NUCLEAR

SP IN RELAXATION

IN SUPERCONDUCTORS

g(x) causes the second order term in Eq. (12) to be small, so that for po suSciently large,

I2

2(x'+vtp')e

"' Ei(Psip)
(16)

(x+stp)

[1+exp( x) j'

The values of p versus otp are needed for /=0. 05 and P=0.2. A connecting function is needed for each value of P to connect the expression for Ip in Eq. (9), valid for small stp, with that in Eq. (16), valid for suKciently large go. The function

I2=

o) oo (] +C)alto

l&oo 1+( e

connects the points and slopes to within 1% using the following parameters: for P=0.05, C= 2.380, b=5.560, and a=4. 305; for /=0. 2, C= 1.735, 6=4.72, and a=2. 93. Investigation of the second order error term in Eq. (12) for I& shows that Is is represented by its asymptotic form, Eq. (16), to better than 2% for sip& 1.50 for both values of p.

The relation between rtp= ep(T)/kT and temperature, which was used in compiling Table I, was taken from HCS theory using their temperature dependence of ep(T) and ep(0) = 3.5kT, . A change in either factor would change only the temperature scale of Table I and Fig. 1, since all calculations scale as rtp itself (i.e. , only the values of po themselves appear in the equation). Thus, the results in Table I may easily be extended to the case of an experimental gap width by using the new relation between qo and temperature. One should note that the value of ep(0)/6 assigned to the table might also be changed. The energy level breadth, 6, was assumed to be independent of temperature. [Such an assumption allows 6 to be interpreted also as an anisotropy in ep(0) in k space. The effect of a temperature dependent 3 can be obtained by moving from one value of ep(0)/6 to another in changing

temperature.
ACKNOWLEDGMENTS

The author wishes to thank Dr. i%i. I.ax for several suggestions concerning suitable approximation methods
for I2.

PHYSICAL REVIEW

VOI UMF.

116, NUMBER

OCTOBER 1, 1959

Origin of the Characteristic

Electron Energy Losses in Magnesium*

Westerrt Australia

C. J. PowELL AND J. B. SwAN DePartmeet of Physics, University of Westerl Australia, itredlartds,

(Received May 12, 1959)

The characteristic electron energy loss spectrum of magnesium has been measured by analyzing the energy distribution of 750-, 1000-, 1505-, and 2020-ev electrons scattered by a, n evaporated specimen through 90'. The spectra were similar in form to those previously obtained with aluminum targets, in that the observed loss peaks were composed entirely of combinations of two elementary energy losses, These two losses, of magnitude 7.1 and 10.6 ev, were identified, respectively, with the lowered plasma loss proposed by Ritchie and the plasma loss proposed by Bohm and Pines.

INTRODUCTION

' "N a previous paper, ' the origin of the characteristic

~ ~ electron energy losses in aluminum was discussed in some detail. It was shown there that the characteristic loss spectrum of aluminum is made up of combinations of elementary 10.3- and 15.3-ev losses, the former, the low-lying loss, being identified with the lowered plasma loss proposed by Ritchie, ' and the latter with the plasma loss proposed by Bohm and Pines. ' The chara, cteristic loss spectrum of magnesium has

been studied by a number of workers, ' and the observed loss values a, re shown in Table I. The shape of the magnesium spectrum has been found to be similar to that of aluminum, in that a sharp energy loss is observed, together with one or more lines due to multiples of this elementary loss. This sharp line, 10 ev in magnesium, has generally been occurring at regarded as arising from collective electron excitation'

"

' L. Marton
~

* Work supported by the Research Grants Committee of the University of Western Australia. ' C. J. Powell and J. B. Swan, Phys. Rev. 115, 869 (1959).
and D. Bohm, Phys. Rev. SS, 338 (1952); D. Pines, Revs. Modern Phys. 28, 184 (1956); P. Nozieres and D. Pines, Phys. Rev. 109, 1062 (1958).
3

and L. B. Leder, Phys. Rev. 94, 203 (1954). L. Marton, Phys. Rev. 95, 1345 (1954). ' H. Watanabe,and Phys. Soc. Japan 9, 1035 (1954). J. 7 w. Kleinn, optik 11, 226 (1954). 'Blackstock, Ritchie, and Birkhoff, Phys. Rev. 100, 1078

L. B. Leder

' R. H. Ritchie,
D. Pines

Phys. Rev. 106, 874 (1957).

(1955). 9 G. W. Jull, Proc. Phys. Soc. (London) 869, 1237 (1956). "V. I. Milyutin and A. I. Xabanov, Uspekhi Fiz. Nauk. 61, 673 (1957). "N. B. Gornyi, Izvest. Akad. Nauk. S.S.S.R. Ser, Fiz. 22, 475 (1958).

C.

J.

POWELL AND

J. B.

SWAN

TABLE I. Observed values of the characteristic electron energy losses in magnesium (in ev). The column headings are reference numbers.
10
11a

9.7
20.3

9.7
20.3

31.2 41.5

10.3 20.7

4.7 10.3
22. 1

rations of magnesium were used to getter the target chamber. Target oxidation and carbonaceous contamination rates were again small enough to permit a spectrum to be recorded and then repeated without any significant change taking place.
RESULTS

10.5 21.5 31.6

10.0 20.3

9.5
20.2 31.8

Unknown

43.2

' See note

added in proof.

(the predicted plasma loss is 10.9 ev assuming two free electrons per atom) though Leder, Mendlowitz, and Marton" have suggested that the energy losses might be due to interband electronic transitions and find some degree of correlation between the energy losses and the displacements of the absorption maxima from the E x-ray absorption edge. It should be noted, however, that Townsend" reports anomalous x-ray absorption on the high-energy side of the L edge in magnesium films that have been evaporated onto zapon backings, and attributes the effect to surface contamination of the magnesium 61m at the metal-backing interface. The only low-lying loss reported for magnesium t that of 4. 7 ev observed by Kleinn, 7' could conceivably be a lowered plasma loss though the present authors consider that, from the shape of his published spectrum, this loss is probably due to the presence of a substantial quantity of magnesium oxide on the specimen surface. ' The necessity for the specimen to be of high volume and surface purity was demonstrated in the previous investigations of the characteristic loss spectrum of aluminum. The work to be reported here was undertaken in order to make studies of the losses in magnesium in a reQection-type experiment, the advantages of which have been discussed. ' It was considered that, under these conditions, a low-lying loss might be observed and that its properties could be compared with Ritchie's theory.
EXPERIMENTAL PROCEDURE
were obtained by analyzing, with the 127' electrostatic electron spectrometer and auxiliary apparatus previously described, ' the energy distribution of electrons scattered by the target through 90'. The target surfaces were prepared by evaporation of spectrographically pure magiiesium from a multistrand tungsten conical basket onto a metallic substrate at room temperature. The evaporation time was made short in order to reduce agglomeration, a fresh film of estimated thickness 50 to 100 A being deposited in a few seconds; such films were found to give spectra similar to those of thicker films. As in the aluminum work, several preliminary evapo-

"

The characteristic electron energy loss spectrum of magnesium was measured for primary electron energies of 750, 1000, 1505, and 2020 ev. The normalized spectra are shown in Fig. 1 with the intensity scales adjusted so that the peak of elastically scattered electrons has an intensity of 25 units in each case. Mean energy loss values are shown in Table II together with the probable error and number of measurements associated with each mean. No significant difference was found in the energy loss values at the diferent primary energies and, with the exception of the peak of lowest energy loss at 2020-ev primary energy, no correction on account of the background has been applied to any of the measured peak positions. It is evident that the loss spectra are composed entirely of combinations of elementary 7. 1- and 10.6-ev energy losses. The former is identified as the lowered plasma loss proposed by Ritchie and the latter with 10-ev loss observed by other workers. As was the found for aluminum, the ratio of these energy losses is 1.49, again slightly greater than the predicted ratio of
12

10

10

IZ
LLI

IZ

The characteristic

loss spectra

)
LLI

4J
Ct

4
2

70

60 50 40
EL EC TRON

30 20

10

ENERGY

L055 (ev)

'f H. Watanabe,

Leder, Mendlowitz, and Marton, Phys. Rev. 101, 1460 (1956). "J. R. Townsend, Phys. Rev. 92, 556 (1953).
See note added in proof. J. Phys. Soc. Japan 9, 920 (1954).

12

nesium

FIG. 1. Characteristic electron energy loss spectra of magfor primary electron energies of 750, 1000, 1505, and 2020 ev. The intensity scales have been adjusted so that the peak of elastically scattered electrons (not shown) has an intensity of 25 units in each case.

ELECTRON ENERGY LOSSES IN

TAaI. K II. Measured mean values of the characteristic electron energy losses in magnesium, together with the probable error and number of measurements associated with each mean. The energy losses and errors are expressed in ev, and m and rI, represent the number of multiples of 7. 1- and 10.6-ev losses, respectively, that give rise to each observed energy loss.
Loss
value

Mg

relative to the peak of elastically scattered electrons, so that on the average d/XM must increase with primary energy. It is also seen that the ratio of the area under the 1.0.6-ev peak to that under the 7.1-ev peak increases
with primary energy. This change is qualitatively in accordance with the predictions of Ritchie. It should be emphasized, however, that as the shapes of the electron paths in the target are not known in detail, it is not possible to measure the change of excitation probability over these paths, and quantitative comparisons with the theory cannot be made. It is found that the width at half maximum intensity of the 10.6-ev line is 1.4&0.1 ev, taking into account the spectrometer resolution and the energy spread of the primary beam. It is interesting to note that this width is not significantly different from the width of the 15.3-ev aluminum loss peak previously reported, ' and presumably arises from similar considerations.
CONCLUSION

7.1

Prob.

01$

lZ

1,0

10.6
0, 1

17.8 1,1

21.3 28.4
0, 2

1,2

31.9 38.6
0, 3

1,3

42.5 52.7 64. 1

0, 4

0, 5

0, 6

error No.

0, 1 0.1 0.1 0. 1 0, 2 0.1 135 145 124 142 65 140

0.2
7

111

0.2

49

0.2

12

0.2

1.41. The discrepancy could well be due to partial agglomeration in the target 61m, though it is surprising and perhaps significant that this same ratio is observed for both aluminum and magnesium. It may also be seen that, apart from differences which are the consequence of the better resolution of the more widely separated aluminum peaks, the form of the loss spectra of magnesium is very similar to that of aluminum. If it is assumed that multiples of the 7.1- and 10.6-ev losses occur only in single quantum loss processes, and that the probabilities of excitation per unit distance traveled in the specimen are constant (for a given primary electron energy) over the total path in the specimen, then the relative total probability of any loss can be calculated in terms of X7, Xio and d, where X7 and X10 are the mean free paths for the 7.1- and 10.6-ev losses, respectively, and d is the average distance traveled by an electron in the target. ' As the area under each peak is proportional to the appropriate of these areas should in probability, measurement principle enable a number of independent determinations of the ratios d/Xq and d/Xyo to be made. It was found, however, that no background could be drawn which would give consistent values of both d/X7 and d/Xyo. The simple assumptions made above must therefore be considered incorrect. The observations are consistent with Ritchie s prediction that X7 and Zip should vary with depth of penetration, in opposite directions, ili the vicinity of the specimen surface. As the primary electron energy is increased, more peaks which are multiples of the 10.6-ev loss appear, and all multiples of the 10.6-ev loss increase in intensity
"The predicted ratio of 1.41 compares the energy losses in the forward direction only, whereas the observed ratio compares the mean energy losses resulting from the dispersion relation and the distribution of loss intensity over the small range of scattering angles. '

"

The characteristic electron energy loss spectrum of magnesium has been measured in a reAection type experiment with primary electron energies of 750, 1000, 1505, and 2020 ev. Each spectrum was found to be composed entirely of combinations of elementary 7.1and 10.6-ev energy losses. The positions of these losses and observations of their relative intensities indicate that the 10.6-ev loss arises from plasma excitation and that the 7.1-ev loss is a lowered plasma loss of the type
predicted by Ritchie.
ACKNOWLEDGMENT

the One of us (C.J.P.) wishes to acknowledge tenure of the I.C.I.A.N. Z. Research Fellowship and a C.S.I.R.O. Senior Postgraduate Studentship at different times during this work. $
f. Note added ie proof. The energy losses observed by Gornyi" in a reflection type experiment with primary electron energies between 30 and 100 ev are now known to be 4.5, 7.5, 10.5, 14 23, 25.5, and 31.5 ev. All these losses were interpreted by him as being due to interband electronic transitions. The losses of 7.5 and 10.5 ev agree well with the low-lying and plasma losses of 7.1 and 10,6 ev, respectively, found in the present work. Gornyi's spectra show the 7.5 ev loss more prominent than the 10.5 ev loss, as would be expected from the change in relative intensity of the fundamental losses with primary energy described above. The 4.5 ev loss is considered to be due to the presence of oxide on the specimen surface; the origin of such a loss will be discussed in a future paper describing the effects of specimen oxidation. The other energy losses found by Gornyi are of much smaller intensity and might be due to interband transitions.