.. V . VV . V V . V V. .V V. V. V . ..V..V“ V V . ...”mwwmumwk 1 . . . ._ V V V V .. . . 3 am . V . .. . V V V WV .. W N _ V . w W .W .3 V V g V m w _. _. V . V .m. m w. E n. . ._ . C V _ 5 A” Wm. EV . . V M _ W . ..uw. C . . V N 8 Vi... . . V V V . V . _ m, V . . V. m .4 V . . ‘ u . V . n V . ‘ . . V m V. .. V _ ...... H V . -... V .. .....lVV} ._ H V. Va ”flu-s: 9 This is to certify that the thesis entitled A Development of the Character Tables for Certain Classes of Unitary and Linear Groups presented by William Albert S impson has been accepted towards fulfillment of the requirements for Ph 0 D 0 degree in Mathematics Q, mea 34W Major pmfessor Date (LU/5‘”); 3),) fill 0-7639 / 1/1. ABSTRACT A DEVELOPMENT OF THE CHARACTER TABLES FOR CERTAIN CLASSES OF UNITARY AND llNEAR GROUPS . By William A. Simpson The classical groups U(n,q2) and GL(n,q) occur as in- finite families of groups indexed by a dimension n and a prime power q. It is convenient to develop what might be called 'abstract' character tables whose entities are written as func- tions of n and q and which describe the characters for the entire family of such groups. It is too difficult to work.with both n and q arbitrary, so n is fixed and the character table which holds for all q is found. In 1955 a method for con- structing the character table of GL(n,q) for a given n and arbitrary q was developed by Green [2]. Since then the only 'abstract' character tables constructed have been those for U(2,q2), U(3,q2) by Ennola [l] and Sp(4,q), q odd, by Srinivasan [3]. In this paper ten abstract character tables are developed, representing the group families SL(n,q), PSL(n,q), SU(n,q2), PSU(n,q2) n 8 2,3 and PSL(4,q) d = 1. Of these, seven have never been published. The tables for these groups are of particular value because they often appear as important subgroups in other larger groups such as many of the Sporadic simple groups,cu' William A. Simpson are themselves simple, or are used to construct other simple groups. This paper has several purposes: 1. The character tables are made available in the sense that the range of all parameters are given explicitly so that the user can easily generate the desired table for any Specific q without searching through the paper for the definition of the various entries in the table. This degree of explicitness is not present for the three tables now in print and for this reason they have been included in this paper. 2. A standardized notation is used for all the tables which should facilitate comparisons and other inter-connecting uses. 3. The procedure discussed in section III, together with the main theorem developed in section II should enable one to more easily work out a Specific character table for any of the groups SL, PSL, SU, PSU not covered by this paper. 4. A very interesting and potentially important conjecture made by Ennola [1} relating the generalized character tables of GL(n,q) and U(n,q2) is extended, in section IV, to the special and pro- jective Special groups. It is demonstrated that a change of q a -q in the character table for SL(n,q) or PSL(n,q) will yield the table for SU(n,q2), PSU(n,q2) reSpectively for the case n = 2,3. The converse, of course, is also valid for these cases. The main theorems of the paper, proved in Chapter II, are the following: IE, If HR is a central product of groups H, K where K is abelian, then every irreducible character of HR remains irreducible when restricted to H. From this theorem, we can William A. Simpson obtain the following Theorem for SL(n,q) which is a striking improvement on Clifford's Theorem. Th_ Any irreducible character x of GL(n,q) when re- stricted to SL(n,q) is either irreducible or Splits into t conjugate, irreducible characters of SL(n,q) of multiplicity l where t‘d and d = (n,q-l). X has 133%1 associates in GL(n,q) relative to SL(n,q) if X is reducible and q-l associates if irreducible. (An identical theorem holds for SU(n,q2) with (q-l) replaced by (q+l).) This is the main working theorem of the paper as it tells us just how characters of GL will split when restricted to SL. Clifford's theorem only says that the number of components, t, will be a factor of (q-l) which is of no help. But the above theorem says that t = 2 or 3 for n = 2 or 3 respectively. In section III the technique used to formulate the char- acter tables is discussed. It is shown that each conjugacy class of SL can be indexed by a Jordan canonical form. The characters of GL(n,q) are restricted down to SL and the inner product (x,x) = Téil- 2 x(g) x(g) is calculated. gESL If (x,x) = 1 the restricted character is irreducible. Otherwise, it is reducible and splits up as indicated by the theorem just discussed. Once the character table for SL is determined we can easily generate the table for PSL, since every character of SL which is constant on Z(SL) is an irreducible character of PSL 8 SIJZ(SL), and every irreducible character of PSL can be so obtained. William A. Simpson In the sections V through XV, the character tables are developed. Many of the routine calculations are done once in detail and merely mentioned or entirely omitted in later sections. In section XVI the results obtained for PSL(4,q) d = 2 are given and some Space is devoted to discussing the problems which prevented further progress, problems which were primarily attributable to the complexity of the table for GL(4,q). REFERENCES l. V. Ennola, Characters 2£_Finite Unitary Groups, Ann. Acad. Scien. Fenn., Ser. A, 323 (1963), 120-155 2. J.A. Green, Characters g£_the Finite General Linear Groups, Trans. Am. Math. Soc., 80 (1955), 402-477 3. B. Srinivasan, Characters pf the Symplectic Group Sp(4,q) 9 odd, Trans. Am. Math. Soc., 131 (1968), 489-525 ' A DEVEmmENT OF THE CHARACTER TABLES FOR CERTAIN CLASSES OF UNITARY AND LINEAR GROUPS BY ‘_ a. {.\\..-‘ '- William At Simpson A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Mathematics 1971 ACKNOWIEDGEMENTS I feel greatly indebted to my major Professor J. Suther- land Frame for his patience, encouragement, and generous assis- tance during the period of research. His insight and suggestions at critical points were inestimable in value. I thank the members of my committee, Professors W.E. Deskins, D.W. Hall, R. Hamelink, and N.L. Hills, for their time Spent in reviewing this paper. I also wish to thank Professor Arunas Rudvalis for his interest in this problem and valuable suggestions on several occasions. Acknowledgement must be made of the kind assistance given by Professors Robert Steinberg and H. Brinkman through correSpon- dence. ii TABLE OF CONTENTS Page LIST OF TABLES v LIST OF CHARACTER TABLES V1 1. INTRODUCTION 1 1.1 History 1 1.2 Recent Interest 2 1.3 The Problem 3 1.4 General Theory 6 II. A MODIFICATION OF CLIFFORD'S THEOREM 13 2.1 Clifford's Theorem 2 13 2.2 A Theorem Applicable to SL(n,q), SU(n,q ) 15 III. TECHNIQUE 21 3.1 Determination of Conjugacy Classes for the Linear Groups 21 3.2 Determination of Conjugacy Classes for the Unitary Groups 24 3.3 Development of Character Tables 25 Iv. ENNOLA'S CONJECTURE 28 V. CHARACTER TABLE FOR SL(2,q), d = l 31 5.1 Character Table for GL(2,q) and U(2,q2) 31 5.2 Conjugacy Class Structure 33 5.3 Calculation of Characters 34 VI. CHARACTER TABLE FOR SL(2,q), d = 2 36 6.1 Conjugacy Class Structure 36 6.2 Calculation of Characters 38 VII. CHARACTER TABlE FOR PSL(2,q), d = 2 46 7.1 Conjugacy Class Structure 46 7.2 Calculation of Characters 49 iii VIII. IX. XI. XII. XIII. XIV. CHARACTER TABLE FOR SU(3,q2), d = 1 8.1 Character Table for GL(3,q), U(3,q2) 8.2 Conjugacy Class Structure 8.3 Calculation of Characters CHARACTER TABLE FOR SU(3,q2), d = 3 9.1 Conjugacy Class Structure 9.2 Calculation of Characters CHARACTER TABLE FOR PSU(3,q2), d = 3 CHARACTER TABLE FOR SL(3,q), PSL(3,q), d = 1,3 11.1 Character Table for SL(3,q), d = 1 11.2 Character Table for SL(3,q), d = 3 11.3 Character Table for PSL(3,q), d = 3 CHARACTER TABLE FOR PSL(4,q), d = 1 12.1 Conjugacy Class Structure 12.2 Calculation of Characters CONJUGACY CLASS STRUCTURE FOR U(4,q2), PSU(4,q2), d I 1 13.1 U(4,q2) 13.2 PSU(4,q2), d = 1 CHARACTER DEGREES AND FREQUENCIES FOR PSL(4,q), d B 2 Summary of Results BIBLIOGRAPHY GENERAL REFERENCES iv Page 51 51 51 57 6O 60 61 67 71 71 74 77 79 79 81 95 95 97 100 107 109 111 Table 10 11 12 13 14 15 16 l7 18 19 20 21 22 23 LIST OF TABLES Character Table for GL(Z,Q) and U(2:qz) Character Table for SL(2,q) a PSL(2,q) a PSU(2,q2) 3 SU(2,q2), d = 1, q even Conjugacy Class Structure for SL(Z.Q), d = 2 Induce-Restrict Table for GL(2,q) - SL(2,q), d = 2 Preliminary Character Table for The Character Table for SL(2,q) a SU(2,q2), d Conjugacy Character SL(2,q), d = 2 II N Class Structure of PSL(2,q), d = 2 Table for PSL(2,q) a PSU(2,q2), d Character Table for GL(3,q) Character Character Table Table Induce-Restrict for U(3,q2) 2 for SU(3,q2) a: PSU(3,q ), d II M II H Table for U(3,q2) - SU(3,q2), d = 3 Partial Character Table for SU(3,q2), d = 3 Character Character Character Table Table Table Induce-Restrict Character Character Character Conjugacy Conjugacy Character Table Table Table Class Class for SU(3,q2), d = 3 for PSU(3,q2), d = 3 for SL(3,q)sz PSL(3,q), d = 1 Table for GL(3,q) - SL(3,q), d = 3 for SL(3,q), d = 3 for PSL(3,q), d = 3 for PSL(4,q), d = 1 Structure of U(4,q2) Structure of PSU(4,q2), d a 1 Degrees and Frequencies for V PSI-(49(1): d = 2 Page 32 35 37 4O 42 45 48 50 52,53 54,55 58,59 63 65,66 69,70 72,73 74 75,76 77,78 86-93 96,97 98,99 106 Character Table and/or Conjugacy Class Table GL(2.q) GL(3,q) SL(2.9). SL(2.q). SL(3.q). SL(3.q). SL(4.q). PSL(2.q). PSL(2.q). PSL(3.Q). PSL(3.q). PSL(4.q). PSL(4.q). U(2.qz) u<3.q2) U(4.qz) SU(2.92). SU(2.qz). 800.3). SU(3.q ). SU(4,q2), Psuamz). PSU(2.q2>. PSU(3,q ). PSU(3.q ). PSU(‘hq ). momma. N NNN manque. mammo- ammo-o. HWHNH NHU’HNt—I HWHNH F‘ U3 P‘ h) P4 LIST OF CHARACTER TABLES Page 32 52,53 35 45 72,73 75,76 86-93 35 50 72,73 77,78 86-93 107 32 54,55 96,97 35 45 58,59 65,66 98,99 35 50 58,59 69,70 98,99 I. INTRODUCTION 1.1 History Towards the end of the 19th century G. Frobenius develOped the foundations for the theory of group representation in a series of papers dating from 1895 to 1911. In 1911 Burnside published his book.Theory pf Groups g£_Finite Order in which he independently rediscovered some results of Frobenius. More importantly he applied this embryonic theory and obtained some surprising results in group theory regarding the solvability of certain groups which gained more serious attention for this new concept. After this initial Success, however, the theory of group representation lay relatively dormant for nearly 25 years. It appeared that the applications of this theory as develOped up to that point had been exhausted. This disinterest was not universal. Physicists quickly grasped the importance of representation theory and soon began using it to advantage. Since representation theory reduces the abstract properties of groups to numbers, they found that this enabled them to apply group theory to any system possessing symmetry that was too complex to handle by classical analysis. Such fields as thermodynamics, crystallography, wave equations, quantum mechanics, molecular and nuclear structure are but a few of the many areas where representation has played an important role. Thus the theory of representations developed steadily in the area of applications long after the pure mathematics of its source ceased to offer much promise. Indeed, during the period of 1905 to 1955 the growth of new developments in group theory proceeded at a very slow pace. Felix Klein felt this was due to the abstract nature of group theory which isolated it from physical phenomenon which motivated SO much mathematics of that time and as a reSult only a certain type of mathematician was attracted to the field. He felt this homogeneity of the researchers was largely responsible for its slow rate of development, unmarked by imaginative leaps forward. If this was the reason for the decline in activity in group theory then the successes won by representation theory in the physical sciences may well have lessened the stigma of abstractness and helped in restoring vitality to the field of group theory. 1.2 Recent Interest In the period 1925-55 a major advance was made by R. Brauer in the study of modular representations. In 1955 when activity in group theory suddenly increased, Brauer's work provided some powerful tools with which to study group structure. Thompson and Feit in their long paper on groups of odd order used modular theory with telling effect and thus rekindled interest in representation theory. M. Suzuki has used considerable character theory in his work on the classification of simple groups by the structure of involution normalizers. At the present time significant advances in group theory are being made in the area of classifying all simple groups and here representation theory continues to play a large role since nearly all theorems on groups using characters establish the exis- tance of normal subgroups. The first step in this classification problem was to find all simple groups of order less than some large number. Initially all the simple groups which were found could be placed in one of several classes of groups, all groups of a class sharing some common properties. However, this convenient pattern was broken when, starting in 1964, people began finding some large simple groups which didn't fit into any of the known classes. These groups became known as the 'Sporadic' simple groups and they attracted considerable attention. At this date the char- acter tables for nearly all these Sporadic groups are known, to- gether with some of their larger Subgroups. 1.3 The Problem There are several reasons for wanting to know the char- acter table for a group. The most obvious reason is to be able to study better the structure of the group. Sometimes one has the character table of a group and desires to know some of its larger subgroups (see p. 66 [2]). Methods are available to find the character table of such subgroups (see p. 150 [16]). One may then be able to recognize this subgroup's character table as being the same as some known table. Thus a good stockpile of character tables would aid in identifying subgroups. The classical groups turn up often as subgroups and so their tables are particularly valuable. Wales'paper [22] furnishes an excellent example of the current manner in which characters of groups, in this case PSL(2,17), can be used to obtain surprising results. Another use for character tables lies in using them to test out various conjectures which might arise. For this purpose one would like to have tables for a wide variety of groups. Sometimes groups are Constructed by extensions of several groups. If the tables of these subgroups are known, it is often possible to obtain the character table of the constructed group. Very frequently the groups used in this construction are of the type discussed in this paper. This paper deals with the character tables for PSL(n,q), PSU(n,q2), SL(n,q) and SU(n,qZ) for n = 2,3 and some results for n = 4. The case of n = 2 was originally done by Frobenius [12], Schur [18] and later independently by Jordan [15]. This case is redone here for several reasons. The methods used by these earlier researchers are considerably different and more complex than those used in this paper. Also the resulting tables are written in such a manner that in order to use them, one would have to go back through the paper to find the definitions and range of values for the many parameters. In addition, the difference in notation from one table to another would make comparisons very difficult. The table for PSL(3,q) was done by Brinkmann [3] but the results were never published. Thus it was considered advisable to work out all the tables for n = 2,3 whether they had been done before or not. The final result is a set of 10 character tables, only 3 of which have been previously published, each of which describes an infinite class of finite groups and all unified by common nota- tion with the range of parameters explicitly given to facilitate using the tables. The procedure used to obtain these tables is quite likely the most straight forward one possible and yet little can be done on the cases for n = 4, not because the method breaks down but because the character tables become so large and complex that the details are exhausting. The character tables develOped here are of a somewhat unusual nature and a word or two should be said about them. Some- times it is possible to find a character table which can serve as the table for a whole class of groups by changing some parameters. Apart from the economy of Such an 'abstract' character table it is valuable in that it provides a strong linkage among the groups. There have been few such tables developed and so it seems that a few more additional tables would be of value. One of the most valuable classes of groups for which char- acter tables are available is the class of symmetric groups S (also An). Although no abstract character table has been developed for the entire class an equivalent formulation has been obtained, namely, a comparatively simple construction method for building the character table for any of the Specified groups. In 1896 Frobenius determined the generalized character table for the groups GL(2,q). In 1949 Steinberg [19] worked out the tables of GL(n,q) for n = 2,3,4 in a particularly straight forward fashion. These results and some later work done by Steinberg led to the develOpment by Green [14] in 1955 of a method to calculate the characters of GL(n,q) for any n. In 1963 Ennola [6] found the characters of U(n,q2) for n = 2,3 and recently Srinivasan [21] obtained the table for Sp(4,q) q odd. This sums up all the results on 'abstract' tables as of this date. 1.4 General Theory In the following short section an attempt is made to run quickly over the key definitions and theorems of character theory which are employed in this paper or necessary for its understanding. DEFINITION. A representation of a group G is a homomorphism p: G a GL(n,() where GL(n,[J is the group of all n X n, nonsingular matrices over the complex numbers; the group operation is ordinary matrix multiplication. Every group has at least one representation, the most trivial one being p1: G d l. The totality of all representations of a given group can be narrowed down by defining the following equivalence relation: DEFINITION. If p and O are n-dimensional representations of group G and there exists a nonsingular matrix A s.t. C(g) = A p(g)A-1 for all g E G then we say 9 and o are equivalent. The next step is to show that certain representations of G are the 'building blocks' from which all the non-equivalent representa- tions can be constructed. DEFINITION. A representation 9 is reducible if it is equivalent * * to a representation with matrices of the form (-6-+-;70 01(8) Lug) 0 I 02(8) 1.e. p: g a It follows that 01(g) and 02(g) are also representations of G. If a representation is not reducible then it is said to be irreducible. THEOREM 1.1 If the representation 9 (over(£) is equivalent to a reducible representation a . 01(8) 1 NS) 1.e. o: g .. 0 \02(8) then 9 is equivalent to the direct sum of the representation 01 and 02. From this theorem it follows that: THEOREM 1.2 If p is a representation for G then 9 is equi- valent to the direct sum of irreducible representations of G, for p over 0;. This key theorem says that once we know all the irreducible rep- resentations of a group G, we have all the possible representa- tions since they can be constructed by adding together irreducible representations. The next surprising theorem tells us that the number of irreducible representations of a finite group is strictly limited. THEOREM 1.3 The number of non-equivalent irreducible representa- tions of G equals the number of conjugacy classes of C. It would be very cumbersome, and as it turns out, unnecessary to write down all the irreducible representations of a given group. To avoid doing this we look for some way to designate each of these representations. This leads to the next definition. DEFINITION. If p is a representation of G, then the character x of p is the mapping from G to a: defined by: x(8) = trace 9(8)- If g1,g2 E G are elements of the same conjugacy class and p is a representation of G, then p(g1) is similar to p(g2) and since similar matrices have the same trace we can state: THEOREM 1.4 Any character x of G is constant on conjugacy classes of G. Also: THEOREM 1.5 Characters of equivalent representations are identical. THEOREM 1.6 If the representation p of G is reducible and equal to the Sum p1 + 92 +...+ pk of irreducible represen- ations, then X = x + X +...+-x . In particular, P 91 92 Pk every character of G is the sum of irreducible characters of G. The above theorems mean that we can identify a representation 9 by specifying its character X which is an r-row column vector X(81) X(82) each entry being the trace of a representative matrix xigr) 9(81) from each conjugacy class of G. We call x(l) the degree of the character. Theorem 6 says we need only Specify the irreducible characters of the group since all the others are simply integral linear combinations of these. Thus we can write down a square matrix for a group G called the character table for G. The columns of this table are the n irreducible characters of G. Customarily the identity character 1 is written as the first 1 column (or row) and the first conjugacy Class, i.e. the first row (or column) is the identity of the group. The ijth entry of the table is the value XJ(81X gi being an element of the ith con- 3 being the jth irreducible character of G. jugacy class and x It should be noted in passing that G 25H implies G and H have the same character table but the converse is go_t true. The quaternion group and the octic group are the smallest counter examples. An amazing number of relationships exist between the rows and columns of a character table. In fact, these relationships are so restrictive that in some cases it is possible to construct the character table for a given simple group knowing no more than the order of the group. There is no universal procedure to develop a character table. After one has armed himself with a thorough knowledge of the various character relationships and some Special number theory concepts, he can then proceed only on intuition and experience. No computer program has yet been devised to calculate character tables for all groups of order s 200 and yet tables for groups of this size can be done reasonably quickly by hand. Frame's paper [5)] displays in detail many of the techniques used to generate a character table of a group from that of a subgroup. Although individuals have found character tables of groups with orders as large as ‘G‘ = 47,377,612,800 the most notable success measured in terms of magnitude has been the generation of the table for Conway's group of order 221-39-54-7-11-13-23 by computer. 10 The method used some of the unique properties of the group and in no way represented a general procedure which would work for other groups. Before the more useful properties of the character table are stated, the following definition is necessary. DEFINITION. Let x1,x2 be characters of group G. Let ( ) dgf 2 x1(gi)x2(si) X1’x2 . ]N(gifT 1 where gi is an element from the ith conjugacy class and \N(gi)‘ is the order of the centralizer N(gi) of gi. In the following let Xi’xj be irreducible characters of G and 9 be a reducible character. THEOREM 1.7 (Orthogonality relations on columns) (1). (Xi’xj) = 6ij iff xi,xj are irreducible char- acters. (ii). (xi,9) = a1 where 31 is the number of times the irreducible character xi is contained in 9. (iii). (9,9) = 2 a? = sum of squares of the multiplicities of all the irreddcible characters appearing as components of 9. The above relations enable one to determine if a character is irreducible or not and in the case of a reducible character, to break it up into its irreducible components. Note the following: (i) implies that if (Y ,Y ) = 1 then Yi is irreducible. If (9,9) I 2 or 3 then (iii) implies that 9 is composed of 2 or 3 irreducible characters respectively (this case comes up often in this paper). 11 THEOREM 1.8 (Row Orthogonality) f’ . . 0 1f g,h are not In the same conjugacy class M7? . x‘(g>xi(h> = < 1 l ‘N(g)‘, if g,h are in the same Lconjugacy class Note: A special case of the above is: c o 2 E x1(1)x1(l) = 2 (degree of ith character) = ‘G‘ . i i The next major topic is the question of how characters of the group are related to characters of its subgroups. If x is an irreducible character of G we can easily get a character x‘H of H s G by restricting x to H. x‘H(g) = x(g) for g being any element in a conjugacy class of H. X‘H is not necessarily irreducible and often it isn't. If Y is an irreducible character of H s G we can obtain a character YG of G by a process called inducing, which is G H Y1 x; G th formulated by: = E where C, is the j con- \NG\ H G NH 3 j Cf:cj x jugacy class of G and C: are all conjugacy classes of H which are contained in the conjugacy class C3. Again we note that the induced character YG need not be irreducible in G. The induced and restricted characters of H and G are bound together by the following theorem: THEOREM 1.9 (Frobenius' Reciprocity Theorem) _ G (lemH — (xd’ )G . This theorem says that the number of times the restricted 12 character “It contains y as a component, equals the number of times that the induced character YG contains X as a com- ponent. Thus if H s G, and one has available the character table for C(H), it may be possible to find the table for H(G) by restricting (inducing) all characters of C(H) to H(G) and using the orthogonality properties and Frobenius' Theorem to determine how to Split up all the reducible characters into their irreducible components. The preceding has been a summary of character theory which is applicable to all groups. In the following section some theorems are discussed which apply only to Special groups. II. A MODIFICATION OF CLIFFORD'S THEOREM In this section we introduce some results known collec- tively as Clifford's Theorem, concerning characters of normal sub- groups. The particular structure of SL(n,q) is analyzed and a theorem is obtained which enables us to extend Clifford's Theorem and achieve even more information concerning the characters of GL(n,q) restricted to SL(n,q). 2.1 Clifford's Theorem Nearly every textbook which contains a section on Clifford's Theorem, including most of Clifford's paper [4], states the results in terms of irreducible G-modules and G-submodules. In this section we restate these theorems in terms of characters. Such a change in vieWpoint can be found in Feit [8] and Lomont [17]. The following are various definitions, notations and con- cepts used in the discussion of Clifford's Theorem. Let G be a group and suppose p is an irreducible representation of G and x its character. DEFINITION. If C E Aut(G) then the representation p0 defined by: po(g) dgf p(g°)‘v g E G is said to be conjugate to p. Likewise the character, x°(g) dgf x(g°) V g E G is conjugate to the character x of G. 13 14 It can be shown that if x is an irreducible character, then x0 is also. If H'd G and X is an irreducible character of G then 8 xjfl, x are said to be conjugate relative to G. Conjugate char- in acters should not be confused with characters x, X' which are complex conjugate characters defined by x'(g) = x(g) . DEFINITION. Let H < C. Two irreducible representations of G, 91 and 92 are associates if they have an irreducible component in common when restricted to H. The following theorems are all modifications of, or an outgrowth of Clifford's work [4]. THEOREM 2.1 Clifford's Theorem If H'd G and p is any irreducible representation of G, then the restriction of p to H denoted by pH is either irreducible or reducible into irreducible components of the same degree. If pg is any such irreducible com- ponent of pH, then all other components are conjugates of pfi relative to G and in addition, every such con- jugate of 9H appears as a component of 9“. This theorem says that if a character XH Splits in H then it Splits into characters of equal degree and which are identical on all classes of H which are complete classes of G. The sets of values of the characters over the other classes of H are identical but the values are permuted over these classes. Now we will take up a theorem which says something about the number of components and their multiplicities. Suppose XH is reducible. Then associated with this char- acter is a subgroup G* s.t. H < G* < C. If G/H is cyclic then the following theorem provides some interesting information. 15 THEOREM 2.2 Clifford's Theorem Let H be a normal subgroup of G Such that G/H is cyclic of order k. Then the irreducible components of XH each have multiplicity 1. The number of components of X is m, and the number of distinct associates of X relative to H is k/m, where m = [G:G*] and H S.G* S G. Since the scalar matrices of GL form a cyclic group, GL/SL, we know that if a character of CL Splits in SL it will split into m pieces of multiplicity l, where m divides [GL:SL] = (q-l). Also, since the number of associates is (q-1)/m, this means that if XSL is irreducible (i.e. m = 1) then there are (q-l) characters of GL that restrict down to the same char- acter of SL. If XSL is reducible into m components Y1,...,Ym then there are (q-1)/m characters of GL which, when restricted to SL, also split into the same components. 2.2 A Theorem Applicable to SL(n,q), SU(n,q) Let X be an irreducible character of GL(n,q). By clifford's Theorem 2.2, if X‘SL is reducible then it splits into m conjugate irreducible components of multiplicity 1 where m divides [GL:SL] = (q-l). In this section we prove that m divides d I (n,q-1), and this determines m exactly for all the groups considered in this paper. DEFINITION. Let H,K,M be groups with M S Z(H) and suppose there exists an isomorphism 9 of M into Z(K). Then if we identify M with its image 9(M), there exists a group G of the form C = HK with M = H 0 K S Z(G) 16 such that H centralizes K. Such a group G is said to be a centralgproduct of H and K w.r.t M. In diagram form: T T Z(H) Z(K) M a 901) Lemma 2.3 If H X K is the direct product of a group H with a group K and Y,9 are irreducible characters of H,K reSpectively then the character X defined by X(hk) = Y(h)-9(k) is an irreducible character of H X K. Con- versely, every irreducible character of H X K is equi- valent to such a product of characters of H and K. pf: see Littlewood [16] Lemma 2.4 If H X K is a direct product of a group H with an abelian group K and X is an irreducible character of H X K, then X‘H is an irreducible character of H. Pf: By lemma 2.3 X(g) = Y(h)-9(k) where g = hk, h E H, k 6 K and Y,9 are irreducible characters of H,K respectively. K is abelian so all its characters are of degree 1. x(h) = X(h-1)= ‘l’(h)~9(1) = ‘i’(h)-1 = ‘i’(h) for all h E H. Thus X‘H = Y which is an irreducible character of H. Lemma 2.5 If HR is a central product, then there exists a homomorphism f:H X K —+ HK pf: Gorenstein [13] Lemma 2.6 Let f be a homomorphism f:G ~ G'. If X is an irreducible character of G/Ker f then X' defined by x'(g) = X(f(g)) is an irreducible character of G. 17 Pf: Consider (X',X'): 1 (x'.x') = Z x'(g) |Gl 86G ‘Ker f‘ = -—TET——' 2 X(8)X(8) gEG/Ker f x'1 = Maj-l) = (q-l) s-_1 Z(SL) = 9‘ d )k ‘ ‘ _ k=Luqd> zmm =d (Sal)k p J Consider: Z(M(d)) SL M a: Z (SL) (1) M(d) = SL-Z(M(d)) Pf: Let A G SL, B E Z(M(d)), then det (AB) = k det Acdet B = l-pd = AB 6 M(d) = SL-Z(M(d)) s M(d). 19 16L] _ ° (q-1> But \SL-2(M(d))\ = LSL‘ngéfimni = (q 1) d = \M(d)‘ .n SL-Z(M(d)) = M(d). (ii) SL 0 Z(M(d)) = Z(SL) Pf: Let A E SL 0 Z(M(d)) then det A 8 l and A is a scalar matrix so A E Z(SL). (iii) Z(M(d)) centralizes SL Pf: Every matrix of Z(M(d)) is scalar and so commutes with every element of SL. Conditions (i), (ii), (iii) = M(d) is a central product of SL and Z(M(d)). We now put this all together to obtain the following results. THEOREM 2.11 Every irreducible character X of GL(n,q) when restricted to SL(n,q) is either irreducible or Splits into t conjugate, irreducible characters of SL of multiplicity 1 where t divides d. X has Xiill associates in GL w.r.t. SL. Pf: GL SL x Z(M(d)) d M(d) B SL-Z(M(d)) SL'_1 d\ SL Suppose X is an irreducible character of GL(n,q). Now M(d)‘< GL by Lemma 2.8 and GL/M(d) is cyclic by Lemma 2.9 so by Clifford's Theorem 2.2 X‘M(d) is either irreducible in M(d) or it Splits into t conjugate irreducible characters X1,...,Xt of multiplicity 1 where t divides [GL:M(d)] = d. Since by Lemma 2.10, M(d) is a central product of SL and the abelian group Z(M(d)), then by Theorem 2.7 every irreducible character of M(d) is also an 20 irreducible character of SL under restriction. We see from this theorem that any splitting undergone by the characters of GL takes place in restricting from GL to M(d). This means that the characters Split into t conjugates where t‘d. This is marked improvement over Clifford's Theorem which requires only that t‘(q-1). We thus have the immediate Corollary: Corollary 2.12 The characters of GL(2,q), GL(3,q), GL(4,q), d = 2 must either be irreducible when restricted to SL or split into 2,3, or 2 irreducible conjugate char- acters respectively. Every character X of GL(2,q), GL(3,q), GL(4,q), d = 2 has 3&1 associates w.r.t. SL where t = 1 if X‘SL is irreducible or t = 2,3, or 2 reSpectively if X‘SL is reducible. Theorems 2.11 and 2.12 can be restated for the characters of U(n,q2) restricted down to SU(n,qz). We need only define M(d) by: M(d) = {A e U(n,q2)\det A = pd'uq‘l) k = 1,...,9§l} where d = (n,q+l) and p is a primitive element of GF(q2). III. TECHNIQUE In this section we describe the procedures by which the character tables are obtained by restriction from GL(n,q) and U(n.q2)- 3.1 Determination of Conjugacy Classes for the Linear Groups Dickson [5] did considerable work on determining the con- jugacy classes of the general linear groups. The following is a brief description of how the class structure is found. The elements of GL(n,q) are matrices, with each of which there is associated a characteristic equation. A familiar theorem tells us that if matrix A is similar to matrix B then the characteristic equation of A equals the characteristic equa- tion of B. Thus we can find all possfllle characteristic equa- tions and correSponding to them we can associate a conjugacy class of GL having that equation. However, since the converse of the above theorem is not true, then some of the characteristic equations will correSpond to several different conjugacy classes having that same equation. We can get a full separation by writing down the possible Jordan canonical forms for each type of characteristic equation in an appropriate extension field. Each canonical form corresponds to exactly one conjugacy class and every class has such a canonical form. 21 22 In [14] Green shows that if A E GL(n,q) and A has k 1R2 ‘34 1 f2 fN where f1,...,fN are distinct irreducible polynomials over GF(q) then for every characteristic.polynomialf possible partition v1,...,vN of the exponents k1""’kN there correSponds a distinct canonical form and thus a distinct conjugacy class. This means that the number of conjugacy classes of GL(n,q) can be determined by counting the number of possible partitions. The final result is a generating function which for a given n is a polynomial in q with constant rational coefficients expres- sing the number of classes for GL(n,q). For example, if we want to find the conjugacy class structure of GL(3,q) we first write down the possible factoriza- tions of the characteristic equation of degree 3 and then for each factorization we write down the possible Jordan forms. Types of characteristic polynomials Jordan Canonical Form irreducible cubic (T q 2) T Tq uadratic factor linear factor k L q ’ p a QB O k k k 3 equal linear factors pk k (9 pk K)(e p k 9 pk 9 19 k k 2 equal linear factors pk k p p L 9 L1 9 p 3 different linear factors pk p; m) P PGGFW) 2 UEGF(Q) 3 T E GF(q ) 23 The number of conjugacy classes can now be easily counted by con- sidering the valid range of the exponents on the primitive field elements. Grouping the conjugacy classes in sets according to the possible Jordan Canonical forms is not only a convenient way to count the classes, but also serves several other purposes. First of all, the values which the exponents k,l,m,n assume can be used to determine the character value for each class, so that great economy of notation results; each whole set of conjugacy classes requires only one entry in each character. Thus the char- acter table of GL(3,q) is composed of only 8 rows, one for each canonical form. Also, it turns out, as first noted by Steinberg in [19], that there is a l-l correspondence between the sets of characters of GL(n,q) having the same degree and the sets of conjugacy classes of the same canonical form. The order of a given set of conjugacy classes is equal to the order of the corresponding set of characters. Thus for the group GL(3,q) we see that all the characters are of only 8 different degrees and it is particularly convenient to set up the character table so as to exhibit this 1-1 correSpondence between classes and characters. We remark that this l-l correspondence between sets of classes and characters doesn't hold in SL and SU if d # 1. Dickson determined the order of the centralizer of an element for each conjugacy class type of GL(n,q) n = 2,3,4, by counting the number of matrices commuting with a given canonical form. 24 Finding the conjugacy class structure of SL(n,q) is accomplished by determining what canonical forms have determinant 1. Since SL‘Q GL, then SL is composed of complete conjugacy classes of GL. Thus every conjugacy class of SL has the same order as it did in GL with the exception of the classes of GL which Split in SL. These splitting classes were determined by Dickson for n = 2,3. 3.2 Determination of Conjugacy Classes for the Unitary Groups In [23] Wall obtains results on the conjugacy classes of some classical groups, the unitary group, U(n,qz), included. His main results for U(n,qz) are: THEOREM. (1) X E GL(n,qZ) is similar to an element of *-1 U(n,qz) iff X ~ X ('~' indicates Simi- larity and '*' indicates the conjugate transpose) .. 2 2 (11) Two elements of U(n,q ) are conjugate in U(n,q ) iff they are similar in GL(n,qz) (iii) The number of conjugacy classes in U(n,qz) is m A the coefficient of tn in n -l4i-E—i . x=11-qt He also gives a formula by which the order of each conjugacy class can be calculated. The above statements say that all the elements of U(n,qz) which are similar to some A 6 GL(n,qZ) such that A ~ (A*)-1, form exactly one conjugacy class. In [7] Ennola refines the formula for calculating the order of the conjugacy classes. He also Shows that if ARE GL(n,qZ), q +l _ e — * - and A ” (A ) 1 and e is an eigenvalue of A, then 1 25 where k is odd. Thus we can use Jordan Canonical forms with marks of GF(qZR), fitting the above conditions, to represent the conjugacy classes of U(n,qz). We use the same type Jordan forms employed as class representatives for GL(n,q) except now we use elements such as pk, ck, wk: where p = 0(q-1), T = TiqB-l) and a, T1 are primitive elements of GF(qz), GF(q6) reSpectively. As in GL(n,q) we assemble the conjugacy classes of U(n,qz) into sets, each set corresponding to a Jordan Canonical form. ‘We likewise place the irreducible characters into sets according to their degrees. As before, there is a 1-1 corre- spondence between the sets of classes and sets of characters in U(n,qz). 3.3 Development of the Character Tables The character tables for GL(n,q) n = 2,3 are given in Steinberg [l9] and for U(n,qz) n = 2,3 in Ennola [6]. In the following we shall describe the procedure for finding the char- acter table for SL(n,q) and PSL(n,q), with the understanding that the same procedure works for the unitary case; just replace GL, SL, PSL by U, SU, PSU throughout the discussion. The conjugacy classes of GL(n,q) are represented by all the possible Jordan Canonical Forms. We first find the con- jugacy class representatives for SL(n,q) by selecting only those representatives of GL with determinant 1. Then the irreducible characters Xi of GL are restricted down to SL one at a time by using only the values Xi(g) where g is from a conjugacy class of SL. The inner product (Xi‘SL’xi‘SL) is calculated. 26 If (xi‘SL’xi‘SL) = 1 then Xi‘SL is irreducible and we are done with this character. Otherwise, we know it Splits into d irre- ducible components Y1,...,Yd which are all equal on those classes of SL which are complete classes of GL. The values of the Ti on the other classes can be later determined, after all the char- acters of GL have been restricted down to SL, by using the orthogonality relations Theorems 1.7 and 1.8.and Gaussian Sums. After the character table for SL is found, we can obtain the table of PSL I SL/Z(SL) by using the following well known theorems due to Frobenius. THEOREM 3.1 If Hid G and X is an irreducible character of G, then X is a character of G/H if and only if it has equal values for any two elements of G which are eQUi- valent modulo H. THEOREM 3.2 If Hld G then every character of G/H is a char- acter of C. To determine the conjugacy classes of PSL we simply fuse into a single class of PSL all the classes of SL which are equivalent under multiplications by scalar matrices from Z(SL). Picking out a set of classes of SL which are all equi- valent to one class in PSL, we then select the characters of SL which are constant over this set of classes. These are the characters of PSL. If one is working with a Specific group character table then this procedure for finding the character table of a known sub- group works rather well. The restricted characters can be tested for reducibility and the use of the character orthogonality relations 27 and Frobenius' reciprocity theorems would enable one to Split the reducible characters into irreducible components. It works well because the job of taking inner products (X,X) is not too arduous a numerical calculation. However, if one is working with a 'generalized' character table Such as GL(n,q) or U(n,qz) the entries are not numerical but are poly- nomials in q and variable roots of unity, so that the task of calculating inner products becomes extremely tedious. It would be almost hopeless if many of the restricted characters were re- ducible with many components. Theorems 2:11 and 2L1; are thus very important because they show that the Splitting is of a very Simple nature, and thereby make this process feasible. The pro- cedure can be best understood by following through one of the table derivations in detail as done in one of the first sections of the paper. IV. ENNOLA 's CONJECTURE In [20] Steinberg noted that if the conjugacy classes of GL(n,q) are partitioned into sets according to the type of Jordan canonical form to which they were similar, and the irreducible characters were also put into sets according to their degrees, then there was a l-l correspondence between the conjugacy class sets and the character sets. Also, the correSponding sets had the same order. Ennola noticed that if this same partitioning is done on U(n,qz) then the number of resulting sets is the same as for GL(n,q). In a general character table for GL(n,q) and U(n,qz) we write down only one character for each such set of irreducible characters and this character has only a single entry for each set of conjugacy classes. Thus the character tables for U(n,qz) and GL(n,q) contain the same number of 'generalized' entries. Ennola conjectured for all n, and proved for n = 2,3, that if q is everywhere replaced by -q in the character table for GL(n,q) and the resulting character changed in sign if the degree function is negative, then the character table U(n,qz) results. He used Brauer's characterization theorem [1] to prove it for these two cases but the complexities of GL(n,q) prevent one from carrying forward his method. If this conjecture is true, then Green's method for constructing characters for GL(n,q) 28 29 would serve for U(n,qz) also. One would only have to use -q in place of q in all the class functions constructed and change the sign if the degree becomes negative. Since GL(n,C) is closely related to Un (GL is a topological product of U and the Space of positive definite Hermitian matrices) we might expect some connection between their character tables, but not such a simple one. Ennola's conjecture appears so trivial that one would expect an equally simple proof. However, like many other theorems involving character relations, such a simple proof is not forth- coming. It is quite possible that the most direct proof would be through the use of Lie theory. One interesting observation which indicates some difficulty is that a permutation character of GL is 22£_transformed into a permutation character of U. Thus some reducible characters split up differently than their images under the q a -q transformation. In this paper we extend Ennola's conjecture to SL and PSL and show that it holds for all cases under consideration. A result which is similar to the above conjecture, but which is proveable is the following: Un n GL THEOREM 4.1 Ci (q) = (-1) C)\ (-q) U where an(q) = the polynomial in q which expresses the order of the Ath conjugacy class of 2 U(n,q ) GL . . . CA (q) = the polynomial In q which expresses the order of the xth conjugacy class of GL(n.q) 30 and the xth conjugacy class of GL has the same type canonical representative as the xth class of U(n,qz). This says that q a -q will transform the polynomials expressing the conjugacy class orders of GL into those for U(n,qz). Com- paring Lemma 2.4 in [14] and Definition 2 in [6] gives the above theorem. It appears that Ennola used this to obtain his tables for U(2,q2) and U(3,q2) but he never explicitly mentions it. We shall use this theorem in Chapter 15. V. THE CHARACTER TABLE FOR SL(2,q) d = 1 In this section we develop the character for SL(2,q) d = 1, q = 2K. This is the most trivial case because for d = l GL(2,q) = SL(2,q) X Z(GL) and every character of GL is irre- ducible upon restriction to SL. This one table serves for all the linear and unitary Special and projective special groups 2 since SL(2,q):2 PSL(2,q) a PSU(2,q )2: SU(2,q2) for d = l. 5.1 Character Tables for GL(2,q) and U(2,q3) All the character tables developed in the next three chapters are obtained from the table for SL(2,q) which is found in Steinberg [19]. Although the table for U(2,q), as found in [6], is not used, we furnish it for completeness. It is con- venient to combine these two character tables as in Table 1 on the following page. This also demonstrates Ennola's conjecture regarding these two character tables. 31 32 Table 1. Character Tables for GL(2,q) and U(zeqz) conjugacy canonical parameters number of centralizer class representative classes order (K) p" a C. ( p") k-1,---. (q-S) q- «r q(q-5) (q+5) m) (9" ‘) 2. 3) 1“ P k-1,"',(q-a’) q-J q(q-b’) (K. (P 91) k,’-l,"-, (q- 5) .1. z. 3 k < 1 2(q-6)(q-1-6) (ca-6) (K) H c. <‘ a“) ;;;.(---3am QM Only the classes C; require any calculation. Cék’L) O 'Ik,L = l,...,q-l is equivalent to 0 pt,....ok' = (1.1.1.1)‘V t (11) x“’ . (q,2t O et(k-k),_etk(Q-1) q ‘SL 8 Yq )= (qs0313'1)V t 0 There is no need to test for irreducibility Since SL is a quotient group of GL for d - 1 and so by Frobenius' Theorem 3.1 we know that all the characters of GL restricted to SL are irreducible. (111) xéil‘SL a Y(t) q+I )e(t+u)(q-l),e(t+u)(q-l),etk-uk +_€-tk+uk = ((q+1 .0) '3 (q+1:133k(t-U) + 6-k(t-u),0) (t) kt -kt let t = (t-u) then Yq+1 - (q+19133 + 6 ,0) so t = 1,...,%(q-2) -1 1 -1 1 k -1 (iv) Xéfl‘SL . wit: - ((q_1)nt -nt(q )(q+ ’.0.-(n (q )c -k e + nktq = ((q-1>,-1.0.-> where 35 q+1 e - 1. Thus t runs over the values l,.. .,’5q. The resulting character table appears below. I Table 2. Character Table for SL(2,q) a: PSL(2,q) a PSU(2,q2) e: SU(2,q2), d = l, q even. conjugacy canonical parameters number of centralizer class representative classes order | c C. (e O) 1 q(q+1)(q-l) I e C! (a I) 1 q (m a c, (g 3...) k-1.---.s(q-2) é(q-2) (1-1 C20 (can-u $4?» k-1,---,§q h q+1 parameters t.1,...’;(q_2)F t-l,"',§q (t)-(-t) number of characters 1 1 %(q-2) 43 c ar W ‘ (17 (0 class W. ‘ W9.— W14 C: 1 q (1+1 q-l Ca 1 0 l t -1 - K C;” 1 1 E!!! j c 0 no (It ~13 4 1 '1 ° -6 - 9 q-l p - primitive element of GF (q) s a 1 a - primitive element of GP (<12) eq+l - 1 v1. CHARACTER TABLE FOR SL(2,q) d = 2 In this section the character table for SL(2,q).a SU(2,q2) d = 2 is develOped. In some reSpects this is a difficult table to handle since there is a Sign fluctuation throughout the table which depends on whether q = : 1(mod 4). 6.1 Conjugacy Class Structure (1) Consider classes of the form (pk pk) . Since 21(Q-1) then k = (q-l) and k = %(q-1) will be the only forms with determinant 1. In SL the 1-1 correspondence be- tween sets of classes of the same canonical form and sets of characters with the same degree does not exist and so we will separate CEq-l) and cik(q-l)) into classes labeled Simply C', C" . 1 k (ii) Consider the classes of the form (9 k) . Again the 1 p only ones in SL are k = q-l, £(q-l). We know from Schur [18] that this type of class will Split into k k k (p k) s (p k) . We Shall label the form (9 k) 1 p p 9 k 1 9 k = $(Q-1):(Q-1) as Cé(k) and the form (9 k) 9 P k = %(q-1).(q-l) as 03(k). k (iii) In order for (p 91,) to have determinant l the following conditions must hold on the indices: 36 L,k = l,...,q-l 37 k+L = q-l k.t # q-l. %(q-1) C(k) = Ca.) (k) k which is equivalent to C3 (p p-k) k = l .,%(q-3). k (iv) The indices on (é qu are = (q-l),2(q-1),...,q(q-1) k 9‘ %(q+1)(q-1) C(k) = C(kq) . k(q-1) This is equivalent to [ O-k(q-l) k = l,...,%(q-l). The above calculations give us the classes as listed in Table 3. Table 3. Conjugacy Class Structure for SL(2,q) d = 2 conjugacy canonical parameters number of centralizer class represrntative classes order ’(K\ 1 C. It's-1 1 0(q --D “(to P" . (p :) kném-l) 1 (1011-!) mo 9" 0 z I 9" k'%(q—1),(q-1) 2 2C4 "m p" 0 C2 9 k=%(q-1),(q-1) 2 2e k,2.1,.u’q-1 It!) " . .. ‘ P M q 1 %(q-3) q-1 3 k ,H q-1,%(q-1) éiKl: cl.) (K) 6x6 k-(q-l) 2(q-1).. C q(q-1) %(q-1) q+1 4 1 k¢ g(q+1)(q-1) Cum, claw 9,0 = primitive elements of GF(q), GF(qZ) respectively 38 6.2 Character Table The characters of GL are again restricted down to SL as in Section 5.3, except that since some of the characters may now be reducible, the inner product (Y,Y) must be calculated on all the characters. 2t(q-l) 2t(1%l) 2kt k=q-1 2kt k=q-1 e ’6 ’6 teem-1):e k=s’ t(k-k) tk(q-l) 6 ,6 (i) xiWSL = < ) (1.1.1.1.1.1.1.1) = ‘1’1 (11) xi” = (q.q.0.0.0.o,1.-1> x xit) in CL. ('1) _ ._ Thus Xq ‘SL " (q,q,0,0,0,0,1,'1) - Yq 2 2 eqmq) 2—-‘1—q(q 4) + 2 .““.1,<-1>““.1. ek(t-u) + €-k(t-u),0) replace (t-u) by t; since t-u = t+u (mod 2) we can also replace (t+u) by t and get: ’31 = (q+1.(q+1)(-1>‘,<-1>t.1.(-1)t,1 .6“ + e'tkfi) t = l,...,q-2 t¥%(Q-1) t=1,...,3;—3 YCC) = Y('t) 39 Now we check for reducibility: 2 (q-3)/2 2(q+l) 4 2kt -2kt (Y ,Y ) = --§--+ -' + z (e +'e + 2)/(q-1) +1 +1 2 q q q/(q+1> cm -1) k=1 9:1 _ 4 2(2) _1__(‘1+1)/2(92kt+ -2kt ___2_ q+1 (q+1) q+1 R; 9 ) q+1 +1 _ 1 ‘1 2kt — 1 + q+1 g 9 .. g = ail (vq_1,wq_l) 1 + 1 2 if t c( 2 ) = 1 + o = 1 if t # (SEE) . Thus Y(t) = Y +-Y' for all t = multiple of 9:13 which q-1 9:.1 £1.21 2 -1 2 2 2 (t) occurs 35- times. The remaining %(q-1) characters xq_1 restrict down in sets of q-l to the %(q-1) irreducible char- acters, Yéfi, of SL. The restrict table for CL to SL is given below. Table 4. The Induce-restrict Table for GL(2,q) - SL(2,q) d = 2 GL(2:9} %(q—1)(q-2) eds-1) I end q—l elf-use) ans-1) yea-1) Hq-I) 114°” '11-. l l 41 51;}. 9:1 (2) <2) _ At this p01nt we know that xq+1 and xq-l split . I d I into 2 characters of SL, Y +1, Y +1 an Y _1, Y _1 2 2 2 2 reSpectively. By Theorem 2.12 Yi(g) = Y;(g) = Xéfil for i = q+1 and q-l on all conjugacy classes of SL which are complete -1 +4 (L) (L5) classes of GL. Thus and Xq_1 Split in half on all 2 xq+1 C£(k), C3(k). We now fill in the known classes of SL except portions of the character table and try to determine the missing values. The Sign alternation on some of the characters depends on whether 1%; is odd or even. It is convenient to let iii 6 = q(mod 4) and replace (-1) 2 by ;:6. Initially we will set up a separate character table for 6 =;: l and later combine them. At this point the character table is as follows: 42 Table 5. preliminary Character Table for SL(2,q) d = 2 6: 1 lasscmr k” we webs) Wei?) WM“) Wail?" iwi‘i" 1- Vi, ‘1‘” / 1 q q+1 q-l %(q+1) %(q+1) 2(q-1) 2(9-1) C; 1 q (q+1)(-1)t (q-1)(-llr 93rd) %(q+1) -%(q-1) Jam-1) I'Hr') 1 o l -1 k,+kz k\-k2 2,412 x.-9e C243”) 1 o (-l)t -(-l)t t3-k4 k3+k4 2,414 51,4, (3’ ‘1'” 1 o l -l k,-k,_ k,+kz 51,-12 R, +92 (fl—(12") 1 0 (-1)t -(-1)t k3+k4 kB-kq $1344 25-24 2““ l 1 EM+ £5“ 0 (--1)K (--l)K o o :3") 1 -1 o - K: M o o —(-1)K -(-1)'< C4 6 = *1 ‘ cl:::r WI We, WE)! W1: Winn) Li’s-(w; Win-n Win-l) C, 2km) 3.;(q+1) ‘5;(q-l) =}(q-l) C; -1r(q+1) -i(q+1) Jim-1) 2(0-1) c" ‘7'") k. +k1 k."kz Rule 9.412.. I? 1;." k3'k 4 k3+k4 Q3411 1344 C7- (same as above) C'z' ‘7’") k, -k2 k. +k2 SI. 4?, 1, +51, Z (1;) k3+k4 k3-k4 513%, 23-314 (3" (-1)" (-1)K o 0 C5,” 0 0 +1)" +1)“ 43 The missing entries in Table 5 can be filled in using the orthogonality prOperties of the character table; however, the details are rather laborious. A more elegant approach is the use of Gaussian sums. The elements in classes C2'(k), 03(k) are of order q. . . | We know that the miSSIng entries for Y +1, Y +1 are all composed 2 2 t of sums of $(q+l) q h roots of unity and the entries for Ygzlf Yézl are composed of sums of 5(Q-1) qth roots of unity. 2 2 th , We let 6 be a q root of unity . 2 61 (the exponents are quadratic residues i=k2(mod q) of q) i e (the exponents are non-residues of q). Also let x and let y 2 i¢k2(mod q) Thus x and y are sums of %(q-l) roots of unity. In partic- ular they are Gaussian sums and the following theorem applies. THEOREM 6.1 With x and y defined as above, then Xy = %(1 - 6Q) x = %(-l +q/Eq) y = %(-1 -,/Eq) where 6 = q(mod 4) = i l . For the entries of the characters Y , Y' we use x,y as 51;].- Ll 2 2 . I = _ _ _ - l = defined above if Y _1 +'Y _1 l and x, y if Y _1 +-Y _1 +1. 2 2 2 2 If we let x' = x+l and y' = y+l then x',y' are sums of %(q+l) roots of unity and we have the following corollary. Corollary 6.2 With x' and y' defined as above, then 44 x' +-y' = l X'y' = 3&(1 - 5Q) X' = 5(1 +/VI) y' = 52(1 mffi) - For the entries of the characters Y +1, Y'+1 we use x',y' if 2 2 ' = .. .. ' ' = .. Y +1 +Y +1 1 and x, y if Y +1 +Y +1 1. 2 2 2 2 Using this method the missing table entries can be easily filled in for the cases 6 = l and 5 = -l and then these two cases can be merged to complete the block as it appears in Table 6. Now that we have determined the table for SL, there is no need to use notation such as C5, C3 etc. to link SL with GL. The resulting character table is more simply written as follows. 45 Table 6. The character Table for SL(2,q) a: SU(2,q2) d = 2 conjugacy canonical parameters number of centralizer class representative classes order C1 (23 f) 1 QUIZ -l) C, 11) (l) 1 q(q"-1) (K) (-].)K K C3 ( 1 (4)) k-1.2 2 2q (K) (~1)k K C4 < P H) ) “'1’?- 2 2‘1 K) x C; ( P p") k'1."'.%(q-3) MCI-3) q-l C(GK) (‘Kflrip-mr”) kill," .,%(q-1) %(q_1) Q+1 parameters t-l, - ' ' ,Mq-B) t-l, ' ' ',%(q-1) number of characters 1 1 2 2 flq-B) §(q-l) ha . {I classc r “’1 W1, Yin“) Wilt-S) W‘éfi W21). C1 1 q %(q+ 6 ) %(q- S ) q+1 q-l C2. 1 «1 sure) -%(q-6) (-1)t (q+1) (4)"(«1-1) K C‘,“ 1 o harm) give-$21371) Hf" <-1>““' Cf,“ 1 0 was?) 9:21)“ (4153) (-1)‘ K (-1)“"" l - Ci,“ 1 1 (nine) (:1? (1-5) L)?“- T" o ‘2' 2. | tl< ) + K+ 't .. C2“ 1 -1 (51521-6) g1) (1+6) 0 11,12“. U1 2. ‘Z 6 = q(mod 4) (Ag-1 - l £J§+1 = 1 p,a = primitive elements of CF (q) , CF (qz) respectively VII. CHARACTER TABLE FOR PSL(2,q), d = 2 In this section the character table for PSL(2,q)sa PSU(2,q) d = 2 is constructed from the table of SL(2,q) given in Tab 1e 6. 7.1 Conjugacy Class Structure As in SL, the fact that q =3: lOmod 4) creates a prob- lem not only in finding the characters themselves but also in determining the conjugacy classes. The scalar matrices of SL -1 O l 0 are (0 1) and ( 0 _1) so the classes of GL will combine in pairs to give the classes of PSL except for the self-equivalent classes (these are classes of SL which map into themselves under matrix multiplication by one of the above two scalar matrices). (k) (k) We note that the number of classes of type C5 (C6 ) in Table 6 is 'even' when 6 = -l (6 = +1) and 'odd' when 6 = +1 (6 = -l), where 6 - q(mod 4). Any time the number of classes of a certain type is 'odd' this means that one such class is self-equivalent. k Thus there is a self-equivalent class of type p O_ > when 0 p Ok(Q'1) 0 6 = +1 and one of type k 1 when 6 = -1. We 0 a- (q- ) also note that the order of a self-equivalent class is a half of its order in SL Since the class collapses. We now calculate the number of classes of type Cék): 46 47 if 6 = -1 no. classes 5(3539 = 33:22.22; 4 = 9-4-5 4 if 5 = +1 no. classes = a9;- - 1) = Elia—Ll Similarly for Cgk); ' = = 11 = (51-22 + 1 if 6 +1 no. classes %( 2 ) 4 = -2+6 4 if 5 = -1 no. classes = 5(iél - 1) = S3;£%—l—l The value of the index k for which the class cék) or cék) is self equivalent is: C(k) (when 6 = +1): k = %(3-—3 + 1) = g..._1 5 2 4 k = 3:9 Cék) (when 6 = -l): k = $(3%1-+ 1) = 321 We can now determine the form of the conjugacy class of PSL which consists of the self-equivalent class of SL. Let C' denote this class. 6 9-1 P 4 -1 w 4 +1 then - 92- ~ ( 05-1) Q) 6 GF(q} Q) = 1 if 6 p is self-equivalent (q-1)(q+1) 4 if 6 = -1 the“ O _ (Q'1)(g+h ~ ¢ '1 o 4 K ¢ 2 4 o E GF(q ) o = 1 . Thus Cé is of the form (CD -1> $4 = 1 and Q) E (GF(q; if 5 = 1 ¢ GF(q ) if b = -1 The conjugacy class structure of PSL is given on the following page. The classes of SL which combine in PSL are indicated. 48 Table 7. Conjugacy Class Structure of PSL(2,q) 2 conjugacy conjugacy number of centralizer class in class in classes order SL PSL I o C' (O ') C] (I '0) (Lira) ' (3 l 1 3Q(Q1-1) C2 0". (K) f-l)K (3 (u’ (I (D) C3 I (4)" Z ' ' 1 q K 4 p (a) 3 p l 1 ‘1 (K) 10“ o /(K) 9" 0 -K C ( .K 5 O P 4 0 P k-l, 3%:(q-3) k-1,'~,i-(q-h-5) 31(q-b-5) ism-1) (K) («(14) 0 Cl“) (Incl—c) 0 C6 (0 ~?(‘Z‘l) 5 (O ':‘K(?")) k'ls s%(9']) k'1:"3%(Q'?+d;) %(Q'2+d;) 2(q+1) / ¢ 0 (1; ( C) 954) 9’)" '1 1 (q- 5) ‘CEGF(q) inf-+1 ¢1Gr'(q2) if Jul p = primitive element of GF(q) Q ll primitive element of GF(qZ) 49 7.2 Calculation of Characters Every character of SL which is constant on the classes C1 and C2 is a character of PSL and even more, every char- acter of PSL can be obtained in this way. Thus the characters X1: qu X +6’ ii: for t even, X(E ) for t even, are all 2 characters of PSL. The only difficulty is in finding the values for these characters on the class C6 since the canonical form of C6 and hence the character values, depend on 6. We will concern ourselves with the character values Xi(8) where g 6 C6 so K = 3:9 as shown in 7.1. 4 (i) (g) - +1 (if 5 = +1) xq } x (g) = -1 (if a = -1) q 9:_1 1:1 .. (t) _ kt -kt _ (t) _1_+g (4” ‘(efl' (11))tq+1(g)1+(,)1 (o — +1) x q+1(g g) — ( 2 )(ISI +w1 ) = 0 (6 = -1> = (Hm-1)t 1__ +1 (t) _ (t) (z, )t "(i— "‘)l (1.11) Xq_1(8) =0 (6 = +1.) xq_1(s) = (T )(H +li) ) = «(32“ +6.5?) (5 = -1) = (o-1)(-1) 211 11 k 1 1 1 4 (iv) xme) = (-1) 45—5— = <-1) “(J-L): (-1) 1% 2 1 El+_1_1 1 1 L2? = H) - <-1>k+1 —,9 (a = -1>= H) (j) = <-1> where [%] = greatest integer S % . We can now write down the character table for PSL(2,q). 50 2 Table 8. Character Table for PSL(2,q)sz PSU(2,q ), d = 2 conjugacy canonical parameters number of centralizer class representative classes order 1 2 C. ( 1) 1 ‘%q(q -1) l (:2 (l l) l q 1 (:3 (F’ 1) 1 q C4“, (pk or k=1.-~,+(q-11-5) inn-11-5) ice-1) dug-n- C500 ( 6””) 1<=11"'.%(q-2+ c5) Tim-2+5) %(q+1) Cc (at as" 1 (q-S) parameters t-l’ - ° - I talf . . . ' ' 3%(9‘13‘3) " 21(Q'2+S) number of , l J‘( 5) characters 1 l l L 4(q-1:’-5 ) 4 ‘27:: Mam!“ \H W1 w‘tl‘i-HT) Y-sz’) W221 wet-1 C. 1 q luv-6) %(q+5) q+1 q-l C: 1 0 so J53) Jew ~53) 1 -1 c1 1 0 1(5 4271) in +1371) 1 -1 K -Kt C4." 1 1 %(-1)K(1+6) 2(43‘ (1+5) (113.11), 0 + +1 ~Kf C?“ 1 1 1m; ‘(1-5) an? (1-6) 0 «If-(«A 3r 2! C; 1 5 (-1)[ 4] (-l)[ 4] (4; (1+6) (-1)t (5 -1) 2 . p,a = primitive elements of GF(q), GF(q ) reSpectively wim'l) = 1 93““) = 1 o4 1 ¢ 6 GF(q) if 5 +1 5 = q(mod 4) o e GF(qZ) if 5 II I ....s VIII. CHARACTER TABIE FOR SU(3,q2) d = 1 2 8.1 Character Table for GL(3,q), U(3,q ) In the next six sections all the character tables will 2 be derived from GL(3,q) and U(3,q ) as given in Steinberg [l9] and Ennola [6]. For convenience these are listed on the following pages. 8.2 Conjugacy Class Structure The determination of the range of the class parameters be- comes more complicated when we move up to 3 X 3 matrices so it is best to indicate how these calculations are made. The fact that d = 1 implies 3 * (q+1) and the condition that the determinant must be unity for all matrices in SD impose strict limitations on the values which the class parameters can take on. The only classes of type in SD are 9 (k) (k) C C2 (k) 1’ Ca those for k = q + l. The classes have determinant 1 only if (k1L) (k1t) Ca ’ C5 2k +-L a 0(mod q+1) i.e. L = -2k(mod q+1). Once k is selected, L is uniquely determined. We can discard the L parameter and the value k = q+1 (Since k a -2k = L for this value). Thus for Cik), Cék) in SU the valid range is k = l,...,q. 51 52 32538.3 Amaze .Amsveo 2318 no obsess? 2:35.:— u e.o.a Ammmuoote. shrew were .Jeru tr Sung 3+33136m. H+o+uooos o .E a? ..F QU u o e e at. C: .H n. so. Hflx XL( «7% V05.» U n U 2.3 No-3 Teams... sea see b 6 IO 1 I «.... n . .2 Va Hamw.....#tm a ..e e e AMIUV EV“ vx O» has seesaw. sews"; ewe; .... sot Cu Halo one H”:— a x XQ AEdfikV 2e 85 at. .o o a ... NAHIEU 313313 angina-«ox r so. aye/1v a} m e. AH+GV nAHIUVW ANUUV AH'UV HUG”. a. «HImQU‘ Au an XQV «WJCU Q a :13 e To 72:31 1 o a n ~ y we CcU a Aauov o H1o H1o..:.H1x . a a N N n y yQ CCU :1ov31n331n3no To To. 2:1 so ..a V .9 ya CC noose nomnmao oefipmpcononaoa nano sonaamapcoo mo hopes: mumpeemama Hmowcocmo homwsncoo 3.8.5 won some noooowoso .m ma non. affix mo macaw—weapon 5m uo>o saw u mason—neaxuw .N..A.x mo maoHumua-uuom “Jacks ecu .38 Sam a moose 3.».wi >. H w I nu Q " IV I nu 7 I u w H+e+~e 2+5: at. + N3 3+5 o a v w H 7% a 7%: a To m + + $va use.» rep 0 O O 0 row wiI wi CSU 1 .. A. O ~$q3P+>bxvawl O WHCSN Wh+v5w 3+»:ww O Stavw 3.5g Tina Tait: 115...: e O O E>+q>+wiN ~£+§v+xawN hisucvtaw MW aélirvw aiécswmw A£+~+£+w AEaSU 3.3.3 w+ m 0 w: WN w 013.73; 0 w w . U a» 3:33 + Eta? 3 cc ~33..qu 91) $233 §h+y3+¢w «a rep 241$ wV+ w+ e mi own eats... are: me mm m u no 1 a . 3.33.3 «233% qv+xpa+twm qw+x3+twm 3+5: 3+3; 2 any G 5 ..5. 5w 333cm: rainbow O 53$an 0 0 Saw)! ¥n3+1vvw1l ¥fi>+3+91WC+VNV ¥n3+v5wmu &»3+.vdmm O kuPWM m m wi a; ypn+vwwai xn>+b+twmm ¥n3+un~vw aw 2332 n— kvanmu xvmwmv “NA QL am A“ Q m.” m T: X 3.3x i ESVCX 3.3x 3.3x ask ”hex. o ...d o m noon. Amnfiv no no no no no no no 230220 3. v: v _ o: c . .3132-on a V3 V 3 VS v a H a mo sense 1? OE; «The. more: T. “a“! 1.3. «Kingsley 31.305 0.5 avovw n mosses.» Tan...HH.-p T1555 {name 31w come 3.5 a a a I a a I I I .o I The as» Hand T: H 3le a p 3. 3 H s a 3 3 . ... 9 2325.3.“ A. o. oaoov a 3.3.... 54 3n cw .Suvvb u a ..AHaZuowammu $5.5 .A vamo mo ucmamam 95353 a :30 m ..anos. a ..T u ... n .2... -3»: .... ..w. .r m 3+an Snug”... ...} w mafiob m... .50 Tn a 7.. xx... 3.... a 605 U HIU afiwoufigu 9% h 2+3:va 8-3 3+3m 3- «3.... T“ x «b 23:0 N H+dq...n HIX v.0 M e v d vx sew w . M73 2...... a... .55 5.... a E. iv Tanzii... ... . 2.3.. 3.3.. a... ... . . w 72:37qu you ..a 8.va 33:7qu 23:. at. a. .a G N H+Ua ....Hluqx v.0 no: ...... ...-.. .. M. . a X rm Axv ~3+vv hr 4...? H+w q . . .«HIM A x a xm «V NU 0. Q. v. 33333:..an 7c H.153... \ .... 0. U x .. .. .... nacho non-«Ho upwaupcmuonaou muaao #332950 no non-B: gnaw-aha Havacocao hoawannoo $28: .8 033. .3836 .3 39.9 55 n3»; mo 9533:8qu 3m MES 82.. m mcwgmuééw $.53" mo acoHumuH-FHHXH 0.393 as”. .320 as» w acquaiaaw r u U u U I u a u ' u u u u H+vauv H-or Ha+ «av Ha+av a HH UV u H H+hw> H H-~u¢ H ~+Vm m- n- T .. .. o ruin snip 3v 0 Q Q wi qu yaw Hva b O “qswwst vaw o q: uva ~41wa 3;wa O 3.»...vw 2.va $3.3 Hide: 7.... c: u 1 - - HwH w O O Eiqsf‘www T+33+xwww Asia? $.va Aiéfivm 1:433; “5+:va aEicCU Cane: ”w + m w w .. w‘ Z + x3 w . Q 3:3 as $.33 «353$ «3+ 225% o ataiw 33? a EU 3.5.3 w m + w + v o w m w W; its» T123 av m .T m U 5 .3 .v I Grim: . Wu. .3. q>+k + v «3+xca+$..w ..H a134¥~3+w~w L 835$ 333V 2 5 n kvlw I v15+vv w v;>+5+3w y—zwrtm O O XVMN CSU N yum m 21+va 333+va c 93 yadtdw my xcfiiwxl Q yvnww... 5% 3:0 . Xv w ....m ¥p3+vuw mm v1>+3+vww n: v¢3u+uvw QM ycfl‘iwl xvnnwmuu vaw .39 XVMM HEU sum um “I .5 m n». .3. . mango to x 2.13 x $.35 X. 2.8 Hit X. 3.x. Ex. Ex. ago J. 0 on o 3... «3U mm: ANIUV ~3+3 mu 3:830 &. 31.36 4730 .70 H+U .70 HoaAODEu—c A T. M 10...; ~75.an .nvogv :QWI» “..sz «twt lust» E .76 v03 0 +5 > v a v a moo.- o .3 7a. 5??» CL“ 925 >33 23d 85 5.5 A H+nua......7p 74.2....le H+0.....HI>.HH.9 3+3. 5.793 3+3. ...?Tp 20.5.59!» 3.. 28% 3 HH ea. 56 (hum) 6 and L. With k,L selected there is a unique value of m such +1 For classes C there are (q2 ) ways to select k that k +'L +'m 0(mod q+1). However there are q choices of k and L which will force m to equal k or L so we must discard these values. Thus the number of classes equals q+1 _(._2__)_:_:=l(-1) 3 6qq ' In order for classes of type Cék’L) to have determinant unity, pko-L(q-1) = pkg-L = 1 which implies L = k(mod q+1). Thus L is determined once k is chosen. Now L i 0(mod q-l) or else db = ot(q?1) = pt and we are back into the classes of a type already considered. Thus Cék) in SU is indexed by k = l,...,q2-1 k i 0(mod q-l) (k) {-qk) C = C The number of such classes is %[(q2-1) - (q+1)] = %(q-2)(q+l). In order for classes of type Cék) to have determinant l we have the following: 3 4 2 6 k -1 + -1 Tm )(q q+1)___1=,rq 2 2 3 Tk(q + q+1) (q - q+1) = Tq +1 1 which implies k E 0(mod q+1). Thus k = (q+1),2(q+l),...,(q3+l) and so the number of classes k 1‘ 0(mod q2 - q+1) 2 4 C00 = C(kq ) = C(kq ) l 2 1 equals §[(q - q+1) - 1] = 3 q(q-l). The complete class structure has thus been determined and appears in Table 11. 57 8.3 Calculation of Characters The characters of U(3,q2) are now restricted down to SU(3,q2) by letting k = (q+1) on classes Cik), Cék), Cék); L = -2k for Czk’c), Cék’t); and L = k for Cék’L). We keep in . +1 . mind that eq = 1. In several cases we can eliminate one of the character variables t,u,v by a suitable Substitution. In X(C’U)‘SU (t,u) u . (t,u) - d and x ‘S we replace (t u) by t an in XS ‘S we substitute t' - (q-1)t for u and obtain a simplification of the characters. (t.U) The calculation of XSp ISU is done below as an example. XS(f),u) ‘SU = (sp e(t-lu) (q+1) ’€(t+u) (q+1) , e:(t-l-u) (q+1) ’Seuk-Zkt euk-Ztk tk uk -quk 90:6 (n + n )30) = (Sp’1,1’86(u-2t)k’€(u-2t)k,o,nk[(q-l)t+u] + T‘k[(q-1)t--qu],0) (replace u by t' - (q-1)t) t'k t'k kt' -kqt' = (Sp31,1,S€ ,6 ’0’“ + n ,0) = Y(t ) SP None of the characters can be reducible so inner products are not 2 calculated. The character table for SU(3,q } d = 1 appears in Table 11. 58 p=o ,T=T (q-l) 3 (q -1) 1 Table 11. Character Table for SU(3,q2 )_ PSU(3,q2 ),d - 1 'conjugacy canonical parameters number of centralizer class representative classes order l 1 C. ( ' A 1 q3rs p | 3 CZ I I I 1 q 3 l ‘ 1 C3 ( i I) 1 q (K) p‘ K 2 C4 9 (52k k-l, ,q q qrs (K) p“ K [5 I P ém k-1.°°',q q q s (mflgn) PK 9 k,2,m-l,°'°,(q+l) 2 C P 1.. Rd‘ m z'q(q-1) s 6 P k+9+m=0 (mod q+1) m P“ K ) k'16”’,(qz-1) M >( ) ~K kf mod q+1 -2 +1 rs C7 6 6 ‘5 Cm) C(-1K) (modif- 0 2. q q 7* . k=(q+1), 2(q+1)," C“) 7"“ ,3(q +1) qu(q-1) p 8 m2“ kfimult. q“ -q+ +1 wal. cik1_)- C(“fi ) . . . 2 6 , o,T1 = primitive elements of GF(q ), GF(q ) reSpectively ~N.>.xu . affix «0 253338qu Sm um>o Sam m momma W Suffix mo acowuuuauoa 0393 0:» ~36 83.0. m momma? SCN .n a a a a a I a a I HuueueauwonaHuH+a>HuHuuFHnH+vw To «can H+vumavuu m m. N ..wvkafybuxkl o O O O T. T _ emu . a. Q uvxwe +83% 0 va new _ O _ C: o o 3...: A . .. I . ......zbw. .. are Passes _- N _ gin »>.3..v.v t I m 0 wxw 1175;;wa .vxw vyww ....uxw Q _ _ CCU 3.3.9. 0 .vxmwm A>Nij+vv¥wNn. vkquW+vxiwi uracw +3...w.T. w Ll _ (“Q w T _ T o _ o o _ .0 m- _ TS m. .T o m- _ «u . km am E can a us if _ .u _ Lam am Q... an. a. w L» . DnMHO i .3 25 \7 3.3.3 3.} 2% a} ‘7 ~7- hafio 2-3% 2.33-3“. 2-3%. a a H a a 2.32% we 3&5: i h . .... 2.: 3.3. y 33 , Avutluwt it :3 . a cane ours...» AH+UV AU+NUVQ.: 08v O‘fi .P V” vu. Haasbkussbvlp HMU.....HI¢ nova. 3.75.59: 6.. . . . ...—I» nahopofiaha A.u.ucouV aa manna IX. CHARACTER TABLE FOR SU(3,q2), d = 3 9.1 Conjugacy#Class Structure (i) For classes Cik),...,C§k) the only values of k for which the determinant is unity are %(q+1), %(q+l), and (q+1). k Dickson [5] shows that classes of type 2 pk in k 1 P GL(3,q) will split into three conjugacy classes in SL(3,q). Since the classes of U(3,q2) are indexed by certain classes of GL(3,q2) gk) of U(3,q2) will this means that the corresponding classes, C , Split in SU(3,q2) into three classes with 1,3,52 as the off diagonal entries, where B is any non-cube root of unity in G(q2). (ii) For the classes Cék’b),C§k’L) we require that 2k +-L a 0 (mod q+1), that is, L = ~2k(mod q+1). Also k a9 multiple of %(q+l), since this value would result in a class of type C1,C2 or C3. 1) ways Cék’L’m) we select k,L in (iii) For the classes (q: and m is uniquely determined since k +-L +-m E 0(mod q+1). However, there are (q+1) - 3 values of k and L for which the determined m value is the same as k or L. We must discard these values. We thus get the class count: fi-[cq’z’b - (cl-2)] = %l = %(q-2>(q+1>. (k) (v) As shown in Section VIII, k, for the classes of type C8 , must be a multiple of (q+1) and not a multiple of (q2 - q+1) of which there are 3 in l,...,q3+l. The total class count is %{(q2 - q+1) - 3] = %(q-2)(q+l). A complete listing of these results can be found in Table 14. 9.2 Calculation of Characters The process of restricting the characters of U(3,q2) down to SU(3,q2) is little different from that of the case d = 1. We let k = l(q+1), %(q+l),(q+1) for classes cik), (320‘), (:00. 3 3 , L = -2k for Cék’L), Cék’b); L = k for C§k’L) and keep in mind that eq+1 = 1. Making these substitutions we can obtain the restricted characters. A further substitution of t' = (t-u) and I = - (tau) (tau) t (t+2u) made in both Xp ‘SU and XQp ‘SU w(t,u) = w21+1-2-22 2 rp rp q3r82p q38 3q 2 = (18 +'2 £1 +-22)/s where 21 ' 2 \ Z e(t+u-2v)k‘2’ 22 g 2 l 2 etkfiuLivm‘Z k=1 (t,u,v) k,L,mfll [t,u,v] 160(3/3) klSU a Y“? s r s r k # multiple of 3. For such a value of t we get: 4 2 2 B 35 r 35 9 9(2-32 _ s r s r q rs p q 8 3q 1 otherwise will occur when t = g2 Thus Y(§) Splits into three irreducible components when t =.§ r r 63 and a different set of three components when t - éfi . We can now construct the restrict table. Table 12. Induce-Restrict Table for U(3,q2) - SU(3,q2) d = 3 veil E;Q(q+l)(q-1) ‘;q(q+1)(q-1) é(q-2) (q+1? sum) sow) %(q-2) (q+1)z (q+1) vs.- - ls. is. in mm )(a. x... m. «...-m. 1+! 1+, .... ,.. ‘sz s“ WSW, LPS‘I’Is Why, WSW, Wit/3 [133/5 \er ‘H—p Wm, Wye/3 “PH; =1; (q+1)(q-2) 3 3 é(q-2) (q+1) 3 SU(3,q2) U(3,q2) q+1 q+1 q+1 q(q+1) q(q+1) ski-2) (q+1):L ”Kn/XI ax‘L-I’X‘L Xian-Wt, ‘X‘KP... ’er ’XSP... ’Xsp 1" 1“ 1“ \t“ w w y. W. ”“1. we LE" ‘11? was” \KP 1 1 1 q q §(q+1)(q-2) SU(3.q2) The reducible characters er and Y 2 will Split into thirds s r on all classes of SU which are complete classes of U. leaves only the values on the classes 05(k), 63(k), cg(k) determined. The missing block is: This to be 64 Table 13. Partial Character Table for SU(3,q2) d = 3 L ‘ '- 1. L“ We Wr- 9/3 git-F73 \PSIH) LPS 173 W5 ‘73 q/St‘ya W5 "/3 W8 ‘7 3 2K (K) 1 i K 2 K 2 1 2K 2 2k J—r L is' sraf srkl sra) sraJ gru’ ark) C. 3 p 3113 3 p s “:5 ‘3 ‘3 3 3 (K) x x x LK 2K 1k - - 2 -1 - -qu -qu - a) —shJ -qu C2 (2‘; 1) (2‘1 1) £1: ”5“” ’5 =3 % 3' =3 C/M a b c a’ b ’ c ’ ” b” c” 3 . / / / 0 CW" c a b c/ a/ b c / a / b 3 // C?“ b c a b/ cl a’ b c” a” (K) r -;__ 3..-”;lw M. m_linnn. _ “-mu,_i_-.. __i 4 (K) 5 -l -l -l 0 CW") 1 or 1 or 1 or 6 -2 -2 -2 (:lK) 7 o M C8 1 l l l l 1 The easiest way to fill in this block is to look at the table for III PSL(4,q), where only the characters Y' 3, ng/3, er/B appear. rp/ It can be shown that the values for these characters on C5, 3, G; must all be rational values. Thus this 3 X 3 block of values can be quickly found using the orthogonality relations. The values of er/3 on C5’ Cg, Cg suggest that the reducible characters Y 2 Split up in a similar manner. The remaining 3 x 6 block 3 r can then be determined using orthogonality relations. The only remaining detail which needs to be mentioned is | u III rp/3’ er/3 the values which er/B’ V have on the various classes of type Cék’L). The reducible character Y:;,u,v), t = 1,...,%(q+l), u=t+%(q+l), - - k, v=t+§1q+l), equals - 2 e 9 2 q” (K) 6" k 1 ( 1) ‘ . .U q+ C5 1 9 2K kiO,mod’ §(q+l) 9’2 ‘13 pK k Q m.]_ ((1,1) Comm) 9’ (Jun, ’ é(q+1)(q-2)+1 32 5 55" k+a+m=o (mod q+1) 9“ k 1.. ‘ ' 9 sq ’1 "" ( s“ ) ky‘Omod ‘13 1.- rs 7 6' “(K’ C‘ (mod q -l) (k) 5K ,_ k-l . . qu Ca ( 6”” «1.4) kiO mod 3(q -q+1) 5(q‘2)(9*1) P 5 Cut): c(K1‘)=C(x14) 2 a = primitive element of GF(qz), p = 09.1, sq -q+l = l 9 E GF(qZ) 93 ¢ 1 N.). a H affix mo 98:32:qu HHm .35 .56 m mcmwem ca w .N..H.x mo mcoHumusHauma oHHomo on”. uo>o Sam m 959: HNTQN " 3 a u > a a u w a I Q a n I nu H m H a H u Huwuc. H H+_u H+v NV u H+d a m H a .H H, H - bu»- .. - o o .wxm.m:c vs 0 o o o .H H .H .ymu . o n . O 0 m5... +kh~¢ wi 39W . H O H ...:U mu .8 H o H>53 u s I I w . c O £>+QS+¥VWNI 0 EW «9&ka ivwvhwarww H H Hfdvcu HI 311...: m 0 H)N-3*t¥w ”NV..- yuw kvw I kw~mw +39% 0 H H CCU 7 h H>.3.S + I I ¢ 0 H>ua<1vvxw MW... row a xuuww 3W; 39%.“le U h d CCU M- m: n 6330. v.3». n41 Xu3l YH>o5+33n v.92 O yu3 O O H MU :32 3am: um: am: y... ... x3 . . x+31 xH>+5+Z3I 3.3 O .53 O O .H ring - n - n - m. = 53: 5.. .3: 3T 2 as. £3- v.:,..\.£3.. ..3 o :3 o o H 2.6 m- m- - . 6 53.. .3. 3 as...“ :3? .33275 :3 :2. :3..- o a. H :d m X 6 V 33 .H~n-.. y3u~nm mum 53f» ~35:ch PH 53% 53 no 533 ma .3 H .U C: n\.:n «Kim Q9. ..nm of mm a... c v in 330 \} 5 & .a t } «>H3ahv . .v_} Aug} va} & } y ohwfi H H H H~-ocHH.ovm AN-.VHH.OV«.W~-OVHH.UVH a a H H H mussoauago Mo hmpfizc n S+§1KH "go Hnsoocmszan coping ohm Qua??? Gnu . HU 3330 so 24.55 0.3 5.5.5 5 t H vOXQ thsoocwuHHEHm x+\{..v W VT ‘25 H i it - R 85 0.5 ...w a! {.93 HHua 33.93 m .5 >v av.» v03 0.3 6363.7» H+d.....H.->.:.a H..«0.....H..a o.... .7» 232.33 H.a.ucouV qH mHan x. CHARACTER TABLE FOR PSU(3,q2) d = 3 The center of SU(3,q2) consists of the scalar matrices (1 l 1) , C” m w) , (92 w2 w2) w3 = 1. To determine the conjugacy classes of PSU we identify classes of SU which are equivalent to each other under multiplication by the above three scalar matrices since PSU = SU/Z(SU). Thus, except for a single case, the classes of SU combine in sets of 3 to give the classes of PSU. The exception lies in the set of classes of type Cék’b’m). The number of such classes in SL is %(q+1)(q-2) +-1 which is not divisible by 3 and so we know that one class is self-equivalent +1 2 +1 (‘13, (3‘ ).q+1) 6 2 . which has the canonical representative (w w i) . This class 18 under scalar multiplication. This class is C broken out and relabeled as C9 in PSU. To find the characters of PSL we take all characters X of SU Such that x (wk wk wk) is constant for k = 1,2,3. Thus Y1, qu, and Yq3 are characters of PSU. When t is a multiple of 3 then Y(t), Y(t) Y(t), Ylt) are characters of PSU. All 9 qp ’ rp r28 three characters Y and all Y(t,u,v) Sp/3 when t + u +-v is Sp , < > a multiple of 3, are characters of PSL also. The table for PSL is thus very easily obtained once we have the characters of SL. 2 We note that the tables for PSU(3,q ) d = 1,3 agree with Frame's results in[J£fl on the number and order of the conjugacy 67 68 classes and the frequency and degrees of the characters. The number and order of the conjugacy classes for SU(3,q2) d = 1,3 given by Dickson [5], page 571, agree with our results for these groups. Several character tables for specific values of q were generated and checked against existing tables for errors. The character entry Yfit,u,v).Hu F.Hu w.H+vI qu.H+vum.HIduH H H I H\a HIHc H+e H m «I? H. o «Ian o HI H I... H ..U I I I o o o o HI HI H o o trek vxvh ..vh 3:0 o o H. o H I H. I o :36? It»? rend gnaw 0: U n>.$.vu I J Nlho .2. o o I I I w. NI H N H m E> +q$+xww NI thaw 5m :8 a: x Hidinv . I I I I I n H 0 IVan 2.:an ranmw Xena 33w 83M Rum k6 H H 3:0 + .9 on O F 6* \W U HI O .n O «...»an Ian Iw ganV xvnw 83M +5! xv? +wi+ H HIVU 1...? HI HI H o H o o H nu . 2T HI HI H o H o o H J Seem HI HI H o H o o H J O \ 7 H781... ....I 2&3 H a .7 o 7 H «u 9.3.. an» PH an no a no .5 H .U 01.. Ira ... an a e _ . 33.0 m HT .39 338+. 2‘7 to? H3} #4.? s; .310 w m HHEZNIS % HHLHXNIUV J... AHEXNISw Hal—5m 303m. 4. H H ““0211“ w. ‘ a . I I Q. ' ' m 3+0 v03 owwt?» . 23683?!» H I. t I m .... m nxm +9 .3 . i 3 9 Wm .IVHC m . 3300:1530 H w «H: m m . a . 3.: $3.» “:5 ...i. 3.3 I thnuwnu 3350 so «$31.3»th . pv HH v» HIHv vol ox» 003: one union 79% .535 H3. .7593 xHuuvwgzkHla HNIEM. . ....Hua obi H.e.ucouV HH IHHuH x1. CHARACTER TABLES FOR SL(3,q), PSL(3,q) d = 1,3 The structure of the character tables for the unitary and linear groups are so similar in form that the details of obtaining the characters of SL and PSL from GL differ very little from those described in Sections VIII-X for the corresponding unitary groups. Once the general method is established the differences between the unitary and linear cases can easily be handled. For this reason no calculations will be given for any of the tables in this section. 11.1 Character Table for SL(3,q) d = l The calculations to determine the number of conjugacy classes of each type and the range of the class parameters are of course quite different from the calculations for SU(3,q2) but they are of the same nature. Since no restricted characters can Split, the development of the character table poses no difficulty. Table 16 is the character table for SL(3,q2) d = 1. 71 72 Table 16. Character Table for SL(3,q) EEPSL(3,Q) d = l conjugacy canonical parameters number of centralizer class representative classes order I Cl ( ' I ) l qzrzsp I C2 ' | l ) 1 (131‘ I (:3 (I l ‘) 1 Q: (K) PK PK 2 C4 (51" k'L'” ,(q-2) q-2 qr s (k) f“ K C.5 " p'ZK k-1.~-,(q-2) q-2 qr (K’p'l’H) pK Q k,9,m-1,... ’(q-l) 2. C6 9 PM Man g.(q-2)(q-3) r k+9+m-O (mod q-l) II . T C(K) (p 0,-K K) k-lg.”3q '1 1 -.L .- 7 5' EtMOJHE-SKg'il §q(q 1) 1‘8 ) ‘1'" . C(K 7K1 k 1 k‘(Q"1),2(Q"1), I I 8 ... 7- m ( ;',(q‘tq)(q-1) ijq(q+1) p (: : qu'= Cfxt) p,a,T are primitive elements of GF(q), GF(qZ), GF(qa) respectively + P ' 09 1 2 _ Tq +q+l r B q - 1 s = q+1 p=q2+q+1 73 a .u HI.HI_U HIHIu: H ...u Hcvw m N 32,3033» >55 mo mcoHumuHBHwa J? .88 .0393 may H.436 as» a mwuuoHccHnw.>.xuMW. 3.».st ufiubevbesxk O O Q Q _ T _ GNU I I r 0 mix? 87: O ume pxw T O _ CCU .335“. w + + + O O F.) +NJ+XNNIN «use bowwxWV “sew yaw VXWV _ .N _ Act—..XVU H>5Ht I m 0 gym 373+:wa uxw nyIw +uxw O _ . H50 H>.3.v f I Q Ivy. 3741th m kaIwwawm uXHIW+vxwm mu m _ Hst _ _I _ O _ O 0 _ n9 AI T THIN w m. 0 w _ NU mm o I. mm mm Q Ma mw _ .0 ....1 a; an .3 a v .3 H nomHo 7: KT .3 ‘7 H113 ‘J .3 ‘7 .3} n ‘7 & hm.wo Arse...” 3.5 ...h 2-38-3». we E H H H 830220 c Hopes: 3:1: dHIm . AHMEICLZB A To cosy ou>+s+u 2-33%.... 85 35 55» 273373..» HIIU.....HI.H Hawaii» 35.3 aofifina H6. “.83 S «HHS 74 11.2 Character Table for SL(BLq) d = 3 The induce-restrict table for GL(3,q) to SL(3,q) given in Table 17 demonstrates the close similarity between the unitary and the linear cases. Table 17. Induce-Restrict Table for GL(3,q) - SL(3,q) d = 3 . GL(3.Q) q-1 I q-1 ]q -1 l(q-1)(q-2) [(q-1> gqquqm +(q-II>’Ié(q-1>‘ Lam-1) Xsp ""XsP l-X\"’Xse XI~15""XI-‘S Yr's " ' XI-‘s ’XI-‘x XV‘S Wu: 44.» IW‘P’J 9,, o,| LR?”g lH‘s/3 why, WV‘S/J [pt-{Vs KHIS/3 3 JI::(q--1I) (q-1)l 3 Ifio Sam a 9:58 355W HUG). H“ 7:. HUAHIUVU H" 3 H+V+ Una dun—Luna HIUNH Hum m N I 0 H H o ave; irwkflwh o o o o H H H EU 0 o o - - .. T o N N 23.? EP yew 29w 6:9 N .8 T o 2.3.3 o + + H N H s «fixatcowHN EM unwuzw EM 3w wi .Ewe; O H O a).3.9v ' o H H m A)». -Sflxw .N Xvw rum xvuw +wi C: n use-.3 on U n H V o EunszvwaWm yew xuuwwaetwm xvuw {yum 3:0 u m | m I u n - u n x»3 _ and.H _ 4 v33 k33+t3 :3 o 53 o o A 333 u M I ah - an: n n v1.3 . v.3 . . 53 £3333 ¥w3 o _ :3 o o .— 9:0 h m w ,n I n 533 0N3 233+ N3. 2,63%.“ vc3 532:3 53 o 53 o o a a? ml I z I I. N ¥N3H _ YJHm Afi+vmsw 3P3“ Xn>+3+vjfififo~v v.93 39.30 3N3” O U H «v30 mu m. m . . 53$." . visas." . an . runanu £2.19ng 293“...” 53%. 539 no no H C: U n\n l 3%.. £3 m L. um .: a a r . no .- N A” x} ‘7 Wait 3.3 3.3.33 3.} EN. 3.3 yr”. 3.. 33:0 a ..a m -o ..a M ..a -a L. 9.38.3.3 m n m AN. 5 v. 3 v: v. 2 yum N N H H A no ~25: 918 . .. J 3o Sansomcmuasswn “.00:an Ohm n\..«+.u } 6.0.6 3330 no nuance 93 03...»? s: \\ \ [‘3 iangmcmfigu 31.5 3 a?» 3.... 33.: A 7?; An .83 ot— xdf.» 3.... SEIEW 3 >v .7 a .85 0.3 a+«a......7a .70....3‘3.» 740......79 $0.???» Fugue—chug _ ’ A. P ”:88 3 3 any for unitary case. SL(3,q) 77 11.3 Character Table for PSLLQLQ) d = 3 The table for PSL(3,q) The table follows. is quickly obtained from the table in the same manner as described previously for the 1‘ Table 19. Character Table for PSL(3,Q) d = 3 conjugacy canonical parameters number of centralizer class representative classes order 1 Cu 1 1 1 J3 Q3 1:: 3p 1 1 C2 1 1 1 éqar / 1'. 1 2. C3 1 1 1 q n (5 ) 1 2 C3 0 1 1 q (a ) I 1 L '5 6‘ 1 1 q (K) Pr k P . K | I C‘ ( 91 ) k'lf”, §(q-h) 3 (Cl-h) 3 gr s (m 2.9“ ) C, p. k-1,'“,3 (q-h) é(q-h) éqr k,?,m-l, sq'l c”w”) p" 9 k. H nu C. H u w H n. 3 +V ..T U u n— +7 “ m «\a H-Hu HH-uv m H H H m .8 H- 0 ms t a o m n H N H to I D o «attain-xv» o o o o H H H EU .- - I H.- O H - O o O .135? ever. swam 23w 2U «3.3.9... QU N .3 .7 o HE) 4.2. ..xaw N o tawny ”aw gamuuMufiw H w H HES; - o H H a H o «ken-u +33nw + eta-UV 5% team {How ..xunu EU a n v n o «run-w Nam. ”Sen-«Va ...:w .— xtmvwcwm sue-w ....«Mn H. .50 .H H...- - H H H- c H o o H no . .. m- H H H- o H o o H ”0 3:5 m H H H- o H o o H .6 3+3}.- .H- .73.. H... c a o a H «U new nah an an no a no no a .U n\nm mu; an nr .3. a . .30 LJ .3} 3.3.3.7 .33 9‘7 3...; Hyde? .30 M AAIUV ANOUVW Aalcvadldv 04. AHIUVU ..w. @7va Aaldvmr H H H HO bong: nu nu. nu 3:50 Tn .. bi .. - - 53853.5. «.5 +91 .333: .mbfvt: ax...» 05 32:8— 3nnooa3dlna¢muvonv H55»- BE. H; V up... “25 $3 Hrs-E H H 2... a-.. 85 o 2:5 Nu Vyna vyuat 'vflva thvsos Tow . ....Hua 7a. 97.7.5.» 313m. 345 3.3% . .....7» i P H6. 28v 3 3 any affix mo 2033:8qu Sm H96 saw a «some widow .NQJH mo mcoHuMuaumm 0:93 23 .Ho>o EH5 m momma—figoc-W HlUnu-A x11. CHARACTER TABLE FOR PSL(4,q) d - 1 In his thesis, [20], Steinberg calculated the character table for GL(4,q). This table has 22 types of classes and char- acters and many of the entries are quite complicated. Due to its size the table will not be reproduced here. Since SL(4,q) a PSL(4,q) for d = 1 there are no class or character splittings and no need to calculate inner products. In Section 12.1 the parameters on the canonical representatives are determined. In Section 12.2 a single example is given to show how these parameters are used to restrict the characters of GL down to PSL. In many cases an unnecessary parameter on the character entries can be eliminated by a suitable substitution; however, such a substitution is in some cases not readily apparent. 12.1 Conjugacy Class Structure Considering only the diagonal elements of the canonical representatives of GL(4,q), we have eleven types with which to work. k k (i) det (f 9 pk k) = 1 implies 4k = 0(mod q-l). But 2,4 p do not divide (q-l) so k = (q-l) is the only possible (k) (k) 1 value. Thus for C ,...,CS k = (q-l). k (ii) det (p pk pk I.) = 1 implies (3k + L) E 0(mod q-l) 9 thus L = -3k. Since k = l,...,(q-l) - 1 there are 79 (iii) (1V) (V) (vi) (vii) (viii) 80 (q-2) classes of each of the class types C(k) C(k) C(k). k 6 ’ 7 ’ 8 det (9 pk pL L) = 1 implies (2k + 2L) E 0(mod q-l) D so L = -k. Since C(k) = C(-k) we see that k = l,...,aéz' and so there are %(q-2) classes of each type Cék), Gig), eff). k det (p pk 9L m) = 1 implies (2k +~L +-m) E 0(mod q-l). p . We thus get m = -(2k + L). There are (q-l)(o-2) ways to select k,L. But -(2k + L) = k (which gives a class of type C6) for (q-2) choices of k,L so these values must be discarded. Thus the number of classes equals 1 l 1 -2-[(q-1>(q-2> - (q-m - 5(q-2) = § det (0 9L pm a) = 1 implies (k + L +-m + n) E 0(mod q-l). 9 Since k < L < m < n k,L,m,n = l,...,q-l we get %Z(q-2)(q-3)(q-4) classes of type C(k,L,m,n). 14 det (p pk 0L qt) = 1 implies (2k + L) = 0(mod q-l). 0 Thus L = -2k and there are %-q(q-1) classes of type (R) 00 C15 ’ C16 ° k det (9 pL om HR) = 1 implies (k +-L +~m) a 0(mod q-l). 0’ Thus m = -(k +AL) and k,L = l,...,qZ-l, k # L and k +-L # multiple (q+1). This gives %~q(q-l)(q-2) classes of type ny’L). k det (G okq ck kq) = 1 implies (2k) 0(mod q-l). But 2.* (q-l) so k E 0 mod(q-l). Since k = l,...,qZ-l and k i 0 mod(q+l) there are q values which are multiples of (q-l). Since C(k) = C(RQ) we get %-q classes of type 81 (k) 00 C18 ’ C19 ' k (ix) det (0 qu CL LQ) = 1 implies (k +.L) E 0(mod q-l) so 0' 1 (k) L = -k. We get g-q(q-2)(q+l) classes of type C20 . k (x) det (9 TL TLq 2) = 1 implies (k + L) =—.-. 0(mod q-l) TLq 1 (k) so L = -k. We get :3’q(q-l)(q+1) classes of type 021 . k 2 (Xi) det (u) wkq qu 3) = 1 implies (q3 + q + q + 1)k a w kq w 0(mod q-l) which means k = 0 mod(q-l). There are %-q2(q+1) classes of type C§:)' These results are tabulated in Table 20. 12.2 Calculation of Characters There are 22 types of Jordan canonical forms for PSL(4,q) and if we look only at the diagonal of these types, as above, there are only 11 types to consider. In the same manner we can consider the characters of PSL to fall into 22 sets consisting of char- acters of the same degree. However some of these sets are related in the sense that the character parameters t,u,v,w run over the (t) X(t) ’ X(g) same range, e.g. sf’ all contain (q-2) characters. qs f q sf There is a rather close resemblance among these sets of characters and they behave in much the same way upon restriction to SL. Thus, in a sense, there are only eleven different type characters to handle. The details of restricting GL to SL are not of Sufficient interest to warrent writing out more than one example calculation. (t a“ 9") We consider the character Xsfp of GL(4,q). Each entry xsfr(gi) is understood to be the value associated with the corresponding class Ci’ 82 _ 2 i = l,...,22. As before, 8 = q+1, r = q-l, i = q +1, 2 p = q + q+1. (t,u,v> = sfp sfp €(v+u+2t)k where t,u,v = l,...,(a-l) ()R (q3 + 3q2 + 2q+1)€ t < u < v 32€( )k x(t,u,v) = X(t,v.u) (2C1+1)e( ” ( >k 6 (t+u+v)k+tL (v+2t)k-+uL e + p e 2 (U,V) 3P (2q+1)€( )k+tt+s 2 €( )R'HJL (u,v) ( )k+tL ( )k+uL e + Sum/)6 S2 2 e(v+t)k+(u+t)L + s E (u,v) (k,L) (v+t)k+(u+t)L 2tk+(v+u)L e + e e(v+u)k+2tL s[ E + e(v+u)k+2tL (U,V) ( )k+( )L ( )k+q+{. Z(u,v)e +tz(kaé)€ ] ( [(v+t)k+uL+tm] + e[(v+t)k+tL-l-um]) S Z(uav)€ + s e(v+u)k+tL+tm + Z e2t+k+vL+um (U-V) 2:(u,v)( )+ J 1+ '2: 8‘: ] 2 evk+uL+tm+tn [kaLaman] (v+u)k+tL 3 € ( )k+tL e vk+uL+tm uk+vL+tm a +6 i0 83 We now use the relations between the various class parameters (k) to restrict Xsfp to SL. For classes Cik)"°°’CS k = q-l; for Cék’L)...Cék’L) L = -3k; for Cék’L)...C{:’L) L = -k; for 0105””), cg’L’m) m = -(2k + L); for cfz’é’m’“) n = -(k + L + m); for Cig’L), Ci:’£) L = ‘Zk; and for ka,L,m) m = -(k +-L). Since xsfp(g) = O for C18,...,C22 we need not be concerned with the parameters for these classes. Using the above relations and keeping in mind that eq-l = l we obtain the following restricted character. (t U-V) _ Xsfp ‘SL - Sfp (q3 + 3q2 + 2q + l) 2 s (2q+l) l (u+v-2t)k (v-3u+2t)k s + P e P 2(u’v)€ (2C1+1)e[ 14-ng 1 Si 1+2; 1 2 (v-u)k (v+u-2t)k -(v+u-2t)k s 2%”- + s[e + e 3 +u-2t k - 503.} 3+5" ))+e() 2 e[ 1 + e[ ] +‘e-E ] (v-t)k+(u-t)L + €(v+t-2u)k+(t-u)L) + S €(v-lu-2t)k S 2 (e (U-V) 2(t -u)k+(v-u)L + 2mm“ 2:( )+eE 3+ze[ 3 (v-t)k+(u-t)L +6(v-t)k+(u-t)m +_e(v-t)L+(u-t)m 2(uav) + Z(k-L-m)€ (v+u-2t)k S e (u-vn-(v-t) (1mg 2 €(v-un-(v-t) (lem) (k,L 3m) 84 ( )k e (v-t)k+(u-t)L (u-t)k+(v-t)L e + e 0 We now observe that the substitution t' = (u-t), u' = (v-t) will eliminate one of the character parameters and simplify the char- acter o ' ' (t ’u )‘SL 3 Vii-3U) = Sfp 3 2 (q + 3q +-2q+l) 52 where u,t = l,...,(q-Z) (2q+l) l Sp e(u+t)k + p 2(u.t)‘(u-3t)k (2q+1)e(u+t)k + e(u-3t)k e(u+t)k 8 2 +823( )k 32 z c(u-t)k + sum-ml- + e-(u+t)k 2 e(ml-Dk + 8 C(u-Pcn- + €-(u+t)k (u-t)k + e(--+c)1- + e-(u-O-t)k 1 S Z(uat)€ uk-HZL + ea-zmwu, €(u+t)k ) + s (u-t)L-2tk + ~’3tm +’€uL+tm (u+t)k 6 ukfitL Z(u,t)€ 00000 The remaining characters of PSL(4,q) d = 1 are developed in a similar manner. The complete character table appears in Table 20. 86 Table 20. Character Table for PSL(4,q) d = l conj. canonical parameters class representative ggmzlzsses 2322:8117” Cl ( 11 j C (1 1 1 1 q‘r’s‘fp z 1 ) C (l 1 1) 1 q‘rzsp 3 1 i 1 1 q'rs C (- - ) c (1 ) ‘ g l 1 1 q, (x) P“ C < P" . _ ) k--3k ‘ P P "‘ k'1-'° ' .q-2 q-2 q‘r'sp on F" a C7 (1 P P" 6):) u q,rt C2" ('3’ f9" ) .13 I p q r C30 (9' P: p" ) C“). CH0 o“ g a. p k l, $(q-2) k--k §(q_2) q‘r’s" C(K) (git PI ) Cf”: COR) p" .. . .. m P k l, - ,q-2 k'-k q-2 qtrls (00 (q' p“ can: Cc-n) " 7 P") ”If" -§(q-2) 10'“ liq-2) (1‘!- (K.” (9.9" P, ) l ,W ' ,q-1 . dun-D k .. '- 9 Ca}! Sm‘glC-QKBJ MCI-2H?” qrss (ml) 9" c ( I P" ' > :3 9 -(a~o!) 9 qr1 “-9-” ”) (pip! Pm ) :1<’§7:an'11 ' ' 39‘]. men J- - - " f’a k+2+m+nlo (mod q-l) 24“; ”(Q ”(q-h) r, C‘m (9 9'0"“ , ) 1:16? A -1 w W W cm) 3023;“ kHz-l) ("‘81 P" on I <1 w > G'- n qrs ad.) 9' ’ k 9 a C". ( 9 tum) ) kgalo"':q '1 4M) ' - - 6 ‘ k+9i0 (mod Q+1) IQ“ 1‘)(q 2) r‘s Pad-T-w - primitive elements of GF(q). GF(qz), GF(qB), GF(qa) . u,(c1+1) (q2+1) 2 ‘rq «1+1 r - (1'1: reap. a - q+1, f-02+1, p=q2+q+1 87 Table 20 (cont'd.) A conj. canonical Parameters number of centralizer class representative classes order (U0 6“ «g k-l,-",qz-l Cue ( d 6" Sn) kvo (mod q+1) in qzrazf k-O (mod q-l) Co” («'63‘ .- qzs M c“ .- k-1,---,q"-1 C20 ( 6' 1‘-“ «g “‘0 (mod qgl) gq(q-2)(q+1) ra" ‘ kH- (mod q-l) P“ 'K k-l .o- 3-1 W) 'T - 9 sq 5:. ( T "Zr-«6) kfo (mod g‘+q+1) §q(q-1)(q+1) :1) can: C‘“‘ = c‘K“) m ‘4" n . k- ‘ 3 n - ) 2 a q qr 3 ’ 6 9¢0'(mod'q-1).C"’= c‘ ‘ («,I) 1 C.. (\ ) H II qmz (m...) exp: Eg'l'm'qd 2 M C” v 6“" m-1,'” ’91-]. (m) (flu-0‘ §q(q-2)(q+l) 1'33 m‘O (mod q-l) C = 97 Table 21 (cont'd.) conjugacy canonical parameters number of centralizer class representative classes order m ‘K “<1. k-l, ,q -1 («1 2 2 z m ( a qkq'“) ”‘0 (”0‘1 ‘14) Cm;- 'C ‘ §(Q*1)(q-2) q r s f (.0 «K W-K‘ u ' u 2 C... (1 1 6‘64“.) q rs ‘ - k, R-l q -1 Cw") 6 ° “1 k3” 'k.93 0 (mod q-I) fi< ) 1 . m v-q q+1 q-2 rs C” 6 «men kiddo (mod q-l) A C1092), C(‘flO 99 Table 22 (cont'd.) conjugacy canonical parameters number of centralizer class representative classes order (K) «K 41‘ k 1 ... 1' ' a sq ‘1 C'5 ( “K "“1) Rio (mod q-l) }(q-2) sirzsf «I k-O (mod q+1) Cr) (1 (“‘W‘ -K‘) II §( -2) 2r 1 o q q 2. (R) 0'“ - b1." " ,q -1 a, ( °' fl -.. ) ks‘O (mod q-l) 601-2) (qz-q-h) r28 ‘3' 3H. Cw, CH‘L Com: CM" 9‘ q: k'l "' q3+l ( j 0 D C; ( 7' qut‘ ,> k*0 (mod q‘-9+1) §q(q-1)(q+1) op .Kt q. no wk are k'(q+1)s2(q*1)s ‘ " sq‘ ‘1 2 C21 W th‘l 3 k'fiO (mod q“+l) iq (q-l) rt “1 O’Tl’w a primitive elements of GF(qz), GF(q6), GF(qa) respectively P , 0(q-1) ,T='r (QB-1) 1 3 r = Q‘l, 2 s = q+1. p = q - q+1, f = q2+l XIV. CHARACTER DEGREES AND FREQUENCIES FOR PSL(4,q) d = 2 The existing character tables for PSL(4,q) and PSU(4,q2) d = 2 are those having q = 2 or q = 3. These tables are of a reasonable size and most of the entries are integers. The non- a ifi/b integer entries, if irrational, are of the form -_:T—_' and if complex, are simple linear combinations of cube roots of unity. Thus one is easily misled into thinking that a generalized char- acter table for these groups would involve entries on the same order of complexity as was the case for n = 2,3, that is, most of the entries would be constants or polynomials in q. These hopes are soon forgotten when we consider the table for PSL(4,q) d = l which we have just developed. We see for example that the entry Y35(015) = 2 for PSL(4:2) COrreSponds to the entry retk _ (“-Ztk + n-Ztkq ) in the generalized table and realize that the simple nature of the existing tables is due entirely to the fact that the various combinations of roots of unity, which are dependent on q, are relatively elementary when q = 2 or 3. We also note that the number of classes and characters can be expressed as poly- nomials in q. At the level n = 4 these polynomials are in gen- eral of a higher degree than those for n = 2 or 3. As a result, the number of classes and characters increases at a much faster rate with an increase in q. Thus we find that PSL(4,2) has 8 char- acters, PSL(4,3) has 29, but PSL(4,4) has 82 characters and 100 101 PSL(4,7) has 407. The smallest projective special groups PSL(4,q) for which no character tables have been developed are PSL(4,7) and PSL(4,5) for d - 2,4 reSpectively. These are very large groups, having over 100 conjugacy classes, and so their character tables would be rather difficult,to work with even if they were available. If it were necessary to have the table or a partial table for one of these groups it would be much easier to develop it from. GL(4,q) with q equal to the appropriate value rather than trying to get the abstract table for arbitrary q because the calculation of inner products involving numerical values is far easier than performing the same calculation with polynomials in q and sums of (q-l)St roots of unity. Aside from these practical considerations there are some theoretical difficulties which would impede progress. For the cases d 8 2,4 both classes and characters undergo some splitting. Pre- liminary checks reveal‘this splitting is much more extensive than on the n = 2,3 levels. There is no correlation between the classes which Split and the splitting characters. We have already seen in the case of SL(3,q) d = 3 that classes of type C Split 3 while the characters which Split correspond to the classes C6 and C Thus, determining the splitting of the classes and characters 8. are two separate problems. For the cases n = 2,3 we had Dickson's results to give us the class splits, but he did not work out the case for n = 4. In the case for d = 2, if a class of GL Splits in SL we do not know if it splits into 2 or 4 classes. 102 Taking inner products to determine which characters are reducible is very difficult with the abstract characters of GL(4,q) because they are so much more complicated than the previous cases considered. Also, for the case d = 4, characters can split into 2 or 4 components and so the inner products must be calculated with more care than previously. We finally note that after we determine which characters split, we cannot immediately fill in their values on the classes which are not complete classes of GL. Since there are many splitting classes and characters this means we are left with numerous large holes in the table to be filled in by the tedious process of using the orthogonality properties of the table. In this section an attempt is made to at least determine the degrees and frequencies of the characters for PSL(4,q) d = 2. We now use a different route and proceed from GL to PCL.a:GIJZ(GL) to PSL. GL. q-l q-l PGL a- GL/Z (GL) \7‘ PSL e: SL/Z(SL) The work is based entirely on Clifford's theorem which tells us how many associates the irreducible and reducible restricted characters have. Since PSLIQ PGL and [PGL : PSL] = 2 we can apply Clifford's theorem. If a character of PGL is irreducible upon restriction to PSL then it has 2 associates. If the character is reducible then 103 it has only 1 associate and splits in half. We now assume tentatively that if 2 divides the number of characters of PGL of the same degree, then they all restrict down in sets of 2 to irreducible characters of PSL. If 2 does not divide the number of characters of the same degree we will pull off one character and assume this character Splits in PSL. We do this for q - 3,7,ll,... and hOpe that a clear pattern emerges. In Steinberg's thesis [20] the frequency table for the characters of PGL(4,q) is given. The only characters which require any Special consideration ('2) (t) are X 2 and X . We consider then separately. 3 fp r fp The number of characters of degree szfp alternates from even to odd. This means that either the number of Splitting char- acters alternates from O to l with every q, or the number increases by one for every q value. This situation did not occur in the tables for n = 2,3 so we have no precedent by which to go; however, the second alternative seems most likely. 104 number of characters q of degree 8 fp in PGL number of characters number of %Z(q-3)(q2-6q+11) in PSL reducible chars. 3 o o 01 7 3 2 l 1~_\i 1 ? n-LB ll 22 20 10 4 1 l ___1 2 l l ———1 19 172 168 84 ———l l ————____~_.______________1 l 1.=:; 1 It 1 1 1 1 From this we conclude that of the %Z(q-3)(q2 - 6q + 11) characters of degree Szfp in PGI(4,q) d = 2, 3i2- of them are reducible, when restricted to PSL and the remaining %Z(q3 - 9q2 + 23q - 15) will restrict down in sets of 2 to irreducible characters of PSL. The same situation occurs for the characters of degree 2 r fp except that the number of characters alternates from odd to even out of phase with the number of x s fp 105 number of characters q of degree r fp in PGL number of characters number of l 2 in PSL red cible chars. §