An interesting and surprising finding is that oxides of cations so diverse in ionic radius, oxygen coordination number, and polarizability as Zn2+, B3+, Ti1+, and Pb2+ can form with bismuth oxide a discrete phase of the same symmetry. An appealing explanation is the concept of a clathrate- or cage-type structure, in which as postulated by Sillén [1] for Si2Bi24040, central Si atoms are surrounded by spheres of Bi12020 atoms. For the case of a central ion with valence different from 4, charge balance would be achieved through cation or oxygen adjustments.

Regarding the metastable b.c.c. phase, the monovalent ion Rb+ formed the phase on cooling in the high temperature x-ray experiments. The compositions containing NiO, MnO, and Nb2O5, which in the process of preparation were ground in alcohol, also formed the b.c.c. phase, metastably. These results would seem to support the conclusion that an impure b.c.c. phase of Bi2O, might be formed (metastably) with most cations, under the proper conditions of composition, grinding, and heating schedules.

It is seen from figure 7 that with the exception of Pbo the unit cell dimensions of the metastable phases are larger than those of the stable phases. The cell dimensions of the metastable impure phases, also, are less than those for the b.c.c. metastable phase of pure Bi2O. Therefore, the compositions of the impure metastable phases cannot be that of pure Bi2O. However, contrary to the case of the stable b.c.c. phases, no correlation exists between unit cell dimensions and ionic radius for the metastable b.c.c. phases. The x-ray diffraction patterns of the b.c.c. phases of the stable and metastable impurity forms are similar in d spacings and intensities to the pattern for pure Bi2O3. It is a reasonable assumption that the structures are similar.

To summarize (see fig. 7), the b.c.c. phase of pure Bi2O has the largest unit cell dimensions, and the addition of a foreign ion to Bi2O, tends to decrease the dimensions. This decrease is least for the larger ions, which tend to form the metastable b.c.c. phase. The decrease in cell dimensions is greatest for the smaller ions, which tend to form the stable b.c.c. phase. Whereas the stable b.c.c. phases show correlation with ionic radius, the metastable phases do not. These findings are compatible with a cage-type structure in which a central cation, including Bi, is surrounded by a sphere of atoms of approximately Bi12020 composition.

4. Summary

The important phase equilibria relationships for the bismuth-rich portions of the phase diagrams are shown schematically in figure 8. Elements in boldfaced type refer to the respective oxide mixtures studied. Elements enclosed in heavy outlines represent oxides which formed the stable b.c.c. phase with bismuth oxide. Composition of the b.c.c. phase was found to be variable for different systems, but most

nearly approached the ideal 12Bi: 1Me ratio for oxides of the tetravalent ions Si4+, Ti1+, and Ge1+. Designations in the upper right-hand corners of the boxes for the stable b.c.c. phases refer to the nature of melting, e.g., congruent, incongruent, or decomposition. position. Oxides of elements which formed the metastable b.c.c. phase are indicated by an M in the upper right-hand corner of the box, and those that formed the rhombohedral solid solution phase, by Rh. The nature of the liquidus curves is indicated by a designation in the lower right-hand corner of each box, as follows: E, simple eutectic; ssr, solid solution type with liquidus and solidus raised; SSL, solid solution type with liquidus and solidus lowered.

The authors acknowledge their sincere appreciation to Robert Friedman, who as a guest worker, summer 1961, from the University of Chicago, prepared the compositions and obtained some of the unit cell dimensions.

5. References

[1] L. G. Sillén, Arkiv fr Kemi, Mineral. Geol. 12A [18] 113 (1937).

[2] W. C. Schumb and E. S. Rittner, J. Am. Chem. Soc. 65, 1055-1060 (1943).

[3] G. Gattow and H. Schröder, Z. anorg. u allgem. Chem. 318 [3-4] 176–189 (1962).

[4] E. M. Levin and C. L. McDaniel, J. Am. Ceram. Soc. 45 [8] 355-360 (1962)

[5] E. M. Levin, C. R. Robbins, and J. L. Waring, J. Am. Ceram. Soc. 44 [2] 87-91 (1961).

[6] E. M. Levin and F. A. Mauer, J. Am. Ceram. Soc. 46 [1] 59-60 (1963).

[7] B. Aurivillius, Arkiv. Kemi, Mineral. Geol. 16A, No. 17, 1-13 (1943).

[8] L. G. Sillén and B. Aurivillius, Z. Krist. 101, 483-495 (1939).

[9] L. G. Sillén and B. Sillén, Z. Phys. Chem. B49, 27-33 (1944).

[10] L. Belladen, Gazz. chim. ital. 5211, 160-164 (1922). [11] See also, E. M. Levin, H. F. McMurdie, and F. P. Hall "Phase Diagrams for Ceramists", The American Ceramic Soc., Inc., fig. 97, p 58 (1956).

[12] K. K. Kelley, U.S. Bur. Mines Bull., No. 393, 166 pp (1936); p 26; Ceram. Absts., 16 [5] 162 (1937).

[13] P. Royen and K. Swars, Angew. Chem. 69 [24] 779 (1957).

[14] R. S. Roth, J. Research NBS 56 [1] 17-25 (1956); RP 2643.

[15] B. Aurivillius and L. G. Sillén, Nature 155, No. 3932, 305-306 (1945).

[16] G. Gattow, Z. anorg. Chem. 298, 65-71 (1959)

[17] R. S. Roth and J. L. Waring, J. Research NBS 66A [6] 451-463 (1962).

1. Introduction

The search for regularities among spectral lines emitted by tungsten arcs and sparks was initiated at the National Bureau of Standards in 1925 when O. Laporte and I were both and I were both new (temporary) new (temporary) members of the Spectroscopy Section. By exploiting existing data on wavelengths, intensities, and Zeeman patterns, Laporte [1]2 identified the first spectral terms and classified lines of WI. Zeeman data confirmed the designation of the six lowest energy levels as "D from 5d 6s2 and 7S from 5d6s1. This term analysis and quantum interpretation of W I was later greatly extended by Laporte and Mack [2] until it included 300 energy levels and 2,378 classified lines ranging in wavelength from 2008.64 Å to 11477.97 Å.

Shortly after Laporte found the low-energy terms in WI, I began a search for the theoretical low D (from 5d 6s1) and 6S (from 5d5) terms in W II, assuming that ionization removed one s electron. This initial search failed because of insufficient information about the spectrum W II. Then I undertook to reobserve the ultraviolet arc and spark spectra of tungsten, to distinguish WII from WI, to refine the wavelengths, and to observe Zeeman patterns of W II lines. From these observations, 27 even levels (including D and 6S) and 50 odd levels were derived from 500 W II lines ranging in wavelength from 1961.43 Å to 4348.13 Å; these results were published [3] in 1938 as a preliminary report on lines and levels of the spectrum W II. From time to time during the following quarter century, further improvements in observations and extensions of analyses have been made; the purpose of the present paper is to present the final data on 62 even levels, 132 odd levels, and 2,173 classified lines of W II, ranging in wavelength from 1756.6 Å W, to 6219.77 Å.

1 Former member of Spectroscopy Section, NBS; present address: Kiel, Wis. 2Figures in brackets indicate the literature references at the end of this paper.

2. Experimental Procedure

Small rods of very pure tungsten were used as electrodes in the arcs and sparks that served as light sources. The arcs were operated with applied potential of 220 v and 6 amp de, the sparks with 30,000 v, about 30 ma, ac, and 0.006 uf capacitance.

The first spectrograms were made in 1925 with a Hilger E 1 (Littrow) quartz spectrograph which could be effectively employed in the ultraviolet down to wavelength 1900 A. Later, these spectrograms were supplemented by a new series obtained by C. C. Kiess with the first Hilger E 185 (Littrow) quartz spectrograph that was delivered to the Bureau of Standards in 1932. This larger spectrograph has a focal length of 3 m, and reciprocal dispersion, or plate factor, of 0.3 Å/mm at 2100 Å and 1 Å/mm at 3000 A. These observations of tungsten lines on photographic plates exposed in quartz spectrographs were presented by me as a Master's Thesis at the University of Chicago [4]. Additional spectrograms were made in 1933 by Kiess with a concave grating of 22-ft radius, and 30,000 lines per inch. In a Wadsworth mounting, this grating produced spectra down to 1900 Å with a plate factor of 2.44 Å/mm in the first order, and with a plate factor of 0.88 Å/mm in the second order from 3200 to 4400 Å. This grating spectrograph was employed also to photograph tungsten spectra throughout the visible range.

Since all the above-mentioned spectrographs were stigmatic, movable diaphragms before the slits permitted placing strips of juxtaposed spectra on photographic plates. The usual procedure was to photograph the iron-arc spectrum containing standard wavelengths, then the tungsten arcs and sparks in adjacent strips and another iron spectrum beside the tungsten spark. The positions of tungsten spectral lines relative to iron lines were measured on Gaertner comparators, in 1925-26 at the Bureau of Standards; and in 1933-34 at Marquette University by courtesy of the late Fr. Joseph Carroll, S. J. Linear scale readings allowed the wavelengths of tungsten lines to be interpolated between iron

standards in the case of prismatic spectra with the aid of Hartmann's dispersion formula, and from grating spectrograms by assuming linear dispersion. In both cases, small corrections were made when the interpolated values of iron wavelengths departed from their standard values. For purposes of analysis, the final wavelengths of tungsten lines were converted to vacuum wavenumbers with the aid of the Table of Wavenumbers published by Coleman, Bozman, and Meggers [5].

At the same time that the positions or wavelengths of tungsten lines were being measured, their relative intensities were estimated in both sources. Although this intensity scale, ranging from 1 for a barely discernible line to 1000 for the strongest, is more or less arbitrary, it is primarily useful in separating W II from W 1, and also qualitatively supports the subsequent analysis in which intensities are related to quantum numbers. In some cases WII lines appear to coincide with W 1, and if a W II line sometimes appears stronger in I, the arc than in the spark, it may be an example of pole effect. Unfortunately, no deliberate attempt to exclude or exploit pole effect in the tungsten arc was made.

Finally, some information about tungsten spectra in the vacuum ultraviolet was obtained by measuring two spectrograms made by J. C. Boyce at the Massachusetts Institute of Technology in 1940 and kindly presented to the Bureau of Standards. These spectrograms were obtained with an evacuated grating spectrograph having a plate factor of 4.26 Å/mm. The spectra were measured down to 1200 Å. Although the excitation conditions were altered, the differences between spectra were

so slight that only above 1756 Å was it possible positively to identify W II lines. The wavelengths from this point to 1950 Å were finally corrected to the scale of values calculated from atomic energy levels.

My preliminary paper [3] of 1938 reported Zeeman-effect measurements of spectrograms made with a Weiss magnet and quartz spectrographs at the Bureau of Standards. These data have been superseded by those obtained with the more powerful Bitter magnet and grating spectrographs at the Massachusetts Institute of Technology. These results will be mentioned in the next section.

3. Results

The main result of the experiments mentioned above is that the wavelengths of about 13,000 spectral lines characteristic of arcs and sparks between tungsten electrodes have been measured. Among these, about 4,000 have been attributed to W II and 2,173 of these lines, including 90 percent of the total intensity, have been shown to arise from transitions between 194 derived energy levels, 62 even and 132 odd. The even levels are displayed in table 1 and the odd levels in table 2. In these tables, the electron configuration is given in col

umn 1, the spectral term designation in column 2, total angular momentum, J, in column 3, relative values of energy levels in column 4, intervals between levels of complex terms in column 5, and observed magnetic splitting factor, g, in column 6. Most of these data were supplied in 1956 for the third volume of Atomic Energy Levels [6] which contains J and level values for 45 even and 90 odd levels. In that volume, g-factors were reported for 38 even levels and for 58 odd. These were derived from MIT Zeeman spectrograms measured by J. E. Mack and Mrs. Taschek at the University of Wisconsin. Since 1956, the number of even levels has increased from 45 to 62 and odd levels from 90 to 132. Likewise, g-factors are now available for 46 even and 71 odd levels as compared with 38 and 58 respectively, in 1956.

In Atomic Energy Levels [6], it is stated that because of departure from LS-coupling it is difficult to assign term designations except for the lowest terms. Consequently, only six even terms, comprising 21 levels between 0 and 17437 K, were completely grouped and designated, and 24 higher even levels, between 18000 and 26929 K, were tentatively designated, guided by analogy with Mo II, g-factors, and theoretical calculations of the late R. E. Trees. Likewise, two odd terms, comprising six levels, were tentatively designated F and 2S, but the remaining 84 levels were given as miscellaneous. In the present paper, additional levels and g-factors are reported but no revision or extension of designations has been made.

The spectral lines of tungsten from which all the information in tables 1 and 2 was derived, are listed in table 3 in order of increasing wavelength. Each angstroms, by its estimated intensity in a tungsten line is described by its measured wavelength in arc and/or spark, by its appropriate vacuum wavenumber in kaysers (1 K=1 cm-1), by the difference (0-0) between the observed wavenumber and that calculated from the combination of energy levels shown in the next-to-last column. Abbreviated notes about the observed Zeeman patterns for 307 lines appear in the final column.

Table 3 contains 2,173 classified lines of W II, including 42 doubly classified. The only other table of classified WII lines was published [3] in 1938; it contained 500 lines, including 6 doubly classified. In addition to 2,173 classified lines, table 3 also contains wavelengths, intensities, and wavenumbers of 25 of the strongest W II lines still unclassified; these are included with the hope that the blank spaces under "combinations" may eventually be filled in.

The "Intensity" column of table 3 shows that the great majority of these WII lines were observed in arc as well as spark sources but with generally higher intensity in the latter when spectrogram exposures were chosen to make WI lines appear in both sources with nearly equal intensities. In a few cases, WII lines appear only, or stronger, in arc spectra; most if not all such deviations may be explained by coincident W 1 lines, pole effects, or air lines that occa

sionally mask metal lines in sparks operated at atmospheric pressure. Where the letter A follows an arc intensity, a line with nearly identical wavelength has actually been classified as a combination of W1 energy levels. The letter d in table 3 indicates a double line.

The observed minus calculated wavenumbers (0-C) have an average value of 0.1 K for wavelengths measured in air, that is, above 2000 A. This means that the average error in these wavelength measurements is of the order of 0.01 Å, and there is high confidence that the great majority of the combinations are real, rather than accidental. However, in the vacuum ultraviolet, below 2000 Å, the average 0-C is considerably greater, in a few cases as large as 5 K, which entails an error in wavelength of the order of 0.2 Å. In such cases, confidence that the combinations are genuine is greatly reduced but they have been tentatively retained for the following reasons. Since the scale of wavenumbers is greatly compressed in the vacuum region, and the tungsten spectra below 2000 Å were photographed with relatively small dispersion and measured without adequate standard wavelengths, it was inevitable that these measured wavelengths of WII lines would be more or less uncertain. Furthermore, it appears that most of these WII short waves involve the lowest and adjacent metastable energy levels, which is precisely what should be expected of strong lines with high energy. Therefore, these combinations with relatively large 0-C may be regarded as real pending wavelength measurements of higher

I

| patterns of W 1 and W II on MIT spectrograms were measured at the University of Wisconsin by J. E. Mack and Mrs. Taschek who supplied most of the magnetic magnetic splitting factors (g-factors) for tungsten levels that appear in Atomic Energy Levels [6]. Recently, those spectrograms were kindly loaned to me for examination. Without giving any details, I have merely written "res" to indicate resolved Zeeman patterns that agree with the combinations of W II lines in table 3. Unresolved Zeeman patterns are numbered 4, 5, 6, and 7, according to the notation of Back and Landé [7]. Incidentally, Zeeman types 4, 5, and 6 uniquely characterize, and fix J-values of, those responsible for WII lines. All the Zeemanenergy levels belonging to even multiplicities such as effect spectrograms are thickly covered by patterns of W I lines, and the W II patterns are often obscured, or vice versa, but despite these difficulties, the combinations are generally verified.

Finally, reference is made to the Tables of SpectralLine Intensities by Meggers, Corliss, and Scribner [8] which list 1,300 tungsten lines, 1168 W 1 and 132 W II, observed in an arc between copper electrodes containing 0.1 atomic percent of tungsten. It is noteworthy that no WII lines appear among the first 28 lines of highest intensity; they are all WI. Tungsten is similar to rhenium in this respect, but unlike thorium or uranium in which spark lines predominate in this type of arc source. Ionization potentials may be responsible for this difference, since the I.P. of copper is less than that of tungsten or rhenium but greater than that of thorium or uranium. In the list of all observed lines of tungsten [8], all of the W II lines are now classified, and will be given in this paper.

In conclusion, let me say that since my association with the Bureau of Standards in 1925-26, when this investigation began, and again in 1943-44, when I collaborated on the uranium project [9], this work on W II has been pursued as a part-time occasional avocation with the goal of finding order in a complex spectrum.

5d1 (5D) 6s

a 6D

02

1/2

212

32

42

5d5

a 6S

22

5d4 (3F) 68

a 4F

1/2

21/2

3/2