Magnetic behavior in Li-doped La2CuO4.

Author(s): Le, LP; Heffner, RH; MacLaughlin, DE; Kojima, K; Luke, GM; Nachumi, B; Uemura, YJ; Sarrao, JL; Fisk, Z | Abstract: We report muon-spin-relaxation studies in the Li-doped cuprates (Formula presented)(Formula presented)(Formula presented)(Formula presented) for x=0.01, 0.05, 0.10, 0.45, and 0.50. For low Li concentrations (x⩽0.10) we find a rapid suppression of (Formula presented) as x increases, but little change in the magnitude and temperature dependence of the antiferromagnetic order parameter. This indicates that Li doping effectively destroys antiferromagnetism (similar to Sr doping, but different from Zn doping) without strongly affecting either the on-site Cu moments or the shape of the spin-wave excitation spectrum. For high Li concentrations we find that the majority of the sample volume is nonmagnetic, suggesting possible singlet-state formation. © 1996 The American Physical Society.

The interplay between magnetism and superconductivity in the cuprate superconductors continues to be a subject of considerable interest. The parent compound La 2 CuO 4 is an antiferrmagnetic ͑AFM͒ insulator (T N Ϸ325 K͒ with a perovskite structure. When an out-of-plane dopant, such as Sr or Ba, is substituted for La, holes are doped into the CuO 2 planes and the AFM order is destroyed. Increased doping leads to superconductivity. Doping into the CuO 2 plane itself destroys both superconductivity and magnetic order. Generally, the in-plane dopants, such as Zn or Mg substituted for Cu, rapidly depress T c , 1 but only modestly depress T N . 2 Recently, the effects of substituting Li for Cu have begun to be investigated. 3 Naively, the addition of a Li atom is roughly equivalent to adding both a Sr and a Zn atom, in the sense that Li both removes a Cu moment ͑as does Zn͒ and adds a hole ͑as does Sr͒. One key difference between Sr and Li doping, however, is that the Sr-added holes are very mobile, while the Li holes are not. This is evident because La 2 Cu 1Ϫx Li x O 4 remains an insulator for 0рxр0.5. 4 For xϭ0.5 the Li sublattice is ordered, and the bulk magnetic susceptibility is negative ͑due to core diamagnetism͒. 5 In this paper we use muon spin relaxation (SR͒ as a microscopic probe of the Li concentration dependence of both the loss of magnetic order and the onset of a nonmagnetic state. Because SR is extremely sensitive to very small magnetic moments and is easily carried out in zero applied field, the technique is well suited to such a study. We find that a small amount of Li doping (р10%͒ will strongly suppress the AFM ordering temperature without drastically reducing the on-site Cu moments. The temperature dependence of the sublattice magnetization is essentially unchanged. Furthermore, there is an inhomogeneous magnetic phase which forms for xϭ0.05-0.10 and has a low ordering temperature.
Eventually the Cu moments are destroyed completely for x near 0.5; we set an upper limit of 10 Ϫ3 B for the size of any residual copper moment for xϭ0. 5.
The SR measurements were carried out at TRIUMF, Canada, using the M15 and M20 surface muon channels. Polycrystalline samples of La 2 Cu 1Ϫx Li x O 4 were prepared by standard solid state reaction techniques at Florida State University. The samples were carefully annealed so that they contained no oxygen deficiency, pressed to pellets about 25 mm in diameter and 3 mm thick, and then mounted in a He-gas-flow cryostat.
For the lightly doped systems a spontaneous muon spin precession is observed at low temperatures in zero applied field, indicating the onset of magnetic ordering. The spectra below T N can be fit by the function G͑t ͒ϭ 2 3 cos͑2 tϩ͒exp͑Ϫt ͒ϩ 1 3 exp͑ϪЈt ͒, ͑1͒ where the frequency is a measure of the sublattice magnetization M , and the spectral weights for the oscillating and nonoscillating signals ͑2:1͒ are due to the powder average for a single muon lattice site. The temperature dependence of for xϭ0 ͑taken from Ref. 6͒, 0.01, 0.05, and 0.10 is given in Fig. 1͑a͒. The rapid suppression of T N is evident. However, one sees that despite the reduction of T N by a factor of 60 ͑from 300 K to 5 K for xϭ0 and xϭ0.10, respectively͒, the zero-temperature frequency declines only by 1/3 ͑from 5.8 MHz to about 4 MHz͒. This indicates that the Li doping efficiently breaks down the strength of the magnetic correlations, but only weakly reduces the on-site Cu moments. ͑The cause of the decrease in M will be examined below.͒ Other SR experiments with different dopants, excess oxygen, 6 doping Sr on the La site, 7 and Zn on the Cu site, 8 also show a much larger reduction in T N compared to the on-site moments. Figure 1͑b͒ shows the normalized linewidth /2 vs temperature for various Li concentrations. We see that the spread in local fields becomes larger as the Li concentration is increased, indicating increasing microscopic inhomogeneity. The observed linewidth in La 2 Cu 1Ϫx Li x O 4 could be due either to a dilution of the Cu moments or to a spread in the magnitude of the Cu moments. To differentiate between these two scenarios, we have computed the field distribution for a dilute spin system, using the La 2 CuO 4 -type magnetic structure with a frozen moment of 0.5 B . 6 We also assume a random distribution of nonmagnetic sites and a muon lattice site of (0.253, 0, 0.162) in the tetragonal notation, 9 with respect to a Cu site at the origin. Figure 2 shows the field distribution (B) for a 10%-spin-diluted system. The small peak near 80 G has a spectral weight of 10% and corresponds to a missing Cu spin at the origin. Compared to the internal field in the undoped system ͑a single line indicated by the arrow in Fig. 2͒, the large peak shows some broadening, but no frequency shift. Except for the small tails, the line shape of this peak can be reasonably described by a Lorentzian distribution with a normalized linewidth of 0.06 ͑the dashed line in Fig. 2͒. Because Li doping removes one spin from the Cu site and also creates a hole, x Li atoms will create at most 2x nonmagnetic sites. Since for xϭ0.05-0.10 neither the decrease in ͑20-30%͒ nor the broader linewidth ͑0.15-0.30͒ is explained by spin dilution alone, Li doping must decrease the average on-site moments and broaden the moment distribution. These effects, however, are much weaker than the reduction of T N (x) with x.
We now examine the effects of Li doping on the magnetic transition temperature. Figure 3 shows a comparison of the magnetic phase diagrams in La 2 Cu 1Ϫx Li x O 4 , La 2 Cu 1Ϫx Zn x O 4 , and La 2Ϫx Sr x CuO 4 . We see that Li doping strongly depresses the AFM order, similar to Sr doping. If we fit the small-x Li data to T N (x)/T N (0)ϭ1Ϫ(x/x c ) 2 phenomenologically, 10 we obtain a critical value x c ϭ0.03, slightly larger than 0.02 for Sr doping, but much smaller than 0.12 for Zn doping. 2 The difference between Sr or Li and Zn is that Zn only removes spins from the Cu sites, so that the remaining spins are still magnetically coupled if the doping concentration is not large. Although x c ϭ0.12 for Zn doping is smaller than the percolation threshold of 0.41 expected for a two-dimensional ͑2D͒ Heisenberg square lattice, 12 the magnetic order does survive up to a relatively high doping concentration compared to Sr or Li. On the other hand, both Sr and Li create holes in the CuO 2 planes which frustrate the antiferromagnetic coupling between the neighboring spins. Thus the presence of holes in the CuO 2 planes is much more effective in destroying the AFM order than simple dilution of the Cu moments.
There are differences between Sr and Li doping, however. First, Li doping removes a Cu spin while Sr doping does not. As discussed above for the case of Zn, this has only a weak effect on T N (x). Of greater importance is that Li doping creates a more localized hole than Sr doping. This conclusion is mainly derived from the behavior in the heavily doped compounds; namely, Sr doping leads to superconductivity, while Li doping leads to insulating diamagnetism at xϭ0.5. However, in the low Li-doping range the hightemperature resistivity actually becomes smaller with increasing Li concentration and reaches a minimum near xϷ0. 10. 4 Similarly, the residual susceptibility 0 initially increases with increasing x and reaches a maximum also near xϷ0. 10. 10 These results indicate that as long as x is small, which is the range of importance for magnetism, Li doping may introduce somewhat delocalized holes in the CuO 2 plane, similar to Sr doping. This is consistent with our observation that Sr doping is only marginally more effective than Li doping in destroying the AFM coupling.
As shown in Fig. 3, T N (x) changes behavior for xу0.03 in both the Sr-and Li-doped systems; namely, the magnetic ordering temperature is small but persists to a rather high concentration, of the order of 10%. To investigate the nature of this intermediate magnetic state, we conducted zero ͑ZF͒ and longitudinal ͑LF͒ field measurements. The SR time spectra observed at 3K in the 10% Li-doped system are shown in Fig. 4͑a͒. We see that the ZF spectrum clearly shows oscillations, accompanied by strong damping ͓comparing with the 1% system, as shown in Fig. 4͑b͔͒. The frequency is 3.1 MHz, corresponding to a muon local field of 230 G. When a longitudinal field of 1 kG is applied, the oscillation is decoupled completely, indicating a static local field, at least during the muon lifetime. ͑The upper limit on the spin fluctuation rate is about 10 6 s Ϫ1 .͒ We note that magnetic correlations for x Li у0.03 ͓Fig. 4͑a͔͒ have not been observed in susceptibility or neutron scattering measurements. In the former case, we find a Curielike upturn in around 50 K, which would hide any AFM signature at low temperatures. For neutron studies, a shortranged spin correlation (Ϸ40 Å͒ is found below room temperature, but no 3D magnetic ordering is observed down to 5 K in the xϭ0.05 system. 13 The absence of magnetic Bragg peaks together with our SR observations implies either short-range magnetic correlations or possibly an incommensurate magnetic structure. Features similar to those seen in Fig. 4͑a͒ are also found with SR in the Sr-doped compounds. 11 The observation of magnetic correlations, particularly in the superconducting region (x Sr Ϸ0.10), has led to much debate on whether magnetism coexists with superconductivity 14 or whether the observed correlations are due to magnetic impurities. 15 Unfortunately, it is not clear at this point whether the correlations seen in Sr-doped La 2 CuO 4 are static, because no longitudinal field measurements have been conducted. For lower Sr concentration, a more careful measurement combining both SR and neutron studies on a sample with x Sr ϭ0.06 ͑Ref. 16͒ showed the formation of a spin glass-like state below 6 K. Our SR data for a sample with x Li ϭ0.05-0.10 show similar results, although the field distribution in our samples seems more consistent with either a broad width around a finite mean value or an incommensurate magnetic structure, rather than a spin glass state. This is particularly true for the sample with x Li ϭ0.05, where the normalized linewidth is about 0.15, much smaller than 0.42 expected for a spin glass ͑assuming a Gaussian line shape͒.
One of the interesting observations in our experiments is the universal scaling of the temperature dependence of the frequency for 0рxр0.10, as shown in Fig. 5. This scaling behavior is remarkable because it holds over a wide range of T N , which includes a possible change in the magnetic struc- ture, as discussed above. Note that this result does not hold for all other dopants. For example, doping with excess oxygen stiffens the magnon excitations, leading to a flattening of M (T) at low temperatures. 17 The parent La 2 CuO 4 compound can be viewed as a 2D Heisenberg antiferromagnet, with a small interplane coupling J Ќ and a small XY anisotropy J XY . 17 J Ќ is primarily due to the orthorhombic distortion of the CuO 2 lattice, which lifts the frustration of the body-centered Cu spin. When doping with Li or Zn, the Cu sites are no longer completely filled with spins. Nevertheless, if x is small, the M (T) data may still be explained by considering an effective intraplane coupling J ʈ , as well as J Ќ and J XY . The observed scaling behavior for M (T) implies that all three exchange constants decrease proportionally as x Li increases, so that T N (x) ϰJ ʈ (x) and M (T/T N ,x) is independent of x. 18 This result agrees qualitatively with a decrease in the orthorhombic strain ͑and hence J Ќ ) with increasing x Li . 3,10 Finally, we discuss our results obtained in the high Li concentration regime (xϭ0.45 and 0.50͒. We find that the ZF-SR spectra at high temperatures are well described by a Kubo-Toyabe relaxation function 19 due to nuclear dipoles alone. Below about 200 K, however, a fast-relaxing component is observed, which grows both in amplitude and relaxation rate as the temperature is reduced. At the lowest measured temperature, this component corresponds to about 15% of the sample volume and a muon static magnetic field of 260 G. Both the magnitude of the muon static field and the onset temperature are comparable to those obtained in the lightly doped systems. Thus, the formation of the fastrelaxing signal is most likely due to magnetic clusters with partially frozen Cu moments; these clusters are inhomogeneously distributed throughout the sample volume. The remaining slow relaxation component ͑about 85% of the sample volume͒ is essentially the same as that found at temperatures above 200 K. We estimated that the zero-field re-laxation rate due to the electronic magnetism is less than 0.08 s Ϫ1 , which corresponds to an upper limit for the residual Cu moment of 10 Ϫ3 B , a reduction by more than three orders of magnitude. This is consistent with a spinsinglet state in the 50% doped system.
In conclusion, we have performed SR measurements on the Li-doped cuprates La 2 CuO 4 . We obtained the magnetic phase diagram for doping concentrations which vary from xϭ0 to xϭ0.5. In the low doping range (xр0.03) a rapid suppression of T N is seen, similar to Sr doping, but different from Zn doping, indicating that the presence of holes in the CuO 2 planes is much more effective in destroying antiferromagnetic correlations than simple dilution of the Cu moments. In the intermediate doping range (xϭ0.05-0.10͒, we found that the magnetic correlations persist, but with increasing inhomogeneity. A change of T N vs x behavior near xϷ0.03 may indicate a possible change in the nature of the magnetic state of La 2 Cu 1Ϫx Li x O 4 . This is also suggested by the form of the relaxation function ͑Fig. 4͒. For xр0. 10 we found a modest 20-30 % change in the magnitude of the on-site Cu moments and no change in the temperature dependence of the magnetic order parameter, despite the fact that T N declines by a factor of 60. This suggests that both the interplane and intraplane coupling constants scale with T N . As the doping concentration increases further, we find no evidence of magnetic relaxation in the majority of the sample volume, consistent with the formation of the singlet ground state for xϭ0.45-0.50.