NUMBER OF COMPLETE SUBGRAPHS OF PEISERT GRAPHS

AND FINITE FIELD HYPERGEOMETRIC FUNCTIONS

arXiv:2205.03928v1 [math.CO] 8 May 2022

ANWITA BHOWMIK AND RUPAM BARMAN

Abstract. For a prime p ≡ 3 (mod 4) and a positive integer t, let q = p2t .

Let g be a primitive element of the finite field Fq . The Peisert graph P ∗ (q) is
defined as the graph with vertex set Fq where ab is an edge if and only if a−b ∈
hg 4 i ∪ ghg 4 i. We provide a formula, in terms of finite field hypergeometric
functions, for the number of complete subgraphs of order four contained in
P ∗ (q). We also give a new proof for the number of complete subgraphs of
order three contained in P ∗ (q) by evaluating certain character sums. The
computations for the number of complete subgraphs of order four are quite
tedious, so we further give an asymptotic result for the number of complete
subgraphs of any order m in Peisert graphs.

1. introduction and statements of results

The arithmetic properties of Gauss and Jacobi sums have a very long history
in number theory, with applications in Diophantine equations and the theory of
L-functions. Recently, number theorists have obtained generalizations of classical
hypergeometric functions that are assembled with these sums, and these functions
have recently led to applications in graph theory. Here we make use of these func-
tions, as developed by Greene, McCarthy, and Ono [10, 11, 14, 16] to study sub-
structures in Peisert graphs, which are relatives of the well-studied Paley graphs.
The Paley graphs are a well-known family of undirected graphs constructed from
the elements of a finite field. Named after Raymond Paley, they were introduced
as graphs independently by Sachs in 1962 and Erdős & Rényi in 1963, inspired by
the construction of Hadamard matrices in Paley’s paper [17]. Let q ≡ 1 (mod 4)
be a prime power. Then the Paley graph of order q is the graph with vertex set as
the finite field Fq and edges defined as, ab is an edge if a − b is a non-zero square
in Fq .
It is natural to study the extent to which a graph exhibits symmetry. A graph
is called symmetric if, given any two edges xy and x1 y1 , there exists a graph
automorphism sending x to x1 and y to y1 . Another kind of symmetry occurs if
a graph is isomorphic to its complement, in which case the graph is called self-
complementary. While Sachs studied the self-complementarity properties of the
Paley graphs, Erdős & Rényi were interested in their symmetries. It turns out that
the Paley graphs are both self-complementary and symmetric.
It is a natural question to ask for the classification of all self-complementary
and symmetric (SCS) graphs. In this direction, Chao’s classification in [5] sheds

Date: 9th May 2022.

2020 Mathematics Subject Classification. 05C25; 05C30; 11T24; 11T30.
Key words and phrases. Peisert graphs; clique; finite fields; character sums; hypergeometric
functions over finite fields.
1
2 ANWITA BHOWMIK AND RUPAM BARMAN

light on the fact that the only such possible graphs of prime order are the Paley
graphs. Zhang in [21], gave an algebraic characterization of SCS graphs using
the classification of finite simple groups, although it did not follow whether one
could find such graphs other than the Paley graphs. In 2001, Peisert gave a full
description of SCS graphs as well as their automorphism groups in [18]. He derived
that there is another infinite family of SCS graphs apart from the Paley graphs,
and, in addition, one more graph not belonging to any of the two former families.
He constructed the P ∗ -graphs (which are now known as Peisert graphs) as follows.
For a prime p ≡ 3 (mod 4) and a positive integer t, let q = p2t . Let g be a primitive
element of the finite field Fq , that is, F∗q = Fq \ {0} = hgi. Then the Peisert graph
P ∗ (q) is defined as the graph with vertex set Fq where ab is an edge if and only if
a − b ∈ hg 4 i ∪ ghg 4 i. It is shown in [18] that the definition is independent of the
choice of g. It turns out that an edge is well defined, since q ≡ 1 (mod 8) implies
that −1 ∈ hg 4 i.
We know that a complete subgraph, or a clique, in an undirected graph is a set
of vertices such that every two distinct vertices in the set are adjacent. The number
of vertices in the clique is called the order of the clique. Let G(n) denote a graph
on n vertices and let G(n) be its complement. Letkm (G) denote the number  of
(n) (n)
cliques of order m in a graph G. Let Tm (n) = min km (G ) + km (G ) where
the minimum is taken over all graphs on n vertices. Erdős [7], Goodman [9] and
Thomason [19] studied Tm (n) for different values of m and n. Here we note that
the study of Tm (n) can be linked to Ramsey theory. This is because, the diagonal
Ramsey number R(m, m) is the smallest positive integer n such that Tm (n) is
positive. Also, for the function km (G(n) ) + km (G(n) ) on graphs with n = p vertices,
p being a prime, Paley graphs are minimal in certain ways; for example, in order
to show that R(4, 4) is atleast 18, the Paley graph with 17 vertices acts as the only
graph (upto isomorphism) such that km (G(17) ) + km (G(17) ) = 0. What followed
was a study on km (G), G being a Paley graph. Evans et al. [8] and Atansov et
al. [2] gave formulae for k4 (G), where G is a Paley graph with number of vertices
a prime and a prime-power, respectively. One step ahead led to generalizations of
Paley graphs by Lim and Praeger [13], and computing the number of cliques of
orders 3 and 4 in those graphs by Dawsey and McCarthy [6]. Very recently, we
[4] have defined Paley-type graphs of order n as follows. For a positive integer n,
the Paley-type graph Gn has the finite commutative ring Zn as its vertex set and
edges defined as, ab is an edge if and only if a − b ≡ x2 (mod n) for some unit x of
Zn . For primes p ≡ 1 (mod 4) and any positive integer α, we have also found the
number of cliques of order 3 and 4 in the Paley-type graphs Gpα .
The Peisert graphs lie in the class of SCS graphs alongwith Paley graphs, so it
would serve as a good analogy to study the number of cliques in the former class
too. There is no known formula for the number of cliques of order 4 in Peisert
graph P ∗ (q). The main purpose of this paper is to provide a general formula for
k4 (P ∗ (q)). In [1], Alexander found the number of cliques of order 3 using the
properties that the Peisert graph are edge-transitive and that any pair of vertices
connected by an edge have the same number of common neighbors (a graph being
edge-transitive means that, given any two edges in the graph, there exists a graph
automorphism sending one edge to the other). In this article, we follow a character-
sum approach to compute the number of cliques of orders 3 and 4 in Peisert graphs.
 NUMBER OF COMPLETE SUBGRAPHS OF PEISERT GRAPHS 3

In the following theorem, we give a new proof for the number of cliques of orders 3
in Peisert graphs by evaluating certain character sums.
Theorem 1.1. Let q = p2t , where p ≡ 3 (mod 4) is a prime and t is a positive
integer. Then, the number of cliques of order 3 in the Peisert graph P ∗ (q) is given
by
q(q − 1)(q − 5)
k3 (P ∗ (q)) = .
48
Next, we find the number of cliques of order 4 in Peisert graphs. In this case, the
character sums are difficult to evaluate. We use finite field hypergeometric functions
to evaluate some of the character sums. Before we state our result on k4 (P ∗ (q)),
we recall Greene’s finite field hypergeometric functions from [10, 11]. Let p be an
odd prime, and let Fq denote the finite field with q elements, where q = pr , r ≥ 1.
Let Fc
×
q be the group of all multiplicative characters on F× . We extend the domain
q

of each χ ∈ Fc ×
q to Fq by setting χ(0) = 0 including the trivial character ε. For
A


multiplicative characters A and B on Fq , the binomial coefficient B is defined by
 
A B(−1)
:= J(A, B),
B q
X
where J(A, B) = A(x)B(1 − x) denotes the Jacobi sum and B is the character
x∈Fq

inverse of B. For a positive integer n, and A0 , . . . , An , B1 , . . . , Bn ∈ Fc

×
q , Greene
[10, 11] defined the n+1 Fn - finite field hypergeometric function over Fq by
      
A0 , A1 , . . . , An q X A0 χ A1 χ An χ
n+1 Fn | x := ··· χ(x).
B1 , . . . , Bn q−1 χ B1 χ Bn χ
c
χ∈F×
q

For n = 2, we recall the following result from [10, Corollary 3.14]:

  X
A, B, C
3 F2 |λ = AE(x)CE(1 − x)B(y)BD(1 − y)A(x − λy).
D, E
x,y∈Fq

Some of the biggest motivations for studying finite field hypergeometric functions
have been their connections with Fourier coefficients and eigenvalues of modular
forms and with counting points on certain kinds of algebraic varieties. For example,
Ono [15] gave formulae for the number of Fp -points on elliptic curves in terms of
special values of Greene’s finite field hypergeometric functions. In [16], Ono wrote
a beautiful chapter on finite field hypergeometric functions and mentioned several
open problems on hypergeometric functions and their relations to modular forms
and algebraic varieties. In recent times, many authors have studied and found
solutions to some of the problems posed by Ono.
Finite field hypergeometric functions are useful in the study of Paley graphs, see
for example [6, 20]. In the following theorem, we express the number of cliques of
order 4 in Peisert graphs in terms of finite field hypergeometric functions.
Theorem 1.2. Let p be a prime such that p ≡ 3 (mod 4). For a positive integer t,
let q = p2t . Let q = u2 + 2v 2 for integers u and v such that u ≡ 3 (mod 4) and p ∤ u
when p ≡ 3 (mod 8). If χ4 is a character of order 4, then the number of cliques of
4 ANWITA BHOWMIK AND RUPAM BARMAN

order 4 in the Peisert graph P ∗ (q) is given by

  
∗ q(q − 1) 2 t 2 χ4 , χ4 , χ34
k4 (P (q)) = 2(q − 20q + 81) + 2u(−p) + 3q · 3 F2 |1 .
3072 ε, ε
Using Sage, we numerically verify Theorem 1.2 for certain values of q. We list
some of the values in Table 1. We denote by 3 F2 (·) the hypergeometric function
appearing in Theorem 1.2.

p k4 (P ∗ (q))
q u q 2 · 3 F2 (·) k4 (P ∗ (q)) 3 F2 (·)
(by Sage) (by Sage) (by Theorem 1.2)
3 9 0 −1 10 0 0.1234 . . .
7 49 2156 7 −30 2156 −0.0123 . . .
3 81 21060 7 −62 21060 −0.0094 . . .
11 121 116160 7 42 116160 0.0028 . . .
19 361 10515930 −17 522 10515930 0.0040 . . .
23 529 49135636 23 930 49135636 0.0033 . . .
Table 1. Numerical data for Theorem 1.2

We note that the number of 3-order cliques in the Peisert graph of order q
equals the number of 3-order cliques in the Paley graph of the same order. The
computations for the number of cliques of order 4 are quite tedious, so we further
give an asymptotic result in the following theorem, for the number of cliques of
order m in Peisert graphs, m ≥ 1 being an integer.
Theorem 1.3. Let p be a prime such that p ≡ 3 (mod 4). For a positive integer
t, let q = p2t . For m ≥ 1, let km (P ∗ (q)) denote the number of cliques of order m
in the Peisert graph P ∗ (q). Then
km (P ∗ (q)) 1
lim = m .
2( 2 ) m!
q→∞ q m

2. preliminaries and some lemmas

We begin by fixing some notations. For a prime p ≡ 3 (mod 4) and positive
integer t, let q = p2t . Let g be a primitive element of the finite field Fq , that is,
F∗q = Fq \ {0} = hgi. Now, we fix a multiplicative character χ4 on Fq of order 4
(which exists since q ≡ 1 (mod 4)). Let ϕ be the unique quadratic character on Fq .
Then, we have χ24 = ϕ. Let H = hg 4 i ∪ ghg 4 i. Since H is the union of two cosets
of hg 4 i in hgi, we see that |H| = 2 × q−1 q−1
4 = 2 . We recall that a vertex-transitive
graph is a graph in which, given any two vertices in the graph, there exists some
graph automorphism sending one of the vertices to the other. Peisert graphs being
symmetric, are vertex-transitive. Also, the subgraphs induced by hg 4 i and ghg 4 i
are both vertex transitive: if s, t are two elements of hg 4 i (or ghg 4 i) then the map
on the vertex set of hg 4 i (or ghg 4 i) given by x 7−→ st x is an isomorphism sending s
to t. The subgraph of P ∗ (q) induced by H is denoted by hHi.
Throughout the article, we fix h = 1 − χ4 (g). For x ∈ F∗q , we have the following:

2 + hχ4 (x) + hχ4 (x) 1, if χ4 (x) ∈ {1, χ4 (g)};
= (2.1)
4 0, otherwise.
We note here that for x 6= 0, x ∈ H if and only if χ4 (x) = 1 or χ4 (x) = χ4 (g).
 NUMBER OF COMPLETE SUBGRAPHS OF PEISERT GRAPHS 5

We have the following lemma which will be used in proving the main results.
Lemma 2.1. Let q = p2t where p ≡ 3 (mod 4) is a prime and t is a positive
integer. Let χ4 be a multiplicative character of order 4 on Fq , and let ϕ be the
unique quadratic character. Then, we have J(χ4 , χ4 ) = J(χ4 , ϕ) = −(−p)t .
Proof. By [12, Proposition 1], we have J(χ4 , χ4 ) = −(−p)t . We also note that by
Theorem 2.1.4 and Theorem 3.2.1 of [3], where the results remain the same if we
replace a prime by a prime power, we see that J(χ4 , ϕ) = χ4 (4)J(χ4 , χ4 ) = a4 +ib4 ,
where a24 + b24 = q and a4 ≡ −( q+1
2 ) (mod 4). Hence, a4 ≡ 1 (mod 4) and a4 =
−(−p)t , b4 = 0. Thus, we obtain J(χ4 , ϕ) = J(χ4 , χ4 ) = −(−p)t . 
Next, we evaluate certain character sums in the following lemmas.
Lemma 2.2. Let q ≡ 1 (mod 4) be a prime power and let χ4 be a character on Fq
of order 4 such that χ4 (−1) = 1, and let ϕ be the unique quadratic character. Let
a ∈ Fq be such that a 6= 0, 1. Then,
X
χ4 ((y − 1)(y − a)) = ϕ(a − 1)J(χ4 , χ4 ).
y∈Fq

Proof. We have
X X
χ4 ((y − 1)(y − a)) = χ4 (y ′ (y ′ + 1 − a))
y∈Fq y ′ ∈Fq
X X
= χ4 ((1 − a)y ′′ )χ4 ((1 − a)(y ′′ + 1)) = ϕ(1 − a) χ4 (y ′′ (y ′′ + 1))
y ′′ ∈Fq y ′′ ∈Fq
X
= ϕ(1 − a) χ4 (−y ′′ (−y ′′ + 1))
y ′′ ∈F q

= ϕ(1 − a)J(χ4 , χ4 ),
where we used the substitutions y − 1 = y ′ , y ′′ = y ′ (1 − a)−1 , and replaced y ′′ by
−y ′′ . 
Lemma 2.3. Let q ≡ 1 (mod 4) be a prime power and let χ4 be a character on Fq
of order 4 such that χ4 (−1) = 1. Let a ∈ Fq be such that a 6= 0, 1. Then,
X
χ4 (y)χ4 (a − y) = −1.
y∈Fq

Proof. We have
X X
χ4 (y)χ4 (a − y) = χ4 (ay ′ )χ4 (a − ay ′ )
y∈Fq y ′ ∈F q
X
′ ′
= χ4 (y )χ4 (1 − y )
y ′ ∈Fq
X

= χ4 y ′ (1 − y ′ )−1
y ′ ∈Fq
X
= χ4 (y ′′ ) = −1,
y ′′ ∈F q ,y ′′ 6=−1

where we used the substitutions y ′ = ya−1 and y ′′ = y ′ (1 − y ′ )−1 , respectively. 

6 ANWITA BHOWMIK AND RUPAM BARMAN

Lemma 2.4. Let q = p2t , where p ≡ 3 (mod 4) is a prime and t is a positive

integer. Let χ4 be a character on Fq of order 4 and let ϕ be the unique quadratic
character. Let J(χ4 , χ4 ) = J(χ4 , ϕ) = ρ, where ρ = −(−p)t . Then,
X
χ4 (x)χ4 (y)χ4 (1 − y)χ4 (x − y) = −2ρ (2.2)
x,y∈Fq ,x6=1

and
X
χ4 (x)χ4 (y)χ4 (1 − y)χ4 (x − y) = 1 − ρ. (2.3)
x,y∈Fq ,x6=1

Proof. By Lemma 2.3, we have

X X X
χ4 (y)χ4 (1 − y) χ4 (x)χ4 (x − y) = χ4 (y)χ4 (1 − y) [−1 − χ4 (y − 1)]
y6=0,1 x6=0,1,y y6=0,1
X
= −ρ − χ4 (y)ϕ(1 − y) = −2ρ,
y

which proves (2.2). Next, using the substitution x′ = xy −1 , we have

X X
χ4 (x)χ4 (x − y) = χ4 (x′ y)χ4 (x′ y − y)
x x′
= ϕ(y)ρ. (2.4)
So, using (2.4), we find that
X X
χ4 (y)χ4 (1 − y) χ4 (x)χ4 (x − y)
y6=0,1 x6=0,1,y
X
= χ4 (y)χ4 (1 − y) [ϕ(y)ρ − χ4 (y − 1)]
y6=0,1
X X
=ρ χ4 (y)χ4 (1 − y) − χ4 (y)
y y
= −ρ + 1.
This completes the proof of the lemma. 

We need to evaluate several analogous character sums as in Lemma 2.4. To this

end, we have the following two lemmas whose proofs merely involve Lemmas 2.2
and 2.3 (as in Lemma 2.4).
Lemma 2.5. Let q = p2t , where p ≡ 3 (mod 4) is a prime and t is a positive
integer. Let χ4 be a character on Fq of order 4 and let ϕ be the unique quadratic
character. Let J(χ4 , χ4 ) = J(χ4 , ϕ) = ρ, where ρ = −(−p)t . Then, we have
X
χi41 (y)χi42 (1 − y)χi43 (x − y)
x,y∈Fq ,x6=1

 −2ρ, if (i1 , i2 , i3 ) ∈ {(1, 1, 1), (−1, −1, −1)};
= 2, if (i1 , i2 , i3 ) ∈ {(1, 1, −1), (−1, −1, 1)};

1 − ρ, if (i1 , i2 , i3 ) ∈ {(1, −1, 1), (1, −1, −1), (−1, 1, 1), (−1, 1, −1)}.
Lemma 2.6. Let q = p2t , where p ≡ 3 (mod 4) is a prime and t is a positive
integer. Let χ4 be a character on Fq of order 4 and let ϕ be the unique quadratic
 NUMBER OF COMPLETE SUBGRAPHS OF PEISERT GRAPHS 7

character. Let J(χ4 , χ4 ) = J(χ4 , ϕ) = ρ, where ρ = −(−p)t . Then, for i1 , i2 , i3 ∈

{±1}, we have the following tabulation of the values of the expression given below:
X
Ax · χi41 (y)χi42 (1 − y)χi43 (x − y). (2.5)
x,y∈Fq ,x6=1

For w ∈ {1, 2, . . . , 8} and z ∈ {1, 2, . . . , 7}, the (w, z)-th entry in the table corre-
sponds to (2.5), where Ax is either χ4 (x), χ4 (x), χ4 (1 − x) or χ4 (1 − x) and the
tuple (i1 , i2 , i3 ) depends on w.
Ax
i1 i2 i3 χ4 (x) χ4 (x) χ4 (1 − x) χ4 (1 − x)
1 1 1 −2ρ −2ρ −2ρ −2ρ
1 1 −1 1−ρ 1−ρ 1−ρ 1−ρ
1 −1 1 ρ2 + 1 2 ρ2 − ρ 1−ρ
1 −1 −1 1−ρ ρ2 − ρ 2 ρ2 + 1
−1 1 1 ρ2 − ρ 1−ρ ρ2 + 1 2
−1 1 −1 2 ρ2 + 1 1−ρ ρ2 − ρ
−1 −1 1 1−ρ 1−ρ 1−ρ 1−ρ
−1 −1 −1 −2ρ −2ρ −2ρ −2ρ
For example, the (3, 6)-th position contains the value ρ2 −ρ. Here w = 3 corresponds
to i1 = 1, i2 = −1, i3 = 1; z = 6 corresponds to the column Ax = χ4 (1 − x). So,
X
χ4 (1 − x)χ4 (y)χ4 (1 − y)χ4 (x − y) = ρ2 − ρ.
x,y∈Fq ,x6=1

Proof. The calculations follow along the lines of Lemma 2.2 and Lemma 2.3. For
example, in Lemma 2.4, one can take χ4 (x), χ4 (x− 1) or χ4 (x− 1) in place of χ4 (x)
in (2.2) and (2.3) (which we denote by Ax ), and easily evaluate the corresponding
character sum. 
Lemma 2.7. Let q = p2t , where p ≡ 3 (mod 4) is a prime and t is a positive
integer. Let χ4 be a character of order 4. Let ϕ and ε be the quadratic and the
trivial characters, respectively. Let q = u2 + 2v 2 for integers u and v such that
u ≡ 3 (mod 4) and p ∤ u when p ≡ 3 (mod 8). Then,
   
χ4 , χ4 , χ4 χ4 , χ4 , χ4
3 F2 | 1 = 3 F2 |1
ε, ε ε, ε
   
χ4 , χ4 , χ4 χ4 , χ4 , χ4
= 3 F2 | 1 = 3 F2 |1
ϕ, ε ϕ, ε
1
= 2 [−2u(−p)t ].
q
Proof. Let χ8 be a character of order 8 such that χ28 = χ4 . Now, Proposition 1 in
[12] tells us that J(χ4 , χ4 ) = −(−p)t and hence it is real. Again, by Theorem 3.3.3
and the paragraph preceeding Theorem 3.3.1 in [3], J(χ8 , χ28 ) = χ8 (−4)J(χ4 , χ4 ),
where χ8 (4) = ±1 and thus, is also real. By [10, Theorem 4.37], we have
      5  5 
χ4 , χ4 , χ4 χ8 χ8 χ8 χ8
3 F2 |1 = +
ε, ε χ28 χ38 χ28 χ8
χ8 (−1)
= [J(χ8 , χ68 )J(χ8 , χ58 ) + J(χ58 , χ68 )J(χ58 , χ8 )].
q2
(2.6)
8 ANWITA BHOWMIK AND RUPAM BARMAN

Using Theorems 2.1.5 and 2.1.6 in [3] we obtain

J(χ8 , χ68 ) = χ8 (−1)J(χ8 , χ8 ),
J(χ8 , χ58 ) = χ8 (−1)J(χ8 , χ28 ),
J(χ58 , χ68 ) = χ8 (−1)J(χ8 , χ8 ).
Substituting these values in (2.6) and using [6, Lemma 3.6 (2)], we find that
 
χ4 , χ4 , χ4 χ8 (−1)
F
3 2 | 1 = [J(χ8 , χ8 )J(χ8 , χ28 ) + J(χ8 , χ8 )J(χ8 , χ28 )]
ε, ε q2
1
= 2 J(χ8 , χ28 ) × 2Re(J(χ8 , χ8 )) × χ8 (−1)
q
1
= 2 [−2u(−p)t ]. (2.7)
q
   
χ4 , χ4 , χ4 χ4 , χ4 , χ4
Since 3 F2 | 1 is the conjugate of 3 F2 | 1 , so
ε, ε ε, ε
both are equal as the value given in (2.7) is a real number. Using Lemma 4.37 in
[10] again, we have
      3  3 
χ4 , χ4 , χ4 χ8 χ8 χ8 χ8
3 F2 |1 = +
ϕ, ε 2
χ8 χ8 χ82 χ8 3
χ8 (−1)
= [J(χ8 , χ8 2 )J(χ8 , χ8 ) + J(χ38 , χ8 2 )J(χ38 , χ38 )].
q2
(2.8)

Recalling Theorem 2.1.6 in [3] gives J(χ8 , χ8 ) = J(χ38 , χ38 ). Also, Theorem 2.1.5 in
[3] gives J(χ38 , χ8 2 ) = J(χ58 , χ28 ) = J(χ8 , χ28 ) = J(χ8 , χ28 ). Hence, (2.8) yields
 
χ4 , χ4 , χ4 1
3 F2 |1 = 2 J(χ8 , χ28 ) × 2Re(J(χ8 , χ8 )) × χ8 (−1)
ϕ, ε q
1
= 2 [−2u(−p)t ],
q
which is thesame real numberwe found in (2.7). Hence, its complex conjugate,
χ4 , χ4 , χ4
namely 3 F2 |1 is also real and has the same value. This completes
ϕ, ε
the proof of the lemma. 

Next, we note the following observations given in the beginning of the sixth
section in [6]. We state it as a lemma since we shall use it in proving Theorem 1.2.
Greene [10, 11] gave some transformation formulae which we list here as follows.
Let A, B, C, D, E be characters on Fq . Then, we have
   
A, B, C BD, AD, CD
3 F2 |1 = 3 F2 |1 , (2.9)
D, E D, ED
   
A, B, C A, AD, AE
F
3 2 |1 = ABCDE(−1) · F
3 2 |1 , (2.10)
D, E AB, AC
   
A, B, C BD, B, BE
F
3 2 |1 = ABCDE(−1) · F
3 2 |1 , (2.11)
D, E BA, BC
 NUMBER OF COMPLETE SUBGRAPHS OF PEISERT GRAPHS 9

   
A, B, C B, ECA,
3 F2 |1 = AE(−1) · 3 F2 |1 , (2.12)
D, E ABD, E
   
A, B, C A, DB, C
3 F2 |1 = AD(−1) · 3 F2 |1 , (2.13)
D, E D, ACE
   
A, B, C AD, B, C
3 F2 |1 = B(−1) · 3 F2 |1 , (2.14)
D, E D, BCE
   
A, B, C AD, BD, C
3 F2 |1 = AB(−1) · 3 F2 |1 . (2.15)
D, E D, DEAB

Let X = {(t1 , t2 , t3 , t4 , t5 ) ∈ Z54 : t1 , t2 , t3 6= 0, t4 , t5 ; t1 + t2 + t3 6= t4 , t5 }. To each

of the transformations in (2.9) to (2.15), Dawsey and McCarthy in [6] associated a
map on X; for example, the transformation in (2.9) gives that
 t1   t2 −t4 
χ4 , χt42 , χt43 χ4 , χ4t1 −t4 , χt43 −t4
3 F2 |1 = 3 F2 |1 ,
χt44 , χt45 χ4−t4 , χt45 −t4
so it induces a map f1 : X → X given by
f1 (t1 , t2 , t3 , t4 , t5 ) = (t2 − t4 , t1 − t4 , t3 − t4 , −t4 , t5 − t4 ).
Similarly, the other transformations in (2.10) to (2.15) led to the construction of
the maps f2 to f7 .
Lemma 2.8. Let X = {(t1 , t2 , t3 , t4 , t5 ) ∈ Z54 : t1 , t2 , t3 6= 0, t4 , t5 ; t1 + t2 + t3 6=
t4 , t5 }. Define the functions fi : X → X, i ∈ {1, 2, . . . , 7} in the following manner:
f1 (t1 , t2 , t3 , t4 , t5 ) = (t2 − t4 , t1 − t4 , t3 − t4 , −t4 , t5 − t4 ),
f2 (t1 , t2 , t3 , t4 , t5 ) = (t1 , t1 − t4 , t1 − t5 , t1 − t2 , t1 − t3 ) ,
f3 (t1 , t2 , t3 , t4 , t5 ) = (t2 − t4 , t2 , t2 − t5 , t2 − t1 , t2 − t3 ) ,
f4 (t1 , t2 , t3 , t4 , t5 ) = (t1 , t2 , t5 − t3 , t1 + t2 − t4 , t5 ) ,
f5 (t1 , t2 , t3 , t4 , t5 ) = (t1 , t4 − t2 , t3 , t4 , t1 + t3 − t5 ) ,
f6 (t1 , t2 , t3 , t4 , t5 ) = (t4 − t1 , t2 , t3 , t4 , t2 + t3 − t5 ) ,
f7 (t1 , t2 , t3 , t4 , t5 ) = (t4 − t1 , t4 − t2 , t3 , t4 , t4 + t5 − t1 − t2 ) .
Then the group generated by f1 , . . . , f7 , with operation composition of functions, is
the set
F = {f0 , fi , fj ◦fl , f4 ◦f1 , f6 ◦f2 , f5 ◦f3 , f1 ◦f4 ◦f1 : 1 ≤ i ≤ 7, 1 ≤ j ≤ 3, 4 ≤ l ≤ 7},
where f0 is the identity map.
Moreover, the group F acts on the set X.
 Ift1 we associate the5-tuple (t1 , t2 , . . . , t5 ) ∈
χ4 , χt42 , χt43
X to the hypergeometric function 3 F2 |1 , then each orbit of the
χt44 , χt45
group action consists of a number of 5-tuples (t1 , t2 , . . . , t5 ), and the corresponding
3 F2 terms have the same value.

Proof. For a proof, see Section 6 of [6]. 

In order to prove Theorem 1.3, the following famous theorem, due to André Weil,
serves as the crux. We state it here.
10 ANWITA BHOWMIK AND RUPAM BARMAN

Theorem 2.9 (Weil’s estimate). Let Fq be the finite field of order q, and let χ be
a character of Fq of order s. Let f (x) be a polynomial of degree d over Fq such that
f (x) cannot be written in the form c · h(x)s , where c ∈ Fq . Then
X √
χ(f (x)) ≤ (d − 1) q.
x∈Fq

The rest of the article goes as follows. In Section 3, we prove Theorem 1.1. In
Section 4, we prove Theorem 1.2. Finally, in Section 5 we prove the asymptotic
formula for the number of cliques of any order in Peisert graphs. To count the
number of cliques in Peisert graphs, we note that since the graph is vertex-transitive,
so any two vertices in the graph are contained in the same number of cliques of
a particular order. We will also use the following notation throughout the proofs.
For an induced subgraph S of a Peisert graph and a vertex v ∈ S, we denote by
k3 (S) and k3 (S, v) the number of cliques of order 3 in S and the number of cliques
of order 3 in S containing v, respectively.

3. number of 3-order cliques in P ∗ (q)

In this section, we prove Theorem 1.1. Recall that F∗q = hgi and H = hg 4 i∪ghg 4 i.
Also, hHi is the subgraph induced by H and h = 1 − χ4 (g).
Proof of Theorem 1.1. Using the vertex-transitivity of P ∗ (q), we find that
1
k3 (P ∗ (q)) =× q × k3 (P ∗ (q), 0)
3
q
= × number of edges in hHi. (3.1)
3
Now,
1 XX
the number of edges in hHi = × 1, (3.2)
2
χ4 (x−y)∈{1,χ4 (g)}

where the 1st sum is taken over all x such that χ4 (x) ∈ {1, χ4 (g)} and the 2nd sum
is taken over all y 6= x such that χ4 (y) ∈ {1, χ4 (g)}. Hence, using (2.1) in (3.2), we
find that
the number of edges in hHi
1 X
= (2 + hχ4 (x) + hχ4 (x))
2 × 43
x6=0
X
× [(2 + hχ4 (y) + hχ4 (y))(2 + hχ4 (x − y) + hχ4 (x − y))]. (3.3)
y6=0,x

We expand the inner summation in (3.3) to obtain

X
[4 + 2hχ4 (y) + 2hχ4 (y) + 2hχ4 (x − y) + 2hχ4 (x − y) + 2χ4 (y)χ4 (x − y)
y6=0,x
+ 2χ4 (y)χ4 (x − y) − 2χ4 (g)χ4 (y(x − y)) + 2χ4 (g)χ4 (y(x − y))]. (3.4)
We have
X X
χ4 (y(x − y)) = χ4 (xy)χ4 (x − xy) = ϕ(x)J(χ4 , χ4 ). (3.5)
y6=0,x y6=0,1
 NUMBER OF COMPLETE SUBGRAPHS OF PEISERT GRAPHS 11

Using Lemma 2.3 and (3.5), (3.4) yields

X
[(2 + hχ4 (y) + hχ4 (y))(2 + hχ4 (x − y) + hχ4 (x − y))]
y6=0,x

= 4(q − 3) − 4hχ4 (x) − 4hχ4 (x) − 2χ4 (g)ϕ(x)J(χ4 , χ4 ) + 2χ4 (g)ϕ(x)J(χ4 , χ4 ).

(3.6)
Now, putting (3.6) into (3.3), and then using Lemma 2.1, we find that
the number of edges in hHi
1 X
= [(2 + hχ4 (x) + hχ4 (x))(4(q − 3) − 4hχ4 (x) − 4hχ4 (x))]
2 × 43
x6=0
1 X
= [8(q − 5) + (4h(q − 3) − 8h)χ4 (x) + (4h(q − 3) − 8h)χ4 (x)]
2 × 43
x6=0
(q − 1)(q − 5)
= .
16
Substituting this value in (3.1) gives us the required result. 

4. number of 4-order cliques in P ∗ (q)

In this section, we prove Theorem 1.2. First, we recall again that F∗q = hgi and
H = hg 4 i ∪ ghg 4 i. Let J(χ4 , χ4 ) = J(χ4 , ϕ) = ρ, where the value of ρ is given by
Lemma 2.1. Let q = u2 + 2v 2 for integers u and v such that u ≡ 3 (mod 4) and
p ∤ u when p ≡ 3 (mod 8). Let χ8 be a character of order 8 such that χ28 = χ4 .
Note that in the proof we shall use the fact that χ4 (−1) = 1 multiple times. Recall
that h = 1 − χ4 (g).
Proof of Theorem 1.2. Noting again that P ∗ (q) is vertex-transitive, we find that
q
k4 (P ∗ (q)) = × number of 4-order cliques in P ∗ (q) containing 0
4
q
= × k3 (hHi). (4.1)
4
Let a, b ∈ H be such that χ4 (ab−1 ) = 1. We note that
1 XX
k3 (hHi, a) = × 1, (4.2)
2
χ4 (x−y)∈{1,χ4 (g)}

where the 1st sum is taken over all x such that χ4 (x), χ4 (a − x) ∈ {1, χ4 (g)} and
the 2nd sum is taken over all y 6= x such that χ4 (y), χ4 (a − y) ∈ {1, χ4 (g)}. Hence,
using (2.1) in (4.2), we find that
k3 (hHi, a)
1 X X
= 5
[(2 + hχ4 (a − x) + hχ4 (a − x))
2×4
x6=0,a y6=0,a,x

× (2 + hχ4 (a − y) + hχ4 (a − y))(2 + hχ4 (x − y) + hχ4 (x − y))

× (2 + hχ4 (x) + hχ4 (x))(2 + hχ4 (y) + hχ4 (y))].
Using the substitution Y = ba−1 y, the sum indexed by y in the above yields
k3 (hHi, a)
12 ANWITA BHOWMIK AND RUPAM BARMAN

1 X X
= [(2 + hχ4 (a − x) + hχ4 (a − x))
2 × 45
x6=0,a Y 6=0,b,ba−1 x

× (2 + hχ4 (Y − b) + hχ4 (Y − b))(2 + hχ4 (Y − ba−1 x) + hχ4 (Y − ba−1 x))

× (2 + hχ4 (x) + hχ4 (x))(2 + hχ4 (Y ) + hχ4 (Y ))]
1 X X
= 5
[(2 + hχ4 (a − x) + hχ4 (a − x))
2×4
Y 6=0,b x6=0,a,ab
−1 Y

× (2 + hχ4 (Y − b) + hχ4 (Y − b))(2 + hχ4 (Y − ba−1 x) + hχ4 (Y − ba−1 x))

× (2 + hχ4 (x) + hχ4 (x))(2 + hχ4 (Y ) + hχ4 (Y ))].
Again, using the substitution X = ba−1 x yields
k3 (hHi, a)
1 X X
= [(2 + hχ4 (b − X) + hχ4 (b − X))
2 × 45
Y 6=0,b X6=0,b,Y

× (2 + hχ4 (b − Y ) + hχ4 (b − Y ))(2 + hχ4 (X − Y ) + hχ4 (X − Y ))

× (2 + hχ4 (X) + hχ4 (X))(2 + hχ4 (Y ) + hχ4 (Y ))]
= k3 (hHi, b).
Thus, if a, b ∈ H are such that χ4 (ab−1 ) = 1, then
k3 (hHi, a) = k3 (hHi, b). (4.3)
Let hg 4 i = {x1 , . . . , x q−1 } with x1 = 1 and ghg 4 i = {y1 , . . . , y q−1 } with y1 = g.
4 4
Then,
q−1 q−1
X
4 X
4

k3 (hHi, xi ) + k3 (hHi, yi ) = 3 × k3 (hHi). (4.4)

i=1 i=1

By (4.3), we have
k3 (hHi, x1 ) = k3 (hHi, x2 ) = · · · = k3 (hHi, x q−1 )
4

and
k3 (hHi, y1 ) = k3 (hHi, y2 ) = · · · = k3 (hHi, y q−1 ).
4

Hence, (4.4) yields

q−1
k3 (hHi) =[k3 (hHi, 1) + k3 (hHi, g)]. (4.5)
12
Thus, we need to find only k3 (hHi, 1) and k3 (hHi, g). We first find k3 (hHi, 1).
We have
k3 (hHi, 1)
1 X
= 5
[(2 + hχ4 (1 − x) + hχ4 (1 − x))(2 + hχ4 (x) + hχ4 (x))]
2×4
x6=0,1
X
[(2 + hχ4 (1 − y) + hχ4 (1 − y))(2 + hχ4 (x − y) + hχ4 (x − y))
y6=0,1,x

× (2 + hχ4 (y) + hχ4 (y))]. (4.6)

 NUMBER OF COMPLETE SUBGRAPHS OF PEISERT GRAPHS 13

Let i1 , i2 , i3 ∈ {±1} and let Fi1 ,i2 ,i3 denote the term χi41 (y)χi42 (1 − y)χi43 (x − y).
Using this notation, we expand and evaluate the inner summation in (4.6). We
have
X
[2 + hχ4 (y) + hχ4 (y)][2 + hχ4 (1 − y) + hχ4 (1 − y)][2 + hχ4 (x − y) + hχ4 (x − y)]
y6=0,1,x
X
= [8 + 4hχ4 (y) + 4hχ4 (y) + 4hχ4 (1 − y) + 4hχ4 (1 − y) + 4hχ4 (x − y)
y6=0,1,x

+ 4hχ4 (x − y) + 4χ4 (y)χ4 (1 − y) + 4χ4 (y)χ4 (1 − y) + 4χ4 (y)χ4 (x − y)

+ 4χ4 (y)χ4 (x − y) + 4χ4 (1 − y)χ4 (x − y) + 4χ4 (1 − y)χ4 (x − y)
2
+ 2h2 χ4 (y)χ4 (1 − y) + 2h χ4 (y)χ4 (1 − y) + 2h2 χ4 (y)χ4 (x − y)
2 2
+ 2h χ4 (y)χ4 (x − y) + 2h2 χ4 (1 − y)χ4 (x − y) + 2h χ4 (1 − y)χ4 (x − y)
+ h3 F1,1,1 + 2hF1,1,−1 + 2hF1,−1,1 + 2hF1,−1,−1 + 2hF−1,1,1 + 2hF−1,1,−1
3
+ 2hF−1,−1,1 + h F−1,−1,−1 ]. (4.7)
Now, P
referring to Lemmas 2.2
P and 2.3, we can easily check that any
P term of the
form χ4 (·)χ4 (·) gives −1, χ4 ((y −1)(y −x)) gives ϕ(x−1)ρ and χ4 (y(y −x))
y y y
gives ϕ(x)ρ. Hence, (4.7) yields
X
[2 + hχ4 (y) + hχ4 (y)][2 + hχ4 (1 − y) + hχ4 (1 − y)][2 + hχ4 (x − y) + hχ4 (x − y)]
y6=0,1,x

= A + Bχ4 (x) + Bχ4 (x) + Bχ4 (x − 1) + Bχ4 (x − 1) − 4χ4 (x)χ4 (x − 1)

2
− 4χ4 (x)χ4 (x − 1) − 2h2 χ4 (x)χ4 (x − 1) − 2h χ4 (x)χ4 (x − 1)
+ h3 F1,1,1 + 2hF1,1,−1 + 2hF1,−1,1 + 2hF1,−1,−1 + 2hF−1,1,1 + 2hF−1,1,−1
3
+ 2hF−1,−1,1 + h F−1,−1,−1
=: I, (4.8)
where A = 8(q − 8) and B = −12h.
Next, we introduce some notations. Let
B1 = 16(q − 9) + 6B + Bh2 ,
D1 = 2B − 8h + Bh2 − 4h3 ,
E1 = 8(q − 9) + 4Bh,
F1 = 16(q − 9) + 4Re(Bh).
For i ∈ {1, 2, 3, 4} and j ∈ {1, 2, . . . , 8}, we define the following character sums.
X X
Tj := χi41 (y)χi42 (1 − y)χi43 (x − y),
x6=0,1 y
X X
Uij := χl4 (m) χi41 (y)χi42 (1 − y)χi43 (x − y),
x6=0,1 y
X X
Vij := χl41 (x)χl42 (1 − x) χi41 (y)χi42 (1 − y)χi43 (x − y),
x y
14 ANWITA BHOWMIK AND RUPAM BARMAN

where

1, if i is odd,
l=
−1, otherwise;

x, if i ∈ {1, 2},
m=
1 − x, otherwise;
and


 (1, 1), if i = 1,

(1, −1), if i = 2,
(l1 , l2 ) =

 (−1, 1), if i = 3,

(−1, −1), if i = 4.
Also, corresponding to each j, let (i1 , i2 , i3 ) take the value according to the following
table:

j i1 i2 i3
1 1 1 1
2 1 1 −1
3 1 −1 1
4 1 −1 −1
5 −1 1 1
6 −1 1 −1
7 −1 −1 1
8 −1 −1 −1

Then, using (4.8) and the notations we just described, (4.6) yields
1 X
k3 (hHi, 1) = [2 + hχ4 (x) + hχ4 (x)][2 + hχ4 (1 − x) + hχ4 (1 − x)] × I
2048
x6=0,1
1 X h
= 32(q − 15) + B1 χ4 (x) + B1 χ4 (x) + B1 χ4 (x − 1) + B1 χ4 (x − 1)
2048
x6=0,1

+ 4Re(Bh)ϕ(x) + 4Re(Bh)ϕ(x − 1) + D1 χ4 (x)ϕ(x − 1) + D1 χ4 (x)ϕ(x − 1)

+ D1 ϕ(x)χ4 (x − 1) + D1 ϕ(x)χ4 (x − 1) + E1 χ4 (x)χ4 (x − 1) + E1 χ4 (x)χ4 (x − 1)
i
+ F1 χ4 (x)χ4 (1 − x) + F1 χ4 (x)χ4 (x − 1)
1 h 3 3
+ 5
4h T1 + 8hT2 + 8hT3 + 8hT4 + 8hT5 + 8hT6 + 8hT7 + 4h T8
2×4
2
+ 2h4 U11 + 4h2 U12 + 4h2 U13 + 8U14 + 4h2 U15 + 8U16 + 8U17 + 4h U18
2 2 2 4
+ 4h2 U21 + 8U22 + 8U23 + 4h U24 + 8U25 + 4h U26 + 4h U27 + 2h U28
2
+ 2h4 U31 + 4h2 U32 + 4h2 U33 + 8U34 + 4h2 U35 + 8U36 + 8U37 + 4h U38
2 2 2 4
+ 4h2 U41 + 8U42 + 8U43 + 4h U44 + 8U45 + 4h U46 + 4h U47 + 2h U48
+ h5 V11 + 2h3 V12 + 2h3 V13 + 4hV14 + 2h3 V15 + 4hV16 + 4hV17 + 4hV18
3
+ 2h3 V21 + 4hV22 + 4hV23 + 4hV24 + 4hV25 + 4hV26 + 4hV27 + 2h V28
 NUMBER OF COMPLETE SUBGRAPHS OF PEISERT GRAPHS 15

3
+ 2h3 V31 + 4hV32 + 4hV33 + 4hV34 + 4hV35 + 4hV36 + 4hV37 + 2h V38
3 3 3 5
i
+ 4hV41 + 4hV42 + 4hV43 + 2h V44 + 4hV45 + 2h V46 + 2h V47 + h V48 .
Using Lemmas 2.4, 2.5 and 2.6, we find that
1 
k3 (hHi, 1) = 32(q 2 − 20q + 81)
2048
+ h5 V11 + 2h3 V12 + 2h3 V13 + 4hV14 + 2h3 V15 + 4hV16 + 4hV17 + 4hV18
3
+ 2h3 V21 + 4hV22 + 4hV23 + 4hV24 + 4hV25 + 4hV26 + 4hV27 + 2h V28
3
+ 2h3 V31 + 4hV32 + 4hV33 + 4hV34 + 4hV35 + 4hV36 + 4hV37 + 2h V38
3 3 3 5
i
+4hV41 + 4hV42 + 4hV43 + 2h V44 + 4hV45 + 2h V46 + 2h V47 + h V48 . (4.9)
Now, we convert each term of the form Vij [i ∈ {1, 2, 3, 4}, j ∈ {1, 2, . . . , 8}] into
2
its equivalent
 t1 q · t32F2 form. We use the notation (t1 , t2 , . . . , t5 ) ∈ Z54 for the term

t3
χ4 , χ4 , χ4
q 2 · 3 F2 |1 . Then, (4.9) yields
χt44 , χt45
1 
k3 (hHi, 1) = 32(q 2 − 20q + 81)
2048
+ h5 (3, 1, 1, 2, 2) + 2h3 (1, 1, 3, 2, 0) + 2h3 (3, 1, 1, 0, 2) + 4h(1, 1, 3, 0, 0)
+ 2h3 (3, 3, 1, 0, 2) + 4h(1, 3, 3, 0, 0) + 4h(3, 3, 1, 2, 2) + 4h(1, 3, 3, 2, 0)
+ 2h3 (3, 1, 3, 2, 2) + 4h(1, 1, 1, 2, 0) + 4h(3, 1, 3, 0, 2) + 4h(1, 1, 1, 0, 0)
3
+ 4h(3, 3, 3, 0, 2) + 4h(1, 3, 1, 0, 0) + 4h(3, 3, 3, 2, 2) + 2h (1, 3, 1, 2, 0)
+ 2h3 (3, 1, 3, 2, 0) + 4h(1, 1, 1, 2, 2) + 4h(3, 1, 3, 0, 0) + 4h(1, 1, 1, 0, 2)
3
+ 4h(3, 3, 3, 0, 0) + 4h(1, 3, 1, 0, 2) + 4h(3, 3, 3, 2, 0) + 2h (1, 3, 1, 2, 2)
3
+ 4h(3, 1, 1, 2, 0) + 4h(1, 1, 3, 2, 2) + 4h(3, 1, 1, 0, 0) + 2h (1, 1, 3, 0, 2)
3 3 5
i
+4h(3, 3, 1, 0, 0) + 2h (1, 3, 3, 0, 2) + 2h (3, 3, 1, 2, 0) + h (1, 3, 3, 2, 2) . (4.10)
Next, we use Lemma 2.8 alongwith the notations therein. We list the tuples
(t1 , t2 , . . . , t5 ) in each orbit of the group action of F on X, and then group the
corresponding terms in (4.10) together. The orbit representatives (1, 1, 1, 0, 0),
(3, 3, 3, 0, 0), (1, 3, 3, 2, 0), (3, 1, 1, 2, 0) and (1, 1, 3, 0, 0) given in the proof of Corol-
lary 2.7 in [6] are the ones whose orbits exhaust the hypergeometric terms in
(4.10). We denote the q 2 · 3 F2 terms corresponding to these orbit representatives
as M1 , M2 , . . . , M5 respectively. Then, (4.10) yields
1 
k3 (hHi, 1) = 32(q 2 − 20q + 81)
2048
+ h5 M4 + 2h3 M1 + 2h3 M1 + 4hM5 + 2h3 M1 + 4hM5 + 4hM1 + 4hM3
3
+ 2h3 M4 + 4hM5 + 4hM2 + 4hM1 + 4hM5 + 4hM5 + 4hM5 + 2h M3
3
+ 2h3 M4 + 4hM5 + 4hM5 + 4hM5 + 4hM2 + 4hM1 + 4hM5 + 2h M3
3 3 3 5
i
+ 4hM4 + 4hM3 + 4hM5 + 2h M2 + 4hM5 + 2h M2 + 2h M2 + h M3 .
(4.11)
16 ANWITA BHOWMIK AND RUPAM BARMAN

Using Lemma 2.7 (note that we could not reduce M5 ), (4.11) yields
  
1 2 t 2 χ4 , χ4 , χ4
k3 (hHi, 1) = 2(q − 20q + 81) + 2u(−p) + 3q · 3 F2 |1 .
128 ε, ε
(4.12)
Returning back to (4.5), we are now left to calculate k3 (hHi, g). Again, we have
k3 (hHi, g)
1 X X 
= (2 + hχ4 (g − x) + hχ4 (g − x))(2 + hχ4 (g − y) + hχ4 (g − y))
2048
x6=0,g y6=0,g,x

× (2 + hχ4 (x − y) + hχ4 (x − y))(2 + hχ4 (x) + hχ4 (x))(2 + hχ4 (y) + hχ4 (y)) .
(4.13)
Using the substitutions Y = yg −1 and X = xg −1 , and then using the fact that
hχ4 (g) = h, (4.13) yields
k3 (hHi, g)
1 X X 
= (2 + hχ4 (1 − x) + hχ4 (1 − x))(2 + hχ4 (1 − y) + hχ4 (1 − y))
2048
x6=0,1 y6=0,1,x

× (2 + hχ4 (x − y) + hχ4 (x − y))(2 + hχ4 (x) + hχ4 (x))(2 + hχ4 (y) + hχ4 (y)) .
Comparing this with (4.6) we see that the expansion of the expression inside this
summation will consist of the same summation terms as in (4.6) except that the
coefficient corresponding to each summation will become the complex conjugate of
the corresponding coefficient of the same summation. This means that, to calculate
the coefficient of each summation after expanding the expression in (4.13), we need
to replace each corresponding coefficient in (4.11) by its complex conjugate. Now,
(4.12) is the final expression from (4.11), and we see that (4.12) contains three
summands, two of them being real numbers and the other being a 3 F2 term whose
coefficient is also a real number. Then by the foregoing argument, (4.13) yields the
same value as given in (4.12). Thus, (4.5) gives that
  
q−1 2 t 2 χ4 , χ4 , χ4
k3 (hHi) = 2(q − 20q + 81) + 2u(−p) + 3q · 3 F2 |1 .
768 ε, ε
Substituting the above value in (4.1), we complete the proof of the theorem. 

5. proof of theorem 1.3

Let m ≥ 1 be an integer. We have observed that the calculations for computing
the number of 4-order cliques in P ∗ (q) become very tedious. However, we can have
an asymptotic result on the number of cliques of order m in P ∗ (q) as q → ∞. The
method follows along the lines of [20] and so we prove by the method of induction.
Proof of Theorem 1.3. Let F∗q = hgi. We set a formal ordering of the elements
of Fq : {a1 < · · · < aq }. Let χ4 be a fixed character on Fq of order 4 and let
h = 1 − χ4 (g). First, we note that the result holds for m = 1, 2 and so let m ≥ 3.
Let the induction hypothesis hold for m − 1. We shall use the notation ‘am 6= ai ’
to mean am 6= a1 , . . . , am−1 . Recalling (2.1), we see that
X X Y 2 + hχ4 (ai − aj ) + hχ3 (ai − aj )
4
km (P ∗ (q)) = ···
a <···<a
4
1 m 1≤i<j≤m
 NUMBER OF COMPLETE SUBGRAPHS OF PEISERT GRAPHS 17


1 X X

Y 2 + hχ4 (ai − aj ) + hχ34 (ai − aj )
= ···
m a <···<a 4
1 m−1 1≤i<j≤m−1

X m−1
Y
1
{2 + hχ4 (am − ai ) + hχ34 (am − ai )} (5.1)
4m−1
am 6=ai i=1

In order to use the induction hypothesis, we try to bound the expression

X m−1
Y
{2 + hχ4 (am − ai ) + hχ34 (am − ai )}
am 6=ai i=1

in terms of q and m. We find that

X m−1
Y
J := {2 + hχ4 (am − ai ) + hχ34 (am − ai )}
am 6=ai i=1

= 2m−1 (q − m + 1)
X
+ [(3m−1 − 1) number of terms containing expressions in χ4 ] (5.2)
am 6=ai

Each term in (5.2) containing χ4 is of the form

′ j′
2f hi h χ4 ((am − ai1 )j1 · · · (am − ais )js ),
where 
0 ≤ f ≤ m − 2, 



0 ≤ i′ , j ′ ≤ m − 1, 
i1 , . . . , is ∈ {1, 2, . . . , m − 1}, (5.3)


j1 , . . . , js ∈ {1, 3}, and 


1 ≤ s ≤ m − 1.
Let us consider such an instance of a term containing χ4 . Excluding the constant
′ j′
factor 2f hi h , we obtain a polynomial in the variable am . Let g(am ) = (am −
ai1 )j1 · · · (am − ais )js . Using Weil’s estimate (Theorem 2.9), we find that
X √
| χ4 (g(am )) |≤ (j1 + · · · + js − 1) q. (5.4)
am ∈Fq

Then, using (5.4) we have

′ j′ X ′ ′ √
|2f hi h χ4 (g(am ))| ≤ 2f +i +j (j1 + · · · + js − 1) q
am
√
≤ 23m−4 (3m − 4) q
√
≤ 23m · 3m q. (5.5)
Noting that the values of χ4 are roots of unity, using (5.5), and using (5.3) and the
conditions therein, we obtain
′ j′ X
| 2 f hi h χ4 (g(am )) |
am 6=ai
( )
′ j′ X
=| 2f hi h χ4 (g(am )) − χ4 (g(a1 )) − · · · − χ4 (g(am−1 )) |
am
18 ANWITA BHOWMIK AND RUPAM BARMAN

√
≤ 23m · 3m q + 22m−3
√
≤ 22m (1 + 2m · 3m q),
that is,
√ ′ j′ X √
−22m (1 + 2m · 3m q) ≤ 2f hi h χ4 (g(am )) ≤ 22m (1 + 2m · 3m q).
am 6=ai

Then, (5.2) yields

√
2m−1 (q − m + 1) − 22m (1 + 2m · 3m q)(3m−1 − 1)
≤J
√
≤ 2m−1 (q − m + 1) + 22m (1 + 2m · 3m q)(3m−1 − 1)
and thus, (5.1) yields
√ 1
[2m−1 (q − m + 1) − 22m (1 + 2m · 3m q)(3m−1 − 1)] × km−1 (P ∗ (q))
m × 4m−1
≤ km (P ∗ (q))
√ 1
≤ [2m−1 (q − m + 1) + 22m (1 + 2m · 3m q)(3m−1 − 1)] × km−1 (P ∗ (q))
m × 4m−1
(5.6)
Dividing by q m throughout in (5.6) and taking q → ∞, we have
√
2m−1 (q − m + 1) − 22m (1 + 2m · 3m q)(3m−1 − 1) km−1 (P ∗ (q))
lim lim
q→∞ m × 4m−1 × q q→∞ q m−1
∗
km (P (q))
≤ lim
q→∞ qm
√
2 m−1
(q − m + 1) + 22m (1 + 2m · 3m q)(3m−1 − 1) km−1 (P ∗ (q))
≤ lim lim
q→∞ m × 4m−1 × q q→∞ q m−1
(5.7)
Now, using the induction hypothesis and noting that
√
2m−1 (q − m + 1) ± 22m (1 + 2m · 3m q)(3m−1 − 1)
lim
q→∞ m × 4m−1 q
1
= 2m−1
m × 4m−1
1
= ,
m × 2m−1
we find that both the limits on the left hand side and the right hand side of (5.7)
are equal. This completes the proof of the result. 

Taking m = 3 in Theorem 1.3, we find that

k3 (P ∗ (q)) 1
lim = .
q→∞ q3 48
We obtain the same limiting value from Theorem 1.1 as well.
Taking m = 4 in Theorem 1.2 and Theorem 1.3, we obtain the following corollary
which is also evident from Table 1.
 NUMBER OF COMPLETE SUBGRAPHS OF PEISERT GRAPHS 19

Corollary 5.1. We have

 
χ4 , χ4 , χ34
lim 3 F2 |1 = 0.
q→∞ ε, ε
Proof. Putting m = 4 in Theorem 1.3, we have
k4 (P ∗ (q)) 1
lim 4
= . (5.8)
q→∞ q 1536
Putting m = 4 in Theorem 1.2, we have
 
k4 (P ∗ (q)) 1 χ4 , χ4 , χ34
lim = + 3 × lim 3 F2 |1 . (5.9)
q→∞ q4 1536 q→∞ ε, ε
Combining (5.8) and (5.9), we complete the proof. 

6. Acknowledgements
We are extremely grateful to Ken Ono for previewing a preliminary version of
this paper and for his helpful comments.

References
[1] J. Alexander, Selected results in combinatorics and graph theory, Thesis (Ph.D.) University
of Delaware (2016).
[2] R. Atanasov, M. Budden, J. Lambert, K. Murphy and A. Penland, On certain induced
subgraphs of Paley graphs, Acta Univ. Apulensis Math. Inform. 40 (2014), 51–65.
[3] B. C. Berndt, K. S. Williams and R. J. Evans, Gauss and Jacobi sums, Canadian Mathemat-
ical Society Series of Monographs and Advanced Texts, A Wiley-Interscience Publication,
John Wiley & Sons, Inc., New York, 1998.
[4] A. Bhowmik and R. Barman, On a Paley-type graph on Zn , Graphs and Combinatorics 38
(2022), no. 2, Paper No. 41, 25 pp.
[5] C. Chao, On the classification of symmetric graphs with a prime number of vertices, Trans-
actions of the American Mathematical Society 158 (1971), 247–256.
[6] M. L. Dawsey and D. McCarthy, Generalized Paley graphs and their complete subgraphs of
orders three and four, Research in the Mathematical Sciences 8 (2021), no. 2, Paper No. 18,
23 pp.
[7] P. Erdös, On the number of complete subgraphs contained in certain graphs, Magyar Tud.
Akad. Mat. Kutató Int. Közl 7 (1962), 459–464.
[8] R.J. Evans, J.R. Pulham and J. Sheehan, On the number of complete subgraphs contained in
certain graphs, Journal of Combinatorial Theory, Series B, 3 (1981), 364–371.
[9] A. W. Goodman, On sets of acquaintances and strangers at any party, The American Math-
ematical Monthly 66 (1959), 778–783.
[10] J. Greene, Hypergeometric functions over finite fields, Trans. Amer. Math. Soc. 301 (1987),
77–101.
[11] J. Greene, Character Sum Analogues for Hypergeometric and Generalized Hypergeometric
Functions over Finite Fields, Ph.D. thesis, Univ. of Minnesota, Minneapolis, 1984.
[12] S. A. Katre and A. R. Rajwade, Resolution of the sign ambiguity in the determination
of the cyclotomic numbers of order 4 and the corresponding Jacobsthal sum, Mathematica
Scandinavica 60 (1987), 52–62.
[13] T. K. Lim and C. E. Praeger, On generalised Paley graphs and their automorphism groups,
Michigan Mathematical Journal 58 (2009), 293–308.
[14] D. McCarthy, Transformations of well-poised hypergeometric functions over finite fields, Fi-
nite Fields and Their Applications, 18 (2012), no. 6, 1133–1147.
[15] K. Ono, Values of Gaussian hypergeometric series, Trans. Amer. Math. Soc. 350 (1998), no.
3, 1205–1223.
[16] K. Ono, The web of modularity: arithmetic of the coefficients of modular forms and q-series,
CBMS Regional Conference Series in Mathematics, 102, Amer. Math. Soc., Providence, RI,
2004.
20 ANWITA BHOWMIK AND RUPAM BARMAN

[17] R. E. A. C. Paley, On orthogonal matrices, Journal of Mathematics and Physics 12 (1933),

311–320.
[18] W. Peisert, All self-complementary symmetric graphs, Journal of Algebra 240 (2001), 209–
229.
[19] A. G. Thomason, Partitions of Graphs, Ph.D. thesis, Cambridge University (1979).
[20] N. Wage, Character sums and Ramsey properties of generalized Paley graphs, Integers 6
(2006), article number 18.
[21] H. Zhang, Self-complementary symmetric graphs, Journal of Graph Theory 16 (1992), 1–5.

Department of Mathematics, Indian Institute of Technology Guwahati, North Guwa-

hati, Guwahati-781039, Assam, INDIA
Email address: [email protected]

Department of Mathematics, Indian Institute of Technology Guwahati, North Guwa-

hati, Guwahati-781039, Assam, INDIA
Email address: [email protected]