Phase Kickback
Phase kickback is the effect where applying a controlled operation to a target in an eigenstate leaves the target unchanged and instead imprints the eigenvalue's phase onto the control register.
What it means
Suppose |ψ⟩ is an eigenstate of a unitary U with eigenvalue e^{iφ}, i.e., U|ψ⟩ = e^{iφ}|ψ⟩.Apply controlled-U with the control in superposition: the |0⟩ branch does nothing, while the |1⟩ branch picks up the factor e^{iφ}.Because the target returns to exactly |ψ⟩ in both branches, the phase cannot be attributed to the target — it appears as a relative phase in the control register: (|0⟩ + |1⟩)/√2 becomes (|0⟩ + e^{iφ}|1⟩)/√2.The information has been 'kicked back' from target to control.The textbook example uses an oracle with target |−⟩ = (|0⟩ − |1⟩)/√2, an eigenstate of the XOR oracle: Uf|x⟩|−⟩ = (−1)^{f(x)}|x⟩|−⟩, so each input branch x acquires the sign (−1)^{f(x)} while the helper qubit stays |−⟩ throughout.This converts function values into amplitude signs, exactly the form interference can act on.Phase kickback is the working core of the Deutsch–Jozsa algorithm, quantum phase estimation (and hence Shor's algorithm), and Grover's oracle marking step.Everyday analogy
Common misconceptions
- The phase does NOT end up on the target qubit — the target is an eigenstate and exits the operation exactly as it entered. Bookkeeping that assigns the phase to the target misses the point: the physically meaningful change is the relative phase in the control register.
- Phase kickback is not a new physical interaction or a violation of causality — it is ordinary linearity: each control branch evolves as unitarity dictates, and the eigenvalue phase, being a scalar, factors out to wherever the superposition lives.
- Kickback would be invisible if the control were in a definite basis state — a phase on a single branch is a global phase with no physical effect. The control must be in superposition for the kicked-back phase to become a measurable relative phase.
Key takeaways
- If U|ψ⟩ = e^{iφ}|ψ⟩, then controlled-U maps (|0⟩ + |1⟩)|ψ⟩/√2 to (|0⟩ + e^{iφ}|1⟩)|ψ⟩/√2: the phase lands on the control.
- With the helper qubit in |−⟩, an XOR oracle acts as Uf|x⟩|−⟩ = (−1)^{f(x)}|x⟩|−⟩, turning function values into amplitude signs.
- Phase kickback converts an oracle's output into phases that interference can then exploit — it is the engine room of Deutsch–Jozsa, phase estimation, and Grover marking.
- The effect requires the target to be in an eigenstate and the control to be in superposition.
Check your understanding
An XOR oracle Uf is applied with the input register in (|0⟩ + |1⟩)/√2 and the helper qubit prepared in |−⟩. Where does the information about f appear afterward?
- A.In the helper qubit, which flips to |+⟩ when f = 1
- B.As relative signs (−1)^{f(x)} on the input register's branches, while the helper stays |−⟩
- C.Nowhere — the state is unchanged
- D.In a classical register inside the oracle
Show the answer
Answer: B. As relative signs (−1)^{f(x)} on the input register's branches, while the helper stays |−⟩
Why: Because |−⟩ is an eigenstate of the XOR action, Uf|x⟩|−⟩ = (−1)^{f(x)}|x⟩|−⟩. The helper qubit is unchanged; each branch x of the input register acquires the sign (−1)^{f(x)}. That is phase kickback: the function value is imprinted as a phase on the register that asked.
Builds on
Primary source: Nielsen & Chuang, Quantum Computation and Quantum Information (2010), doi:10.1017/CBO9780511976667
Graded 2026-07-10 (human sign-off): established — phase kickback as used in Deutsch–Jozsa and phase estimation per Nielsen & Chuang (2010) §1.4.3–§1.4.4 and §5.2, Preskill Ph219, and the IBM Qiskit Textbook chapter on phase kickback. Pending human grading.
Learn it hands-on
This concept is part of a 46-level curriculum with an interactive simulator and Lumen, a tutor whose answers are verified before you see them. Levels 1–5 are free.