Quantum Eigenvalue Transformations for Arbitrary Matrices
Xabier Gutiérrez, Lorenzo Laneve, Mikel Sanz · 2026
Read Quantum Eigenvalue Transformations for Arbitrary Matrices by Xabier Gutiérrez, Lorenzo Laneve, Mikel Sanz (2026) 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-2604-19688.txt.
Introduction
Gutiérrez, Laneve, and Sanz extend quantum signal processing to general square matrices, closing a gap in quantum algorithm design. Signal processing handles only unitary matrices through eigenvalue polynomials, while the singular value method covers Hermitian matrices alone, so neither reaches a non-Hermitian matrix that cannot be diagonalized. Their method acts on eigenvalues, giving a simple route to transform matrices of the general kind.
Catalog record
| CREATOR | Xabier Gutiérrez, Lorenzo Laneve, Mikel Sanz |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 29954 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.19688
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Quantum Eigenvalue Transformations for Arbitrary Matrices Xabier Gutiérrez, Lorenzo Laneve, Mikel Sanz Abstract Quantum Signal Processing (QSP) and Quantum Singular Value Transformation (QSVT) provide an efficient framework for implementing polynomials of block-encoded matrices, and thus offer a systematic approach to quantum algorithm design. However, despite a number of recent advances, important limitations remain. In particular, QSP can only transform unitary matrices, by applying a polynomial to their eigenvalues, while QSVT is a singular-value transformation and thus one can only obtain the polynomial of Hermitian matrices. As a consequence, these techniques do not directly apply to an arbitrary non-Hermitian matrix that is not diagonalizable. In this work, we propose a simple yet powerful method to extend these ideas to arbitrary square matrices by acting on their eigenvalues. To this end, we introduce the notion of an n n -regular block encoding, namely, a block encoding whose k k -th power reproduces the k k -th power of the encoded matrix for every 0 ≤ k ≤ n 0\leq k\leq n . We show that applying QSP to any unitary with this property is equivalent to applying a polynomial of degree at most n n to the block-encoded matrix, independently of its internal structure. Moreover, we provide a simple construction that transforms any block encoding into an n n -regular one using only 𝒪 ( log n ) \mathcal{O}(\log n) ancillary qubits and operations. Finally, we show that this construction induces the desired transformation on the eigenvalues associated with the Jordan normal form of the matrix. Introduction— Quantum algorithms have the potential to deliver substantial speedups over their classical counterparts, yet until very recently their design had no truly systematic framework and instead relied largely on case-by-case ingenuity. In this context, the method of quantum signal processing (QSP) 20 ; 25 has emerged as a unifying paradigm, linking algorithm design with techniques from numerical analysis and, in particular, with the implementation of polynomials of matrices encoded through unitary transformations. This perspective enabled a more abstract and systematic approach to algorithm design, and it has already been applied in a variety of settings, including Hamiltonian simulation 20 ; 21 , quantum search 25 , quantum linear systems 5 , and quantum state preparation 18 ; 9 ; 26 ; 16 . In QSP, one has access to a signal z —typically a complex number of unit modulus or, depending on the convention, a real number in [-1,1] —encoded in a single-qubit unitary known as the signal operator. By interleaving applications of this operator with a sequence of operations that is independent of z —the processing operators—one can construct essentially any polynomial transformation P(z)=\sum_{k=0}^{n}a_{k}z^{k} of the signal, which may, for instance, appear in the upper-left entry of the resulting unitary 24 ; 11 ; 27 . Moreover, efficient and numerically stable methods are known for constructing a single-qubit protocol realizing a prescribed polynomial P 2 ; 1 ; 17 ; 28 . By leveraging the spectral theorem, this construction can be lifted so that the scalar signal z is replaced by a unitary operator U 27 . The resulting circuit then implements the non-unitary transformation P(U)=\sum_{k=0}^{n}a_{k}U^{k} , thereby applying P directly to the eigenvalues of U in a black-box fashion, that is, without the need to explicitly construct or estimate the corresponding eigenstates. Remarkably, this requires only n+1 single-qubit operations and n controlled uses of U . On the other hand, when a matrix A , not necessarily square, appears in the upper-left block of a unitary U_{A} that we can implement—a so-called block-encoding —one can similarly apply a QSP polynomial to its singular values by a closely related construction, known as quantum singular value transformation (QSVT) 11 ; 29 . This method has received particular attention because of its practical usefulness. It is important, however, to note that a polynomial of a square matrix does not, in general, coincide with the polynomial transformation of its singular values. Despite their rapid development, these techniques still suffer from important limitations. In particular, the existing frameworks can perform either an eigenvalue transformation of a unitary matrix, or a singular value transformation of a block-encoded matrix A , but they do not directly provide an eigenvalue transformation of a block-encoded matrix, except in the Hermitian case, where singular values and eigenvalues coincide up to sign. Consequently, for a general non-Hermitian matrix, neither method can be used directly to implement its polynomial transformation. Several approaches beyond the QSP framework have been proposed to partially bridge this gap, especially the construction of Chebyshev and Faber history states 23 , and the method of linear combination of Hamiltonian simulations 4 , which has been shown to be useful for the simulation of non-unitary dynamics 12 ; 22 . In particular, the latter approach has been used to block-encode eigenvalue transformations by expressing them as inverse Laplace transforms 3 . However, both methods rely on the linear combination of unitaries (LCU) technique 6 , which predates QSP but exhibits several limitations. First, the complexity of implementing a target transformation f(z) depends strongly on its specific form and typically requires a non-trivial synthesis of the corresponding quantum circuit. Moreover, a linear combination of N terms generally requires an N -dimensional register, so the qubit overhead can grow rapidly, especially in the aforementioned approaches, where LCU is used to approximate complicated integrals to a prescribed precision. Another drawback of LCU-based methods is that the resulting block-encoded matrices can suffer from severe subnormalization—that is, instead of obtaining a block-encoding of A , one obtains a block-encoding of A/\alpha for some potentially very large \alpha . This leads to small success probabilities and, consequently, to further overhead in the overall complexity. In this Article, we propose a method to carry out transformations of eigenvalues, understood as solutions to the characteristic polynomials of a matrix, of an arbitrary square matrix, even non-diagonalizable ones. In order to do this, we introduce the concept of n - regular block-encoding as a block encoding whose k -th power reproduces the k -th power of the encoded matrix for every 0\leq k\leq n , and prove that this kind of block-encoding is always possible by adding only \mathcal{O}(\log n) ancilla qubits and \mathcal{O}(\mathrm{poly}\log n) extra operations. We then show that a standard application of QSP yields the eigenvalue transformation when this property holds. This allows us to obtain matrix transformations by remaining in this established framework. This technique extends the QSP formalism to several applications, such as simulation of general inhomogeneous differential equations and non-unitary dynamics, which arise in a broad range of settings, including dissipative dynamics treated via Carleman linearization 19 ; 15 , partial differential equations 7 , and open quantum systems 31 , among others. Block-encodings and (generalized) quantum signal processing — Here we summarize the main definitions and results that we are going to use in the rest of the work. Let us start by clarifying the notation: we will use \mathbb{D} to denote the complex unit disk \mathbb{D}=\quantity{ z \in\C: \abs{z} < 1 } , while \mathbb{T}=\partial\mathbb{D} will be the unit circle. Additionally, if not said otherwise, \norm{\cdot} refers to the operator norm. Let us now provide the definition of a block encoding: