This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Ce-site dilution in the ferromagnetic Kondo lattice CeRh6Ge4

Jia-Cheng Xu Center for Correlated Matter and Department of Physics, Zhejiang University, Hangzhou 310058, China    Hang Su Center for Correlated Matter and Department of Physics, Zhejiang University, Hangzhou 310058, China    Rohit Kumar Center for Correlated Matter and Department of Physics, Zhejiang University, Hangzhou 310058, China    Shuai-Shuai Luo Center for Correlated Matter and Department of Physics, Zhejiang University, Hangzhou 310058, China    Zhi-Yong Nie Center for Correlated Matter and Department of Physics, Zhejiang University, Hangzhou 310058, China    An Wang Center for Correlated Matter and Department of Physics, Zhejiang University, Hangzhou 310058, China    Feng Du Center for Correlated Matter and Department of Physics, Zhejiang University, Hangzhou 310058, China    Rui Li Center for Correlated Matter and Department of Physics, Zhejiang University, Hangzhou 310058, China    Michael Smidman Center for Correlated Matter and Department of Physics, Zhejiang University, Hangzhou 310058, China Zhejiang Province Key Laboratory of Quantum Technology and Device, Department of Physics, Zhejiang University, Hangzhou 310058, China    Hui-Qiu Yuan [email protected] Center for Correlated Matter and Department of Physics, Zhejiang University, Hangzhou 310058, China Zhejiang Province Key Laboratory of Quantum Technology and Device, Department of Physics, Zhejiang University, Hangzhou 310058, China State Key Laboratory of Silicon Materials, Zhejiang University, Hangzhou 310058, China
Abstract

The heavy fermion ferromagnet CeRh6Ge4 is the first example of a clean stoichiometric system where the ferromagnetic transition can be continuously suppressed by hydrostatic pressure to a quantum critical point. In order to reveal the outcome when the magnetic lattice of CeRh6Ge4 is diluted with non-magnetic atoms, this study reports comprehensive measurements of the physical properties of both single crystal and polycrystalline samples of LaxCe1-xRh6Ge4. With increasing xx, the Curie temperature decreases, and no transition is observed for xx >> 0.25, while the system evolves from exhibiting coherent Kondo lattice behaviors at low xx, to the Kondo impurity scenario at large xx. Moreover, non-Fermi liquid behavior (NFL) is observed over a wide doping range, which agrees well with the disordered Kondo model for 0.52 \leq xx \leq 0.66, while strange metal behavior is revealed in the vicinity of xcx_{c} = 0.26.

PACS numbers

71.27.+a, 75.40.-s, 72.15.Qm

The highly tunable competition between the Ruderman–Kittel–Kasuya–Yosida (RKKY) interaction and the Kondo effect in heavy fermion systems makes them one of the most widely studied testbeds for quantum critical behavior in condensed matter physics 1si2010heavy ; 2gegenwart2008quantum ; 38High-pressure ; 39Heavyfermionsmagneticfields ; 40TuningtheHeavyFermion ; 47Heavy . Among the non-thermal tuning parameters, dilution of the magnetic Kondo lattice via substitution of non-magnetic atoms is an important method for tuning the magnetic transition 36thompson2011holes ; 37kumar2014kondo . In Ce-based compounds, La doping can be used to dilute the magnetic lattice composed of Ce ions. One interesting example is that of CePd3 7lawrence1996kondo where La doping leads to Kondo hole behavior. The physical properties of the heavy fermion compound CeNi2Ge2 evolves from coherent Fermi liquid (FL) behavior at low La doping, to a non-Fermi liquid (NFL) at intermediate doping levels and local Fermi liquid behavior at high La doping 8pikul2012single ; 9pikul2010lack . A similar evolution of the physical properties is observed in La-doped CeRhIn5 10pagliuso2002site .

However, unlike antiferromagnetic quantum critical points (QCPs) 1si2010heavy ; 2gegenwart2008quantum ; 38High-pressure ; 39Heavyfermionsmagneticfields ; 40TuningtheHeavyFermion ; 47Heavy , where numerous cases have been found experimentally, ferromagnetic (FM) QCPs in clean systems have hardly been observed. Recently, it was found that CeRh6Ge4 belongs to a rare class of clean systems whose FM transition can be continuously suppressed by hydrostatic pressure to reveal an FM QCP 13shen2020strange . In CeRh6Ge4, strange metal behaviors such as a linear-TT resistivity, and a logarithmic divergence of the specific heat coefficient were observed close to the FM QCP 13shen2020strange . A subsequent study of quantum oscillations in CeRh6Ge4 demonstrated that the 4ff electrons remain localized at ambient pressure and hence do not contribute to the Fermi surface 42wang2021localized . An angle-resolved photoemission spectroscopy study at elevated temperatures provided spectroscopic evidence for the presence of anisotropic hybridization between the Ce-4ff and conduction elections in CeRh6Ge4 41wu2021anisotropic , and it was inferred that the anisotropy of the low energy crystalline electric field (CEF) orbitals plays an important role in bringing about the anisotropic cfc-f coupling 43shu2021magnetic . In such a heavy-fermion ferromagnetic material, Ce-site substitution by La will not only suppress the RKKY interaction but also weaken the Kondo effect, which motivates us to study the effects of La doping on the magnetism and the critical behavior of CeRh6Ge4.

In this letter, the synthesis and physical properties of LaxCe1-xRh6Ge4 are reported 11vosswinkel2012bismuth ; 13shen2020strange . Polycrystalline samples were synthesized by arc melting the constituent elements in a stoichiometric ratio. The as-cast samples were wrapped in Ta-foil, sealed in evacuated quartz ampoules and then annealed. Single crystals of LaxCe1-xRh6Ge4 were grown using a Bi flux method 11vosswinkel2012bismuth ; 13shen2020strange . The elements were combined in a molar ratio of LaxCe1-xRh6Ge4:Bi of 1:100, and sealed in an evacuated quartz tube. The tube was heated to 1400 K for 10-13 hours and then cooled to 773 K at 2 K/hour. The ampoules were then centrifuged to remove excess Bi. The remaining Bi on the surface of the crystals was removed by placing them in a 1:1 molar mixture of H2O2 and glacial acetic acid for several days. The crystal structure of the polycrystals was characterized by powder X-ray diffraction (XRD) using a Rigaku Ultima IV diffractometer with Cu-Kα1 radiation. The composition of all doped samples was checked by energy-dispersive X-ray spectroscopy which was carried out using a Hitachi SU-8010 scanning electron microscope. Electrical transport and specific heat measurements were performed using a Quantum Design Physical Property Measurement System (PPMS).

Refer to caption
Figure 1: (a) The XRD patterns of powdered polycrystals with xx = 0, 0.05, 0.10, 0.25, 0.40, 0.60. The peak positions corresponding to CeRh6Ge4 are marked by the red vertical lines. (b) Enlargement of the strongest Bragg peak, in which the peak position shifts to lower angle with increasing xx. (c) The lattice constants aa, cc and the unit cell volume VV of LaxCe1-xRh6Ge4 versus xx. (d) The crystal structure of the parent compound, CeRh6Ge4.

XRD patterns of the powdered polycrystalline samples are shown in Fig. 1(a). Here, the peak positions are compared to those expected for stoichiometric CeRh6Ge4, which has the LiCo6P4 type crystal structure 14buschmann1991darstellung . All doped compositions shown in Fig. 1(a) have very similar Bragg positions to that of CeRh6Ge4 [P6¯\bar{6}m2 (No. 187, D3h1{}^{1}_{3h})], suggesting that the doped samples also belong to the same crystal structure. After annealing, only a few very small peaks corresponding to impurity phases could be resolved. In Fig. 1(b), the peaks with the highest intensity are displayed for all compositions, demonstrating that the peaks shift to lower angles as the La doping level is increased, suggesting the expansion of the lattice. Figure 1(c) displays the variation of the lattice parameters aa, cc and the volume of the unit cell VV as a function of La doping.

Refer to caption
Figure 2: Temperature dependence of the normalized resistivity for (a) polycrystalline Ce-rich compositions, and (b) La-rich single crystals. The inset is an enlargement for clarity. (c) Temperature variation of the magnetic contribution per Ce atom to the resistivity (ρmag\rho_{mag}) of all samples, where ρmag\rho_{mag}=(ρxρLa\rho_{x}-\rho_{La})/(1x-x). Data for polycrystals and single crystals are represented by open triangles and filled circles, respectively. The inset is an enlargement showing the \simlog(1/T1/T) behavior of ρmag(T)\rho_{mag}(T) for xx = 0.66 and 0.87.

The temperature dependence of the resistivity normalized by the value at 300 K is shown in Figs. 2(a) and (b). Note that due to the poor quality of La-rich polycrystalline samples, for xx >> 0.30 the data for single crystals is displayed. It is clear that the two sides of the series exhibit qualitatively different behaviors. Because of the disorder effect induced by elemental substitutions, the residual resistivity increases with doping on the Ce-rich side whereas, it gradually decreases for the La-rich side. The resistivity of La0.66Ce0.34Rh6Ge4 and La0.87Ce0.13Rh6Ge4 exhibits an upturn below 30 K. To study the doping evolution of the magnetic contribution to the resistivity per Ce (ρmag\rho_{mag}), the resistivity of LaRh6Ge4 was subtracted from the doped samples, which is defined as ρmag\rho_{mag}=(ρxρLa\rho_{x}-\rho_{La})/(1x-x), and is displayed in Fig. 2(c) for all compositions. In order to make sure that there is no fundamental difference between the physical properties of single and polycrystalline samples at the same composition, single crystals corresponding to low doping levels (xx = 0.13 and 0.24) were also grown and characterized. It is evident from the data of Ce-rich crystals (xx = 0.13 and 0.24), that the behavior is highly consistent between polycrystalline and single crystalline samples [see Fig. 2(c)].

All curves in Fig. 2(c) exhibit a maximum near 100 K, which has also been observed in stoichiometric CeRh6Ge4 13shen2020strange . Above the high temperature maximum (TmaxT_{max}), the ρmag(T)\rho_{mag}(T) of all samples behaves similarly, except for the value of the TmaxT_{max}, which shifts to slightly lower temperature with increasing xx. The TmaxT_{max} can be related to the combined effects of the low lying CEF excitations and the Kondo scattering. On the contrary, the behavior of ρmag(T)\rho_{mag}(T) below TmaxT_{max} exhibits a strong composition dependence. In the Ce-rich region (xx \leq 0.30), ρmag\rho_{mag} decreases significantly with decreasing temperature as a result of coherent Kondo lattice behavior. For xx = 0.52, after a slight decrease, the ρmag\rho_{mag} becomes almost temperature independent down to 0.5 K while for xx = 0.66 and 0.87 there is a logarithmic divergence [ρmag\rho_{mag} \sim log(1/TT)] from 20 K to 10 K, as shown in the inset. The low temperature log(1/TT) behavior is attributed to incoherent Kondo scattering from the fully occupied CEF ground state, as is typical for Kondo impurity systems 3bauer1991anomalous ; 22cox1988transport , in contrast to the coherent scattering in the case of a well ordered Kondo lattice. In general, as xx increases, LaxCe1-xRh6Ge4 evolves from Kondo lattice behaviors in the Ce-rich region to the Kondo impurity regime in the La-rich region. This also can be observed in other dilute Kondo systems, such as CexLa1-xCu6 20onuki1985kondo , CexLa1-xNi2Ge2 8pikul2012single ; 9pikul2010lack and CexLa1-xTiGe3 44lee2019suppression ).

Refer to caption
Figure 3: Low temperature resistivity (a) and specific heat (b) for Ce-rich samples. Data for polycrystals and single crystals are represented by open triangles and filled circles, respectively. In (a), the resistivity curves are vertically shifted by 0.2 μΩ\mu\Omega cm for clarity. The black dashed lines represent a linear fit to the data and the red solid curves show fits to a quadratic temperature dependence for a Fermi liquid, ρ(T)\rho(T) \sim T2T^{2}. The inset shows the derivative dρ\rho/dTT for samples with doping concentration xx \leq 0.25, where the peak position corresponds to TCT_{C}. In (b), solid arrows mark the positions of TCT_{C}, while the dashed arrows highlight the existence of a magnetic transition for the xx = 0.24 and 0.25 samples, while no transition is detected in C/TC/T for xx = 0.30.

The low temperature resistivity is shown in Fig. 3(a), and the first derivative of the resistivity dρ\rho/dTT is displayed in the inset, which is utilized to define the Curie temperature TCT_{C}. A pronounced transition in dρ\rho/dTT [Fig. 3(a)] and the specific heat [Fig. 3(b)] is observed for compositions up to xx = 0.15, while for xx = 0.24 and 0.25 just an onset of magnetic transition is visible as a small hump in C/TC/T [Fig. 3(a)] and as an upturn in dρ\rho/dTT [Fig. 3(b)], but the determination of the Curie temperature TCT_{C} for these compositions requires further measurements down to lower temperatures. In Fig. 3, the behavior of C/TC/T at low doping levels is similar for polycrystals and single crystals (xx = 0.13 and 0.24), again showing that the properties of single crystals and polycrystalline samples are similar. Since the interaction between local magnetic moments is responsible for the ferromagnetic order in CeRh6Ge4, weakening of the RKKY interaction by La doping is the primary reason for the suppression of Curie temperature. A transition is not visible for the xx = 0.30 composition down to 0.3 K, and instead, the resistivity decreases linearly with decreasing temperature [Fig. 3(a)], while C/TC/T increases [Fig. 4(b)]. For xx \leq 0.3, the resistivity in the paramagnetic state shows non-Fermi liquid behavior which is manifested in the linear-TT dependent behavior of the resistivity (marked with black dashed lines), whereas in the magnetic state a quadratic dependence [ρ(T)\rho(T)=ρ0{\rho_{0}}+AAT2{T^{2}}, marked with solid red lines] suggestive of coherent Fermi liquid behavior is observed.

Refer to caption
Figure 4: (a) Low temperature resistivity for xx \geq 0.52. The data are all shifted vertically by 0.07  μΩ\mu\Omega cm for clarity, except for xx = 0.52 which is shifted by 1.05 μΩ\mu\Omega cm. The black dashed lines highlight a linear increase of ρ(T)\rho(T) with the decreasing temperature. The red solid curve shows the fit of ρ(T)\rho(T) to the local Fermi liquid behavior [ρ\rho \sim (ρ0α\rho_{0}-\alphaT2T^{2})] below 2.5 K for xx = 0.87. For xx = 1, the slight increase of ρ(T)\rho(T) at low temperatures is likely attributed to a tiny amount of magnetic impurities. (b) Low temperature C/TC/T for xx \geq 0.30 on a logarithmic temperature scale. Data for polycrystals and single crystals are represented by open triangles and filled circles, respectively. The log(1/T1/T) behavior highlighted by the black dashed lines is observed in the doping range xx = 0.30 - 0.66.

In Fig. 4(a), low temperature transport and thermodynamic measurements of La-rich alloys are displayed. The black dashed lines are used to highlight the NFL behavior, where ρ\rho \sim (ρ0AT\rho_{0}-A^{\prime}T) down to 0.3 K for xx = 0.52 and 0.66 but only down to 2.5 K for xx = 0.87. For the latter, the data below 2.5 K can be fitted with ρ\rho~{}\sim (ρ0α\rho_{0}-\alphaT2T^{2}), which is characteristic of incoherent (local) Kondo systems 25nozieres1974fermi . Therefore, similar to diluted CexLa1-xNi2Ge2 9pikul2010lack , this diluted sample (xx = 0.87) exhibits local FL behavior below 2.5 K, suggesting a weak intersite coupling between the local 4ff moments.

In Fig. 4(b), the logarithmic temperature dependence (as shown by black dashed lines) of the specific heat [C/TC/T \sim log(1/T1/T)] for xx = 0.3, 0.52 and 0.66 is displayed, while C/TC/T has no upturn and is almost temperature independent below 5 K for xx = 0.87. The behavior of the latter is similar to that observed in LaRh6Ge4, which is a paramagnetic metal without the Kondo effect. Hence, this temperature-independent behavior of C/TC/T implies the weak Kondo screening of local moments in the dilute limit 26oliveira1981specific . The logarithmic divergent C/TC/T along with the TT-linear resistivity [ρ\rho \sim TT, as shown in Fig. 3(a)] for xx = 0.30 is similar to the strange metal behavior seen under pressure in CeRh6Ge4 13shen2020strange . For 0.66 \geq xx \geq 0.52, the samples exhibit NFL behavior with ρ\rho \sim (ρ0AT\rho_{0}-A^{\prime}T) and C/TC/T \sim log(1/T1/T), which can be described within the framework of the disordered Kondo model 32miranda1996kondo ; 28de1998evidence .

Refer to caption
Figure 5: (a) Concentration dependence of the specific heat coefficient γ0.35K\gamma_{0.35\text{K}} (as C/TC/T at 0.35 K). (b) TT-xx phase diagram of LaxCe1-xRh6Ge4. The open squares and triangles denote TCT_{C} derived from the specific heat and dρ\rho/dTT, respectively. The black dashed line is the extrapolation of the initial slope of TCT_{C}(xx), which reaches zero temperature at xcx_{c} \approx 0.26. The shaded blue region corresponds to coherent Fermi liquid behavior and ferromagnetic order at low doping, while the local FL in the dilute limit is in yellow. NFL behavior is observed over a wide doping range (0.3 \leq xx \leq 0.66).

In Fig. 5(a), γ0.35K\gamma_{0.35\text{K}} which is defined as C/TC/T at 0.35 K, is plotted against the doping concentration xx. A pronounced maximum is observed at the composition xx \approx 0.24. It is noted that the values of γ0.35K\gamma_{0.35\text{K}} for xx = 0.24 and 0.25 are taken from the points near the magnetic transition, which exceed the Sommerfeld coefficient at zero temperature. On the other hand, γ0.35K\gamma_{0.35\text{K}} near xcx_{c} = 0.26 is underestimated in comparison with the corresponding Sommerfeld coefficient. Therefore, the Sommerfeld coefficient is likely peaked near the critical concentration where the ferromagnetic order is suppressed. In Fig. 5(b), the temperature-composition phase diagram derived from all the measurements is displayed. The transition temperatures determined from the specific heat and resistivity are in good agreement. For dense Kondo systems with low La-doping (xx \leq 0.15), coherent FL behavior is observed in the FM state, while for dilute Kondo alloys (xx \geq 0.87), the low-TT behavior can be described as a local Fermi liquid, represented by the yellow region. Similar to CexLa1-xNi2Ge2 8pikul2012single ; 9pikul2010lack ), these two FL regimes are separated by a NFL region (0.25 << xx \leq 0.66), wherein the area near xcx_{c} = 0.26 shows behaviors similar to the strange metal phase observed in the case of pressured induced FM QCP 13shen2020strange , while at higher doping the behavior corresponds to the disordered Kondo model.

At low doping levels (xx \leq 0.15), the evolution of magnetism in LaxCe1-xRh6Ge4 is similar to that of another dilute Kondo ferromagnet CexLa1-xTiGe3, where TCT_{C} is linearly suppressed as xx increases 44lee2019suppression . If it is assumed that TCT_{C} continues to decrease linearly with xx, it will be completely suppressed near xcx_{c} = 0.26, as highlighted by the black dashed line in Fig. 5(b), which is close to where there is a maximum in the Sommerfeld coefficient. It is noted that a ferromagentic transition still survives for xx = 0.24 and 0.25 with an onset transition above 0.5 K. However, these two points are not shown in the phase diagram since their transition temperatures are below the temperature limit of measurements. It is obvious that Ce/La substitutions in CeRh6Ge4 result a rich phase diagram. Further measurements are badly required in order to establish the existence of a FM QCP induced by chemical doping and understand its unusual behavior.

To conclude, we have studied single and polycrystalline samples of La-doped CeRh6Ge4. The effect of La doping is to increase the lattice volume and dilute the Kondo lattice, leading to a crossover from coherent Kondo lattice behaviors at low doping levels, to the Kondo impurity regime at higher doping. La doping also leads to suppression of the long-range magnetic order, resulting in NFL behavior. Based on the physical property measurements, we construct a phase diagram to map the evolution of the low temperature properties with La-doping. With increasing La-concentration, the ferromagnetic order is suppressed near xcx_{c} = 0.26, at which pronounced strange metal behaviors with ρ\rho \sim TT and C/TC/T \sim log(1/T)(1/T) are observed. At higher doping (0.52 << xx \leq 0.66), the NFL behavior deviates from the quantum critical behaviors of pressure-tuned CeRh6Ge4 13shen2020strange , and instead can be described in terms of a disordered Kondo model. In contrast to the absence of NFL behavior in CexLa1-xTiGe3 44lee2019suppression , the observation of strange metal behavior in LaxCe1-xRh6Ge4 near the critical concentration reflects the presence of strong quantum fluctuations, indicating the need for further studies of the nature of the ground state in this region.

Acknowledgements.
This work was supported by the National Natural Science Foundation of China (No. 12034017, and No. 11974306), the National Key R&\&D Program of China (No. 2017YFA0303100 and No. 2016YFA0300202) and the Key R&\&D Program of Zhejiang Province, China (2021C01002).

References

  • (1) Si Q M and Steglich F 2010 Science 329 1161
  • (2) Gegenwart P, Si Q M and Steglich F 2008 Nature Phys. 4 186
  • (3) Chen Y, Weng Z F, Smidman M, Lu X and Yuan H Q 2016 Chinese Physics B 25 077401
  • (4) Smidman M, Shen B, Guo C Y, Jiao L, Lu X and Yuan H Q 2019 Chinese Physics B 28 017106
  • (5) Xia X B, Shen B, Smidman M, Chen Y, Lee H and Yuan H Q 2018 Chinese Physics Letters 35 067102
  • (6) Xie W, Shen B, Zhang Y J, Guo C Y, Xu J C, Lu X and Yuan H Q 2019 Acta Physica Sinica 68 177101
  • (7) Thompson J D 2011 Proceedings of the National Academy of Sciences 108 18191
  • (8) Kumar P and Vidhyadhiraja N S 2014 Physical Review B 90 235133
  • (9) Lawrence J M, Graf T, Hundley M F, Mandrus D, Thompson J D, Lacerda A, Torikachvili M S, Sarrao J L and Fisk Z 1996 Physical Review B 53 12559
  • (10) Pikul A P, Stockert U, Steppke A, Cichorek T, Hartmann S, Caroca-Canales N, Oeschler N, Brando M, Geibel C and Steglich F 2012 Physical Review Letters 108 066405
  • (11) Pikul A P, Geibel C, Oeschler N, Macovei M E, Caroca-Canales N and Steglich F 2010 Physica Status Solidi (b) 247 691
  • (12) Pagliuso P G, Moreno N O, Curro N J, Thompson J D, Hundley M F, Sarrao J L, Fisk Z, Christianson A D, Lacerda A H, Light B E et al 2002 Physical Review B 66 054433
  • (13) Shen B, Zhang Y J, Komijani Y, Nicklas M, Borth R, Wang A, Chen Y, Nie Z Y, Li R, Lu X et al 2020 Nature 579 51
  • (14) Wang A, Du F, Zhang Y J, Graf D, Shen B, Chen Y, Liu Y, Smidman M, Cao C, Steglich F et al 2021 Science Bulletin 66 1389
  • (15) Wu Y, Zhang Y J, Du F, Shen B, Zheng H, Fang Y, Smidman M, Cao C, Steglich F, Yuan H Q et al 2021 Physical Review Letters 126 216406
  • (16) Shu J W, Adroja D T, Hillier A D, Zhang Y J, Chen Y X, Shen B, Orlandi F, Walker H C, Liu Y, Cao C et al 2021 arXiv:2102.12788v1 [cond-mat.str-el]
  • (17) Voßwinkel D, Niehaus O, Rodewald U C and Pöttgen R 2012 Zeitschrift für Naturforschung B 67 1241
  • (18) Buschmann R and Schuster H U 1991 Zeitschrift für Naturforschung B 46 699
  • (19) Bauer E 1991 Advances in Physics 40 417
  • (20) Cox D L and Grewe N 1988 Zeitschrift für Physik B Condensed Matter 71 321
  • (21) Ōnuki Y, Shimizu Y, Nishihara M, Machii Y and Komatsubara T 1985 Journal of the Physical Society of Japan 54 1964
  • (22) Lee J, Rabus A, Coutts C and Mun E 2019 Physical Review B 99 045135
  • (23) Nozieres P 1974 Journal of low température physics 17 31
  • (24) Oliveira L N and Wilkins J W 1981 Physical Review Letters 47 1553
  • (25) Miranda E, Dobrosavljevic V and Kotliar G 1996 Journal of Physics: Condensed Matter 8 9871
  • (26) De Andrade M C, Chau R, Dickey R P, Dilley N R, Freeman E J, Gajewski D A, Maple M B, Movshovich R, Neto Castro A H, Castilla G et al 1998 Physical Review Letters 81 5620