This dissertation has been microfilmed exactly as received 6 8-8827
G R E E N E , Lawrence Conde, 1913- SERIES ASSIGNMENTS IN THE FLUORESCENCE LINE SPECTRA OF HIGH PURITY CADMIUM SULFIDE.
The Ohio State University, Ph.D„ 1967 Physics, solid state
University Microfilms, Inc., Ann Arbor, Michigan SERIES ASSIGNMENTS IN THE FLUORESCENCE LINE SPECTRA
OF HIGH PURITY CADMIUM SULFIDE
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosphy in the Graduate School of The Ohio State University
by
Lawrence Conde,Greene, B.S., M.A.
* * * * * *
The Ohio State University
1967
Approved by
Department of Physics ACKNOWLEDGMENTS
We wish to thank D, E, Johnson for his work in drawing preparation,
H, A, Wilson for the crystal preparation, L, W. Croft for the spectro- graphy, as well as Professor Richard C, Nelson and the late Professor
Charles H, Shaw for careful criticism of the original manuscript. VITA
October 23. 1913 B o m - Milton, Wisconsin
1948 B.S., Western Michigan University, Kalamazoo, Michigan
1950 M,A,, Kalamazoo College, Kalamazoo, Michigan
1951-1967 .. Research Physicist, Aerospace Research Laboratories, Wright-Patterson AFB, Ohio
PUBLICATIONS
'’Emulsion Studies of Cosmic Ray Stars in Metal Foils", co-author I. Barbour, Phys, Rev. 79, 406 (15 Jul 1950).
"Properties of a CdS Photorectifier", co-authors D. C. Reynolds and L. L. Antes, J. Chem, Phys. 25, 1177 (Dec 1956),
"Crystal Growth Mechanism in Cadmium Sulfide Crystals" co-author D, C, Reynolds, J. Appl, Phys, 219, 559 (Mar 1958),
"Growth and Properties of CdS Crystals", co-author D, C. Reynolds, pres. International Conf, on Solid State Physics, Brussels, Belgium (June 1958),
"Method for Growing Large CdS and ZnS Single Crystals", co-authors D. C, Reynolds, S, Czyzak and W, M, Baker, J, Chem, Phys, 29, 1375 (Dec 1958), “
"An Improved Furnace for the Growth and Treatment of II-VI Compound Single Crystal Platelets", co-author Charles R. Geesner, J. Appl, Phys, 38, 3662 (Aug 1967).
iii TABLE OF CONTENTS
Acknowledgments...... ii
V i t a ...... iii
Tables ...... vi
List of Figures . . .vii
SECTION
I. INTRODUCTION...... 1
II. BACKGROUND...... 4
Excitons Phonon Assisted Edge Emission Crystal Growing
III. CRYSTAL GROWTH AND CONTROL OF CRYSTAL DEFECT STRUCTURE...... 23
Apparatus The Edge Emission Bands Pure Argon as the Carrier Gas Argon and Hydrogen Sulfide as the Carrier Argon and Cadmium Vapor as the Carrier Argon and Oxygen as the Carrier
IV. DISCUSSION OF DEFECT STRUCTURE AS RELATED TO CRYSTAL GROWTH...... 37
V. THE LOW TEMPERATURE EMISSION LINE SPECTRA OF CADMIUM SULFIDE ...... 42
Experimental Procedure Results The I2JJ Line The S Series The T Series The V and W Series The U and U' Series
iv Crystal Stoichiometry
The Relation of the Emission Lines to the Defect Structure Concluding Remarks
v TABLES
Page TABLE I The S Series in the Fluorescence of Cadmium Sulfide at 4 . 2 ° K ...... 52
TABLE II The T Series in the Fluorescence of Cadmium Sulfide at 4 . 2 ° K ...... 54
TABLE III The V and W Series in the Fluorescence of Cadmium Sulfide at 4.2°K ...... 57
TABLE IV The U and U 1 Series in the Fluorescence of Cadmium Sulfide at 4 . 2 ° K ...... 58
vi LIST OF FIGURES
Figure Page
1. Crystal Lattice for the Wurtzite Structure ...... 6
2. First Brillouin Zone for Cadmium Sulfide...... 8
3. Band Extrema in Cadmium Sulfide ...... 9
4. Cross Section Diagram of Platelet Furnace ...... 25
5. Furnace Temperature Profile Suitable for Growing Undoped CdS Crystals...... 26
6. CdS Platelets ...... 28
7. Fluorescence Spectra of CdS Grown with Pure Argon as the Carrier, Quenched...... 30'
8. Fluorescence Spectra of CdS'.E^S, Quenched...... 32
9. Fluorescence Spectra of CdSjE^S, Slowly Cooled.. . 33
10. Fluorescence Spectra of CdS:Cd, Quenched ...... 35
11. Fluorescence Spectra of CdS:Cd, Not Quenched ...... 36
12. Energy Level Diagram for CdS Showing Proposed Defect Level Assignments...... 38
13. Emission Spectrum of Cadmium Sulfide (4.2°K) Showing Dominant L i n e s ...... 46
14. Emission Spectrum of Cadmium Sulfide (4.2°K) Showing I £g Line...... 49
15. Emission Spectrum of Cadmium Sulfide (4.2°K) Showing First Three Lines of S Series ...... 50
16 Emission Spectrum of Cadmium Sulfide (4.2°K) Showing First Three Lines of T Series ...... 53
17 Emission Spectrum of Cadmium Sulfide (4.2°K) Showing V and W Series...... 56
vii Figures Page
18. Emission Spectrum of Cadmium Sulfide (4.2°K) Showing U and U1 Series...... 59
19. Emission Spectra of Cadmium Sulfide (4.2°) Doped with Sulfur...... 61
20. Emission Spectra of Cadmium Sulfide (4.2°K) Doped with Cadmium ...... 62
viii I. INTRODUCTION
The subject of this dissertation is the experimental investigation of the line structure seen in the fluorescence spectra of single cry stals of cadmium sulfide. The origin of these lines has been explained in terms of the widely accepted exciton picture, which has been quite extensively investigated by optical reflection, absorption and Zeeman splitting measurements. In this paper the primary interest will be in the separation of these spectra into groups or series of lines of common origin and the study of the nature of the center associated with each series.
The fluorescence spectra of many compounds having the wurtzite structure have been shown to lie in a region just to the long wavelength side of the absorption edge which contains a very large number of lines, many of them quite narrow. This structure has been explained largely on the basis of exciton decay, with phonon assistance. An exciton is the bound electron-hole system formed by a Coloumb attraction between an excited electron and a hole remaining in the valence band. Such a charge system, formed when an electron is excited from the valence band to the conduction band, may move away from the region of formation through normal diffusion processes, giving up its energy of formation on recombination, but cannot transport charge since it is electrically neutral. It can however transport energy and mass, Exciton lines are best seen at very low temperatures since they broaden rapidly with increasing temperature, and for most materials the excitons are thermal ly dissociated at roan temperature. 1 The exciton spectra in semiconductors and insulators can provide a great deal of information concerning the details of the electronic structure of the material. In the past a number of optical methods have been used in such investigations, Among these methods are trans mission, reflection, photoconductivity and fluorescence. Fluorescence at low temperatures has probably been the most widely used of these methods.
It will be the purpose of this paper,' (1) to report on the detail ed structure of the fluorescence spectra near the absorption edge of cad mium sulfide in pure crystals, that is crystals containing no added foreign impurity; (2) to study the relationship between known defects and inpurities and the spectral line structure and from these relationships to make source assignments for certain line series whose origin has been in question.
The line structure near the absorption edge in cadmium sulfide has been studied sporadically for a number of years. Some of the more prominant lines have been given assignments as due to intrinsic excitons, some as due to excitons bound to certain types of charged or neutral centers, and others as being phonon-assisted satellites. However, very little has been done toward making definite defect assignments for the centers associated with the bound excitons.
The excitons responsible for the fluorescence lines studied in this work are considered to be those bound to lattice defects or for eign impurities. It is the purpose of this paper to relate these
lines to specific defects by attempting to introduce or remove them 3 preferentially. Although these bound exciton lines have been under study for some time, it has been only recently that crystals of suffic iently high quality have become available to make possible a really definitive investigation.
In these studies all the measurements were made on single crystal platelets of CdS grown from the vapor phase. The crystals were un doped, that is they contained no deliberately added impurity, but were grown under conditions such that the presence of native or stoich iometric defects was to be expected. II. BACKGROUND
Cadmium sulfide in its crystalline form has received much atten tion because of its position midway between the ionic insulating cry stals such as the alkali halides and the covalent semiconductors such as silicon or germanium. Cadmium sulfide crystals of high purity and of a high degree of structural perfection may have a dark resistivity of as much as 10^ ohm-cm. On the other hand very small additions of suitable inpurities of structural defects may reduce the resistivity to less than an ohm-cm, thus giving it many of the properties of norm al semiconductors. It is as a photoconductor, however, that cadmium sulfide first attracted widespread attention. The presence of certain defects or inpurities can have the result of increasing by many orders of magnitude the lifetime of photoexcited carriers, with resultant gain in photosensitivity. The ratio of carriers in the external cir- 4 cuit to photons absorbed may be as high as 10 , In addition, when pro vided with suitable electrodes, cadmium sulfide crystals can be made to serve as very sensitive photovoltaic cells.
In recent years the luminescence of cadmium sulfide crystals has become the subject of intensive investigation, The sulfides and selen- ides of cadmium and zinc as well as their solid solutions have been studied for many years because of their importance in commercial appli cations such as television tubes.
4 5 At atmospheric pressure cadmium sulfide may exist in the zinc- blende structure with A ■ 5.82A and nearest interatomic distance
2,52A, or in the wurtzite structure with A ■ 4.14A, Co ■ 6.72A and nearest interatomic distance » 2.52A, In general usage when one speaks of single crystal cadmium sulfide one refers to the wurtzite struc ture and this usage will be continued in this paper. Other properties of cadmium sulfide are:
Density 4,82 g/cm3
Indexes of refraction at 5893A: 2,506(w), 2,529(£)
Low frequency dielectric constant 11,6
Band gap at room temperature 2,48 eV 2 Electron mobility at room temperature 300 cm /volt-second
Effective mass of free electrons Me » 0.2m
Effective mass of holes % 11 ■ 5 m, ■ 0,7m Where 11 and are the effective masses of the hole 11 and J_ to
the c axis.
Cadmium sulfide crystallized in the wurtzite structure has a c/a ratio 1,624 which is somewhat less than the ideal close-packed sphere value of 1,633, The lattice has no center of symmetry. Each atom of one species is surrounded by four atoms of the other species, (Fig. 1),
The unit cell contains two sulfur atoms and two cadmium atoms. The point group is C6y which corresponds to the symmetry of the hexagonal plane. The unit vectors are directed along the trirectangular axes
Qxf Oy, Oz, The hexagonal Bravais lattice is defined by the vectors 6
Fig. 1 Crystal Lattice for Wurtzite Structure. The reciprocal lattice which is also hexagonal is defined by the vectors
"Si = 4? (?T+ IT), "E, = 4a7 , “B, = fk .
The Brillouin zone is a prism with a hexagonal base, (Fig* 2),
It has all the symmetries of the point group of the crystal.
The structure of the energy bands of cadmium sulfide has been extensively studied in optical reflection, absorption and fluorescence spectra. It is generally assumed that the valence band is made up of p-levels and that the conduction band is made up of s-levels. The minima of both valence and conduction band occur at k ■ (0,0,0) in the
Brillouin zone. The crystalline field and spin-orbit coupling remove the degeneracies of the p-levels of the valence bands so that three distinct bands are formed, (Fig, 3), Using this model as a basis, selection rules have been formulated from the symmetry properties assoc- iated with the bands. From these selection rules it is found that optical transitions corresponding to the passage of an electron from the highest valence band (r9* ^ T 7) are allowed only if the absorbed light is polarized perpendicular to the c-axis of the crystal. Trans itions corresponding to the passage of an electron from one of the lower valence bands to the conduction band(IV r7 )are independent of polar ization of the absorbed light.*
^Additional discussion of the crystallography and band structures of cad mium sulfide can be found in, Aven and Prener (Editors) ,. Physics and Chemistry of II-VI Compounds (North-Hoiland Publishing Company, Amsterdam, 1967.) A k
Fig. 2 First Brillouin Zone for Cadmium Sulfide. 2 .5 8 2 eV
0.016 eV *•
0 .0 5 7 e V
k = 0
BAND EXTREMA I n'C dS
Fig. 3 Band Extrema in CdS 10
Excitons
When an insulating crystal is radiated with photons of energy greater than that of the forbidden gap of the crystal, electrons will be raised from the valence band to the conduction band, and holes will remain in the valence band. One or both of these can act as charge carriers, and photoconductivity will result if a voltage is applied across the crystal. If there are centers in the crystal which can increase the effective lifetime of one of the carriers to a value great er than that of the carrier transit time, then photocurrent gain is obtained, and very large photocurrents may result.
After excitation, recombination must occur. This is a very much more complicated problem involving traps and recombination centers in the forbidden band. In the process of recombination the carrier loses the excitation energy in the form of optical radiation and lattice vi brations. Ordinarily the direct transition from the conduction band to the valence band is forbidden so that the electron must return in steps by way of recombination centers. It is these recombination cen ters that determine the structure of much of the fluorescence spectra of semiconducting and insulating crystals.
A most important and informative type of recombination transi tion is that occuring through exciton states. An exciton is an elec- tron-hole pair. The two oppositely charged particles are bound to gether into hydrogen like states, with discrete energy levels. The decay of such an exciton produces very characteristic sharp lines in the fluorescence spectrum. n The existence of such nonconducting excited electronic states in solids probably was first proposed by Frenkel.* He proposed that the
1. J, Frenkel, Phys, Rev, 37_ 17, 1276 (1931), individual atoms in the solid might be excited by the absorption of a photon and that this excitation might then migrate from atom to atom in 2 a particle like manner. Later Wannier suggested in an alternate hy-
2. G. H. Wannier, Phys. Rev. 52, 191 (1937), pothesis that the electron and hole are weakly bound, with a separation
large compared with the lattice constant. Because of the coulomb attrac tion between them a hydrogen like state is formed with an energy smaller than that of the band gap.
The actual existence of such a quasi-particle was open to question for a considerable period of time. However, in recent years the experi- 3 4 5 ments of Hayashi and Katsuki , Gross and Nikitine have proved the
3. M. Hayashi and K, Katsuki, J. Phys. Soc. Japan jj, 381 (1950),
4. E, E, Gross, Nuovo Cim. Suppl, £, 672 (1956),
5. S, Nikitine, J, Phys. Radium 17, 817 (1956), existence of the exciton beyond any doubt. It was first seen as a hy drogen- like series of narrow lines in the absorption spectrum of cuprous
oxide at low temperatures. These series converged to a limit and showed
line separations which could be determined from energy level assignments
derivable from a modified Rydberg formula. Very many crystals show this
complex structure in the absorption spectra, and also in fluorescence,
reflectivity and photoconductivity spectra. 12 If we look at the exciton in the Wannier formulation as an electron-
hole pair bound by a Coulomb potential,
v = - SL € r where £ is the dielectric constant of the medium, we see that we
should have bound states of the exciton system lying below the bottom
of the conduction band« These exciton levels will then be given by the
Bohr formula:
~/x e44 tn= 2ti2 * where Y\ is the principal quantum number and ft is the reduced mass of the exciton I _ _j _____ L f j * ‘ m e m h In the decay of such an exciton one ought then to see lines at energies determined by h v x\ = ^gap -t-gap- E — — h n 2 where E gap is the band gap energy and Eg is the exciton binding energy *« /A* E4 e b = 2 fi2 € 2 Complexes in which excitons are bound to neutral or charged donors 6 7 or acceptors have been described by Haynes and Lampert. In crystals 6, J, R, Haynes, Phys. Rev, Letters £, 361 (I960), 7, M, A, Lampert, Phys* Rev. Letters, 1, 450 (1958), with a direct band gap, such bound excitons are seen as sharp absorption or emission lines at longer wavelengths than those of the intrinsic ex citons. The four possible simple point defect exciton complexes may be represented by @ — 4- Exciton bound to a charged donor © ■+" " Exciton bound to a charged acceptor © — H- Exciton bound to a neutral donor O H Exciton bound to a neutral acceptor Excitons bound to neutral donors or neutral acceptors are always possible and frequently both are seen in the same material and even in O the same specimen. However, it can be shown that in a given material excitons should not bind to both ionized donors and ionized acceptors, 8, J. J. Hopfield, "Proc. of the 7th Intern. Conf. on Phvs. of S e m i c o n d u c t o r s (Academic i’ress, hew York, Wb4j p /2b and may bind to neither if the electron-hole band mass ratio is near one. Bound excitons may also be associated with various defect com plexes, such as donor-acceptor pairs. Phonon Assisted Edge Emission Many of the II-VI compounds show one or more series of fluorescence peaks on the long wavelength side of the absorption edge. The peaks in a series are equally spaced in energy, the spacing being equal to the longitudinal optical phonon energy for the particular lattice. These fluorescnece bands are broad as compared with bound exciton transitions. In undqped cadmium sulfide there are two such series. One of these series appears only at low temperatures, near that of liquid helium, and for this reason is often designated simply as the low temperature green series. The zero phonon peak is at 5175A and the longitudinal optical phonon separation in cadmium sulfide at 4,2°K is 0.038 eV, The other series usually occurs only at higher temperatures being seen very well at 77°K, The zero phonon peak occurs at 5130A, The green fluorescence bands in cadmium sulfide which have just g been described were probably first seen by Kroger, Other investiga- 9. F, A. Kroger, Physica Tj 1 (1940), tors also examined this fluorescence at 77°K and at 4,2°K, so that the spectral positions and temperature behavior of the bands have become well established. A model involving recombination between free elec trons and holes trapped at inpurity centers was suggested for silver 10 11 1 •* doped cadmium sulfide by Schoen and later by Klasens. Later, Lambe 12 and Klick , on the basis of studies on the relative decay rates of 10. M. Schoen, J. Phys. 119, 463 (1941). 11. H. A. Klasens, Nature 158, 306 (1946). 12. J. J. Lambe and C. C. Klick, Phys. Rev. 9£, 909 (1955). fluorescence and photoconductivity in silver doped cadmium sulfide, re versed the mode of interaction between electrons and holes, and sug gested that the transition for the characteristic silver emission was the recombination between trapped electrons and free holes. In a later paper, Lambe, Klick and Dexter^ extended this interpretation to 13. J, J. Lambe, C, C. Klick and D. L, Dexter, Phys, Rev. 103, 1715 (1956). account for the green fluorescence bands in undoped cadmium sulfide. Crystal Growing Since the beginning of recorded history the beauty lying in the color and symmetry of natural crystals has held a continual fascina tion for men. From ancient times artisans have attempted to duplicate such crystals in glass. However, it has only been in mo d e m times that they have succeeded in actually duplicating crystals found in nature and only in the last few years have scientists surpassed nature in color and variety, extending the catalogue of crystalline materials far beyond the bounds of those found in nature. Of the many different types of crystals known to the crystallo- grapher, a few have become of special interest to other physicists because of some physical property; electrical properties quite differ ent from those of the metals, optical properties whose study permitted an extension of the then recently developed quantum theory to an under standing of the electronic structure of solids, as well as mechanical and acoustical effects of tremendous interest. Increased understanding of these properties and phenomena led in turn to the development of a wide variety of useful applications such as photocells, rectifiers, transistors, and acousto-electric and thermo-electric transducers to mention only a very few of the almost unbelievable number of solid- state components used in present day technology. There are probably as many techniques for growing crystals as there are materials to be crystallized. However, there are a few basic methods which will include all of these. Probably the earliest was the growth from a solution. At first this meant growth from an aqueous solution, or perhaps from some other common solvent liquid at room temperature. This is still a popular method for crystallizing substances that might decompose chemically under other methods. Where solubility is too low, it may be increased by heating under pressure. This leads to the hydrothermal method used for the production of syn thetic quartz crystals. Again a molten salt may be used as a solvent as in the growth of barium titanate crystals, A somewhat later development was crystallization from a melt. Here the molten charge might be dropped or pulled through a temperature gradient as in the Bridgman-Stockbargers method. In the Czochralski 17 method a seed crystal is dipped into the melt and then slowly with drawn, while in the Vemeuil process the molten material is dripped onto a seed crystal. Almost any material that can be dissolved with out chemical change or melted without vaporizing or decomposing, can, in principle at least, by crystallized by one of these methods or by variants of them. However, there still remains a rather large class of materials that cannot be conveniently dissolved or melted. Some of these cannot be melted because of their very high vapor pressure at their melting temperatures, These are the materials that are crystallized using vapor phase techniques which will be described in more detail. Among the crystals of particular interest because of unusual elec trical and optical properties, cadmium sulfide, greenockite in its natural crystalline form, very early received widespread attention. Cadmium sulfide was first extensively studied as a luminiphore, but its extraordinary electro-optical properties attracted attention even before single crystals were available. Lorenz14 was probably the first to report growth of cadmium sulfide crystals. Allen and Crenshaw15, 14, R, Lorenz, Chem, Ber, 24_, 1509 (1891), 15, E. J, Allen and J, L, Crenshaw, Am, J, Sci, 34_, 310 (1912), reported producing microcrystals from a solution. However, the first method to produce crystals of usable size was the dynamic vapor phase method, reported by Frerichs,16 He heated cadmium metal in a small 16. R. Frerichs, Phys, Rev. 72, 594 (1947), 1S porcelain boat inside a quartz tube to a temperature of 800 to 1000°C. The cadmium vapor formed was carried by a slow stream of hydrogen into a reaction zone where it was mixed with hydrogen sulfide brought in through a separate tube. The furnace was slowly cooled with the gases flowing and crystals of very pure cadmium sulfide were deposited on the quartz walls as the proper temperature was reached. Crystals produced by this method are generally of the plate or ribbon type, often in mica like bundles which can be separated into large thin sheets. Crystals as much as a centimeter in length can be obtained in a sufficiently large system. The crystals are always of the wurtzite structure with the "c" axis in the plane of the plate. An improvement of Frerichs method was introduced by Bishop and 17 Liebson, In their method cadmium metal and sulfur were heated sep- 17, M. E, Bishop and S, H, Liebson, J, Appl, Phys.24, 660 (1953), arately and the vapors carried into a reaction chamber by a flow of argon. Crystals of several square centimeters area and as much as millimeter thick were grown inside the sphere. The methods of Frerichs and of Bishop and Liebson had the advantage of considerable purification of the components before reaction, but had serious disadvantages of a technical nature, particularly in the area of suitable furnace design. As increasingly pure cadmium sulfide be came available it naturally seemed advantageous to grow crystals direct- 18 ly from the powdered compound, Such a method is described by Stanley, 18, J, M, Stanley, J. Chem. Phys, 24, 1279 (1956), 19 High purity cadmium sulfide is sublimed in a slowly moving stream of an inert gas and the vapor carried to a cooler region of the system for deposition. Both prism and platelike crystals are grown by this method, but one ordinarily seeks conditions which will produce thin platelets. The three methods described in the preceding paragraphs are refer red to as dynamic growth methods because they involve the transfer of vapors by a carrier gas. The crystals produced will be largely plate like or needle like. In an attempt to develop methods for growing bulk 19 crystals which could be machined or cleaved, Reynolds and Czyzak 19, D. C. Reynolds and S. J, Czyzak, Phys, Rev, 79, 543 (1950), devised a "static" method. This was first applied to zinc sulfide, but except for the temperatures used, the procedure is precisely the same for cadmium sulfide. In this method a charge of highly purified powder was distributed over a length of several inches in a quartz tube. This charge was then placed in the center of a horizontal resistance furnace. The tube was filled with hydrogen sulfide and sealed. The furnace was held at a temperature of 1000° (for CdS) for periods ranging from 48 to 96 hours. Crystals formed directly on the charge in the form of hexa gonal prisms. Prisms as large as one cubic centimeter have been grown. A second static method20 has made it possible to grow crystals 20 20. L. C. Greene, D. C. Reynolds, S. J. Czyzak and W. M. Baker, J. Chem. Phys. 29, 1375 (1958). weighing as much as 300 gm and more than 1*1/2 inches in diameter, the only limitation of the size of the crystals being the size of the fur nace. Cadmium sulfide powder, previously dried and densified by a series of heat treatments, is placed in a quartz tube about two inches in diameter and closed at one end by a flat quartz disk. A second sim ilar but slightly smaller tube is inserted such that a small space is left between the charge and each quartz plate. This assembly is placed in a larger ceramic tube and the whole placed in a specially designed furnace, A pressure of inert gas is maintained in the tube until the furnace has attained operating temperature, about 1250° for cadmium sulfide, and then the tube is sealed. The design of the furnace is such that a small temperature differential exists between the center of the charge and the quartz plates. The charge migrates from the charge to these seed plates and deposits there as single crystal grains. It is believed that nucleation occurs on dislocation like defects on the quartz surfaces. The grains apparently grow rapidly parallel to the seed plate, until a mosaic is formed and then grow perpendicular to this direction. Grains of less favored orientation may be crowded out so that a few large single crystal grains remain at the end of the growing period. 2J Crystals grown by this seed plate method have different growth habits from those grown by other techniques. The growth direction is perpendicular to the seed plate if the temperature distribution is optimum. However, the c-axis of the crystal is tilted by as much as 14° to the seed plate normal. In the other methods described the growth surfaces are crystallographic planes. A modification of the bulk crystal growth technique was later made 21 by Piper and Polich, Instead of a flat seed plate they used a cone 21, W, W, Piper and S, J, Polich, J, Appl, Phys. 3^2, 1278 (1961) shape surface provided with a quartz heat sink for initial nucleation. This seed cone was moved slowly from the hot region of the furnace into a cooler region. One might call this a horizontal vapor phase Bridgeman arrangement. A somewhat different vapor phase growth technique for CdS has come into limited use in the recent years. It is referred to as the "chem ical transport method". Here a halogen such as chlorine or iodine and a small pressure of hydrogen sulfide is sealed with the charge in a quartz tube. The tube is placed in a temperature distribution such that at the charge position the cadmium halide vapor is stable, whereas at the deposition area the cadmium sulfide is stable. In effect the halogen picks up the cadmium from the charge, deposits it as cadmium sulfide and then goes back for more. This method is used at much lower temperatures with the advantage of lower defect concentrations. It has the serious disadvantage of halogen contamination. In this investigation is was necessary to use crystals having un damaged surfaces. Such "as grown" surfaces are available in prisms grown by the static method or in platelets grown by the dynamic method. The later method was chosen because of its simpler provision for intro ducing vapors of sulfur or cadmium over the growing crystals. The de sign and operation of the furnace developed for this purpose will be discussed in detail in the next section. III. CRYSTAL GROWTH AND CONTROL OF CRYSTAL DEFECT STRUCTURE In the investigation of the fluorescence properties of semicon ductors (and certainly other optical as well as electrical properties) control of the defect structure is essential. It is customary to try to achieve this control in cadmium sulfide by the application of a ser ies of treatments in furnaces separate from the growth furnace. Before beginning these treatments the specimen is etched to remove any surface contamination which might diffuse into the crystal. The crystal is then baked at a suitable temperature in a vacuum to remove any stoich iometric excess of either component element. When cold it is sealed in a capsule with a small amount of cadmium metal, sulfur or hydrogen sul fide, and again heated for several hours. In the present investigation the crystals were grown in a carrier gas of carefully controlled composition, which could be neutral, oxydiz- ing or reducing, or an atmosphere containing a doping element such as cadmium or sulfur. After the crystal growth was completed the crystals were given such treatment as was necessary without removal from the furnace. Since it is recognized that the defects produced by a given carrier gas composition applied during growth need not necessarily be the same as those produced by exposure to the same atmosphere in a post- growth diffusion process, crystals grown in an inert carrier were given a vacuum bake at 350-400°C and then exposed to the chosen composite gas at 600-800°C, At the end of the treatment the crystals were either cool ed very slowly over a period of several days to anneal out less stable 23 defects, or quenched quickly with chilled argon gas to freeze in the high temperature equilibrium concentrations of these defects. Apparatus The body of the furnace consists of porous firebrick, the inner bricks being cut so as to form a chamber with a square cross section, (Fig. 4). Silicon carbide resistance heaters are used in vertical pairs. Three separately controlled heating zones make it possible to maintain a desired temperature profile. Figure 5 shows an example of such a profile, one used for pure cadmium sulfide. The growing chamber itself consists of a mullite cylinder with glass joints cemented to its ends. Inside the mullite cylinder is a quartz cylinder open at one end and with an exhaust vent at the other. The charge and dopant boats are placed near the open end of the quartz tube as in Fig, 4, Each boat contains a thermocouple well. The thermocouple tubes leave the furnace through rubber glands so that they can be used for positioning the boats. The progress of the.crystal growth can be observed through a glass window (not shown in the illustration), In operation, the boats are loaded with weighed quantities of the raw material and dopant and positioned as in Fig, 4. In this paper we will be discussing cadmium sulfide, but the procedures are equally applicable to the other wurtzite II-VI compounds. The carrier gas is introduced into the outer tube of the growing chamber through a stop cock (Fig. 4). CdS PLATELET FURNACE 9 * DEPOSITION REACTION ' ' ZONE *: z o n e:; EXHAUST -T.C. -T.C. THERMOCOUPLES CARRIER GAS LEGEND - CHARGE - DOPANT • - GLOBAR [•':-;v'-.v/| _f ir e b r ic k V//////A - CORE Fig. 4 Cross section diagram of platelet furnace. UlfO 5 Furnace temperature profile suitable for growing undoped CdS CdS undoped growing for suitable profile temperature Furnace 5 Crystals, DEGREES CENT. 1000 1200 1100 0 0 9 800 E G R A H C E N O Z (INTOCAVITY) INCHES TC DEPOSITION E N O Z 20 25 27 The gas passes the full length of the furnace before entering the growing chamber to assure temperature equilibrium. The cadmium sulfide and dopant vapors are picked up by the hot carrier and transferred to a region of a temperature suitable for deposition. The material is de posited as thin plates varying from a few micrometers to a few tenths of a millimeter in thickness and having surface areas of as much as a square centimeter, (Figure 6), The Edge Emission Bands, The visible fluorescence spectrum of cadmium sulfide at 4,2°K is divided into several fairly well-defined regions. These are: (1) the intrinsic exciton region below 4700A, (2) the bound exciton region from 4500A to 4900A, (3) a region from 4900A to 5100A containing many lines, most of which are phonon wings of the bound excitons, (4) the region from 5120A to 6000A which contains the much studied green phonon assist ed edge emission bands, and (5) one or two rather broad bands in the region from 6000A to 7000A, Any given specimen will not in general show all of these. The purpose here will be to observe the changes in the green and orange bands as the growth conditions and carrier gas com positions are varied however, as these bands have been associated with specific defects by other writers,^ 22, L, S, Pedrotti, and D, C, Reynolds, Phys. Rev,, 120, 1664 (1960) B. A. Kulp and R. H, Kelley, J. Appl. Phys. 31, 1057 (1960) B. A. Kulp, Phys, Rev. 125, 1865 (1962) 29 Pure Argon as the Carrier Gas - The charge used in these experi ments was high purity cadmium sulfide synthesized from the purified elements* The center of the charge region was held at 1200°C and the center of the deposition region at 1000”C. The carrier gas flow was held at 75 cc/min per square centimeter of growth chamber cross section. The total transfer time was one and one-half hours. In one series of runs the crystals were quenched to room tempera ture in a flow of cold argon. Specimens were selected immediately and mounted in a glass optical dewar for observation. The fluorescence was excited with a Hanovia 250 watt Hg-Xe arc lamp with a Kopp No. 41 U.V. filter. The spectra were photographed with a Cenco 87102 spectrograph equipped with Gaertner high precision slit and Bausch and Lomb 2" x 2" grating having 15,000 lines per inch. The reciprocal dispersion of this instrument is 17A/MM, Figure 7 shows the spectra at 77°K and 4,2°K at comparable exposure. It is to be observed that the green bands are very much more intense at 77°K than at 4,2°K and that there are two series of bands at 4,2°K, There is no emission in the 6000A to 7000A region. In a second series of runs pure argon was again used as the carrier, i and growth conditions were the same as in the above runs except that the crystals were cooled slowly at about 25° per hour in the furnace for 48 hours. The fluorescence spectra of these crystals showed only very weak green bands, most specimens none at all. None of the samples showed orange bands, Fig. 7 Fluorescence spectra of CdS grown with pure argon as the as argon pure with grown CdS of spectra Fluorescence 7 Fig. 4900 REL. LOG INTENSITY carrier quenched. carrier 5200 m in 505800 5500 ANGSTROMS 4.2 #K 77° 77° K 6100 406700 6400 30 Argon and hydrogen sulfide as the carrier - Numerous runs were made using argon with hydrogen sulfide in varying proportions. It was found that when the hydrogen sulfide exceeded about one part in eight of the mixture, the crystal structure of the deposit was adversely affected, A typical run using 90 parts of hydrogen sulfide to 750 parts of argon will be discussed. Here again two series of runs were made, one series being quenched in cold argon, the other slowly cooled over a 48-hour period again in pure argon. The spectra resulting from the quenched run is given in Fig. 8, showing again the green bands at both 4,2°K and 77°K, Here the two series of bands are of comparable intensity in contrast to those shown in Fig. 7 where the longer wave length series was much more intense. Again there is no emission in the 6000A to 7000A region, The spectra from the slowly cooled crystals are given in Fig, 9, The double series of green bands seen at 4,2°K in all of the quenched runs has now become a single well defined series. Argon and cadmium vapor as carrier - In order to introduce cadmium vapor into the carrier, a weighed quantity of purified cadmium metal was placed in the dopant boat. Both boats were kept in the cool region of the furnace until its operating temperature was reached, at which time the boats were positioned in their proper temperature zones. It was found to be most important that the cadmium supply should not be ex hausted while the furnace is at operating temperature as this results in a reduction in the amount of excess cadmium in the crystals. 4900 REL. LOG INTENSITY 15A^£0-^5225 5 A 2 2 5 ^ - 0 £ ^ 5145 A 5200 Fig. 8 Fluorescence spectra of CdSjI^S, quenched. CdSjI^S, of spectra Fluorescence 8 Fig. 5135 A A 5135 ,LA 5215 A 0 0 3 5 5 A 3 3 5 A L H 3 5 5500 ANGSTROMS 5800 77" K 77" 6100 6400 6700 4900 REL. LOG INTENSITY Fig. 9 Fluorescence spectra of CdSst^S, slowly cooled. slowly CdSst^S, of spectra Fluorescence 9 Fig. 5200 5145 A 5145 5215 5215 A 0 A 0 3 5 5500 ANGSTROMS 5800 6100 4.2* K 6400 33 6700 34 The fluorescence spectra of CdS:Cd crystals quenched in cold argon are shown in Fig. 10. The green bands at 4.2° are much narrower than those in the spectra from the crystals grown in argon or argon- hydrogen sulfide mixtures. If excess cadmium were present in the form of cadmium interstitials one would expect to see the orange bands be- 23 lieved to be associated with this defect. None are observed, 23. B. A. Kulp, Phys. Rev., 125, 1865 (1962) Some crystals of CdSjCd slowly cooled showed a very striking spec tral change upon annealing at room temperature, In Fig. 11 the upper spectrum was made using a CdSjCd crystal fresh from the furnace. It shows a very dense orange band, but no green. The lower spectrum was made from the same crystal which had been kept at room temperature for three days. The orange is completely gone while the double green series predominates, Argon and oxygen as the carrier - A mixture of argon and oxygen was used as the carrier in a series of crystal growing experiments. In succeeding runs the proportions of oxygen was varied from a few parts per million to 0,1$, The crystals were cooled slowly to room tempera ture over a period of 20 hours. The fluorescence spectra of these crystals showed no lines or bands to which one might assign oxygen as the responsible impurity, In the green edge emission region the spectra were identical to that of crystals grown in argon-cadmium vapor mixtures as shown in Fig, 10, 4900 REL. LOG INTENSITY ---- Fig, 10 Fluorescence spectra of CdS:Cd, quenched. CdS:Cd, of spectra Fluorescence 10 Fig, 5200 1 ______ m m 5500 I ______ ANGSTROMS 5800 I ______ 6100 I ______ 6400 I ______ 35 6700 36 £ £ 2 5130 A 52IOA 4900 5200 5500 5800 6100 6400 6700 ANGSTROMS Fig. 11 Fluorescence spectra of CdSjCd, not quenched. Upper spectrum is from a specimen fresh from the furnace showing the dense orange band. The lower spectrum is from the same crystal three days later. Both spectra were made at 4.2°K. IV, DISCUSSION OF DEFECT STRUCTURE AS RELATED TO CRYSTAL GROWTH It is not the purpose of this paper to discuss the theory of de fect emission in CdS, but rather to show that the defect structure involved in the fluorescence can be controlled in the crystal growing furnace. To this end a brief discussion of a model explaining the observed emission is useful. The energy level diagram shown in Fig. 12 is based on the two level model proposed by Pedrotti and Reynolds 24 24, L. S. Pedrotti and D. C. Reynolds, Phys. Rev. 120, 1664 (1960). as explaining the temperature dependence of the green bands in CdS. 25 It has been shown that the green bands at 77UK are associated with the presence of the sulfur interstitial. This center, (x) would then 25, B* A. Kulp and R, H, Kelley, J. Appl. Phys. 31^, 1057 (1960) be expected to be an acceptor, so that the other center, (Y), ought to be a donor, that is a sulfur vacancy or cadmium interstitial. Now it 9ft has also been shown that the orange emission is associated with the cadmium interstitial. However, the orange and green bands are not al ways seen in the same specimen, so that the sulfur vacancy would seem to be the more logical choice as the donor involved in the green emission. Ii 26, B, A. Kulp, Phys. Rev, 125 1865 (1962). 37 3S CONDUCTION BAND Cd: < < iD O ^ ro 5 * •D S 2 3o ID < O O ID cm' 2 VALENCE BAND Fig. 12 Energy level diagram for CdS showing proposed defect level assignments. / 39 The representative spectra of cadmium sulfide crystallized in an inert gas and quenched, was seen in Fig. 7. Assuming the validity of the model, the bands seen at 77°K correspond to the free-to-bound transition involving the sulfur interstitial; at 4.2°K these bands as well as those arising from the bound-to-bound transition are seen. In the slowly cooled crystals all of these bands are very generally weakened, or completely missing at both temperatures. This certainly seems to indicate that the bands are associated with sulfur Frenkel defects which are frozen in by the fast cooling and annealed out by the slow cooling. In Fig, 8 the fluorescence spectra of cadmium sulfide crystals grown in a flow of argon containing H 2S and quenched are seen. The same transitions are seen as before except that here the free-to- bound transition is comparatively very much stronger. In the spectra of slowly cooled crystals shown in Fig. 9 only the free-to-bound transition at both temperatures are seen. These data might very well indicate that in the quenched crystals there exist quenched-in sulfur Frenkel defects plus an additional concentration of sulfur interstit ials. In the slowly cooled crystals most of the Frenkel defects are annealed out leaving the sulfur interstitials predominant. The crystals grown in argon-l^S mixture both show a predominance of sulfur Frenkel defects in the quenched runs. When the crystals are grown in an argon-cadmium vapor atmosphere and quenched, these defects appear only very weakly. In Fig. 10 a series of very narrow green bands representing a center to center transition is seen. One would 40 expect either sulfur vacancies or cadmium interstitials under these growth conditions; however, the orange band characteristic of the cad mium interstitial is not seen. It seems very reasonable to believe, then, that the transition responsible for these bands may involve the sulfur vacancy, as has been assumed. In the quenched Cd doped cadmium sulfide the narrow green bands at 4.2° are seen, but in the slowly cooled crystals only the strong orange emission characteristic of interstitial cadmium (Fig. 11) appeared when the crystals are fresh, changing to a spectrum which is characteristic of the Frenkel defect when the crystals have been held at room temperature for a few days. It would seem that annealing oc curs during the slow cooling in which the centers responsible for the green bands are replaced by cadmium interstitials. Further annealing would occur at room temperature with the replacement of cadmium inter stitials by sulfur interstitial-vacancy pairs. That the spectra seen in crystals grown in an excess of cadmium vapor should appear in crystals grown in the presence of oxygen seems at first surprising. However, cadmium vapor acts as a reducing agent, and it has been shown27 that oxygen also is a reducing agent toward 27, H. Woodbury, J. Phys. Chem Solid 27, 1257 (1966). cadmium sulfide in the temperature range 1000° to 1200°C. Cadmium sulfide crystals grown in the presence of oxygen might then be ex- " pected to have sulfur vacancies as the dominant defects. This ex pectation gives weight to the belief that the green bands seen at 4.2°K in these crystals and in those grown in the presence of cadmium vapor are indeed associated with the sulfur vacancy. 41 Regardless of the interpretation put on the spectra, we can say with reasonable assurance that with the furnace described here we are able to grow cadmium sulfide platelets with a greatly improved control of the defect structure. Previous work indicated that this furnace should be equally applicable to the growth of cadmium selenide, zinc sulfide and antimony trisulfide. V. THE LOW TEMPERATURE EMISSION LINE SPECTRA In Section III a furnace of improved design was described. In this furnace it was possible to grow single crystal platelets of high purity cadmium sulfide. It was shown that the native defect structure of these crystals can be influenced by the use of suitable growth con ditions and post-growth treatment. Since the crystals are of the platelet form with nearly perfect surfaces, they can be used to study the low temperature emission line spectrum of cadmium sulfide and to investigate the relationship between the spectrum and the defect struc ture of the crystal. At low temperatures the ultraviolet excited fluorescence spectrum of high purity single crystal cadmium sulfide shows a complex line and band structure in the visible region. Between W O O A and 5100A there 9ft occur many sharp lines ° which arise from the decay of excitons bound 28. E. Grillot, J. Phys. Rad. IT 822 (1956) to defect or impurity centers, either directly or with phonon assistance, while in the region between 5100A and 5700A there appear the various series of phonon assisted green edge emission bands, which are various- 29 ly interpreted as arising from distant pair recombination processes 30 or from nearest neighbor pair recombination. 29. D. G. Thomas and J, J, Hopfield, Proc. 7th Intern. Conf. Phys. of Semiconductors. Paris (1965). p. 67 30. 0. Goede and E. Gutsche, Phys. Stat. Sol, 17, 911 (1966) k2 43 Complexes in which excitons are bound to neutral or charged defect centers have been described by Haynes1 and by Lampert. Thomas and 31. J, R, Haynes, Phys. Rev. Letters _4, 361 (1960) 32. M. A. Lampert, Phys. Rev. Letters 1^ 450 (1958) Hopfield^ as well as Reynolds and Litton^ have discussed bound ex- citon spectra in cadmium sulfide, but considered mostly the magneto optical behavior of the more prominent lines. 33. J. J, Hopfield and D. G, Thomas, Phys. Rev. 122, 35 (1961) D. G. Thomas and J. J, Hopfield, J, Appl. Phys. 33, 3243 (1962) “ D. G. Thomas and J. J. Hopfield, Phys. Rev. 128, 2135 (1962) 34. D. C. Reynolds and C. W. Litton, Phys. Rev. 132, 1023 (1963) Of the many lines appearing in the complex bound exciton spectra of cadmium sulfide crystals a very intense line appears, in both ab- 35 sorption and emission at 4869.1A. Handelman and Thomas refer to this 36 line as I2g. Zeeman studies indicate that this line is associated 35. E, T. Handelman and D. G. Thomas, J. Phys. Chem. Solids 26, 1261 (1965) with a transition involving an exciton bound to a neutral center. On the basis of spectral position and g-value it is suggested that this center is a donor. 44 36 Closely associated with I2fi are the lines I2C at 4870*2A ^ *2A at 4867*2A# Tiiese are thought also to be associated with 36. D. C. Reynolds and C, W, Litton, Phys. Rev. 132, 1023 (1963) 37 38 transitions involving excitons bound to a donor, Yee and Condas 37. J. J. Hopfield and D. G. Thomas, Phys. Rev. 122, 35 (1961) 38. J. H, Yee and G. A, Condas, Solid State Electronics 10, 257 (1967) “ have reported the observation of phonon wings of the emission line 4867.2A at 4941A and 5019A. 39 A very sharp line, 1 ^ is often seen at 4888.5A. At sufficiently 39. J. J. Hopfield and D. G, Thomas, Phys. Rev. 122, 35 (1961) D. G. Thomas and J. J. Hopfield, Phys. Rev. 116, 573 (1959) high resolution this line is seen to consist of two components at 4888,1A and 4888,7A, and has therefore been referred to as a zero field split doublet.40 It has been shown 41 that in a magnetic field both components of 1^ show linear Zeeman splitting and so would arise from 40. J, J. Hopfield and D. G. Thomas, Phys. Rev. 122, 35 (1961) 41. D, G. Thomas and J, J. Hopfield, Phys. Rev. 128, 2135 (1962) a complex having one unpaired spin in both the excited and ground states, which would be an exciton bound to neutral donor or acceptor . Measure ments made by the same authors on the thermal behavior of these Zeeman split lines in absorption have shown that the center involved is an 1+5 acceptor. Longitudinal optical phonon wings of 1^ have teen observed in emission at 1+962A and 5039A by the same authors. Experimental Procedure The cadmium sulfide crystals used in this study were grown by subli mation in a carrier gas flow in the furnace described in Section III. The carrier gas was always argon with carefully measured additions of hydrogen sulfide or cadmium vapor. The argon and the hydrogen sulfide were metered into the furnace through precision flowmeters. The de sired pressure of cadmium vapor was maintained by keeping the cadmium boat at a separately controlled temperature, the cadmium vapor pres- li 2 sure being determined from published tables. 1+2. An N, Nesmayanov, Vapor Pressure of the Elements. (Academic Press, New York, 1963) Results The low temperature fluorescence spectrum of high purity cadmium sulfide ordinarily contains a very complex array of narrow lines in the spectral region from 1+800A to 5100A as is shown in Fig. 13. Most of these lines are believed to be associated with the decay of bound excitons, either directly, or with phonon assistance. It should be possible to separate these lines into phonon assisted series, but it has previously been difficult to make such a separation unambiguously because of the overlapping of the several series and because of the non-reproducibility of the spectra from specimen to specimen. Attempts to classify the lines on the basis of multiphonon assignments using 1+3 phonon energies obtained from infrared absorption studies produce completely ambiguous results. Fig, 13 Emission spectrum of cadmium sulfide (4.2°K) showing dominant dominant showing (4.2°K) sulfide cadmium of spectrum Emission 13 Fig, 4800 RELATIVE PLATE DENSITY lines. 90 00 5100 5000 4900 AEEGH (ANGSTROMS) WAVELENGTH 4.2° K ro 10 5116 A (T 3) 46 HT 43, M, Balkanski and J, M, Besson, J, Appl. Phys, 3£, 2292 (19613 W, G. Spitzer, J. Appl. Phys. 34, 792 (1963). S. S. Mitra, Phys. Letters 249 (1963) In this study we have been able to grow cadmium sulfide crystals that show several of the various series separately, so that it be comes possible to make the series assignments by observing which lines always appear together. The spectra of several hundred specimens were studied in order to be certain that the grouping assignments were real and not merely fortuitous. It is to be emphasized that it is not the purpose of this paper to discuss the observation of emission lines not previously reported, but rather to show empirically that the complex emission line spectrum of cadmium sulfide can be resolved into well defined groups and to show the relationship between these groups and the stoichiometry of the cry stal. Furthermore, this resolution was not carried out on the basis of phonon-like line separations, but strictly on the basis of co-occur rence, The dominant emission lines were resolved into seven groups, of which six were clearly defined series. In each of these series it was indeed found that the lines were separated by energies near that 44 of the longitudinal optical phonon energy of 0,038 eV. In some of the series phonon assistance is obvious; in others it is strongly in dicated, but differences in line separations leave unanswered questions, 44. R. J. Collins, J. Appl. Phys. 30, 1135 (1959) 48 The I2B Line ■ In all of the cadmium sulfide crystals which we have studied the I2B line at 4869,1A has been present; in most speci mens it dominated the spectrum. It was our purpose to determine if there were phonon wings associated with I2B, A careful study of the line spectra of many specimens revealed some that contained only I i_ , and a companion line at about 4866A, The line spectrum of Z D ^ such a crystal is shown in Fig. 14. This spectrum contains no lines which can be associated with I2B as phonon wings, nor did increased exposure reveal such lines. We conclude then that I2B is not part of a phonon series. The S series - A group of lines which we have observed to always occur together are those at 4857A, 4928A, 5005A and 5086A. These lines are seen most prominently in sulfur doped crystals. If the cry stal is not too heavily doped with sulfur, the series may appear accompanied only by the lines near 4869A, as shown in Fig. 15, We have chosen to designate this group of lines as the S series. The SO line at 4857A seems not to have been discussed in the liter ature, possibly because it so nearly coincides in wavelength with the r 8 exciton line at 4854,5A and lies at the same wavelength as the T 6 line at 4857A.^ However SO cannot be associated with the 45, J, J. Hopfield and D* G, Thomas, Phys, Rev, 122, 35 (1961) decay of an intrinsic exciton because its appearance is very strongly dependent upon the stoichiometry of the crystals as will be shown 49 THE I2 GROUP 4.2° K >- CM »- U>z Id O Id 0. Id > H < -J Id tr 4850 4900 4950 5000 5050 WAVELENGTH (ANGSTROMS) Fig. 14 Emission spectrum of cadmium sulfide (4.2°K) showing I9R lines. Fig. 15 Emission spectrum of cadmium sulfide (4.2°K) showing first showing (4.2°K) sulfide cadmium of spectrum Emission 15 Fig. 4850 RELATIVE PLATE DENSITY in to 0> three lines of S series. S of lines three eV 2 7 3 0 . 0 AEEGH (ANGSTROMS) WAVELENGTH 4900 CM to 905000 0 0 5 4950 eV 5 7 3 0 . 0 H S SERIES S THE m in CM CO 5050 51 below* Furthermore, it is strongly polarized E J. C whereas the T 6 line at the same position is known to be polarized EIIC. The line separa tions (Table I) are clearly suggestive of longitudinal optical phonon assistance. The T series - The group of lines, which for convenience we have chosen to call the T series, consists of lines at 4888.5A (Ij), 4962A, 5039A, and 5116A, Table 2. These lines have been discussed extensive ly in the literature^ and the first two line separations carefully measured. ^ We have found this series to occur predominantly in cry- 46. J, J, Hopfield and D. G. Thomas, Phys. Rev. 122, 35 (1961) 47. D. W. Langer, Y. S. Park and R. N. Euwema, Phys. Rev. 152, 788 (1966) stals heavily doped with sulfur, and always accompanied by the S series as shown in Fig. 16, However, since the S series frequently occurs without the T series the two cannot be presumed to be directly associated. The V and W series - In Fig. 17 there are, in addition to the lines already discussed, two clearly defined series. One of these, which we have called the V series, with lines at 4941A, 5019A and 48 5098A, has been recognized as phonon assisted with the zero phonon line at 4867.2A ( ^ a ^* zero P*1011011 I*11® *s not resolved in Fig. 17 because of its very intense neighbors. 48. J, H, Yee and G. A. Condas, Solid State Electronics 10, 257 (1967) *“ 52 Table I. The S Series in the Fluorescence of Cadmium Sulfide at 4.2°K line X(A) E(eV) A E (eV) SO 4857 2.5524 0.0372 SI 4929 2.5152 0.0375 S2 5005 2.4767 0.0387 S3 5085 2.4380 Fig. 16 Emission spectrum o£ cadmium sulfide (4.2°K) showing first showing (4.2°K) sulfide cadmium o£ spectrum Emission 16 Fig. 4850 RELATIVE PLATE DENSITY in three lines of T series. of T lines three WAVELENGTH (ANGSTROMS) 4900 0.0377 eV 0.0377 4950 H T SERIESTHE T .° K 4.2° 0.0377 eV 0.0377 5000 5050 54 Table II, The T Series in the Fluorescence of Cadmium Sulfide at 4,2°K line XCA) E(eV) AE(eV) TO 4888,5 2,53585 0.0377 T1 4962 2.4981 0.0377 T2 5039 2.4604 0.0373 T3 5116 2.4231 55 The other group, which we have called the W series has lines at 4916A, 4990A and 5068A, We have observed no lines at shorter wave length that might assumed to be the leading line for the series. Furthermore, although the V series and the W series usually occur to gether, we have seen many spectra in which one or the other appears alone. Hence, it cannot be assumed that they are different series associated with the same transition. For these reasons we have desig nated the 4916A lines as the leading line, WO, of the series. Table III shows the wavelengths, energies and line separations of these series. The V and W series appear prominently in crystals grown in a reducing atmosphere. The U and U* series - The lines discussed above occur in pure cadmium sulfide in various combinations in most specimens. There are, in addition, two series which are not often seen. These are the U series with lines at 4895.^A, 4969A, 5049 and 5133A, and the U' series with lines at 4916A and 5060A as shown in Fig. 18. As in the groups discussed, the line separations (Table IV) clearly suggest phonon assistance although these energies differ considerably from the accept ed value of 0.377eV,^ The U and U' series are seen most often in crystals grown in a carrier containing a very small addition of either oxygen or hydrogen. 49. D, W. Langer, Y. S. Park and R. N. Euwema, Phys. Rev. 152, 788 (1966) 4850 Fig. 17 Emission spectrum of cadmium sulfide (4.2°K) showing V and W and V showing (4.2°K) sulfide cadmium of spectrum Emission 17 Fig. RELATIVE PLATE DENSITY series. 0 .0 3 8 5 eV 5 8 3 .0 0 AEEGH (ANGSTROMS) WAVELENGTH o 4950 - * 0 .0 3 8 6 eV 6 8 3 .0 0 5050 THE Vand W CM .° K 4.2° SERIES if) 5150 56- 57 Table III. The V and W Series in the Fluorescence of Cadmium Sulfide at 4,2°K, line XCA) ECeV) AE(eV) VO 4867 2.5472 0,0385 VI 4941 2.5087 0,0386 V2 5018 2.4700 0.0377 . V3 5097 2,4323 WO 4916 2,5218 0.0369 W1 4989 2,4849 0,0363 W2 5069 2,4460 58 Table IV. The U and U' Series in the Fluorescence of Cadmium Sulfide at 4«2°K, line A (A) E(eV) AE(eV) UO 4895.5 2.5321 0.0373 U1 4969 2.4948 0.0400 U2 5050 2.4549 0.0400 U3 5134 2.4148 U'l 4981 2.4888 0.0389 U'2 5060 2.4499 59 28 0 . 0 3 7 3 e V THE U and U 0 . 0 4 0 0 eV SERIES 4.2° K 4 9 6 9 A (U I) >- h- 0 . 0 4 0 0 eV V)z L&J a 5 0 5 0 A (U 2) LlJ 0 . 0 3 8 9 eV I- < a.-J UJ > 513 4 A (U 3) H < -J 4981 A (U'O) UJ 01 5 0 6 0 A (U'l) 4850 4950 5050 5150 WAVELENGTH (ANGSTROMS) Fig. 18 Emission spectrum of cadmium sulfide (4.2°K) showing U and U' series. 6 0 Crystal stoichiometry - In order to obtain a better understanding of the relationship between the various emission series and the crystal stoichiometry, crystals were grown and treated as above using differ ent gas mixtures* The proportion of dopant (cadmium vapor or hydro gen sulfide) in the carrier gas was varied from 0*011 to 10%, In each run, after the deposition was complete, the crystals were soaked at 900°C for the sixteen hours in the same gas mixture used during growth. During the soak period the cadmium sulfide charge was also held at 900°C in order to provide an equilibrium atmosphere over the crystals. This treatment has been found to give improved reproducibility since it permits nonequilibrium defect structure introduced during growth to anneal out. Figure 19 shows the emission spectra at 4.2K of two cadmium sul fide crystals doped with sulfur. One specimen was lightly doped, being taken from a run in which the carrier gas was argon with 0,01% hydrogen sulfide. The other specimen is heavily sulfur doped. Its carrier contained 1,0% hydrogen sulfide. In both specimens I^g and the S series are dominant. The T series is very strong in the heavily doped specimen but does not ap pear in the more lightly doped one. The V series and the W series are very weak in both spectra. Figure 20 shows the emission spectra at 4,2K of cadmium doped crystals; one lightly doped and one heavily doped. The carrier gas compositions used were argon with 0,01% cadmium vapor and argon with 1,0% cadmium vapor. Fig. 19 Emission spectra of cadmium sulfide (4.2°) doped with with (4.2°) doped sulfide cadmium of spectra Emission 19 Fig. 80 90 00 5100 5000 4900 4800 RELATIVE PLATE DENSITY ______sulfur. SO SO 2B 1 > w TO AEEGH (ANGSTROMS) WAVELENGTH XI00 I\XIOO O W S2 S2 V2 T2 ___ 0.01%H,S 10%H* S I * S3 Fig. 20 Emission spectra of cadmium sulfide (4.2°K) doped with cadmium. with (4.2°K) doped sulfide cadmium of spectra Emission 20Fig. 4800 RELATIVE PLATE DENSITY SO WAVELENGTH (ANGSTROMS) 0 0 9 4 XIOO WO wo Wl 5100 0 0 0 5 S2 S2 V2 0.01% Cd0.01% .0% Cd .0% ,Xt) l 62 6 3 • In these spectra IgB is again prominent. The V and W series are strong while the S series, although present, is of low intensity. The T series is not present at all. From the spectra in Figures 19 and 20, relationships between the line intensities of the various series and the growth conditions are clearly seen. IgB dominant in all four spectra but it neverthe less can be seen to increase in intensity with decreasing hydrogen sul fide and increasing cadmium vapor pressures. This trend, although not strong, is clearly and consistently seen in the many spectra which we have studied. Both the V series and the W series show a similar but much stronger dependence on carrier gas compositon. It would appear, then, that the centers responsible for I2fi and the V and W lines are associated with cadmium excess defects. SO and the S series although present in all but the most heavily cadmium doped crystals show a strong dependence on carrier gas compo sition, increasing in intensity with decreasing cadmium pressure. The T series appears only in the spectrum of the specimen most heavily doped with sulfur. It appears then that the S series and the T series are associated with sulfur excess defects. The relationship of the emission lines to the defect structure. In their studies on low temperature emission from cadmium sulfide Pedrotti and Reynolds'^ observed two apparently quite different types of cry stals. One type whose emission appeared blue to the eye showed a 50. L. S. Pedrotti and D. C. Reynolds, Phys. Rev. 119. 1897 (i9 6 0 ). 44 characteristic spectrum whose lines were referred to by these authors as the blue or ”B” lines. The other type showed in addition to the "B" lines many other blue and blue-green lines and in addition the green edge emission bands. These green bands caused the emission to appear green to the eye, and for this reason all of the lines and bands not seen in the "blue” crystals were called "green” or "G” lines. No explanation was given for the difference between the two types. From the data presented in this paper it becomes clear that this difference is just a difference in stoichiometry. In the cadmium sulfide crystals grown and treated under the con ditions used in this study I was always dominant in the emission spectrum regardless of the composition of the furnace atmosphere. Figures.19 and 20 show an increasing intensity of I2q with decreasing sulfur and increasing cadmium pressure. On the basis of this rather weak dependence one might hesitate to assign a cadmium excess defect as the center associated with this line. However, it has been shown*** 51. J, J, Hopfield and D. G, Thomas, Phys. Rev, 122, 35 (1961) D. C, Reynolds and C, W, Litton, Phys, Rev. 132, 1023 (1963) that the center is a donor and hence might reasonably be expected to be associated with a cadmium excess defect, that is, either a cadmium interstitial or a sulfur vacancy as we have assumed, SO at 4857A and its associated lines of the S series also dominate in all of the specimens which we have studied except those most heavily doped with cadmium. Figures 19 and 20 show the variation of the in tensity of the lines of the S series with crystal stoichiometry. The 65 lines increase in strength with decreasing cadmium pressure. The evidence indicates that the center involved is a sulfur excess de fect, that is, either a sulfur interstitial or a cadmium vacancy, and hence an acceptor.There are at present no magneto-optical measure ments available to verify this point. When heated in Cd vapor, pure cadmium sulfide shows n-conduc- 52 tivityj when heated in sulfur vapor it becomes insulating. It has been assumed52 that in the cadmium doped material the defect situation is dominated by ionized sulfur vacancies compensated by free electrons, 52. W. Van Gool, Principles of the Defect Chemistry of Crystalline Solids, (Academic frress. New York. 19b6) whereas in the sulfur doped material the compensation occurs as the result of the presence of cadmium vacancies rather than free electrons. This situation can be described by the following equations: Cdfl ^V S+ e1 + Cdcd n[v;]«KCdPCd 0 =V| + Vid K<] $6, where Ks is the Schottky constant* When PC£j is increased, n and Vg increase and V C(| decreased. The overall result is an increase in n, the number of free electrons as seen in the equation n * [vj ]« Kct Pct With increasing sulfur pressure over the crystal, the cadmium pressure is suppressed according to the equilibrium equations Cdcd Sg ”2 Cdg PCd ^ w ^CdS so that from the equations [Vci]** [vs] = n = KC(j Kg * Pea it can be seen that increased sulfur pressure produces an increased concentration of cadmium vacancies and as a result a decreased density of free electrons* 6,7 Now if one were to assume that the center associated with I_D Zd were a sulfur vacancy and that the center associated with SO and the S series were a cadmium vacancy, then this explanation of the resistiv- ity-stoichiometry relationship would equally well explain the intensi ty- stoichiometry relationships for these lines. This model tells us that as the equilibrium between crystal phase and gas phase changes from high cadmium toward high sulfur, the density of sulfur vacancies decreases while the density of cadmium vacancies increases on the high cadmium side; the densities of both types of defects remains nearly unchanged on the high sulfur side. This behavior of the sulfur and cadmium vacancy densities very closely parallels the observed behavior of the I„T. (cadmium excess) line and SO (sulfur excess) line. Id The T series appears predominantly in sulfur doped crystals. Its line intensity is strongly dependent on sulfur pressure as is seen in Figure 19. The center involved is known to be an acceptor53 and so it is reasonable to assume that it is a sulfur excess defect. 53, D, G. Thomas and J, J, Hopfield, Phys, Rev, 128, 2135 (1962) We have observed, as have others,^ that TO and the high tempera ture green edge emission bands always appear together in pure cadmium sulfide. That a common center is involved has already been recog- 6 S nized,5^ Furthermore, it has been shown 55 that the center associated 54. E. T, Handelman and D. G. Thomas, J. Phys, Chem, Solids 26, 1261 (1965) 55. B. A. Kulp and R. H. Kelley, J. Appl. Phys. 31, 1057 (1960) with the high temperature green bands is the sulfur interstitial. These observations, then, lead us to the conclusion that the center responsible for the T series is also the sulfur interstitial. The V series with VO at 4867,7A is quite clearly associated with a cadmium excess defect, but the data available does not indicate whether the center is a sulfur vacancy or a cadmium interstitial. The same is true of the W series, but the longer wavelength of the leading line, WO, as compared with those of the other series would indicate the probability that the lines are associated with an exciton bound to a complex, possibly a donor-acceptor pair involving a cadmium ex cess defect as a component. The U and U' series are seen in crystals grown in a carrier con taining either hydrogen or oxygen. It is known56 that oxygen acts as a reducing agent toward cadmium sulfide under the temperature condi- 56. H, Woodbury, J, Phys, Chem, Solids 27_, 1257 (1966) tions used in vapor phase crystal growth. For this reason it seems most likely that the defect introduced by the presence of the hydro gen or the oxygen will be associated with a cadmium excess center. However, again, the longer wavelengths of the leading lines suggest the involvement of excitons bound to complexes, 6 9 Concluding Remarks In this paper the complex line spectrum of cadmium sulfide was simplified by separation into groups or series of lines whose co occurrence indicated a common origin. This was made possible by the development of a crystal growing furnace of a new design from which platelet type crystals with nearly perfect surfaces and with control led stoichiometry and native defect structure were obtained. It was shown that each of the groups of lines could be associated with a particular stoichiometric situation and, indeed, evidence was presented to indicate that several of the groups were associated with specific defects. These results open the way for a far more intensive study of the low temperature fluorescence of cadmium sulfide than has been previous ly possible since it now becomes possible to grow crystals with those emission series that are to be studied. It would be particularly in formative to study lines arising from the excited states of the bound excitons associated with the series which have been discussed. It would be difficult to identify these very weak series in the complex spectra of crystals in general use, but with the use of crystals such as are described in this paper such studies would be greatly simplified, The discussions in this paper have been limited to the A excitons. It may now become possible to study the series associated with the B and C excitons. Since these lines are associated with transitions to the B and C valence bands, they are so weak that they are completely obscured by their more intense neighbors in uncontrolled crystals. 70 Finally the understanding of the emission spectrum of undoped cadmium sulfide which has been provided by this paper makes possible a meaningful study of the emission of cadmium sulfide with foreign doping. Such doping may introduce new lines or bands which can now be identified, and may also enhance or surpress known lines or bands by compensation processes. No attempt has been made in this paper to discuss all of the low temperature emission lines in cadmium sulfide. There are for example, 57 strong lines at 4864A, 4866A, 4870.2A (I J and 4890A as well as 57, E, T, Handelman and D. G, Thomas, J, Phys, Chem, Solids 26 1261 (1965) many less intense lines. The discussion has been limited to those dominant lines which were clearly members of a series or which showed a definite relationship to the crystal stoichiometry. It would have been desirable to be able to make a more quantita tive presentation of the data relating the structure of the emission line spectra to the crystal growth conditions. Interfering phenomena such as non-reproducible equilibration of defects during cooling, deterioration of the crystal surface as a result of exposure to the atmosphere, and change of occupation state of the various centers while the spectra were being photographed, made such presentation impractical. It has been necessary then, for the writer to rely entirely on visual comparison between the various series rather than on a compari son of the measured intensities of individual lines.