Quantum Phase Estimation (QPE)

Overview

Quantum Phase Estimation (QPE) is the standard quantum method for extracting an eigenphase from a unitary operation.

In simplified terms:

• You have a unitary $U$.
• You have an eigenstate $|\psi\rangle$ such that $$U|\psi\rangle = e^{2\pi i\phi}|\psi\rangle,$$ where $\phi$ is a number in $[0,1)$.

Goal: estimate $\phi$ efficiently.

Why it matters:

• Many problems reduce to “learn something about an eigenvalue.”
• QPE is the engine behind algorithms for simulation, phase-related estimation tasks, and the period-finding backbone of Shor’s algorithm.

Most importantly for intuition: QPE explains how phase becomes measurable information.

Intuition

Phase is invisible if you only measure immediately in the computational basis. If a state picks up a global phase, you can’t detect it.

QPE makes phase observable by:

• creating a superposition of different “time steps” (different powers of $U$)
• letting each branch accumulate a different amount of phase
• interfering those branches so the phase shows up as a measurable bitstring

A good mental model:

• Controlled-$U$ gates “write phase” into a control register.
• The control register is then processed to turn that phase pattern into a readable number.

So QPE is a phase-to-bits converter.

Algorithm Structure

High-level steps:

1. Prepare two registers:
   • an $m$-qubit phase (control) register
   • a target register containing an eigenstate $|\psi\rangle$
2. Put the phase register into a uniform superposition (Hadamards on all $m$ qubits).
3. Apply controlled powers of the unitary:
   • controlled-$U^{2^0}$, controlled-$U^{2^1}$, ..., controlled-$U^{2^{m-1}}$
4. Apply an inverse-QFT-like interference step on the phase register (conceptually: convert phase into binary digits).
5. Measure the phase register to obtain an $m$-bit estimate of $\phi$.

We won’t derive the inverse-QFT here. You just need to understand its role: it is the decoding step.

Formal Description

Eigenstates make QPE clean

QPE assumes the target is an eigenstate:

$$U|\psi\rangle = e^{2\pi i\phi}|\psi\rangle.$$

Then applying $U$ repeatedly just multiplies by a phase:

$$U^k|\psi\rangle = e^{2\pi i k\phi}|\psi\rangle.$$

This is exactly what QPE exploits.

What the controlled powers accomplish

Each control qubit effectively asks:

• “What happens if I apply $U$ this many times?”

Because the phase register is in superposition, the circuit builds a structured phase pattern across the register.

The inverse-QFT stage then converts that structured phase pattern into a computational-basis bitstring.

Key point in words:

• The target register stays (approximately) in $|\psi\rangle$.
• The phase information migrates into the phase register as something you can measure.

If the target is not exactly an eigenstate

If the target is a superposition of eigenstates, QPE behaves like:

• it “samples” one eigenphase according to the state’s overlap with each eigenstate

That is useful in some contexts, but the cleanest story is the eigenstate case.

Worked Example

Use a 1-qubit unitary with a known eigenstate, and estimate a simple phase.

Suppose $U$ has an eigenstate $|\psi\rangle$ with eigenvalue $e^{2\pi i\phi}$ where $\phi = 1/4$. That means applying $U$ multiplies the eigenstate by $e^{i\pi/2}$.

Take $m=2$ phase qubits. The ideal output should be the 2-bit binary approximation of $1/4$, which is:

• $0.01$ in binary

So QPE should output the bitstring 01 with high probability.

Interpretation:

• The controlled-$U$ powers create phase shifts that depend on $\phi$.
• The decoding step maps those shifts to the binary digits of $\phi$.
• Measurement reads the digits.

Turtle Tip

Common Pitfalls

Quick Check

What’s Next

QPE is a core primitive. Next we use it to explain, at a high level, why Shor’s algorithm can factor integers: factoring is reduced to period finding, and period finding is powered by phase estimation and Fourier-style interference.