Spin-Peierls State versus Neel State in Doped CuGe03

The competition between nonmagnetic spin-Peierls (SP) and magnetic Neel ground states has been investigated in CuGe03 doped with Zn, Ni, Mn, and Si. The analysis of specific heat, C„, data in CuGe03, measured as a function of magnetic field, supports a SP transition at Tsp = 14 K. The replacement of Cu ' by Zn +, Ni ', or Mn ' reduces Tsp and stabilizes a Neel state, not a spin-glass phase as previously suggested. No significant difference in C„was observed for Zn-and Ni-doped samples. We suggest that the Neel state depends on ~S' — S~, where S' is the spin of the dopant and S=2.

. AF chains of Cu ' along the c axis are coupled through 0 ions and separated by Ge-0 chains.This structure suggests 1D AF magnetic coupling.Recent neutron scattering data [5] have been interpreted as due to a continuous twisting of the oxygens connecting the Cu atoms to the a-b plane between room temperature and T -Tsp.This structural distortion was claimed to abruptly end at Tsp, suggesting that it drives the SP transition, i.e. , the motion of the oxygens modulates the spin-spin interaction along the c axis, and they are responsible for the coupling between the singlet spin pairs and phonons.
A model proposed by Imada [6], using the concept of the resonating valence bond (RVB) introduced by Anderson [7], may describe the SP phase in CuGe03.In his model, the spin gap is not a consequence of a static lattice dimerization, but the result of dynamic SP fIuctuations.
Nevertheless, two-magnon scattering data are not in agreement with the SP interpretation [8], but agree with a strong interchain ferromagnetic (FM) interaction in CuGeOs, contrary to the negligible interchain interaction expected for SP systems.
CuGe03 is interesting because it can be used in order to address the problem of doping in one-dimensional magnetic systems.This is a subject of current interest where little experimental work is available [9], primarily because of the difficulty of finding chain magnets that can be doped without large changes in their symmetry and magnetic interactions.
CuGe03 is a unique system because it allows us to study the effects of partial ] replacement of Cu by 5 4 2 ions.The effect of doping has not been investigated in organic SP systems because of the low solubility of dopants.Hase et al. recently reported [10] a rapid decrease of Tsp in CuGeO, upon doping with Zn and the appearance of a new magnetic state for concentrations of Zn ~0.02.From their g data, they concluded that the new state is a spin-glass (SG) phase.On the other hand, Lu, Su, and Yu [11]suggested that Zn doping results in the collapse of the spin gap, and predicted the existence of a gapless SP state for Zn -0.03.
In this Letter we present specific heat, |"~, data mea- sured as a function of magnetic field (H) in pure CuGeO, and the effects of Zn, Ni, Mn, and Si doping.Mag- netic specific heat has been useful in understanding organic SP systems [12].The data are analyzed in terms of a mean-field model and compared with ~and EPR measurements on the same samples [13].
0.2 tion in air.Large single crystals were grown by the slow cooling of stoichiometric melts.The specific heat was measured between 1.4 and 20 K using a small-sample re- laxation calorimeter [14].Data for C"/T vs T for several compounds are shown in Fig. 1.A A-type anomaly is observed in every case.Cĩ s similar for 4%%uo Ni or Zn doping.A more rapid reduction of Tsp was found for Mn than for Ni or Zn doping.In con- trast, Si which substitutes for Ge has little effect on Tsp.
Applied fields of up to 10 T shift Tsp by -1.7 K.As seen in the inset of Fig. 1, Cup9gZnp p2Ge03 exhibits anomalies at -3 and -10 K.In order to separate the lattice contri- bution, CL, from the magnetic contribution, C, we have assumed that the magnetic transition, the applied magnetic field, and the doping have only minor effects on CL.It is a reasonable assumption because independent fits for the different samples provide comparable values for CL.This is clear from a simple inspection of the curves shown in Fig. 1 which remain essentially parallel everywhere except close to the anomalies.The data for T && Tsp have been fitted by Cp --P, T' + c exp( -bTsp/T), where the first term accounts for CL.A better fit for T )4 K is obtained by adding a term P2T5 The d. ata can then be represented above and below Tsp by means of the same CL.A mean-field analysis of an SP system is analogous to a BCS analysis for the superconducting state; thus we fit the magnetic specific heat data below Tsp using a BCS-like exponential gap function [15].The data above Tsp were fitted by where the linear term yT is characteristic of 1D homo- geneous AF systems with y = 2Nk~/3J [3,12].More weight was given to the parameters obtained for Eq. ( 1) using CuGe03 data.A larger T range is covered when T ( Tsp for the undoped sample; consequently, the errors are smaller. Using the same argument, the parameters .obtained from fitting the 4% Zn and Ni data by Eq. ( 2) were given more weight for T ) Tsp.These parameters were determined independently for each compound and fell within experimental error of each other.The best fits were obtained with Pi = 0.66 ~0.05 m J/molK~, P2 = -0.00077~10 5 m J/molK6, and y = 45 5 m J/molK .Pt corresponds to a Debye temperature Oo -240 K.In Fig. 1 we display the fit by Eq. ( 1) of the CuGe03, and by Eq. ( 2) of Cuo96Znoo4GeO~data, with H = 0.For the sake of clarity, fits to the other samples are not given.
If y = 45 ~5 m J/molK2 is substituted in the expression y = 2Nkli/3J, a value of J/kii -123 K is obtained that is in agreement with the value of -125 K derived from high magnetic field data [16].The discrepancy with J/kii = 88 K, obtained from y data [2], may be associated with the poor agreement between the g data and the model used to fit them [17].
In order to study the region of the transition in more detail we analyze it within the mean-field model, as in the case for the organic SP compounds [12].In Fig. 2 we present C = C~-CL for CuGe03.The A- shaped anomaly is approximated by means of a triangular function, with the same entropy gain under this function as under the measured anomaly.
The baseline of the triangle, represented by the term yT for the 4% Zn and Ni specimens, shows the entropy gain as compared to the case where no transition occurs.The mean-field transition temperature, Tsp, is found to be -14 K, in excellent agreement with Tsp obtained from g and EPR data [2, 13].The temperature of the peak, T", is -13.3K, i.e. , indicating a transition width of -0.7 K.The jump in the triangular function, -920 ~100 m J/molK, can be compared with the jump predicted by the BCS model, 5C = 1.43'Tsp.This expression yields a value of y 46 ~5 m J/mol K, in excellent agreement with the value of y obtained by extrapolation above Tsp.Below Tsp, the magnetic term C was fitted by c exp( -bTsp/T).
The fit yields c -20 J/mol K and b -2.6.These values can be compared with those expected for a BCS model, c 10yTsp atid b 1.5 [15].For CuGe03, values of c -6 J/molK and b -1.5 are obtained by using y-45 m J/mol K and Tsp" -14 K, these values are smaller than the experimental ones.Similar discrepancies were found in organic SP systems [12].
A test for the SP model is to study the change of Tsp with magnetic field.The behavior in a field can help in discriminating between a simple structural transition and a SP transition.A large difference in the decrease of Tsp is expected between the two cases, hT -(H/J)2 for a structural transition and ATsp -(H/Tsp) for a SP transi- tion [18].In the inset of Fig. 2 we present the fits of the C", data measured at 0, 5, and 10 T by c exp( bTsp/T)- The best fits are obtained with c = 20 ~2, 16 ~2, and 9 ~2; and bTsp = 37 + 1, 32 ~1, and 25 ~1 K for H = 0, 5, and 10 T, respectively.When analyzing the data within the mean-field model, values of Tsp = 13.6 and 12.3 K and T~= 12.8 and 11. 5 K were found for H = 5 and 10 T, respectively.A decrease of bTsp with H is expected because of the increase of the Zeeman energy and the reduction of the energy gap in the spin- wave spectrum.For fields with ppH ( 0.5kpTsp, a de- crease of Tsp by b.Tsp/Tsp = o.x(1+ x + . . .), where x = (gpiIH/2kiITsp) and g (gyromagnetic factor for Cu +) is -2.17 for a powder sample [13], has been predicted [18 -20].The expression reduces to the first term for p, &H « k&T.Using the Tsp" and Tp given above, values of n = 0.41 and 0.40 for 5 and 10 T are obtained, respectively.They agree well with those derived from a Hartree-Fock approximation, which yields n = 0.44 [18,19] and the calculation by Cross which pre- dicts Ix = 0.38 [20].For fields larger than 5 T, the agree- ment is possibly fortuitous.
A similar agreement was found in the organic SP systems when measuring at high fields [21].Our data analysis supports a SP transition, in disagreement with the two-magnon scattering conclusions [8].A strong FM interaction between chains, as suggested in Ref. [8], should produce large g shifts of the Cu ' EPR line as a function of T, but they were not observed [13].
The influence of doping upon the magnetic specific heat can be observed in Fig. 3, in which C /T ratios are plotted as a function of temperature.The anomaly associated with the SP transition shifts to lower T with doping.As previously suggested by Hase et al. [10], and measured at 0 (O), 5 (V), and 10 (0) T, and at H = 0, for 2% (0) and 4% g ) Zn and 4% (x) Ni doping.Inset (a) shows C,"Q,), and the zero-field-cooled (ZFC) (O) and field-cooled (FC) (4) y for Cuo96Zno04Ge03, measured in 50 Oe.Inset (b) shows C (x) and y (6) for Cuo96NioII4GeO, measured in 50 Oe.
seen in Figs. 1 and 3, there is a second anomaly at -3 K for Cuo98Zn002GeO&.For larger concentrations of Zn or Ni, the temperature of the second anomaly increases to -4 K with the complete disappearance of the SP anomaly.There is no significant difference between the specific heat for 4%%uo Zn and Ni materials.We calculated the excess entropies for 4% Zn and Ni with ~S' -S~= The fit yields AS -0.15 J/molK, which accounts for -4% of 8 ln(2~S' -S~+ 1) = 5.76 J/molK.One would expect that replacing Cu by Zn would result in an increase of g due to Cu ions that do not dimerize in a singlet ground state.However, only a small increase in g, that weakly depends on doping, is measured.This can be understood as follows: Zn can be thought of as an on-site spin vacancy that introduces a localized magnetic moment which is "polarized" by the staggered magnetization.
As it sees the local field due to the staggered order, no Curie term appears [9].The consequence of doping with nonmagnetic Zn + or magnetic Ni ', S = 1, is almost the same.The increase in y due to Ni doping is small compared with the contribution expected from Ni + free ions.The effect of doping with Mn +, 'S = 2, was studied by EPR.The EPR data can be explained as Mn + doping depresses Tsp about twice as rapidly as Zn and Ni.From these observations we may infer that the difference between the spin value of the doping impurity and the S = 2 spin of Cu ' is an important parameter in determining the properties of the system.Furthermore, the size of the ion substituting for Cu + = 0.73 A (Zn~= 0.74 A, Ni + = 0.69 A) does not seem to play a significant role.
Replacing Ge + by up to 10% of Si"' reduces Tsp by just -1 K.
Our data analysis does not agree with the speculation of Ref. [10] that a SG phase resulted from Zn doping.As seen in the insets of Fig. 3 for 4% of Zn or Ni, a peak in C is found at the same T, within the experimental error, where g shows a maximum.It is well known that the magnetic specific heat of a SG system shows no indication of a cooperative peak, discontinuity, or any broad anomaly close to the freezing temperature, Tso, where ~displays its characteristic sharp peak [22].For a SG system, C increases linearly with T showing only a rounded maximum at T well above Tso.Furthermore, contrary to Ref. [10], and as can be seen in inset (a) of Fig. 3, our low field ~measured in 0 = 50 Oe for Cu 0 96Zn 0 p4GeO 3 does not display a difference between zero-field cooling and field cooling below the g cusp, as would be expected for a SG compound [22].Another argument against the SG description is that frustration is required for its occurrence, which is not possible within 1D chains, unless a strong interchain interaction is induced by doping [10].The peak in C and the drop in the y below its maximum are strong indications of a Neel state, rather than a SG.The presence of two peaks in the Cp of Cu o 98Zn 0 02GeO 3 suggests the coexistence of a SP and a Neel state, for which we do not have an explanation.
In addition, the existence of the two peaks appears to be inconsistent with the gapless SP state proposed by Lu, Su, and Yu [11].
As mentioned above, neutron scattering data [5] indi- cate that the SP transition in CuGe03 is not a consequence of a static lattice dimerization, but is due to dynamic Auctuations of the oxygens, a possibility proposed earlier by Imada [6] using the concept of RVB [7].Doping could destroy the fluctuations by pinning the impurities, which could explain the rapid decrease of Tsp with Cu-site sub- stitution.Inagaki and Fukuyama [23] studied the possibil- ity of having a magnetic Neel or a SP as a ground state for a quasi-1D Heisenberg AF.They obtained a phase dia- gram where either the SP or the Neel state is the ground state depending on the ratio between the interchain and intrachain exchange interaction.As Cu substitution modi- fies the chain length, it changes the spin-phonon interaction, and possibly increases the intrachain-interchain ratio, then a Neel state ground state could be stabilized.For CuGe03 the boundary between the SP state and the Neel state is -3% for Zn or Ni and -1% for Mn.
New x-rays and neutron studies on CuGeO3 were reported after the completion of this work [24].Their authors postulate the existence of oxygen displacements in the a-b plane and critical fluctuations above Tsp for CuGe03.Their data analysis supports a SP transi- tion [24].In summary, the analysis of the field dependent specific heat data of CuGe03 indicates the occurrence of a SP transition, and the specific heat data suggest that the second magnetic anomaly observed for doped samples is associated to a Neel and not a SG ground state.This research was sponsored at San Diego State Uni- versity by NSF Grant No. DMR-91-17212 and at Los Alamos National Laboratory under the auspices of the United States Department of Energy.One of us (B.A. ) was supported by TUBITAK.We wish to thank D.

FIG. 3 .
FIG. 3. Magnetic specific heat, C /T, as determined from the subtraction of the lattice contribution, CL.For pure CuGe03