1: %% ****** Start of file com.tex ****** %
2: %%
3: %%
4: %% This file is part of the APS files in the REVTeX 4 distribution.
5: %% Version 4.0 of REVTeX, August 2001
6: %%
7: %%
8: %% Copyright (c) 2001 The American Physical Society.
9: %%
10: %% See the REVTeX 4 README file for restrictions and more information.
11: %%
12: %
13: % This is a template for producing manuscripts for use with REVTEX 4.0
14: % Copy this file to another name and then work on that file.
15: % That way, you always have this original template file to use.
16: %
17: % Group addresses by affiliation; use superscriptaddress for long
18: % author lists, or if there are many overlapping affiliations.
19: % For Phys. Rev. appearance, change preprint to twocolumn.
20: % Choose pra, prb, prc, prd, pre, prl, prstab, or rmp for journal
21: % Add 'draft' option to mark overfull boxes with black boxes
22: % Add 'showpacs' option to make PACS codes appear
23: % Add 'showkeys' option to make keywords appear
24: \documentclass[aps,twocolumn,groupedaddress]{revtex4}
25: %\documentclass[aps,prl,preprint,superscriptaddress]{revtex4}
26: %\documentclass[aps,prl,twocolumn,groupedaddress]{revtex4}
27:
28: % You should use BibTeX and apsrev.bst for references
29: % Choosing a journal automatically selects the correct APS
30: % BibTeX style file (bst file), so only uncomment the line
31: % below if necessary.
32: %\bibliographystyle{apsrev}
33:
34: \usepackage{graphicx}
35:
36: \begin{document}
37:
38: % Use the \preprint command to place your local institutional report
39: % number in the upper righthand corner of the title page in preprint mode.
40: % Multiple \preprint commands are allowed.
41: % Use the 'preprintnumbers' class option to override journal defaults
42: % to display numbers if necessary
43: %\preprint{}
44:
45: %Title of paper
46: \title{Wireless communications with diffuse waves}
47:
48: % repeat the \author .. \affiliation etc. as needed
49: % \email, \thanks, \homepage, \altaffiliation all apply to the current
50: % author. Explanatory text should go in the []'s, actual e-mail
51: % address or url should go in the {}'s for \email and \homepage.
52: % Please use the appropriate macro foreach each type of information
53:
54: % \affiliation command applies to all authors since the last
55: % \affiliation command. The \affiliation command should follow the
56: % other information
57: % \affiliation can be followed by \email, \homepage, \thanks as well.
58: \author{S.E. Skipetrov}
59: \email[]{Sergey.Skipetrov@grenoble.cnrs.fr}
60: %\homepage[]{Your web page}
61: %\thanks{}
62: %\altaffiliation{}
63: \affiliation{CNRS/Laboratoire de Physique et Mod\'elisation des Milieux Condens\'es,\\
64: 38042 Grenoble, France}
65:
66: %Collaboration name if desired (requires use of superscriptaddress
67: %option in \documentclass). \noaffiliation is required (may also be
68: %used with the \author command).
69: %\collaboration can be followed by \email, \homepage, \thanks as well.
70: %\collaboration{}
71: %\noaffiliation
72:
73: \date{\today}
74:
75: \begin{abstract}
76: Diffuse, multiple-scattered waves can be very efficient for information transfer
77: through disordered media, provided that antenna arrays are used for both
78: transmission and reception of signals.
79: Information capacity $C$ of a communication channel between
80: two identical linear arrays of $n$ equally-spaced antennas,
81: placed in a disordered medium with diffuse scattering,
82: grows linearly with $n$ and can attain considerable
83: values, if antenna spacing $a \agt \lambda/2$, where $\lambda$ is the wavelength.
84: Decrease of $a$ below $\lambda/2$
85: makes the signals received by different antennas partially
86: correlated, thus introducing redundancy and reducing capacity of the
87: communication system. When the size of antenna arrays is well below
88: $\lambda/2$, the scaling of $C$ with $n$ becomes logarithmic and capacity
89: is low.
90: \end{abstract}
91:
92: % insert suggested PACS numbers in braces on next line
93: \pacs{}
94: % insert suggested keywords - APS authors don't need to do this
95: %\keywords{}
96:
97: %\maketitle must follow title, authors, abstract, \pacs, and \keywords
98: \maketitle
99:
100: Wireless communications in a disordered environment have
101: recently attracted considerable attention \cite{mous00,andrews01,simon01}.
102: At first glance, scattering from randomly distributed heterogeneities disturbs
103: the signal carried by the scattered wave (either acoustic or electromagnetic) and
104: is expected to
105: reduce the efficiency of communication.
106: Indeed, the rate
107: of error-free information transfer per Hertz of frequency bandwidth
108: for a scalar, narrow-band communication channel between a transmitter and
109: a receiver is bounded by the \textit{channel information capacity}
110: $= \log_2 [ 1 + \left| G \right|^2 / N ]$
111: measured in bits per second per Hertz (bps/Hz) \cite{shannon48,cover91}.
112: Here $G = G(\nu)$ is the Fourier transform of the impulse response function $G(t)$,
113: describing the signal at the receiver due to an infinitely short pulse emitted by the
114: transmitter, and $N$ is the power of the Gaussian white noise at the receiver.
115: Obviously, the decrease of the signal power $\left| G \right|^2$ due
116: to scattering results in a reduction of the channel capacity.
117: However,
118: in addition to the overall decrease of $\left| G \right|^2$, scattering introduces
119: random fluctuations of $G$ in space.
120: It was recently realized \cite{foschini98} that one can make use of these fluctuations
121: to overcome the reduction of $C$ due to the drop of the signal power,
122: provided that the communication system contains a large number of antennas
123: \cite{mous00,simon01}. This issue is very promising for applications in wireless communications
124: (mobile telephony in cities, indoor wireless local-area networks in
125: buildings with complex structure,
126: underwater communications with acoustic waves, etc.).
127:
128: Assuming that the information about
129: the scattering environment is available at the receivers (but not at the transmitters)
130: and that uncorrelated noises at different receiving antennas have the same power $N$,
131: the average capacity of a communication channel between arrays of
132: $n$ transmitting and $n$ receiving antennas \footnote{For simplicity, we consider
133: the transmitting and receiving arrays to consist of the same number $n$ of antennas.}
134: can be defined as \cite{mous00}
135: \begin{eqnarray}
136: C =
137: \max\limits_{\mathbf{Q}} \left< \log_2 \det \left[ \mathbf{I}_n + \mathbf{G} \,
138: \mathbf{Q} \, \mathbf{G}^+ / N \right]
139: \right>,
140: \label{c}
141: \end{eqnarray}
142: where
143: $\mathbf{I}_n$ is the $n \times n$ unit matrix, $\mathbf{G}$
144: is a Green matrix ($G_{i \alpha}$
145: gives the signal received by the receiver $\alpha$ due to the transmitter $i$),
146: and $\mathbf{Q}$ is a non-negative definite covariance matrix
147: describing correlations between the transmitted signals (with the
148: constraint on the maximum transmitted power $\mathrm{Tr}\, \mathbf{Q} \leq n$).
149: Angular brackets in Eq.\ (\ref{c}) denote averaging over realizations of disorder.
150: A rigorous analysis of the information capacity $C$ in a disordered medium
151: cannot be based uniquely on the arguments of the information theory
152: \cite{shannon48,cover91} and requires
153: the physical understanding of scattering undergone by the waves carrying the
154: information from transmitters to receivers. In the present paper we
155: analyze $C$ in the framework of the model of diffuse multiple scattering, resulting
156: in a complicated, seemingly random spatial distribution
157: of scattered wave fields (so-called ``speckles'').
158: For the extreme cases of (a) a single speckle spot covering the whole antenna array
159: ($\mathbf{G}$ has perfectly correlated entries) and (b) different antennas
160: situated inside different speckle spots (entries of $\mathbf{G}$ are uncorrelated),
161: we derive complete analytical expressions for $C$.
162: We show that $C$ increases during the continuous
163: transition from the case (a) to the case
164: (b) upon increasing antenna spacing (or, equivalently, upon decreasing
165: correlations between the entries of $\mathbf{G}$) and
166: changes its asymptotic scaling with $n$ from $C \propto \ln n$
167: to $C \propto n$.
168:
169: To be specific, we consider two identical linear arrays of equally-spaced
170: transmitting/receiving antennas placed in a disordered medium.
171: The distance $L$ between the arrays is assumed to be much greater than the
172: mean free path $\ell$ for waves in the medium, while the array size
173: $d = (n-1) a \ll \ell$ and $d \ll (\lambda L)^{1/2}$ (Fresnel limit),
174: where $a$ is the spacing between adjacent
175: antennas and $\lambda$ is the wavelength.
176: In the majority of practically important cases $\lambda \ll \ell$,
177: and hence the propagation of waves from transmitters to
178: receivers is diffusive \cite{ishim78}.
179: Entries $G_{i \alpha}$ of the matrix
180: $\mathbf{G}$ can be then treated as complex Gaussian random variables with zero mean,
181: equal variances,
182: and possibly nontrivial correlations \cite{shapiro86}
183: \begin{eqnarray}
184: \left< G_{i \alpha} G_{j \beta} \right> =
185: \sigma^2 \frac{\sin(k \Delta r_{i j})}{k \Delta r_{i j}}
186: \frac{\sin(k \Delta r_{\alpha \beta})}{k \Delta r_{\alpha \beta}},
187: \label{corr}
188: \end{eqnarray}
189: where
190: $k = 2 \pi / \lambda$,
191: $\Delta r_{i j} = \left| i - j \right| a$,
192: and $\Delta r_{\alpha \beta} = \left| \alpha - \beta \right| a$.
193:
194: \begin{figure}
195: \includegraphics[width=8cm]{fig1.ps}
196: \caption{\label{fig1} Average information capacity per antenna of
197: a communication channel between two arrays of $n$ antennas in a disordered
198: medium assuming statistical independence of the entries of the Green
199: matrix $\mathbf{G}$:
200: results obtained using the random matrix theory (solid line)
201: and Monte Carlo simulation (symbols) are shown for the signal to noise ratio
202: $S/N = 100$. Horizontal dashed line shows the asymptotic
203: value of $C/n$ for $n \rightarrow \infty$. Inset: Capacity per antenna as
204: a function of $S/N$ for single transmitting and single receiving antennas
205: (dashed line) and an infinitely large number of antennas (solid line).}
206: \end{figure}
207:
208: The simplest cases to consider are (a) $k d \rightarrow 0$ and
209: (b) $k a = m \pi$ ($m = 1, 2, \ldots$).
210: In the former case, all $G_{i \alpha}$ are
211: perfectly correlated and one finds
212: $C = \exp[1/(n^2 S/N)] E_1[1/(n^2 S/N)] / \ln 2$,
213: where $E_1$ is the exponential integral function,
214: $S = n \sigma^2$
215: is the average signal power received at each receiver assuming
216: independent signals from transmitters, and
217: $C \propto \ln n$ for $n^2 S/N \gg 1$.
218: In the case (b),
219: $\mathbf{Q} = \mathbf{I}_n$ in Eq.\ (\ref{c})
220: \cite{telatar99} and
221: the machinery of the random matrix theory \cite{mehta91} can be employed
222: for further analysis of the problem. In particular, it is useful to rewrite
223: Eq.\ (\ref{c}) as
224: \begin{eqnarray}
225: C = \left< \sum\limits_{i = 1}^n \log_2 \left[ 1 + (S/N) \mu_i \right] \right>,
226: \label{c1}
227: \end{eqnarray}
228: where $\mu_i$ are the squares of the singular values of the matrix
229: $\mathbf{G}^+/ S^{1/2}$
230: with the joint probability density function \cite{mehta91}
231: \begin{eqnarray}
232: p_n(\mu_1, \ldots, \mu_n) = A_n
233: \exp\left( -n \sum\limits_{i=1}^n \mu_i \right)
234: \prod\limits_{i < j} (\mu_i - \mu_j)^2,
235: \label{rmt}
236: \end{eqnarray}
237: where $A_n$ is a normalization constant. The moment generating function $F(x)$ of
238: a random variable $c$, defined by Eq.\ (\ref{c1}) without
239: averaging, is obtained by averaging
240: $\exp(x c)$ with help of Eq.\ (\ref{rmt}) \cite{sengupta00},
241: and $C = \left< c \right>$ is then calculated as
242: \begin{eqnarray}
243: C = \frac{\mathrm{d}}{\mathrm{d} x} F(x) \Big|_{x = 0} =
244: \mathrm{Tr} \left( \mathbf{P}^{-1} \mathbf{R} \right),
245: \label{c2}
246: \end{eqnarray}
247: where $\mathbf{P}$ and $\mathbf{R}$ are $n \times n$ matrices:
248: \begin{eqnarray}
249: P_{i j} &=& (i + j - 2)!\, n^{1-i-j},
250: \label{p} \\
251: R_{i j} &=& \int_0^{\infty} \mathrm{d} \mu \,
252: \ln\left[ 1 + (S/N) \mu \right] \mu^{i+j-2} \exp(-n \mu).
253: \label{q}
254: \end{eqnarray}
255: Equations (\ref{c2})--(\ref{q}) provide an efficient way for capacity calculation
256: at arbitrary $n$ and $S/N$ (see the solid line in Fig.\ \ref{fig1}) and agree
257: perfectly with the direct Monte Carlo simulation of capacity using
258: Eq.\ (\ref{c}) (symbols in Fig.\ \ref{fig1}).
259: Note that the growth of $C$ with $n$ is linear at large $n$ and hence is
260: much faster than in the case (a).
261:
262: An alternative way of calculating capacity consists in performing averaging
263: directly in Eq.\ (\ref{c1}) using Eq.\ (\ref{rmt}). This yields
264: \begin{eqnarray}
265: C/n = \int_0^{\infty} \mathrm{d} \mu \log_2
266: \left[ 1 + (S/N) \mu \right] f_n(\mu),
267: \label{c3}
268: \end{eqnarray}
269: where
270: $f_n(\mu_1) = \int_0^{\infty} \mathrm{d} \mu_2 \ldots \mathrm{d} \mu_n \,
271: p_n(\mu_1, \ldots, \mu_n)$ can be evaluated by a direct integration, at least at
272: moderate $n$:
273: $f_1(\mu) = \exp(-\mu)$,
274: $f_2(\mu) = 2 \exp(-2 \mu) [1 - 2 \mu + 2 \mu^2]$,
275: $f_3(\mu) = 3 \exp(-3 \mu) [1 - 6 \mu + 18 \mu^2 - 18 \mu^3 + (27/4) \mu^4]$,
276: etc.
277: The values of capacity obtained then from Eq.\ (\ref{c3}) coincide exactly
278: with that following from Eq.\ (\ref{c2}).
279: An asymptotic expression for $f_n(\mu)$ at $n \gg 1$ can be
280: found in the framework of the random matrix theory \cite{mehta91,sengupta99}:
281: $f_{\infty}(\mu) = (2 \pi)^{-1} (4/\mu - 1)^{1/2}$
282: for $0 < \mu < 4$ and $f_{\infty}(\mu) = 0$ otherwise. Eq.\ (\ref{c3}) then yields
283: \begin{eqnarray}
284: C/n &=&
285: (S/N) \,_3F_2 \left( 1, 1, 3/2;\, 2, 3;\, -4\, S/N \right) / \ln 2,
286: \label{cinf}
287: \end{eqnarray}
288: where $_3F_2$ is the generalized hypergeometric function.
289: This result is shown in
290: the main plot of Fig.\ \ref{fig1} by a dashed horizontal line.
291: We find $C/n \simeq \log_2(S/N) - \log_2 e$ for $S/N \gg 1$
292: (see also the inset of Fig.\ \ref{fig1}).
293: It is worthwhile to note that $C/n$ decreases monotonically with $n$,
294: while the difference between the values of $C/n$ at $n=1$ and $n > 1$
295: never exceeds 7\%.
296:
297: \begin{figure}
298: \includegraphics[width=8cm]{fig2.ps}
299: \caption{\label{fig2} Average information capacity per antenna of
300: a communication channel between two identical linear arrays of $n$ antennas
301: in a disordered medium assuming antenna spacing $a$ and $S/N = 100$.
302: Solid lines correspond to $n=1$ (exact result), $n=2$ (Monte Carlo simulation)
303: and $n=100$ (asymptotic large-$n$ result).
304: Dashed line shows the result obtained using the asymptotic
305: large-$n$ theory at $n=2$.
306: Inset: Capacity per antenna as
307: a function of $n$ for different antenna spacings $a$. Results for
308: $ka = \pi/2$ and $\pi/4$ are asymptotic large-$n$ results.}
309: \end{figure}
310:
311: We now allow $k a$ to take arbitrary values, thus introducing
312: correlations between the entries of the Green matrix $\mathbf{G}$.
313: Eq.\ (\ref{c}) can again be reduced to Eq.\ (\ref{c1})
314: with $\mu_i$ denoting the squares of the singular values of the matrix
315: $\mathbf{Q}^{1/2} \mathbf{G}^+/ S^{1/2}$,
316: where $\mathbf{Q}$ is chosen to
317: maximize the result.
318: The joint probability density function of $\mu_i$,
319: analogous to Eq.\ (\ref{rmt}), is not known in this case,
320: but one can still calculate $C$ in
321: the large-$n$ limit \cite{mous00,sengupta00}.
322: The idea is to represent the moment generating function $F(x)$
323: of a random variable $c$, defined by Eq.\ (\ref{c1}) without averaging,
324: as a multiple Gaussian integral
325: (the so-called ``replica trick'') and then to perform the integrations using
326: saddle point methods in the limit $n \gg 1$. The maximization
327: of $C = (\mathrm{d} / \mathrm{d} x) F(x) |_{x = 0}$
328: over
329: $\mathbf{Q}$ is then
330: accomplished by requiring $\delta C \leq 0$ for all allowed small variations
331: $\delta \mathbf{Q}$ of the optimal matrix $\mathbf{Q}$.
332: This yields a system of nonlinear equations for the eigenvalues of
333: $\mathbf{Q}$ and some auxiliary variables that can be solved numerically.
334: We refer the reader to Refs.\ \onlinecite{sengupta00} and \onlinecite{sengupta99}
335: for an exhaustive account of the theoretical approach and the algorithm of
336: the numerical calculation.
337: The results that we obtained for identical linear arrays of equally-spaced
338: antennas are presented in Fig.\ \ref{fig2}.
339: As follows from the figure, at $ka > \pi$ correlations between
340: the entries of $\mathbf{G}$ are too weak to affect $C$ significantly and
341: the latter remains very close to its maximum value, given by Eqs.\ (\ref{c2})
342: or (\ref{c3}).
343: In contrast, correlations become important
344: when $ka$ decreases below $\pi$, leading to a significant drop of $C$.
345: The dashed line in Fig.\ \ref{fig2} results from the asymptotic large-$n$ theory
346: \cite{mous00,sengupta00,sengupta99} with $n = 2$ and is shown for
347: illustration purposes only. Its closeness to the Monte Carlo result
348: (solid line for $n=2$) testifies a qualitative validity of the large-$n$ theory
349: even at small $n$.
350: As $k a$ increases, the scaling of $C$ with $n$ changes from $C \propto \ln n$ at
351: $ka=0$ to $C \propto n$ at $ka \agt \pi$
352: (see the inset of Fig.\ \ref{fig2}). We note that even
353: at $0 < ka < \pi$,
354: there is still a significant gain in capacity as compared to the case of
355: $ka = 0$: e.g., at $n = 100$, $C/n$ for $ka = \pi/2$ ($\pi/4$)
356: is almost 20 (13) times larger than for $ka = 0$.
357:
358: In conclusion, we have studied the information capacity $C$ of a wireless
359: communication channel in a disordered medium, assuming multiple diffuse scattering
360: of waves that carry information, and taking into account fluctuations of
361: wave fields in space (speckles).
362: Although multiple scattering reduces
363: the received signal, it allows for a dramatic increase of capacity in the case of
364: communication between two antenna arrays, provided that antenna
365: spacing $a \agt \lambda/2$, where $\lambda$ is the wavelength. Namely,
366: for identical linear arrays of $n$ equally spaced receiving/transmitting antennas,
367: scaling of $C$ with $n$ changes from $C \propto \ln n$ for $kd \ll \pi$
368: to $C \propto n$ for $ka \agt \pi$, where $d = a (n-1)$ is the array size and
369: $k = 2 \pi / \lambda$. Even at $0 < ka < \pi$ an important
370: gain in capacity is possible as compared to $ka = 0$.
371:
372: The author is indebted to Prof. R. Maynard for helpful discussions
373: and critical reading of the manuscript.
374:
375: \begin{thebibliography}{99}
376:
377: \bibitem{mous00}
378: A.L. Moustakas, H.U. Baranger, L. Balents, A.M. Sengupta, S.H. Simon,
379: Science \textbf{287,} 287 (2000).
380:
381: \bibitem{andrews01}
382: M.R. Andrews, P.P. Mitra, and R. deCarvalho,
383: Nature \textbf{409,} 316 (2001).
384:
385: \bibitem{simon01}
386: S.H. Simon, A.L. Moustakas, M. Stoytchev, and H. Safar,
387: Phys. Today \textbf{54}(9), 38 (2001).
388:
389: \bibitem{shannon48}
390: C.E. Shannon,
391: Bell Syst. Tech. J. \textbf{27,} 379 \& 623 (1948).
392:
393: \bibitem{cover91}
394: T.M. Cover and J.A. Thomas,
395: Elements of Information Theory (Wiley, N.Y., 1991).
396:
397: \bibitem{foschini98}
398: G.J. Foschini and M.J. Gans,
399: Wireless Personal Communications \textbf{6,} 311 (1998).
400:
401: \bibitem{ishim78}
402: A. Ishimaru,
403: Wave Propagation and Scattering in Random Media
404: (Academic, N.Y., 1978).
405:
406: \bibitem{shapiro86}
407: B. Shapiro,
408: Phys. Rev. Lett. \textbf{57,} 2186 (1986).
409:
410: \bibitem{telatar99}
411: I.E. Telatar,
412: Eur. Trans. Telecommun. \textbf{10,} 585 (1999).
413:
414: \bibitem{mehta91}
415: M.L. Mehta,
416: Random Matrices (Academic, N.Y., 1991).
417:
418: \bibitem{sengupta00}
419: A.M. Sengupta and P.P. Mitra,
420: physics/0010081 (2000).
421:
422: \bibitem{sengupta99}
423: A.M. Sengupta and P.P. Mitra,
424: Phys. Rev. E \textbf{60,} 3389 (1999).
425:
426: \end{thebibliography}
427:
428: \end{document}
429: %
430: % ****** End of file com.tex ******
431:
432: