Neutron scattering study of CeNiSn

PHYSICA[

Physica B 186-188 (1993) 409-411 North-Holland

Neutron scattering study of CeNiSn M. Kohgi a, K. Ohoyama a, T. Osakabe a, M. Kasaya a, T.

T a k a b a t a k e b a n d H . Fujii b

"Department of Physics, Tohoku University, Sendai 980, Japan bFaculty of Integrated Arts and Science, Hiroshima University, Hiroshima 730, Japan Neutron scattering studies on the gap-type Kondo system CeNiSn as well as on its reference materials CePdSn and CePtSn are reported. CeNiSn shows a Lorentzian-like broad magnetic response centered at E = 0 with an anomalous peak at about 4 meV. Well-defined crystal field excitations have been observed for CePdSn and CePtSn.

1. Introduction The e-TiNiSi-type compound CeNiSn attracts attention because it shows strong anomalies in magnetic and transport properties below about 10 K which are characterized by the formation of a gap of 2-6 K at the Fermi energy [1,2]. In a previous paper [3], we reported that the inelastic neutron scattering spectrum of CeNiSn is quite anomalous compared with that of the reference material CePdSn. The spectrum, which was observed with a chopper spectrometer, is rather structureless like the tail of a Lorentzian function above about 5 meV. However, the detailed structure of the spectrum below 5 meV was not clear because of the resolution limitations of the spectrometer. The main purpose of the present paper is to report the result of a high resolution neutron scattering study on CeNiSn to obtain more precise information on the magnetic response at the low energy transfer region. We also report the result of a neutron scattering measurement on another reference material, CePtSn. This material shows quite similar magnetic properties to those of CePdSn [4], but has been suggested to have a much stronger crystal field [5].

2. Experimental details

there are no appreciable phases other than the target material in the samples. The high resolution neutron scattering experiments on CeNiSn were performed on the LAM-D and LAM40 crystal analyzer spectrometers at the pulsed neutron source KENS in the National Laboratory for High Energy Physics. Both spectrometers employ several large solid-angle pyrolytic graphite energy-focusing type analyzers, each with its own berryllium filter and detector and with final energy of 4.6 meV. LAM-D views an ambient temperature water moderator, while LAM-40 views a solid methane moderator. Thus, the former has a high sensitivity at the energy transfer of the thermal energy region, the latter at the low energy region, but with same resolution of 0.35 meV (FWHM) at E = 0. The inelastic neutron scattering experiments on CePtSn were performed on the chopper spectrometer INC at KENS. The experimental condition is same as those for the previous experiments [3]; the incident neutron energy was 60meV, and the resolution (FWHM) is 3.3 meV at E ( = h~o)= 0. The raw data were corrected for monitor normalization, background subtraction, absorption correction and detector efficiency, and put to the absolute scale by using the data obtained for a vanadium standard sample. The phonon contribution in the raw data was estimated from the high scattering angle (-120 °) data and subtracted from the low angle data as described previously [3].

Polycrystalline samples of CeNiSn and CePtSn of about 60 g each were used in the experiments. They were made by arc melt. The CeNiSn sample was a part of the same sample that was used in a previous work [3]. The X-ray powder diffraction analysis showed that

3. Results

Correspondence to: M. Kohgi, Department of Physics, Tohoku University, Sendai 980, Japan.

First, we show the result of the INC experiment on CePtSn together with those of CeNiSn and CePdSn.

0921-4526/93/$06.00 © 1993 - Elsevier Science Publishers B.V. All rights reserved

M. Kohgi et al. / Neutron scattering study of CeNiSn

410

Figures l(a), (b) and (c) show the inelastic magnetic response of CeNiSn, CePdSn and CePtSn, respectively, at about 20K taken in the low angle bank of detectors (20 = 5-12°). The phonon contribution is subtracted from the raw data. The insets in the figures show the raw data and the estimated phonon contribution for each compound. Well-defined crystal field excitations were observed for CePtSn as in the case of CePdSn. The solid curves in figs. l(b) and l(c) show the best fits to the data using a sum of two Lorentzians as the spectral function. The excitation energies obtained for CePdSn are 17.4 and 25.8 meV with line widths (HWHM) of 2.2 and 2.7meV, while those for CePtSn are 23.6 and 36.6 meV with line widths of 2.0 and 3.6 meV. As cited above, the spectrum of CeNiSn is rather structureless above about 5 meV. The solid curve in fig. l(a) shows the best fit to the data using a single Lorentzian centered at E = 0 as the spectral function. The fit gives a width of 2.2 meV. Figure 2 shows the spectra of CeNiSn at 3 and 15 K

i

i

i

CeNiSn

19 K

\

o I.

~

s

-c ~

4

~

2 0

~D 10

/

. . . . aata"]

V~" " ~ i ~

Ei = 60 meV

k;7,

10

I

s 13.....~ V~°"°n3

o

^

(a)

10

20

30

40

i

i

i

u

CePdSn 16K

l

,'0 = - : ,

o ~bs~ =

u

u

~,,,

INC

~k

4

i

101.~

. ~e~

'[~.' ~

,

- 50- - -

60

i

,

n

]

0.6

o.s

u

-.-7.

~ P:+.

= 0.4

['~

0.3

u

u

~

o 3K

~o~'*

•~ + ~ "

10

20

30

40

50

I

i

u

i

u

CePtSn 18K

,~

t0~ [~

~ 05

'

60

LAMD

~"~

0.2

E f = 4.6 meV

0.1 0.0

I 10

I 20

I 30

20

30

obtained with the LAM-D spectrometer at the scattering angle of 35°. The raw data were corrected for the wavelength variation of the incident spectrum, detector efficiency and absorption effect. No correction was done for the phonon contribution. The result shows that the spectrum below about 10 meV is not a monotonic function of energy but has a peak at about 4 meV which is strongly temperature dependent and cannot be seen in the INC data at 19 K. The features of the spectrum above about 10 meV are not so different from the raw data of the INC experiment (see the inset in fig. l(a)). The apparent difference between the two spectra is due to the difference of the scattering triangles, of phonon contribution and of the resolution. Figure 3 shows the response functions of CeNiSn at 3, 15 and 50 K at small energy transfer obtained with the LAM-40 spectrometer. After the same correction as in the case of the LAM-D data, the raw data, which

[=~

~

40

- SO

60

(meV)

Fig. I. Magnetic responses of (a) CeNiSn, (b) CePdSn and (c) CePtSn at about 20 K observed with the INC spectrometer. T h e solid curves indicate the results of the least squares fitting to the data using Lorentzian(s) as the spectral function. Insets show the raw data and the estimated p h o n o n contribution for each compound.

60

Fig. 2. Spectra of CeNiSn observed with the L A M - D spectrometer.

(e)

Energy

I 50

Energy (meV)

-"~,= ~

10

I 40

u

1.5

~=~

~

1.0 : . ~ & +~~

. 0 . 5 ......~

0.0

I

eNiSn

~

-~ /

u

CeNiSn

m

0

n

50K

. . . . phonon

"\~.~ ~ ' ~ ~X,-~ A ~ ^

I 2

LAM-40 E f = 4.6 meV 20 = 24° " 72°

^

4

6

Energy (meV) Fig. 3. Spectra of CeNiSn observed with the LAM-40 spectrometer. T h e thicker solid, dotted and chained lines indicate the results of the Lorentzian fit to the data between 0.7 and 2.5 meV at 3, 15 and 50 K, respectively. Broken line represents the estimated p h o n o n contribution.

M. Kohgi et al. / Neutron scattering study of CeNiSn are the sum of the counts at the 24 ° , 40 ° , 56 ° and 72 ° detectors, were corrected for the kr/k i term and the factor hto/[1-exp(-hto/3)] in the cross-section to obtain the response function (=Im[x(Q, to)/to]+ phonon background). In this correction, we neglected the resolution effect because it is small enough for the present purpose. The data at energy transfers less than 0.7 meV are omitted in the figure because of the strong contamination of the incoherent elastic scattering. The broken line in the figure indicates the phonon contribution estimated from the measurement on LaNiSn with the same experimental conditions as the measurement on CeNiSn. It is clear that the phonon contribution is small in this energy region. The overall height of the observed magnetic response function decreases with increasing temperature. This is qualitatively consistent with the behavior of the reported temperature dependence of the bulk susceptibility. Each spectrum below about 2.5 meV increases monotonically with decreasing energy transfer though a small hump is seen around 1.1 meV. The solid, dotted and chained lines in the figure show the best fit Lorentzian functions centered at E = 0 to the data between 0.7 and 2.5meV at 3, 14 and 50K, respectively. The agreement is rather good in the energy region. The widths of the Lorentzians were 1.6, 2.1 and 3.0 meV, respectively. Note that the width of 2.1 meV at 15 K is comparable with the result of the Lorentzian fit to the INC data at 19 K which gives a width of 2.2 meV. However, the spectra show more complicated features at higher energy transfer. A t 3 K, the spectrum starts to deviate from the Lorentzian above about 2.5 meV and shows a broad peak around 4 meV. At 15 K, the spectrum also shows a broad peak around 4.5-7 meV: however, the intensity around 4meV is considerably smaller than that at 3 K, whereas the intensity around 7 meV is the same as that at 3 K. A t 50 K, no peak is seen around 4meV. Instead, the spectrum starts to deviate from the Lorentzian around that energy.

4. Discussion The experimental results demonstrate that the 4f electron state in CeNiSn is quite different from those of CePdSn and CePtSn. In the latter compounds, two well-defined crystal field excitations were observed,

411

and this indicates that the 4f electron is well localized. The reason for the rather large difference in the crystal field splitting between these compounds is not known. No well-defined crystal field excitations were observed for CeNiSn. Instead, it shows a Lorentzian-like broad response centered at E = 0 with a broad peak at about 4meV. The temperature dependence of the peaks is quite anomalous. Similar anomalous spectra for CeNiSn were also reported by other groups [2,6] but the agreement with our results is not satisfactory. The magnetic response might be understood as that of over-dumped crystal field excitations. But, it is difficult to explain the temperature dependence of the peak around 4 meV by any simple model. We suggest that the anomalous temperature dependence of the peak around 4-7 meV corresponds to the development of a short range magnetic correlation at the low temperatures.

Acknowledgements The authors would like to thank T. Otomo for assistance with the experiments at the INC spectrometer, and T. Kamiyama and K. Shibata for assistance with the experiments at the L A M - D and LAM-40 spectrometers.

References [1] T. Takabatake, F. Teshima, H. Fujii, N. Nishiguri, T. Suzuki, T. Fujita, Y. Yamaguchi, J. Sakurai and D. Jaccard, Phys. Rev. B 41 (1991) 9607; T. Takabatake and H. Fujii, Physical Properties of Actinide and Rare Earth Compounds, ed. T. Kasuya (Jpn. J. Appl. Phys., Tokyo, 1993) JJAP Series 8, p. 254 and references therein. [2] F.G. Aliev, V.V. Moshchalkov, M.K. Zalyalyutdinov, G.I. Pak, R.V. Scolozdra, P.A. Alekseev, N.N. Lazukov and I.P. Sadikov, Physica B 163 (1990) 358. [3] M. Kohgi, K. Ohoyama, T. Osakabe and M. Kasaya, J. Magn. Magn. Mater. 108 (1992) 187. [4] M. Kasaya, T. Tani, H. Suzuki, K. Ohoyama and M. Kohgi, J. Phys. Soc. Jpn. 60 (1991) 2542. [5] T. Takabatake, H. Iwasaki, G. Nakamoto, H. Fujii, H. Nakkote, F.R. de Boer and V. Sechovsky, Physica B 186-188 (1993) 734. [6] G. Appli, E. Bucher and T.E. Mason: Physical Phenomena at High Magnetic Fields (Addison-Wesley, Reading, MA, 1992) p. 175.