⚛️ quantum physics

A Quantum Circuit for Gaussian Elimination

This paper presents a garbage-free quantum circuit for Gaussian elimination over any finite field, improving upon previous $\mathrm{GF}(2)$ -restricted works while maintaining optimal asymptotic Toffoli depth.

Original authors: Hochang Lee, Kyung Chul Jeong, Panjin Kim

Published 2026-09-25

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Hochang Lee, Kyung Chul Jeong, Panjin Kim

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the quiet, high-stakes world of quantum computing, researchers are constantly trying to teach machines how to solve problems that would take classical computers millennia to finish. To do this, they must translate complex mathematical tasks into a language of quantum bits, or qubits, which can exist in multiple states at once. One of the most fundamental tools in mathematics is a method called Gaussian elimination, a systematic way of untangling a web of linear equations to find a single, clear answer. Imagine a massive spreadsheet filled with numbers; this method is the process of clearing out rows and columns until the solution stands alone. For decades, scientists have known how to run this process on standard computers, but getting a quantum computer to do the same thing has been a stumbling block. The difficulty lies in the fact that quantum operations must be perfectly reversible, meaning no information can be lost or discarded during the calculation, a rule that makes the process much harder to design than its classical counterpart.

A team of researchers at the Aﬃliated Institute of ETRI in South Korea has now built a new quantum circuit that performs this elimination process, but with a significant upgrade over previous attempts. While earlier designs were limited to working only with the simplest kind of numbers, essentially just zeros and ones, this new design is flexible enough to handle any finite field of numbers. This is a crucial distinction because many real-world cryptographic systems and complex data problems rely on more complicated number sets than just binary digits. The researchers developed a way to organize the data so that the quantum computer can perform the necessary steps without leaving behind any "garbage" data. In quantum computing, garbage refers to extra bits of information that are created as a byproduct of a calculation and must be stored or erased later, which wastes precious resources. By ensuring that the final result overwrites the initial input cleanly, the team has created a circuit that uses the absolute minimum amount of memory space required to reverse the operation.

The paper details how the team achieved this efficiency by introducing a specific structure they call a "pseudo row echelon form." In simpler terms, this is a way of arranging the numbers in a grid so that the most important information is preserved in a pattern that looks like a staircase, while the less critical parts of the grid are used to store the secret instructions needed to undo the process later. This clever arrangement allows the computer to solve the system of equations without needing a large amount of extra storage space, a problem that plagued earlier versions of the algorithm. The researchers proved that their method works for any size of matrix, provided the matrix is full of useful information, and they showed that the time it takes to run the calculation is comparable to the best classical methods, even when accounting for the extra steps required to keep the process reversible.

When the researchers compared their new circuit to the best existing designs that worked only with simple binary numbers, they found their approach was superior in almost every way. It required fewer complex logic gates to perform the same task and used less time to complete the calculation, measured by the depth of the circuit. Perhaps most importantly, it did so without needing any extra "garbage" space, a feature that previous designs lacked. This means that as quantum computers grow larger and more powerful, this method will scale efficiently, allowing them to tackle larger and more complex problems without running out of memory. The work represents a generalization of a known technique, proving that the constraints of quantum mechanics do not force scientists to accept inefficient solutions, even for tasks as fundamental as solving linear equations.

The significance of this work extends beyond just the numbers. By demonstrating that a reversible, garbage-free construction is possible for any finite field, the researchers have removed a major bottleneck for future quantum applications. This includes tasks like breaking certain types of encryption or simulating complex chemical reactions, where the ability to manipulate large matrices efficiently is essential. The team did not just propose a theoretical idea; they provided a concrete blueprint for how to build the circuit, detailing exactly how many operations are needed and how they can be arranged in parallel to save time. Their findings suggest that the path to practical quantum advantage in these areas is clearer than before, as the fundamental building blocks for these calculations have been optimized to a level that matches the efficiency of classical computing, all while adhering to the strict rules of quantum reversibility.

Technical Summary: A Quantum Circuit for Gaussian Elimination

Problem Statement

Gaussian elimination is a fundamental algorithm for solving systems of linear equations, computing matrix rank, determinants, and inverses. While abstract quantum algorithms are defined by unitary operations, their practical implementation often requires reversible classical constructions of bit-string mappings. A significant challenge arises because Gaussian elimination is inherently non-injective (many input matrices map to the same row-echelon form), necessitating a workspace (garbage) of at least $x-y$ qubits to maintain reversibility, where $x$ and $y$ are input and output bit lengths.

Previous quantum implementations of Gaussian elimination were restricted to the binary field $GF(2)$ . These existing works faced two primary limitations:

Garbage Accumulation: They either failed to meet the minimum theoretical work space requirements or produced significant "garbage" (supplemental space) that could not be uncomputed.
Complexity Mismatch: They often failed to achieve the same asymptotic complexity as classical implementations, particularly regarding Toffoli depth and qubit count.

Furthermore, no generalized quantum circuit existed for Gaussian elimination over arbitrary finite fields $GF(p^l)$ , limiting the applicability of quantum linear algebra in broader cryptographic and algebraic contexts.

Methodology

The authors propose a reversible quantum circuit for Gaussian elimination that operates over any finite field $GF(p^l)$ . The methodology centers on three core innovations:

1. Pseudo Row Echelon Form

To address the non-injectivity of standard Gaussian elimination, the authors introduce a Pseudo Row Echelon Form.

Definition: A matrix $A'$ is in pseudo row echelon form if its upper triangular elements match those of the standard row echelon form of $A$ . The lower triangular and diagonal parts are not strictly defined by the form but are utilized to store information.
Mechanism: This transformation converts the inherently non-injective mapping into a one-to-one (bijective) transformation. The information required to reverse the operation is preserved within the lower triangular and diagonal entries, allowing the result to be overwritten onto the input qubits while utilizing a minimal number of temporary qubits.

2. Algorithmic Structure

The proposed algorithm (Algorithm 1) iterates through columns to transform the matrix. It relies on four reversible subroutines:

Labeling: Identifies the pivot index (the row with the first non-zero entry in the current column) and encodes it in unary format. The authors utilize a recursive, tournament-like structure to parallelize this process, reducing depth.
Pivoting: Swaps the first row with the pivot row. This is implemented using a tournament-style swapping mechanism that can be parallelized using "borrowed qubits" (ancilla qubits in unknown states that are restored to their initial state by the end of the operation).
Row Reduction: Divides the pivot row elements by the pivot element and subtracts the pivot row from other rows. This involves field arithmetic (multiplication, inversion, addition) which is parallelized using borrowed qubits to minimize depth.
Deleting the Pivot Index: Reverses the Labeling and partial Pivoting steps to restore the work qubits to zero, ensuring the circuit is garbage-free regarding supplemental space.

3. Parallelization with Borrowed Qubits

A key technical contribution is the systematic use of borrowed qubits to parallelize operations.

Instead of requiring clean ancilla qubits, the circuit utilizes qubits that may hold unknown values.
The authors adapt Gidney's approach to transfer shared control values or operands to these borrowed qubits, perform parallel operations, and then apply correction steps to restore the borrowed qubits to their original states.
This technique is applied to controlled swaps (in Pivoting) and arithmetic operations (in Row Reduction), significantly reducing the circuit depth.

Key Contributions and Results

1. Generalization to Arbitrary Finite Fields

Unlike previous works limited to $GF(2)$ , this design is generalized to any finite field $GF(p^l)$ . This allows the circuit to handle more complex algebraic structures relevant to various cryptographic primitives and coding theory problems.

2. Garbage-Free Construction

The developed circuit achieves a garbage-free construction regarding supplemental space.

The result of the Gaussian elimination is overwritten directly onto the input qubits.
The circuit requires $m-1$ temporary work qubits for intermediate calculations, which are fully restored to zero by the end of the operation.
Crucially, the circuit occupies no supplemental (garbage) space beyond these temporary qubits and the input qubits. This contrasts with previous works (e.g., [20]) that required supplemental space proportional to $O(n^2)$ or $O(mn)$ which was not returned to zero. The paper defines supplemental space as work qubits that are initially zero but not returned to zero at the end; this design ensures that count is zero.

3. Complexity Improvements

The paper provides a quantitative comparison of circuit complexities for a matrix of size $m \times n$ (where $m \ge n$ ) in $GF(2)$ , as summarized in Table 1 of the paper:

Metric	Previous Works ([9], [20], [5])	This Work
Toffoli Count	$O(n^2m)$ to $O(n^2m^2)$	$O(n^2m)$
Toffoli Depth	$O(nm)$ to $O(n^2m \log^3 m)$	$O(m \log^2 n + n \log^4 m)$
Temporary Space	$0$ to $m-2$	$m-1$
Supplemental (Garbage) Space	$0$ to $O(n^2)$	$0$

In $GF(2)$ : The circuit achieves the best known asymptotic Toffoli depth up to logarithmic factors while matching the classical arithmetic complexity of $O(n^2m)$ .
In General Fields: The complexity scales with the field size parameter $\alpha = l \lceil \log_2 p \rceil$ , maintaining the same structural efficiency.

Significance and Claims

The authors claim that this work represents a significant optimization in the quantum implementation of linear algebra:

Optimization of Trade-offs: The design successfully optimizes the trade-off between time (depth) and space (qubits). It matches the classical arithmetic complexity while strictly limiting the use of extra space, a critical factor for near-term quantum hardware where qubit count is a bottleneck.
Reversibility without Garbage: By introducing the pseudo row echelon form, the paper demonstrates that Gaussian elimination can be performed reversibly without accumulating supplemental garbage qubits, a feat not achieved in prior $GF(2)$ implementations, while utilizing only $m-1$ temporary qubits.
Broad Applicability: The generalization to arbitrary finite fields expands the scope of quantum algorithms for problems involving linear systems over non-binary fields, such as certain post-quantum cryptanalysis tasks and coding theory applications.

The paper concludes that while further optimizations (such as reducing multiplication depth) are possible, the current design sets a new benchmark for reversible Gaussian elimination, balancing low qubit overhead with competitive time complexity.