Approximate eigenvalue determination of geometrically frustrated magnetic molecules
Geometrically frustrated magnetic molecules have attracted a lot of interest in the field of molecular magnetism as well as frustrated Heisenberg antiferromagnets. In this article we demonstrate how an approximate diagonalization scheme can be used in order to obtain thermodynamic and spectroscopi...
Gespeichert in:
Datum: | 2009 |
---|---|
Hauptverfasser: | , , |
Format: | Artikel |
Sprache: | English |
Veröffentlicht: |
Інститут фізики конденсованих систем НАН України
2009
|
Schriftenreihe: | Condensed Matter Physics |
Online Zugang: | http://dspace.nbuv.gov.ua/handle/123456789/119999 |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Назва журналу: | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
Zitieren: | Approximate eigenvalue determination of geometrically frustrated magnetic molecules / R. Schnalle, A.M. Läuchli, J. Schnack // Condensed Matter Physics. — 2009. — Т. 12, № 3. — С. 331-342. — Бібліогр.: 35 назв. — англ. |
Institution
Digital Library of Periodicals of National Academy of Sciences of Ukraineid |
irk-123456789-119999 |
---|---|
record_format |
dspace |
spelling |
irk-123456789-1199992017-06-11T03:04:07Z Approximate eigenvalue determination of geometrically frustrated magnetic molecules Schnalle, R. Läuchli, A.M. Schnack, J. Geometrically frustrated magnetic molecules have attracted a lot of interest in the field of molecular magnetism as well as frustrated Heisenberg antiferromagnets. In this article we demonstrate how an approximate diagonalization scheme can be used in order to obtain thermodynamic and spectroscopic information about frustrated magnetic molecules. To this end we theoretically investigate an antiferromagnetically coupled spin system with cuboctahedral structure modeled by an isotropic Heisenberg Hamiltonian. Геометрично фрустрованi магнiтнi молекули привертають багато уваги як у дiлянцi молекулярного магнетизму, так i в дiлянцi фрустрованих антиферомагнетикiв Гайзенберга. У цiй статтi ми демонструємо, як наближена дiагоналiзацiйна схема може бути використана для отримання термодинамiчної i спектроскопiчної iнформацiї про фрустрованi магнiтнi молекули. М и теоретично дослiджуємо антиферомагнiтно взаємодiючу спiнову систему з кубоктаедричною структурою, модельовану iзотропним гамiльтонiаном Гайзенберга. 2009 Article Approximate eigenvalue determination of geometrically frustrated magnetic molecules / R. Schnalle, A.M. Läuchli, J. Schnack // Condensed Matter Physics. — 2009. — Т. 12, № 3. — С. 331-342. — Бібліогр.: 35 назв. — англ. 1607-324X PACS: 75.10.Jm, 75.50.Xx, 75.40.Mg, 75.50.Ee DOI:10.5488/CMP.12.3.331 http://dspace.nbuv.gov.ua/handle/123456789/119999 en Condensed Matter Physics Інститут фізики конденсованих систем НАН України |
institution |
Digital Library of Periodicals of National Academy of Sciences of Ukraine |
collection |
DSpace DC |
language |
English |
description |
Geometrically frustrated magnetic molecules have attracted a lot of interest in the field of molecular magnetism
as well as frustrated Heisenberg antiferromagnets. In this article we demonstrate how an approximate
diagonalization scheme can be used in order to obtain thermodynamic and spectroscopic information about
frustrated magnetic molecules. To this end we theoretically investigate an antiferromagnetically coupled spin
system with cuboctahedral structure modeled by an isotropic Heisenberg Hamiltonian. |
format |
Article |
author |
Schnalle, R. Läuchli, A.M. Schnack, J. |
spellingShingle |
Schnalle, R. Läuchli, A.M. Schnack, J. Approximate eigenvalue determination of geometrically frustrated magnetic molecules Condensed Matter Physics |
author_facet |
Schnalle, R. Läuchli, A.M. Schnack, J. |
author_sort |
Schnalle, R. |
title |
Approximate eigenvalue determination of geometrically frustrated magnetic molecules |
title_short |
Approximate eigenvalue determination of geometrically frustrated magnetic molecules |
title_full |
Approximate eigenvalue determination of geometrically frustrated magnetic molecules |
title_fullStr |
Approximate eigenvalue determination of geometrically frustrated magnetic molecules |
title_full_unstemmed |
Approximate eigenvalue determination of geometrically frustrated magnetic molecules |
title_sort |
approximate eigenvalue determination of geometrically frustrated magnetic molecules |
publisher |
Інститут фізики конденсованих систем НАН України |
publishDate |
2009 |
url |
http://dspace.nbuv.gov.ua/handle/123456789/119999 |
citation_txt |
Approximate eigenvalue determination of geometrically frustrated magnetic molecules / R. Schnalle, A.M. Läuchli, J. Schnack // Condensed Matter Physics. — 2009. — Т. 12, № 3. — С. 331-342. — Бібліогр.: 35 назв. — англ. |
series |
Condensed Matter Physics |
work_keys_str_mv |
AT schnaller approximateeigenvaluedeterminationofgeometricallyfrustratedmagneticmolecules AT lauchliam approximateeigenvaluedeterminationofgeometricallyfrustratedmagneticmolecules AT schnackj approximateeigenvaluedeterminationofgeometricallyfrustratedmagneticmolecules |
first_indexed |
2025-07-08T17:03:16Z |
last_indexed |
2025-07-08T17:03:16Z |
_version_ |
1837099066980302848 |
fulltext |
Condensed Matter Physics 2009, Vol. 12, No 3, pp. 331–342
Approximate eigenvalue determination of geometrically
frustrated magnetic molecules
R. Schnalle1∗, A.M. Läuchli2, J. Schnack3
1 Universität Osnabrück, Fachbereich Physik, D–49069 Osnabrück, Germany
2 Max Planck Institut für Physik komplexer Systeme, Nöthnitzerstr. 38, D–01187 Dresden, Germany
3 Universität Bielefeld, Fakultät für Physik, Postfach 100131, D–33501 Bielefeld, Germany
Received May 29, 2009
Geometrically frustrated magnetic molecules have attracted a lot of interest in the field of molecular mag-
netism as well as frustrated Heisenberg antiferromagnets. In this article we demonstrate how an approximate
diagonalization scheme can be used in order to obtain thermodynamic and spectroscopic information about
frustrated magnetic molecules. To this end we theoretically investigate an antiferromagnetically coupled spin
system with cuboctahedral structure modeled by an isotropic Heisenberg Hamiltonian.
Key words: magnetic molecules, Heisenberg model, geometric frustration, irreducible tensor operator
technique, approximate diagonalization, cuboctahedron
PACS: 75.10.Jm, 75.50.Xx, 75.40.Mg, 75.50.Ee
1. Introduction
The complete understanding of small magnetic systems such as magnetic molecules is compulso-
rily connected to the knowledge of their energy spectra. From the energy spectra all spectroscopic,
dynamic, and thermodynamic properties of the spin systems can be obtained. Unfortunately, an
exact calculation of the spectrum is often restricted due to the huge dimension of the Hilbert
space even if one works within the most simple isotropic Heisenberg model. The dimension grows
exponentially for a system of N spins with spin quantum number s and is (2s + 1)N .
In order to get insight into the properties of large magnetic molecules one can access several
numerical methods which were developed in the past. Of course, the ultimate method of choice
would be an exact numerical diagonalization yielding the complete energy spectrum. In recent
years there has been an enormous progress on extending the range of applicability of the exact
numerical diagonalization of the Heisenberg model. To this end the use of spin-rotational symmetry
[1,2] in combination with point-group symmetries [3–6] can be of great advantage with respect to a
reduction of computational requirements, i. e. a need of hardware resources and computation time.
Apart from the exact numerical diagonalization technique the magnetism of magnetic molecules
can be very well investigated using complementary methods such as Density Matrix Renormali-
zation Group (DMRG) [7–9], Lanczos [10], or Quantum Monte Carlo (QMC) [11–13] techniques.
Nevertheless, these methods also suffer from theoretical limitations, QMC for instance in systems
with geometric frustration.
Currently magnetic molecules which exhibit geometric frustration are of special interest due to
the richness of physical phenomena like plateaus and jumps of the magnetization for a varying field
as well as special features of their spectra such as low-lying singlets [14–17]. In this respect a lot
of insight has been obtained by investigating molecular representations of archimedean-type spin
systems [18,19], i. e. systems in which participating spins occupy the vertices of an archimedean
solid. Such representations were already synthesized several years ago and they exist for example as
{Cu12La8} [20] (cuboctahedron, s = 1/2) and {Mo72Fe30} [21,22] (icosidodecahedron, s = 5/2). In
∗E-mail: rschnall@uos.de
c© R. Schnalle, A.M. Läuchli, J. Schnack 331
R. Schnalle, A.M. Läuchli, J. Schnack
this context the molecular compound {Mo72Fe30} is probably one of the most thoroughly investi-
gated magnetic molecules. However, a theoretical explanation of its special physical properties has
been so far given mostly by considering purely classical models [23,24] or directly related quantum
mechanical counterparts like the rotational-band model [25,26].
In this paper we want to show how the approximate diagonalization technique which was
developed and applied to unfrustrated, i. e. bipartite, magnetic molecules in [6] can be used in
order to determine the energy spectrum of geometrically frustrated magnetic molecules. The idea
of this technique is to diagonalize the full Hamiltonian in a reduced basis set. The basis set itself
is an eigenbasis of the rotational-band Hamiltonian. Such an ansatz has already been used by
Oliver Waldmann [27] in order to interpret inelastic neutron spectra of {Mo72Fe30} [18]. We will
demonstrate that in contrast to bipartite systems for a frustrated spin system, the rotational-
band states of all non-trivially different sublattice colorings have to be taken into account in
order to achieve a reliable convergence of energy levels. Throughout this paper we use a spin
system with cuboctahedral structure and spin quantum numbers s = 1 and s = 3/2 as well as an
antiferromagnetic coupling as an archetypical example of a frustrated magnetic spin system.
This paper is organized as follows. In section 2 the general theoretical description of the system
within the Heisenberg model as well as general remarks on the use of irreducible tensor opera-
tors and point-group symmetries are given. In section 3 the theoretical basics of the approximate
diagonalization within the isotropic Heisenberg model are briefly reviewed and specified for the
cuboctahedral spin system. The convergence behaviour of the approximate diagonalization is dis-
played and deeply discussed as well as the specific heat and zero-field magnetic susceptibility for a
cuboctahedron with s = 3/2. Furthermore, a numerically based finding of an approximate selection
rule is reported. This paper closes with a summary in section 4.
2. Theoretical method
In order to model the physics of antiferromagnetic molecules it has been shown that an isotropic
Heisenberg Hamiltonian with an additional Zeeman-term and an antiferromagnetic nearest-neighbor
coupling provides the dominant terms. Such a Hamiltonian looks like
H∼ = −
∑
〈i,j〉
Jij s∼(i) · s∼(j) + gμBS∼ · �B. (1)
The indices of the sum are running over all pairs 〈i, j〉 of interacting spins i and j. The first part
consisting of the sum over single spin operators s∼(i) at sites i interacting with the coupling strength
Jij < 0 refers to the Heisenberg exchange whereas the second part – the Zeeman-term – couples
the total spin S∼ to an external magnetic field �B.
Taking without loss of generality Jij = J for interacting spins, the Heisenberg part assumes
the following form
H∼Heisenberg = −J
∑
〈i,j〉
s∼(i) · s∼(j), (2)
where each coupling is counted only once. Since due to SU(2) symmetry the commutation relations[
H∼Heisenberg, S∼
]
= 0 hold, a common eigenbasis {|ν〉} of H∼Heisenberg, S∼
2 and S∼z can be found and
the effect of an external magnetic field �B = B · �ez can be included later, i. e.
Eν(B) = Eν + gμBBMν . (3)
Here Eν denotes the energy eigenvalues, |ν〉 the eigenstates, and Mν denotes the corresponding
magnetic quantum number.
For the matrix representation of the Heisenberg Hamiltonian (2) the irreducible tensor operator
technique is used [1,2,6]. The Heisenberg Hamiltonian is expressed in terms of irreducible tensor
332
Approximate eigenvalue determination of geometrically frustrated magnetic molecules
operators and its matrix elements are evaluated using the Wigner-Eckart-theorem
〈α S M |T∼(k)
q |α′ S′ M ′〉 =
(−1)S−M 〈α S||T∼(k)||α′ S′〉
(
S k S′
−M q M ′
)
. (4)
Equation (4) states that a matrix element of the q-th component of an irreducible tensor operator
T∼
(k) of rank k is given by the reduced matrix element 〈α S||T∼(k)||α′ S′〉 and a factor containing a
Wigner-3J symbol [28]. The basis in which the Hamilton matrix is set up is of the form |α S M〉.
α refers to a set of intermediate quantum numbers given by addition rules when coupling single
spins s∼(i) to a total spin S∼ with spin quantum number S. The underlying spin-coupling scheme
directly effects the form of the irreducible tensor operator T∼
(k) and further it effects the successive
calculation process of the reduced matrix elements in equation (4).
By using irreducible tensor operators and the Wigner-Eckart-theorem it is possible to drastically
reduce the dimensionality of the problem, i. e. of the Hamilton matrices which have to be diago-
nalized numerically. The total Hilbert space H can be decomposed into subspaces H(S, M = S).
Such a decomposition results in block factorizing the Hamilton matrix where each block can be
labelled by the total spin quantum number S.
Additionally point-group symmetries lead to a further reduction of the matrices. Apart from
that, states are labelled by the irreducible representation they are belonging to. This is very
helpful in understanding the physics of the system. The incorporation of these symmetries results
in symmetrized basis states which are constructed by the projection operator [29]
P(n)|α S M〉 =
(
ln
h
∑
R
(
χ(n)(R)
)∗
G∼(R)
)
|α S M〉 . (5)
Here ln denotes the dimension of the n-th irreducible representation of the point-group G which is
of the order of h. G∼(R) refers to the symmetry operations of G and χ(n)(R) denotes its character
with respect to n. The effect of the symmetry operation G∼(R) on basis states of the form |α S M〉
is discussed in [3,5,6].
3. Approximate diagonalization for frustrated systems: the cuboctahedron
Figure 1. Geometrical structure of the cuboctahedron [30].
In this section we follow a method of calculating approximate eigenstates and eigenvalues of
the Heisenberg Hamiltonian in equation (2) and determine approximately the energy spectrum
of a spin system with cuboctahedral structure. In this system 12 spins of spin quantum number
s occupy the vertices of a cuboctahedron interacting along its edges. The geometrical structure
of the cuboctahedron is shown in figure 1. A detailed description of the approximation scheme is
333
R. Schnalle, A.M. Läuchli, J. Schnack
given in [6] where it was successfully applied in order to determine approximate energy spectra and
thermodynamic properties of spin rings, i. e. bipartite systems. This approximation rests on the
idea of diagonalizing the full Hamiltonian in a reduced basis set. The basis set itself is an eigenbasis
of the rotational-band Hamiltonian which from the point of view of perturbation theory can be
understood as an approximation to the full Hamiltonian. For bipartite systems this (zeroth order)
approximation is already very good [25].
The Heisenberg Hamiltonian can be decomposed into two parts like
H∼ = H∼RB + H∼
′ , (6)
where H∼RB is the rotational-band Hamiltonian [25,26,31] and H∼
′ is an operator containing the re-
maining terms. The rotational-band Hamiltonian which is an effective quantum mechanical Hamil-
tonian based on classical assumptions looks like
H∼RB = −DJ
2N
[
S∼
2 −
Ns∑
n=1
S∼
2
n
]
. (7)
Here N denotes the number of spins within the system and Ns denotes the number of sublattices
the classical ground state of the system is composed of. The prefactor −DJ/(2N) can be seen as
the effective coupling strength between the total spin S∼ and the sublattice spins S∼n which arise
from coupling all the single spins s∼(i) belonging to the n-th sublattice. The value of D directly
depends on the system. It is chosen such that the energy of the ferromagnetic state of the system
is matched; for the cuboctahedron it is D = 6.
The full Heisenberg Hamiltonian is now diagonalized within a reduced set {|φi〉}, i = 1, . . . , nred,
of basis states of H∼RB yielding approximate eigenstates and eigenvalues of H∼ . The set of approxi-
mate basis states is energetically ordered.
Figure 2. Part of the energy spectra of the rotational-band Hamiltonian for a cuboctahedron
with twelve spins s = 1 (left) and s = 3/2 (right). Seven super-bands are colored.
Before discussing the results of our approximate diagonalization we would like to characterize
the eigenbasis of H∼RB. Figure 2 shows the low-lying part of the energy spectra of H∼RB for a
cuboctahedron with s = 1 and s = 3/2 given by the rotational-band model. They exclusively
consist of parabolas – so-called rotational bands. The eigenvalues ERB(S1, S2, S3, S) of H∼RB depend
only on the spin quantum numbers of the sublattice spins Sn, with n = 1, 2, 3, and of the total spin
S. Corresponding eigenstates are trivial and analytically given as |α S1 S2 S3 S M〉. The additional
quantum number α refers to a set of intermediate spin quantum numbers which appears when
coupling single spins s∼i of the same sublattice to the corresponding sublattice spins S∼i and further
on coupling the sublattice spins to the total spin S∼. With regard to the set of intermediate spin
quantum numbers α the number of all possible ways of constructing states characterized by fixed
334
Approximate eigenvalue determination of geometrically frustrated magnetic molecules
values of the sublattice and total spin quantum numbers determine the degeneracy of energy levels
in the spectrum of the rotational-band Hamiltonian.
In the case of s = 1 the lowest band in figure 2 is given by states |α S1 S2 S3 S M〉 with sublattice
spin quantum numbers S1 = S2 = S3 = Smax = 4 · 1 = 4 while the second band is given by a
deviation of one sublattice spin of 1, i. e. |α (Smax − 1)Smax Smax S M〉 and permutations thereof.
The other bands can then be constructed by introducing additional deviations of the sublattice
spin quantum numbers. The energy spectrum with s = 3/2 shown in figure 2 can be constructed
accordingly.
Additionally several energy levels in figure 2 are colored. This coloring refers to the so-called
super-bands. A super-band consists of those rotational bands for which the sum of sublattice
spin quantum numbers is the same. In contrast to bipartite systems (see [6]) the spectrum of the
rotational-band Hamiltonian for the cuboctahedron is much denser at low energies and only the
first three super-bands are well separated from the others.
The classical ground state of the system plays the key role within the approximate diagonali-
zation. The better the quantum mechanical system can be approximated by a classical picture of
the ground state the more effectively the approximate diagonalization works, i. e. the faster the
approximate energy eigenvalues converge towards the exact values. From a purely classical point
of view the ground state of a cuboctahedron exhibits a three-sublattice structure and is infinitely
degenerate since there coexist coplanar and non-coplanar vector orientations [32]. Here it should
be emphasized that within the approximate diagonalization, i. e. within a quantum mechanical
treatment, coplanarity of the classical ground state does not play a role, but the coloring of the
classical ground state does, since there is a direct impact on the set of intermediate spin quantum
numbers α in the rotational-band states |α S1 S2 S3 S M〉 which are taken into account as basis
states in the approximate diagonalization.
Figure 3. Families of the classical ground state of the cuboctahedron. Coloring of the Γ(q = 0)-
family (upper left). Coloring of the three equivalent states of the M -family (upper right and
bottom). The drawing is a schematic planar projection. The labels correspond to y: yellow, r:
red, and b: blue.
The classical ground state of the cuboctahedron exhibits 24 colorings of the spins which can
be – by group theoretical considerations – decomposed into two families that are invariant under
operation of the full point-group symmetry Oh [19]. Following [19] these families will be denoted
335
R. Schnalle, A.M. Läuchli, J. Schnack
as Γ(q = 0)-family and M -family. It has also been shown that those irreducible representations of
Oh which form the Γ(q = 0)- and M -families are found in the low-lying part of the spectrum of the
quantum cuboctahedron with half-integer spins s = 1/2, 3/2, 5/2 [19]. Figure 3 shows the colorings
of the different classical ground state families of the cuboctahedron. In a classical picture, each
color refers to a sublattice with all spins pointing towards the same direction. The angle between
classical spins belonging to different sublattices is 120◦.
In order to calculate the approximate spectrum of the system one is now left with the con-
struction of basis states of the form |α S1 S2 S3 S M〉, i. e. quasi-classical states. Therefore, spins
belonging to the same sublattice have to be coupled to yield the total sublattice spins S∼1, S∼2 and
S∼3. Afterwards these total sublattice spins are coupled to the total spin S∼. The underlying coupling
scheme is given by a classical ground state, i. e. a coloring from figure 3 and incorporated in the
quantum number α. In what follows, the resulting basis states will be labelled with respect to
the classical reference state. To this end, one has to distinguish between basis states of the form
|γ S1 S2 S3 S M〉Γ, |μ1 S1 S2 S3 S M〉M1 , |μ2 S1 S2 S3 S M〉M2 and |μ3 S1 S2 S3 S M〉M3 . The nota-
tion of the set of intermediate quantum numbers and the subscript of the states now directly point
to the classical ground state colorings, i. e. the underlying coupling scheme. It is important to note,
that each of these four basis sets spans the same Hilbert space H(S, M).
In the case of the M -family, three different colorings exist. When approximately diagonalizing
the Hamiltonian and additionally using point-group symmetries one has to restrict to those sym-
metry groups where the symmetry operations do not alter the sublattice structure. A symmetry
operation has no impact on the sublattice structure if the corresponding spin permutation results
in recoloring of spins where all spins of a given sublattice maintain the same color, i. e. subscript (r:
red, y: yellow, b: blue). For example an operation which leads to a cyclic permutation of colorings
like
r → y → b → r (8)
has no impact on the sublattice structure.
While the sublattice structure of the Γ(q = 0)-family is left invariant under all symmetry
operations of Oh, the sublattice structure (i. e. coloring) of the M -family is not. In order to restore
the sublattice invariance of the classical ground state belonging to the M -family the basis states
for the approximate diagonalization can be constructed from a usually overcomplete set M of basis
states by an orthonormalization procedure. This set consists of rotational-band eigenstates from
each coloring of the M -family, i. e. it is given by
M =
{|μ1 S1 S2 S3 S M〉iM1
}
+
{|μ2 S1 S2 S3 S M〉iM2
}
+
{|μ3 S1 S2 S3 S M〉iM3
}
, (9)
while the index i is taking the values 1, . . . , nred. Here nred reflects the overall number of states
contained in the incorporated rotational bands which is independent of the choice of the coupling
scheme.
Since the underlying coupling scheme is different regarding the chosen coloring, these states
have to be converted into an – in general – arbitrary reference scheme A before calculations can be
performed. A transition between states belonging to different coupling schemes can be calculated
using general recoupling coefficients [33,34]. For the transition of a basis state |μ1 S M〉M1 referring
to a coloring M1 of the M -family into the scheme A one yields
|μ1SM〉M1 =
∑
α,S1,S2,S3
A〈α S M |μ1 S M〉M1 |α S M〉A , (10)
where the summation indices indicate that the summation is running over all valid combinations
of values for the intermediate spin quantum numbers α as well as for the sublattice spin quantum
numbers. The transitions between states of colorings M2 and M3 into the scheme A can be obtained
analogously.
Figure 4 shows the classification of the quasi-classical states with lowest energies within S = 0
and S = 1 subspaces according to irreducible representations of Oh. It corresponds to an ap-
proximate diagonalization of H∼Heisenberg using only a reduced basis set of states from the lowest
336
Approximate eigenvalue determination of geometrically frustrated magnetic molecules
0 1
S
-31.5
-31
E
/|J
|
A1g
A2g
Eg
T2u
Γ(q=0)-family
0 1
S
M-family
Figure 4. Classification by irreducible representations of Oh of the quasi-classical states with
lowest energies of the Γ(q = 0)-family and M -family within S = 0 and S = 1 subspaces. The
given energies refer to a diagonalization of the Heisenberg Hamiltonian exclusively within these
quasi-classical states in the system with s = 3/2.
rotational band in the system with s = 3/2. Obviously, the classification of the quasi-classical
states is independent of the single spin quantum number s. Solely, the energies are changed when
approximately diagonalizing the Hamiltonian of a system with different single spin quantum num-
bers. Based on a classical ground state belonging to the Γ(q = 0)-family the states belong to the
irreducible representations A1g (S = 0) and A2g, Eg (S = 1). Taking the M -family as a starting
point for the construction of basis states the states belong to A1g, Eg (S = 0) and A1g, Eg, T2u
(S = 1) where the T2u-state has apart from its intrinsic degeneration, i. e. the dimension of the
irreducible representation T2u, a twofold multiplicity.
The difference in the geometric properties of the lowest states of the chosen ground-state family,
expressed by their decomposition into different irreducible representations, should directly lead
to a convergence behaviour which depends on the choice of the underlying coloring. Since the
approximate basis set contains energetically ordered states (from lowest to higher), low-lying states
are expected to converge faster with growing number of basis states used for the approximate
diagonalization [6].
3.1. Convergence of individual colorings
In figure 5 the approximate eigenvalues of a cuboctahedron with s = 1 and s = 3/2 depending
on the number of incorporated bands within the S = 0 subspaces are shown. The Γ(q = 0)-family
as well as one member of the M -family have been chosen as underlying classical ground states for
each system. In every subfigure the last column refers to the results from a complete diagonalization
within H(S = 0). Since the Γ(q = 0)-family is invariant under Oh, the full point-group symmetry of
a cuboctahedron (Oh) was used in order to classify the states. As mentioned before, the individual
members of the M -family are not invariant under Oh but under D2. Thus, in this case the D2
point-group symmetry was applied.
Looking at figure 5 it becomes apparent that the convergence is directly dependent on the
choice of the underlying classical ground states. In agreement with theoretical expectations, the
states which correspond to low-lying eigenstates of the rotational-band model converge faster than
higher-lying states. Additionally, by comparing the spectra of s = 1 and s = 3/2 it can be seen that
the ground state energy converges more rapidly with increasing single spin quantum number s.
Nevertheless, apart from some states which exhibit a quite regular and fast convergence behaviour,
the convergence is – overall – rather poor. Especially the so-called low-lying singlet1, which ap-
pears in the case of half-integer spins s below the first triplet contributions, converges very slowly.
This means that it is poorly approximated by one or by a few eigenstates of the rotational-band
Hamiltonian, but instead is a superposition of very many basis states.
1Here: the lowest Eg-state in the system with s = 3/2 based on a classical ground state coloring of the Γ(q = 0)-
family.
337
R. Schnalle, A.M. Läuchli, J. Schnack
2 4 6 8 10 12 14 16 18 20 22
# bands
-18
-17
-16
-15
-14
-13
E
/|J
|
A1g
A2g
Eg
T1g
T2g
A1u
A2u
Eu
T1u
T2u
2 4 6 8 10 12 14 16 18 20 22
# bands
-18
-17
-16
-15
-14
-13
E
/|J
|
A1
A2
B1
B2
5 10 15 20 25 30 35 40 45 50
# bands
-36
-35
-34
-33
-32
-31
-30
-29
E
/|J
|
A1g
A2g
Eg
T1g
T2g
A1u
A2u
Eu
T1u
T2u
2
4
6
8 10 12 14 16
18
5 10 15 20 25 30 35 40 45 50
# bands
-36
-35
-34
-33
-32
-31
-30
-29
E
/|J
|
A1
A2
B1
B2
Figure 5. Low-lying energy spectra of a cuboctahedron with s = 1 (upper row) and s = 3/2
(lower row) within S = 0 subspaces using an increasing number of bands (states). The underlying
classical ground states belong to the Γ(q = 0)-family (left) and to one coloring of the M -family
(right) while Oh (left) and D2 (right) point-group symmetries were used in order to classify the
states. The dashed boxes in the lower left figure indicate regions, i.e super-bands, in which the
energy of the ground state is considerably lowered. The numbers refer to the number of magnons
of the super-bands.
One significant property which becomes evident when tracing single energy levels is that the
convergence is stepwise – at least in the low-lying part of the spectra. Regions in which the ground
state is considerably lowered are exemplarily marked with dashed boxes in the case of a cubocta-
hedron s = 3/2 assuming a classical reference state of the Γ(q = 0)-family. The numbers below the
boxes refer to the number of magnons existing in the states of those rotational bands which are
incorporated within the marked regions. Obviously, the ground state energy is lowered whenever
super-bands with an even number of magnons are incorporated in the approximate diagonaliza-
tion. The observation of a stepwise convergence leads to a helpful additional approximation given
by an approximate selection rule which is discussed below (see section 3.4). It should be menti-
oned here that qualitatively the same convergence behaviour can be observed within subspaces of
S > 0. However, additional information cannot be extracted from graphical visualizations of the
convergence in these subspaces. Thus, they will not be presented here.
3.2. Convergence of combined colorings
In order to improve the convergence of the energy levels in comparison with the behaviour
shown in figure 5 the decomposition of the quasi-classical states in figure 4 directly leads to the
starting point. Since the quasi-classical states, that result from different colorings, can be classified
according to irreducible representations of 2A1g and Eg, it is a straightforward task to use linear
combinations of rotational-band eigenstates of the Γ(q = 0)-family as well as of the M -family as
basis states for the approximate diagonalization. The set of basis states results from an extension
338
Approximate eigenvalue determination of geometrically frustrated magnetic molecules
of M in equation (9) to
G =
{|μ1 S1 S2 S3 S M〉iM1
}
+
{|μ2 S1 S2 S3 S M〉iM2
}
+
{|μ3 S1 S2 S3 S M〉iM3
}
+
{|γ S1 S2 S3 S M〉iΓ
} (11)
and subsequent orthonormalization of the incorporated states.
2 4 6 8 10 12 14 16 18 20 22
# bands
0
1
2
3
4
5
(E
-E
0
(S
))
/|J
|
2 4 6 8 10 12 14 16 18 20 22
0
2
4
A1g
Eg
T2g
T1u
1 2 3 4 5 6 7 8 9 10 full
# bands
0
0.5
1
1.5
2
2.5
3
(E
-E
0
(S
))
/|J
|
A1g
Eg
T2g
T1u
Figure 6. Energy difference to the ground state of selected low-lying energy levels of the cuboc-
tahedron s = 1 within S = 0 subspace dependening on the number of incorporated rotational
bands. The reduced set of basis states is taken from rotational-band states of the Γ(q = 0)-family
only (left) and from linear combinations of states of all four colorings (right). The classification
is according to Oh point-group symmetry. The dotted red line refers to the exact difference from
the Eg-state.
In figure 6 the energy difference between the selected low-lying energy levels and the ground
state of the system is displayed for a cuboctahedron s = 1 within S = 0 subspace. The difference
is shown depending on the number of incorporated rotational bands. In both subfigures the last
column refers to the exact values, taken from a complete diagonalization. The dotted red line
indicates the energy difference to the Eg-state.
In the first case (left subfigure) only rotational-band states of the Γ(q = 0)-family are taken
into account for the approximate diagonalization. In the second case (right subfigure) linear com-
binations of states of all four colorings are used. The first noticeable difference is that when taking
basis states of all colorings into account the low-lying levels start and remain in close proximity to
their final (true) values. This is especially obvious when comparing the convergence of the lowest
Eg-state in both subfigures. The second difference is that when using linear combinations of states
of all four colorings the convergence is smoother and more rapid. This can be traced back to the
fact that the different colorings contribute to the low-lying states of different irreducible represen-
tations which in the other families would only be available as high-lying states. Therefore, it is
advantageous to use fewer bands but more basis states of all classical colorings for the approximate
diagonalization.
A short explanation might be appropriate in order to understand why the energy differences
sometimes increase when taking more states into account. This is due to the fact that the ground
state in such cases converges more rapidly than the excited states. Although all approximate
eigenvalues improve on an absolute scale the differences get worse for the moment.
3.3. Results
In this section we should like to present how well thermodynamic observables can be approx-
imated. Figure 7 shows the specific heat C(T, B) (left) and the magnetic susceptibility dM/dB
(right) for zero field B = 0 of a cuboctahedron s = 3/2. The energy spectrum of this system can
be completely calculated using irreducible tensor operator technique in combination with a D2
point-group symmetry [35]. In figure 7 the exactly calculated specific heat and the susceptibility
339
R. Schnalle, A.M. Läuchli, J. Schnack
are compared with the results from an approximate diagonalization using only states of the lowest
rotational band in each coloring in order to set up the basis. We also show how these observables
look like if evaluated with only the lowest rotational band (L-band) of H∼RB (red colored energy
levels in figure 2).
0 0.2 0.4 0.6
T/|J|
0
0.5
1
1.5
C
/(
2s
k
B
)
s=3/2: exact
s=3/2: L-band
s=3/2: 1 band
0 0.5 1 1.5 2 2.5 3
T/|J|
0
1
2
3
4
5
dM
/d
B
(
a.
u.
)
s=3/2: exact
s=3/2: L-band
s=3/2: 1 band
Figure 7. Comparison of approximate and exact specific heat C(T, B) (left) and zero-field suscep-
tibility dM/dB (right) for a cuboctahedron s = 3/2. The L-band refers to magnetic properties
evaluated with the lowest band of the rotational-band Hamiltonian. For the approximate diag-
onalization, only the states of the lowest rotational band but of each coloring are taken into
account.
As one can see, the low-temperature specific heat is very sensitive to the structure of low-lying
levels. In the rotational-band model the lowest band (L-band) is gapped and describes a rotor.
Therefore, the specific heat is suppressed at small temperatures and at higher temperatures it
approaches 3/2 kB. The approximate diagonalization, although only taking the lowest (L-band)
states of all four coloring into account, achieves already an improvement for low-temperatures.
The little Schottky-peak is the result of low-lying singlets and rearranging triplets. At higher
temperatures this approximation displays the same behavior as the pure L-band which is expected
since only L-band-like states are incorporated.
In contrast to the specific heat which directly reflects the density of states within a certain
energy interval, the magnetic susceptibility reflects the density of magnetic states. To this end, it
is not surprising that the magnetic susceptibilities calculated from the approximate spectra do not
differ considerably. Since the exact spectrum also possesses a rotational-band like behaviour at the
lower edge, the contributions from the approximate diagonalization and from the L-band very well
reproduce a steep rise of susceptibility for low temperatures. Obviously, the exact thermodynam-
ical properties cannot be properly reproduced in the case of increasing temperature because the
approximate spectra only contain a fraction of the energy levels of the full Hilbert space H.
3.4. Approximate selection rule
As mentioned before (section 3.1), the stepwise convergence behaviour leads to an approximate
selection rule. Using this selection rule, the computational effort when setting up the Hamilton
matrices can be further reduced. In figure 5 it was shown that when diagonalizing in the Γ(q = 0)-
family, regions can be marked in which single energy levels are affected by taking into account
additional rotational-band states. The same can be done for the M -family. As it was already
mentioned, the marked regions coincide with the incorporation of rotational-band states belonging
to the same super-band.
Looking at figure 5 the approximate energy eigenvalues of the ground state are lowered whenever
states |γ S1 S2 S3 S M〉 are additionally included which belong to a super-band with an even number
of magnons. Since the convergence behaviour is state-sensitive, the energy of the lowest states
belonging to the 3-dimensional irreducible representation T1u, for example, is considerably lowered
whenever super-bands with an odd number of magnons are included. This relation between the
340
Approximate eigenvalue determination of geometrically frustrated magnetic molecules
incorporation of super-bands and the immediate affection on the energies of certain approximate
eigenstates can also be observed if the classical ground state belongs to the M -family.
As a result of the aforementioned observations a simple rule can be conjectured for low-lying
energy levels |S1,b S2,b S3,b S M〉. The matrix elements 〈S1,a S2,a S3,a S M |H∼ |S1,b S2,b S3,b S M〉 con-
necting these states with other states |S1,a S2,a S3,a S M〉 are one magnitude (or more) bigger than
other matrix elements if
|na − nb| mod 2 = 0, (12)
where na =
∑3
i=1 Si,max −
∑3
i=1 Si,a and nb (similarly evaluated), denote the number of magnons
of the super-bands the states are belonging to. The Hamilton matrices approximately split up ac-
cording to equation (12). Thus, computation time is saved and at the same time the dimensionality
of the problem is reduced.
4. Summary
In this paper an approximate diagonalization scheme was proposed and used in order to deter-
mine the energy spectra of geometrically frustrated spin systems. As an example, the cuboctahe-
dron was discussed. It was shown that the convergence is clearly dependent on the coloring of the
underlying classical ground state, i.e on the coupling scheme. Therefore, the approximation can be
improved by using an adapted set of basis states that originates from rotational-band states of all
possible sublattice colorings. Furthermore, it was shown that the approximate diagonalization is
useful in order to study the low-temperature thermodynamics of geometrically frustrated systems,
even in its simplest form when only states from one rotational band within each coloring are taken
into account.
Although the necessary calculations are rather involved, especially for huge frustrated systems,
the approximate diagonalization can be a valuable method. Numerically the evaluation of recou-
pling coefficients constitutes the strongest challenge, i. e. calculations are rather limited by runtime
than by available memory. However, recent developments show that highly parallelized program
code or public resource computing can help to overcome this barrier.
Acknowledgements
Computing time at the Leibniz Computing Centre in Garching is greatfully acknowledged
as well as helpful advices from Mohammed Allalen. We also thank Boris Tsukerblat for fruitful
discussions about the irreducible tensor operator technique. This work was supported within a
Ph.D. program of the State of Lower Saxony in Osnabrück.
References
1. Gatteschi D., Pardi L., Gazz. Chim. Ital., 1993, 123, 231.
2. Borras-Almenar J.J., Clemente-Juan J.M., Coronado E., Tsukerblat B.S., Inorg. Chem., 1999, 38,
6081.
3. Waldmann O., Phys. Rev. B, 2000, 61, 6138.
4. Bostrem I.G., Ovchinnikov A.S., Sinitsyn V.E., Theor. Math. Phys., 2006, 149, 1527.
5. Sinitsyn V.E., Bostrem I.G., Ovchinnikov A.S., J. Phys. A: Math. Theor., 2007, 40, 645.
6. Schnalle R., Schnack J., Phys. Rev. B, 2009, 79, 104419.
7. White S.R., Phys. Rev. B, 1993, 48, 10345.
8. Exler M., Schnack J., Phys. Rev. B, 2003, 67, 094440.
9. Schollwöck U., Rev. Mod. Phys., 2005, 77, 259.
10. Lanczos C., J. Res. Nat. Bur. Stand., 1950, 45, 255.
11. Sandvik A.W., Kurkijärvi J., Phys. Rev. B, 1991, 43, 5950.
12. Sandvik A.W., Phys. Rev. B, 1999, 59, R14157.
13. Engelhardt L., Luban M., Phys. Rev. B, 2006, 73, 054430.
14. Schulenburg J., Honecker A., Schnack J., Richter J., Schmidt H.-J., Phys. Rev. Lett., 2002, 88, 167207.
341
R. Schnalle, A.M. Läuchli, J. Schnack
15. Schröder C., Schmidt H.-J., Schnack J., Luban M., Phys. Rev. Lett., 2005, 94, 207203.
16. Schmidt R., Schnack J., Richter J., J. Magn. Magn. Mater., 2005, 295, 164.
17. Schnack J., Schmidt R., Richter J., Phys. Rev. B, 2007, 76, 054413.
18. Garlea V.O., Nagler S.E., Zarestky J.L., Stassis C., Vaknin D., Kögerler P., McMorrow D.F., Nieder-
mayer C., Tennant D.A., Lake B., Qiu Y., Exler M., Schnack J., Luban M., Phys. Rev. B, 2006, 73,
024414.
19. Rousochatzakis I., Läuchli A.M., Mila F., Phys. Rev. B, 2008, 77, 094420.
20. Blake A.J., Gould R.O., Grant C.M., Milne P.E.Y., Parsons S., Winpenny R.E.P., J. Chem. Soc.-Dalton
Trans., 1997, 485.
21. Müller A., Sarkar S., Shah N., Bögge H., Schmidtmann M., Sarkar S., Kögerler P., Hauptfleisch B.,
Trautwein A.X., Schünemann V., Angew. Chem., Int. Ed., 1999, 38, 3238.
22. Müller A., Luban M., Schröder C., Modler R., Kögerler P., Axenovich M., Schnack J., Canfield P.C.,
Bud’ko S., Harrison N., Chem. Phys. Chem., 2001, 2, 517.
23. Axenovich M., Luban M., Phys. Rev. B, 2001, 63, 100407.
24. Schröder C., Prozorov R., Kögerler P., Vannette M.D., Fang X., Luban M., Phys. Rev. B, 2008, 77,
224409.
25. Schnack J., Luban M., Phys. Rev. B, 2000, 63, 014418.
26. Schnack J., Luban M., Modler R., Europhys. Lett., 2001, 56, 863.
27. Waldmann O., Phys. Rev. B, 2007, 75, 012415.
28. Varshalovich D.A., Moskalev A.N., Khersonskii V.K. Quantum Theory of Angular Momentum. World
Scientific Publishing, 1988.
29. Tinkham M. Group Theory and Quantum Mechanics. Dover Publications, New York, 2003.
30. Weisstein E. Mathworld. URL http://mathworld.wolfram.com
31. Waldmann O., Phys. Rev. B, 2001, 65, 024424.
32. Schmidt H.-J., Luban M., J. Phys. A: Math. Gen., 2003, 36, 6351.
33. Fack V., Pitre S.N., van der Jeugt J., Comp. Phys. Comm. 86 (1995) 105
34. Fack V., Pitre S.N., van der Jeugt J., Comp. Phys. Comm., 1997, 101, 155.
35. Schnack J., Schnalle R. Polyhedron, doi:10.1016/j.poly.2008.10.017
Наближене визначення власних значень геометрично
фрустрованих магнiтних молекул
Р.Шнале1, А. Лойхлi2,Ю.Шнак3
1 Унiверситет Ознабрюка, вiддiлення фiзики, D-49069 Ознабрюк, Нiмеччина
2 Iнститут Макса Планка фiзики складних систем, Ньотнiтцерштрасе 38, D–01187 Дрезден, Нiмеччина
3 Унiверситет Бiлєфельда, факультет фiзики, поштова скринька 100131, D–33501 Бiлєфельд,
Нiмеччина
Отримано 29 травня 2009 р.
Геометрично фрустрованi магнiтнi молекули привертають багато уваги як у дiлянцi молекулярного
магнетизму, так i в дiлянцi фрустрованих антиферомагнетикiв Гайзенберга. У цiй статтi ми демонст-
руємо, як наближена дiагоналiзацiйна схема може бути використана для отримання термодинамi-
чної i спектроскопiчної iнформацiї про фрустрованi магнiтнi молекули. Ми теоретично дослiджуємо
антиферомагнiтно взаємодiючу спiнову систему з кубоктаедричною структурою, модельовану iзо-
тропним гамiльтонiаном Гайзенберга.
Ключовi слова: магнiтнi молекули, модель Гайзенберга, геометричнi фрустрацiї, технiка
незвiдних тензорних операторiв, наближена дiагоналiзацiя, кубоктаедрон
PACS: 75.10.Jm, 75.50.Xx, 75.40.Mg, 75.50.Ee
342
|