Proc. Natd. Acad. Sci.

USA

Vol. 85, pp. 8376-8380, November 1988 Chemistry

Unified structure theory of icosahedral quasicrystals: Evidence from neutron powder diffraction patterns that AlCrFeMnSi, AlCuLiMg, and TiNiFeSi icosahedral quasicrystals are twins of cubic crystals containing about 820 or 1012 atoms in a primitive unit cube
(Al74Mn2lSis/AlsloCuLMg,*,/atomic volume)
LINUS PAULING
Linus Pauling Institute of Science and Medicine, 440 Page Mill Road, Palo Alto, CA 94306

Contributed by Linus Pauling, July 29, 1988

ABSTRACT

Aunifiedstructuretheoryoficosahedralquasi-

crystals, combinig the twinned-cubic-crystal theory and the Penrose-tiling-six-dimensional-projection theory, is described.
Values of the primitive-cubic lattice constant for several quasicrystals are evaluated from x-ray and neutron diffraction data. The fact that the low-angle diffraction maxima can be indexed with cubic unit cells provides additional support for the twinnedcubic-crystal theory of icosahedral quasicrystals.

The difference between the two alternative theories is that in the twinned-cubic-crystal theory there are, for the crystallites, three translational symmetry operations, so that the cluster-cluster vectors include those with components that are integral multiples of the cube edge a, whereas in the other theory the vectors have only random magnitudes and directions, except for those within a cluster.
The Unified Theory

Since their discovery (1) 4 years ago, the question of the nature of the so-called icosahedral quasicrystals has remained unresolved. Most of the investigators who have studied this matter have concluded that some sort of randomness in structure has resulted in the formation of quasicrystals with fivefold axes and other symmetry elements that give icosahedral symmetry to the substances, down to atomic levels (2-5). A few, however, have suggested that icosahedral twinning of conventional crystals could explain all of the experimental observations. I, in particular, have contended that crystals composed of very large clusters of atoms with approximate icosahedral symmetry can form cubic crystallites that by icosahedral twinning form grains with icosahedral point-group symmetry, the so-called icosahedral quasicrystals (6-8). During the past 2 years there has been widespread agreement that many icosahedral quasicrystals contain the large icosahedral triple-shell complex discovered in 1952 in the compound Mg32(Al,Zn)49 by Bergman, Waugh, and Pauling (9, 10). This complex consists of a central atom (which sometimes is missing) surrounded by an icosahedral shell of 12 atoms, a second shell of 32 atoms, and a third shell of 60, 72, 80, or 92 atoms (all retaining icosahedral point-group symmetry). I described two new structures involving 8 of these clusters in the ,-W positions (primitive cubic unit cell), one, for the icosahedral quasicrystal Al6Mn (6), with 104atom clusters in the 8 positions, and the other, for the icosahedral quasicrystal Al6CuLi3 (7), with two 104-atom clusters and six 136-atom clusters in these positions. (These structures are described as the 820-atom structure and the 1012-atom structure, although the number of atoms in the unit cube may vary somewhat.) Among other investigators who have discussed this cluster in connection with icosahedral quasicrystals are Audier et al. (11), Ma et al. (12), Henley and Elser (13), Elswijk et al. (14), and Van Smaalen et al. (15). All investigators assume that the large icosahedral clusters are present in nearly parallel orientation.
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. 1734 solely to indicate this fact.

Over the past year I have formulated a theory that unifies the two theories about the nature of the icosahedral quasicrystals. According to this unified theory (7), there are in the quasicrystals cluster-cluster vectors of both kinds: those provided by the translational identity operations of the cubic units of structure, and also others that involve certain kinds of randomness. Moreover, the unified theory, by invoking the cubic crystallites with structures closely similar to those of known intermetallic compounds, has the desirable feature of conforming to the known principles of structural chemistry. The formation of large icosahedral clusters is expected for an alloy of two or more metallic elements with different radii, and these clusters, probably with a range of values of the number of atoms in the outer shell, can be expected to be present in the molten alloy. If the molten alloy were to be quenched extremely rapidly it would form a metallic glass, with the clusters having the same random orientation as in the melt and being piled together as in the melt. With somewhat less rapid quenching the clusters would begin to form crystallites. These crystallites might in some cases, as for Al6CuLi3, have the thermodynamically stable structure. Usually, however, as for Mg32(Al,Zn)49, the thermodynamically stable structure involves sharing the atoms in the outer shell between clusters, which requires time, so that instead the alloy crystallizes with a metastable structure that involves an easily assumed arrangement in which the clusters that were present in the liquid pile together in a crystallographically simple way, such as cubic close packing of the clusters, 4 clusters in a face-centered cube. In fact, no such 4-cluster quasicrystals have been found; instead, both the 820-atom and the 1012-atom structure are based on the p-W arrangement, with the unit cube containing 8 clusters. The reason for the selection of the P3-W arrangement is obvious. It is that the Bergman-Waugh-Pauling triple-shell cluster of 104 to 136 atoms has icosahedral point-group symmetry and accordingly strives to coordinate 12 other clusters about itself at the corners of an icosahedron. Each of the clusters at 0 0 0 and 1/2 Y2 1/2 in the p-W unit is ligated in this way to its 12 neighbors, at 0 1/4 Y2 and equivalent positions. I have pointed out that the pseudo-icosahedral character of these cubic crystallites would cause them to undergo icosa8376

Chemistry: Pauling
hedral twinning to 20 orientations, corresponding to identity of one threefold axis and three planes of symmetry of each crystallite with those of the icosahedral cluster (6). A feature of this multiply twinned aggregate is that all of the icosahedral clusters in all of the crystallites have nearly the same orientation. The diffraction maxima in such a grain have intensities and positions determined by three structure factors: the structure factor of the cluster, the structure factor of the cubic crystallite, and the structure factor representing the interaction of one crystallite with the other. The first two of these can be handled by the methods of conventional crystallography. The structure factor of the cubic crystal leads to Q vectors in reciprocal space ending at or near the lattice points, the closeness being determined inversely by the size of the crystallites. I have found that the Q values for many quasicrystals, especially for the weak peaks, can be accounted for by a primitive cubic unit cell; this fact strongly supports the twinned-crystal theory and the unified theory. The structure factor of the cluster also determines the intensities of the diffraction maxima, and, if the crystallites are small, can move the maxima appreciably from the lattice points. It is likely, however, that this effect is in the main caused by crystallite-crystallite interaction, the third structure factor. The third structure factor can be discussed by consideration of atom-atom pair vectors. Let us consider the vectors between an atom in one crystallite and two atoms in a cluster in another crystallite. Neither of the two vectors is related in a well-defined way to any translational symmetry operation, but their difference vector remains the same from cluster to cluster, because the clusters have nearly the same orientation. It has been shown by many groups of investigators that this treatment leads to sharp diffraction maxima at Q values related by the golden number for example, by Levine and Steinhardt (2), using Penrose tiling to introduce the randomness in the vectors, and by Kalugin et al. (5), using projection from a six-dimensional icosahedral crystal. It was shown by Elswijk et al. (14) that the Bergman-Waugh-Pauling cluster (7, 8) leads with this calculation to good values for the positions and intensities of the 37 strongest diffraction maxima of Al6CuLi3. I point out that this calculation applies also, at least approximately, to a single crystal with 8 clusters in a cubic unit. For each cluster in one unit there is only one in another unit related to it simply by the translational operations; for the other seven there are quasirandom differences in the vector lengths. There is the possibility that for sufficiently rapid quenching of some alloys the cubic microcrystals would be so small (less than one unit of structure) that it would be unjustified to call them crystals, in that there would be no translational identity operations, but that the intercluster interactions that lead to the formation of the crystals and to the process of twinning would still operate to align the clusters into parallel orientations, giving rise to a structure with icosahedral point-group symmetry of the sort considered by many quasicrystal investigators. The quasicrystals that I have studied all provide evidence for the presence of twinned cubic crystallites. Analyses of x-ray and neutron diffraction patterns of several quasicrystals supporting the theory that twinned cubic crystals are present are described in earlier papers (68). Additional evidence is presented in the following sections.
r,

Proc. Natl. Acad. Sci. USA 85 (1988)

8377

atoms are replaced by Cr and Fe atoms. The data consist of intensity curves from Q (= 2Xr/d) = 1 A-' to 8 A-' and computer evaluations of maxima from 0.68 A' to 6.8 A'1. The five curves show many weak peaks that were ignored by the computer. The peaks appear on all five curves at nearly the same Q value, but with greatly different intensity, because of the different neutron F factors of the different mixtures of Cr, Fe, and Mn. Values of the interplanar distance d for the 11 weak peaks and 2 strong peaks with the smallest values of Q for the A174Mn21Si5 curve are shown in Table 1. Nearly the same values are found for the other four curves. The smallest cubic unit of structure that accounts for the observed spacings has edge a equal to 22.99 A. Many more peaks, with larger Q, can be seen. They can all be fitted, to within the error in measurement, by the same unit of structure. Values of d for the first 20 peaks listed by the computer are given in Table 2. They are compatible with the same cubic unit, yielding 23.01 A for a. The 33 peaks with larger Q values are also compatible with this unit. These alloys can be assigned the 820-atom structure, with complexes of 104 atoms (20 condensed Friauf polyhedra) at the B-W positions.

A161Cu102Li337
In table Ila of ref. 16 values are given of Q and Ifor 17 maxima in the high-intensity x-ray powder diffraction pattern of the icosahedral phase A1561Cu102Li337. It is to be expected that this phase would consist of cubic crystals with a primitive unit cube with edge nearly the same as for Al6CuLi3 (7), 25.70 A. Measurement of 14 weak peaks (8) with small Q values on a published x-ray diffraction curve of A1561CU102Li337 gave the value a = 25.69 A. It was found that the 17 maxima given in ref. 16, extending to Q = 4.8 A-', could be indexed with the same unit, as is shown in Table 3, leading to the average value a = 25.81 A. Pulsed Neutron Powder Diffraction Data

Shen et al. (16, 21) have published curves and tables of values of Q and the intensity I for the stronger maxima obtained by pulsed neutron powder diffraction by four AlCuLiMg alloys. S. J. Poon kindly sent me the data, values of I at intervals of
Table 1. Analysis of 13 maxima in the neutron diffraction pattern of the rapidly quenched alloy A174Mn2lSi5

1*
1 1 1 4 2 1 1 50 1 1 1 40 1

d,tA

hkl

a,A

Quenched Alloys

AM74Mn2lSis and Other Al74(Cr,FeMn)21Si5 Rapidly

Christian Janot (Max von Laue-Paul Langevin Institute, Grenoble, France) kindly sent me detailed high-resolution neutron powder diffraction data for rapidly quenched Al74Mn21Si5 and four other alloys in which some of the Mn

23.0 321 6.16 22.9 400 5.72 23.0 33 1 5.27 23.0 332 4.90 22.9 500 4.57 22.9 521 4.175 23.01 440 4.068 23.04 610 3.788 22.97 541 3.545 23.00 622 3.468 22.89 63 1 3.375 22.96 444 3.314 23.05 642 3.080 22.99 Average of last 7 I is the intensity, d is the interplanar distance, h k I is the set of indices, and a is the edge of the unit cube; Q = 22.99 A. *The values of I are visual estimates. All very weak peaks are given the value I = 1. tValues of d are from measurements made on the plotted curve; the uncertainty in a decreases from about 0.1 to 0.05 A.

8378

Chemistry: Pauling
Al55CuoLi53
I* 16 20 6 6 173 130 27 21 24 163 26 68 125 19 25 63 40 5 6 35

Proc. Natl. Acad. Sci. USA 85 (1988)

Table 4. Analysis of 14 strong maxima in the pulsed neutron powder diffraction pattern of the icosahedral quasicrystal

Table 2. Analysis of 20 computer-evaluated maxima for the powder neutron diffraction pattern of the rapidly quenched alloy

A174Mn2ISi5
a, A hkI 211 (22.0) 320 6.403 23.09 4.940 332 23.17 4.596 500 22.98 22.92 3.820 600 444 22.95 3.313 731 22.87 2.977 2.810 733 23.00 830 23.06 2.699 920 2.502 23.06 2.442 23.03 922 2.379 932 23.06 2.328 941 23.04 2.272 10 1 1 22.95 2.154 871 23.00 10 5 0 23.01 2.058 11 3 0 22.97 2.015 12 0 0 23.06 1.922 12 2 2 23.03 1.868 12 3 3 23.03 1.810 23.01 Average of 19 I is the intensity, d is the interplanar distance, h k I is the set of indices, and a is the edge of the unit cube; a = 23.01 A. *These values of I were obtained in the computerized analysis. tCalculated from values of 20. *The variation in 20 for the five alloys with composition A174 (Cr,Fe,Mn)2lSi5 indicates the uncertainty in a to be 0.1-0.05 A. This maximum may be caused by another phase. SNot included in average.

d,t A 8.98

3271 12 6 0 25.77 3587 12 8 3 25.80 3832 14 6 4 25.82 3927 16 3 1 25.80 4192 16 6 2 25.79 4493 18 4 0 25.79 4811 19 5 2 25.79 25.79 Average of last 11 I is the intensity, Q is the reciprocal length 2ir/d (in A-1), with d the interplanar distance, h k I is the set of indices, and a is the edge of the unit cube; the average value of a is 25.79 A. *Values of I are 1000 times those given in table Ia of ref. 16. tFrom table Ia of ref. 16. *Not indexed. Each of these very strong peaks probably represents two overlapping peaks.

I* 265 3257 2053 915 2970 577 20254 5585 1087 7047 1460 1270 3481 4437

1000 Qt 1066 1483

h kl 331 610 642 11 3 0 1140

a, A 25.69 25.78
25.78 25.77 25.82

1711t
1824 2780 2848

3148t

0.01 A-` in Q from about 0.50 A-, to 10 A-' or 25 A-'. In order to prevent subjective error, I did not interpolate to obtain more accurate values of Q for the peaks, but instead recorded in the tables as a peak every value of Q for which
Table 3. Analysis of 17 maxima in the high-intensity x-ray powder diffraction pattern of the icosahedral quasicrystal

the intensity was greater than for each of the two adjacent entries. A few entries in Tables 5-8 are shown with an additional significant figure for Q. These entries are taken from tables in ref. 16; they are for the stronger peaks, which comprise about 25% of all the peaks. Values of the intensity, I, for these peaks are those given in ref. 16. For the weak peaks I obtained values for I by subtracting from the value at Q(peak) a background value got by linear interpolation from the adjacent minima and multiplying by a factor given by the strong peaks.
Table 5. Analysis of 17 maxima in the pulsed neutron powder diffraction pattern of the icosahedral quasicrystal Al55CuoLi,5 hkl a, A 100Qt l* 222 26.22 83 37 25.91 100 410 170 25.69 331 265 106.6 116* 48 25.75 55 122 332 25.94 440 165 137 531 25.99 143 175 25.77 610 3257 148.3 25.79 621 770 156 165 631 25.79 1800 25.78 182.4 642 925
25.84 202 821 25.76 214 832 25.75 221 910 25.82 223 842 25.80 231 930 25.76 844 239 25.81 Average of last 15 is the reciprocal length 2ir/d (in A-'), with d I is the intensity, Q the interplanar distance, h k I is the set of indices, and a is the edge of the unit cube; the average value of a is 25.81 A. *Values of f are as described in the text. tValues of Q correspond to the peaks at intervals of 0.01 A when I is greater than for the two adjacent values. Seven extremely weak peaks, I < 30, have been omitted as probable fluctuations (100 Q = 46, 53, 56, 58, 68, 71, and 125). Values of Q with one extra decimal place are from table Ia in ref. 16. *This peak is not identified. 1030 60 70 140 190 2000

Al,6Cu102Li337
14.33*) 620 25.86 1059 25.82 1480 25.77 1707 25.86 1818 25.74 2264 25.80 10 4 1 2634 25.77 11 2 2 2769 25.806 11 6 3 3137 5231 25.810 13 3 1 3257 971 25.825 12 8 3 214 3584 14 7 1 25.798 3820 939 25.799 16 2 0 50 3927 25.826 17 1 0 4193 295 25.7% 17 7 0 4478 4892 25.813 18 6 4 4720 1206 25.817 18 8 0 4794 2599 25.810 Average of last 9 I is the intensity, Q is the reciprocal length 2ir/d (in A-'), with d the interplanar distance, h k I is the set of indices, and a is the edge of the unit cube; the average value of a = 25.81 A. *Values of I are 1000 times those given in table IHa of ref. 16. tFrom table Ha of ref. 16. *This peak indicates the presence of about 0.1% of the 160-atom cubic phase. Not included in average. An earlier analysis (8) led to the less accurate value, a = 25.69 A.
47 523 75 91 25 794 3723

1* 1

lo00 Qt

h kl (11 0t 331 610 700 642 921

a, A

Chemistry: Pauling
Table 6. Analysis of 20 maxima in the pulsed neutron powder diffraction pattern of the icosahedral quasicrystal Al55CujOLi30Mg5

Proc. Natl. Acad. Sci. USA 85 (1988)

8379

I*
6 3 <1 28 1 2 8 4 1 <1 243 211 <1 10 13 2500 6 6 665 1010

100Q
44t 49 51t 59

hkl

a,A

(1 11 24.7) 25.64 200 24.6) (2 0 0 211 26.09 24.4) 63t (2 1 1 26.18 72 300 25.73 311 81 25.91 222 84 26.12 90 321 24.1) (3 2 0 94t 26.08 331 105 510 25.84 124 24.5) 128t (5 0 0 521 25.68 134 441 25.97 139 610 25.82 148 541 25.94 157 26.02 162 630 700 25.87 170 25.83 642 182 25.91 Average of 15 I is the intensity, Q is the reciprocal length 27r/d (in A-1), with d the interplanar distance, h k I is the set of indices, and a is the edge of the unit cube; the average value of a is 25.91 A. *See footnote * to Table 5. tNot indexed on the basis of the 25.93-A unit; these very weak peaks may be caused by a small amount of a phase with cube edge about 24.3 A.

The curves show shelves and inflections indicating other diffractions. I have not included these unresolved peaks in the tables, except for the few that are given in the tables of ref. 16. The data for sample SEPD 1590 (Table 8), extending to Q = 10 A-1, show 60 peaks, 20 of which are included in Table 8. The 40 peaks beyond Q = 2.90 A-' are not included because the density of values of h2 + k2 + 12 becomes so great relative to the uncertainty 0.005 A-' in Q as to make the assignment of indices unreliable. They have, however, all been assigned indices with allowed values of h2 + k2 + 12 compatible with the values found for a. The three other sets of neutron diffraction data extend to Q = 25.00 A-1. Sample SEPD 1591 shows 166 peaks (of which 25 are given in Table 7), sample SEPD 1592 shows 164 peaks (20 in Table 6), and sample SEPD 1595 shows 173 peaks (17 in Table 5).

Table 7. Analysis of 25 maxima in the pulsed neutron powder diffraction pattern of the icosahedral quasicrystal A1510CU125Li235 Mg130 with two cubic crystals 100 Qt h kl a, A I* 60.8t 12 24.02 74 220 1 25.80 77 310 1 14.13 111 4 79 24.00 400 104.7 21 25.63 115 332 2 1 118 1 125$ 21 132S 13.96 300 100 135 25.76 440 138 100 13.79 310 80 14211 25.95 147.3 610 768 25.97 153 620 40 25.79 156 621 10 25.77 541 158 50 13.98 162 321 60 25.87 170 700 768 14.04 40 179** 400 25.86 155 181.8 642 25.78 811 20 198 25.84 202 821 100 25.86 831 209 40 25.84 662 20 212 752 25.93 214 10 14.02 Average of 5 25.83 Average of 14 I is the intensity, Q is the reciprocal length 2Ir/d (in A-'), with d the interplanar distance, h k I is the set of indices, and a is the edge of the unit cube. Values found for a are 14.02 A and 25.83 A. Maxima are assigned to one cube or the other to give the better agreement. *Values of I obtained as described in the text. tValues given to an additional significant figure are from table I1b of ref. 16. tNot indexed; a broad peak, probably overlap of 1 1 0 of 14.01 A and 2 1 1 of 25.81 A. Assignment uncertain; possibly a very small amount of the 820-atom cube, a = 24.01 A. Not indexed. I1This peak may also be indexed as 5 3 0, a = 25.80 A. **This peak may also be indexed as 7 2 1, a = 25.79 A.

AM510Cu125Li235Mg130
The pulsed neutron powder diffraction data for A1510CU125Li235Mg130 (Table 7) were largely compatible with a cubic unit with a = 25.83 A. Several lines indicated the presence of another phase, with a = 14.02 A. This phase, primitive cubic, may have the 160-atom structure of Mg32(Al,Zn)49, but with the clusters at 0 0 0 and '/2 1/2 '/2 antiparallel rather than parallel.

AS55CuIOLi35
Table 4 contains values of I and Q for 14 of the stronger peaks in the pulsed neutron powder diffraction pattern of Al55CuOLi35, from table Ia of ref. 16. Because of the replacement of some of the Al atoms by the larger atoms of Li, the lattice constant can be predicted to be slightly larger than that of Al6CuLi3, which is 25.70 A. From Table 4 it can be seen that 12 of the 14 reported strong peaks can be indexed in this way, leading to the average value 25.79 A for a. Table 5 includes 3 peaks from Table 4 and 14 others, mainly weak. Their analysis leads to a = 25.81 A, in good agreement with the Table 4 value.

AM510Cu12SMgMS
The pulsed neutron powder diffraction data for A1510Cu125Mg365 were analyzed with two primitive cubic phases, with a = 25.98 A and a = 14.07 A (Table 8). The second phase probably has the primitive 160-atom structure.

AI55Cu1OLi3OMg5
Analysis of 20 peaks, mainly weak, on the pulsed neutron powder diffraction pattern of Al55CuOLi30Mg5 (Table 6) gives the value a = 25.91 A. A table of Q and I values for this quasicrystal was not given
in ref. 16.

Ti56Ni23Fe.SiN6
Dunlap et al. (17) have recently reported the preparation of the icosahedral quasicrystal Ti56Ni23Fe5Si16 and its analysis in terms of the cut-and-projection method from six dimensions. I measured the values of Q for the first 35 peaks on their

8380

Chemistry: Pauling

Proc. Natl. Acad. Sci. USA 85 (1988)

assumption that there is little change in average atomic volume when the compounds are formed, lead to approximately 820 or 1012 atoms in the unit cubes, supporting the

Table 8. Analysis of 20 maxima in the pulsed neutron powder diffraction pattern of the rapidly quenched alloy A1510CU125Mg365 with two cubic crystals hkI lo Qt I* a, A 7 45 100 13.96 245 110 14.10 63.2 4 84 222 25.91 22 91 321 25.83 95) 20 10 105 331 26.08 8 109 211 14.12 111 421 25.94 3 4 113 332 26.08 1 422 25.94 119 141 310 28 14.09 400 13 179 14.04 191 73 2 25.90 3 811 2 1% 26.04 840 168 215 26.14 422 14.06 219 3 227 500 14.11 98 441 257.2 14.04 2775 610 14.05 4456 271.9 54 1 7 14.09 289 14.07 Average of 10 25.98 Average of 10 I is the intensity, Q is the reciprocal length 2ir/d, with d the interplanar distance, h k 1 is the set of indices, and a is the edge length of the unit cubes. Maxima are assigned to one cube or the other to give the better agreement. Average values of a are 14.07 A and 25.98 A. *Values of I were obtained as described in the text. tValues given to an additional significant figure are from table lb of ref. 16. tNot indexed; a very broad peak, with shelves indicating four
components.

theory of icosahedral twinning of cubic crystallites in the icosahedral quasicrystals.

I thank Christian Janot, S. J. Poon, Dorothy Munro, Zelek Herman, and Jerry Latter for their help. 1. Shechtman, D., Blech, I., Gratias, D. & Cahn, J. W. (1984) Phys. Rev. Lett. 53, 1951-1953. 2. Levine, D. & Steinhardt, P. J. (1984) Phys. Rev. Lett. 53, 24772480. 3. Cahn, J. W., Shechtman, D. & Gratias, D. (1986) J. Mater. Res. 1, 13-26. 4. Bak, P. (1986) Scr. Metall. 20, 1199-1204. 5. Kalugin, P. A., Kitaw, A. Y. & Levitov, L. S. (1985) Pis'ma Zh. Eksp. Teor. Fiz. 41, 119-121. 6. Pauling, L. (1987) Phys. Rev. Lett. 58, 365-368. 7. Pauling, L. (1988) Proc. Natl. Acad. Sci. USA 85, 36663669. 8. Pauling, L. (1988) Proc. Nail. Acad. Sci. USA 85, 45874590. 9. Bergman, G., Waugh, J. L. T. & Pauling, L. (1952) Nature (London) 169, 1057. 10. Bergman, G., Waugh, J. L. T. & Pauling, L. (1957) Acta Crystallogr. 10, 254-259. 11. Audier, M., Sainfort, P. & Dubost, B. (1986) Philos. Mag. B54, L105-L111. 12. Ma, Y., Stern, E. A. & Gayle, F. W. (1987) Phys. Rev. Lett. 58, 1956-1959. 13. Henley, C. L. & Elser, V. (1986) Philos. Mag. B53, 159-163. 14. Elswijk, H. B., DeHosson, J. T. M., van Smaalen, S. & de Boer, J. L. (1988) Phys. Rev. Lett. 57, 4261-4264. 15. van Smaalen, S., Bronsveld, P. & deBoer, J. L. (1987) Solid State Commun. 63, 751-753. 16. Shen, Y., Poon, S. J. Dmowski, W. Egami, T. & Shiflet, G.-J. (1987) Phys. Rev. Lett. 58, 1440-1443. 17. Dunlap, R. A., O'Handley, R. C., McHenry, M. E. & Chatteijee, R. (1988) Phys. Rev. B37, 8484-8487. 18. Pauling, L. & Kamb, B. (1986) Proc. Natl. Acad. Sci. USA 83, 3569-3571. 19. Pauling, L. (1987) Proc. Natl. Acad. Sci. USA 84, 4754-4756. 20. Pauling, L. (1988) Phys. Rev. B, in press. 21. Shen, Y., Dmowski, W., Egami, T., Poon, J. & Shiflet, G.-J. (1988) Phys. Rev. B37, 1146-1152.

Also 5 3 0, 25.98
a

A.

published curve and found that they all are compatible with primitive unit cube with edge 24.24 A (20).
Conclusion
The values of the cube edge a given in the tables, together with values of the atomic volume of the elements and the