TCSLib

113 Error-Correcting Codes — MRRW Bound

113.1 Overview

This chapter formalizes the binary first McEliece–Rodemich–Rumsey–Welch (MRRW) bound on the asymptotic rate of binary codes: for every \(\delta \in [0,1/2]\),

\[ R(\delta ) \; \le \; H\! \left(\tfrac 12 - \sqrt{\delta (1-\delta )}\right), \]

where \(R\) is the asymptotic binary code rate and \(H\) is the binary entropy function. The development supplies the maximum code size \(A(n,d)\), the binary Krawtchouk polynomials (both as an explicit alternating sum and as genuine polynomials over \(\mathbb {R}\) defined by the three-term recurrence), Delsarte’s linear-programming bound, the truncated Christoffel–Darboux kernel \(\Phi _{n,t}\) together with its feasibility properties, and the asymptotic ingredients needed to assemble the main theorem.

113.2 Declarations

113.2.1 Statement and normalization

Definition 113.1 Maximum size of a binary code with given distance
#

For naturals \(n\) and \(d\), \(A(n,d)\) is the supremum of the cardinalities \(k\) for which there is a finite set \(C\) of binary codewords \(\mathrm{Fin}\, n \to \mathrm{Bool}\) with \(|C| = k\) and such that any two distinct \(x, y \in C\) differ in at least \(d\) coordinates.

Definition 113.2 Binary entropy in base \(2\)
#

The binary entropy function

\[ H(x) \; =\; -x\log _2 x - (1-x)\log _2(1-x), \]

using the base-\(2\) logarithm, with the convention \(0\log _2 0 = 0\) inherited from \(\log _2 0 = 0\).

Definition 113.3 Asymptotic binary code rate
#

For \(\delta \in [0,1]\), the asymptotic rate is

\[ R(\delta ) \; =\; \limsup _{n\to \infty } \frac{1}{n}\log _2 A(n, \lfloor \delta n \rfloor ). \]

113.2.2 Krawtchouk polynomials

Definition 113.4 Binary Krawtchouk polynomial at integer arguments
#

For naturals \(n\), \(j\), \(x\),

\[ K_j^{(n)}(x) \; =\; \sum _{i=0}^{j} (-1)^i \binom {x}{i}\binom {n-x}{j-i}, \]

where the subtraction \(n-x\) is natural-number subtraction; the definition gives the intended values for \(x \le n\).

Definition 113.5 Krawtchouk polynomials over \(\mathbb {R}\)
#

The family of real polynomials defined by the three-term recurrence \(K_0 = 1\), \(K_1 = n - 2X\), and

\[ (j+2)\, K_{j+2} \; =\; (n - 2X)\, K_{j+1} - (n-j)\, K_j, \]

extending the integer-argument definition to real arguments.

Theorem 113.6 Value of the zeroth Krawtchouk polynomial
#

For all \(n\) and \(x\), \(K_0^{(n)}(x) = 1\).

Theorem 113.7 Krawtchouk polynomial at \(x = 0\)
#

For \(j \le n\), \(K_j^{(n)}(0) = \binom {n}{j}\).

Theorem 113.8 First Krawtchouk polynomial
#

For \(x \le n\), \(K_1^{(n)}(x) = n - 2x\).

Theorem 113.9 Krawtchouk generating function
#

For \(x \le n\) and every real \(z\),

\[ \sum _{j=0}^{n} K_j^{(n)}(x)\, z^{j} \; =\; (1-z)^{x}(1+z)^{\, n-x}. \]
Theorem 113.10 Orthogonality of Krawtchouk polynomials
#

For \(r \le n\) and \(s \le n\),

\[ \sum _{x=0}^{n} \binom {n}{x} K_r^{(n)}(x)\, K_s^{(n)}(x) \; =\; \begin{cases} 2^{n}\binom {n}{r}, & r = s,\\ 0, & r \ne s. \end{cases} \]
Theorem 113.11 Three-term recurrence for Krawtchouk polynomials
#

For \(1 \le j\), \(j + 1 \le n\) and \(x \le n\),

\[ (j+1)\, K_{j+1}^{(n)}(x) \; =\; (n - 2x)\, K_j^{(n)}(x) - (n - j + 1)\, K_{j-1}^{(n)}(x). \]
Theorem 113.12 Agreement of the sum formula and the polynomial form

For \(j \le n\) and \(x \le n\), evaluating the polynomial \(K_j^{(n)}\) at the real number \(x\) gives the value \(K_j^{(n)}(x)\) of the alternating-sum definition.

113.2.3 Delsarte’s linear programming bound

Theorem 113.13 Delsarte LP inequality
#

Let \(F(x) = \sum _{j=0}^{n} F_j K_j^{(n)}(x)\) for real coefficients \(F_j\). If \(F_0 \gt 0\), if \(F_j \ge 0\) for all \(1 \le j \le n\), and if \(F(x) \le 0\) for every integer \(x\) with \(d \le x \le n\), then

\[ A(n,d) \; \le \; \frac{F(0)}{F_0}. \]

113.2.4 Christoffel–Darboux kernel and feasibility

Definition 113.14 Truncated Christoffel–Darboux kernel
#

For \(t \le n\) and real \(a, x\), the truncated reproducing kernel is

\[ \Phi _{n,t}(a,x) \; =\; \sum _{j=0}^{t} \frac{K_j^{(n)}(a)\, K_j^{(n)}(x)}{\binom {n}{j}}. \]
Theorem 113.15 Christoffel–Darboux identity for Krawtchouk polynomials
#

For \(t + 1 \le n\) and real \(a \ne x\) there is a scalar \(c \gt 0\) with

\[ \Phi _{n,t}(a,x) \; =\; c\, \frac{K_t^{(n)}(a)K_{t+1}^{(n)}(x) - K_{t+1}^{(n)}(a)K_t^{(n)}(x)}{a - x}. \]
Theorem 113.16 Positivity of Krawtchouk values below all zeros

Fix \(t + 1 \le n\) and a real \(a\) that is strictly smaller than every real zero \(\xi \) of \(K_t^{(n)}\). Then \(K_j^{(n)}(a) \gt 0\) for every \(j \le t\).

Theorem 113.17 Nonpositivity of the kernel beyond a threshold

Under the same hypotheses (\(t + 1 \le n\) and \(a\) strictly below every real zero of \(K_t^{(n)}\)), there exists a threshold \(\eta \) such that \(\Phi _{n,t}(a,x) \le 0\) for every integer \(x\) with \(\eta \le x \le n\).

Theorem 113.18 Finite-\(n\) MRRW bound

Let \(t + 1 \le n\), let \(a\) be strictly below every real zero of \(K_t^{(n)}\), and suppose \(\Phi _{n,t}(a,x) \le 0\) for every integer \(x\) with \(d \le x \le n\). Then

\[ A(n,d) \; \le \; \Phi _{n,t}(a,0). \]

113.2.5 Asymptotic choice of parameters

Theorem 113.19 Asymptotics of the smallest Krawtchouk zero

Let \(0 \lt \tau \lt 1/2\) and let \(t_n\) satisfy \(t_n / n \to \tau \). Then there is a sequence \(a_n\) such that, eventually in \(n\), \(a_n\) lies strictly below every real zero of \(K_{t_n}^{(n)}\), with \(a_n/n \to \tfrac 12 - \sqrt{\tau (1-\tau )}\), and such that eventually \(\Phi _{n,t_n}(a_n,x) \le 0\) for all integers \(x\) with \(\lfloor (\tfrac 12 - \sqrt{\tau (1-\tau )})n \rfloor \le x \le n\).

Theorem 113.20 Entropy growth rate of the objective

Let \(0 \lt \tau \lt 1/2\), let \(t_n/n \to \tau \), and let \(a_n\) eventually lie strictly below every real zero of \(K_{t_n}^{(n)}\). Then

\[ \limsup _{n\to \infty } \frac{1}{n}\log _2 \Phi _{n,t_n}(a_n,0) \; \le \; H(\tau ). \]
Theorem 113.21 Entropy asymptotics for binomial tails
#

For \(0 \le \tau \le 1/2\),

\[ \frac{1}{n}\log _2\! \left(\sum _{j=0}^{\lfloor \tau n\rfloor } \binom {n}{j}\right) \; \longrightarrow \; H(\tau ) \qquad (n \to \infty ). \]

113.2.6 Conclusion of the MRRW bound

Theorem 113.22 Involution between \(\delta \) and \(\tau \)
#

For \(0 \le \delta \le 1/2\), setting \(\tau = \tfrac 12 - \sqrt{\delta (1-\delta )}\) gives back \(\delta = \tfrac 12 - \sqrt{\tau (1-\tau )}\); that is, the map is an involution on \([0,1/2]\).

Theorem 113.23 Binary entropy at \(0\)
#

\(H(0) = 0\).

Theorem 113.24 Binary entropy at \(1/2\)
#

\(H(1/2) = 1\).

Theorem 113.25 Binary first MRRW bound
#

For every \(\delta \) with \(0 \le \delta \le 1/2\),

\[ R(\delta ) \; \le \; H\! \left(\tfrac 12 - \sqrt{\delta (1-\delta )}\right). \]