1: %#!platex journal.tex
2: %%
3: %% analysis_phase_estimation.tex %%
4: %%
5: %% Last modified: 05/03/25 17:58:41 on dell148
6: %%
7:
8: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
9: \section{Discussion on phase estimation algorithms}
10: \label{sec:phase_estimation}
11:
12: In this section,
13: we extend our results with respect to
14: robust implementation against the decoherence
15: %on the quantum counting
16: to the phase estimation.
17: We also discuss
18: robustness of
19: semi-classical implementation
20: %known as efficient implementation.
21: against the decoherence.
22:
23: %%%%%%%%%%%%%%%%%%%%
24: \subsection{Decoherence on phase estimation algorithms}
25: \label{sec:decoherence_phase_estimation}
26:
27: \begin{figure}[t]
28: \begin{center}
29: \scalebox{0.90}{\includegraphics{figures/circuit_phase_estimation.eps}}
30: \end{center}
31: \caption[Circuit for the phase estimation]{Circuit for the phase estimation
32: with the descending-order.
33: % \hspace{7zw}
34: Controlled rotations $R_j$ are defined as
35: $R_j :=${\scriptsize
36: $\left( \begin{array}{cc} 1 & 0 \\ 0 & \phi_j \end{array} \right)$}
37: with $\phi_j := e^{-2\pi{}i/2^j}$.}
38: \label{fig:circuit_phase_estimation}
39: \end{figure}
40:
41: Phase estimation is one of key quantum algorithms,
42: used in Shor's factoring~\cite{Shor94factorization}.
43: %In this section,
44: %we consider how to implement
45: %the phase estimation
46: %to be robust against the decoherence.
47: Suppose that a unitary operator $U$ has
48: an eigenvector $\ket{u}$
49: with eigenvalue $e^{2\pi{}i\varphi}$,
50: where the value of $\varphi$ is unknown.
51: The goal of the phase estimation is to
52: find the phase $\varphi$.
53: A circuit for the phase estimation is shown
54: in Figure~\ref{fig:circuit_phase_estimation}.
55: The phase estimation is performed
56: by application of controlled-$U$ operations
57: to the second register prepared
58: to the corresponding eigenvector $\ket{u}$ initially.
59: Like the quantum counting,
60: the phase estimation has such {\itshape equivalent} implementations
61: as the ascending-order circuit and the descending-order circuit,
62: by changing ordering of application of controlled-$U$ operations
63: instead of controlled-$G$ operations in the quantum counting.
64: %
65: \begin{claim}
66: The phase estimation can
67: find the desired phase
68: on the ascending-order circuit
69: with higher probability
70: than on the descending-order one
71: in the presence of decoherence on the second register.
72: % if some decoherence error occurs on the second register
73: % in the phase estimation.
74: \end{claim}
75:
76: As mentioned in Subsection~\ref{sec:analysis_second},
77: influences of the decoherence error on the second register
78: in the quantum counting
79: propagate only to lower qubits of the first register
80: with exponentially high probability on the ascending-order circuit,
81: whereas these influences propagate to
82: higher qubits of the first register on the descending-order circuit.
83: This propagation of decoherence
84: is applicable not only to the quantum counting
85: but also to the phase estimation,
86: though positions of wrong peaks caused by the decoherence
87: are determined by a unitary operator $U$ on the phase estimation.
88:
89: %%%%%%%%%%%%%%%%%%%%
90: \subsection{Decoherence on efficient implementations for the phase estimation}
91: \label{sec:analysis_semi-classical}
92:
93: \begin{figure}
94: \begin{center}
95: \scalebox{0.9}{
96: \includegraphics{figures/circuit_semi_classical.eps}
97: }
98: \end{center}
99: \caption[Semi-classical circuit for the phase estimation]
100: {Semi-classical circuit for the phase estimation.
101: $R'_j$ is defined with the results of previous measurements:
102: % $R'_j$ are combinations of the rotations $R_j$ and
103: % the results of previous measurements:
104: $R'_j =${\scriptsize $\left(\begin{array}{cc} 1&0\\ 0&\phi'_j
105: \end{array} \right)$}
106: with $\phi'_j := e^{-2\pi{}i\sum_{k=2}^{j} m_{j-k}/2^k }$.}
107: \label{fig:circuit_semi_classical}
108: \end{figure}
109:
110: An efficient implementation for the phase estimation,
111: called the {\itshape semi-classical}
112: implementation, was demonstrated
113: by Parker et al.~\cite{PP00efficient_factorization}.
114: We show the semi-classical circuit for the phase estimation
115: in Figure~\ref{fig:circuit_semi_classical},
116: corresponding to the circuit
117: in Figure~\ref{fig:circuit_phase_estimation}.
118: On this circuit, inverse quantum Fourier transform
119: and measurements are applied
120: as fast as possible
121: so that only single qubit is required for the first register.
122: By using semi-classical technique,
123: Beauregard~\cite{Beauregard022nplus3}
124: reduced the number of qubits
125: for the factoring about half as implementations
126: in Figure~\ref{fig:circuit_phase_estimation}.
127: It sounds so good for quantum computation
128: since handling many qubits
129: is considered to be difficult.
130: %The semi-classical QFT with single qubit on fiber optics
131: %was experimented in~\cite{TN03mqft}.
132:
133: Here we discuss robustness of the semi-classical
134: implementation against the decoherence.
135: This implementation saves the number of qubits
136: dramatically, although depth of circuits
137: is almost the same as usual implementation.
138: Since influences of the decoherence is considered
139: to increase in proportion to product
140: of the depth of circuits and the number of qubits,
141: the semi-classical implementation is more robust
142: against the decoherence
143: from this point of view.
144:
145: We next focus on the ordering of application
146: of controlled-$U$ operations.
147: On the semi-classical circuit,
148: controlled-$U$ operations must be applied
149: from controlled-$U^{2^{p-1}}$ operations
150: just as the descending-order case
151: so that measurements are done as fast as possible.
152: Since the semi-classical implementation
153: has single qubit for the first register,
154: most influences of the decoherence are caused
155: by the second register.
156: It means that the semi-classical circuit
157: is less robust against the decoherence
158: like the descending-order.
159: %Consequently,
160: %whether the semi-classical implementation
161: %is more robust against the decoherence
162: %than the normal one or not
163: %is determined by an oracle
164: %because influences of the decoherence
165: %on the second register depends on an oracle.
166:
167: We finally note the robustness
168: of the semi-classical implementation
169: for the quantum counting
170: especially in order to solve
171: {\bfseries NP}-complete problems.
172: Checking whether the number of solutions for these problems
173: is zero or non-zero by the quantum counting
174: helps us solve the problems.
175: As we have shown in Proposition~\ref{thm:position_outputs_second},
176: the descending-order circuit has wrong peaks at
177: 0 or $N$ independently of oracle.
178: Since the semi-classical circuit
179: is restricted to descending-order,
180: it may not suitable for such problems.
181: