1: %#!platex journal.tex
2: %%
3: %% analysis_second.tex %%
4: %%
5: %% Last modified: 05/03/25 20:34:08 on dell148
6: %%
7:
8: \subsection{Decoherence on the second register}
9: \label{sec:analysis_second}
10:
11: %%% 第1レジスタは qubit ごとに表現されているが，
12: %%% 第2レジスタは \ket{b}, \ket{g} によってしか
13: %%% 表現できないから．
14:
15: We then deal with the case
16: that the decoherence error occurs once on the second register
17: in the quantum counting.
18: %Similar to the analysis of the first register,
19: %we investigate the decoherence
20: %by calculating probability distributions.
21: %For calculations,
22: %we begin by representing the error
23: %on the second register.
24: %
25: In Section~\ref{sec:analysis_first},
26: we treat the error on the first register,
27: which consists of control gates.
28: In that case, error affects only the number of application $m$
29: of controlled-$G^m$.
30: On the other hand, the error on the second register
31: modifies the state on which Grover operator acts.
32: Because the states $\ket{b}$ and $\ket{g}$ depend
33: on the quantum oracle,
34: we can not specify how the second register is disturbed
35: by the decoherence.
36: We need to consider the disturbance
37: and action of Grover operator on a disturbed second register.
38:
39: We first show action of Grover operator
40: on an arbitrary quantum state
41: in order to deal with the operator on a disturbed
42: second register.
43: Any quantum state $\ket{\phi} \in \mathcal{H}$ is decomposed
44: as follows
45: %For $\ket{\phi}$,
46: %the following unique decomposition
47: %is always obtained
48: by means of Gram-Schmidt orthogonalization:
49: \begin{equation}
50: \ket{\phi} :=
51: \fourGroverstate{\phi},
52: \label{eq:phi}
53: \end{equation}
54: where $\up, \vp, \uep, \vep \in \mathbb{C}$,
55: $\ketebp \in \mathcal{H}_b, \ \ketegp \in \mathcal{H}_g$, and
56: $\ketebp$ and $\ketegp$ are determined
57: such that $\inner{b}{\ebp} = \inner{g}{\egp} = 0$.
58: % depends on $\ket{\phi}$,
59: %As stated in Section~\ref{sec:counting},
60: %Grover operator is defined
61: %by a combination of two unitary operations
62: %as $G:=U_2U_1$,
63: %where
64: %\begin{eqnarray*}
65: % U_1 &:= \sum_x(-1)^{f(x)}\mixed{x}{x}, \\
66: % \ U_2 &:= 2\mixed{s}{s}-I, \
67: % \ket{s} %:= \amp{N}\sum_{i=0}^{N-1}\ket{i}
68: % := \frac{\sqrt{t}}{\sqrt{N}}\ket{g}
69: % + \frac{\sqrt{N-t}}{\sqrt{N}}\ket{b}.
70: %\end{eqnarray*}
71: %By $\inner{s}{\ebp} = \inner{s}{\egp} = 0$ and
72: %application of $U_1$ and $U_2$ operations
73: %to new bases states $\ketebp$ and $\ketegp$:
74: % \begin{align*}
75: % \begin{cases}
76: % U_1\ketebp = \ketebp \\
77: % U_1\ketegp = -\ketegp
78: % \end{cases}, \qquad
79: % \begin{cases}
80: % U_2\ketebp = -\ketebp \\
81: % U_2\ketegp = -\ketegp
82: % \end{cases},
83: % \end{align*}
84: %we have
85: % \begin{align}
86: % G\ketebp &= -\ketebp, \nonumber \\
87: % G\ketegp &= \ketegp.
88: % \end{align}
89: By definition of Grover operator,
90: we obtain the following lemma.
91: %
92: \begin{lemma}
93: \label{lem:G_four_bases}
94: For any quantum state $\ket{\phi}$,
95: Grover operator $G$ can be rewritten as:
96: \begin{equation}
97: G \equiv
98: \left(
99: \begin{array}{cccc}
100: \cos \theta & -\sin \theta & 0 & 0 \\
101: \sin \theta & \cos \theta & 0 & 0 \\
102: 0 & 0 & -1 & 0 \\
103: 0 & 0 & 0 & 1
104: \end{array}
105: \right),
106: \label{eq:G_four_bases}
107: \end{equation}
108: on four-dimensional space
109: spanned by the basis states
110: $\ket{b}, \ket{g}, \ketebp$, and $\ketegp$,
111: which satisfy
112: % which are a priori uniquely determined by a quantum oracle
113: % and the state $\ket{\phi}$,
114: % where
115: %\begin{eqnarray}
116: $\ketebp \in \mathcal{H}_b, \ \ketegp \in \mathcal{H}_g, %\nonumber \\
117: \inner{b}{\ebp} = \inner{g}{\egp} = 0.$
118: %\label{eq:definition_four_bases}
119: %\end{eqnarray}
120: \end{lemma}
121:
122: Before disturbance, the second register is a superposition
123: of the states $\ket{b}$ and $\ket{g}$.
124: If some error occurs on the register,
125: two states are disturbed into
126: \begin{eqnarray}
127: \ket{b} &\rightarrow %\stackrel{E}{\rightarrow}
128: \fourGroverstate{b}, \nonumber \\
129: \ket{g} &\rightarrow %\stackrel{E}{\rightarrow}
130: \fourGroverstate{g},
131: \label{eq:disturbed_bg}
132: \end{eqnarray}
133: satisfying
134: \begin{eqnarray}
135: \ketebb, \ketebg \in \mathcal{H}_b, \
136: \ketegb, \ketegg \in \mathcal{H}_g, \nonumber \\
137: \inner{b}{\ebb} = \inner{b}{\ebg} =
138: \inner{g}{\egb} = \inner{g}{\egg} = 0.
139: \label{eq:six_properties}
140: \end{eqnarray}
141: %It gives the following property with respect to the disturbed Grover space.
142: %
143: \begin{lemma}
144: \label{lem:Grover_six_bases}
145: Any disturbed second register by the decoherence
146: can be represented by the superpositions of
147: $\ket{b}, \ket{g}, \ketebb, \ketebg, \ketegb$, and $\ketegg$,
148: satisfying Equation~(\ref{eq:six_properties}).
149: Action of Grover operator is a rotation by $\theta$
150: on two-dimensional space spanned by $\ket{b}$ and $\ket{g}$,
151: by $\pi$ on $\ketebb$ and $\ketebg$,
152: and by $0$ on $\ketegb$ and $\ketegg$.
153: \end{lemma}
154:
155: %\noindent
156: %In the analysis of probability distributions,
157: %we use two states $\ket{\pm} := \amp{2}(\ket{b}\mp \rmi\ket{g})$
158: %instead of $\ket{b}$ and $\ket{g}$
159: %for convenience,
160: %which are also orthogonal to the other four basis states.
161: %
162: %The relation between $\ketebb$ and $\ketebg$
163: %is usually unknown because
164: %these bases are determined by a quantum oracle.
165: %If we assume a decoherence model as the depolarizing channel,
166: %we can show some properties of the disturbed second register.
167: %We deal with the case of the depolarizing channel
168: %in~\ref{sec:appendix_dep},
169: %and
170: %have no assumption of a type of the decoherence errors
171: %on our analysis later.
172:
173: % Consider angles of rotations of Grover operator $G$.
174: % On three subspaces $\{ \ket{b}, \ket{g} \}, \ket{e_b}$, and
175: % $\ket{e_g}$,
176: % the angles are $\theta, \pi$ and $0$, respectively.
177:
178: \begin{figure}[t]
179: \begin{center}
180: \scalebox{1.0}{ \includegraphics{figures/circuit_error2nd.eps} }
181: \end{center}
182: %\vspace{-1zw}
183: \caption[Position of some error on the second register]{Some error $E$ occurs on the second register
184: after application of controlled-$G^k$ operations with the $j$-th
185: control qubit
186: $(0 \le j \le p-1, 0 \le k \le 2^j)$. }
187: \label{fig:circuit_error2nd}
188: \end{figure}
189:
190: Like the case of the first register,
191: we consider that
192: some error $E$, not necessarily the depolarizing channel,
193: is applied on the second register
194: after application of controlled-$G^k$ operations
195: with the $j$-th control qubit,
196: as shown in Figure~\ref{fig:circuit_error2nd}.
197: \begin{figure}[Ht]
198: %\vspace{-13mm}
199: \begin{center}
200: \begin{minipage}{0.45\linewidth}
201: \begin{center}
202: \hspace{-2mm}
203: \scalebox{1.0}{\includegraphics[angle=270,width=\textwidth]{figures/6_5_1000_lower_s_afterG_1d3_t.ps}}
204:
205: output $\tilde{t}$ \\
206: (a)
207: \end{center}
208: \end{minipage}
209: \begin{minipage}{0.45\linewidth}
210: \begin{center}
211: \scalebox{1.0}{\includegraphics[angle=270,width=\textwidth]{figures/6_5_1000_lower_l_afterG_1d3_t.ps}}
212:
213: \hspace{1.5mm}
214: output $\tilde{t}$ \\
215: \hspace{1.5mm}(b)
216: \end{center}
217: \end{minipage}
218: \vspace{1mm}
219: \caption[The output in the second register case]{
220: The probability distributions of the output $\tilde{t}$
221: by numerical calculations
222: $10^3$ trials with the error rate $d=4\times 10^{-3}$.
223: (a): on the ascending-order circuit,
224: (b): on the descending-order circuit.
225: }
226: \label{fig:err2_outputs}
227: \end{center}
228: \end{figure}
229: %\begin{figure}
230: % \vspace{1zw}
231: % \begin{center}
232: % \begin{minipage}{0.45\linewidth}
233: % \begin{center}
234: % \hspace{-2zw}
235: % %\scalebox{0.87}{\includegraphics{figures/err2s.eps}} \\
236: % \scalebox{1.0}{\includegraphics[angle=270,width=\textwidth]{figures/6_5_1000_lower_s_afterG_1d3_m.ps}}
237: %
238: % measurement result $\tilde{m}$ \\
239: % (g)
240: % %\vspace{-2mm}\hspace{4mm}\scalebox{1.00}{$\tilde{m}$}\\
241: % %\mbox{\hspace{3mm} (a)}
242: %% \vspace{2mm}
243: % \end{center}
244: % \end{minipage}
245: %% \hspace{2mm}
246: % \begin{minipage}{0.45\linewidth}
247: % \begin{center}
248: % \vspace{0.0zw}
249: % \scalebox{1.0}{\includegraphics[angle=270,width=\textwidth]{figures/6_5_1000_lower_l_afterG_1d3_m.ps}}
250: %
251: % \hspace{1.5zw}
252: % measurement result $\tilde{m}$ \\
253: % \hspace{1.5zw}(h)
254: %% \vspace{-2mm}\hspace{4mm}\scalebox{1.00}{$\tilde{m}$}\\
255: %% \mbox{\hspace{3mm} (b)}
256: %% \vspace{2mm}
257: % \end{center}
258: % \end{minipage}
259: % \vspace{1zw}
260: % % \vspace{-3mm}
261: % \caption[Graphs of measurement result in the second register case]{
262: % Graphs of probability distributions
263: % of $\tilde{m}$
264: % by numerical calculations.
265: % (g) and (h) corresponding to
266: % Figure~\ref{fig:err2_outputs}~(a) and (b),
267: % respectively.
268: %% corresponding to
269: %% $10^3$ trials with the error rate $d=4\times 10^{-3}$.
270: %% (a): on the ascending-order circuit,
271: %% (b): on the descending-order circuit.b
272: % }
273: % \label{fig:err2_prob}
274: % \end{center}
275: %\end{figure}
276: %
277: In this case,
278: a probability distribution
279: before measurement
280: on the ascending-order circuit
281: has peaks near
282: $m' \simeq f, 2^p-f$ that are
283: the same as ones in no error case,
284: as detailed in \ref{sec:appendix_second}.
285: On the other hand,
286: a probability distribution
287: on the descending-order circuit
288: has peaks not only near $m' \simeq f, 2^p-f$
289: but also at $m' = 0, 2^p/2$
290: with high probability,
291: independently of the quantum oracle.
292: By calculating an output $t'=N\sin^2 (\pi m'/2^p)$ of the quantum counting,
293: we obtain the following proposition.
294: \begin{proposition}
295: \label{thm:position_outputs_second}
296: The following wrong outputs of the quantum counting
297: related to the first order term of
298: the error rate
299: are obtained with high probability
300: if some decoherence error occurs on the second register:
301: \begin{itemize}
302: \item Wrong outputs near $t$
303: in the ascending-order case.
304: \item Wrong outputs 0 and $N$
305: in the descending-order case,
306: independently of the quantum oracle.
307: \end{itemize}
308: \end{proposition}
309:
310: We show two graphs of outputs of
311: the quantum counting
312: with the ascending-order and the descending-order
313: by numerical calculations
314: in Figure~\ref{fig:err2_outputs}~(a) and (b),
315: respectively.
316: These experiments were done $10^3$ trials
317: with the same conditions as the first register case
318: except that the decoherence error occurs on the second register.
319: %Compare these figures to
320: %Figure~\ref{fig:no_error_outputs}
321: %in the case of no decoherence.
322: %As seen in Figure~\ref{fig:err2_outputs}~(a),
323: %on the ascending-order circuit,
324: %the position of a single peak
325: %is the same as the no decoherence case,
326: %although
327: %the probability to obtain
328: %outputs near the peak is higher.
329: %On the contrary, Figure~\ref{fig:err2_outputs}~(b)
330: %on the descending-order circuit
331: %shows the interesting behavior
332: %that two strong outputs $0$ and $N$
333: %are obtained with high probability
334: %other than the correct output.
335: %The positions of two wrong peaks
336: %are always $0$ and $N$
337: %independently of a quantum oracle.
338:
339: %誤りのピークの位置の違いは，
340: %直観的にはエラーの伝搬方向による．
341: %
342: %第2レジスタにエラーが発生すると，
343: %エラーの影響は第1レジスタに伝搬する．
344: %このとき昇順回路は観測結果の上位ビットに対応する
345: %qubitから演算を行なうので，
346: %エラーは観測結果の下位ビットに向かって伝搬することになる．
347: %
348: %下位ビットをcontrol qubitとしたGrover演算子の適用には，
349: %上位ビットに比べて指数時間かかるので，エラーは
350: %下位ビットで発生しやすい．
351: %したがって，昇順回路では下位ビットだけがエラーの影響を
352: %受けやすくなる．
353: %
354: %それに対して，降順回路ではエラーが上位ビットに向かって伝搬するので，
355: %下位ビットだけでなく上位ビットもエラーの影響によって
356: %正しい状態からずれてしまい，まったく異なるピークが出現しやすく
357: %なるのである．
358: %\begin{remark}
359: % \label{rem:intuitively}
360: The difference of positions of wrong peaks
361: between two quantum counting circuits
362: can be intuitively considered as follows:
363: If a decoherence error occurs
364: on the second register,
365: influences of the error propagate to
366: the first register by controlled-$G$ operators.
367: On the ascending-order circuit,
368: $G$s are applied from the controlled-$G^{2^0}$
369: with the $0$th control qubit
370: corresponding to the MSB of a measurement result.
371: The influences therefore propagate
372: the $0$th control qubit(MSB) to $(p-1)$th control qubit(LSB).
373: % from higher qubits
374: % of the first register to lower qubits.
375: Since application of controlled-$G^m$ operations
376: with the LSB needs
377: more time exponentially than with the MSB,
378: decoherence error occurs
379: with exponential higher probability
380: on controlled-$G^m$ operations with low control qubits.
381: It follows that influences of the error
382: propagate to only low control qubits.
383: On the other hand, in the descending-order case,
384: the influences propagate from
385: the LSB to the MSB of the first register
386: because of reversed ordering of application of
387: controlled-$G$ operations.
388: Therefore not only low control qubits but also
389: high qubits are affected by the decoherence.
390: %\end{remark}
391:
392: %%% 第2レジスタに関しての確率
393: We finally consider the probability to obtain
394: the correct output for the quantum counting.
395: Peaks in the probability distribution on the descending-order circuit
396: are distributed to four peaks
397: whereas the probability distribution on the ascending-order circuit
398: has only the correct two peaks.
399:
400: %%% 正しい解の得られる確率に関する定理
401: \begin{proposition}
402: \label{thm:higher_probability_second}
403: The correct output of the quantum counting
404: is obtained
405: with higher probability
406: on the ascending-order circuit
407: than on the descending-order circuit
408: if the decoherence error occurs on the second register.
409: \end{proposition}
410:
411: %%% 証明
412: %\begin{proof}[Outline of proof]
413: % As shown in~\ref{sec:appendix_second},
414: % all six terms in the probability distribution
415: % on the ascending-order circuit
416: % have the same two correct peaks $m' \simeq f, 2^p-f$,
417: % whereas
418: % on the descending-order circuit,
419: % only two terms $\ket{\pm}$
420: % have the correct peaks,
421: % and the other four terms have wrong peaks
422: % at $m'=0, 2^p/2$.
423: %\endp{roof}
424: