betaFeedback

WorksPhysics and Astronomy

End-to-End Quantum Algorithm for Differential Equations: Application to Simulating a Biomolecular System

Abstract

Quantum algorithms for differential equations have the potential for substantial algorithmic speedups over classical methods. However, they rely on certain assumptions and abstract oracle specifications that leave their practical feasibility and computational advantages for realistic applications unclear. Here we formalize the concept of an end-to-end quantum algorithm for differential equations as one that accepts classical inputs and returns classical outputs, and provide implementations for algorithmic primitives in constructing such algorithms. We present a tailored end-to-end quantum algorithm for the linearized Poisson-Boltzmann (LPB) equation, a partial differential equation (PDE) with variable coefficients widely used in simulating a biomolecular system. We explicitly implement all aspects of the algorithm, from preparing an initial state to generating a quantum solution and estimating properties of the system. In particular, we develop a new discrete representation for variable-coefficient PDEs on a grid that enables efficient implementation of the operator oracle, and build on a recent technique based on optimal wavelet preconditioning to generate a quantum solution with a cost independent of the condition number of the discretized operator. We show the algorithm can be implemented by quantum circuits of size polylogarithmic in both the number of grid points NgN_g and the inverse of the desired accuracy ε\varepsilon, with an overall gate cost polylogarithmic in NgN_g and quasilinear in 1/ε1/\varepsilon. We also show the classical preprocessing to construct the circuits has polylogarithmic cost in NgN_g and 1/ε1/\varepsilon. Our work provides a concrete end-to-end quantum algorithm for a PDE of practical value, demonstrates an exponential improvement in scaling with NgN_g over classical methods, and provides a foundation for assessing its potential quantum advantage in practice.

Introduction

Partial differential equations (PDEs) are the most universal descriptions of the natural physical laws and capture various problems in science and engineering. This makes them an ideal case to probe the potential utility of quantum computers for various applications. Many quantum algorithms have therefore been developed for differential equations over the years [1, 2, 3, 4, 5, 6, 7, 8, 9]. However, these algorithms mainly remain at an abstract level. Specifically, quantum PDE solvers typically assume access to certain oracles that specify a discretized version of the PDE, such as an oracle for preparing an appropriate initial state and an oracle for encoding the discretized PDE operator. The cost of the algorithm is then assessed in terms of the number of queries to these oracles. However, implementing these oracles from a classical specification of the PDE can introduce substantial gate costs that are not captured by the query cost.

Furthermore, unlike classical algorithms that explicitly return the solution vector of the discretized PDE, quantum PDE solvers generate a quantum state that encodes an approximation of the solution. It is therefore assumed that physically relevant properties of the solution vector can be extracted from this state, although the cost of such post-processing of the quantum solution can be significant. Moreover, quantum PDE solvers assume that the discretized PDE operator is well conditioned to achieve a polylogarithmic cost in the system size, whereas for typical PDEs, its condition number generally grows polynomially with the system size. These algorithms also commonly assume prior knowledge of additional problem-dependent parameters [9, 10, 11, 12], such as a sufficiently tight upper bound on the norm of the discretized operator, which may be computationally costly to obtain. Considering these assumptions, the practical feasibility and the computational advantage of quantum algorithms for realistic PDEs remain unclear.

In this paper, we formalize the concept of an end-to-end quantum algorithm for differential equations, by which we mean a quantum algorithm that accepts classical inputs and returns classical outputs (Figure 1). To this end, we present a set of broadly applicable algorithmic primitives for realizing the main subroutines of an end-to-end quantum algorithm. In particular, we present explicit quantum circuits for loading a quantum state into a quantum register and estimating properties of a quantum solution, construct implementations for different encodings of differential operators with constant and variable coefficients, and rigorously analyze the computational costs of these primitives. We then develop a tailored end-to-end quantum algorithm for the linearized Poisson-Boltzmann (LPB) equation, a second-order PDE with variable coefficients that is widely used in computational biochemistry to model the electrostatic potential around a biomolecule immersed in an ionic aqueous solution. The electrostatic potential and related quantities obtained from this model are used to study how the surrounding solution affects the stability of biomolecules, such as proteins, and their binding to other molecules, such as potential drugs [13, 14].

Figure 1. Common steps of an end-to-end quantum algorithm. The algorithm accepts classical inputs and returns classical outputs, denoted by double lines. In the first step, an initial state is prepared. The initial state is then evolved to a final state. The final step is to perform a measurement to obtain an expectation value of an observable with respect to the final state. These steps may need to be repeated to obtain multiple samples for more accurate computation of the desired expectation value.

A central contribution of our work is an efficient and fully explicit construction of an end-to-end quantum algorithm directly from a classical specification of the biomolecular system, without assuming access to abstract oracles or prior knowledge of additional problem-dependent parameters commonly taken as inputs to quantum algorithms. To achieve this, we exploit the mathematical and physical structure of the biomolecular system to develop suitable representations and quantum circuits tailored to the problem. In particular, we introduce a new discrete representation of multidimensional differential operators with variable coefficients that enables us to construct the oracle for the discretized PDE operator from simple components, which can be efficiently implemented by directly exploiting the geometry and physical properties of the biomolecular system. Similarly, we exploit the structure of the problem to develop an efficient quantum circuit for preparing an appropriate initial state from the classical specification, without explicitly loading the high-dimensional discretized data. We further identify physically relevant properties of the biomolecular system and construct efficiently implementable observables whose expectation values with respect to a quantum encoding of the solution yield these properties.

The differential operator of the LPB equation is a multidimensional generalization of a Sturm–Liouville operator with variable and bounded coefficients. When discretized, operators of this form yield matrices whose condition numbers scale polynomially with the matrix size NgN_g. However, such operators can be optimally preconditioned using a simple diagonal wavelet preconditioner [8], meaning that the condition number of the preconditioned matrix becomes independent of the matrix size. We therefore build on the quantum PDE algorithm introduced in [8], which uses wavelet preconditioning and a quantum linear-system solver to generate a distributed quantum encoding of the solution for the discretized PDE with a query cost independent of the condition number of the original discretized operator.

This favorable query cost, however, does not by itself guarantee an efficient quantum implementation, as the overall cost also depends on the normalization and gate cost of implementing a block encoding of the preconditioned matrix. The algorithm of [8] assumes access to a suitably normalized block encoding of the discretized PDE operator, without accounting for the normalization and gate cost associated with constructing such an encoding from the classical specification of a particular PDE. A key technical contribution of our work is that we construct an explicit and efficient block encoding of the wavelet-preconditioned matrix needed by this algorithm. To this end, we exploit the multiscale structure of the wavelet-transformed matrices to achieve a block encoding with a logarithmic normalization and polylogarithmic gate cost in the matrix size. Beyond constructing the required oracles, we make the quantum algorithm for generating the distributed quantum solution state explicit by implementing fixed-point amplitude amplification using a known lower bound on the success probability, and provide a detailed analysis of the computational cost and error of this procedure. These constructions enable an efficient gate-level realization of the algorithm and allow the favorable scaling of its query cost to be translated into the overall gate-cost scaling.

In constructing our algorithm for the biomolecular system, we treat the number of grid points NgN_g and the desired algorithmic error ε\varepsilon as independent parameters. Here, ε\varepsilon denotes the error tolerance in estimating a solution property relative to its ideal value for the discretized problem. It accounts for errors in generating a quantum solution state and estimating the expectation value of the corresponding observable, but not the error in spatial discretization of the continuous PDE. This treatment is motivated by the physical structure of the biomolecular system. Specifically, in spatial discretization of such systems, the grid resolution is constrained by the molecular geometry, particularly by the positions and van der Waals radii of the atoms. Consequently, the grid resolution, characterized by the number of grid points NN per spatial dimension, is not determined as a function of the desired accuracy ε\varepsilon, as is done in [9] to control the discretization error for second-order PDEs with constant coefficients. In our application, NN can thus be much larger than 1/ε1/\varepsilon. We therefore treat NgN_g and ε\varepsilon as separate parameters and analyze the computational cost of our algorithm with respect to both parameters.

We rigorously analyze the gate cost and error of the complete end-to-end algorithm using several methods for property estimation. Our analysis reveals different tradeoffs between circuit size, ancilla count, and the number of circuit runs. In particular, depending on the method used for property estimation, the end-to-end algorithm can be implemented using quantum circuits of size polylogarithmic in both NN and 1/ε1/\varepsilon, but estimating a desired property to accuracy ε\varepsilon needs a number of circuit runs that scales quadratically with 1/ε1/\varepsilon. Alternatively, methods using larger circuits can reduce the number of runs and achieve an overall gate cost quasilinear in 1/ε1/\varepsilon, with the circuit size remaining polylogarithmic in NN. We also account for the classical preprocessing needed to construct the quantum circuits for the end-to-end algorithm directly from the classical specification of the biomolecular system. We show that, for a fixed number of atoms, the computational cost of this preprocessing is at most polylogarithmic in both NN and 1/ε1/\varepsilon, and therefore does not eliminate the favorable scaling of the gate cost of the quantum algorithm.

Our results thus demonstrate an exponential improvement in scaling with the number of grid points compared with known classical algorithms that solve the same discretized problem. By providing explicit circuit constructions and accounting for their costs, our work establishes a concrete realization of an end-to-end quantum algorithm for a realistic PDE of practical value and provides a foundation for assessing its potential quantum advantage in practice.

Organization. The remainder of the paper organized as follows. We begin § 2 by presenting a general framework for an end-to-end quantum algorithm for differential equations. We first discuss the classical specification of a general linear PDE and introduce a new discrete representation for variable-coefficient PDEs that is suitable for efficient implementation of various oracles of the discretized operator. We then present several algorithmic primitives, which have applications beyond the scope of this work, and provide a detailed analysis of their computational costs. In particular, we present an explicit approach for loading a quantum state onto a quantum register, construct various implementations for different oracular encodings of matrices obtained from discretizing a differential equation, and discuss several methods for estimating properties of a quantum state.

In § 3, we present our tailored end-to-end quantum algorithm for the LPB equation used in simulating a biomolecular system. We describe this system and develop tailored implementations for preparing the initial state, generating a quantum solution, and estimating properties of interest. We analyze the computational cost of these implementations and provide a detailed cost analysis of the complete end-to-end algorithm. We conclude with a discussion of our results in § 4. Technical proofs and detailed analyses are provided in the appendices.

End-to-end quantum algorithm for differential equations

An end-to-end quantum algorithm for a PDE accepts classical inputs and returns classical outputs. The classical inputs fully specify the PDE together with the desired properties of its solution, and the classical outputs are estimates of those properties (see Figure 1). In § 2.1, we elaborate the full specification of a general linear PDE and the solution properties that serve as classical inputs for an end-to-end quantum PDE solver. We present a new representation for PDEs with variable coefficients in § 2.3 and discuss how this representation provides a simple and efficient implementation for the oracles of the PDE operator. We then use this representation to implement various oracle encodings of operators in § 2.4 and analyze the cost of these implementations. We conclude in § 2.5 by presenting several approaches for extracting properties of the quantum solution state with a rigorous cost analysis.

Specification of the discretized PDE

When discretized, a differential equation together with its boundary conditions, i.e., a boundary-value problem, becomes a finite-dimensional algebraic problem. For a linear PDE Au(x)=b(x)\mathcal{A}u(x)=b(x), the discretization leads to a system of linear equations Au=b\mathcal{A}u=b. The vector uu contains the discrete degrees of freedom used to represent the approximate solution. These degrees of freedom could be, for example, the values of u(x)u(x) at grid points in the finite-difference method, or the unknown expansion coefficients in the spectral or wavelet method, where the solution is expressed as a linear combination of basis functions of a finite-dimensional space. The matrix AA represents the action of the differential operator A\mathcal{A} on this space for the chosen discretization, and the vector bb represents the corresponding discretization of b(x)b(x). The boundary conditions are encoded in the linear system through appropriate modifications to AA and bb, for example by adding boundary rows in AA with the corresponding boundary value in bb, or by wrap-around entries in AA for periodic boundary conditions.

The discretized PDE problem is therefore fully specified by the matrix AA and the vector bb. Quantum algorithms for PDEs often assume quantum encodings of AA and bb are given as input. For the vector bb, the quantum encoding is often a procedure that prepares a quantum state that encodes elements of bb on its amplitudes. More specifically, it is assumed that the quantum algorithm takes as an input a procedure Pb\mathcal{P}_b that prepares the initial state ∣b⟩:=∑jbˉj∣j⟩\lvert b\rangle:= \sum_j \bar{b}_j \lvert j\rangle with the normalized coefficients bˉj:=bj/∥b∥\bar{b}_j := b_j/\lVert b\rVert. That is to say, PbP_b maps the all-zero state to the state ∣b⟩\lvert b\rangle. The underlying assumption for the procedure PbP_b is that it can be constructed by access to a quantum oracle Ob∣j⟩∣0⟩↦∣j⟩∣bˉj⟩O_b\lvert j\rangle\lvert0\rangle\mapsto\lvert j\rangle\lvert\bar{b}_j\rangle that returns coefficients bˉj\bar{b}_j of the initial state ∣b⟩\lvert b\rangle. Having an implementation for this oracle, PbP_b can be constructed, for example, using the Grover method [15] for quantum-state preparation or by the inequality-test method [16], which significantly reduces the quantum arithmetic cost of the Grover method.

The oracle ObO_b or the procedure PbP_b can be regarded as a quantum specification of bb. For an end-to-end algorithm, a classical specification of bb is needed as input for implementing ObO_b and PbP_b. In general, bb needs to be sparse or structured to efficiently prepare the state ∣b⟩\lvert b\rangle, as otherwise the quantum speedup of a full quantum PDE solver may be eliminated by the cost of the initial-state preparation. The classical specification could be the nonzero entries of bb if it is sparse or an efficient description of the structure of bb. For example, if elements of bb follow a Gaussian distribution, then bb is fully specified by the mean and standard deviation of the Gaussian distribution, and these parameters will be the classical specification of bb. In § 3, we discuss the classical inputs that specify the initial state of the LPB equation and provide a tailored implementation for preparing the initial state.

The common quantum oracle encodings for the matrix AA are the sparse encoding and the block encoding. The sparse encoding is specified by two quantum oracles that yield the location and the value of nonzero elements in AA. Specifically, let r(ℓ,j)r(\ell,j) be a function that returns the row index of the ℓ\ellth nonzero element in the jjth column of AA. The sparse encoding of AA is then described by two oracles

LOC⁡∣ℓ⟩∣j⟩=∣ℓ⟩∣r(ℓ,j)⟩,VAL⁡∣0⟩∣ℓ⟩∣j⟩=R(θ(ℓ,j))∣0⟩∣ℓ⟩∣j⟩(1)\operatorname{LOC}\lvert\ell\rangle\lvert j\rangle= \lvert\ell\rangle\lvert r(\ell,j)\rangle,\qquad\operatorname{VAL}\lvert0\rangle\lvert\ell\rangle\lvert j\rangle= R(\theta(\ell,j))\lvert0\rangle\lvert\ell\rangle\lvert j\rangle \tag*{(1)}

where R(θ)R(\theta) is the single-qubit rotation gate around the yy-axis with angle θ\theta, R(θ)∣0⟩=cos⁡(θ/2)∣0⟩+sin⁡(θ/2)∣1⟩R(\theta)\lvert0\rangle= \cos(\theta/2)\lvert0\rangle+ \sin(\theta/2)\lvert1\rangle, and where cos⁡(θ(ℓ,j)/2)=Ar(ℓ,j),j/∥A∥max⁡\cos(\theta(\ell,j)/2) = A_{r(\ell,j),j}/\lVert A\rVert_{\max} with ∥A∥max⁡\lVert A\rVert_{\max} the max norm of a matrix. The location oracle LOC yields the location (i.e., the row index) of the ℓ\ellth nonzero element in the jjth column of AA and the value oracle VAL encodes the value of the nonzero element in the rotation angle of the single-qubit register. The sparse encoding strongly depends on the function r(ℓ,j)r(\ell,j), which itself depends on the matrix structure.

In the block encoding, the nonunitary matrix AA is embedded as a block of a larger unitary matrix as

UA=[A/α∗∗∗](2)U_A = \begin{bmatrix} A/\alpha& * \\ * & * \end{bmatrix} \tag*{(2)}

where α≥∥A∥\alpha\ge\lVert A\rVert, and the blocks denoted by ∗* could be any matrix as long as UAU_A is a unitary matrix. The block encoding does not need to be exact and can be accurate up to an error ε>0\varepsilon> 0. More specifically, an (α,a,ε)(\alpha,a,\varepsilon)-block-encoding of an nn-qubit matrix AA is an (n+a)(n+a)-qubit unitary UAU_A such that ∥A−α((⟨0a∣⊗In)UA(∣0a⟩⊗In))∥≤ε\lVert A-\alpha((\langle0^a\rvert\otimes I_n)U_A(\lvert0^a\rangle\otimes I_n))\rVert\le\varepsilon.

The sparse and block encodings described above provide a quantum specification of the discretized PDE matrix AA. In practice, however, implementing these encodings requires a suitable classical specification of AA. The classical specification of AA is the classical data needed to implement these quantum encodings. The form of this classical specification depends on the structure of AA. For a banded or sparse matrix with a fixed stencil pattern, it can be the list of distinct nonzero entries and their locations, or an efficiently computable function for generating them. If AA is represented as a linear combination of Pauli operators, then the relevant classical data is a description of the Pauli strings and their coefficients. More generally, the classical specification should describe AA in a form that allows efficient implementation of a quantum encoding of AA. Otherwise, the overhead of constructing the quantum encoding could dominate the overall cost of a quantum PDE solver.

The last step of an end-to-end quantum PDE solver is to extract physically or computationally relevant properties of the solution vector uu. In general, a solution property can be expressed as the expectation value ⟨u∣M∣u⟩\langle u|M|u\rangle of a Hermitian operator MM with respect to the quantum state ∣u⟩|u\rangle that encodes uu. The operator MM encodes the desired property and it is often a physical observable of the system under study. More generally, the desired property can be a function of the solution vector, for example a weighted average of the solution vector. In such cases, the measurement operator MM is constructed so that its expectation value with respect to ∣u⟩|u\rangle yields the desired property. Quantum algorithms for estimating expectation values are often based on a quantum encoding of the measurement operator MM. Therefore, similar to the matrix AA, a suitable classical specification of MM is needed to construct this quantum encoding efficiently.

Loading an initial state onto a quantum register: the binary-tree approach

In this section, we present a procedure for loading a general mm-qubit initial state onto an mm-qubit quantum register using only a single ancilla qubit. The approach we present here is suitable when the amplitudes of the quantum state are given explicitly as classical data, as in our application later in § 3.2, not implicitly by a function or an oracle. Specifically, we consider a general mm-qubit state

∣ψ⟩:=∑j=02m−1aj∣j⟩,∑j∣aj∣2=1,(3)|\psi\rangle:= \sum_{j=0}^{2^m-1} a_j |j\rangle,\qquad\sum_j |a_j|^2 = 1, \tag*{(3)}

whose amplitudes {aj}\{a_j\} are given explicitly as an array of numbers. The binary-tree approach we present here provides a quantum circuit for loading this state using only CNOT and single-qubit rotation gates with classically precomputed rotation angles (Figure 2). First, we construct the circuit and then show the number of gates in the circuit scales as Θ(2m)\Theta(2^m). We also present a procedure for classical pre-computation of the rotation angles used in the circuit with a cost of Θ(2m)\Theta(2^m). This gate cost is expected for loading an arbitrary quantum state; note that just reading the input amplitudes already costs Ω(2m)\Omega(2^m). However, the construction here gives an explicit quantum circuit and a recipe for computing the angles, and is advantageous for small register size, where m∈O(1)m \in O(1) or scales logarithmically with some input parameter.

Figure 2. Quantum circuit for loading a general mm-qubit initial state. (a) The circuit first loads the state with magnitude of the amplitudes using uniformly controlled rotations. The phases of the amplitudes are then added using a uniformly controlled zz-rotation with angle 2αj2\alpha_j on an ancilla qubit initiated at ∣1⟩\lvert1\rangle. (b) The circuit for loading the four-qubit magnitude state using rotation and CNOT gates; all rotation angles are classically precomputable. (c) An example of a uniformly controlled rotation implemented using an alternating sequence of a single-qubit rotation and CNOT gates.

Writing each amplitude as aj=∣aj∣exp⁡(iαj)a_j = |a_j| \exp(\mathrm{i}\alpha_j), we load the state in two steps. In the first step, we load the magnitude state whose amplitudes are {∣aj∣}\{|a_j|\}. In the second step, we incorporate the phase factors by implementing an operation that maps each basis state as ∣j⟩↦exp⁡(iαj)∣j⟩|j\rangle\mapsto\exp(\mathrm{i}\alpha_j)|j\rangle, thereby loading the phases of the original amplitudes.

The magnitude-loading step is based on a binary tree of probability weights. Let us define the probabilities pj:=∣aj∣2p_j := |a_j|^2 and write the computational basis as ∣j⟩=∣jm−1⋯j1j0⟩|j\rangle= |j_{m-1}\cdots j_1j_0\rangle, where qubits are ordered from right to left, j0j_0 is the least significant bit in binary representation of jj, and ∣jℓ⟩|j_\ell\rangle is the state of the ℓ\ellth qubit qℓq_\ell. Consider a binary tree with m+1m+1 levels ℓ=0,…,m\ell=0,\ldots,m, where the root node of the tree is at the top level (ℓ=0)(\ell=0) and the leaf nodes of the tree are at the bottom level (ℓ=m)(\ell=m). At level ℓ<m\ell<m, we set the state of qubit qℓq_\ell by applying a rotation gate. Let x=xℓ−1⋯x0x=x_{\ell-1}\cdots x_0 be the bitstring of the rightmost ℓ\ell qubits qℓ−1⋯q0q_{\ell-1}\cdots q_0 already prepared before level ℓ\ell; note that no qubit is prepared at ℓ=0\ell=0. We use x=0,…,2ℓ−1x=0,\ldots,2^\ell-1 as the index of the three nodes at level ℓ\ell and assign the probability weight

w(ℓ,x)=∑j: right ℓ bits of j=xpj=∑j:j mod 2ℓ=xpj(4)w(\ell,x)=\sum_{j:\,\text{right }\ell\text{ bits of }j=x}p_j=\sum_{j:j\bmod2^\ell=x}p_j \tag*{(4)}

for the node (ℓ,x)(\ell,x). Here w(ℓ,x)w(\ell,x) is the total probability weight of all basis states ∣j⟩|j\rangle whose lower ℓ\ell bits jℓ−1⋯j0j_{\ell-1}\cdots j_0 are equal to xx. For instance, we have nodes x=x0∈{0,1}x=x_0\in\{0,1\} at level ℓ=1\ell=1 and

w(1,0)=∑j:j0=0pj=∑j evenpj,w(1,1)=∑j:j0=1pj=∑j oddpj,(5)w(1,0)=\sum_{j:j_0=0}p_j=\sum_{j\ \mathrm{even}}p_j,\qquad w(1,1)=\sum_{j:j_0=1}p_j=\sum_{j\ \mathrm{odd}}p_j, \tag*{(5)}

References

  1. [1]D. W. Berry, High-order quantum algorithm for solving linear differential equations, J. Phys. A Math. Theor. 47, 105301 (2014).DOI
  2. [2]D. W. Berry, A. M. Childs, A. Ostrander, and G. Wang, Quantum algorithm for linear differential equations with exponentially improved dependence on precision, Commun. Math. Phys. 356, 1057 (2017).DOI
  3. [3]A. M. Childs and J.-P. Liu, Quantum spectral methods for differential equations, Commun. Math. Phys. 375, 1427 (2020).DOI
  4. [4]A. Montanaro and S. Pallister, Quantum algorithms and the finite element method, Phys. Rev. A 93, 032324 (2016).DOI
  5. [5]H. Krovi, Improved quantum algorithms for linear and nonlinear differential equations, Quantum 7, 913 (2023).DOI
  6. [6]D. Fang, L. Lin, and Y. Tong, Time-marching based quantum solvers for time-dependent linear differential equations, Quantum 7, 955 (2023).DOI
  7. [7]S. Jin, N. Liu, and Y. Yu, Quantum simulation of partial differential equations via schrödingerization, Phys. Rev. Lett. 133, 230602 (2024).DOI
  8. [8]M. Bagherimehrab, K. Nakaji, N. Wiebe, G. K. Brennen, B. C. Sanders, and A. Aspuru-Guzik, Fast quantum algorithm for differential equations, Phys. Rev. Res. (2026), arXiv:2306.11802 .arxiv.org/abs/2306.11802
  9. [9]A. M. Childs, J.-P. Liu, and A. Ostrander, High-precision quantum algorithms for partial differential equations, Quantum 5, 574 (2021).DOI
  10. [10]D. Jennings, M. Lostaglio, S. Pallister, A. T. Sornborger, and Y. Subaşı, Randomized adiabatic quantum linear solver algorithm with optimal complexity scaling and detailed running costs, PRX Quantum 6, 040373 (2025).DOI
  11. [11]A. Alase, R. R. Nerem, M. Bagherimehrab, P. Höyer, and B. C. Sanders, Tight bound for estimating expectation values from a system of linear equations, Phys. Rev. Res. 4, 023237 (2022).DOI
  12. [12]F. Golse, S. Jin, and N. Liu, Quantum algorithms for uncertainty quantification: application to partial differential equations (2022), arXiv:2209.11220 [quant-ph] .arxiv.org/abs/2209.11220
  13. [13]J. M. Antosiewicz and D. Shugar, Poisson—Boltzmann continuum-solvation models: applications to pH-dependent properties of biomolecules, Mol. BioSyst. 7, 2923 (2011).DOI
  14. [14]F. Fogolari, A. Brigo, and H. Molinari, The Poisson–Boltzmann equation for biomolecular electrostatics: a tool for structural biology, J. Mol. Recognit. 15, 377 (2002).pubmed.ncbi.nlm.nih.gov/12501158
  15. [15]L. Grover and T. Rudolph, Creating superpositions that correspond to efficiently integrable probability distributions (2002), arXiv:quant-ph/0208112 [quant-ph] .DOI
  16. [16]Y. R. Sanders, G. H. Low, A. Scherer, and D. W. Berry, Black-box quantum state preparation without arithmetic, Phys. Rev. Lett. 122, 020502 (2019).DOI
  17. [17]M. Möttönen, J. J. Vartiainen, V. Bergholm, and M. M. Salomaa, Quantum Circuits for General Multiqubit Gates, Phys. Rev. Lett. 93, 130502 (2004).DOI
  18. [18]V. Vandaele, Asymptotically optimal quantum circuits for comparators and incrementers (2026), arXiv:2603.12917 [quant-ph] .arxiv.org/abs/2603.12917
  19. [19]A. Gilvén, Y. Su, G. H. Low, and N. Wiebe, Quantum singular value transformation and beyond: exponential improvements for quantum matrix arithmetics, in Proceedings of the 51st Annual ACM SIGACT Symposium on Theory of Computing (2019) pp. 193–204.
  20. [20]D. Camps, L. Lin, R. Van Beekmen, and C. Yang, Explicit quantum circuits for block encodings of certain sparse matrices, SIAM J. Matrix Anal. Appl. 45, 801 (2024).arxiv.org/abs/2203.10236
  21. [21]A. M. Childs and N. Wiebe, Hamiltonian simulation using linear combinations of unitary operations, Quantum Inf. Comput. 12, 901 (2012).
  22. [22]E. Knill, G. Ortiz, and R. D. Somma, Optimal quantum measurements of expectation values of observables, Phys. Rev. A 75, 012328 (2007).DOI
  23. [23]C. Quan, B. Stamm, and Y. Maday, A domain decomposition method for the poisson–boltzmann solvation models, SIAM J. Sci. Comput. 41, B320 (2019).arxiv.org/abs/1807.05384
  24. [24]N. A. Baker, D. Sept, M. J. Holst, and J. A. McCammon, The adaptive multilevel finite element solution of the Poisson-Boltzmann equation on massively parallel computers, IBM J. Res. Dev. 45, 427 (2001).DOI
  25. [25]C. Kwewu, L. Feng, M. Stein, and P. Benner, Fast solution of the linearized Poisson–Boltzmann equation with nonaffine parametrized boundary conditions using the reduced basis method, Comput. Vis. Sci. 23, 1 (2020).
  26. [26]N. Baker, M. Holst, and F. Wang, Adaptive multilevel finite element solution of the Poisson–Boltzmann equation II. refinement at solvent-accessible surfaces in biomolecular systems, J. Comput. Chem. 21, 1343 (2000).doi.org/10.1002/1096-987x(20001130)21:15<1343::aid-jcc2>3.0.co;2-k
  27. [27]M. Holst, N. Baker, and F. Wang, Adaptive multilevel finite element solution of the Poisson–Boltzmann equation I. algorithms and examples, J. Comput. Chem. 21, 1319 (2000).
  28. [28]M. J. Holst, The Poisson–Boltzmann equation: Analysis and multilevel numerical solution (1994), unpublished report; updated and extended version of the author’s Ph.D. thesis.
  29. [29]A. Jha, M. Nottoli, A. Mikhalev, C. Quan, and B. Stamm, Linear scaling computation of forces for the domain-decomposition linear Poisson–Boltzmann method, J. Chem. Phys. 158, 104105 (2023).DOI
  30. [30]Charge discretization in finite-difference Poisson–Boltzmann calculations, APBS Documentation; accessed August 10, 2026.
  31. [31]R. C. Harris, A. H. Boschitsch, and M. O. Fenley, Influence of grid spacing in Poisson–Boltzmann equation binding energy estimation, J. Chem. Theory Comput. 9, 3677 (2013).pubmed.ncbi.nlm.nih.gov/23997692
  32. [32]M. Bagherimehrab and A. Aspuru-Guzik, Efficient quantum algorithm for all quantum wavelet transforms, Quantum Sci. Technol. 9, 035010 (2024).DOI
  33. [33]J. Nie, W. Zi, and X. Sun, Quantum circuit for multi-qubit toffoli gate with optimal resource (2024), arXiv:2402.05053 [quant-ph] .arxiv.org/abs/2402.05053
  34. [34]T. J. Yoder, G. H. Low, and I. L. Chuang, Fixed-point quantum search with an optimal number of queries, Phys. Rev. Lett. 113, 210501 (2014).DOI
  35. [35]wwPDB consortium, Protein Data Bank: the single global archive for 3D macromolecular structure data, Nucleic Acids Res. 47, D520 (2019).DOI
  36. [36]L. Lapworth and C. Sündenhauf, Preconditioned block encodings for quantum linear systems, Quantum Sci. Technol. 10, 045064 (2025).DOI
  37. [37]S. Jin, N. Liu, C. Ma, and Y. Yu, Quantum preconditioning method for finite difference discretizations of the poisson equation via Schrödingerization (2026), arXiv:2505.06866 [math.NA] .arxiv.org/abs/2505.06866
  38. [38]M. Bagherimehrab, Algorithmic quantum-state generation for simulating quantum field theories on a quantum computer, Ph.D. thesis, University of Calgary, Calgary, AB (2022).DOI
  39. [39]M. Bagherimehrab, Y. R. Sanders, D. W. Berry, G. K. Brennen, and B. C. Sanders, Nearly optimal quantum algorithm for generating the ground state of a free quantum field theory, PRX Quantum 3, 020364 (2022).DOI
  40. [40]D. J. George, Y. R. Sanders, M. Bagherimehrab, B. C. Sanders, and G. K. Brennen, Entanglement in quantum field theory via wavelet representations, Phys. Rev. D 106, 036025 (2022).DOI
  41. [41]R. M. Gharat, G. Muraleedharan, D. W. Berry, and G. K. Brennen, Quantum algorithm for solving differential equations using SLAC derivatives (2026), arXiv:2605.04861 [quant-ph] .arxiv.org/abs/2605.04861
  42. [42]T. Häner, M. Roetteler, and K. M. Svore, Optimizing quantum circuits for arithmetic (2018), arXiv:1805.12445 [quant-ph] .arxiv.org/abs/1805.12445
  43. [43]R. Qi and R. Luo, Robustness and efficiency of poisson–boltzmann modeling on graphics processing units, Journal of chemical information and modeling 59, 409 (2018).DOI
  44. [44]M. Holst and F. Saied, Multigrid solution of the Poisson–Boltzmann equation, J. Comput. Chem. 14, 105 (1993).DOI
  45. [45]J. Wang and R. Luo, Assessment of linear finite-difference Poisson–Boltzmann solvers, J. Comput. Chem. 31, 1689 (2010).DOI
  46. [46]W. L. Briggs, V. E. Henson, and S. F. McCormick, A Multigrid Tutorial, 2nd ed. (Society for Industrial and Applied Mathematics, 2000).
  47. [47]B. K. Berntson, D. Jennings, M. Lostaglio, and S. Parker, An end-to-end quantum algorithm for weakly nonlinear plasma physics with superquadratic speedup (2026), arXiv:2607.14308 [quant-ph] .arxiv.org/abs/2607.14308
  48. [48]A. M. Childs, L. Johnston, B. Kiedrowski, M. Vempati, and J. Yu, Quantum algorithms for heterogeneous pdes: The neutron diffusion eigenvalue problem (2026), arXiv:2604.05098 [quant-ph] .arxiv.org/abs/2604.05098

Paper details

vlnc:W_5sp5dnqm
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

AI tools

View statement
Looking for related papers…
Contents