Physique
48
(1987)
1663-1670
OCTOBRE
1987,
1663
Classification
Physics Abstracts
75.25
-
75.30
76.50
Solitons in
in
magnetic field :
second
(ER
CNRS
216),
(Reu
le 14 mai 1987,
accept le
24 jccin
1987)
Nous proposons un formalisme de seconde quantification pour les modes de solitons dans les Rsum. dans le cadre chanes antiferromagntiques dIsing de spin s =1/2. Nous donnons les expressions exactes des facteurs de structure dynamique en prsence dun champ magntique, dune approximation un soliton parallle ou perpendiculaire la chane. Nous montrons que le champ magntique donne lieu un ddoublement des modes de solitons. Nous calculons aussi la fonction de corrlation de spin relative au mode antiferromagntique dans le cas de chaines finies ou infinies.
2014 2014 2014
We propose a second quantization formalism for the soliton modes in the s =112 Ising-like antiferromagnetic chain. We give the exact expressions, within the one-soliton approximation, for the dynamical structure factors in presence of a magnetic field, either parallel or perpendicular to the chain. We show that the field gives rise to a splitting of the soliton modes. We also derive the spin correlation function relative to the antiferromagnetic mode for finite and infinite chains. Abstract.
2014
1. Introduction.
Antiferromagnetic Ising-like chains are known to offer good examples of soliton excitations. As shown first by Villain, an exact derivation for the fluctuations induced by the solitons can be obtained in the case of quantum spins s = 1/2 [1]. For classical spins different derivations have been proposed, which all refer to the sine-Gordon model. Recently, Haldane
has established that, in this case, additional internal modes are expected to occur in the solitons [2]. Moreover, the nature of these internal modes should depend on whether the spin value is integer or half integer. In the same line, Affleck has related these predictions to the dyon concept which is used in particle physics. It was also suggested that the application of an external magnetic field in the spin direction might offer a useful way of disentangling the dyon levels [3]. In the present work, we consider explicitely the case of quantum spin chains in an external magnetic field which is applied either parallel or perpendicular to the spins. The second quantization approach that we propose provides a clear understanding of the effects of such a field on the solitons. Rather
different behaviours are actually observed depending on the field direction. In the following, the dynamical structure factors sa ( q, w) (a = x, y, z ) associated with the solitons are calculated within the onesoliton approximation in the full Brillouin zone (- 7T q 7r). Concerning the antiferromagnetic mode near q ir, resulting from the soliton-induced spin flippings in the magnetic sublattices, new statistical arguments are given, which reinforce the phenomenological description proposed by Maki [4].
=
2. Second
quantization.
is the
The
starting point
following
Hamiltonian :
with s, Hi IJ and H_LIJ 1. The first term where J is the exchange coupling describes the Ising energy. It is responsible for the antiferromagnetic ground state with the spins aligned along the z direction (which here coincides with the chain axis). Following Villain [1], we consider chains with an odd number of
1664
spins and cyclic conditions. This situation forces the presence of an odd number of alternance defects or topological solitons in the chains. The diagonalization of the Hamiltonian in equation (1) has been performed by Shiba and Adachi [5] in order to 0. Our calculate the dynamical susceptibility at q second quantization approach allows an extension of this work to any value of q. LetIp) (p half integer) be the state with one soliton between sites p - 1/2 and p + 1/2. In the subspace with one soliton, the matrix elements of the spin operators are :
=
Similarly,
the Hamiltonian
can
be rewritten
as :
The
With
are
to
a
no magnetic field, the k) andI 7T - k) states degenerate and the Brillouin zone can be reduced - 2r k -- - 2,T. The effect of a magnetic field is g
degeneracy. As a result, one obtains of the soliton energy in two branches as splitting shown in figure 2.
seen
to lift this
The phase convention chosen in equation (2) is such that the soliton is positive (extra spin up) for p - 1/2 odd and negative (extra spin down) for p - 1/2 even (see Fig. 1). We now introduce cp and
Fig.
1. soliton
(b)
and
state
(a), positive
Fig.
a
2.
without
parallel
Band-structure of the one-soliton excitations magnetic field (a), and with a magnetic field (b) and perpendicular (c) to the chain.
given by :
3. Parallel field. The
spin operators
are
expressed
as :
(Hl
0),
the
and
evaluated to be :
where
(a = - I )
states
correspond
to a combination of
1665
k + ) states. Moreover, the energy splitting is independent of the wave vector k (Fig. 2b). Using equations (3) and (8), the spins operators are now given by :
Thus,
obtains for H parallel, a simple Zeeman splitting favoring the negative solitons in the ( k - ) states with respect to the positive solitons in the
one
and the
as :
spin
correlation functions
can
be
expressed
where nku defines the thermal population of the I ku) state. It can be expressed as nku ns P kU/ ZI , where ns ~ exp (- (3J) is the soliton density and :
=
In the last equation, Io (x ) is the zero-order modified Bessel function. The dynamical structure factors SI(q, ( ) are obtained by Fourier transformation of equation (10). After an integration over k, one gets :
two energy branches are
are
thermal
weighting
factors which
are
given by :
thermal factors are ns(1 - ns) and B (q) 0. At lower temperature, ns can be neglected compared to 1 and for Hjj 0, one recovers the results previously given by Villain for the z-component [1] and by Nagler et al. for the transverse spin component [6]. Finally, at very low temperature (kT .c Hi, EJ), only the lower states of the lower band are occupied : nku = ns 5 k, ,,, /2 8 u, - and equations (11) are replaced by &-functions which can be easily derived from
=
equations (10).
All the dynamical structure factors in equation (11) exhibit a square root divergence. This corresponds to the nesting condition characteristic of one-dimensional bands. It is also worthnoting that Sf (q, w ) is unaffected by the external magnetic
1666
field. This is expected as this function corresponds to transitions inside each energy band of figure 2b. On the opposite, for the transverse spin fluctuations which describe transitions from one band to the other, the position of the divergence is shifted by the (12 q + HI) with = interband splitting : This results in a doubling of the soliton modes, 1. as shown in figure 3 where we have drawn the function :
f2 q 6
w (i.e. for 0 and q notices that, at q f2q 0), the spreaded distribution of frequencies merges into &-functions. In particular the ESR mode Sr(q 0, w ) (a = x or y), which corresponds to vertical transitions between the two subbands in figure 2b is peaked at the conventional Larmor frequency:
= = =
a) =
HII -
4. In
Perpendicular field.
perpendicular field (HII = 0), the Hamiltonian (Eq. (4) can be diagonalized in the initialk) basis :
with - 7T -- k 7r. Again, two distinct energy branches are obtained in the reduced Brillouin zone as shown in figure 2c. The states in the upper branch
in the magnetic field (a) magnetic parallel (b) and perpendicular (c) to the chain. The following parameters have been used : J 100 K, T 25 K, EJ 10 K. The wavevectors are labelled in reciprocal lattice unit : q q/7r.
Fig.
3. absence of
-
Dynamical
structure factor
Sy(q, w )
a
=
field
and with
equation (13) ; they can be viewed as triplet states (positive combination of adjacent positive and negative solitons), while the lower branch k equation (13)) can be considered as singlet states. Using equation (3), the correlation functions are expressed as :
correspond
to
kI
7r
in
:> 7r in
where
Exact calculations of the dynamical structure factors can also be performed (see Appendix). However, their expressions are too complicated to allow a simple formulation. As an example, the function Sy L (q, has been calculated; it is shown in figure 3c for H 1. &/. A doubling of the soliton mode is also observed, which, however differs drastically from the case of parallel field. The position of the square root singularities can be expressed (see Appendix) as a function of the
=
scattering
wave
vector q
as :
singularities are obtained which correspond to 7y 1 in equation (15) ; however, for hq > 1, there are only two singularities obtained for Ty + 1. This remarkable result is displayed in figure 4 where the dispersion relation of the singularities of Sy L (q, l) and S L (q, w ) are shown for different values of H_Ll eJ. The twice degenerated singularity line for H 1. 0 is splitted into two distinct lines for H1. 0. While a gap with frequency w 0 for one line, the second 2 H 1. is open at q line is now restricted to the interval comprised between q 2 sin-1 (H1./8 EJ) and q ir. With H1. increasing, this second line is reduced to a
For
h -- 1,
=
four
>
1667
by
as
independent
random
events. The z-component of the spin at site n is fixed by two conditions : i) the parity of n and ii) the total
number of solitons on the left hand side of the site n. The corresponding correlation function can be written as :
equation, we have used (- )N ei7TN and we have defined AN (n, t) N (n, t ) - N (0, 0), where N (n, t ) is the operator which describes the number
In this
=
=
of solitons present at time t between the left end of the chain and the site n. This number is the result of a random process, which accounts for the presence or the absence of solitons in a given k) state. At any time, AN can be considered as the sum of two terms AN AN + + AN - , corresponding to solitons which give positive (+ 1 ) and negative (-1 ) contributions. In equation (17), one can actually replace this sum by the difference AN N+ - N_, which remains a positive quantity. The correlation function appears therefore to be the characteristic function (or the Fourier transform) 0 (n, t ; u ) of the probafor u 7r : bility distribution p
=
(AN )
Fig.
Dispersion relation of the singularities of and Sz L (q, - ) for different values of the perpendicular field : H 1.. 0 (full line), H 1.. 2 e7 (dotted line), H 1.. 4 EJ (dotted-broken line) and H 1.. 8 sJ (broken line). The singularity lines for S(q, w ) are obtained by changing q in 7r - q.
4.
-
Sl (q, w )
=
In order to evaluate the moments of this distributions, AN is expressed as a function of the creation-annihilation operators :
smaller and smaller zone near q ir, where it condensates for H.1. 8 EJ. For larger field, only the first singularity line remains present. A similar behaviour is observed for the function Sl (q, m ) by changing q in zr - q. For this function, the gap 7r and the condensation point at occurs at q 0. Finally, one notices that, contrary to the q parallel field case, the ESR modes Sr (q 0, w ) and Sz L (q 0, w ) consist of a continuum of resonance frequencies and present square-root divergences at the somewhat unusual values : co = 2 H.1. which actually correspond to the nesting condition for vertical transitions in figure 2c.
= =
infinite system, the second term in the parenthesis occuring in equation (19) can be dropped out since it gives vanishing contribution when integrated over K if N -+ oo. As the contributions of the different k) states are statistically independent, the first two moments of are given by :
an
In
p (,N)
5.
Antiferromagnetic
mode.
The one-soliton approximation becomes insufficient to describe the fluctuations in the z direction near the antiferromagnetic point q 7T, as evidenced by
=
factor in equations (10) and divergent 2 (14). To go beyond this approximation, we calculate the fluctuations resulting from the flipping induced
the
1/cos2 1 q
1668
one
obtains :
where
approximations : 1- nk + x =1 and Ek+K - Ek = vk K. The minimum function Min (x, y ) results from an exact integration over K [11]. To account for a zero first moment and a finite second moment, we can tentatively assume a Gaussian distribution for AN(0, t ). The corresponding characteristic function is :
we
have
again
made the
time, the time dependence of the correlation function is the same for finite and infinite chains. However, at long times, it goes to a constant
At short
In the last step of equation (21), we have made the approximations 1nk , K == 1 and Ek + x - Ek Vk K. Otherwise the integration over K is exact [7]. It appears therefore that the two first moments of the distribution are equal. This result agrees formally with p the assumption of a Poisson distribution and justifies in the quantum case, the description first proposed by Krumhansl and Schrieffer [8] for solitons in the classical 04 model. Since the characteristic function associated with the Poisson distribution is 0 (u)
where vx
dEx/dk.
in finite chains :
(AN)
on the average number of solitons ns l located between the end of the chain and the considered site. Of course, in real systems, it would be necessary to take into account the random distribution of the segment lengths [12].
depends
6. Conclusion.
exp[M1(eiu -1)],
we
obtain :
This
expression was first given by Maki [4], but not explicitely demonstrated. It can be viewed as an
extension to this quantum s = 1/2 case of the soliton gas model preiously used for classical s = oo sineGordon solitons [9, 10]. In this derivation, the effect of the magnetic field has not been considered. However, it is easy to realize that the only effects are related to changes in the thermal population nk and soliton velocity vk in perpendicular field. Finally, we consider the case of finite chains. The second term in parenthesis of equation (19) can no more be dropped out. It follows that the timedependent contribution to the first moment van-
the fact that the k) states are no more progressive, but stationnary waves. To simplify, we consider only the autocorrelation function (n 0), which is relevant for neutron spin echo and NMR techniques. The related second moment is given by :
=
The present description in terms of creation-annihilation soliton operators provides a comprehensive picture of the effects of an external magnetic field on the soliton excitations in Ising-like quantum spin chains. New specific features have been established for both field parallel and perpendicular to the spins. For parallel field, an additional Zeeman splitting is found. For perpendicular field, one is led to a simultaneous change of both the energy and the velocity of the solitons. The corresponding dynamical structure factors have been calculated in the onesoliton model. It is known that this approximation is quite insufficient to describe realistic spin systems [1, 10, 13]. Interactions between solitons, between solitons and magnons and with impurities are expected to round off the square-root singularities [14]. However, the main feature - the doubling of the soliton modes - established by our formulation would be maintained. The lineshapes displayed in figure 3 have been evaluated for the physical parameters J = 100 K, T = 25 K and EJ = H =10 K, which are realistic values for the compounds CsCoCl3 and CsCoBr3. For these compounds the doubling of the soliton modes should be accessible experimentally. In principle, neutron scattering measurements would permit a direct observation of the S"(q, w ) in the full Brillouin zone [6, 14-16]. Alternatively, the magnetic resonance techniques can provide a check of the Zeeman splitting and of the gaps of the singularity lines expected at q 0, for HII and H1- respectively [17, 18]. The present results have been obtained in the case of quantum spin chains. We may wonder if they can be extrapolated to classical spin chains. The Zeeman splitting obtained for Hp might be the analog of the dyon levels described by Affleck. It would remain to
=
1669
find the classical analog of the soliton doubling for H 1.. Similar calculations in Ising-like classical spin chains are highly desirable. Finally, the present work which concerns a spin 1/2 system can be a first contribution to the crucial question concerning the different behaviours expected for antiferromagnetic chains with integer and half-integer spins.
dynamical
with ki being
the solutions of
f (k )
0. This
gives :
factors ; it is, however, easy to have them calculated with a computer by replacing ki in equation (A.1) with the solutions obtained from
ture
equation (A.3).
as
On the other hand, the singularities are obtained the poles of equation (A.1) which are given by :
ki (again ki has been changed into 1 7T - k for S"). Among the two existing solutions :
with
v
=
sin
into Ir k,.
an
for Sx and
taking
unique equation :
with 71 1, both are valid for h -- 1, but only that with = + 1 is correct for hq > 1 because of the conditionI v ,1. The general expression for the singularity frequencies (Eq. (15)) immediately follows using equation (A.2). Moreover, it can be checked by making a development near d2 that the singularities are of square root type as in parallel field.
=
where
defined by standart equations (16). methods, but only the solutionsI u I -- 1 should be retained. As the solutions of this fourth degree equation are quite complicated, we will not give the explicit expressions for the resulting dynamical strucn
=
( /2 n q
It
can
and be
hq
are
solved
by
References
VILLAIN, J., Physica B 79 (1975) 1. HALDANE, F. D. M., Phys. Rev. Lett. 50 (1983) 1153. AFFLECK, I., Phys. Rev. Lett. 57 (1986) 1048. MAKI, K., Phys. Rev. B 24 (1981) 335. SHIBA, H. and ADACHI, K., J. Phys. Soc. Jpn 50
(1981)
3278.
S. E., BUYER, W. J. L., ARMSTRONG, A. L. and BRIAT, B., Phys. Rev. B 28 (1983) 3873. GRADSHTEYN, I. S. and RYZHIK, I. M., Table of Integrals, series and products (Academic Press, New York) 1965, p. 371.
1670
[8] KRUMHANSL,
Rev. B 11
Phys.
[14] BOUCHER,
[9] MIKESKA, H. J., J. Phys. C 13 (1980) 2913. [10] MAKI, K., J. Low Temp. Phys. 41 (1980) 327. [11] Reference [7], p. 451. [12] NAGLER, S. E., BUYERS, W. J. L., ARMSTRONG, R. L. and RITCHIE, R. A., J. Phys. C 17 (1984)
4819.
REGNAULT, L. P., ROSSAT-MIGHENRY, Y., BOUILLOT, J. and STIRG., Phys. Rev. B 31 (1985) 3015. [15] YOSHIZAWA, H., HIRAKAWA, K., SATIJA, S. K. and SHIRANE, G., Phys. Rev. B 23 (1981) 2298. BUYERS, W. J. L., HOGAN, M. J., ARMSTRONG, [16] R. L. and BRIAT, B., Phys. Rev. B 33 (1986)
NOD, J., LING, W.
J. P.,
1727.
[13] SAZAKI,
257.
Phys.
Rev. B 35
(1987)
[17] ADACHI, K.,J. Phys. Soc. Jpn 50 (1981) 3904. [18] BOUCHER, J. P., RIUS, G. and HENRY,Y., to appear in Europhys. Lett.