1. Introduction
A partially balanced incomplete block design is a triple (
P,
B,
I), where
P is a set of elements called points and
B a family of sets called blocks, together with a point-block symmetric incidence relation
I ⊆
P ×
B, where the size of
P is
v, the size of
B is
b, each point of
P is contained in
r blocks of
B, each blocks of
B contains
k points of
P and two points are contained in
λ blocks,
λ ∈ {
λ1,
λ2, …,
λm}. A symmetric relation of association between two points is established: two points are
ith associates for some
i, with
I ∈ {1, 2, …,
m}, if they are contained in exactly
blocks. For this reason, a partially balanced incomplete block design is also called an
m-class association scheme, cf. [
1,
2,
3]. The number of
ith associates of each point is
ni. If
p and
q are two points which are
ith associates, then the number of points which are
jth associates of
p and
kth associates of
q is
and it is independent of the pair of
ith associates
p and
q. The numbers
v,
b,
r,
k,
λ1,
λ2, …,
λm are called parameters of first kind and the numbers
ni’s and
’s are called parameters of second kind. They satisfy the relation:
vr =
bk;
v ≤
b;
;
;
if
i ≠
j;
if
i =
j;
. If
b =
v, and so
r =
k, i.e., the number of points is equal to the number of lines, the partially balanced incomplete block design is said to be symmetric. Throughout the paper, we only consider connected incidence structures, where any two elements of
P∪
B are connected via a path of incident elements. A 2-(
v,
k,
λ) balanced incomplete block design is a partially balanced incomplete block design where every pair of points occurs in exactly
λ blocks. The parameters
v,
b,
r,
k and
λ satisfy the relation:
λ (
v−1) =
r(
k−1). A (
vr,
bk)-configuration is a partially balanced incomplete block design such that two different points can be in at most one block. A symmetric configuration is denoted by (
vk). A
t-(
v,
k,1) design is a (
vr,
bk)-configuration such that any
t distinct points belong to exactly one block. A
t-(
v,
k,1) design is called a Steiner system S (
t,
k,
v). A S (3,
q + 1,
q2 + 1) is said to be a finite inversive plane (also known as Möbius plane) of order
q. A 3-(
v,4,1) design is called a Steiner Quadruple system and denoted SQS (
v). Designs exhibit many remarkable properties and appear to be closely interconnected, cf. [
4,
5]. A peculiar relationship in this sense occurs when one design completely determines another one, cf. [
6,
7]. The mutual relation between the projective plane of order four and the three-dimensional projective space of order two by the factorizations of the complete graph
K6 on six vertices was first investigated by Beutelspacher, cf. [
8]. Three new geometric descriptions for one of the two projective equivalence classes of 15-sets of type (3,6)
2 in PG(3,3) were provided by Tondini in [
9]. In [
10] and [
11], the close connection between the Desargues and the Cremona–Richmond configurations is provided. The Steiner Quadruple system with ten points, SQS(10), is one of the most important point-line incidence structures. It consists of ten points and thirty blocks, with four points on each block, twelve blocks passing through any point and at most one block through two different points. Up to isomorphism, there is a unique SQS(10); see [
12]. The Cremona–Richmond configuration, also known as the generalized quadrangle GQ(2,2), is the only triangle-free symmetric (15
3) configuration. Edge was the first to describe the fundamental properties of the elliptic quadric of PG(3,3). The 30 points lying not on the elliptic quadric of PG(3,3) are joined by lines called chords. There are exactly 45 such chords. In [
13], he divided the 30 points not on the quadric into two symmetric groups of 15 points and arranged each group into a set of 6 pentagons, with each of the 15 points being the common vertex of two of them. Tutte demonstrated that the 45 chords of an elliptic quadric in PG(3,3) form the edges of the highly symmetric, triangle-free cubic graph, cf. [
14]. In [
15], Coxeter connected the projective geometry of an elliptic quadric in PG(3,3) with the combinatorial properties of the symmetric group by proving that the automorphism group of the chords’ graph is isomorphic to the group of projective transformations under which the elliptic quadric is invariant. More recently, Brier and Bryant [
16] constructed SQS(10), where the points are the ten triangle factors of
K6 and the blocks are the fifteen edges of
K6 and the fifteen 1-factors of
K6 and the close connection of SQS(10) and GQ(2,2) is given by showing that the two types of blocks correspond with the points and the lines of the Cremona–Richmond configuration. The purpose of this research is to highlight, from another point of view, the relationship between the Steiner Quadruple system with ten points and the Cremona–Richmond configuration by the incidence properties of the elliptic quadric of PG(3,3). The relation between the external points of a quadric and partially balanced incomplete block designs is not a new one and D. K. Ray-Chaudhury, in his pioneering paper [
17], constructed two associate class partially incomplete block designs by linear flats contained in quadrics. In [
3], the authors consider degenerate quadrics and construct a family of four-class association schemes. In [
1], association schemes on the anisotropic points of classical polar spaces are studied. Geometric construction of some families of two-class and three-class association schemes from non-degenerate quadrics in characteristic two is provided in [
2]. The paper is organized as follows.
Section 2 introduces the preliminary result. In
Section 3, by the Singer representation of PG(3,3), we obtain the two symmetric Cremona–Richmond configurations of the points not belonging to the elliptic quadric. In
Section 4, by the Singer representation of PG(3,5), we obtain the two symmetric partially balanced incomplete block designs of the points not belonging to the elliptic quadric. Finally,
Section 5 concludes with remarks and possible directions for future research.
2. The Incidence Properties of the Elliptic Quadric of PG(3,q), q Odd
In this section, we prove a theorem showing the close connection between a finite inversive plane of order
q and a partially balanced incomplete block design. In PG(3,
q), a non-degenerate quadric
Q is defined by the vanishing of a quadratic form
in four variables. When
q is odd, the quadric
Q can be classified based on the determinant
of the associated matrix: a hyperbolic quadric
Q+(3,
q) if
is a square in
, and an elliptic quadric
Q−(3,
q) if
is not a square in
. When evaluating the quadratic form at a fixed point
not lying on the quadric, the value of the quadratic form
determines whether the point belongs to the set of squares or non-squares of the field
. The points
P not belonging to
Q for which
is a square are called squares, while the points
P not belonging to
Q for which
is not a square are called non-squares. The polar planes with respect to
Q of square points are said to be square planes, while the polar planes of non-square points are said to be non-square planes, cf. [
1].
Theorem 1. An elliptic quadric Q−(3,q) of PG(3,q), q odd, defines a Möbius plane and two isomorphic partially balanced incomplete block designs.
Proof. Let Q = Q−(3,q) be an elliptic quadric of PG(3,q), q odd, having quadratic form . A secant plane intersects the elliptic quadric Q in a non-singular conic containing exactly q + 1 points. Since three non-collinear points are contained in exactly one plane, the points of an elliptic quadric Q with the secant planes define a Möbius plane, i.e., a S (3, q + 1, q2 + 1) Steiner system. The q (q2 + 1) points of PG(3,q) not belonging to Q are partitioned into two sets of equal size: square, say S, and non-square, N. In each secant plane, the remaining q2 points, not on Q, within the plane are partitioned into square and non-square points. There are q (q2 + 1) secant planes in total, which are polar planes of points not belonging to Q−(3,q). The geometric and algebraic properties of the polar plane perfectly mirror the quadratic nature of its pole. They are divided equally into two classes: square planes and non-square planes. A square plane contains square points and non-square points; conversely, a non-square plane contains square points and non-square points. The reason for this imbalance lies in the restriction of the quadratic form to the polar plane and in the reciprocity theorem. When the quadratic form of PG(3,q) is restricted to the three-dimensional vector subspace defining the polar plane, it becomes a quadratic form in three variables of a specific type. We establish the behavior for both types of points.
If the pole is a non-zero square, the discriminant of this restricted form dictates that the non-degenerate conic behaves like a hyperbolic conic within the algebraic structure of that plane. To prove the distribution of points in the polar plane
of a non-zero square point
P, we use the algebraic properties of the bilinear form associated with the elliptic quadric in PG(3,
q),
q odd. Let
be the vector space associated with PG(3,
q) and let
be the quadratic form defining the elliptic quadric. Let
be the symmetric bilinear form obtained by polarizing
:
A point P in PG(3,q) corresponds to a one-dimensional subspace .
If P is a non-zero square point, then
. The polar plane
is the three-dimensional orthogonal complement
with respect to
:
To analyze the points inside
, we examine the restriction of
to the three-dimensional subspace
. Let
be this restricted quadratic form. Since P does not belong to the quadric, the subspace
does not contain
p, and
V decomposes into a direct sum:
. For any vector
, we can write
, where
and
. Evaluating the global quadratic form
on
yields:
Since
, the cross term
vanishes, simplifying to:
The determinant (discriminant) of a quadratic form determines the distribution of its values. Let:
From the matrix representation of the direct sum decomposition, the global discriminant factorizes as:
The global discriminant of an elliptic quadric in PG(3,q) is non–square in (this is the defining algebraic property that ensures that an elliptic quadric has no lines and contains exactly q2 + 1 points).
In a finite field , the product of a non-zero square with a non-square is always a non-square. Thus, if P is a square point, we get that is square and is non-square in .
A three-dimensional quadratic form over
, with a non-square discriminant, defines a non-degenerate conic in the projective plane
. For such a ternary form
, standard finite field character sums yield the precise number of vectors
mapping to each value
. For any non-zero value
, If
is non–square in
, the number of vector solutions to
depends on whether
z matches the character of
. Since
dictates the geometry, the number of non-zero vector solutions in
is explicitly given by:
where
is the Legendre symbol, i.e., a function that equals 1 if
is a square, or x1 if
is a non-square.
Because is a non-square in , evaluating this for all squares and non-squares in the three-dimensional subspace gives the total vector counts:
zero vectors (on the conic): q2 vectors ⇒ q + 1 projective points (since );
non-zero square vectors: exactly vectors;
non-square vectors: exactly vectors.
To convert these vector counts within into projective points in , we divide the non-zero vector totals by :
This proves algebraically that the polar plane
of a non-zero square point contains a surplus of exactly
q square points over non-square points:
If we choose the pole P to be a non-square point, the algebraic symmetry of the finite field causes the entire distribution inside the polar plane to completely flip.
If P is a non-square point, then is a non-square in . With the same notation as above, the global discriminant is a non-square in (because the quadric is an elliptic quadric) and is a non-square (because P is a non-square point).
In a finite field
, the product of two non-squares is always a square. Thus, if
P is a non-square point, we get that
is non-square and
is a square in
. When the discriminant
of a three-variable quadratic form is a square, the geometric nature of the conic section within the plane
remains non-degenerate, still containing exactly
q + 1 points. However, the internal distribution of the field values reverses. The character sum formula for the number of vector solutions in
changes its sign behavior:
Since is a square, the character depends entirely on . This flips the counting results between squares and non-squares at the vector level:
zero vectors (on the conic): q2 vectors ⇒ q + 1 projective points (since );
non-zero square vectors: exactly vectors;
non-square vectors: exactly vectors.
To convert these vector counts within into projective points in , we divide the non-zero vector totals by :
This proves algebraically that the polar plane
of a non-square point contains a surplus of exactly
q non-square points over non-zero square points:
This proves the beautiful reciprocity of Galois geometries: a pole always commands a majority of its own algebraic type inside its own polar plane.
To recap, in PG(3,q) with q odd, the polar plane of a point not on the elliptic quadric is a secant plane. It intersects the quadric in a non-degenerate conic containing exactly q + 1 points.
When the pole is a non-zero square point, its polar plane contains an asymmetric surplus of non-zero square points, partitioned as follows:
q + 1 points lie on the quadric, forming the intersection conic;
points are non-zero squares;
points are non-squares.
When the pole is a non-square point, its polar plane contains an asymmetric surplus of non-square points, partitioned as follows:
q + 1 points lie on the quadric, forming the intersection conic;
points are non-squares;
points are non-zero squares.
To determine how many square and non-square planes pass through a line in PG(3,q), q odd, we must classify the line with respect to the elliptic quadric Q. In the dual two-dimensional projective space (the pencil of planes passing through a line), the classification of the planes (square or non-square) exactly reflects the signature of the points on the polar line ⊥. Since the quadric is elliptic, the polar line ⊥ always has the opposite or inverted nature relative to . The number of planes for each type depends exclusively on the behavior of relative to Q, falling into three possible geometric cases. If is external, it contains exactly square points and non-square points. By the principle of polar reciprocity, its polar line ⊥ is a secant line to the quadric. A secant line contains two points of the quadric, square points and non-square points. Since the planes passing through correspond bijectively to the points of ⊥, and the classification (square/non-square) is inverted upon passing to the dual in the elliptic case, we have that exactly square planes and non-square planes pass through an external line . If is a secant line, it contains exactly square points and non-square points. Its polar line ⊥ is a line external to the quadric. An external line contains exactly square points and non-square points. By reversing the nature of the points on ⊥ to find the planes through , we get that exactly square planes and non-square planes pass through a secant line . Now, suppose that is a tangent line at a point T. If we establish a system of homogeneous coordinates in the tangent plane by placing the origin at T, every line passing through T can be identified by its direction. It contains exactly square points and non-square points. The tangent lines t whose direction is a square are called squares, while the tangent lines t whose direction is not a square are called non-squares. A square tangent line contains exactly square points and non-square points. A non-square tangent line contains exactly square points and non-square points.
Exactly square planes pass through a square tangent line t. This geometric result is established by applying the principle of polar reciprocity to the elliptic quadric. For an elliptic quadric, the polarity associates each tangent line r with a polar line t⊥, which is also a tangent line sharing the same point of contact T. The key property of the elliptic polarity is that it inverts the nature of points and planes when passing to the dual space:
The planes passing through the line t correspond bijectively to the points lying on its polar line t⊥.
A square point on t⊥ corresponds to a non-square plane passing through t.
A non-square point on t⊥ corresponds to a square plane passing through t.
The point of contact T (which lies on the quadric) corresponds to the unique tangent plane .
If the tangent line t is a square tangent line, its polar line t⊥ preserves the same geometric character, meaning it is also a square tangent line.
By applying the dual inversion rule, we can count the planes in the pencil of t based on the points of t⊥:
- ○
The square planes correspond to the non-square points of t⊥. Since there are exactly non-square points on t⊥, there are exactly square planes in the pencil.
- ○
The non-square planes correspond to the square points of t⊥. Since there are square points on t⊥, there are non-square planes in the pencil.
Exactly square planes (and ) non-square planes) pass through a square tangent line.
Exactly square planes pass through a non-square tangent line t. This configuration is the dual counterpart to the previous case, operating under the same rules of elliptic polar reciprocity.
Under the polarity defined by an elliptic quadric, the nature of points and planes is inverted when shifting to the dual space:
Planes passing through the line t correspond bijectively to the points on its polar line t⊥.
A square point on t⊥ maps to a non-square plane through t.
A non-square point on t⊥ maps to a square plane through t.
The contact point T on the quadric maps to the unique tangent plane .
When the tangent line t is a non-square tangent line, its polar line t⊥ is also a non-square tangent line passing through the same contact point T.
We determine the composition of the q + 1 planes in the pencil of t by looking at the types of points on t ⊥ and applying the inversion rule:
The square planes correspond directly to the non-square points on t ⊥. Since t ⊥ contains non-square points, there are exactly square planes passing through t.
The non-square planes correspond to the square points on t ⊥. Since t ⊥ contains square points, there are exactly non-square planes passing through t.
Thus, the square points with the non-square planes define a symmetric partially balanced incomplete block design and the non-square points with the square planes define a symmetric partially balanced incomplete block design. The two designs are isomorphic. □
Since we are interested in the connection between SQS(10) and GQ(2,2), the order q must be equal to three. If q = 3, the proof of Theorem 1 ensures the existence of two isomorphic (153) symmetric configurations: the 15 square points with the 15 polar planes of the non–square points and the 15 non-square points with the 15 polar planes of the non–zero square points. In the next section, by the cyclic structure of the three-dimensional projective space of order three, PG(3,3), we explicitly show the two isomorphic symmetric configurations listed below in two tables.
3. The Connection Between SQS(10) and GQ(2,2)
In this section, we highlight the relationship between the Steiner Quadruple system with ten points and the Cremona–Richmond configuration by the incidence properties of the elliptic quadric of PG(3,3). To write the cyclic structure of PG(3,3), let
w be a primitive element of
over
and let
p (
x) =
x4 +
x − 1 be its minimal polynomial over
. The polynomial
p (
x) =
x4 +
x − 1 is primitive on
. Its companion matrix
C (
p) is
. Let us consider the point
We get:
w1 =
w0 C (
p) =
; by continuing in this way and by denoting the points represented by
wi simply by
i, we obtain the cyclic structure of PG(3,3), as listed in
Table 1. Thus, the Singer group is isomorphic to the additive group Z
40, the integers modulo 40.
Since under the action of a cyclic collineation group of a finite projective space, the point set and the plane set have the same cyclic structure, select any plane; for example, we choose the plane
π0: =
x0 = 0, which contains the 13-set of points listed in
Table 2, where the first entry is the label of the plane label, not a point, and the others 13 entries are the points of the plane. Note that the plane
π0 does not contain the point 0 because the equation of
π0 is
x0 = 0, while the point 0 has homogeneous coordinate
x0 = 1.
The remaining planes of space are found by adding 1 to each point of the preceding plane, beginning with
π0 and using addition modulo 40. Let us consider an elliptic quadric
Q−(3,3) of PG(3,3). The canonical equation of
Q−(3,3) is
. Let us color the points of PG(3,3) with red, blue, and green. Red, if they belong to the elliptic quadric, that is, if the quadratic form evaluates to zero; blue, if the quadratic form evaluates to a non-zero square; and green, if the quadratic form evaluates to a non-square. The colored points are listed in
Table 3.
We get three sets of points:
N = {1, 4, 5, 6, 7, 14, 16, 18, 19, 22, 25, 26, 30, 37, 38}.
Q = {2, 3, 9, 10, 11, 20, 29, 31, 32, 36}.
S = {0, 8, 12, 13, 15, 17, 21, 23, 24, 27, 28, 33, 34, 35, 39}.
The colored points of the planes of PG(3,3) are listed in the rows of
Table 4, where the first entry is the label of the plane label, not a point, and the others 13 entries are the points of the plane.
The elliptic quadric
Q−(3,3) defines a polarity
Let us color the planes of PG(3,3) with red, blue, and green. Red, if the plane is the polar plane of a point belonging to the elliptic quadric; blue, if it is the polar plane of a non-zero square point; and green, if it is the polar plane of a non-square point. The colored planes of PG(3,3) are listed in
Table 5.
Let us consider the green points of the blue planes listed in
Table 6. It is simple to verify that we get a Cremona–Richmond configuration by the direct representation in
Figure 1.
Let us consider the red points of the blue and green planes listed in
Table 7. It is easy to verify that we get a SQS(10).
Now, let us consider the blue points of the green planes listed in
Table 8. It is simple to verify that we get a Cremona–Richmond configuration by the direct representation in
Figure 2.
For the next odd order, q = 5, the proof of Theorem 1 ensures the existence of two isomorphic symmetric partially balanced incomplete block designs: the 65 square points with the 65 polar planes of the non-square points and the 65 non-square points with the 65 polar planes of the non-zero square points. In the next section, by the cyclic structure of the three-dimensional projective space of order five, PG(3,5), we explicitly show the two isomorphic symmetric partially balanced incomplete block designs listed below in two tables.
4. The Connection Between the Steiner System S(3,6,26) and a Symmetric (65,10) Point-Block Incidence Structure
In this section, we highlight the relationship between the Steiner Quadruple system with twenty-six points and two isomorphic symmetric partially balanced incomplete block designs by the incidence properties of the elliptic quadric of PG(3,5). The finite inversive plane of order five, S(3,6,26), is unique, cf. [
18] and [
19]. It consists of twenty-six points and one hundred and thirty blocks, with six points on each block, thirty blocks passing through any point and exactly six blocks through two different points. To write the cyclic structure of PG(3,5), let
w be a primitive element of
over
and let
p (
x) =
x4 +
x2 + 2
x + 2 be its minimal polynomial over
. The polynomial
p (
x) =
x4 +
x2 + 2
x + 2 is primitive on
. The companion matrix
C (
p) is
. Let us consider the point
We get:
w1 =
w0 C (
p) =
; by continuing in this way and by denoting the points represented by
wi simply by
i, we obtain the cyclic structure of PG(3,5), as listed in
Table 9. Thus, the Singer group is isomorphic to the additive group Z
156, the integers modulo 156. Let us consider an elliptic quadric
Q−(3,5) of PG(3,5). The canonical equation of
Q−(3,5) is
. Let us color the points of PG(3,5) with red, blue, and green. Red, if they belong to the elliptic quadric, that is, if the quadratic form evaluates to zero; blue, if the quadratic form evaluates to a non-zero square; and green, if the quadratic form evaluates to a non-square. The colored points are listed in
Table 9.
We get three sets of points:
N = {3, 6, 9, 10, 12, 14, 16, 18, 21, 26, 29, 30, 31, 35, 37, 39, 41, 46, 49, 50, 51, 53, 56, 59, 61, 62, 63, 64, 68, 70, 71, 75, 76, 78, 83, 85, 86, 88, 89, 90, 92, 93, 94, 95, 96, 97, 98, 99, 103, 112, 114, 119, 122, 123, 130, 131, 134, 136, 138, 140, 143, 144, 147, 148, 155}.
Q = {5, 7, 23, 24, 27, 28, 33, 36, 43, 44, 65, 72, 73, 79, 84, 107, 109, 110, 117, 121, 125, 126, 133, 150, 153, 154}.
S = {0, 1, 2, 4, 8, 11, 13, 15, 17, 19, 20, 22, 25, 32, 34, 38, 40, 42, 45, 47, 48, 52, 54, 55, 57, 58, 60, 66, 67, 69, 74, 77, 80, 81, 82, 87, 91, 100, 101, 102, 104, 105, 106, 108, 111, 113, 115, 116, 118, 120, 124, 127, 128, 129, 132, 135, 137, 139, 141, 142, 145, 146, 149, 151, 152}.
Since under the action of a cyclic collineation group of a finite projective space, the point set and the plane set have the same cyclic structure, select any plane; for example, we choose the plane
π0: =
x0 = 0, which contains the 31-set of points listed in
Table 10, where the first entry is the label of the plane label, not a point, and the others 31 entries are the points of the plane. Note that the plane
π0 does not contain the point 0 because the equation of
π0 is
x0 = 0, while the point 0 has homogeneous coordinate
x0 = 1.
The remaining planes of space are found by adding 1 to each point of the preceding plane, beginning with
π0 and using addition modulo 156. The colored points of the planes of PG(3,5) are listed in the rows of
Table A1,
Appendix A, where the first entry is the label of the plane label, not a point, and the others 31 entries are the points of the plane. Let us consider the green points of the blue planes listed in
Table A1. A direct check shows that we get a symmetric partially balanced incomplete block design with parameters
v =
b = 65
r =
k = 10,
m = 3,
λ1 = 0,
λ2 = 2,
λ3 = 3,
n1 = 24,
n2 = 30,
n3 = 10,
,
, and
, whose points and blocks are listed in the rows of
Table 11.
Let us consider the red points of the blue and green planes listed in
Table A1. It is easy to verify, by checking
Table 12, that the finite inversive plane of order five, S(3,6,26), is obtained.
Let us consider the blue points of the green planes listed in
Table A1. A direct check shows that we get a symmetric partially balanced incomplete block design with parameters
v =
b = 65
r =
k = 10,
m = 3,
λ1 = 0,
λ2 = 2,
λ3 = 3,
n1 = 24,
n2 = 30,
n3 = 10,
,
, and
, whose points and blocks are listed in the rows of
Table 13.