|^2}
2752: \ee
2753: with corresponding normalized eigenstates
2754: \be \label{degestates}
2755: \Psi_\pm = \frac{1}{\sqrt{1+c_\pm^2}}\Bigl( \Psi_1 + c_\pm \Psi_2 \Bigr), \;\;\;
2756: c_\pm = \frac{ \frac{E_2-E_1}{2} \pm \sqrt{\big(\tfrac{E_1-E_2}{2}\big)^2 +
2757: |\langle \Psi_1 |V_{ee} |\Psi_2\rangle|^2 } }{\langle \Psi_1 |V_{ee} |\Psi_2\rangle}.
2758: \ee
2759: These formulae together with eq. (\ref{H0PT}), Tables \ref{Tab:VeeMatrix1}--\ref{Tab:VeeMatrix3} and
2760: Lemma \ref{L:VeeElements} immediately yield:
2761:
2762: \begin{theorem} \label{mainresult} For $N=3,...,10$, and $Z>0$, the
2763: energy levels of the PT Model (\ref{PT}), (\ref{H0H'}) are as given in
2764: Tables \ref{Tab:PTEigenspaces1}, \ref{Tab:PTEigenspaces2}. Each eigenspace has
2765: the minimal dimension $(2L+1)(2S+1)$ possible for its spin and angular momentum quantum numbers $L$ and $S$ (see Lemma \ref{symlem}),
2766: and the up to normalization unique corresponding
2767: eigenstate with zero $L_3$ and maximal $S_3$ is as shown in the Tables. Moreover the levels and eigenstates in the Tables provide the
2768: leading order asymptotic terms of the true Schr\"odinger levels as $Z\to\infty$, in the sense described in Theorem \ref{T:isolimit}.
2769: \end{theorem}
2770:
2771: \noindent
2772: Note that the ordering of the PT levels is independent of $Z$, since the spectral gaps are linear in $Z$.
2773:
2774: \begin{table}[htbp]
2775: \hspace{5mm}
2776: \rotatebox{90}{ %
2777: \resizebox{0.9\textheight}{!}{
2778: % \begin{center}
2779: \begin{tabular}{|c|c|c|l|c|r|r|}
2780: \hline & Symm. & $\Psi$ &
2781: \multicolumn{1}{|c|}{$E$} &
2782: \multicolumn{1}{|c|}{$c$} & \multicolumn{1}{|c|}{$E$ (num.)} &
2783: \multicolumn{1}{|c|}{$c$ (num.)} \\
2784: \hline
2785: %
2786: Li & $^2S$ & $\Psi_1$ & $-\tfrac{9}{8}Z^2+\tfrac{5965}{5832}Z$ & &-7.0566 & \\
2787: & $^2P^\circ$ & $\Psi_2$ & $-\tfrac{9}{8}Z^2+\tfrac{57397}{52488}Z$ & &-6.8444 &\\ \hline
2788: %
2789: Be & $^1S$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_1 + c \Psi_2)$
2790: & $-\frac{5}{4}Z^2 +
2791: \frac{1}{1679616}(2813231-5\sqrt{1509308377})Z$
2792: & $-\frac{1}{59049}(2\sqrt{1509308377}-6981)\sqrt{3}$
2793: &-13.7629 & -0.2311\\
2794: & $^3P^\circ$ & $\Psi_4$ & $-\frac{5}{4}Z^2 + \frac{1363969}{839808}Z$ & &-13.5034 &\\
2795: & $^1P^\circ$ & $\Psi_3$ & $-\frac{5}{4}Z^2 + \frac{2826353}{1679616}Z$& &-13.2690 &\\
2796: & $^3P$ & $\Psi_5$ & $-\frac{5}{4}Z^2 + \frac{1449605}{839808}Z$ & &-13.0955 &\\
2797: & $^1D$ & $\Psi_6$ & $-\frac{5}{4}Z^2 + \frac{14673197}{8398080}Z$ & &-13.0112 &\\
2798: & $^1S$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_1 + c \Psi_2)$
2799: & $-\frac{5}{4}Z^2 +
2800: \frac{1}{1679616}(2813231+5\sqrt{1509308377})Z$
2801: & $\frac{1}{59049}(2\sqrt{1509308377}+6981)\sqrt{3}$
2802: &-12.8377 & 4.3271\\
2803: \hline
2804: %
2805: B & $^2P^\circ$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_3 + c \Psi_4)$
2806: & $-\frac{11}{8}Z^2 +
2807: \frac{1}{6718464}(16493659-\sqrt{733174301809})Z$
2808: & $-\frac{1}{393660}(\sqrt{733174301809}-809747)\sqrt{2}$
2809: &-22.7374 & -0.1671\\
2810: & $^4P$ & $\Psi_6$ & $-\frac{11}{8}Z^2 + \frac{2006759}{839808}Z$ & &-22.4273 &\\
2811: & $^2D$ & $\Psi_7$ & $-\frac{11}{8}Z^2 + \frac{40981549}{16796160}Z$& &-22.1753 &\\
2812: & $^2S$ & $\Psi_1$ & $-\frac{11}{8}Z^2+ \frac{4151299}{1679616}Z$& &-22.0171 &\\
2813: & $^2P$ & $\Psi_5$ & $-\frac{11}{8}Z^2 + \frac{8322281}{3359232}Z$& &-21.9878 &\\
2814: & $^4S^\circ$ & $\Psi_2$ & $-\frac{11}{8}Z^2 + \frac{706213}{279936}Z$ & &-21.7612 &\\
2815: & $^2D^\circ$ & $\Psi_8$ & $-\frac{11}{8}Z^2 + \frac{14301407}{5598720}Z$ & &-21.6030 &\\
2816: & $^2P^\circ$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_3 + c \Psi_4)$
2817: & $-\frac{11}{8}Z^2 +
2818: \frac{1}{6718464}(16493659+\sqrt{733174301809})Z$
2819: & $\frac{1}{393660}(\sqrt{733174301809}+809747)\sqrt{2}$
2820: &-21.4629 & 5.9851\\
2821: \hline
2822: %
2823: C & $^3P$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_6+c\Psi_7)$
2824: & $-\frac{3}{2}Z^2 + \big( \frac{3806107}{1119744}
2825: - \frac{1}{3359232}\sqrt{221876564389} \big)Z$
2826: & $-\frac{1}{98415}(\sqrt{221876564389}-460642)$
2827: &-34.4468 & -0.1056\\
2828: & $^1D$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_9+c\Psi_{10})$
2829: & $-\frac{3}{2}Z^2 + \big( \frac{19148633}{5598720}
2830: -\frac{1}{3359232}\sqrt{221876564389} \big)Z$
2831: & $\frac{1}{98415}(\sqrt{221876564389}-460642)$
2832: &-34.3202 & 0.1056\\
2833: & $^5S^\circ$ & $\Psi_4$ & $-\frac{3}{2}Z^2 +
2834: \frac{464555}{139968}Z $
2835: & &-34.0859 &\\
2836: & $^1S$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_1+c\Psi_2)$
2837: & $-\frac{3}{2}Z^2 + \big( \frac{966289}{279936}
2838: - \frac{1}{1679616}\sqrt{62733275266}\big)Z$
2839: & $-\frac{1}{98415}(\sqrt{62733275266}-230321)$
2840: &-34.1838 & -0.2047\\
2841: & $^3D^\circ$ & $\Psi_{12}$ &
2842: $-\frac{3}{2}Z^2 + \frac{4730843}{1399680}Z$ & &-33.7203 &\\
2843: & $^3P^\circ$ & $\Psi_8$ &
2844: $-\frac{3}{2}Z^2 + \frac{1904147}{559872}Z$ & &-33.5938 &\\
2845: & $^1D^\circ$ & $\Psi_{11}$ &
2846: $-\frac{3}{2}Z^2 + \frac{9625711}{2799360}Z$ & &-33.3688 &\\
2847: & $^3S^\circ$ & $\Psi_3$ &
2848: $-\frac{3}{2}Z^2 + \frac{961915}{279936}Z$ & &-33.3828 &\\
2849: & $^1P^\circ$ & $\Psi_5$
2850: & $-\frac{3}{2}Z^2 + \frac{242119}{69984}Z$ & &-33.2422 &\\
2851: & $^3P$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_6+c\Psi_7)$
2852: & $-\frac{3}{2}Z^2 + \big( \frac{3806107}{1119744}
2853: + \frac{1}{3359232}\sqrt{221876564389} \big)Z$
2854: & $\frac{1}{98415}(\sqrt{221876564389}+460642)$
2855: &-32.7641 & 9.4668\\
2856: & $^1D$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_9+c\Psi_{10})$
2857: & $-\frac{3}{2}Z^2 + \big( \frac{19148633}{5598720}
2858: +\frac{1}{3359232}\sqrt{221876564389} \big)Z$
2859: & $\frac{1}{98415}(-\sqrt{221876564389}-460642)$
2860: &-32.6376 & -9.4668\\
2861: & $^1S$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_1+c\Psi_2)$
2862: & $-\frac{3}{2}Z^2 + \big( \frac{966289}{279936}
2863: + \frac{1}{1679616}\sqrt{62733275266}\big)Z$
2864: & $\frac{1}{98415}(\sqrt{62733275266}+230321)$
2865: &-32.3943 & 4.8853\\
2866: \hline
2867: \end{tabular}
2868: % \end{center}
2869: } % end resizebox
2870: } % end rotatebox
2871: \caption{Asymptotic Schr\"odinger energy levels and eigenstates (= exact PT levels and states) for the Li-C sequences. The
2872: numerical values are for $Z\!=\!N$. $\Psi$ is the up to
2873: normalization unique PT eigenstate with $L_3\!=\!0$ and maximal $S_3$. The $\Psi_i$
2874: are as in Tables \ref{Tab:VeeMatrix1}-\ref{Tab:VeeMatrix2}, labelled by order of appearance, with
2875: $1,2,3,4,5,$ $\overline{1},\overline{2},\overline{3},\overline{4},\overline{5}$ as in Theorem \ref{GStheorem}.}
2876: \label{Tab:PTEigenspaces1}
2877: \end{table}
2878:
2879: \begin{table}[htbp]
2880: \hspace{25mm}
2881: \rotatebox{90}{ %
2882: \resizebox{0.9\textheight}{!}{
2883: % \begin{center}
2884: \begin{tabular}{|c|c|c|l|c|r|r|}
2885: \hline & Symm. & $\Psi$ &
2886: \multicolumn{1}{|c|}{$E$} &
2887: \multicolumn{1}{|c|}{$c$} & \multicolumn{1}{|c|}{$E$ (num.)} &
2888: \multicolumn{1}{|c|}{$c$ (num.)} \\
2889: \hline
2890: %
2891: N & $^4S^\circ$ & $\Psi_2$ & $-\frac{13}{8}Z^2 + \frac{2437421}{559872}Z$ & & -49.1503 & \\
2892: & $^2D^\circ$ & $\Psi_7$ & $-\frac{13}{8}Z^2 + \frac{24551357}{5598720}Z$ & &-48.9288 & \\
2893: & $^2P^\circ$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_4+c\Psi_3)$
2894: & $-\frac{13}{8}Z^2 + \frac{1}{6718464}(30636167 -
2895: \sqrt{1144203315841})Z$
2896: & $-\frac{1}{393660}(\sqrt{1144203315841}-1032821)\sqrt{2}$
2897: &-48.8195 & -0.1324 \\
2898: & $^4P$ & $\Psi_6$ & $-\frac{13}{8}Z^2 + \frac{7549145}{1679616}Z$ & &-48.1630 & \\
2899: & $^2D$ & $\Psi_8$ & $-\frac{13}{8}Z^2 + \frac{76337819}{16796160}Z$& &-47.8103 & \\
2900: & $^2S$ & $\Psi_1$ & $-\frac{13}{8}Z^2 + \frac{3843463}{839808}Z$ & &-47.5888 &\\
2901: & $^2P$ & $\Psi_5$ & $-\frac{13}{8}Z^2 + \frac{15393535}{3359232}Z$& &-47.5478 &\\
2902: & $^2P^\circ$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_4+c\Psi_3)$
2903: & $-\frac{13}{8}Z^2 + \frac{1}{6718464}(30636167 +
2904: \sqrt{1144203315841})Z$
2905: & $\frac{1}{393660}(\sqrt{1144203315841}+1032821)\sqrt{2}$
2906: &-46.5905 & 7.5532\\
2907: \hline
2908: %
2909: O & $^3P$ & $\Psi_5$ & $-\frac{7}{4}Z^2 + \frac{4754911}{839808}Z$ & &-66.7048 &\\
2910: & $^1D$ & $\Psi_6$ & $-\frac{7}{4}Z^2 + \frac{47726257}{8398080}Z$ & &-66.5360 &\\
2911: & $^1S$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_1+c\Psi_2)$
2912: & $-\frac{7}{4}Z^2 +
2913: \frac{1}{1679616}(9884485-\sqrt{89111336179})Z$
2914: & $-\frac{1}{295245}(2\sqrt{89111336179}-572179)\sqrt{3}$
2915: &-66.3421 & -0.1458\\
2916: & $^3P^\circ$ & $\Psi_4$ & $-\frac{7}{4}Z^2 + \frac{1224899}{209952}Z$ & &-65.3265 &\\
2917: & $^1P^\circ$ & $\Psi_3$ & $-\frac{7}{4}Z^2 + \frac{9897607}{1679616}Z$ & &-64.8578 &\\
2918: & $^1S$ & $\frac{1}{\sqrt{1+c^2}}(\Psi_1+c\Psi_2)$
2919: & $-\frac{7}{4}Z^2 +
2920: \frac{1}{1679616}(9884485+\sqrt{89111336179})Z$
2921: & $\frac{1}{295245}(2\sqrt{89111336179}+572179)\sqrt{3}$
2922: &-63.4984 & 6.8592\\
2923: \hline
2924: %
2925: F & $^2P^\circ$ & $\Psi_2$ & $-\frac{15}{8}Z^2 + \frac{11982943}{1679616}Z$ & &-87.6660 &\\
2926: & $^2S$ & $\Psi_1$ & $-\frac{15}{8}Z^2 + \frac{4108267}{559872}Z$ & &-85.8342 &\\
2927: \hline
2928: %
2929: Ne & $^1S$ & $\Psi_1$ & $-2Z^2 + \frac{2455271}{279936}Z$ & &-112.2917 &\\
2930: \hline
2931: %
2932: \end{tabular}
2933: % \end{center}
2934: } % end resizebox
2935: } % end rotatebox
2936: \caption{Asymptotic Schr\"odinger energy levels and eigenstates (= exact PT levels and states) for the N-Ne sequences.
2937: The numerical values are for $Z=N$. $\Psi$ is the up to
2938: normalization unique PT eigenstate with $L_3=0$ and maximal $S_3$.
2939: The $\Psi_i$ are as in Table \ref{Tab:VeeMatrix3}, labelled by order of appearance, with
2940: $1,2,3,4,5,\overline{1},\overline{2},\overline{3},\overline{4},\overline{5}$ as in Theorem \ref{GStheorem}.}
2941: \label{Tab:PTEigenspaces2}
2942: \end{table}
2943:
2944: \section{Comparison with experiment and methods in the physics and chemistry literature}\label{Sec:Comparison}
2945:
2946: The analytical eigenvalues and eigenstates derived in the isoelectronic limit
2947: provide a wealth of information on various quantitites of physical and chemical interest,
2948: and yield a number of insights into the inner working mechanisms of the many-electron Schr\"odinger equation which
2949: are not readily available from numerical simulations.
2950:
2951: We will discuss, in turn, the obtained $L$ and $S$ values, ground state dimensions, ground state energies,
2952: and spectral orderings.
2953:
2954: \subsection{$L$ and $S$ values and the notion of `group' in the periodic table}
2955: The ground states themselves are not accessible from experiment, but
2956: their spin and angular momentum quantum numbers are. As already mentioned, the theoretical values agree with the experimental
2957: values in every case (see Table \ref{Tab:H0Dim}), not just for large $Z$ but all the way down to neutral atoms
2958: ($Z=N$), capturing the nontrivial dependence on the number of electrons.
2959:
2960: An important theoretical feature of $L$ and $S$ values as compared to the more familiar semi-empirical concept
2961: of ``hydrogen orbital configurations'' is that regardless of the approximations made to predict them in practice,
2962: they remain well defined in the full Schr\"odinger equation. See Section \ref{Sec:Basic}.
2963: It would therefore be of value to base quantum mechanical explanations of the periodic table on numbers
2964: such as these. In this context we note that $L$ and $S$ values suffice to explain quantum mechanically
2965: a large part of the notion of ``group'' in the periodic table.
2966: Only five different $(L,S)$ pairs occur mathematically for the first 10 atoms, and experimentally for the first 20. Now
2967: these correspond precisely to group 1 (alkali metals),
2968: the union of groups 2 and 8 (alkaline earth metals and noble gases),
2969: the union of groups 3 and 7 (group 3 metals and halogens), the union of groups 4 and 6 (Carbon group and Oxygen group),
2970: and group 5 (Nitrogen group). See the table below.
2971:
2972: \begin{figure}[ht] \label{F:periodictable}
2973: \begin{center}
2974: \resizebox{11cm}{!}{
2975: \includegraphics{PeriodicBW.eps}
2976: }% end scalebox
2977: \end{center}
2978: \end{figure}
2979: \noindent
2980: Moreover, taking into account the gradients of $L$ or $S$ with respect to
2981: atomic number $N$ would separate the group 3 metals from the halogens, and the Carbon group from the Oxygen group.
2982: Note that $L$ and $S$ gradients are mathematically analogous to ionization energies, which are gradients of energy with respect to $N$.
2983:
2984: %
2985: %
2986: %
2987: \subsection{Ground state dimensions} These dimensions are shown in Table \ref{Tab:H0Dim}. They are interesting
2988: as they are a measure of the `flexibility' within the ground state, in that they specify the number of degrees of freedom which can be varied without
2989: affecting the energy of the state. This flexibility appears to be curiously unexplored in the literature, perhaps in part due to it not being
2990: clearly captured by the semi-empirical Bohr-Slater picture, the Hartree-Fock approximation, or Kohn-Sham density
2991: functional theory.
2992:
2993: On a qualitative level, we expect that an atom with a high-dimensional ground state will form a wider range of molecules than an atom
2994: with a similar number of valence electrons but with a lower dimensional ground state. This should be true both in terms of
2995: molecular geometry (e.g. linear, bent, triangular, tetrahedral) as well as in terms of which atoms it will stably bond with. We plan
2996: to develop this idea in a more mathematical way in a future publication.
2997: %
2998: %
2999: %
3000: \subsection{Ground state energies} \label{Sec:GSE}
3001: %
3002: %
3003: The asymptotic ground state energies, despite being theoretically justified only for strongly positive ions
3004: (see Section \ref{Sec:Pert}), still capture around $90\%$ of
3005: the experimental \cite{NIST} energies of neutral atoms. See the following table.
3006: \\[1mm]
3007: %\begin{center}
3008: \resizebox{\textwidth}{!} %height
3009: {
3010: \begin{tabular}{|l|c|c|c|c|c|c|c|c|}
3011: \hline
3012: Atom & Li & Be & B & C & N & O & F & Ne \\
3013: \hline
3014: $E_{PT}$ & -7.0566 & -13.7629 & -22.7374 & -34.4468 & -49.1503 & -66.7048 & -87.6660 & -112.2917 \\
3015: $E_{exp}$ & -7.4779 & -14.6684 & -24.6581 & -37.8558 & -54.6117 & -75.1080 & -99.8060 & -129.0500 \\
3016: Error & 5.6\% & 6.2\% & 7.8\% & 9.0\% & 10.0\% & 11.2\% &
3017: 12.2\% & 13.0\% \\
3018: \hline
3019: \end{tabular}
3020: } %end resizebox
3021: %\end{center}
3022: %
3023: %
3024: %
3025:
3026: \subsection{Spectral orderings and Hund's rule} \label{Sec:Hund}
3027: The spectral orderings of the asymptotic levels are in spectacular
3028: agreement with the experimental data \cite{Huheey93,NIST}, even for neutral atoms. For the
3029: purpose of these comparisons we consider only the experimental
3030: states attributed to configurations containing only orbitals with $n
3031: \leq 2$. The results differ only by the interchange of two higher levels
3032: in Beryllium (${}^1D$ and ${}^3P$) and Carbon (${}^1D^o$ and ${}^3S^o$).
3033: \\[2mm]
3034: A key virtue of our exact eigenstates is that they allow to trace the spectral
3035: gaps to the size of individual Coulomb and exchange integrals.
3036:
3037: As an example of a $2s$--$2p$ spectral gap, consider the $^2S$ ground
3038: state and $^2P$ first excited state of Lithium. Table
3039: \ref{Tab:VeeMatrix2} shows that the gap is given by
3040: the difference in interaction of the $2p$ and $2s$ orbitals with the
3041: $1s$ shell, $[2(11|33)-(13|31)] - [2(11|22)-(12|21)]$.
3042:
3043: As an example of energy level splitting between two states with an equal number of
3044: $1s$, $2s$ and $2p$ orbitals, consider the $^4S^o$ ground state and $^2D^o$ first
3045: excited state of Nitrogen. A look at Table \ref{Tab:VeeMatrix3}
3046: reveals that the energy difference consists only of
3047: the exchange term $-3(34|43)$, which is present in the ground state
3048: due to the parallel spins of the three $p$-orbitals, but absent in
3049: the excited state.
3050: \\[2mm]
3051: In a large majority of cases, the theoretical orderings also agree with Hund's rules.
3052: In fact, many of Hund's rules are rigorous theorems in first order perturbation theory and
3053: related models, and rely only on the structure of the symbolic
3054: matrices in Tables \ref{Tab:VeeMatrix1}--\ref{Tab:VeeMatrix3}, not their numerical values. This will be discussed elsewhere.
3055:
3056: Let us also describe a
3057: \\[2mm]
3058: %
3059: %
3060: {\it Counterexample to Hund's rules}. Consider the higher Carbon $1s^22s2p^3$ states. Hund's rules would
3061: order their energies, regardless of the choice of nuclear charge $Z$, as
3062: $$
3063: E_{^5S^o}<E_{^3D^o}<E_{^3P^o}<E_{^3S^o}<E_{^1D^o}<E_{^1P^o}.
3064: $$
3065: For large $Z$ this agrees with the PT and experimental orderings. (That the latter two agree with
3066: each other follows from Theorem \ref{T:isolimit}.)
3067:
3068: But experimentally, at $Z=20$ the
3069: $^1D^o$ singlet and the $^3S^o$ triplet are observed to cross, see \cite{NIST} and Figure \ref{F:Splitting}.
3070: (This crossing is beautifully confirmed by theoretical calculations based on a simple CI model designed with
3071: the help of our asymptotic findings here, see \cite{FG09}.)
3072: In particular, in the neutral atom, $Z=N$, the experimental ordering is
3073: $$
3074: E_{^5S^o}<E_{^3D^o}<E_{^3P^o}<E_{^1D^o}<E_{^3S^o}<E_{^1P^o}.
3075: $$
3076: This is an important example because it shows that it is of
3077: value to investigate which of Hund's rules can be justified quantum
3078: mechanically and which ones can not. In this particular case, a
3079: closer look shows that there should be no universal ordering,
3080: neither one way nor the other. The energy difference as read off
3081: from Table \ref{Tab:VeeMatrix2} consists of a $2s$--$2p$ positive
3082: exchange term and a $2p$--$2p$ negative exchange term
3083: \be \label{truediff}
3084: E_{^3S^o} - E_{^1D^o} = (24|42) - 3(34|43),
3085: \ee
3086: and so could have either sign, depending on the orbitals.
3087: %
3088: %
3089: %
3090:
3091: Note that this interesting effect is missed when the states under investigation are modelled
3092: by their aufbau principle configurations. By Hund's rules, these are
3093: $|1\overline{1}23\overline{4}\overline{5}\rangle$ for the singlet
3094: and $|1\overline{1}234\overline{5}\rangle$ for the triplet. A simple
3095: calculation shows that the energy difference is then
3096: $E_{^3S^o} - E_{^1D^o} = - (24|42) < 0$, which is very far from the correct difference (\ref{truediff})
3097: and incorrectly predicts a universal ordering.
3098:
3099: \subsection{Spectral gaps}
3100: The asymptotic energy levels, despite their excellent orderings,
3101: do not give quantitatively useful spectral gaps for
3102: neutral atoms. However, in the regime of highly positive ions in which they were theoretically
3103: justified (see Theorem \ref{T:isolimit}) they beautifully match the experimental gaps,
3104: as shown in Figure \ref{F:Splitting}.
3105:
3106: \begin{figure}[ht]
3107: \begin{center}
3108: \resizebox{8cm}{!}{
3109: \includegraphics{SpectralGaps6-20.eps}
3110: }% end scalebox
3111: \end{center}
3112: \caption{Splitting of noninteracting Carbon ground state energy by electron interaction. Lines: Asymptotic Schr\"odinger levels (this paper); Circles:
3113: experimental data \cite{NIST,Moore70}. For the highest level at $Z=6$
3114: and the fourth level at $Z=18$, we were unable to find experimental
3115: data.}
3116: \label{F:Splitting}
3117: \end{figure}
3118:
3119: \noindent
3120: For the Carbon series ($N=6$, $Z=6,7,8,\dots$), we show the experimental spectral gaps
3121: $$
3122: \frac{E_j(N,Z)}{Z^2} - \frac{E_1(N,Z)}{Z^2}
3123: $$
3124: (circles) and the perturbation-theoretic spectral gaps
3125: $$
3126: \frac{E_j^{PT}(N,Z)}{Z^2} - \frac{E_1^{PT}(N,Z)}{Z^2} = \frac{\Etilde_j^{(1)}}{Z} - \frac{\Etilde_1^{(1)}}{Z}
3127: $$
3128: (lines) against $\frac{1}{Z}$, with the energy of the lowest level (which shifts with $Z$) having
3129: been subtracted for clarity.
3130: %
3131: %
3132: %
3133: %
3134: %
3135: By Theorem \ref{T:isolimit}, the match between PT and Schr\"odinger energy levels would become even
3136: better as $Z$ increases further. But beyond the value $Z=20$ shown here, the match
3137: between Schr\"odinger levels and experiment slowly starts to deviate, due to the onset of relativistic effects, whose
3138: study lies beyond the scope of the present paper.
3139:
3140: \subsection{Overall conclusion}
3141: The principal conclusion of this paper is that the semi-empirical hydrogen orbital configurations
3142: of atoms developed by Bohr, Hund and Slater have a precise
3143: mathematical meaning, as asymptotic limits of the true Schr\"odinger ground states for large nuclear charge. (This holds
3144: up to certain small but interesting corrections, as described in Section \ref{Sec:Aufbau}.)
3145: We hope that the limit eigenstates calculated here (see Table \ref{Tab:PTGS})
3146: can serve as a theoretical alternative to semi-empirical discussions of the periodic table
3147: in the literature.
3148:
3149: Another use of our findings, as benchmark data for the design and validation of computational methods,
3150: is explored in a companion paper \cite{FG09}.
3151: %
3152: %
3153: %
3154: %
3155: \\[4mm]
3156: {\bf Acknowledgements} The research of B.G. was supported by a graduate scholarship from EPSRC. We thank
3157: P.Gill for helpful comments, and Ch.Mendl for careful checking of Tables \ref{Tab:PTEigenspaces1},
3158: \ref{Tab:PTEigenspaces2}.
3159:
3160:
3161: %GATHER{PT2.bib}
3162: % allows WinEdt to autocomplete citations
3163:
3164: \bibliographystyle{alpha}
3165: \bibliography{PT3}
3166:
3167: \noindent
3168: Address of authors: \\[2mm]
3169: Gero Friesecke \\
3170: Center for Mathematics, TU Munich, Germany, {\tt gf@ma.tum.de} \\[2mm]
3171: Benjamin D. Goddard \\
3172: Mathematics Institute, University of Warwick, U.K. {\tt b.d.goddard@warwick.ac.uk}
3173:
3174: \end{document}
3175: