= Solution
Take the <Hadamard gate> to be $H=2^{-1/2}\begin{pmatrix}1&1\\1&-1\end{pmatrix}$ and the <phase gate> to be $P(\phi)=\operatorname{diag}(1,e^{i\phi})$. This fixes the sign convention for the interference phase.
For the coherent <quantum circuit>, the successive <quantum states> are
$$
|0\rangle\longmapsto\frac{|0\rangle+|1\rangle}{\sqrt2}
\longmapsto\frac{|0\rangle+e^{i\phi}|1\rangle}{\sqrt2}
\longmapsto\frac{(1+e^{i\phi})|0\rangle+(1-e^{i\phi})|1\rangle}{2}.
$$
The <Born rule> therefore gives the concise answer
$$
\boxed{P_0(\phi)=\frac{|1+e^{i\phi}|^2}{4}=\frac{1+\cos\phi}{2}=\cos^2(\phi/2).}
$$
With the stated <quantum decoherence> after the <phase gate>, the joint <pure state> before the last <Hadamard gate> is
$$
\frac{|0\rangle|e_0\rangle+e^{i\phi}|1\rangle|e_1\rangle}{\sqrt2}.
$$
After that <Hadamard gate>, its environment coefficient at output zero is $(|e_0\rangle+e^{i\phi}|e_1\rangle)/2$. Squaring its <norm> sums over every unobserved environment outcome, so
$$
P_0=\frac14\left(2+e^{i\phi}\langle e_0|e_1\rangle+e^{-i\phi}\langle e_1|e_0\rangle\right),\qquad
\boxed{P_0(\phi,v,\alpha)=\frac{1+v\cos(\phi+\alpha)}2.}
$$
The <Cauchy-Schwarz inequality> gives $0\le v\le1$. In this <single-qubit interference with conditional environment states>, $v$ is the <interference visibility>: orthogonal environment states erase the fringe, while a unit-modulus overlap shifts its phase without reducing visibility. The plus sign in $\phi+\alpha$ follows from the specified overlap $\langle e_0|e_1\rangle$, rather than its complex conjugate.
If the same <quantum decoherence> occurs before the <phase gate>, the joint <pure state> is first $(|0\rangle|e_0\rangle+|1\rangle|e_1\rangle)/\sqrt2$. The <phase gate> then produces exactly the same state used above. Equivalently, the controlled environment interaction commutes with the diagonal <phase gate>. \b[The probability is unchanged.]
For the <Deutsch algorithm>, use the reversible <Boolean quantum oracle>
$$
U_f|x\rangle|y\rangle=|x\rangle|y\oplus f(x)\rangle.
$$
The algebraic oracle map printed in the source has its arguments misplaced: taken literally, it sends two distinct answer inputs to the same output for constant $f$, so it cannot be a <unitary operator>. The corrected map above agrees with the diagram's input-control, answer-target arrangement. Put the answer <qubit> in $|-\rangle=(|0\rangle-|1\rangle)/\sqrt2$. Since the <Pauli X gate> satisfies $X|-\rangle=-|-\rangle$, <quantum phase kickback> gives
$$
U_f|x\rangle|-\rangle=(-1)^{f(x)}|x\rangle|-\rangle.
$$
Write $b=f(0)\oplus f(1)$. Up to the irrelevant <global phase> $(-1)^{f(0)}$, the input <qubit> acquires relative phase $(-1)^b$, so its ideal final <computational-basis measurement> returns $b$: zero for a constant function and one for a balanced function.
Let the specified <quantum decoherence> occur in the interference portion between the two <Hadamard gates>. It commutes with the oracle's effective diagonal phase on the input, so the same analysis applies whether it occurs just before or just after the query. The two likelihoods are
$$
\Pr(0\mid b)=\frac{1+(-1)^b v\cos\alpha}{2},\qquad
\Pr(1\mid b)=\frac{1-(-1)^b v\cos\alpha}{2}.
$$
Thus the diagram's usual zero-for-constant decision rule has
$$
\boxed{P_{\mathrm{correct}}=\frac{1+v\cos\alpha}{2}.}
$$
This is the reliability for either promised class separately, with no prior needed. For equally likely classes, if $v\cos\alpha$ is known to be negative, reversing the interpretation gives $(1+|v\cos\alpha|)/2$. A known environment phase can instead be compensated by the <phase gate> $P(-\alpha)$ before the final <Hadamard gate>, giving \b[success probability $(1+v)/2$]. In particular, $v=0$ gives no information and $v=1$ permits certainty after phase compensation. These refinements distinguish a shifted fringe from lost <quantum coherence>; with the unmodified diagram a phase shift alone can reverse the apparent answer. <Quantum decoherence> after the final <Hadamard gate> leaves its computational populations unchanged, so that different placement would not spoil the measured answer.
Back to article page
