The importance of paramagnetic impurities to the nuclear magnetic resonance relaxation of ion-conducting glasses

SOLID STATE Nuclear Magnetic Resonance Solid State Nuclear Magnetic Resonance 5 (1995) 145-150

The importance of paramagnetic impurities to the nuclear magnetic resonance relaxation of ion-conducting glasses M. Griine ‘, W. Miiller-Warmuth

*

Institut fiir Physikalische Chemie der Westfiiischen Wilhelms-Universitiit, Schlossplatz 4 /’ 7, D-48149 Miinster, Germany

Received 23 September 1994; revised 23 February 1995; accepted 23 February 1995

Abstract

Paramagnetic impurities have been shown to affect the ‘Li nuclear magnetic resonance relaxation rates in cation-conducting glasses, and wrong data may then be extracted from the experiments. Frequency- and temperature-dependent T;’ studies of lithium borate and thioborate glasses revealed that iron impurities cause frequencyindependent relaxation and “shoulders” of the low-temperature slopes in high fields, whereas manganese produces enhanced relaxation peaks. The results of the T;’ (and some TIQ1 and T; ‘> measurements are discussed. Keywords:

Lithium

‘Li nuclear magnetic resonance; Glasses, borate; Lithium thioborate; Iron; Manganese

1. Introduction Temperatureand frequency-dependent studies of nuclear magnetic resonance (NMR) relaxation rates have often been used to investigate the cation dynamics of glasses. Examples are the ‘Li NMR in lithium silicate, borate and phosphate glasses 111, ‘Li NMR in lithium borate glasses with lithium halides as dopant species [2,3] and in lithium aluminosilicate glasses [4], ‘Li NMR in lithium thioborate [.5], thiosilicate [6,7] and thiogermanate systems [8], as well as 23Na [9,10] and ‘09Ag NMR studies [ll-141 of various glasses. Most studies employed spin-lattice relax-

’ Present address: Institut fiir Molekulare Biotechnologie, D-07708 Jena, Germany. * Corresponding author.

ion-conducting;

Paramagnetic

impurities;

Relaxation

times;

ation rate (T;‘) data to obtain information on the spectral density; occasionally the relaxation frame, the spin-spin time TIP in the rotating relaxation time T2 or the relaxation of the spinspin dipolar energy (T,,) was measured as well. The results revealed important differences from laboratory to laboratory and also within the same laboratory. Reasons are the various preparation techniques including glassmaking procedures, thermal history, and the materials used. It is the purpose of this communication to present new 7Li NMR T, (and some TIP and T2) measurements of glasses in the systems Li,O-B,O,Li,Cl, and Li,S-B,S, which elucidated the importance of paramagnetic impurities to the relaxation. In contrast to crystalline materials, for glasses the frequency and temperature dependence of the relaxation rates is already affected by very small amounts of iron and manganese

0926-2040/95/$09.50 0 1995 Elsevier Science B.V. All rights reserved SSDI 0926-2040(9.5)00025-9

146

M. Griine, W Miiller- Warmuth /Solid

State Nuclear Magnetic Resonance 5 (1995) 145-150

which enter during the preparation either from the chemicals or from the application of twin roller quenching, etc. We realized that the magnitude of the relaxation maximum and the characteristic low-temperature behaviour is particularly affected. Even the “shoulders” on the low-temperature side observed in high magnetic fields [5,15] are connected with impurities. In the light of our results, relaxation data of glasses have to be analyzed very carefully to obtain correct information on the cation dynamics.

2. Experimental

and results

Six glasses whose analysis composition is given in Table 1 were prepared using normal glassmaking procedures; for the sulfide-based glasses, cf. Ref. [5]. One sample of each series (01 and Sl) was prepared with special care. Apart from the utilization of primary products of the highest purity the application of roller quenching was avoided. Without particular attention, impurities are introduced into the glass first of all by impurity-containing boron and by contact with metals. Glass no. 03 contains a well-defined amount of manganese(I1) oxide which was added intentionally. The glass formation was determined by differential scanning calorimetry (DSC) and powder X-ray diffraction. Table 1 contains in addition to the composition and the glass temperature Tg the concentration of iron and manganese ions as determined by inductively coupled plasma atomic emission spec-

Table I Composition, glass transition temperature thioborate glasses studied in this work

trometry (ICP-AES). Mn2+ was also identified by electron paramagnetic resonance CEPR). The ‘Li NMR T, measurements were carried out at 116.64 MHz (7.05 T) and 16.05 MHz (0.97 T) with a Bruker FT-CXP 300 spectrometer in combination with an electromagnet. 180” -7-90” or n 90”-~-90” and sometimes 90”,-T’-64”,(90”,) echo sequences (instead of the 90” detection pulse) were applied. 90” pulses had a width of 4-5 ps, and T was varied appropriately. For most samples exponential time dependences of the nuclear magnetization M,(t) were found. Exceptions are glass no. S2 with a relatively large iron content and glass no. 02 for temperatures below 200 K. For these glasses less accurate Tl values are given, estimated from the relaxation decay of [M,(T = 0) - M,(c~)l to l/e of its value. In the other cases the accuracy of the measurements amounts to about f5%. In addition to T, we present some T,, and T2 data obtained at 117 MHz. TIP was determined as usual by a 90:-(locking pulse), sequence, and T2 by 90” -7-180” -T echo sequences. At low temperatures the magnetization kf,( 7) approaches its equilibrium value as exp[ - (t/T,,>0.5] indicating the absence of spin diffusion [16,17]. At high temperatures the time dependence of Alp(~) showed deviations from mono-exponentiality, which were attributed to diffusion-limited relaxation [18]. Within the limits of the experimental errors T2 was not frequency dependent. All the measurements were carried out betweep temperatures below T, and about 100 K. Four different home-made nitrogen flow-cryo-

and (total) iron and manganese

Glass no.

Composition in mol%

T, K)

01 02 03 Sl s2 s3

28Li,O-56B,O,-16Li,C1, 28Li,O-60B,O,-12Li,C1, 28Li,O-55B,O,-17Li,C1, 46Li,S-54B,.S, 46Li,S-54B,S, 46Li,S-S4B,S3

690 707 682 693

concentration

(from analysis) of the borate and

Concentration in 10m4 mol% Fe

Mn

< 20 68 < 20 < 20 560 110

<2 <2 124 12 10 70

M. Griine, W. Miiller- Warrnuth /Solid State Nuclear Magnetic Resonance 5 (1995) 145-150

stats and high-temperature probeheads were used to control the temperature at the two frequencies and the various ranges. The error was in between & 1 K and +3 K depending on the particular set-up. Fig. 1 shows the spin-lattice relaxation rates of the three oxidic glasses plotted versus reciprocal temperatures. For the purpose of clarity the 16 MHz data of glass no. 01 are omitted, since Fig. 2 gives a complete view of all results for this purest oxide glass. As compared with glasses nos. 01 and 02 the T;’ maxima of glass no. 03 and the relaxation rates of the slopes on the right appear to be four or five times larger. Glass no. 02, on the other hand, has similar relaxation maxima as glass no. 01, but there are the convexities or “shoulders” in the 117 MHz curve which were already previously observed for conventionally produced glasses [5,15]. This latter glass re-

500 T-‘,

-L 5-l-:

’

200

300

147 T/K

150

$@$g

+A

d

d

,

'I

=

f

0 ?o 00

moo

0.17 4

'*.

, 2

,

, L

,

l *.

0

l a

Ooo

, 6

o

,

.

0

0

,

,

0

1031T

. K“

500 $

;

300

200

150 i

T/K

,

#hiA, .

Fig. 2. Experimental ‘Li NMR relaxation rates T;’ with r = 1, 2, lp of the oxidic glass no. 01. From top to bottom: (A) T; -’ extracted from the central transition linewidth, (~)T;‘,(O)Tlp’at50kHz,(o)T;‘at16MHz,(o)T;’at 117 MHz.

veals in addition a rather strong temperature-independent relaxation part at low temperatures. For glass no. 01, in addition to the “true” relaxation rate T;’ measured by the appropriate pulse sequence, relaxation rates TX*_ ’ derived from the linewidths of the central transition of the ‘Li NMR spectra are also plotted in Fig. 2. For this, the signal was either simulated by a Gaussian (T < 260 K) or by a Lorentzian (T > 260 K).

2

1 4

I

81 6

1 6-

" 103/T K-1

Fig. 1. Experimental ‘Li NMR spin-lattice relaxation rates of the oxidic glasses plotted versus reciprocal temperatures. From top to bottom: (A) glass no. 03 at 16 MHz, ( W) glass no. 02 at 16 MHz, (A ) glass no. 03 at 117 MHz, (0) glass no. 02 at 117 MHz, (0) glass no. 01 at 117 MHz. Each point represents an individual measurement.

The behaviour of the thioborate glasses (Figs. 3 and 4) is similar, at least in principle. Glass no. S3, with both manganese and iron impurities, displays again much stronger relaxation rates and a “shoulder” at 117 MHz (Fig. 3). Glass no. S2, with strong iron impurities and low manganese contents, shows near the maximum relaxation rates comparable to those of the purest glass no. Sl, but with “shoulders” and large temperatureindependent rates in addition (Fig. 4).

148

h4. Griine, W. Miiiler- Warm&r

T;’

500 ’

I

A

200

300

150

/Solid

T/K

,

State Nuclear Magnetic Resonance 5 11995) 145-150 5ccl

1

300

200

150

ZolZ

s-1 100:

10:

D 1:

0

11

.

.

E8

. g

2

0.1:

-I,

,

,

,

,

4

2

,

,

6

,

,

8

103/T

Glass no.

extracted

from the 7Li NMR spectra

Spectra

Relaxation

and relaxation

at 16 MHz

2,

01 02

0.27 0.32

205 200

100 108

18.1 20.7

03 Sl s2

0.26 0.11 0.16

200 160 160

470 46 89

24.4 13.4 13-14

s3

0.11 are explained

160 in the text. The apparent

E; (kJ mol-‘1

CT; ‘Lx fs-t)

200 activation

0

0

I 0

a

103/T K-’ no. S2. From top to at 16 MHz, (0) T;’

18.9 energies

measurements Relaxation

MDD 9 s-2 Cli; )

The symbols “shoulders”.

0

.

Table 2 lists some parameters extracted from the measurements. MyD is the dipolar rigid lattice second moment obtained from the central resonance. T, is the temperature where line narrowing occurs which may be used as an estimate of the activation energy by application of the Waugh-Fedin relation [19] E,/kJ mol-’ = 0.156 K/K. U;l>,,,ax and E;; refer to the T;’ value of the peak and to the apparent activation energy obtained from the slope on the right side of the maximum, respectively.

In contrast to the relaxation data, the ‘Li NMR spectra of the oxidic glasses on one hand, and of the sulfidic glasses on the other hand (not shown), all look very similar. They consist of a rather narrow central line superimposed at the bottom by a large quadrupole broadening. At elevated temperatures both the central line and the quadrupole structure become narrowed.

of NMR parameters

6

Fig. 4. Same as Fig. 2, but for glass bottom: tv) T;-‘, (A) TT1, (m) T;’ at 117 MHz.

,I

Fig. 3. Same as Fig. 1, but for thioborate glasses. From top to bottom: (A.) glass no. S3 at 16 MHz, (0) glass no. Sl at 16 MHz, (a) glass no. S3 at 117 MHz, (0) glass no. Sl at 117 MHz.

Table 2 A selection

4

a

.

Ei

marked

U;l),,,, (s-t)

at 117 MHz Ei (kJ mot-‘)

15 18

15.5 =9 *

60 6.4 21

16.2 12.8 =8 *

27 by an asterisk

- 13* refer to temperatures

above the

M. Griine, W. Miiller-Warmuth /Solid

State Nuclear Magnetic Resonance 5 (1995) 145-150

149

3. Discussion and conclusions

sis with glasses nos. 01 and Sl, for which no impurities could be detected.

3.1. Influence of the paramagnetic impurities

3.2. Relaxation behaviour of glasses without impurities

The experimental results of Figs. 1-4 and the parameters of Table 2 reveal distinct differences within both series of those glasses which should be equal if the paramagnetic ingredients are not considered. While the spectra (and the parameters M2 and T, as well) are nearly unaffected by the iron and manganese, relaxation is greatly influenced. Within the systems 01, 02, 03 as well as Sl, S2, S3 the (diffusion) peak heights the slopes of the curves (Ei) and their (T,‘),,,, appearances in detail, and even the temperature position of the maxima may differ. The major conclusions are confirmed by measurements on further glasses containing LiBr and LiI as dopants (not shown). We may distinguish between the effects of iron and manganese impurities. Iron-containing glasses display a strong frequency-independent relaxation contribution which leads to the “shoulders” in the high field. If the iron concentration is not particularly large, cf. for instance glass no. 02, (T;‘),, and Ei (only for 16 MHz) as well as the ratio CT; ‘I,,(16 MHz)/(T,‘),,,(117 MHz) = 7 are similar to glass no. 01 without paramagnetic impurities. However, if the iron concentration is extremely large becomes en(cf., e.g., glass no. S2) (T;l),, hanced as well, and the peak height ratio is smaller. In all cases, because of the shoulders, analysis of the low-temperature side of the curves is dubious. The most striking effect of manganese impurities is to enhance the peak height CT;‘),,,, by a factor of 2 4, cf. glasses nos. 03 and S3 compared with 01 and Sl, respectively. The slopes are also different and yield wrong data for El. However, the wa- ’ frequency dependence of the T; ’ maxima is not influenced by the addition of Mn*+. Glass no. S3 shows the combined influence of iron and manganese impurities. Relaxation studies of glasses require therefore samples which contain less impurities than those generally obtained from normal glassmaking procedures. In our case we are left for a data analy-

Although impurities are considered to be the main subject of this study, a brief discussion of the dominant relaxation mechanism will also be given for the glasses nos. 01 and Sl. Following previous investigations of similar systems [l-51 the frequency and temperature dependence of TIP’ and Tl;’ in the range of the T;’ peak can be explained in terms of the cation translational diffusion, which causes fluctuations of the nuclear quadrupole interaction. In agreement with an analysis of the spectra, the quadrupole interaction appears to be one order of magnitude larger than the dipole-dipole interaction. T; ‘, on the other hand, is governed by dipole-dipole interactions only (Figs. 2 and 4) and it deviates from T,* - ’ extracted from the central linewidth. The order of magnitude of the dipolar interaction is given by MFD of Table 2; a comparable My value for the quadrupole interaction as obtained from the spectra of the rigid lattice amounts to 50 x lo9 SK* for the oxidic glass and 20 X lo9 sP2 for the sulfidic glass. Like for other glassy materials the apparent activation energies E;; of Table 2 do not represent the real activation energy EA for the diffusion process. In disordered systems the correlation time of the motion is distributed and cooperative effects may play a role, Various models have been proposed for the correlation functions (cf. [15] for details), and they all have in common that Ei = aEA holds with a distribution coefficient (Y on the order of 0.3-0.4. The same holds for a novel crystalline thioborate sulfide, whose relaxation behaviour is very similar to those in glasses and which may be considered as a model system with a much larger accessible temperature range [20]. We thus obtain EA = 50 kJ mol-’ for glass no. 01 and EA = 40 kJ mol-’ for glass no. Sl. Both values are slightly larger than those from DC conductivity of similar systems: 46 kJ mol- ’ [21] and 30 kJ mol-’ [22]. Estimation from T, (Table 2) using the Waugh-Fedin relation [19]

150

M. Griine, W Miiller- Warmuth /Solid

State Nuclear Magnetic Resonance 5 (1995) 145-150

leads to an activation energy of 32 kJ mol-’ (01) and 25 kJ mol-’ (Sl). NMR relaxation data suffer from the nonaccessibility of the high-temperature side of the Tr-’ maximum, and the theoretical description is not yet satisfactory. T;’ had been expected to display the real activation energy. Fig. 2 shows, however, that the apparent activation energy varies here from 14 kJ mol-’ CT < 415 K) to 30 kJ mol-’ (T > 415 K) for glass no. 01.

References H. Olyschlager and H. [ll E. Gobel, W. Miiiler-Warmuth, Dutz, J. Magn. Reson., 36 (1979) 371. El 0. Kanert, J. Steinert, H. Jain and K.L. Ngai, J. NonCryst. Solids, 131-133 (1991) 1001. S.W. Martin and F. Borsa, [31 M. Trunnell, D.R. Torgeson, J. Non-Cryst. Solids, 139 (1992) 257. [41 W. Franke and P. Heitjans, Ber. Bunsenges. Phys. Chem., 96 (1992) 1674. W. Miiller-Warmuth, P. zum I51 M. Grime, H. Meierkord, Hebel, B. Krebs and M. Wulff, Ber. Bunsenges. Phys. Chem., 93 (1989) 1313. [61 A. Pradel, M. Ribes and M. Maurin, Solid State Ionics, 28-30 (1988) 762.

J. [71 S.W. Martin, H.K. Pate], F. Borsa and D. Torgeson, Non-Cryst. Solids, 131-133 (1991) 1041. J. Phys. Chem. Solids, P31J. Senegas and J. Olivier-Fourcade, 44 (1983) 1033. and G. 191 K.L. Ngai, J.N. Mundy, H. Jain, 0. Kanert Balzer-Jiillenbeck, Phys. Rev. B, 39 (1989) 6169. 0. Kanert, H. Jain and K.L. Ngai, [lOI G. Balzer-Jollenbeck, Phys. Rev. B, 39 (1989) 6071. H.J. Bischof, M. Mali, J. Roos and D. 1111 S.W. Martin, Brinkmann, Solid State Ionics, 18/19 (1986) 421. M. Mali, A. Pradel and M. 1121 J. Roos, D. Brinkmann, Ribes, Solid State Ionics, 28-30 (1988) 710. 1131 S.H. Chung, K.R. Jeffrey, J.R. Stevens and L. Borjesson, Solid State Ionics, 40/41 (1990) 279; Phys. Rev. B, 41 (1990) 6154. Solids, 131-133 [141 A. Pradel and M. Ribes, J. Non-Cryst. (1991) 1063. Ber. Bunsenges. 1151 M. Grime and W. Miiller-Warmuth, Phys. Chem., 95 (1991) 1068. 1161D. Tse and S.R. Hartmann; Phys. Rev. Lett., 21 (1968) 511. Solids, 66 [171 H.-J. Stockmann and P. Heitjans, J. Non-t&t. (1984) 501 [18] W.E. Blumberg, Phys. Rev., 119 (1960) 79. [19] J.S. Waugh and E.I. Fedin, Sov. Phys. Solid State, 4 (1963) 1633. [20] M. Grime, W. Miiller-Warmuth, P. zum Hebel and B. Krebs, Solid State Ionics, 66 (1993) 165. [21] A. Levasseur, J.-C. Brethous, J.-M. Reau and P. Hagenmuller, Mat. Res. Bull., 14 (1979) 921. [22] M. Menetrier, A. Hojjaji, C. Estournes and A. Levasseur, Solid State Ionics, 48 (1991) 325.