Quantum Phase Estimation (QPE)

Overview

Quantum phase estimation, or QPE, measures an eigenvalue of a unitary operator. If U is a quantum operation and the state u is an eigenvector, so that U acting on u gives e to the power 2 pi i phi times u, then QPE returns a binary estimate of the phase phi. With t ancilla qubits it reaches precision of about 2 to the minus t.

QPE is arguably the most important subroutine in quantum computing. It sits inside Shor's algorithm, where the unitary is modular multiplication, and inside quantum chemistry, where the unitary is the evolution under a molecular Hamiltonian and the phase is the molecule's energy. It also powers the HHL algorithm for linear systems and many quantum counting and sampling methods.

The problem it solves

Determining the precise phase associated with an eigenstate of a unitary quantum operation.

History and background

Alexei Kitaev introduced the idea in 1995 in a paper on the abelian stabilizer problem, using a sequence of measurements to estimate a phase one bit at a time. In 1998 Richard Cleve, Ekert, Macchiavello, and Mosca published the version most textbooks use today, which applies controlled powers of U to a register of ancilla qubits and then performs an inverse quantum Fourier transform.

In 2005, Aspuru-Guzik, Dutoi, Love, and Head-Gordon showed that QPE can compute molecular energies, which launched the field of quantum computational chemistry. Since then, researchers have worked to cut the cost of QPE with iterative variants that use a single ancilla, Bayesian phase estimation, and qubitization methods that reduce the number of gates. Small QPE experiments have been run on photons, ions, and superconducting qubits since 2007.

How it works

Combines controlled application of the unitary operation with the inverse Quantum Fourier Transform to extract the phase to high precision.

The idea behind QPE is phase kickback. If you apply a controlled version of U to a control qubit in superposition and a target in an eigenstate, the eigenvalue phase appears on the control qubit as a relative phase between its 0 and 1 components. The control qubit then carries one bit of information about the phase.

Using t control qubits, each controlling U raised to successive powers of two, the controls together hold a quantum Fourier transform of the phase. Applying the inverse QFT converts them into a binary number that approximates phi. The j-th qubit effectively measures the phase with resolution 2 to the minus j, and combining them gives resolution 2 to the minus t.

The cost of precision grows like 2 to the t applications of U in the standard version, which matches the Heisenberg limit: precision proportional to the inverse of the total evolution time. That is a quadratic improvement over statistical estimation, where error shrinks only as the inverse square root of the number of repetitions.

Step by step

1. Prepare a register of t ancilla qubits in the state 0 and a target register in an eigenstate u of U, or in a state with large overlap with it.
2. Apply Hadamard gates to all ancilla qubits to create a uniform superposition.
3. For each ancilla qubit j, apply a controlled U raised to the power 2 to the j to the target register.
4. Apply the inverse quantum Fourier transform to the ancilla register.
5. Measure the ancilla register to obtain an integer m, and report the phase estimate m divided by 2 to the t.
6. If the target is a superposition of eigenstates, the measurement returns one of the eigenvalues with probability equal to the squared overlap, and the target collapses to the corresponding eigenvector.

A worked example

Take U to be a single-qubit phase gate with U acting on the state 1 as multiplication by e to the power 2 pi i times 3 over 8, so phi equals 3 over 8. With t = 3 ancilla qubits, the controlled operations produce phases that the inverse QFT converts exactly to the binary string 011, which equals 3 over 8. The measurement returns 011 with probability 1.

If phi were 0.3, which is not an exact multiple of 1 over 8, the output would be distributed over the nearest values 0.25 and 0.375, with the most likely one being 0.25. Adding more ancilla qubits improves precision, and with t = n + log of 1 over epsilon qubits the estimate is correct to n bits with failure probability below epsilon.

Complexity and resources

What hardware it needs

• Long coherent circuits: the controlled U to the power 2 to the j gates make the circuit depth grow exponentially with the precision.
• Error correction for useful chemistry problems, with logical qubits in the hundreds to thousands depending on the molecule.
• Preparation of an initial state that overlaps well with the target eigenstate, which is a major challenge for chemistry.

Applications

• Quantum chemistry: computing ground and excited state energies of molecules and materials.
• Shor's algorithm, where QPE on modular multiplication yields the period.
• The HHL linear system solver, quantum counting, and amplitude estimation.
• Metrology tasks that measure physical parameters near the Heisenberg limit.

Limitations and caveats

• It needs deep circuits and fault tolerance, so it is not practical on current noisy devices.
• Success depends on preparing a state with large overlap with the desired eigenstate.
• Controlled powers of U can be expensive; for Hamiltonian simulation, each requires a sophisticated implementation.

Common misconceptions

• QPE does not find the eigenvector, only the eigenvalue, although the target register ends up in the corresponding eigenvector.
• It does not give a free exponential speedup for chemistry. The overlap problem and the circuit cost are significant obstacles.

How it compares

• The variational quantum eigensolver is a near-term alternative that uses short circuits and classical optimization but has no accuracy guarantee.
• Iterative phase estimation reaches the same precision with one ancilla qubit, trading qubits for measurements and time.
• Compared with quantum counting, which applies QPE to a Grover operator, QPE here is the general-purpose tool.

Real-world impact

Used inside Shor's Algorithm, quantum chemistry simulations, and the HHL algorithm for solving linear systems — a true workhorse subroutine.

Why it matters

QPE is the shared core of the most valuable fault-tolerant quantum algorithms. Whenever a company says a quantum computer will compute chemistry energies exactly or break RSA, QPE or one of its descendants is inside.

Its Heisenberg-limited scaling is also a rare provable advantage in estimation tasks, and the resources required for QPE are the main input to cost estimates for useful quantum chemistry.

Frequently asked questions

Keep learning

• Related algorithm: Quantum Fourier Transform (QFT)
• Related algorithm: Iterative Quantum Phase Estimation
• Related algorithm: Shor's Algorithm
• Related algorithm: HHL Algorithm

This page is written for educational purposes. Complexity statements describe the best known results at the time of review (September 30, 2026), and research in this area moves quickly, so check recent papers before relying on a specific bound.

← Back to Algorithms Database