Resource reduction for variational quantum algorithms by non-demolition measurements
D. Melegari, R. Abdul Razaq, G. Minuto, P. Solinas · 2025
Read Resource reduction for variational quantum algorithms by non-demolition measurements by D. Melegari, R. Abdul Razaq, G. Minuto, P. Solinas (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-2503-24090.txt.
Introduction
Dario Melegari, Rabia Abdul Razaq, Giovanni Minuto, and Paolo Solinas compare two ways to run variational quantum methods, one that reads the detector in a way that spares the quantum state. Such methods hunt the lowest energy of a complex quantum system. Their work came out in twenty twenty-five.
In the standard way, direct reads pull out the slope of the energy. In the non demolition way, slope data is stored in a quantum detector that is read at the end. Both ways find the energy low point on lithium based molecules and on harder systems. The new way needs far fewer total resources, and its edge grows in a straight line with system size.
Catalog record
| CREATOR | D. Melegari, R. Abdul Razaq, G. Minuto, P. Solinas |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-24090
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Resource reduction for variational quantum algorithms by non-demolition measurements D. Melegari, R. Abdul Razaq, G. Minuto, P. Solinas Abstract We present a comparative study of two implementations of a variational quantum algorithm aimed at minimizing the energy of a complex quantum system. In one implementation, we extract the information of the energy gradient by projective measurements. In the second implementation, called the Non-Demolition approach, the gradient information is stored in a quantum detector, which is eventually measured. As prototypical examples, we study the energy minimization of the Lithium-based molecules and then extend the analysis systems with increased complexity. We find that, while both approaches are able to identify the energy minimum, the Non-Demolition approach has a clear advantage in terms of the overall computational resources needed. This advantage increases linearly with the complexity of the quantum system, making the Non-Demolition approach the ideal candidate to implement such variational quantum algorithms. * * footnotetext: These authors contributed equally to this work I Introduction Quantum systems are notoriously hard to simulate in classical computers [ 1 ] . This is particularly true of many-body quantum systems that are composed of a huge number of atoms and electrons. In this case, the quantum correlations and entanglement cannot be neglected and it is impossible reduce the simulation space which increases exponentially with the number of particles. As a consequence, the classical computational memory is quickly saturated. This poses a strong limitation on the dimension of quantum systems, which can be simulated with classical hardware. Indeed, the simulations of biological molecules and high-temperature superconductors remain out of the reach of even the most powerful classical computers. In this framework, quantum computers might help us reduce the time and cost of these simulations and make them accessible in the future. Because they exploit the laws of quantum mechanics, the problems of simulation dimensions and correlation complexity are immediately solved. Despite the fact that it is not yet clear for which problems we can have a quantum advantage, there are a few categories which are believed to be well-fitted for the quantum simulations. Among these, the simplest is to determine the minimal energy and the ground state of a quantum system. Since physical systems naturally evolve towards their ground state, in the absence of perturbations or in the low energy/temperature case, it is one in which the system is mostly found. Then, knowing the properties of the ground state gives us invaluable information about its behaviour and properties. Starting from these ideas, the most straightforward algorithms to address this problem are borrowed from the variational algorithms, and are named Variational Quantum Algorithm (VQA) (see Ref. [ 2 ] and Sec. II ). The commonly used approaches to VQA are hybrid quantum algorithms. The complex calculation of the system properties is performed with the quantum computers; the data are then passed to a classical computer which performs a simpler and straightforward step. More specifically, the quantum computer is used to extract information about the gradient of a cost function (energy) in a point of the parameter space of the system. Given this, the classical algorithm decides how to change the system parameters in order to minimize the cost function. Iterating this procedure leads to a (local or global) minimum of the cost function. Traditionally, the measurement of the gradient has been done through projective or Direct Measurements (DM ) [ 3 , 4 , 5 , 6 , 7 , 8 ] . However, this approach is time and resource-consuming since it requires running the quantum computational circuits many times. To reduce the quantum computational resources, recently, an alternative way to extract the gradient information has been proposed [ 9 , 10 ] . This is based on the Quantum non-demolition measurement (QNDM ) [ 11 , 12 , 13 , 14 , 15 , 16 , 17 ] and exploits a quantum detector to store the information about the average derivative of the energy at a point. This considerably reduces the resources needed to measure the gradient and run the VQA [ 10 , 9 ] . At the moment, this approach has been analyzed only in terms of gradient measurement, but its applicability to the optimization procedure has even been discussed. In this article, we fill this gap by comparing the efficiency of the DM and QNDM approaches to the optimization procedure. As a test-bench, we use the problem of finding the minimum energy of some real (although approximated) molecules: H 2 and LiH . We show that, while both algorithms converge to the energy minimum within a reasonable time, the QNDM approach needs fewer resources, making it the best choice for the minimization task. Then, we extend the study to minimize the energy of another (but still physical) system. This allows us to increase the Hamiltonian complexity keeping a sound statistical analysis. This allows us to compare the two approaches in a more realistic situations. In these case, the QNDM advantage in terms of quantum computational logical gates increases further with respect to the DM approach. The article structure is the following. In Sec. II , we present the key ideas of the VQA, and in Sec. III , we discuss the two approaches to measure the gradient of the cost function. Section IV contains the details of the numerical simulations and the results. The conclusions are drawn in Sec. V . II Variational quantum algorithms In this section, we recap the fundamentals of Variational Quantum Algorithms (VQAs) [ 2 , 9 , 8 ] . The problem we want to solve is the following: given a complex Hamiltonian \hat{H} , find the minimum energy. In many practical scenarios, the Hamiltonian \hat{H} in a quantum computer implementation is composed of a sum of Pauli strings \hat{P}_{i}=\prod_{j}\hat{A}_{i}^{j} , where \hat{A}_{j}\in[X_{j},Y_{j},Z_{j},I_{j}] , and X_{j},Y_{j},Z_{j} are the usual Pauli operator acting on the j -th qubit [ 18 , 19 ] . In this case, the Hamiltonian can be written as: \hat{H}=\sum_{i=1}^{J}h_{i}\hat{P}_{i} (1) where J is the total number of Pauli string composing \hat{H} , and h_{i} is the weight associated to the i -th string. The increase in the working space dimension, i.e., the number of qubits used, implies a more accurate model. For an n qubits system, initialized in the state |\psi_{0}\rangle=|0\rangle_{1}\otimes|0\rangle_{2}\otimes\ldots\otimes|0\rangle_{n}\equiv|00...0\rangle , the generic unitary transformation acting on the n- qubit system is denoted as U({\vec{\theta}}) , where \vec{\theta} is a vector in the parameter space, and the quantum state |\psi({\vec{\theta}})\rangle is obtained by applying the transformation U({\vec{\theta}}) to the initial state |\psi_{0}\rangle , i.e. |\psi(\vec{\theta})\rangle=U(\vec{\theta})|\psi_{0}\rangle . The unitary operator U({\vec{\theta}}) can be implemented as a sequence of L parameterized unitary transformations U_{j}({\theta_{j}}) , interleaved with unparametrized entangling transformations V_{j} , with j\in[1,L] , such that: U({\vec{\theta}})=V_{L}U_{L}({\theta_{L}})\dots V_{2}U_{2}({\theta_{2}})V_{1}U_{1}({\theta_{1}})\,. (2) For this discussion, we’ll consider U_{j}({\theta_{j}})=\exp{\{-i\theta_{j}\hat{H}_{j}\}} , where \hat{H}_{j} is the corresponding generator of the transformation. If \hat{H}_{j} is composed by the Pauli operators, \hat{H}_{j}^{2}=\mathds{1} , and we have that U_{j}(\theta_{j})=e^{-i\hat{H}_{j}\theta_{j}/2}=\cos\frac{\theta_{j}}{2}\mathds{1}-i\sin\frac{\theta_{j}}{2}\hat{H}_{j}. (3) This accounts for many relevant cases, such as when H_{j} is a tensor product of any multi-qubit Pauli matrices [ 7 ] . However, to further simplify the discussion and the simulation, we restrict our attention to the case in which U_{j}(\vec{\theta}_{j})=R^{j}_{1}(\theta^{j}_{1})\dots R^{j}_{n}(\theta^{j}_{n})\,, (4)