1: % This file is part of the APS files in the REVTeX 4 distribution.
2: % Version 4.0 of REVTeX, August 2001
3: %
4: % Copyright (c) 2001 The American Physical Society.
5: %
6: % See the REVTeX 4 README file for restrictions and more information.
7: %
8: % TeX'ing this file requires that you have AMS-LaTeX 2.0 installed
9: % as well as the rest of the prerequisites for REVTeX 4.0
10: %
11: % See the REVTeX 4 README file
12: % It also requires running BibTeX. The commands are as follows:
13: %
14: % 1) latex apssamp.tex
15: % 2) bibtex apssamp
16: % 3) latex apssamp.tex
17: % 4) latex apssamp.tex
18: %
19: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
20: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
21:
22: % Some other (several out of many) possibilities
23: %\documentclass[preprint,aps]{revtex4}
24: %\documentclass[preprint,aps,draft]{revtex4}
25: %\documentclass[prb]{revtex4}% Physical Review B
26:
27: \usepackage{graphicx}% Include figure files
28: \usepackage{dcolumn}% Align table columns on decimal point
29: \usepackage{bm}% bold math
30:
31: \newcommand{\ket} [1] {\vert #1 \rangle}
32: \newcommand{\bra} [1] {\langle #1 \vert}
33:
34: %\nofiles
35:
36: \begin{document}
37:
38: %\preprint{APS/123-QED}
39:
40: \title{Experimental quantum communication complexity}% Force line breaks with \\
41:
42: \author{Pavel Trojek$^{1,2}$}
43: % \altaffiliation[Also at ]{}%Lines break automatically or can be forced with \\
44: % \email{}
45: % \homepage{http://www}
46: \author{Christian Schmid$^{1,2}$}
47: \author{Mohamed Bourennane$^{1,2}$}
48: \author{\v{C}aslav Brukner$^3$}
49: \author{Marek \.{Z}ukowski$^4$}
50: \author{Harald Weinfurter$^{1,2}$}
51: \affiliation{$^1$ Max-Planck-Institut f\"{u}r Quantenoptik,
52: D-85748 Garching, Germany
53: \\
54: $^2$ Sektion Physik,
55: Ludwig-Maximilians-Universit\"{a}t, D-80799 M\"{u}nchen, Germany
56: \\
57: $^3$ Institut f\"{u}r Experimentalphysik, Universit\"{a}t Wien,
58: Boltzmanngasse 5, A-1090, Wien, Austria
59: \\
60: $^4$ Instytut Fizyki
61: Teoretycznej i Astrofizyki Uniwersytet Gda\'{n}ski, PL-80-952
62: Gda\'{n}sk, Poland}
63:
64: \date{June 23, 2004}
65:
66: \begin{abstract}
67: %Quantum communication significantly decreases communication costs
68: %in distributed computations.
69: We prove that the fidelity of two exemplary communication
70: complexity protocols, allowing for an N-1 bit communication, can
71: be exponentially improved by N-1 (unentangled) qubit
72: communication. Taking into account, for a fair comparison, all
73: inefficiencies of state-of-the-art set-up, the experimental
74: implementation outperforms the best classical protocol, making it
75: the candidate for multi-party quantum communication applications.
76: \end{abstract}
77:
78: \pacs{03.67.Hk, 42.65.Lm}
79: %\keywords{Suggested keywords}%Use showkeys class option if keyword
80: %display desired
81: \maketitle
82:
83: Quantum information science transgresses limitations of
84: conventional information transfer, cryptography and computation.
85: Recently, significant advantages were recognized when applying
86: quantum phenomena in the field of communication complexity
87: problems (CCP's) \cite{yao79}. There, separated parties,
88: performing {\em local} computations, exchange information in order
89: to accomplish some {\em globally} defined task. Two types of CCP's
90: are distinguished: the first minimizes the amount of information
91: exchange necessary to solve the task with certainty
92: \cite{cleve97,buhrman99,buhrman97}. The second maximizes the
93: probability of successfully solving the task for restricted amount
94: of information \cite{buhrman97,hardy99,brukner02}. Such studies
95: aim, e.g., at a speed up of distributed computations by increasing
96: the communication efficiency, or at an optimization of VLSI
97: circuits and data structures \cite{kushilevitz97}.
98:
99: Quantum CCP protocols, using multi-particle entanglement, were
100: proven to be clearly superior with respect to classical ones
101: \cite{buhrman97,buhrman99,cleve97,hardy99,brukner02}. However, the
102: technology of entanglement based multi-party quantum communication
103: is still in a premature stage. A recent reformulation of quantum
104: CCP's pointed out that even the communication employing single
105: qubits may outperform classical CCP's
106: \cite{buhrman98,raz99,galvao01}. Such a simplification would be of
107: tremendous importance, as it would make a multi-party
108: communication task technologically comparable to quantum key
109: distribution, the only commercial application of quantum
110: information science so far.
111:
112: Here we prove that, for CCP's with restricted communication, the
113: superiority of the single qubit assisted protocols over the
114: corresponding classical ones may increase even exponentially with
115: the number of partners. Furthermore, using parametric
116: down-conversion as a source of heralded single qubits, we
117: experimentally show that quantum protocols solve two exemplary
118: CCP's more efficiently, even with the limited detection efficiency
119: inherent in real single-photon experiments. By solving these CCP's
120: with a sequential transfer of a single qubit only, we demonstrate
121: a generic way of bringing multi-party quantum communication
122: schemes much closer to realistic applications.
123:
124: Let us first introduce the two CCP's analyzed and implemented
125: here. The first one, problem A, is the so called {\em modulo-4
126: sum} problem \cite{buhrman97,buhrman99,galvao01}. Imagine $N$
127: separated partners $P_1, \ldots, P_N$. Each of them receives a
128: two-bit string $X_k$, $\left(X_k=0,1,2,3; k = 1,\ldots,N \right)$.
129: The $X_k$'s are distributed such that their sum is even, i.e.
130: $(\sum_{k=1}^{N} X_k){\textrm{mod} 2} = 0$. No partner has any
131: information whatsoever on the values received by the others. The
132: partners then communicate with the common goal that one of them,
133: say $P_N$, can tell whether the sum modulo-4 of all input strings
134: is equal 0 or 2. That is, $P_N$ announces the value of the
135: dichotomic, i.e. equal $\pm1$, function $T(X_1,...,X_N)$ given by
136: $T_A(X_1,\ldots,X_N) = 1-(\sum_{k=1}^{N} X_k)\textrm{mod} 4$ (for
137: an alternative formulation see footnote \cite{remark1}). The
138: partners can freely choose how to communicate information about
139: their $X_k$, i.e. they can choose between sequential communication
140: from one to the other or any arbitrary tree-like
141: structure ending at the last party $P_N$. However, the total
142: amount of communication is restricted to only $N-1$ bits
143: (classical scenario).
144:
145: Problem B has a similar structure, but now $N$ {\em real} numbers
146: $X_1,\ldots,X_N \in [0,2 \pi)$ with probability density
147: \begin{equation}\label{distrrho}
148: p_B(X_1,\ldots,X_N) = \frac{1}{4(2\pi)^{N-1}}|\cos(X_1+ \ldots +X_N)|
149: \end{equation}
150: are distributed to the partners. Their task is to compute whether
151: $ \cos(X_1+ \ldots +X_N)$ is positive or negative, i.e. to give
152: the value of the function $T_B= S[ \cos (\sum_{k=1}^{N} X_k)]$,
153: where $S(x)=x/|x|$. The communication restriction is the same as
154: for problem A.
155:
156: To find the best performing classical protocols for these CCP's,
157: we first rewrite the random inputs $X_k$. For the task A we put
158: $X_k = (1-y_k) + x_k$, where $y_k \in \{-1,1\}$, $x_k \in
159: \{0,1\}$. For the task B we write $X_k = \pi (1-y_k)/2 + x_k$,
160: with $y_k \in \{-1,1\}$, $x_k \in [0,\pi)$. Note that the
161: dichotomic variables $y_k$ are not restricted by the probability
162: distributions for the $X_k$'s. Thus they are completely random.
163: The global task function $T$ can now be put as $T=\prod_{k=1}^N
164: y_k f(x_1,...,x_N)$, see \cite{reformulation}, and
165: $p(X_1,...,X_N)=2^{-N}p'(x_1,...,x_N)$.
166:
167: Depending on the value of the product of all $y_k$'s the value of
168: $T$ flips between $\pm1$. Thus, if information on $y_k$ of any of
169: the partners is omitted in the course of the protocol, the result
170: is completely random. This implies that all $N$ partners must be
171: involved in an unbroken communication structure. Because of the
172: restriction to maximum $N-1$ bits of communication, each of the
173: partners must send only {\em one} bit, except for the last one,
174: $P_N$, who gives the result \cite{broadcasting}.
175:
176: Each of these one-bit messages encodes the value of a dichotomic
177: function $e_k=\pm1$. It depends on the local input number $X_k$
178: and possibly on information $\{e_l,e_m,\ldots\}$ already received
179: from other partners. Due to the highly restricted form of two
180: valued functions, see \cite{dichotomic}, one can express any $e_k$
181: in the form $e_k = b_k(x_k, e_l, e_m, \ldots) + c_k(x_k, e_l, e_m,
182: \ldots)y_k$. In order to obtain a non-random final result, $e_k$
183: {\it must} depend on $y_k$. Thus, $b_k$ must be equal 0, while
184: $c_k$ itself is now a dichotomic function. Continuing the
185: expansion of $c_k(x_k, e_l, e_m, \ldots)$ and keeping in mind that
186: all previous messages received by the $k$-th partner must be taken
187: into account, we obtain $e_k = y_k a_k(x_k) e_l e_m \ldots $,
188: where $a_k$ is again a dichotomic function depending only on the
189: local input $x_k$. Next, one expands in a similar way $e_l, e_m,
190: \ldots$, which leads to
191: $e_k=y_ka_k(x_k)y_la_l(x_l)y_ma_m(x_m)\ldots$. The final answer,
192: $e_N$, given by $P_N$ must have the same structure as $e_k$ and
193: therefore, it must be equal to $e_N = \prod_{i=1}^{N}a_i(x_i)y_i$,
194: \cite{sequential}.
195:
196: If the answer is correct, then $T = e_N$, and thus $T\cdot e_N =
197: 1$. Otherwise, $T \cdot e_N = -1$. Thus, one can introduce the
198: measure $F$ (fidelity) of the average success in the form
199: \begin{equation}
200: F_c =\left| \sum\limits _{X_1, \ldots, X_N}p(X_1, \ldots,
201: X_N)T(X_1, \ldots, X_N)e_N \right|.
202: \end{equation}
203: For the problem B the summations are replaced by integrations. The
204: probability of success, $P$, is given by $P =(1+F)/2$. For the
205: best classical protocols of the CCP's given above we obtain
206: \begin{eqnarray}
207: F_c&=&\left|2^{-N}\sum\limits _{x_1, \ldots, x_N}\sum\limits
208: _{y_1, \ldots, y_N=\pm1} p'(x_1, \ldots, x_{N}) \prod_{l=1}^N
209: y_l\right.
210: \nonumber\\
211: && \left. \times f(x_1, \ldots, x_{N}) e_N(X_1,...,X_N) \right|
212: \nonumber \\
213: &=&\left|\mathop{\sum}\limits _{x_1, \ldots, x_N}g(x_1, \ldots,
214: x_N)a_1(x_1)\ldots a_N(x_N)\right|,
215: \end{eqnarray}
216: where we denoted the product $p'(x_1, \ldots, x_{N}) f(x_1,
217: \ldots, x_{N})$ by $g(x_1, \ldots, x_{N})$. Since $F_c$ depends on
218: the product of local functions $|a_i(x_i)|\leq1$, it is bounded
219: from above, i.e., $F_c\leq B(N)$ \cite{brukner02}.
220:
221: The bounds $B(N)$ for our problems A and B can be easily
222: calculated. In both cases the fidelity decreases exponentially
223: with number $N$ of parties. For task A one has $F_{c,A}=2^{-K+1}$,
224: where $K=N/2$ and $K=(N+1)/2$ for even and odd number of parties,
225: respectively. This {\em analytic} result, valid for arbitrary $N$,
226: confirms the numerical simulations of \cite{galvao01} for small
227: $N$. For task B we derived $F_{c,B}=(2/\pi)^{N-1}$. Due to the
228: formal analogies the integrals needed to get this result already
229: appeared in the derivation of a Bell inequality involving
230: continuous range of settings \cite{zukowski93}.
231:
232: %For the quantum solution of these CCP's we identify the capacity
233: %of a classical bit according to Holevo's theorem with the capacity
234: %of a qubit and consequently restrict the exchange of information
235: %to $N-1$ qubits.
236: The Holevo bound \cite{holevo73} limits the information storage
237: capacity of a qubit to exactly one classical bit. Thus, we
238: restrict the maximum communication exchange for quantum protocols
239: of the presented CCP's to $N-1$ qubits, or alternatively, to
240: $N-1$-times exchange of a {\em single} qubit. The solution of task
241: A starts with a qubit in the state
242: $\ket{\psi_i}=2^{-1/2}(\ket{0}+\ket{1})$. Parties sequentially act
243: on the qubit with the unitary phase-shift transformation of the
244: form $ \ket{0}\bra{0} + e^{i\pi/2X_k}\ket{1}\bra{1}$, in
245: accordance with their local data. After all $N$ phase shifts the
246: state is
247: \begin{equation}
248: \ket{\psi_f} = \frac{1}{\sqrt{2}}(\ket{0}+e^{i\pi/2(\sum_{k=1}^{N}
249: X_k)}\ket{1}).
250: \end{equation}
251: Since the sum over $X_k$ is even, the phase factor
252: $e^{i\pi/2(\sum_{k=1}^{N} X_k)}$ is equal to the dichotomic
253: function $T_A$ to be computed. Therefore, a measurement of the
254: qubit in the basis $ (\ket{0}\pm \ket{1})/\sqrt{2}$ reveals the
255: value of $T_A$ with fidelity $F_{q,A} = 1$, that is, always
256: correctly.
257:
258: Our quantum protocol for task B starts with a qubit in the same state $
259: \ket{\psi_i}$. Each party performs according to his/her
260: local data a unitary transformation $ \ket{0}\bra{0} + e^{i
261: X_k}\ket{1}\bra{1}. $ Thus, the final state is
262: \begin{equation}
263: \ket{\psi_f} = \frac{1}{\sqrt{2}}(\ket{0}+e^{i \sum_{k=1}^{N}
264: X_k}\ket{1}).
265: \end{equation}
266: The last party makes the same measurement as in task A. The
267: probability for the detection of state $2^{-1/2}(\ket{0} \pm
268: \ket{1})$, which we associate with the result $r=\pm1$, is given
269: by $P(\pm) = [1 \pm \cos(\sum_{k=1}^{N} X_k)]/2$. The expectation
270: value for the final answer $e_N=r$ is $ E(X_1,\ldots,X_N)= P(+) - P(-)$, and reads
271: $\cos(\sum_{k=1}^{N} X_k)$. The fidelity of $e_N$, with
272: respect to $T_B$ is
273: \begin{eqnarray}
274: F_{q,B} & = & \int_0^{2\pi}dX_1 \ldots \int_0^{2\pi}dX_N
275: p_B(X_1,\ldots,X_N) \nonumber\\
276: & & \times T_B(X_1,\ldots,X_N)
277: E(X_1,\ldots,X_N).
278: \end{eqnarray}
279: With the actual forms of $p_B$, $T_B$, and $E$, one gets $F_{q,B}
280: = \pi/4$, i.e., the protocol gives the correct value of $T_B$ with
281: probability $P_{q,B} = (1+\pi/4)/2 \approx 0.892$.
282:
283: For both problems the classical fidelity $F_c$ or the probability
284: of success $P_c$ decreases exponentially with number $N$ of
285: parties to the value corresponding to a random guess of the result
286: of $T$, i.e. to the value achievable without any communication at
287: all. In contrast, $P_q$ remains constant for any $N$ and reaches 1
288: for task A, and $\approx 0.892$ for task B. The simple, one qubit
289: assisted quantum protocol clearly outperforms the best classical
290: protocols
291: %interestingly only, if the partners do not read the message
292: %encoded by the previous partners.
293: without any shared multi-particle entanglement (!), utilizing only
294: the coherence properties of the transmitted qubit.
295: %We see that by utilizing coherence properties of the transmitted
296: %qubit the partners can clearly outperform the best classical
297: %protocols. Interestingly, this requires that they do not read out
298: %the message encoded by the previous partners as they would destroy
299: %thereby the coherence.
300:
301: We implemented the quantum protocols for $N=5$ parties, using a
302: heralded single photon as the carrier of the qubit communicated
303: from one partner to the other \cite{coherence}. The qubit was
304: encoded in polarization. The computational basis, $``0"$ and
305: $``1"$, corresponds to horizontal $H$ and vertical $V$ linear
306: polarization, respectively. The data $X_k$ of each party was
307: encoded on the qubit via a phase shift using birefringent
308: materials. The last party performed a measurement in the
309: $2^{-1/2}(\ket{H}\pm\ket{V})$ basis in order to obtain the final
310: value $T$.
311:
312: The schematic set-up is shown in Fig. \ref{thelabel}. Photon pairs
313: are produced via spontaneous parametric down-conversion (SPDC).
314: The detection of one photon by detector D$_\mathrm{T}$ as a
315: trigger heralds the existence of the other one used in protocol.
316: The narrow gate window of 4 ns for observing the coincidence
317: detection between these two photons along with the single-count
318: rates of $\sim 140000$ s$^{-1}$ at the detectors D$_+$ and D$_-$
319: warrant that the recorded data are due to single photons only.
320: Type-II SPDC in 2 mm thick $\beta$-barium borate (BBO) crystal,
321: pumped by a single-mode laser diode (402.5 nm, 10 mW) is used,
322: emitting pairs of orthogonally polarized photons at $\lambda=805$
323: nm ($\Delta \lambda \approx 6$ nm). Filtering of the vertical
324: polarization of trigger photons by a polarizer, ensures that the
325: protocol photon has horizontal polarization initially. A half-wave
326: plate (HWP$_1$) transforms the state of the photon to
327: $2^{-1/2}(\ket{H} + \ket{V})$ as required in protocol.
328:
329: \begin{figure}[tb]
330: \includegraphics[width=86mm]{exp_setup.eps}
331: \caption[Short caption]{\label{thelabel} Set-up for quantum CCP.
332: Pairs of orthogonally polarized photons are emitted from a BBO
333: crystal via the type-II SPDC process. The detection of one photon
334: as trigger at D$_\mathrm{T}$ indicates the existence of the other
335: one used in protocol. The polarization state is prepared with a
336: half-wave plate (HWP$_1$) and a polarizer, placed in the trigger
337: arm. Each of the five parties introduces a phase-shift by the
338: rotation of a birefringent YVO$_4$ crystal (C$_1$ to C$_5$). The
339: last party performs the analysis of a photon-polarization state
340: using a half-wave plate (HWP$_2$) followed by a polarizing
341: beam-splitter (PBS).}
342: \end{figure}
343:
344: For a fair comparison between the classical protocol and the
345: quantum protocol, no heralded events are discarded, even if the
346: detection of the protocol photon fails. In such a case it is still
347: allowed to guess the value of $T$. This works with probability
348: 1/2, and leads to very demanding experimental requirements for
349: unambiguous demonstration of the enhanced efficiency of
350: qubit-assisted CCP compared to its classical counterpart
351: \cite{galvao01}. In particular, high detection efficiency of the
352: heralded photons, i.e. high coincidence/single ratio for our
353: set-up, is essential.
354:
355: In order to minimize the events where no photon was detected, the
356: yield of heralded photons was maximized by adopting an unbalanced
357: SPDC scheme. That means, we select a restricted spatial mode with
358: well defined polarization of the trigger photons by coupling them
359: into single-mode fibre behind a polarizer, whereas no spatial
360: filtering is performed on the protocol photons. With such
361: configuration we observed $\approx 5000$ trigger events per second
362: with $\approx 2400$ coincident events per second of protocol
363: detections, i.e. an overall detection efficiency of $\approx
364: 0.48$, close to the limit given by the detector efficiency of the
365: avalanche photodiodes used, which was about $55\%$ for our
366: operating wavelength.
367:
368: The individual phase shifts of parties are implemented by rotating
369: 200 $\mu$m thick Yttrium-Vanadate (YVO$_4$) birefringent crystals
370: (C$_i$) along their optic axis, oriented perpendicularly to the
371: beam. An additional YVO$_4$ crystal (C$_{\rm comp}$) compensates
372: dispersion effects. To analyze the polarization state of photons
373: in the desired basis, a half wave-plate (HWP$_2$) followed by
374: polarizing beam-splitter (PBS) is used.
375:
376: The protocols were run many times, to obtain sufficient
377: statistics. Each run took about one second. It consisted of
378: generating a set of pseudorandom numbers obeying the specific
379: distribution, subsequent setting of the corresponding phase shifts
380: by the rotations of YVO$_4$ crystals, and opening detectors for a
381: collection time window $\tau$. The limitation of communicating one
382: qubit per run requires that only these runs, in which exactly one
383: trigger photon is detected during $\tau$, are selected for the
384: evaluation of the probability of success $P_{exp}$. To maximize
385: the number of such runs, $n$, the length of $\tau$ was optimized
386: to 200 $\mu$s assuming Poissonian photon-number distribution of
387: SPDC photons.
388:
389: In order to determine the probability of success from the data
390: acquired during the runs we have to distinguish the following two
391: cases. First, the heralded photon is detected, which happens with
392: probability $\eta$ given by the coincidence/single ratio. Then,
393: the answer on the value of the function $T$ can be based on the
394: measurement result. However, due to experimental imperfections in
395: the preparation of the initial state, the setting of the desired
396: phase shifts and the polarization analysis, the answer is correct
397: only with probability $\gamma$, which must be compared with the
398: theoretical limits given by $P_{q,A}$ and $P_{q,B}$ for the task A
399: and B, respectively. Second, with the probability $1-\eta$ the
400: detection of the heralded photon fails. Forced to make a random
401: guess, one gives the correct answer in half of the cases. This
402: leads to an overall success probability
403: $P_{exp}=\eta\gamma+(1-\eta)0.5$, or a fidelity of $F_{exp}=
404: \eta(2\gamma -1)$.
405:
406: Due to a finite measurement sample, our experimental results for
407: the success probability are distributed around the value $P_{exp}$
408: as shown in Fig. \ref{thelabel1} for both tasks. The width of the
409: distribution is interpreted as the error in the experimental
410: success probability. For task A we obtain a quantum
411: success probability of $P_{exp,A} = 0.711 \pm 0.005$.
412: \begin{figure}[tb]
413: \includegraphics[width=86mm]{succ.eps}
414: \caption[Short caption]{\label{thelabel1} Histograms of measured
415: quantum success probabilities (a) for the task A and (b) for the
416: task B. The bounds for optimum classical protocols are displayed
417: as well.}
418: \end{figure}
419: The bound $P_{c,A} = 5/8$ for the optimum classical protocol is
420: violated by 17 standard deviations. For the task B we reached
421: $P_{exp,B} = 0.669 \pm 0.003$, whereas the classical bound is
422: $P_{c,B} \approx 0.582$. The violation is by 29 standard
423: deviations, \cite{fidelity}. Table \ref{tab} summarizes the
424: relevant experimental parameters $n$, $\eta$ and $\gamma$ for both
425: tasks.
426: \begin{table} [tb]
427: \caption{\label{tab}Experimental parameters}
428: \begin{ruledtabular}
429: \begin{tabular}{cccc}
430: & $n$ & $\eta$ & $\gamma$ \\
431: task A & 6692 & $0.452\pm0.010$ & $0.966\pm0.003$\\
432: task B & 18169 & $0.471\pm0.006$ & $0.858\pm0.004$ \\
433: \end{tabular}
434: \end{ruledtabular}
435: \end{table}
436:
437:
438: In conclusion, we have proven and experimentally demonstrated the
439: superiority of quantum communication over its classical
440: counterpart for distributed computational tasks by solving two
441: exemplary CCP's. For nontrivial CCP's, where the
442: input from all the partners is required in order to obtain a
443: non-random final result, the best classical fidelity goes
444: exponentially to 0 with increasing number $N$ of partners. Yet, the
445: fidelity stays constant and independent on $N$ for our single
446: qubit assisted protocols.
447:
448: In our experimental realization we have reached
449: higher-than-classical performance even when including all
450: experimental imperfections of state-of-the-art technologies. Thus,
451: by successfully performing fair and real comparison with the
452: classical scenario with present-day technology we clearly
453: illustrate the potential of the implemented scheme in real
454: applications of multi-party quantum communication. Most
455: importantly, our method gives a generic prescription to simplify
456: multi-party quantum communication protocols. For example,
457: multi-party secret-sharing protocols employing multi-qubit
458: GHZ-states and local operations only, can now be directly
459: transformed to single-qubit applications, thereby significantly
460: enhancing their applicability \cite{secret sharing}.
461:
462:
463: M.\.{Z}. was supported by Profesorial Subsidy of FNP, and by MNiI
464: grant No PBZ-MIN-008/ P03/ 2003. This work was supported by the
465: DFG, EU-FET (RamboQ, IST-2001-38864), Marie-Curie program and
466: DAAD/KBN exchange program.
467:
468:
469: \begin{thebibliography}{99}
470:
471: %\bibitem{brassard84}
472: %C. Bennett and G. Brassard, in {\it Proceedings of IEEE Int. Conf.
473: %on Computers, Systems and Signal Processing}, 1984, p. 175; C.
474: %Bennett {\it et. al}, J. Cryptol. {\bf 5}, 3 (1992); N. Gisin, G.
475: %Ribordy, W. Tittel, and H. Zbinden, Rev. Mod. Phys. {\bf 74}, 145
476: %(2002); C. Kurtsiefer {\it et. al}, Nature {\bf 419}, 450 (2002).
477: \bibitem{yao79}
478: A. C.-C. Yao, in {\it Proceedings of the 11th Annual ACM Symposium
479: on Theory of Computing}, (ACM Press, NewYork, 1979), p. 209.
480: \bibitem{cleve97}
481: R. Cleve and H. Buhrman, Phys. Rev. A {\bf 56}, 1201 (1997).
482: \bibitem{buhrman99}
483: H. Buhrman, W. van Dam, P. H\o yer, and A. Tapp, Phys. Rev. A {\bf
484: 60}, 2737 (1999).
485: \bibitem{buhrman97} H. Buhrman, R. Cleve, and W.
486: van Dam, e-print quant-ph/9705033.
487: \bibitem{hardy99}
488: L. Hardy and W. van Dam, Phys. Rev. A {\bf 59}, 2635 (1999).
489: \bibitem{brukner02}
490: \v{C}. Brukner, M. \.Zukowski, and A. Zeilinger, Phys. Rev. Lett.
491: {\bf 89}, 197901 (2002); \v{C}. Brukner, M. \.Zukowski, J.-W. Pan,
492: and A. Zeilinger, Phys. Rev. Lett. {\bf 92}, 127901 (2004).
493: \bibitem{kushilevitz97}
494: E. Kushilevitz and N. Nisan, {\it Communication complexity}
495: (Cambridge University Press, England, 1997).
496: \bibitem{buhrman98}
497: Harry Buhrman, Richard Cleve, and Avi Wigderson, in {\it
498: Proceedings of the 30th Annual ACM Symposium on Theory of
499: Computing}, (ACM Press, New York, 1998), p. 63.
500: \bibitem{raz99}
501: R. Raz, in {\it Proceedings of the 31st Annual ACM Symposium on
502: Theory of Computing}, (ACM Press, New York, 1999), p. 358.
503: \bibitem{galvao01}
504: E. F. Galv\~{a}o, Phys. Rev. A. {\bf 65}, 012318 (2001).
505: %\bibitem{grover97}
506: %L. K. Grover, e-print quant-ph/9704012; G. Brassard, e-print
507: %quant-ph/0101005.
508: %\bibitem{galvao03}
509: %E. F. Galv\~{a}o and L. Hardy, Phys. Rev. Lett. {\bf 90},
510: %087902 (2003).
511: \bibitem{remark1}
512: An alternative description, simplifying calculations and making the connection
513: with task B more visible, puts the probability distribution
514: for local data as
515: $p_A(X_1,\ldots,X_N)=2^{-2N+1}|\cos(\frac{\pi}{2}\sum_{k=1}^{N}X_k)|$
516: and the global task function as
517: $T_A(X_1,\ldots,X_N)=\cos(\frac{\pi}{2}\sum_{k=1}^{N}X_k)$.
518: \bibitem{reformulation}
519: Consequently, $f = f_A=\cos(\frac{\pi}{2}\sum_{k=1}^{N}x_k)$, and
520: $f=f_B=S[\cos(\sum_{k=1}^{N}x_k)]$.
521: \bibitem{broadcasting}
522: Broadcasting of bits, i.e., communicating them publicly, does not
523: improve the success rate of the classical protocol.
524: \bibitem{dichotomic} {\em Any} dichotomic function, $\sigma=\pm 1$, of
525: a two valued variable, $\xi = \pm 1$ has the form $\sigma(\xi) = b
526: + c \xi$, with $|b|,|c|=0$ or $1$ and $|b| + |c| = 1$, i.e., only
527: one term is nonzero.
528: \bibitem{sequential}
529: In sequential communication, $e_k$ depends only on the local
530: variables $x_k$, $y_k$ and the message $e_{k-1}$ from the previous
531: partner, with $e_k=y_ka_k(x_k)e_{k-1}$. Recursive expansion of
532: $e_{k-1}$ results in the same $e_N$ as above.
533: \bibitem{zukowski93}
534: M. \.{Z}ukowski, Phys. Lett. A {\bf 177}, 290 (1993).
535: \bibitem{holevo73}
536: A. S. Holevo, Probl. Peredachi Inf. {\bf 9}, 3 (1973) [Probl. Inf.
537: Transm. {\bf 9}, 177 (1973)].
538: \bibitem{coherence}One should not be tempted to exchange the heralded
539: single photons carrying the qubits by bright polarized pulses of
540: light. In such a case, a suitable polarization measurement of the
541: pulses reveals all the encoded input data of any party, i.e. two
542: bits for task A and arbitrarily many for task B. Thus, the
543: communication restriction to $N-1$ bits is violated! The
544: attenuation of pulses to single-photon level does not help either.
545: The efficiency of the protocol would be significantly lowered, as
546: the attenuation causes a lot of non-detection events, forcing one
547: to guess the answer most of the time (see description of
548: experiment in main text).
549: \bibitem{fidelity}
550: Expressing the final results in terms of fidelities, we obtain
551: $F_{exp,A} = 0.421 \pm 0.010$ for task A, and $F_{exp,B} = 0.337 \pm 0.006$ for task B. The best
552: classical protocol reaches $F_{c,A}=0.25$ for A and
553: $F_{c,B}\approx0.164$ for B.
554: \bibitem{secret sharing}
555: C. Schmid {\it et. al}, (to be published).
556:
557: \end{thebibliography}
558:
559: \end{document}
560: