On certain blocking sets and the minimum weight of the code of generalised polygons

The code \mathfrak{C}_\mathbb{F}(\Gamma) is a subspace of the code \mathfrak{C}_\mathbb{F}(\Gamma'), where \Gamma' is the geometry of points and lines from \mathrm{PG}(k,q). It follows from Theorem 4 that q+1 is a lower bound for the minimum weight of \mathfrak{C}_\mathbb{F}(\Gamma), and the incidence vector of any line of \Gamma shows equality.

We clearly have that

Now assume that there is an element z in \Gamma_{\le d_2}(y_1)\cap\Gamma_{\le d_2}(y_2) with z\not\in \Gamma_{\le d_2-d_1}(x). Then, combining the paths from z to y_1, from y_1 to y_2 and from y_2 back to z, we find a cycle of length at most d_2 + 2d_1+ d_2 \le 4m-2d_1-2+2d_1 = 4m-2. This contradicts Theorem 1 which states that the incidence graph of a 2m-gon has girth 4m.

1. This follows from Lemma 1 with d_1 = 1.

2. 1. We have that \delta(L,M_1)=\delta(L,M_2)=2 and \delta(M_1,M_2)=4. Lemma 1 with d_1=2 and d_2=2r-1 shows that

      \Gamma_{\leq 2r-3 }(L)=\Gamma_{\leq 2r-1}(M_1)\cap \Gamma_{\leq 2r-1}(M_2).

      Restricting these sets to their intersection with \mathcal{P} yields the desired result.

   2. This follows from noting that \delta(v_1,v_2) is even and restricting the equality found in part 1 to the point set \mathcal{P}.

1. We prove this by induction on d. The base case d=1 follows by taking M=L. Assume that the statement holds for some k\in \mathbb{N} with 1\le k < m-1 and consider the case d=k+1. Let L\in\mathcal{L} be an arbitrary line of \Gamma. Then there exists an M\in \mathcal{L}_{\le 2k-2}(L) such that C\cap \mathcal{P}_{\le2k-1}(M) = C\cap \mathcal{P}_{1}(L).

   We label the s+1 points on M as \{v_1,v_2,\dots,v_{s+1}\}. For each of these points v_i we can consider a line N_i through v_i and different from M. Note that the lines N_1,\ldots,N_{s+1} are mutually disjoint.

   By Corollary 5(2a), we now observe that for any two of these lines N_i and N_j with i\neq j, we have

   \mathcal{P}_{\le 2(k+1)-1}(N_i)\cap \mathcal{P}_{\le 2(k+1)-1}(N_j)= \mathcal{P}_{\le 2(k+1)-3}(M).

   Since |C|<s+1, at least one of the s+1 disjoint sets \mathcal{P}_{\le 2(k+1)-1}(N_i)\setminus \mathcal{P}_{\le 2(k+1)-3}(M), i \in \{1,2,\ldots,s+1\}, must have an empty intersection with C. Without loss of generality, we may assume

   (\mathcal{P}_{\le 2(k+1)-1}(N_1)\setminus \mathcal{P}_{\le 2(k+1)-3}(M)) \cap C = \emptyset.

   This means that

   C\cap \mathcal{P}_{\le2(k+1)-1}(N_1) = C\cap \mathcal{P}_{\le2(k+1)-3}(M) = C\cap \mathcal{P}_{1}(L),

   where we have invoked the induction hypothesis.

   Since M\in \mathcal{L}_{2k-2}(L) and N_1\in\mathcal{L}_{2}(M), we see that N_1\in \mathcal{L}_{2k-2+2}(L)=\mathcal{L}_{2k}(L)=\mathcal{L}_{2(k+1)-2}(L). The statement follows.

2. Let L\in\mathcal{L} be an arbitrary line of \Gamma. From part (i), we know that there exists a line M\in \mathcal{L}_{\le 2d-2}(L) such that C\cap \mathcal{P}_{\le2d-1}(M) = C\cap \mathcal{P}_{1}(L).

   We label the s+1 points on M as \{v_1,v_2,\dots,v_{s+1}\}.

   By Corollary 5(2b), we have that for i\neq j,

   \mathcal{P}_{\le 2d}(v_i)\cap \mathcal{P}_{\le 2d}(v_j)= \mathcal{P}_{\le 2d-1}(M).

   Since |C|<s+1 points, at least one of the disjoint sets \mathcal{P}_{\le 2d}(v_i)\setminus \mathcal{P}_{\le 2d-1}(M) must have an empty intersection with C. Without loss of generality, we may assume

   (\mathcal{P}_{\le 2d}(v_1)\setminus \mathcal{P}_{\le 2d-1}(M)) \cap C = \emptyset.

   This means that

   C\cap \mathcal{P}_{\le2d}(v_1) = C\cap \mathcal{P}_{\le2d-1}(M) = C\cap \mathcal{P}_{1}(L)

   Since v_1\in \mathcal{P}_1(M) and M\subseteq \mathcal{L}_{\leq 2d-2}(L), we see that v_1\in \mathcal{P}_{1}(M) \subseteq \mathcal{P}_{\leq 2d-1}(L) and the statement follows.

By Corollary 2, L contains a unique point w closest to v and we know that \delta(v,w) \in \{0,2,\dots, 2m-2\}. If \delta(v,w) < 2m-2 then clearly all points on L are in \mathcal{P}_{\leq 2m-2}(v). If \delta(v,w) = 2m-2 then all points on L different from w are at distance 2m from v and therefore not in \mathcal{P}_{\leq 2m-2}(v).

We double count the following set:

Each point in B lies on t+1 lines so we find |X|=|B|(t+1). Because there are |\mathcal{L}| = (1+t)\frac{s^m t^m-1}{st-1} lines (see for instance [Hendrik]), each of which needs to contain at least one point of B we find |X|\ge(1+t)\frac{s^m t^m-1}{st-1}. Combining these results gives us |B|\ge \frac{s^m t^m-1}{st-1}.

Assume to the contrary that |C|<s+1. Lemma 2 (ii) with d=m-1 tells us that

Since \forall x, C\cap \mathcal{P}_{\le 2m-2}(x)\neq\emptyset , we deduce that

Using Lemma 4, we find that |C|\geq \frac{s^mt^m-1}{st-1}\geq s+1, a contradiction.

Let z_1 and z_2 be two different elements in \Gamma_1(x). By Corollary 4 and 5(1),

Assume to the contrary that

then we find that |C|\ge |\Gamma_1(x)\setminus\{y\}| + 2.

There are either s (when x is a line) or t (when x is a point) elements in \Gamma_1(x), different from y. Hence, since s\le t there are at least s elements in \Gamma_1(x) different from y. We find |C|\ge |\Gamma_1(x)\setminus\{y\}| + 2 \ge s+2, a contradiction since |C|=s+1.

Assume to the contrary that there is an element y in \Gamma_{1}(x) with

We prove the following claim by induction on d':

The base case d' = d-1 follows by considering the point y: since \Gamma_{\le d-2}(y) is a subset of \mathcal{P}_{\le d-1}(x), the statement follows from \mathcal{P}_{\le d-1}(x)\cap C = \emptyset and \mathcal{P}_{d-1}(y)\cap C = \emptyset.

Now assume that the statement holds for d' = k with d-1\le k<2m-2. We find an element z_k in \Gamma_{\le k-d+1}(y) such that

Since z_k\in \Gamma_{\le k-d+1}(y) we also get y \in \Gamma_{\le k-d+1}(z_k). Hence, there is an element y' in \Gamma_{1}(z_k) such that y \in \Gamma_{\le k-d}(y'). This implies that

Therefore, \mathcal{P}_{\le k+1}(y') contains at least two points of C. We are now ready to use Lemma 6 with x=z_k and y=y'. This implies the existence of an element z_{k+1} in \Gamma_1(z_k)\subseteq \Gamma_{\le k-d+2}(y) such that

This proves our claim. By setting d'=2m-2, we now find a contradiction with the assumption that C is an \mathcal{X}-blocking set.

Corollary 5(1) shows that for any two elements y_1 and y_2 in \Gamma_{1}(x)

This implies that

Since \mathcal{P}_{\leq d-2}\cap C=\emptyset we see that for every y\in \Gamma_1(x), the set \mathcal{P}_{d-1}(y)\setminus\mathcal{P}_{d-2}(x) contains a different point of C, and the points obtained in this way are mutually distinct. We find that s+1=|C|\geq |\Gamma_1(x)|. If x is a point, |\Gamma_1(x)|=t+1, and by our assumption, t\geq s, so it follows that in this case s=t. Furthermore, it follows that in both the case where x a point or x is a line, each of the sets \mathcal{P}_{d-1}(y)\setminus\mathcal{P}_{d-2}(x) contains exactly one point of C, and C does not contain any further points. It follows that C\subseteq \mathcal{P}_{\leq d}(x), and since \mathcal{P}_{\leq d-2}(x)\cap C=\emptyset, C\subseteq \mathcal{P}_d(x).

Let z_1\neq z_2 be two points of C such that \delta(z_1,z_2)\leq \delta(u,v) for all u\neq v points of C.

By the axioms of the generalised 2m-gon \Gamma, there exists an ordinary 2m-gon \Gamma' containing z_1 and z_2. Since z_1 and z_2 are both points, the distance between them is even, and we can find opposite elements x and y in \Gamma' such that:

• z_1,z_2 \in \mathcal{P}_d(x)

• z_1,z_2 \in \mathcal{P}_{2m-d}(y)

for d=\frac{\delta(z_1,z_2)}{2}.

If \mathcal{P}_{\le d-1}(x) would contain a point of C, then z_1 and z_2 would not have been at the minimal distance. We can therefore use Lemma 7 and Lemma 8 to conclude that all points of C are contained in \mathcal{P}_{d}(x). Since x and y are opposite, \mathcal{P}_{\le d}(x)\cap \mathcal{P}_{\le 2m-d-1}(y) = \emptyset and we get that \mathcal{P}_{\le 2m-d-1}(y) can not contain any point of C. Again, from lemma 7 and lemma 8, we can conclude that C\subseteq \mathcal{P}_{2m-d}(y). We now have C\subseteq \mathcal{P}_{d}(x)\cap\mathcal{P}_{2m-d}(y) proving the statement.

Let \Gamma'=(\mathcal{P}',\mathcal{L}',I') be the projective plane which is the perp-geometry determined by the projective point x. We need to show that for all v\in \mathcal{P}, \mathcal{P}_{\leq 2m-2}(v)\cap T_{2,x,y}\neq \emptyset. If v is opposite x, then T_{2,x,v} and T_{2,x,y} are lines of \Gamma', and since \Gamma' is a projective plane, these two lines have a point w of \Gamma' in common. Since w is a point of T_{2,x,v}, it follows that w\in \mathcal{P}_{\leq 2m-2}(v)\cap T_{2,x,y}.

If v and x are not opposite, \delta(v,x)\leq 2m-2. The second point of a shortest path from x to v, say r, is contained in one of the lines through x, say L. We have that r is contained in \mathcal{P}_{\leq 2m-4}(v), and since L is a line of \Gamma', it meets T_{2,x,y}, which is a line of \Gamma' too, in a point s of \Gamma'. Since \delta(r,s)=2, and r\in \mathcal{P}_{\leq 2m-4}(v) it follows that s\in \mathcal{P}_{\leq 2m-2}(v)\cap T_{2,x,y}. We conclude that T_{2,x,y} is an \mathcal{X}- blocking set.

Let L be a line of \Gamma. By Corollary 2, there exists a unique point w on L closest to v. Let d = \delta(v,w). Observe that d is even, that d \le 2m-2, and that

If d = 2m-2 then we find

If d<2m-2 then

Let L be a line. Using Lemma 10 we see that

Since \mathfrak{C}_\mathbb{F} is generated by the vectors \mathfrak{i}_L, where L\in \mathcal{L}, the statement follows.

Lemma 11 shows that \langle \mathfrak{c}, \mathfrak{c}_{v} - \mathfrak{c}_w \rangle = 0 for all \mathfrak{c}\in \mathfrak{C}_\mathbb{F}.

By Lemma 5 we know that there exists a point v such that

Since the support of \mathfrak{c}_{v} is contained in \mathcal{P}_{\le 2m-2}(v), the statement follows.

Let w be a point of C and assume that |C|\le t+1. It follows that at least one of the t+1 lines through w, say M, does not contain a second point of C. Our assumption implies that there exists a point v\in\mathcal{P} such that |\text{supp}(\mathfrak{c}_{v}\cdot \mathfrak{i}_M)| = s+1 \text{ and } \langle \mathfrak{c}, \mathfrak{c}_{v}\cdot\mathfrak{i}_M \rangle = 0. Since \langle \mathfrak{c}, \mathfrak{c}_{v}\cdot\mathfrak{i}_M \rangle = \langle \mathfrak{c}, \mathfrak{c}_{v}\cdot\mathfrak{i}_{\{w\}} \rangle = \mathfrak{c}(w)\cdot \mathfrak{c}_{v}(w) where both \mathfrak{c}(w) and \mathfrak{c}_{x}(w) are elements of \mathbb{F} different from 0, this is a contradiction.

By Lemma \ref{starLemma} (ii) with d=m-1 we get:

Therefore,

Which implies that

The set C\setminus\{v\} has size s, so we can invoke Lemma 2 with d=m-1 to deduce that there exists a w\in\mathcal{P}_{\le 2m-3}(L) such that

Because v\in C\cap\mathcal{P}_{\le 2m-2}(w)\cap\mathcal{P}_{1}(L), we get

The statement now follows since

Assume that there exists a point v\in \mathcal{P} such that \langle \mathfrak{c}_{v} ,\mathfrak{c}\rangle = 0. By Corollary 6, we then have

Now let v be a point of C. Since we assume s\le t, and |C\setminus\{v\}| = s, at least one of the lines through v does not contain a second point C. Let L be such a line. By Lemma 15, we find that

This means that

Because \mathbb{F} is a field and both \mathfrak{c}_w(v) and \mathfrak{c}(v) are assumed to be non-zero, this forms a contradiction.

Let \mathfrak{c}\in\mathfrak{C}_\mathbb{F} be a codeword with support C and assume to the contrary that |C|<s +1.

By Lemma 12,

while Corollary 6 implies that

We conclude that

From Lemma 14 we get

Combining this, we find

The condition on \mathbb{F} implies that for all v\in \mathcal{P}, supp(\mathfrak{c}_{v})=\mathcal{P}_{\leq 2m-2}(v). Since all points on a line that lies at distance at most 2m-3 from a point v lie at distance at most 2m-2 of v, we see that all lines at distance at most 2m-3 from v are covered by \mathfrak{c}_{v} i.e.:

Therefore, the conditions for Lemma 13 are fulfilled, which gives us |C|>t+1. This contradicts the assumptions that |C|<s+1 and s\le t.

Since for any line L\in \mathcal{L}, the codeword \mathfrak{i}_L has weight s+1, we have shown that the minimum weight of \mathfrak{C}_\mathbb{F} is s+1.

Now assume that |C|=s+1. The condition on \mathbb{F} shows that we can invoke Lemma 16 to find that C forms an \mathcal{X}-blocking set. Theorem 6 now shows that C is a distance trace T_{d,x,y}. Furthermore, if s<t, then d is odd. If additionally \Gamma is a weak generalised quadrangle, then m=2, so since d\in \{1,\ldots,m\} is odd, only the possibility T_{1,x,y} occurs: the support C is the point set of a line. Now assume that C is the point set of a line L. Let v be a point of L. Consider the vector \mathfrak{c}-\mathfrak{c}(v)\mathfrak{i}_L. This is a linear combination of code words in \mathfrak{C}_\mathbb{F}, so it is a code word of \mathfrak{C}_\mathbb{F}, but its weight is at most s. Therefore, it is the zero vector. It follows that \mathfrak{c}=\mathfrak{c}(v)\mathfrak{i}_L.