This last section on magnetic properties will examine in how far the changes in magnetic properties allow conclusions to be drawn regarding the charge transfer and the bonding of the hydrogen atoms in the ternary hydrides. The many results discussed above make it clear that the most spectacular changes involve the 3d electrons. The 3d band magnetism in metals is still the subject of current experimental and theoretical studies and an analysis of the data presented above in terms of the 3d band model is not free of ambiguity. The same argument as already presented elsewhere will be used (Buschow and de Chätel, 1979) in order to show that some general conclusions can nevertheless be reached.
Malik et al. (1978b) and Wallace (1982) explained the reduetion in Co moment after H2 uptake as being due to a further depletion of the 3d band. In view of the discussion on 3d band magnetism given in section 2.2.2 this is a very unlikely explanation. The rather large value of #co, espeeially in LaCo» suggests a filled majority (say, spin-up) subband, in which case the depletion of the d band should lead to an increasing moment. One would have to assume a considerable number of
76 K.H.J. BUSCHOW
d holes per Co atom to meet the condition N(EF) ~" > N(EF)~, which, according to implication (iii) of the discussion in section 2.2.2. is necessary for the mechanism proposed by Malik et al. to give the desired reduction of PCo. As pointed out in section 2.2.2, the number of d holes is not expected to deviate much in R-Co compounds from its value in pure Co. The low electronegativities of Y and La make an increased number of holes especially unlikely in the compounds listed in table 8.
If one sticks to explaining the moment changes in terms of charge transfer, one is left with conclusions like the one given by Kuijpers (1973) that the d band must become gradually filled upon H2 uptake. Here we recall that, as a result of the feedback mechanism mentioned in section 2.2.2, the change in the number of d electrons may actually be somewhat smaller than the change in #co/PB.
We have seen that hydrogen absorption in Fe compounds,,in contradistinction to Co compounds, leads to an increase in 3d moment (see table 9). As the 3d moments quoted in table 9 are generally 1 PB higher than the ones given for PCo in table 8, the same arguments can be used for the Fe compounds as for the Co compounds to show that N(EF)T < N(EF)+ is more lilely to hold than the opposite inequality. Therefore, implication (iii) in section 2.2.2. entails a depletion of the 3d band upon hydrogen absorptiono Similar conclusions have been reached before (Wallace, 1982; Malik,
1978a; Buschow and van Diepen, 1976). However, it should be pointed out again that out explanation does not necessitate a charge transfer of d electrons of the order of ApF~/#B, because of the feedback effect referred to earlier (section 2.2.2). It has already been noted in connection with the analysis of IS in section 5.2.3. that the small positive change in IS implies that the transfer of charge has to be composed of about equal amounts of s and d electrons. If the number of transferred d electrons were significantly below A#Fe/I~B, the total number of electrons transferred could reach a magnitude that would be in keeping with the redistribution of charge based on the electronegativity differences between the composing elements (Miedema, 1973). In other words, because of the feedback effect the discrepancy between the large change in #Fe, and the comparatively small change of the corresponding IS is less strong that it seemed initially. Nevertheless it is doubtful whether this reasoning is still applicable in cases where A~F e becomes of the order o f a whole Bohr magneton or more, such as in ScFe2Hx.
A further difficulty with explanations based on charge transfer arises if one compares the results obtained in Fe compounds with those in Co compounds. If one wishes to ascribe a decisive role to d-band occupancy changes in the decrease of Pco and in the increase of #Fe upon hydrogen absorption, one has to assume that H donates electrons to the Co 3d band but accepts electrons from the Fe 3d band.
Attempts have been made to explain this apparently contradictory situation in terms of electronegativities (Buschow, 1978), the value for metallic H being lower than that of Co and higher than that of Fe (Miedema, 1973). Although the decrease of the Ni moment observed in the meantime in YNi3H x is in keeping with the sequence of electronegativities, the decrease in magnetic susceptibility observed in TiFeH x (Stucki and Schlapbach, 1980) does not support the assumption that the H accepts electrons from the Fe 3d band.
It follows from the foregoing discussion that there is more evidence against major
H , - , 3d charge transfer effects than there is in favour of it. In terms of the 3d band model there is still the possibility of explaining the changes in magnetic properties in terms of changes of the d ~ t interaction parameter. In fact, equally convincing arguments can be found for an increase of the interaction parameter upon hydrogen absorption. If Fe is combined with R metals the 3d electrons of the former will mix or hybridize with the d and S, p electrons of the latter and in this way cause a reduction in the d ~ t interaction parameter and hence a reduction in band splitting.
In the hydrides the R - F e distances are increased and a large part of the R - F e contacts are lost due to the absorbed H atoms. This will reduce the hybridization and partially restore the exchange splitting. Note that the influence of an increasing lattice constant is also beneficial in so far as it leads to rauch narrower 3d bands.
The results obtained on ScFe2 are very illustrative in this respect. Compared to «-Fe the Fe moments in these compounds are strongly reduced. However, in the hydrides the spin imbalance has become almost as large again as that in «-Fe. This means that the presence of the H atoms has neutralized the unfavourable effect of the Sc atoms almost completely. This notion is in keeping with the fact that the H atoms (at not too high H concentrations) are located close to the R atoms and, as it were,
"screen" the R atoms from the 3d atoms (see section 4.1). Increasing hydrogen concentration is found to lead to a larger occupancy of those interstitial holes where the situation is less favourable, i.e., where the area of contact of the H atomic cells with the 3d atomic cells has become larger and that with the R atomic cells has become smaller. One may expect therefore that an increase in hybridization between the 3d electrons and the 1 s electrons of H will take place. In addition more dissimilar interstitial holes will become occupied so that disorder of the lattice is increased. Both occurrances will eventually lead to a reduction of exchange splitting. This agrees with the results obtained for ErFe3Hx and the disappearance of an Fe moment in ErFe2Hx when x > 4.1 (see fig. 49). In off-stoichiometric ScFe2 an effective shielding of the Fe sublattice from the Sc sublattice by means of a symmetric filling of interstitial hole sites with H atoms is not possible. In this case the moment increase upon charging is considerably less than the moment increase in stoichiometric ScFe2 (Smit et al., 1982).
A reasoning similar to that given above can,also be applied to Co (and Ni) compounds. Here we have to take into account that the 3d moment formation in Co and its compounds is rauch more vulnerable. Arguments have been presented elsewhere (Buschow et al., 1980) that the 3d moments in Co intermetallics are less localized than those in Fe intermetallics. The detrimental effect on the 3d moment formation of increasing 3d-H contacts at the expense of 3d-3d contacts upon charging is much stronger and overcompensates the beneficial influence of the reduced hybridization of the 3d electrons with the d and s, p electrons of the R component.
From the results discussed above the following conclusions can be drawn:
(i) If one wishes to explain the controversial changes in 3d moment in Fe and Co compounds by means of a unified model one has to accept that charge transfer is not the main reason for the hydrogen-induced change of moment.
(ii) A relatively large transfer of charge takes place between the H atoms and the
78 K . H J . BUSCHOW
R atoms. This follows from the isomer shift results obtained by means of 151Eu, 161Dy and 161Gd Mössbauer effect spectroscopy, which indicate that the R atoms donate electrons to the H atoms. These findings are in agreement with the fact that there is virtually no difference in electronegativity between H and the 3d metals but a considerable electronegativity difference between R and H.
In terms of Miedema's cellular model, the changes in 3d magnetism can be explained as follows (Buschow et al., 1982a): In the ternary hydride the contact area of a given 3d atom comprises 3d-3d contacts as weil as a substantial portion of 3d-H contacts. In many respects the situation is similar to the one that would be present in the binary 3d transition metal hydrides, so that an analogous magnetic behaviour can be anticipated.
In the case of Ni and Co the introduction of hydrogen leads to a decrease of Te and to a lower spontaneous magnetization (Wagner and Wortmann, 1978). Hence one expects similar effects upon ternary hydride formation of Co and Ni compounds, which agrees with experiments. It should be noted that the reduction of the ferromagnetic properties in the 3d metal hydrides is a result of the magnetic exchange parameters, charge transfer between Ni or Co and hydrogen being very small.
In the case of Mn and Fe the situation is less clear than in the case of Ni and Co.
There are experimental indications that the presence of hydrogen is not an unfavourable factor with respect to magnetic moment formation. Antonov et al.
(1978) found high Curie temperatures and an appreciable spontaneous magnetization in hybrided Fe65(NiMn)3 » alloys containing up to 17 a t ~ manganese. These alloys are non-magnetic without hydrogen. Local moment formation is a delicate matter, however, so that accurate predictions of the change in magnetic properties of Fe and Mn compounds upon hydrogen absorption are hardly possible.
Finally we note that the cellular model also provides a hint as to the direction of possible valence changes in the intermetallics RMn upon hydrogenation, when R is one of the elements Ce, Eu or Yb. Hydrogen absorption results in the formation of an appreciable area of contact between R atoms and H atoms at the expense of R - M contacts. The R atoms therefore will tend to adopt the same valences they have in the binary hydrides, i.e., Ce tends to be trivalent, Eu and Yb tend to be divalent.
This agrees with experimentally observed valence changes (Buschow et al., 1977;
Buschow, 1980b).
Apart from the volume change due to the H2 uptake, orte has an additional volume effect owing to the valence change. It should be emphasized that the valence change is driven by the energy effects associated with the creation of R - H interfaces and the loss of R - M interfaces. This has as a consequence the occurrence of an additional volume effect. Alternatively one could say that the additional volume effect is driven by the valence change and not vice versa, as is sometimes assumed.
5.3. Miscellaneous properties
A serious drawback in the study of the changes in physical properties due to hydrogen absorption is the fact that the material pulverizes during charging. Using standard methods it therefore seems impossible to determine electrical transport
78
_L I _,_ Il _ L I _1_ lI _ i
r " I ~ - i ~ - i ~ - , ~
7L
QL.
7.2
I I I I P r i
0 1 2 3 4 5 6
Time [h}
Fig. 51. Effect of hydrogen sorption (at 50°C) on the electrical resistivity of a 3000 Â. thick LaNi 5 film:
I, absorption; Il, desorption. The results represent data published by Adachi et al. (1981).
properties of ternary hydrides. Adachi et al. (1981) were able to avoid sample pulverization by using for their resistivity measurements a sample of LaNi 5 prepared in the form of a thin film (3000 A thick) by means of flash evaporation. Their results are reproduced in fig. 51, where the change in resistivity during cycling is shown after initial activation and after allowing the changes to become reproducible during the first 20 cycles. It can be seen from the figure that charging results in an initial small increase followed by a strong decrease. The initial increase was interpreted by the authors as being due to chemisorption, dissociation of H2 and dissolution in the film as H - ions. The strong decrease of p was ascribed to the formation of a highly conductive ternary hydride. The authors verified that the properties of a La or Ni film prepared in this way were quite different. They also checked the composition of the films and concluded from X-ray data that they were amorphous (Adachi et al., 1982). The results of Adachi et al. show that the resistivity of LaNi5 remains the same within 10~o after charging. This result is rather surprising in view of the large changes in structure and bonding resülting from the introduction of H atoms. The experimental results of Adachi et al., showing the presence of highly conductive material after charging, are at odds with the findings of Walsh et al. (1976), who concluded from ESR experiments that LaNi5 after charging has quite a low carrier concentration, reminiscent of semiconducting or barely metallic materials.
Although somewhat less interesting from the point of view of solid state physics, the thermal conductivity is of paramount importance in all cases where intermetallic compounds and their hydrides are employed as hydrogen storage materials. In hydrides with a good thermal conductivity the heat released upon charging can easily be dissipated and decomposition can be avoided. Good thermal conductivity is essential if these materials are to be used in heat pump devices. Unfortunately, no investigations of this kind have been performed on rare earth intermetallics and their hydrides.
Because of their relevance to these materials we would like to mention here briefly the results of Suda et al. (1980), who studied the thermal conductivity of TiMnl. » and its hydride. These authors found different thermal conductivity values during
80 K.H.J. BUSCHOW
absorption and desorption, which reflect the presence of hysteresis in the temperature composition relationship. Suda et al. derived an empirical expression for the thermal conductivity, comprising a contribution proportional to the H/M ratio and pressure- dependent contributions of the type ( l n p ) n, where n ranges from 1 to 3.
The effect of hydrogen absorption on superconductive properties was studied on relatively few compounds. These comprise Th7Fe 3 (Malik et al., 1978a), CeCo2 (Buschow and Sherwood, 1978), several La-Ni compounds (Oesterreicher et al., 1976) and several Th-R alloys (Oesterreicher et al., 1977). In all cases it was found that the absorption of hydrogen gas leads to disappearance or lowering of the superconducting state.