Quenched charge disorder in CuO2 spin chains: Experimental and numerical studies

Quenched charge disorder in Cu02 spin chains: Experimental and numerical studies

R. Leidl,^ R. Klingeler,^'^ B. Buchner,^ M. Holtschneider/ and W. Selke^

^ Institut fiir Theoretische Physik, Rheinisch- Westfdlische Technische Hochschule Aachen, 52056 Aachen, Germany
^ Laboratoire National des Champs Magnetiques Pulses, 31432 Toulouse, France
' Leihmz Institute for Solid State and Materials Research IFW Dresden, 01171 Dresden, Germany

(Dated: May 8, 2006)

We report on measurements of the magnetic response of the anisotropic Cu02 spin chains in
lightly hole-doped Lai,(Ca,Sr)i4_i,Cu2404i, x > 5. The experimental data suggest that in magnetic
fields B > 4T (applied along the easy axis) the system is characterized by short-range spin order and
quasi-static (quenched) charge disorder. The magnetic susceptibility x{B) shows a broad anomaly,
which we interpret as the remnant of a spin-flop transition. To corroborate this idea, we present
Monte Carlo simulations of a classical, anisotropic Heisenberg model with randomly distributed,
static holes. Our numerical results clearly show that the spin-flop transition of the pure model
(without holes) is destroyed and smeared out due to the disorder introduced by the quasi-static
holes. Both the numerically calculated susceptibility curves x{B) and the temperature dependence
of the position of the anomaly are in qualitative agreement with the experimental data.

PACS numbers: 74.72.Dn, 75.25.-fz, 75.10.Hk, 05.10.Ln

I. INTRODUCTION

The tendency of charge carriers for self-organization
seems to be an intrinsic property of hole-doped transi-
tion metal oxides. One remarkable example which em-
phasizes the interplay of charge order and antiferromag-
netism is the formation of spatial spin and charge modu-
lations in the high- Tc cuprates.^ Other examples for self-
organization of holes in low-dimensional magnets include
the layered nickelates^^ and manganites^, and the doped
Cu02 spin chain systems such as Sri4_a;Caa;Cu2404i and
Nai+a;Cu02i^'S^ On the other hand, these observa-
tions also suggest that quenched disorder plays an im-
portant role in such systems. In the case of the half-
doped manganites, the quenched structural A-site dis-
order was found to enhance the fluctuation of the com-
peting order parameters, i.e., between the charge /orbital
order and the metallic ferromagnetism.^ The example of
the manganites shows that, in the case of competing
phases, quenched disorder can lead to properties that
are very different from those of a slightly impure mate-
rial.^'' Quenched disorder can also significantly affect the
properties of hole-doped layered cupratesi^i E.g., recent
numerical results suggest that disorder effects are impor-
tant to describe the underdoped regime of the layered
cuprates and the pseudogap in these compounds

In this paper, we report on experimental and numeri-
cal studies of the magnetic response of a cuprate model
system, i.e. the lightly hole-doped Cu02 spin chains in
Laa;(Ca,Sr)i4_2.Cu2404i, with a; > 5. In these com-
pounds, two quasi-one-dimensional (ID) magnetic struc-
tures are realized: CU2O3 spin ladders and Cu02 spin
chains. The former exhibit a large spin gap of Agap ^
400K^ Hence the magnetic response at low tempera-
ture, which is the subject of our study, is due to the
chains. The chains consist of edge-sharing CUO4 plaquet-
tes containing Cu^+ ions with spin 5* = 1/2 and non-
magnetic Zhang-Rice singlets. The concentration of holes

in the spin chains amounts to less than 10% and the Cu
spins in the hole-free chain segments form predominantly
FM fragments since the nearest-neighbor (NN) cou-
pling is ferromagnetic. The NN coupling is anisotropic,
thereby causing an uniaxial anisotropy perpendicular to
the Cu04-plaquettes, i.e. along the crystallographic b
axisii^^i^ In contrast, the magnetic coupling of Cu spins
via a hole is antiferromagnetic (AFM) , as is known from
a comparison with the strongly (i.e., 60%) hole-doped
spin chains of the mother compound Sri4Cu2404iiiSiii
Moreover, there is a finite interchain coupling causing
3D AFM spin order below Tn ~ lOKiiSiia In previous
papers we have argued that the spin ordered phase at
zero magnetic field is presumably also characterized by a
(short-range) charge order External magnetic fields
of the order of a few Tesla suppress the long-range spin
order when applied along the easy axis and cause a short-
range antiferromagnetically spin ordered and charge dis-
ordered phase In the present paper we concentrate
on the properties of the intermediate field phase at sev-
eral Tesla, which is characterized by (i) short-range AFM
spin order, and (ii) quasi-static charge disorder.

II. MOTIVATION OF THE MODEL

As was shown previously, the melting of long-range
AFM spin order at a field B = Bi (depending on
temperature) causes an anomaly in the magnetization
M{B)m^^ This is demonstrated by Fig.QjL, which shows
the magnetization M{B), at fixed temperature T = 2.5K,
of La5.2Ca8.8Cu2404i. If the magnetic field is oriented
parallel to the chain direction, i.e. B\\c, the magnetization
depends linearly on B, except for a small contribution
of free spins. In contrast, two anomalies are observed in
M{B\\b), which become clearly visible if the susceptibility
X = dM/dB in Fig.Hb is considered. At Bi = 3.75T one
recognizes a sharp peak which is attributed to the melt-

2

ing of the long-range spin order. Hence, this anomaly
signals a transition from a spin and (probably short-
ranged) charge ordered phase for low fields B\\b < Bi
into a charge disordered state for B > Bi.

Based on the suggestions described in Ref. various
theoretical studies have been devoted to the phenomena
at B = Bi ,20,23^24,25 ^herc the magnetic degrees of free-
dom were described by Ising spins and the holes were
assumed to move either freely along the chains or under
the influence of a periodic pinning potential stabilizing a
striped structure. These models predict a breakdown of
the striped (charge ordered) phase and may thus explain
the transition a,t B = Bi.

Our present study, however, focuses on the proper-
ties for B > Bi. Previous numerical and experimen-
tal work^SiSLSS implies that this phase is characterized
by short-range AFM spin correlations, and quasi-static
(quenched) charge disorder. The data in Fig. ^ display
an additional broad peak in X: at i?2 = 6.9T > Bi,
which was not captured by the previous theoretical stud-
ies. We attribute this anomaly to the reorientation of the
Cu spins. The idea that the anomaly at B2 is in fact a
"smoothened out" spin-flop transition will be explored in
greater detail in the next section, where we present the
results of our Monte Carlo simulations. We propose that
the anomaly at B = B2 is the relic of a spin-flop tran-
sition, which is smeared out due to the strong disorder
induced by the quasi-static holes. In this sense one may
call the anomaly a "pseudo" spin-flop peak.

In the scenario of the "pseudo" spin-flop transition,
the magnetic field overcomes, for B > B2, the uniaxial
anisotropy whic h is du e to the nearly 90° Cu-O-Cu ex-
change (cf. Refs. ll4ll5|) . Quantitatively, the experimental
value of B2 is consistent with a recent inelastic neutron
scattering study on La5Ca9Cu2404ir& which reported a
spin gap of Agap/ (g/^s) = (7 ± 0.5)T.

III. NUMERICAL SIMULATIONS

A. Definition of the model, choice of interaction
parameters, and simulation method

Taking the scenario of quenched charge disorder and
short-range AFM spin order for B > Bi as our start-
ing point, we adopt a complementary view to the previ-
ous studies which considered mobile charge carriers and
study the influence of quenched charge disorder on the
magnetic properties of the system, ignoring the mobility
of the holes altogether.

We consider a LxL square lattice consisting of L rows,
which we identify with the chains, and L sites per chain.
This choice of lattice geometry is motivated by neutron
scattering experiments indicating that the copper ions in
the Cu02 planes of La5Cai4Cu2404i form a rectangular
arrayiSi We conveniently set the lattice constants along
and perpendicular to the chains equal to unity. Periodic
boundary conditions are employed throughout.

3

B(T)

FIG. 1: Magnetization M (a), and susceptibility % ~ dM/dB
(b), of La5.2Cag.8Cu2404i, at T = 2.5K, vs. magnetic field B
parallel to the 6- and to the c-axis, respectively.™ The data
are corrected by the (/-factor taken from Ref. Ha. In (a), the
small, isotropic contribution due to free defect spins (dashed
curve) has been subtracted, see Ref. 22;, The vertical dashed
line shows the phase boundary between long-range and short-
range antiferromagnetic spin order.

Each site (?, j), where i is the chain index and j labels
the sites along the chain, is either occupied by a spin
(representing a magnetic Cu^^ ion), or a non-magnetic
hole (Zhang- Rice singlet). To describe the hole distribu-
tion, we introduce random variables pij taking the values
Pij = 1 if a spin resides at lattice site (i, j) and pi^j = if
it is occupied by a hole. The spins are modeled by (classi-
cal) three-component vectors Sij = {Sfj,Sfj,Sfj) with

\Sij \ = 1. As discussed in Ref. I2M we expect our results
to remain qualitatively correct if one took the quantum
character of the spins properly into account (although
there would be, of course, quantitative deviations). This
is basically a consequence of the Ising-like anisotropy of
the model which tends to suppress quantum fluctuations,
in particular in the presence of a field applied along the
easy axis as in our case. If a hole is at site (i,j), we set
Si,j — 0. We simulate either the pure system without
holes {pij — 1 for all or employ a fixed hole con-
centration of 10% within each chain. The latter should
resemble the situation in the lightly hole-doped chain sys-
tems of La2;Cai4_2:Cu2404i with cc ~ 5. Moreover, we

3

disallow nearest-neighbor pairs of holes within the same
chain, since such configurations are energetically unfa-
vorable due to the strong Coulomb repulsion. Thus con-
secutive holes along the chains are always separated by
at least one spin.

The configurational energy depends both on the spin
variables {Sij} and the hole distribution described by the
occupation variables {pi,j}- In a field H applied along the
z-axis, the Hamiltonian of our model reads:

Ti. — — Jci (Sij ■ Sij+i + A S^ jSi j^i ] — Jc2 Sij ■ Sij+2 — Jo ^^(1 ~ Pij+i)Sij ■ Sij+2

1,3

hJ

(1)

The interactions of this model are shown schematically
in Fig. [3 The Cu-O-Cu bonding angle of nearly 90° sug-
gests that the nearest-neighbor (NN) intrachain coupling
Jci is ferromagnetic (Jci > 0). Moreover, this coupfing
is anisotropic favoring the alignment of the spins
along an easy axis (the crystallographic 6-axis), which
we take to be the z-axis. The anisotropy parameter is
A > 0, where A = corresponds to the isotropic case.
In Cu02 spin chains one expects next-nearest-neighbor
(NNN) spins to be coupled antiferromagneticallyiSS The
coupling between NNN spins is Jc2 < if they are sepa-
rated by a spin and Jq < if a hole resides between them,
where \Jq\ > \ Jc2 \ ■ The difference is mainly caused by the
smaller size of the Cu'^'*' ions leading to a stronger overlap
of the p-orbitals of the involved oxygen ions. Finally, in
accordance with the experimentally observed long-range
AFM ordering, we assume an interchain coupling <
between NN spins on adjacent chains.

In the following, we turn to the values of the interac-
tion parameters of Eq. To the best of our knowledge,
no direct information on the magnetic coupling param-
eters is available which agrees with our thermodynamic
studies;^ Thus, for the numerical simulations we indi-
rectly estimate the interaction parameters. First, we ad-
dress the NNN interaction across a hole. This parameter
has been determined to be Jo/fcs = — 130K from neu-
tron diffraction studies of the stoichiometric compound
Sri4Cu2404iiiSi^ In contrast to Jq, only little is known
about Jci and Jc2- Qualitatively, the bonding geometry
suggests Jci > and Jc2 < oMi^Si^ In addition, the
presence of ferromagnetic spin order along the chains in
the weakly hole-doped compounds implies |Jci| > |Jc2|-
In order to get a quantitative estimation, we apply the ex-
change parameters which have been extracted for the un-
doped Cu02 spin chains in Li2Cu02. In this compound,
the Cu-O-Cu bonding angles in the Cu02 spin chains are
very similar to those of (Ca,Sr)i4_a;La2:Cu2404i. Thus,
we take Jci/ks = lOOKMiMi^ Moreover, we assume
Jc2/kB — — 35K, which is again estimated from a com-
parison with Li2Cu02 by applying the bond- valence sum
rule and the pressure dependence of Jq in Sri4Cu2404i.
For Sri4Cu2404i, the latter amounts to dJo/dpc —

FIG. 2: Schematic representation of the interactions of our
model Hamiltonian, Eq. Q. Full and open circles denote
Cu and O atoms, respectively. Nearest-neighbor (NN) spins
along the Cu02 chains (c direction) interact via an anisotropic
ferromagnetic exchange (Jci, Jci = (1 + A) Jd), whereas next-
nearest-neighbor (NNN) spins are coupled antiferromagneti-
cally. The strength of the antiferromagnetic coupling depends
on whether the NNN spins are separated by a hole (Jo) or by
another spin (Jc2). Finally, neighboring chains have an AFM
interaction (Ja).

4.2K/GPa,

17

The magnitude of Ja significantly affects the zero-field
ordering temperature of the model (without holes). In
Sri4Cu2404i, the coupling between Cu spins in adjacent
chains amounts to Jq ~ — 20K.^» However, preliminary
diffraction experiments^^ suggest changes of the relative
positions of neighboring Cu02 chains upon La-doping,
which are supposed to strongly affect the interchain cou-
pling constant. According to a recent result on undoped

4

spin chains in Ref. |32

{=)3D

eiD

CW ~ ^cff-

Ja

(2)

where the 3D Curie- Weiss temperature may be estimated
as 6cw — ~8K from a fit to high-temperature suscep-
tibihty dataj?^ while for the ID Curie- Weiss tempera-
ture one has 6cw — 0.23Jci = — 23K from a cluster
calculationiS Using the (approximate) effective number
Zeti = 2 of nearest neighbors at surrounding chains, one
gets a slightly larger value Ja — — 30K as compared
to Sri4Cu2404i. This estimate, however, intimately de-
pends on Zoff, which might be different. In the following,
we set Ja/fcs = — 25K, since this gives rise to a zero-field
ordering temperature which appears to be quite reason-
able as compared to the experiments (see the discussion
in Sec HTTBl .

From the interaction parameters Jd, Jc2j f^nd J^, and
a fit to the experimentally determined spin-wave gap of
Ref. 13 we can calculate the anisotropy parameter A.
This yields A = 0.0255.

The model is simulated employing a single-spin Met-
ropolis algorithm. System sizes range from L = 20 to
L = 240. To obtain good equilibrium data, up to 2 x lO*"
Monte Carlo steps per site are needed for the largest sys-
tems. At the beginning of each run, 20% of the steps are
discarded for thermalization. For the system with holes,
we average over up to 300 randomly generated realiza-
tions of the disorder.

A quantity of primary concern due to its relation to
the experiments is the magnetic susceptibility x^.

1

ksTL^

(3)

. .) denotes the thermal average and =
J is the z-component of the total magnetization.

where

Other observables of interest include the specific heat and
the staggered magnetization. We also record typical spin
configurations generated during the Monte Carlo runs in
order to monitor directly microscopic properties of the
system.

dicular to the chains, we define the quantity

2

1 ^ / 1 ^ \

j=i \ i=i /

(4)

Note that the expression within the parentheses is the
staggered magnetization (per spin) of one column of the
square lattice. We cannot simply take the difference be-
tween the total magnetizations of even and odd rows,
which would be a natural candidate for the AFM order
parameter, since the usual AFM structure is modified
by a helical ordering of the spins along the chains, as
explained below. Thus the total magnetization of each
chain vanishes for H = 0. In the fully (antifcrromagnct-
ically) ordered state, one has = 1.

From our simulational data (Fig. (S} we infer that a
phase with long-range AFM order exists at low temper-
atures. The interchain order parameter seems to
vanish continuously at a Neel temperature T/v , which we
estimate as fcsT/v/l Ja| ~ 0.61 by finite-size extrapolation
of our data. This value is also obtained by analyzing the
peak positions of the specific heat.

0-0L=

:10

L=

:20

:40

:80

L=

:160

k„T / 1 J I

B a

B. Pure (undoped) system

FIG. 3: Interchain order parameter, Eq. JIJ, of the pure
model vs. temperature (at _ff = 0), for different system sizes
L.

Before studying the infiuence of the disorder, it is in-
structive to review some basic properties of the pure sys-
tem without holes. In the present context, the most rele-
vant features of the pure model are (i) the existence, for
low temperatures and fields, of a phase with long-range
AFM order perpendicular to the chains, as well as (ii)
the occurrence of a spin-flop transition upon applying a
magnetic field along the easy axis (i.e., the z-axis).

Let us first discuss the model in zero magnetic field
{H = 0). To measure the long-range AFM order perpen-

Taking the spin S = 1/2 of the Cu^+ ions into account
we obtain an estimate of T/v — 13K for the Neel tem-
perature at zero field, which is reasonably close to the
experimental value for La5Ca9Cu2404i (T/v = 10.5K).i^
However, this result has to be taken with care and should
only be regarded as a rough consistency check. To men-
tion just two points, the real system is not hole-free and
quantum fluctuations (absent in our classical spin model)
certainly alter the value of the ordering temperature (cf.
Ref.EF

5

Whereas neighboring spins on adjacent chains are
aligned antiferromagnetically for T < T/v, the spins
within the chains exhibit a more compHcated structure
due to the competing intrachain interactions Jd >
and Jc2 < 0. At T = 0, this structure can be found
by a ground-state analysis using (and slightly generaliz-
ing) the methods described in Refs. and W?. With-
out anisotropy (A = 0), one would obtain a simple he-
lical ordering within each chain. In that case the spins
rotate, with a constant angle a between two consecu-
tive spins, within a plane whose orientation is fixed in
space. A straightforward calculation yields a ~ 44°, cor-
responding to a wavelength of the helix of approximately
eight lattice constants. The finite exchange anisotropy
A > 0, however, modifies this structure. In order to
minimize the anisotropy energy, the spins rotate in a
plane that contains the z-axis (without anisotropy the
orientation of the plane is arbitrary). Moreover, the ro-
tation angle is not a constant, but somewhat smaller for
spins in the vicinity of the z-axis. The wavelength of the
modified helix, though, changes only little as compared
to the isotropic case. Such a configuration is depicted
schematically in Fig.^. The results of the ground-state
analysis are corroborated by inspection of typical low-
temperature Monte Carlo configurations. We can unam-
biguously identify the type of helical order shown in Fig.

The wavelength of the helix turns out to depend only
weakly on temperature.

FIG. 4: Schematic representation of the (intrachain) spin
configurations for (a) H = 0, (b) < H < H^i, and (c)
Hsf < H < //pm, where Haf is the spin-flop field and -ffpm the
field of the spin-fiop to paramagnetic transition.

A magnetic field H > along the easy axis lifts the
symmetry between the positive and negative z-direction
and the system develops a finite total magnetization
along the z-axis. At T = 0, one can again find the cor-
responding spin configurations by a ground-state analy-
sis. For small fields H < iJgf, where Hgi is the spin-flop
field (see below), the analysis yields a fan- like structure
(Fig. 2)3). At H — Hsi, this structure becomes unstable
against a spin-fiop phase where the spins make a finite
angle with the z axis and rotate on the surface of a cone
(see Fig.^J;). All spins now have the same z-component,
but the X- and y-components are reversed for nearest-
neighbor spins on adjacent chains. This justifies calling
the structure a "spin-flop phase". At H = Hsf, the z-
component of the magnetization (and various other quan-
tities) exhibit a discontinuity. For the parameters of our
model, the value of the zero-temperature spin-fiop field
is given by Hsf/\Ja\ — 0.70. Upon further increasing the

field the opening angle of the cone continuously shrinks
to zero until ai H = iJpm all spins point along the z-axis.
However, this transition from the spin-flop to the param-
agnetic phase occurs at values of the magnetic field much
larger than the highest fields used in the experiments and
will therefore be disregarded in the following.

The above spin structures in a magnetic field H >
can again be found in our finite-temperature Monte Carlo
configurations. Moreover, we observe a sharp peak in
the susceptibility (Fig- Eld which occurs at a field
quite close to the value of the spin-flop field at T = 0,
Hsf/\Ja\ = 0.70 (see above). Apparently, the peak signals
the (presumably first-order) transition towards the spin-
fiop phase. Similar anomalies are found in other quanti-
ties. The spin-flop field Hgf is only weakly temperature-
dependent (for low temperatures).

FIG. 5: Magnetic susceptibility of the pure model at fixed
temperature kBT/\Ja\ = 0.25 for a system of size L = 80.
Note the sharp spin-flop peak at Hsi/\Ja\ — 0.72.

A more detailed investigation of the phase diagram of
the pure model, the nature of the various transition lines,
and the behavior near possible critical and multicritical
points, where the different phases eventually meet, would
certainly be of interest by its own but is beyond the scope
of the present paper.

C. Influence of random, immobile holes

The introduction of randomly distributed, immobile
holes has a drastic impact on the properties of the model.
Any long-range spin order (including the AFM ordering
perpendicular to the chains) gets destroyed, which leads
to a smearing out of all phase transitions discussed in the
previous section.

6

The loss of long-range order already appears in the
ground-state (T = 0). Within a modeling in terms
of Ising rather than Heisenberg spins, it can be shown
analytically that at T = the spin correlation func-
tion within the chains decays exponentially for large dis-
tancesiSi This is intuitively clear since the strong AFM
coupling Jo enforces an antiparallel alignment of two
spins on the left and right sides of a hole. Thus the chain
splits up into (ferromagnetic) fragments separated by an-
tiphase boundaries. If the holes are distributed randomly,
all long-range spin correlations along the chain are thus
destroyed. For the spin correlations perpendicular to the
chains an analytic treatment is much more complicated
due to the frustration of the interchain interactions. The
latter occurs since fragments of neighboring chains will
generally be displaced against each other. In order to
minimize its total energy the system will thus form ad-
ditional antiphase boundaries within the chains in order
to balance the competing intrachain and interchain in-
teraction energies. In any event, one again expects an
exponential decay of the spin correlations. This has been
confirmed numerically.

The above mechanism for the destruction of long-range
correlations may equally well apply if the Ising spins are
replaced by Heisenberg spins. We have checked this by
analyzing low-temperature Monte Carlo data of the AFM
interchain order parameter, Eq. Q, for varying system
sizes L (Fig.EJ. The order parameter seems to extrapo-
late to zero for L — > oo, indicating the absence of long-
range AFM interchain order in the thermodynamic limit,
as expected from the above arguments.

One should keep in mind that despite the lack of long-
range order, the spins will in general still exhibit some
degree of short-range ordering, which reflects the proper-
ties of the corresponding pure phases (without holes) in
the various regions of the T, _ff-plane.

The smearing out of the phase transitions due to the
presence of the randomly distributed holes can be in-
ferred from our simulational data. E.g., for H = the
specific heat of the pure model exhibits a peak whose
height increases with the system size L and whose po-
sition approaches the Neel temperature Tn as L — *■ oo.
For the disordered system, on the other hand, we observe
a non-critical maximum, being almost size-independent
for sufficiently large systems, which occurs at a "pseudo"
Neel temperature ksT^ /\Ja\ — 0.58, as compared to

ksTN/lJal

0.61 for the pure model (see Sec. IIIIBII .
Thus the Neel transition is not only smeared out but also
slightly shifted towards lower temperature. In addition,
the specific heat shows a small anomaly at lower temper-
atures, which is probably due to the incommensurability
of the wavelength of the helical structures with the sys-
tem size. The same conclusions are found by analyzing
other quantities, such as the magnetic susceptibility.

Turning now to the (in our context) more interesting
case of a non- vanishing magnetic field ^ 0, we examine
how the spin-flop transition of the pure system is affected
by the quenched holes. Again, we flnd that the transi-

0.6

0.5

0.4

A

Y 0.3

V

0.2

O.I

0.005 0.0 1

0.015
1/L

I , \

0.02 0.025 0.03

FIG. 6: Interchain order parameter, Eq. 0, for various sys-
tem sizes (L — 40,80,240) at fixed temperature kBT/\Ja\ =
0.1. For 1/L the order parameter seems to extrapolate
to zero (dashed line), indicating the destruction of the long-
range AFM order due to the quenched holes.

tion is transformed into a smooth anomaly. Whereas for
the pure system the susceptibility exhibits a sharp
peak at H = Hsf (cf. Fig. O , we now observe a broad
(and much smaller) maximum at a "pseudo" spin-fiop
field H^^ (depending on system size), see Fig. [7| The
curves display some finite-size dependence for small sys-
tems. E.g., the strong increase, for small system sizes, of
as ^ is weakened significantly for larger systems.
For systems larger than L — 80 the curves change only
little. The inset of Fig. [7| illustrates this for the posi-
tion of the maximum, which approaches H^f/\Ja\ ~ 0.5.
Analogous conclusions apply to the height of the maxi-
mum, which quickly saturates if L > 80. Thus, in the
disordered case, it seems to be sufficient to simulate sys-
tems of size L = 80 to capture the relevant properties
holding in the thermodynamic limit.

A closer look at the Monte Carlo configurations re-
veals local spin-flop structures for H > H^^, while for
H < H^f domains showing helically modified antiferro-
magnetic structures can be identified. In this sense, H^^
marks a smooth crossover from the AFM phase to the
spin-flop state. This qualitative picture can be corrob-
orated by examining a suitable quantity measuring the
local AFM order (see below).

For magnetic flelds smaller than the spin-flop fleld
H < H^f, the disorder fluctuations due to different real-
izations of the hole distribution are signiflcantly larger
than for H > H^^. To reduce the disorder fluctua-
tions one therefore has to average over many realizations.
Since simulating many different hole distributions for all
values of the field requires too much computational time,

7

o

L=20

□

L=40

0.6

o

L=80

0.5

40 80 120 160
L

0.6 0.8
H/IJ I

0.1

G-OL=20
□-OL=40
a^L=80

0.8P^
0.6

0.3

0.6 0.9
H/IJ I

1.2

1.5

FIG. 7; Susceptibility vs. magnetic field at constant tem-
perature kBT/\Ja\ ~ 0.25. The solid lines are guides to the
eye. Up to 300 disorder realizations were used to generate the
data. The error bars are smaller than the symbol sizes to the
right of the maxima and are thus not shown there. The inset
shows the position of the maximum as a function of system
size.

FIG. 8: Square of the z-component of the local staggered
magnetization, Eq. vs. magnetic field at constant tem-
perature kBT/\Ja\ = 0.25, indicating a smooth transition be-
tween the AFM and the spin-flop phase. For the pure model,
the same quantity appears to behave discontinuously at the
spin-flop field (see inset). In all cases, the error bars are much
smaller than the symbol size.

we generated a large number of up to 300 disorder realiza-
tions for lower fields only {H < H^D- For higher fields,
100 realizations usually turned out to be sufficient. In
this way we obtained a reasonably good statistics for all
data points.

To provide further evidence that the broad suscepti-
bility maximum is indeed the remnant of a smeared-
out spin-flop transition, we study the square of the z-
component of the local staggered magnetization, which
is a measure of the degree of local AFM order perpendic-
ular to the chains:

mocf - E E [SI, - {su, + sui,)/2] ' ■
i=i j=i

(5)

Note that the expression under the double sum is (up
to a factor of 1/4) the square of the z-component of
the local AFM order parameter at site (i,j). This lo-
cal quantity is then averaged over the whole lattice. As
exemplified in Fig.|Hl ((^^s^ioc)^) drops down smoothly as
one increases the magnetic field, i.e. the local AFM order
along the z-direction decreases, as one expects for a tran-
sition between an AFM and a spin- flop phase. Moreover,
the slope of the curve is maximal at the same field H^^
where the susceptibility has its maximum (cf. Fig.[7J).
Note that H^^ is somewhat lower than the spin-flop field
Hsf/\Ja\ — 0.70 of the pure model. For the pure system,
((MjfjQ^)^) appears to jump at H — iJgf (see the inset of
Fig. Ell.

We would like to stress that many of the above conclu-
sions are qualitatively insensitive to details of the model
like the precise values of the interaction parameters, pro-
vided that the system exhibits randomly distributed, im-
mobile holes. In fact, we have also carried out simula-
tions using the model proposed in Ref. which was
based on an interpretation of inelastic neutron scatter-
ing data for La5Ca9Cu2404i (an analysis of this model,
with and without mobile holes, may be found in Ref.
|2^ . This model has a different lattice geometry, a single-
ion instead of an exchange anisotropy and quite distinct
values of the interaction parameters. Moreover, we con-
sidered a simplified model with vanishing NNN coupling
within the chains and a ferromagnetic NN interaction
(i.e., Jc2 = and Jd > 0). Essentially all of our con-
clusions concerning the smearing out of the phase tran-
sitions, which transform into (smooth) anomalies when
introducing quenched holes, also hold for these modified
models. On the other hand, when comparing the results
of the simulations with the experimental data, the quan-
titative agreement seems to be most satisfying for our
present model.

However, the choice of the interaction parameters cer-
tainly has an effect on the typical spin configurations
within the chains. E.g., for the simplified model with
Jc2 — 0, Jci > mentioned above the chains order ferro-
magnetically (this also applies to the model of Ref. I2B).
Thus in the disordered system each chain splits up into
ferromagnetic fragments separated by the holes which in-

8

duce antiphase boundaries. But for the present model
the spins form "hehcal" chain fragments due to the com-
peting intrachain interactions (see Sec. IIII B)| . and again
reverse their direction across a hole. In diffraction ex-
periments, no indications of such a residual helical or-
dering in La5Ca9Cu2404i have been found so far. One
should note, however, that hints at incommensurate or-
dered spin structures, which could in principle be ex-
plained by the presence of a (modified) helical phase,
have been reported in LagCagCu2404i^ Note also that
in the closely-related spin-chain system Li2Cu02 the he-
lical ordering is destroyed due to the anisotropy and the
finite interchain coupling according to theoretical calcu-
lationsi2^ Thus the existence or non-existence of helical
structures appears to be a delicate question which de-
pends sensitively on details of the interaction and the lat-
tice geometry (i.e., the coordination number of the inter-
chain interaction, which is different for La5Ca9Cu2404i
and Li2Cu02).

0.33

0.29

0.25

N

0.21

0.171

0.13

O K„ 1 / I J I = 0.1

□ k,T/IJI = 0.25
0.35

0.6 0.9
H/l J I

FIG. 9: Susceptibility curves for different temperatures (and
system size L = 80) . As in Fig. 13 we used a varying number
of up to 300 disorder realizations (depending on the value of
the field) to generate the data points. The solid lines are
guides to the eye.

k„T / 1 J I

B a

FIG. 10: Magnetic phase diagram of the model with
quenched disorder. The curve shows the temperature depen-
dence of the "pseudo" spin-flop field where the susceptibility
maximum occurs (cf. Fig. |^. All data were obtained using
systems of size L = 80.

theoretical values for H^^^ into physical units, taking the
spin value S" = 1/2 of the Cu^+ ions and the correct
g-factors^ properly into account, we obtain a "pseudo"
spin-flop field of approximately 9T for a temperature of
ksT/lJal = 0.1. This compares reasonably well with
the experimental values. An even better agreement may
be reached by fine-tuning of the interaction parameters
(whose precise values are not known yet), taking into ac-
count quantum effects, or allowing for a (partial) mobility
of the holes.

Summarizing, the above findings corroborate the idea
that the experimentally observed broad anomaly in the
susceptibility curves can indeed be understood as a dis-
order phenomenon due to randomly distributed, quasi-
static holes, which lead to a destruction of the long-range
AFM order and, correspondingly, to a smearing-out of
the spin-fiop transition.

Plotting the susceptibility curves for various tempera-
tures (Fig. inj allows us to draw a more detailed compari-
son with the experimental magnetization measurements.
The temperature dependence of both the position and
the height of the spin-fiop anomaly resemble the experi-
mental data which will be presented in Sec. IIVI (Fig. Illll
quite well.

Finally, we depict the "magnetic phase diagram" of our
model in Fig. ^| i.e., the dependence of the "pseudo"
spin-flop field H^f on temperature. Again, we find qual-
itative agreement with the upper line B2 of the exper-
imental phase diagram (Fig. Ill|l . If one converts the

IV. EXPERIMENTAL RESULTS ON THE
"PSEUDO SPIN-FLOP PEAK"

In order to test the numerical predictions of the pre-
ceding section, we here present some of our experimen-
tal data on the magnetic properties of the lightly hole-
doped spin chains in Laa;(Ca,Sr)i4_a;Cu2404i. We stud-
ied single crystals of approximately 0.2cm'^, grown by the
floating zone technique?2£ For the magnetization mea-
surements a vibrating sample magnetometer (VSM) was
used. The measurements were performed in magnetic

9

fields up to 16T. The fields were applied either parallel
to the chain direction (c axis) or perpendicular to the
CUO4 plaquettes of the Cu02 chains, i.e. along the easy
magnetic axis {b axis).

B(T)

6 1 1 I I I 1 I 1 I I I 1 I 1 I

2 4 6 8 10 12 14 16

T(K)

FIG. 11: Susceptibility of La5.2Ca8.8Cu24 04i vs. magnetic
field B\\b parallel to the easy axis for different temperatures,
(a), and position of the maximum in (a) vs. temperature, (b).

Fig.lTTk shows the susceptibility of La5.2Ca8.8Cu2404i
vs. magnetic field along the 6-axis, at different temper-
atures up to 15K > T/v = 10. 5K. The sharp anomaly
at Bi, which signals the melting of the long-range spin
order, is visible for all temperatures below T^r. In con-
trast to the melting of the spin order, the anomaly at B2
is still present for T > T/v, where only short-range spin
correlations exist. This fact agrees with the observation
that at r = 2.5K the anomaly occurs at fields B > Bi,
where also only short-range spin order does exist. Upon
heating, the anomaly is shifted to higher fields. More-
over, the peak both shrinks and broadens drastically at
higher temperatures. Comparing the data in Fig. Illb
with those in Fig. illustrates the similarities between
the experimental data and the numerical results. The
broad peak at B2 is well described by the model calcula-

tions, which strongly reinforces the approach presented
in Sec. cm

The temperature dependence of the " pseudo" spin-flop
transition in La5.2Ca8.8Cu2404i is summarized in Fig.
Illb . Qualitatively, the presence of the broad anomaly
B2 indicates short-range spin correlations up to 15K. The
temperature dependence of the peak maximum depends
roughly linearly on the temperature, in agreement with
the numerical findings fFig. I10|l . At (7±1)K, the curva-
ture of B2{T) slightly changes, which again resembles the
numerical results.

V. CONCLUSION AND OUTLOOK

We have presented measurements of the magnetic
properties of the anisotropic spin chains in lightly hole-
doped Lax(Ca,Sr)i4_a;Cu2404i, x > 5. The experiments
suggest that for fields B > 4T the system is charac-
terized by short-range AFM spin order and quasi-static
charge disorder. The susceptibility as a function of the
magnetic field B (applied along the easy axis) shows a
broad anomaly instead of a sharp peak, as one would have
expected if the system underwent a spin-flop transition
(and as one indeed observes in the related, but hole-free
spin chain compound Li2Cu02, see Ref . 123) . In order to
understand these findings theoretically, we have carried
out Monte Carlo simulations of an anisotropic classical
Heisenberg model with quenched holes. Our numerical
data show that the spin-flop transition of the pure model
is smeared out upon introducing quenched holes. The
susceptibility curves at flxed temperature exibit broad
peaks and resemble the experimental data quite well. At
low temperatures, the peak occurs at a field value slightly
below the corresponding spin-flop field of the pure sys-
tem. Furthermore, the peak position increases with tem-
perature, similarly as it is observed in the experiments.
Taken together, our numerical studies corroborate the
idea that the broad anomaly in the experimental suscep-
tibility curves is essentially a signature of the disorder
due to quasi-static holes.

Nonetheless, there remain several challenging ques-
tions for future (experimental and theoretical) work. One
of them concerns the possible mechanisms for the pinning
of the holes, which would explain the occurrence of quasi-
static (quenched) charge disorder. While the destruction
of the stripe-ordered phase by an effective, fleld-induccd
attraction of the holes (as proposed in Ref. 22) obviously
requires a certain mobility of the holes, the existence of
the broad anomaly in the susceptibility seems to sug-
gest that pinning might play an important role to under-
stand the high-field behavior {B > Bi). If one assumes
the holes to move freely along the chains, the theoretical
models predict a clustering of the holes upon increasing
the field and no broad anomal y in the susceptibility oc-
curs (see the discussion in Ref. l28j) . Thus one may spec-
ulate that as the field becomes large enough {B > Bi)
and the holes start to move around, they get trapped

10

at (randomly distributed) pinning centers and then stay
more or less immobile.

Closely related to the pinning of the holes is the possi-
ble influence of the Coulomb interaction, which has been
neglected in the theoretical models so far. The Coulomb
repulsion of the holes destabilizes the above-mentioned
clustered structures and would tend to distribute the
holes more uniformly across the system. It is unclear,
however, whether inclusion of the Coulomb interaction
between the holes alone would suffice to predict the ex-
istence of quasi-static disorder for fields B > Bi. It
might also be necessary to take the interaction with the
La'^"'" and Ca^+ ions into account. Due to their different
charges and ionic radii, these might introduce additional

disorder into the system which may turn out to be im-
portant for an understanding of the pinning of the holes.

Acknowledgments

Wc thank J. M. Tranquada for a useful discussion.
This work was supported by the Deutsche Forschungsge-
meinschaft (DFG) within SPP 1073 (BU 887/1-3). W.S.,
M.H., and R.K. gratefully acknowledge financial support
by the Deutsche Forschungsgemeinschaft under Grants
No. SE 324/4 and KL 1824/1-1, respectively.

^ S. Kivelson, I. Bindloss, E. Fradkin, V. Ogancsyan, J.

Tranquada, A. Kapitulnik, and C. Howald, Rev. Mod.

Phys. 75, 1201 (2003).
^ J. M. Tranquada, D. J. Buttrcy, V. Sachan, and J. E.

Lorenzo, Phys. Rev. Lett. 73, 1003 (1994).
^ P. Wochner, J. M. Tranquada, D. J. Buttrey, and V.

Sachan, Phys. Rev. B 57, 1066 (1998).

B. J. Sternlieb, J. P. Hill, U. C. Wildgruber, G. M. Luke, B.

Nachumi, Y. Moritomo, and Y. Tokura, Phys. Rev. Lett.

76, 2169 (1996).
^ M. Takigawa, N. Motoyama, H. Eisaki, and S. Uchida,

Phys. Rev. B 57, 1124 (1998).
® C. Hess, H. ElHaes, B. Biichner, U. Ammerahl, M. Hiickcr,

and A. Revcolevschi, Phys. Rev. Lett. 93, 027005 (2004).
^ R. KUngeler, N. Tristan, B. Biichner, M. Hiicker, U. Am-
merahl, A. Revcolevschi, Phys. Rev. B 72, 184406 (2005)
* P. Horsch, M. Sofin, M. Mayr, and M. Jansen, Phys. Rev.

Lett. 94, 076403 (2005).
^ D. Akahoshi, M. Uchida, Y. Tomioka, T. Arima, Y. Mat-

sui, and Y. Tokura, Phys. Rev. Lett. 90, 177203 (2003).
^° J. Burgy, M. Mayr, V. Martin-Mayor, A. Moreo, and E.

Dagotto, Phys. Rev. Lett. 87, 277202 (2001).
" E. Dagotto, Science 309, 257 (2005).

G. Alvarez, M. Mayr, A. Moreo, and E. Dagotto, Phys.

Rev. B 71, 014514 (2005).

R. S. Eccleston, M. Uehara, J. Akimitsu, H. Eisaki, N. Mo-
toyama, and S. Uchida, Phys. Rev. Lett. 81, 1702 (1998).
V. Yu. Yushankhai and R. Hayn, Europhys. Lett. 47, 116

(1999).

V. Kataev, K.-Y. Choi, M. Griiningcr, U. Ammerahl, B.
Biichner, A. Freimuth, and A. Revcolevschi, Phys. Rev.
Lett. 86, 2882 (2001).

L. P. Regnault, J. P. Boucher, H. Moudden, J. E. Lorenzo,
A. Hiess, U. Ammerahl, G. Dhalenne, and A. Revcolevschi,
Phys. Rev. B 59, 1055 (1999).

^"^ U. Ammerahl, B. Biichner, L. Colonescu, R. Gross, and A.
Revcolevschi, Phys. Rev. B 62, 8630 (2000).

^® M. Matsuda, K. M. Kojima, Y. J. Uemura, J. L. Zarestky,
K. Nakajima, K. Kakurai, T. Yokoo, S. M. Shapiro, and
G. Shirane, Phys. Rev. B 57, 11467 (1998).

^® U. Ammerahl, B. Biichner, C. Kerpen, R. Gross, and A.

Revcolevschi, Phys. Rev. B 62, R3592 (2000).
^° W. Selke, V. L. Pokrovsky, B. Biichner, and T. KroU, Eur.
Phys. J. B 30, 83 (2002).

21 T. Kroll, R. Klingcler, ,J. Geek, B. Biichner, W. Selke, M.
Hiicker, and A. Gukasov, J. Magn. Magn. Mater. 290-291,
306 (2005).

22 R. Klingeler, PhD thesis, RWTH Aachen (2003).

23 M. Holtschneider and W. Selke, Phys. Rev. E 68, 026120
(2003).

2* M. Holtschneider, W. Selke, and R. Leidl, J. Magn. Magn.

Mater. 290-291, 326 (2005).
25 W. Selke, M. Holtschneider, and R. Leidl, Cond. Matter

Physics 8, 15 (2005).
2® M. Matsuda, K. Kakurai, J. E. Lorenzo, L. P. Regnault, A.

Hiess, and G. Shirane, Phys. Rev. B 68, 060406(R) (2003).
2^ A. Gukasov, unpublished.

2* R. Leidl and W. Seiko, Phys. Rev. B 70, 174425 (2004).
Y. Mizuno, T. Toliyama, S. Maekawa, T. Osafune, N. Mo-
toyama, H. Eisaki, and S. Uchida, Phys. Rev. B 57, 5326
(1998).

3° R. Leidl and W. Selke, Phys. Rev. B 69, 056401 (2004).

31 M. Matsuda, T. Yosihama, K. Kakurai, and G. Shirane,
Phys. Rev. B 59, 1060 (1999)

32 S.-L. Drcchsler, J. Richter, J. Malek, A.S. Moskvin, R.
Khngcler, and H. Rosner, J. Magn. Magn. Mat. 290-291,
345 (2005).

33 R. KUngeler, B. Biichner, K.-Y. Choi, V. Kataev, U. Am-
merahl, A. Revcolevschi, and J. Schnack, Phys. Rev. B 73,
014426 (2006).

3* Y. Mizuno, T. Tohyama, and S. Maekawa Phys. Rev. B

60, 6230 (1999).
35 C. de Graaf, 1. de P. R. Moreira, F. lUas, 6. Iglesias, and

A. Labarta, Phys. Rev. B 66, 014448 (2002).
3® T. Nagamiya, K. Nagata, and Y. Kitano, Prog. Theor.

Phys. 27, 1253 (1962).

J. M. Robinson and P. Erdos, Phys. Rev. B 2, 2642 (1970).
38 M. Matsuda, K. Katsumata, T. Yokoo, S. M. Shapiro, and

G. Shirane, Phys. Rev. B 54, R15626 (1996).
3® U. Ammerahl and A. Revcolevschi, J. Cryst. Growth 197,

825 (1999).