The connection between quadratic bent-negabent functions and the Kerdock code (Q2280306)

A Boolean function of \(n\) variables \(f: \mathrm{GF}(2)^n\to \mathrm{GF}(2)\) is \textit{bent} if the absolute value of its Hadamard transform is the constant \(2^{n/2}\). Recall the Hadamard transform is defined by \[ H_f(u) =\sum_{x \in \mathrm{GF}(2)^n}(-1)^{f(x)+ \langle u,x\rangle}. \] To define negabent, define the nega-Hadamard transform by \[ N_f(u) =2^{-n/2}\sum_{x \in \mathrm{GF}(2)^n}i^{wt(x)}(-1)^{f(x)+ \langle u,x\rangle}, \] where \(i=\sqrt{-1}\) and \(wt(x)\) is the Hamming weight of the vector \(x\in \mathrm{GF}(2)^n\). The function \(f\) is \textit{negabent} if \(|N_f(u)|=1\) for all \(u\in \mathrm{GF}(2)^n\). If \(f\) is both bent and negabent then it is called \textit{bent-negabent}. From this reviewer's perspective, the paper under review contains three main results. In Corollaries 15 and 16, the authors give an explicit formula for the number of quadratic homogeneous bent-negabent functions. In their Theorem 5, the authors characterize those functions in the Kerdock code that are bent-negabent. Finally, in Theorem 20 the authors construct an explicit family of subgroups of \(\mathrm{GL}(n,\mathrm{GF}(2))\), each containing \(\mathrm{SL}(n)\), whose action preserves bent-negabent functions. These interesting results are too technical to state precisely in this short review, so we refer the reader to the paper for details.