Entrywise transforms preserving matrix positivity and non-positivity
Dominique Guillot, Himanshu Gupta, Prateek Kumar Vishwakarma, Chi Hoi Yip · 2025
Read Entrywise transforms preserving matrix positivity and non-positivity by Dominique Guillot, Himanshu Gupta, Prateek Kumar Vishwakarma, Chi Hoi Yip (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-06076.txt.
Introduction
Dominique Guillot and co-authors classify the functions that preserve a matrix's positivity when applied entry by entry. Feed a positive definite matrix through such a function, cell by cell, and the result must be positive definite exactly when the input was. Earlier classics solved the question when all sizes are allowed at once; this work fixes the size, the harder case.
For matrices of size three or more, the answer is stark: only positive multiples of the field's own continuous symmetries qualify. Size two breaks the pattern, allowing power functions to sneak in, which the team charts over wide complex domains. Extensions cover matrices with forced zeros and a cousin problem about monotone maps, rounding out a complete census of these cell-wise guardians.
Catalog record
| CREATOR | Dominique Guillot, Himanshu Gupta, Prateek Kumar Vishwakarma, Chi Hoi Yip |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 88 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-06076
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Entrywise transforms preserving matrix positivity and non-positivity Dominique Guillot, Himanshu Gupta, Prateek Kumar Vishwakarma, Chi Hoi Yip Abstract . We characterize real and complex functions which, when applied entrywise to square matrices, yield a positive definite matrix if and only if the original matrix is positive definite. We refer to these transformations as sign preservers . Compared to classical work on entrywise preservers of Schoenberg and others, we completely resolve this problem in the harder fixed dimensional setting, extending a similar recent classification of sign preservers obtained for matrices over finite fields. When the matrix dimension is fixed and at least 3 3 , we show that the sign preservers are precisely the positive scalar multiples of the continuous automorphisms of the underlying field. This is in contrast to the 2 × 2 2\times 2 case where the sign preservers are extensions of power functions. These results are built on our classification of 2 × 2 2\times 2 entrywise positivity preservers over broader complex domains. Our results yield a complementary connection with a work of Belton, Guillot, Khare, and Putinar (2023) on negativity-preserving transforms. We also extend our sign preserver results to matrices with a structure of zeros, as studied by Guillot, Khare, and Rajaratnam for the entrywise positivity preserver problem. Finally, in the spirit of sign preservers, we address a natural extension to monotone maps, classically studied by Loewner and many others. 2020 Mathematics Subject Classification 15B48, 47B49 (primary), 15B57, 26A48, 05C50 (secondary) 1. Introduction As we commemorate the 100th anniversary of the seminal observation by Pólya and Szegő in 1925, that every absolutely monotonic function preserves positive semidefiniteness when applied entrywise to real matrices, we address a different facet of the century-old classification problem in fixed dimensions, rooted in celebrated theorems of Schoenberg [ 23 ] , Rudin [ 22 ] , Herz [ 14 ] , and others. This story begins in 1911 with the fundamental Schur product theorem [ 24 ] : for an integer n\geq 1 , and positive semidefinite A=(a_{ij}) and B=(b_{ij})\in\mathbb{C}^{n\times n} , the entrywise product A\circ B:=(a_{ij}b_{ij}) is positive semidefinite. The class of positive semidefinite matrices forms a closed convex cone, which is further closed under taking entrywise conjugates. Thus if A=(a_{ij}) is positive semidefinite, then so is the entrywise product A^{\circ m}\circ\overline{A}^{\circ k}:=(a_{ij}^{m}\overline{a_{ij}}^{k}) for integers m,k\geq 0 , where we set 0^{0}:=1 . Writing f[A]:=(f(a_{ij})) for the entrywise action of a function f:\mathbb{C}\to\mathbb{C} on a matrix A=(a_{ij}) , we immediately conclude that the monomial f(z)=z^{m}\overline{z}^{k} preserves positive semidefiniteness when applied entrywise. Note that this holds independently of the dimension of the matrix A . Clearly, the same holds for nonnegative combinations of such monomials, as well as appropriate limits of these polynomials. This is the observation of Pólya and Szegő, who asked in their famous book [ 20 ] if other functions have the same property. Lemma 1.1 (Pólya and Szegő [ 20 ] , Rudin [ 22 ] ) . Let \Omega\subseteq\mathbb{C} , and let f:\Omega\to\mathbb{C} be given by f(z)\equiv\sum_{m,k\geq 0}c_{m,k}z^{m}\overline{z}^{k}, where c_{m,k}\in[0,\infty) for all integers m,k\geq 0 . Then f[A]:=(f(a_{ij})) is positive semidefinite for all positive semidefinite A\in\Omega^{n\times n}, and all n\geq 1. In the special case where \Omega\subseteq\mathbb{R}, the corresponding set of functions identified above is the set of convergent power series with nonnegative coefficients, also known as absolutely monotonic functions. Pólya and Szegő’s work naturally led to the search for real functions that are not absolutely monotonic, but preserve positivity for all positive semidefinite matrices of all sizes. The problem was resolved 17 years later, through Schoenberg’s work and later refined by many, revealing — quite surprisingly — that no other such positivity preservers exist. To state the result formally, we introduce the following notation. For \Omega\subseteq\mathbb{C} , let \mathbb{P}_{n}(\Omega) denote the set of all n\times n Hermitian positive semidefinite matrices with entries in \Omega , and let \mathbb{P}_{n}:=\mathbb{P}_{n}(\mathbb{C}) . We also define D(a,r):=\{z\in\mathbb{C}:|z-a|<\rho\} , the complex open disk centered at a and of radius \rho . Theorem 1.2 (Schoenberg [ 24 ] , Rudin [ 22 ] , Herz [ 14 ] , Vasudeva [ 25 ] , [ 18 , Chapter 18] ) . Let 0<\rho\leq\infty . and let \Omega=(-\rho,\rho),[0,\rho),(0,\rho) , or D(0,\rho) . Let f:\Omega\to\mathbb{F}, where \mathbb{F}=\mathbb{C} if \Omega\not\subseteq\mathbb{R} and \mathbb{F}=\mathbb{R} otherwise. Then the following are equivalent: (1) f[A]:=(f(a_{ij}))\in\mathbb{P}_{n} for all A\in\mathbb{P}_{n}(\Omega), and all n\geq 1. (2) f(z)\equiv\sum_{m,k\geq 0}c_{m,k}z^{m}\overline{z}^{k}, where c_{m,k}\geq 0 for all integers m,k\geq 0. This result has sparked numerous classical and modern developments, with applications in areas as diverse as metric geometry and high-dimensional covariance estimation. See [ 2 , 3 ] for details. The dimension-free aspect, i.e., the independence from the dimension n of the preservers in Theorem 1.2 naturally leads to the problem of classifying positivity preservers f over positive semidefinite matrices of a fixed dimension n . Other than the n=2 case resolved by Vasudeva [ 25 ] , this fixed-dimensional problem — even after 100 years since Pólya and Szegő’s observation — largely remains open. We refer to recent works of Khare and Tao [ 19 ] and of Belton, Guillot, Khare, and Putinar [ 1 ] for developments involving symmetric function theory. In the same vein of fixed-dimensional results, an impactful classical result of FitzGerald and Horn is worth mentioning. Theorem 1.3 (FitzGerald and Horn [ 7 ] ; see also [ 10 , 17 ] ) . Fix an integer n\geq 2 and \beta\in\mathbb{R} . Then \displaystyle A^{\circ\beta}:=(a_{ij}^{\beta})\in\mathbb{P}_{n}\quad\mbox{for all}\quad A=(a_{ij})\in\mathbb{P}_{n}([0,\infty)) if and only if \beta\in\mathbb{N}\cup[n-2,\infty) . Moreover, if \beta\in(0,n-2)\setminus\mathbb{N} then there exists \epsilon>0 such that \begin{pmatrix}(1+\epsilon ij)^{\beta}\end{pmatrix}_{i,j=1}^{n} is not positive semidefinite. In a different direction, following the work of Cooper, Hanna, and Whitlatch [ 5 ] , Guillot, Gupta, Vishwakarma, and Yip [ 8 ] recently considered positive definite matrices over finite fields \mathbb{F}_{q} . By analogy from \mathbb{R} , we call an element in \mathbb{F}_{q} positive if it is the square of a non-zero element in \mathbb{F}_{q} ; and we call an n\times n symmetric matrix with entries in \mathbb{F}_{q} positive definite if all its leading principal minors are positive in \mathbb{F}_{q} . Guillot, Gupta, Vishwakarma, and Yip [ 8 ] classified the entrywise positivity preservers for matrices over finite fields in the harder fixed-dimensional setting, specifically for matrices of a fixed dimension n\geq 3 1 1 1 In the n=2 case, the preservers are more field dependent and some cases of the problem remain open. . Surprisingly, these are precisely the positive multiples of the field’s automorphisms (i.e., powers of the Frobenius). To formalize this, henceforth, we denote the set of n\times n matrices with entries in a set S by M_{n}(S) . Theorem 1.4 ( [ 8 , Corollary 1.4] ) . For a finite field \mathbb{F}_{q} and a fixed n\geq 3 , the entrywise positivity preservers on M_{n}(\mathbb{F}_{q}) are precisely the positive multiples of the automorphisms of \mathbb{F}_{q} . To the authors’ knowledge, this is the only setting in which the fixed-dimensional entrywise positivity preserver problem has been fully resolved. Remarkably, in this case, the entrywise positivity preservers exhibit an even stronger property: a symmetric matrix A is positive definite if and only if f[A] is positive definite. We call such functions sign preservers . The characterization in Theorem 1.4 reveals