HAL Id: jpa-00247136
https://hal.archives-ouvertes.fr/jpa-00247136
Submitted on 1 Jan 1995
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
A simple Derivation of the Tunneling Splitting for Large Quantum Spins
Françoise Hartmann-Boutron
To cite this version:
Françoise Hartmann-Boutron. A simple Derivation of the Tunneling Splitting for Large Quantum Spins. Journal de Physique I, EDP Sciences, 1995, 5 (10), pp.1281-1300. �10.1051/jp1:1995197�.
�jpa-00247136�
Classification Physics Abstracts
75.60Jp 73.40Gk 76.20+q
A Simple Derivation of trie Thnnefing Splitting for Large Quantum Spins
Flançoise
Hartmann-BoutronLaboratoire de Spectrométrie Physique
(UJF-CNRS,
UA08),
Université J. Fourier/Grenoble
I,B-P. 87, 38402 Saint-Martin d'Hères Cédex, France
(Received
8 June 1995, received and accepted in final form 6July1995)
Abstract. With the help of
a perturbation treatment to lowest order, general expressions for the tunneling splitting have been derived for a quantum spin S. This spm is assumed to have a main axial Harniltonian
(without
appliedlongitudinal
magneticfield),
and to besubmitted to perturbations with orthorhombic, tetragonal, trigonal or hexagonal symmetries, or
to a transverse field. The formulas obtained are valid for general
(not
verysmall)
S. They are compared with available results of the literature. Previous observations of tunneling splittingsin terbium compounds by EPR are mentioned, with the relevant references.
Introduction
The
tunneling
of alarge single
quantumspin
S is of theoretical interest as a mortel for thetunneling
of themagnetization
of small clusters ornanoparticles
[1, 2]. Elaborateapproaches
have been
used,
such asmapping
thelarge
quantum spin into apartiale problem,
which is then solved withhelp
of the conventional W.K.B. method for theparticle problem
[3], orresorting
to instanton
techniques (including
the contribution of theBerry phase)
[4].In what
follows,
we present a naivecalculation,
based on lowest orderperturbation theory,
for a spin submitted to a main axial Hamiltonian7io
"
-AS),
A >0,
and to zerolongitudinal
field. The
tunneling splitting
2r is inducedby perturbations 7ii
of thespin Hamiltonian,
with orthorhombic,tetragonal, trigonal,
orhexagonal
symmetries, orby
a transverse field(which
is theonly possibility
for halfinteger
spin, because of trie Kramerstheorem).
To lowest
perturbation
order, r can berepresented
as the shortest chainconnecting
the twodegenerate
states(e.g.,
(Sz= +S
>).
It can beexpressed exactly,
whatever S, in terms of factonals. Use of theStirling's
formula transforms it into a power of"7ii/7io".
It has been checked that when an exact solution does exist, the "factonal form" is identical to the firstterm of its serres expansion as a function of
"7ii/7io".
On the otherhand,
in the case of atransverse
field,
the "factorial form" and the"power
form" coinciderespectively (exactly
oralmost
exactly)
with results obtainedby
Korenblit and Shender [si andby
Scharf et ai. [3]by
use of different methods
(recurrence
relation andW.K.B.).
© Les Editions de Physique 1995
It is mentioned that
tunneling splitting
hasalready
been observed(although
the name was notused)
in E.P.R. experiments on trivalent terbium inhexagonal yttrium ethylsulfate
[6,ii
and in
tetragonal LiYF4, CaW04,
PbMo04 (8].The detailed
organization
of the paper is as follows:In a first part, the
principle
of the chain method isexplained
on the standard case of an orthorhombicdistortion,
withexplicit examples (S
=
4,
S=
5);
the "factorial form" of r for theground
doublet (Sz = +S > iscompared
with exactsolutions,
when available(S
=
2,
S =3).
Stirling's
transformationleading
to a"power
form" isintroduced;
formulas for excited doublets (Sz = +m > are alsogiven.
The case oftunneling splitting
due to a transversejietd
is thenconsidered;
the "factorial form" of r for theground
state doublet iscompared
with previous results of Korenblit andShender,
obtainedby
a recursionrelationship
for determinants[si,
the
"power
form" iscompared
with results obtainedby
Scharf et ut. withhelp
of a W.K.B.treatment, both for the
ground
doublet and for the excited doublets [3]: the agreement looks verysatisfying.
Formulas are alsogiven
fortetragonat
andtrigonat
distortions. In thehe~agonat
case, there exists a
simple example,
that ofTb~+(J
=
6)
in Y.E.S. for which exact formulasare available in trie literature [6,
ii. Generally speaking however,
search ofsimple tunneling phenomena
in Rare Earthcompounds
raisesdifliculties,
which are listed and discussed.The second part is devoted to the variation
laws,
withS,
of the parameters of thespin
Hamiltonian of thelarge spin
S. Forthis,
it is assumed that S is the sum: S=
£s~,
of N identical individualspins
s, with the maximum value S i= Ns. The parameters of the
spin
Hamiltonian of S areexpressed
in ternis of those of thespin
Harniltonian of an individualspin
s, with the
help
of "transfer coefficients"(analogous
to reduced rnatrix elements of theWigner
Eckart
theorem).
The "transfer coefficients" derived in this second part are used in the hmit S - cc of the first part.1.
Tunneling Splitting
1.I. ORTHORHOMBIC CASE
1.1.1. Justification of trie Chain
Approach by
Use ofSymmetric
andAntisymmetric
Wave Functions. Let us start frorn a Harniltonian:ii
= iio + iii "
-As]
+vis(
+sfj Ill
with A > 0, (v( < A. The lowest state of
7io
is thedegenerate
doublet (Sz = +S >. We want to know whether itsdegeneracy
is liftedby:
7ii =
vis(
+S~
=
2viSj S()
=2vO(iS)
12)where
O(iS)
is a Stevens operatoriAbragam
andBleaney iii,
Table17,
p.867).
The selection rule àm= +2 requires
integer
S. We are not interested in S= 1, for which the
degeneracy
is lifted to first order in v, and we will assume S > 2.Even S. Let us first
study
this case, forexarnple
S= 4. We
only
need to consider the 5 x 5 submatrix of 7icorresponding
to states (Sz = +4 >, (Sz = +2 > and(Sz
= 0 >. Ifone uses the
symmetric
states (4s >,(2s
> and (0 > on the one hand and theantisymmetric
states
(4a
>,(2a
> on the other hand, 7ii has no matrix elements betweensymmetric
andantisymmetric
states, so it further separates into a 3 x 3 matrix 1(0 >,(2s
>,(4s >)
and a 2 x 2 matrix1(2a
>,(4a >).
The 2 x 2 matrices (2s >,(4s
> and(2a
>,(4a
> areidentical,
so allperturbation
terms constructed from them, shift the doublet(4a
>,(4s
> as a whole. Anenergy dilserence between
[4a
> and [4s > canonly
appear when one buildsperturbation
terms goingthrough
[o >, since theseonly
exist for[4s
>. The lowest order term goingthrough
[o >is:
~~ <
4s[7iiÎ2s
><2s[7iiÎo
><o[7iiÎ2s
><2s[7iiÎ4s
>~~
(E4s E2s)(1~4s E0)(E4s E2s)
~
v~l< 4SISÎ
+S~
12S ><2SllSÎ
+ S~ )1°>)~
1-A)3@2)(16)(12)
j~~~~4~
~~ ~~/-
)3~~~jÎ
~ ~ ~~ ~
~)3 6~~[j2
~~ ~~ ~~~in which we have made use of the result:
j
s +M)1
<
S,
MkÎSiÎS'~
~~(5)
(S
+ Mk)!
xÙW
whence:
<
-sjsisj
+ s >=j2s)1 jù)
So state
[4s
> isdisplaced by
2r with respect to[4a
>(tunnel splitting).
Tbeing negative, [4s
> is lower than[4a
>. Note that T is alsoequal
to:~ <
-4j7iij
2 ><-2j7iij
<oj7iij
+ 2 ><+2j7iij
+ 4 >lE4 E2)lE4 Eo)lE4 E-2)
~~~which is the shortest
perturbation
termconnecting directly
thedegenerate
states + 4 >. Itsnumerator has the form of a chain: (4 >- (2 >- (o >- 2 >- 4 > between the two
states + 4 > and 4 >, which can
easily
be visualized on adrawing,
whence the name: "thechain method".
Odd S. For S
= 1 the
degeneracy
is removed to first order. Here we are interested inlarge spins,
inpractice
S > 3. Let usstudy
the case S= 5. We
only
need consider the 6 x 6 submatrix of 7icorresponding
to states + 5 >, + 3 >, +1 >. If one uses thesymmetric
and antisymmetric states, it separates into two 3 x 3 matrices
ils
>,(3s
>,(5s
> and(la
>,(3a
>,(5a
>. These matrices are identical except for triediagonal
matrix elements ofils
>and la >. Indeed:
<
ls(7iiÎls
>= <la(7iiÎla >=<1(S((
-1>=<-1(Sf(
+1 >=SIS +1) #
0It then seems
preferable
to rewrite the Hamiltonian as:7i=7i[+7i[
with7i[
=
7io
+ils
><ls(7iiÎls
><ls(
+(la
><la(7ii (la
><lai (8')
7i[
= 7ii
Ils
><ls(7iiÎls
><ls( (la
><la(7iiÎla
><lai (8")
An energy diiference between
(5s
> and(5a
> willonly
appear when one buildsperturba-
tion terms
going through ils
> and(la
>, because of the diiference in theenergies
of thedenominators. The lowest order such
perturbation
term is for (SS >:,
<
5s(7iiÎ3s
><3s(7ii Ils
><ls(7iiÎ3s
><3s(7iiÎ5s
>~~~~
(E5s E3s)(E5s Els~
<lS(~lÎlS >)(E5s E3s)
~~~or, since <
ls(7iiÎls
> <(E5s Eisl, ôE(~
=ôE~
+ôE5s
with:~~ <
5s(7iiÎ3s
><3s(7ii Ils
><ls(7iiÎls
><ls(7iiÎ3s
><3s(7ii
(SS >~~
lE5s E3s)lE5s Eis)lE5s Eis)lE5s E3s)
~~~~~~55
~5"
j_A)4 (~~)
<
-5(S£(
3 ><-3(S£(
-1 ><-1(S£(
+1 ><+1(Si(
+ 3 ><+3(S£(
+ 5 >~ 16 x 24 x 24 x16
(12)
~~~~
~ÎÎ]2
~
~~~~~
~~ ~~~~Similarly
one can show thatôE(~
=ôE~
+ôE5a
with:ôE5a
" -rState
(5s
> ishigher
or lower than(5a
>depending
on thesign
of v.In this odd S case, r can also be put into the chain form:
~ <
-5(7iiÎ
3 ><-3(7iiÎ
-1><-1(7iiÎl
><1(7iiÎ3
><3(7iiÎ5
>(E5
E3)(E5
Ei)(E5 E-1)(E5 E-3)
~~1.1.2. Generalization
According
to the aboveformulas,
r has the form:<
-S(7ii1
S + 2 >< -S +2(7ii1
S + 4 > < S4(7ii ÎS
2 > < S2(7ii ÎS
>~
(ES ES-2) (ES ES-4). --(ES E-S+4 )(ES E-S+2)
ils)
It has
already
been noticed that the numerator contains12S)!
Since 7io=
-AS),
the denom- inator can also berearranged
as follows:(-A)~~~ lS~ (S 2)~ils~ (S
4)~iiS~ (S
6)~ilS~ (-S
+4)~ils~ 1-S
+2)~i
=
(-A)S-1(2s 2)(2)(2s 4)
x4(2s 6)
x 6 x x 4 x(2s 4)
x2(2s 2)
=
(-A)S-i
x225-2(s i)
x i x(s 2)
x 2 x(s 3)
x 3 x x i x(s -1)
~
~_ ~~s-i~2s-2j~~
~~jj2°~~~~~~°~~'
v~(25)!
(16)
~ # ~chain
"
(_ ~)S-1
y225-2[(
s 1)[j2For even S: ôE«
= 0,
ôEs
=
2r(r
< 0 because it contains(-1)~~~
= -l.
For odd S:
ôE«
= -P,
ôEs
= +P
(where
thesign
of P is that ofv~,
1.e., that ofv).
Remark: In Messiah's notations [9], P coula also be considered as the lowest order off-
diagonal
matrix element between twodegenerate
states:<
-S(7ii ~°
7ii~°
.7ii
~°
7ii1
+ S >i16)
a a a
in which
~°
=
L'~~
~~~~
Matrix elements of this type are
used,
forexample,
whena
Es Em
the
degeneracy
of a level is removedonly
to second orderischiif, Quantum Mechanics,
2nd edition[loi, Eq. (25.24) ).
In our case the tunnelsplitting
coula also be considered asresulting
from the
diagonalization
of a matrix of the form:(S
> S ><
Si
-ôE P(ii)
<
-Si
r -ôEin
which,
to lowestorder,
r is a chainoff-diagonal
matrix element.l.1.3.
Comparison
witl1Exact SolutionsCase S
= 2. Let us measure the
energies
with respect to that of theunperturbed
doublet(Sz
= +2 >. Theeigenvalue equation
can be solvedexactly
with three solutions(refered
tothe energy E = -4A of the
unperturbed doublet):
El
= 2A + 2Afi
+12~ (18)
(perturbed
energy of (0>)
E[
= 0i19)
iperturbed
energy of theantisymmetric
state (2«>) E( ~
= 2A 2A 1 +
12~ (20)
A
_1~~2 ~4 ~6
~~
~A ~
~~A3 ~~~A5
~ ~~~~E(
is theperturbed
energy of thesymmetric
state(2s
>:The first term in the expansion of
ôE(1-12v~
IA) is
identical to 2r btained by the
chain
method.
Note that the exact solution is not derived from thefirst
term ofltenng
its
numericalby
theory
~ ~~~ ~ ~~ ~
~~~~~~~
igiving
the exactE3s(-)
and Eis(+))
~ ~~ ~~ ~
~~~
~~~~
igiving
the exactE3a(-)
andEia(+))
The exact value of the tunnel
splitting
of the lowest doublet isgiven by:
E3s E3a " 12v 4A +
(~
+ 4A(~ (25)
For this same doublet 3s, 3a we have the series expansions:
~~~ ~~
~~Î
~Î
~~ Î~ ~3 ÎÎ
~~
~ ~~~~45 315 405
~~~ ~~
~~A
4 A2 32 A3 ~ 64 A4 ~ ~~~~E3s " E~ +
ôE3s,
E3a " E~ +ôE3s
with:
Ea~ = -9A
-15~ () ~
+...
(28)
(shift
of the doublet as awhole)
~~~~
~~~~ Î ÎÎ
~ ~~~~
(splitting
of thedoublet)
The first term of
ôE3s
coincides with r~ha;n.l.1.4.
Computation
oflbnneling Splitting by
PerturbationExpansion:
Case S= 4. In this
case, there exists an exact solution for E4a~
E4a = -10A
~/36A2
+(4Viv)2 (30)
~~~ ~~~
Î~
~
ÎÎ~ ÎÎÎ~
~ ~~~~
On the contrary,
E4s
is the solution of a third-orderequation.
It can be shownby perturbation theory
that E4s " E4a +ôE4s
with:ôE4s "
-35~
~ x35~
+..(32)
The first term of
ôE4s
coincides with the value of 2r obtainedby
the chain method in thebeginning
of thischapter.
Theorigin
of the second one is morecomplicated.
General
perturbation
expressions to order 4 and 6 can be obtainedby
use of Messiah's formulas [9]. If one assumes that <0(V(0
>= 0 in Messiah's notation, the 4~~ orderperturbed
energy e4 is trie sum of two terms
(Appendix):
e4 =< o
v~°v~°v~°v
o > < o
v~jv
o>< o
v~°vlo
>a a a a a
The second term contributes
equally
to E4a andE4s(+196v~/27A~).
The first term, which contributesonly
toE4s,
is the lowest order "chain term"(2r
=
-35v~/A~)
obtained inil.l).
On the other hand the 6~~ order energy e6 is the sum of nine terms
iAppendix).
Trie 8~~ and9~~ terms contribute
equally
to E4a and E4s andyield (-2744v~/243A~).
Trie 6~~ and fl~ternis
yield
0. The first five terms contributeonly
toE4s(-175v~ /6A~).
At first
sight,
with this standardperturbation theory,
there does not seem to exist anysimple
rule to derive the 6~~ order terms from the 5~~ order terms, etc.Therefore,
forgeneral S,
thetunneling splitting
obtainedby
the "chain method" willonly
be approximate, sincehigher
orderperturbation
terms will bemissing; also,
even ifVIA
< 1, we will have no way to check the convergence of theperturbation
series when S - cc. Otherapproaches iinstanton
or WKB
methods)
may be moreapproppriate
to thislimit,
butthey
are much more intricate.It is nevertheless
reassuring that,
m the case of a transverse field hiinext chapter),
the chain method leads to formulas for all doubletsiground
andexcited)
which are almost identical to those obtainedby
the WKB method [3]. The interest of the chain method rests in itssimplicity
and
adaptability ito phonon
assistedtunneling problems
forexample il1]).
1.1.5. 7Yansformation
by
Use ofStirling's
Formula. We have foundthat,
to lowest order in),
the tunnelsplitting
of theground
statewas
Es
E«= 2r with:
~
i-1)S-Îx
225-2 ~Îl (~~~ÎÎ!]2
~~~~
This can be transformed with the
help
of theStirling's
formula: n! %/ùn"e~"@
into:r à
fils-ilÎS
xIi ls
~)~Î
~134)
When S goes to
infinity:
ls -1)
ji L)S
~@
~~~~Whence:
4AS~/~
v S
4vS~/~
v s-1
~~~~° ~
l-1)~~~@ ~A) l-1)~~~@ ~A)
~~~~in which AS and
)
must bekept
constantisee
Section 2on "transfer coefficients"
1.1.6.
ibnneling Splitting
of an Excited Doublet + m >im
>0)
When m
= 1 the
degeneracy
is removed to first order invit
r~
v).
When m > 1, it can beshown
by
the chain method that the tunnelsplitting
to lowest order isequal
to2rim)
withito
lowest order in)):
~i~)
"(-1)mÎ122m-2
~im-1
~ijm /1)!j2 ÎÎÉ ÎÎ! i~~)
When m
=
S,
one recovers the previous formula forr(S).
Note that:rjm)
m A x(1)~ j38)
with
(~
< l. When mdecreases,
i e., one climbs towards the top of triebarrier, rim)
A
increases.
In
particular,
if m < Sivery
near the top of thebarrier):
[(
+jj(
~s2m (39)
Then:
~~~~
~(-l)m-ÎÎm
1)!]2ÎÎ
~~~~
When S - cc, AS
= const,
VIA
= const,but, ~(
willfinally
becomelarger
than 1. Then 4~
r
+~ const x S~~~~ will become
comparable
to the distance between doublets (Sz = +m' >and
perturbation theory
will break.On the other
nana,
transformation of thegeneral expression (37) by Stirling's
formula requires that both m > 1 and(S m)
> 1, 1-e-, the doublet + m > isfar,
both from the lowest doublet and from the top of the barrer(mid-height).
Then:~~~~ i-1)~~Î2m~~~
2àre21
1ÎÎ Î~
~ ~Î ~
~~~~
Î~Î)m
Î~~~i~Î
~
ÎÎ~Î -ÎÎ2
~Î
~ ~Î
~~ ~ ~~~~Because of the
assumption
on m, this formula does not reduce to that obtained for m= S.
l,1.7. Remark. Because
ris)
orrim)
are powers:ris)
~fi
m v
(j)~~~
m A(j)~
m A()
~~143)
they
cannotreally
be put into aunique
activation form:= exp
1-
~ with aclearly
T To
kBT*
defined
prefactor 1/To.
1.2. TRANSVERSE FIELD CASE
1.2.1. Ground Doublet. Let us start from trie Hamiltonian:
7i =
-AS)
+h~(S+
+ S-(44)
in which h~
=
-~~~~~
< A. If S is an
integer,
the chain goesthrough (S
= 0 >. If S is 2half-integer,
it goesthrough
+ >. The tunnelsplitting
is Es E«= 2r with
(to
lowest 2order in h
IA)
h2S(~ s)j
h2S~s~
(-A)2S~l[(25
1)!]2(-A)2S~l (25
1)! ~~~~For
integer
S: ôE« = 0,ôEs
= 2P < 0. Forhalf-integer
S:ôE«
= -P andôEs
= +r where P has trie sign ofh(~
=
hl~~~l,
1.e., triesign of h~.
Upon application
ofStirling's
formula:~
(-l)~~~~
x/&e
(25 -1)AÎ (-l)2S~l/&@~
(25 -1)AÎ
~~~~When S goes to
infinity:
Il
~~ l 125
(2s)2s (1_
1)25 (25)25
x e~l25
So the first expression for r becomes:
~~~°°
~(~lÎ2~~~ÎÎ ÎÎÎÎÎ
~~~~These
expressions
are to becompared
with those obtainedby
Korenblit et ai. [Si with thehelp
of a diiferent method. Korenblit et ai. use a Hamiltoman:7i=-DSj-H.S=-DSj-HszcosÙ-H~~+)~~~sinÙ
and with
help
of a recurrencemethod, they
find that to lowest order inH/DS,
the tunnelsplitting
isequal
to(their Eqs. (6), (7)):
~ ~
~~~Î ~ (2/~ l)!
~~~~or, when S - co:
~~
@~~
4DS ~~~~
With the substitution D - A and
(-Hs1I1Ù)/2
-h~,
their A isequal
to oursplitting
2r, except for thesign:
A= -2r.
However,
whatthey really
obtain is A~: see theirequation (A7).
Note that these authorsequally
compute the matrix elements of Sz andS+:
within the symmetric
(I)
and antisymmetric(II)
states when 2HS cos 6 < A(see paragraph
below their
Eq. (16))
within the
perturbed
states ~fiI= (S* > aI1d ~fiII
= S* > wheI12HScos6 > A
(their Eq.
(18) ).
It caI1 be showI1from their results that in this last case, the tunnel eifect isdestroyed by
themagI1etic
field componeI1t aloI1g the easy axis Oz. This will be reexamined in more detailin refereI1ce
il ii.
1.2.2. Excited Doublet + m >
(m
>0).
TheI1:h2m
(s
+m)1
~~~~
=(-A)2m-1((2m 1)ij2
~(s m)1
~°~Since
~~
< l, the tunnel
splittiI1g
iI1creases when one chmbs towards the top of the barrier.A
If m < S, 1-e-, very Ilear the top:
(S
+m)! /(S m)!
£tS~~,
wheI1ce:r(m)
à~i~~m-ii~m i~,i~ l~i~l
~~
(Si)
WheI1 S - co, h~ = const, AS = coI1st =
C,
thereforeh~S/A
~Jh~S~/C
willfiI1ally
becomeforger
thon 1. Thenr(m)
~J coI1st x S~~~~ will
diverge,
andperturbation theory
willbreak,
as
already
seen in the orthorhombic case.On the other
hand,
transformation of thegeneral expression
forr(m) by
use ofStirling's
formula requires that both m » 1 and S m »(mid height
of thebarrier);
one gets that:~
~ ~y~~ 1
/~~~@fij
~~
s + y~~ S+Î
~~~°~
(- l)2m-1
~ 27re2A(2m
1)2 S m ~~~~which does Dot reduce to the result
already
established when m= S. After noticing thon when
m is
large
il/(2m 1)~i~~
-e~/(4m~)~~
this leads to:
1 ~~(y~~ 1
/~)
/~~~ s2 y~~2 ~~ s + y~~ S+ Î~~~~~
(-l)2m-1
~ 7r ~ 4Am2 S m ~~~~This is almost identical to the results of Scl~arf et ai.
([3], Eqs. (1.7, 1.8)
with ~ = h~ and 2~/ =
A,
which con be put into the form:~
2m ~~ j
~ ~ =
~
=
~'Î~
~~s + ~2
~ ~
2 ~ ~
(5~)
~ Tm 7r 8~/m2 2 S + ~ m
Their solution was obtained
by
mapping theproblem
of alarge
quantum spin into apartiale problem,
which is then solvedby
usiI1g the coI1ventional W-K-B- methods for thepartiale problem.
The differences between the two formulas are the sigI1 factor in the first one
(but
in the second one, > 0by definition)
and the presence of S + iI1stead of S inside the bracketsTm 2
of the second oI1e. Scharf et ai. are iI1terested in the hmit S - co, ~/ - 0,
~/S
= constant, but
they
still retaiI1 a= ~/ S + ~
~
as a parameter(see
their discussion afterEq. (2.19)
of[3]).
The fact of
keepiI1g
S + iI1stead of S has thepractical
iI1terest ofavoiding divergences
when 2m = S.
When m » 1, m
# S,
the1/2
can bedropped
in the brackets and the results of Scharf et ai. and of the present work become identical.When m
= S in formula
(54)
derived fromequations (1.7-1.8)
of [3], it becomes:AE~ =
~'~~@@
~~($
~~(55)
~ '~
~~
When S - co:
2S
+ -
(25)~~e~/~
2
~~
whence:
(AESÎ
#(~) 4~/SVÎ ()j
~~= 0.5248 x
4~/SVÎ ()j
~~(56)
~r 'f 'f
while Korenblit and SheI1der [Si, as well as the present
work,
fiI1d:~~~~~ ~~Î~ (~~ÎÎÉI ÎÎÎÎÎ
~~ ~'~~~~ ~(~~ÎÎ~l Î~ÎÎÎ
~~ ~~~~The results are
ideI1tical,
except for the small differeI1ce betweeI1 the two Ilumerical factors 0.52 and 0.56 whicl~might
arise from diiferentapproximations
in the calculations.1.3. TETRAGONAL CASE
It
corresponds
to the case of theferrimagI1etic
cluster "Mn12" with S= 10
il,11,12].
1.3.1. Ground Doublet. Let us start from the HamiltoI1ian:
7i =
-AS)
+~1(S(
+Si
=
-AS)
+ 2~1O((S) (58)
where
OI (S)
is the operatorequivalent
of x~6x~y~
+ y~ and ~1 < A.The
degeneracy
of theground
state doublet (Sz = +S > canonly
be liftedby
theOI
term with selection rule Am= +4, if S is even and S > 2. In the chair computation, we have to
distinguish
the cases S= 4n, where the chain goes
through
(0 >, and S= 4n + 2
(S
>2),
where the chain goes
through
+ 2 >:However, as in the orthorhombic situation, we
finally
obtain a tunnelsplitting:
E~ Ea= 2r
with
(to
lowest order in)):
r -
~~i~~~~
~
~ ~~~ x
~ili~~
x~~~2Sjj,~~ (59)
2
whence,
uponapplication
ofStirling's
formula:~ y
8(s ~)@~-s-2
~s4 ~~~~
(-l)~/~~~Vi
A
)
)~Î
~~~~
~ y
8(
s~)@~-S-2
~~s4 S/2(~l)~~~~~~~@
~~A(S )~Î
~~~~
When S - co:
ls4
S/2 s2 S isS s2)S/2
~
~S
~(S
2)2(S 2) ji
2 S e-2 e-2S
whence:
8~s3/2
~~s2 S/2~~~~7/2
~~s2 S/2-1~S~°°
(_i)S/2-1 fi
(fij
(_i)S/2-1~2 fi
~2~ ~~~~For S = 4n + 2 or
S/2
= 2n + one has(-1)~/~-~
= +l and r cc ~1~"+~; the sign of r is that of ~1. For S
= 4n oI1e has
(-1)S/~~l
= -1 and r cc -~1~" < 0 whatever ~1. The
sigI1of
~lisprobably
related to the characteristics of thetetragoI1al
distortion(compression
orelongatioI1).
~s2 Note that when S
- co, the ratio
~
must bekept
constant(see
Section 2 on "transfer coefficients" ).1.3.2.
illnneling Splitting
of an Excited Doublet (Sz = +m >,(m
> 0,even).
Then(to
lowest order in ~1IA)
WheI1 m
= S we recover trie previous formula for
r(S).
WheI1 m
decreases,1-e-,
oI1e dimbs towards the top of thebarrier, r(m)
increases.If m < S
(very
Ilear the top of thebarrier),
~~ ~~~~
E£S~~
aI1d(S m)1
r(n~)
ç~16A
~s4
m/2(-1)m/2-1 [(q _1)lj2
~ 16A(64)
~s2 ~s4 m/2
wh~~ s _ ~ ~s
= ~~~~t = ~~~~t c
cm/2
sm' ' 16A '
16A~
Therefore
r(m)
~J const x Sm-~ will become
comparable
to trie distance between doubletsand
perturbation theory
will break.On trie other
bond,
for m » 1 and S m » 1(mid height
of triebarrier)
ii ispossible
ta transform triegeneral
expressionby
use ofStirling's formula,
whichyields:
~~~°~ ~
(-
)m/2~~~
(m 2)
4111m ~Î21~~~
~ÎÎ
~~~~~ ~~~~(which
does not reduce to the expression ofr(S)
when m=
S).
1.3.3. Orders of
Magnitude
of tl~e Parameters for tl~e Cluster "Mn12"(S
=
10). Experi- meI1tally
A~J 0.6 K [12]. A very
rough
estimate of ~lis to be fouI1d in refereI1ce [11]: let us start from a tentative value of the siI1gle ion coefficientB( (B(
= 584 in
Abragam-BleaI1ey's notation,
seeEqs. (7.2a, 7.88)
of [7]):10~~
cm~~
<(B((
< 5 x10~~
cm~~(66)
and let us use trie fact that in the
coupled
duster MI112, ~1~JB( /2000
[11]; ibis leads to:o-à x
10~~ cm~~
<(~1( < 2.5
x10~~ cm~~ (67)
wheI1ce a tunnel
splittiI1g
in triegrouI1d
state (Sz = +S >:0.4 x
10~~
Hz < ~~ < 1.26 Hz(68)
in agreement with trie lower hmit of the relaxation lime in trie
high
temperature activatedregime: 1/(2irT)
~J 3 x 10~~ Hz
[12,13].
However a more detailed discussion is necessary [11], because there exist various measuremeI1ttechniques,
aI1d becauseparasitic
eifects(Earth's
magnetic
field forexample)
mayplay
a rote.1.4. TRIGONAL CASE. In tl~at case, the tunnel
splittil~g
arises from a termOI
in tl~e spil~HamiltoniaI1
(for
Rare Earths,OI
could alsoplay
arote). AccordiI1g
to Table 17 ofAbragam
aI1d
BleaI1ey
[7], p.869):
O(
=(Sz(S(
+St)
+(S(
+St )Sz (69)
For
simplicity,
we will work with the HamiltoniaI1:7i=7io+7ii =-AS)+r(Sz(S(+Sf)+(S(+Sf)Sz) (70)
S must be an
iI1teger
and 25 must be amultiple
of 3. Therefore S= 3n, S > 3. Because of the preseI1ce of trie Sz operators, trie chaiI1
computation
is morecomplicated
thaI1 in theprecediI1g
cases. Afteradequate
factorizatioI1s aI1dsimplifications
of trie Ilumerator aI1d of trie deI1omiI1ator, trie final result is:~~"~
~~~~~AÎ~~Î~
1 ~
~25/3
~~)~
i),j2
~~~~3
Note that the denomiI1ator coI1taiI1s
l~ )
iI1stead of ~~l)
!, as could beexpected
ai3 3
first
sight.
After
applying Stirling's formula,
ibis becomes:~~~~~ ~~~~~ ~
~~~Î21~~ e2 /Î~ ))1~~~~
~~~~
When S - co:
1 1
s
3)25/3
~s2S/3~-2
Whence:
~~~~~'~~°° ~~~~~
~~Î~ ÎÎ
~~~~~~~~
Accorjjng
to theproperties
of traI1sfer coelIicieI1ts(see
Section2),
when S - co, AS= coI1st.
~~~
A
~°~~~'
l.5. HEXAGONAL CASE. The creation of a tunnel
splitting by
a term of the form:O(
=(J(
+J~ (74)
2
looks
possible
forf
ions like Rare Earths. Ingeneral however,
thesimple
model based on7io
"-AJ)
is Ilotapplicable
to R-E-(see
discussion in Section1.6). Nevertheless,
there existsa favorable case, that of Tb~+
(J
=6)
inyttrium ethylsulfate,
where the chaiI1 calculation is not Ilecessary, since an exactdiagonalization
in the basis(6s
>, (0 >,(6a
> ispossible
andis indeed
presented by Abragam Bleaney
[7] (§ 5.3,Eqs. (5.18-5.22);
5.6,Eq. (5.56) ).
The tunnelsplitting
A= 0.387 cm-~ has been measured
by
E-P-R- in zero defield,
with an RF fieldparallel
to thecrystal
axis(Abragam Bleaney
[7], 5.6, Tables5.9-5.10).
Thelongitudinal
relaxation lime has also been measured
(Abragam BleaI1ey
[7], 10.5,Fig. 10.5). Although
theIlame "tuI1nel eifect" is non
used,
thisexperiment
isquite
similar to thonperformed
on ferritinby
Awschalom et ai.[14-16].
Another favourable casemight
be thon of Terbium diluted inY(OH)3,
if such acompouI1d
caI1 beprepared (data
of the literature are relative to coI1ceI1tratedTb(OH)3 (17]).
Note thon for terbium in
Y.E.S.,
the eifect of itshyperfiI1e coupliI1g
7in=
Ajjszlz
with itsowI1 nuclear spin 1
=
3/2,
is the same as thon of amagnetic
fieldg~~IBH
=
Ajjm.
In zero externatfield,
there are therefore two tunnelfrequencies given by Abragarn
andBleaney
[7],Eq. (5.56)),
for an effective spiI1 Se~r =1/2:
&w
=
((A(fm)~
+A~ ~~~ (75)
WheI1ce,
in terms of the true spiI1:hw
=
((2AjjmS)~
+A~
)~~~(75')
AccordiI1g
to the captioI1 ofFigure
10.5, thesefrequeI1cies
are:u2 # 14.932 GHz for m
= +
vi " 12.018 GHz for m
= +
These
frequeI1cies
arehigh eI1ough
to avoid aI1y deleterious eifects of other Iludearspins (which
are
probably
rather far fromTb3+).
On the otherhand,
relaxation rate measuremeI1ts were carried oui oI1ly dowI1 to Kr- 2u2 +~ 3ui, which is toohigh
to observe aI1ypossible
saturation of1/T,
ai very low temperaturesIl il.
1.6. SPECIFIC PROBLEMS RAISED BY THE SEARCH FOR TUNNEL EFFECT IN RARE EARTH
COMPOUNDS.
GeI1erally speaking,
ail the terms of thecrystalline
electricpotential
con-tribute to the
eigenvalues
and theeigenfunctions
of the R-E- ion. Therefore ils levels caI1I1ot be describedby
asimple
HamiltoI1ian 7io =-AJ)
aI1d the wave fuI1ctions are mixtures of theform ~fi
=
~j
ojM
(JM
>M
For non-Kramers ions, the
ground
state is notIlecessarily
a doublet.WheI1 the
grouI1d
store is adoublet,
ii is Ilotnecessarily (Jz
= +Jmax > or(Jz
= +Jmjn >:
for
Yb~+(J
=
7/2)
inyttrium ethylsulfate,
it is +3/2
>.When a non-Kramers doublet
corresponds
to an irreducible representation of dimension two of 7io + 7ii, or to twodegeI1erate
oI1e-dimensional irreducible representatioI1s(case
of ther3,4
doublet of
Ho3+(J
=
8)
inLiYF4),
itsdegeI1eracy
caI1only
be liftedby
an externatmagnetic field,
as for Kramers doublets.When the tunnel
splitting
is due to aperpendicular
fieldhi,
in order to have an eifect iI1volviI1g virtual excited states, oI1e must bave gi = 0. If trie eI1ergy levelshappeI1
to be welldescribed
by
7io=
-AJ),
theI1A= 2r
~J
h[~ (for example hf~
for J = là/2 (Dy3+, Er3+),
which is Ilot very convenieI1t for
experiments).
In order to bave moderate
A(< cm~~)
and small admixture of trie excited stores into triegrouI1d
statedoublet,
it is Ilecessary that the excited stores behigh eI1ough (>
100cm~~).
Hyperfine
structure eifects are for frombeiI1g Ilegligible,
withcoupliI1gs
of trie order 10~ -103MHz,
aI1d there are 100%~~~Pr, ~~~Tb,
~~~Ho, ~~~Tm. WheI1available,
isotopes withoutIluclear
spiI1would
bepreferable (Dy, Er, Yb).
Up to now
[7,18],
trieonly
reliable observations of tunnelsplittiI1gs
in R-E- ions seem to be:Tb3+(J
=
6)
inhexagonal
YES: A= 0.387 cm-~
([6,
7] and Section1.5) Tb3+(J
=
6)
intetragonal
LiYF4~ A= 27.98 GHz [8]
in
tetragonal
CaW041 A= 8.1 GHz [8]
in
tetragonal
PbMo04~ A= 15.8 GHz [8]
In other cases, distributions of
crystal
distortions are preseI1t(LaC13~Tb~+ [19],
YV04 aI1d similar matrices [20]). FiI1ally, accordiI1g
to MôssbauerSpectroscopy,
there exist somedyspro-
sium
compouI1ds
with agrouI1d
store + là/2
> [21], but information on trie excited states islackiI1g.
2. Transfer Coefficients
2.1. AIM OF THE CALCULATION. Wl~eI~ many spil~ s; are
coupled
togive
a resultant spiI1S,
trie traI1sfer coelIicieI1ts relate trie terms of trie spiI1 HamiltoI1iaI17i(S)
of triecoupled
system, to trie terms of trie spiI1 HamiltoI1iaI1s
~j7i,(s;)
of trie iI1dividualspiI1s
s,. TraI1sfer coefficients areproportioI1al
ta the reduced matrix1 elemeI1ts of theWigner
Eckart theorem([22], Eq. (5A.1)).
Here,
we wiII be iI1terested in the case of N identicalspiI1s
s aI1d of a maximum resuItaI1t spin S=
~j
s, with S= Ns. This wiII
give
us trie way in which trie parametersA,
h~, u, ~1,1
r, iI1troduced in Section 1, vary, wheI1 S
(1.e., N)
goes ta iI1fiI1ity.The operators
Or
in what follows are defined in agreement withAbragam
andBleaney ([7], Appendix B,
Table16). They
areproportional
to(Tz
+Tpm)
or to(Tz Tp~),
where theTz
are operatorequivalents
of thespherical
harmonics.According
to trieWigI1er
Eckarttheorem,
for giveI1 n, the reduced matrix elemeI1ts of the componentsTz
of an irreducibletensor operator Tn are the same whatever m. Ii is ofteI1 easier ta compute them for m
= 0.
2.2. FIRST-ORDER TERMS
(COUPLING
TO AFIELD).
This case is verysimple:
~h.s;=h.~js,=h.S (76)
; 1
The traI1sfer coefficient
relatiI1g
Ls, to S isequal
to 1 and h must bekept
constant wheI1 S - co.2.3. SECOND-ORDER TERMS
2.3.1. Axial Case. We are interested in the transfer coefficient for:
~~i,
= B
~j3sÎ, 8(s
+i)1 (77)
; ,
in which:
3s), s(s +1)
=
O( (s,),
is a component of an irreducible tensor operator of order 2(O(
= operator
equivalent
of 3z~r~). According
to trieWigner
Eckart theorem:B
£ O((s,)
=
bO((S)
or:;
B
~j[3s(, s(s +1)]
=