1: \documentclass[twocolumn]{revtex4}
2: \usepackage{graphicx}
3: \begin{document}
4: \title{Entangled states of light}
5: \author{S.J. van Enk}
6: \affiliation{Bell Labs, Lucent Technologies\\
7: 600-700 Mountain Ave,
8: Murray Hill, NJ 07974}
9: \affiliation{Institute for Quantum Information,\\
10: California Institute of Technology,\\
11: Pasadena, CA 91125
12: }
13:
14: \date{\today}
15:
16:
17: \begin{abstract}
18: These notes are more or less a faithful representation of
19: my talk at the Workshop on ``Quantum Coding and Quantum Computing''
20: held at the University of Virginia. As such it is an introduction
21: for non-physicists to the topics of the quantum theory of light and
22: entangled states of light.
23: In particular, I discuss the photon concept
24: and what is really entangled in an
25: entangled state of light (it is not the photons!).
26: Moreover, I discuss an example that highlights
27: the peculiar behavior of entanglement in an infinite-dimensional
28: Hilbert
29: space.
30: \end{abstract}
31:
32: \maketitle
33: \section{Light}
34: Thanks to a certain class of T-shirts \cite{T} we all know that light
35: is described by Maxwell's equations. These equations
36: determine how the electric and magnetic fields vary in space and time, given
37: electric currents and charges.
38: Fortunately, even if there are no sources,
39: that is, no currents and no charges,
40: there are still
41: nontrivial solutions. It is those solutions that describe light waves.
42: One way to express such a light wave is to write
43: down the electric field vector
44: as a function of position $\vec{r}$ and time $t$,
45: \begin{equation}
46: \vec{E}(\vec{r},t)=
47: \vec{\epsilon}
48: \exp i\big(\vec{k}\cdot\vec{r}-\omega t\big)+{\rm complex}
49: \;{\rm conjugate}.
50: \end{equation}
51: This wave corresponds to a monochromatic plane
52: wave with frequency $\omega$,
53: propagating in the direction
54: of $\vec{k}$, and with a polarization $\vec{\epsilon}$.
55: These quantities are not independent as they satisfy
56: \begin{equation}
57: |\vec{k}|c=\omega;\,\,\vec{k}\cdot\vec{\epsilon}=0.
58: \end{equation}
59: Here $c$ is the speed of, what else, light.
60: The general solution to the source-free Maxwell equations is a sum
61: (or an integral, depending on whether we allow $\vec{k}$ to take
62: continuous values)
63: of
64: plane waves, according to
65: \begin{eqnarray}\label{sol}
66: \vec{E}(\vec{r},t)&=&\sum_{\vec{k}}\sum_{\vec{\epsilon}\perp\vec{k}}
67: A_{\vec{k},\vec{\epsilon}}\,
68: \vec{\epsilon}
69: \exp i\big(\vec{k}\cdot\vec{r}-\omega t\big)\nonumber\\
70: &&+{\rm complex}
71: \;{\rm conjugate},
72: \end{eqnarray}
73: where $A_{\vec{k},\vec{\epsilon}}$ are arbitrary complex
74: coefficients, called amplitudes, that
75: determine the intensity of the light beam.
76:
77: The Maxwell equations and hence its solutions (\ref{sol})
78: are ``classical'', as
79: they do not describe any quantum effects.
80: So, what changes if we are to use a quantum theory of light?
81: Some may think that the frequency $\omega$ becomes quantized, that is,
82: takes on only discrete values. Others may think the polarization vector
83: $\vec{\epsilon}$
84: becomes a quantum object. Pessimists may think that the Maxwell equations
85: are no longer valid and perhaps polarization and frequency are no longer well
86: defined. But none of these expectations are correct.
87: What really happens when one quantizes
88: the electromagnetic field is the following:
89: Just as position and momentum of a particle become
90: operators acting on some complex Hilbert space
91: when one quantizes the particle, so do
92: the electric (and magnetic) fields become operators in a quantum theory of
93: light. It turns out that the Maxwell equations keep in fact the same form.
94: So, the solutions (\ref{sol}) are also still valid,
95: except that ``complex conjugate'' becomes ``Hermitian conjugate.''
96:
97: The difference with the classical solutions (\ref{sol}) is that the
98: amplitudes $A_{\vec{k},\vec{\epsilon}}$ become operators.
99: We typically write the amplitude as the product of a number
100: (with the dimension of an electric field), and a dimensionless operator
101: $a_{\vec{k},\vec{\epsilon}}$. For each value of $(\vec{k},\vec{\epsilon})$,
102: called a {\em mode},
103: there is such an operator $a$
104: (leaving out the subscripts for ease of notation).
105: That operator acts on a complex Hilbert space. In the
106: case of the electromagnetic field, that Hilbert space is infinite-dimensional.
107: It is spanned by states $|n\rangle$ for $n=0,1,2\ldots$, and the operator $a$
108: acts as
109: \begin{equation}\label{a}
110: a|n\rangle=\sqrt{n}|n-1\rangle.
111: \end{equation}
112: The number $n$ is interpreted as the number of photons in the
113: corresponding mode.
114: Thus the state $|2\rangle$ is a state with two photons in a particular mode.
115: The operator $a$ annihilates one photon, and is called the annihilation operator.
116: Similarly, the Hermitian conjugate operator, $a^{\dagger}$ acts as
117: \begin{equation}\label{ad}
118: a^{\dagger}|n\rangle=\sqrt{n+1}|n+1\rangle,
119: \end{equation}
120: and is called the creation operator.
121: The fact that the number states $|n\rangle$ start with
122: $n=0$ is not just a convenient convention:
123: Eq.~(\ref{a}) shows that $a$ acting on the state $|0\rangle$
124: gives nothing, zero, not even a state.
125: We cannot have a state with $-1$ photons.
126:
127: For later use we note that if we define $N=a^\dagger a$, we
128: have in fact managed to define the number operator,
129: \begin{equation}\label{N}
130: N|n\rangle=n|n\rangle,
131: \end{equation}
132: that counts the number of photons in a mode.
133: \subsection{Entangled states of light}
134: To make an entangled state of light we need (at least) two modes.
135: Suppose we have two modes, denoted by $A$ and $B$.
136: Modes $A$ and $B$ may correspond to modes with
137: the same frequency and the same wave vector,
138: but different (orthogonal) polarizations.
139: Or they may have the same polarization and the same frequency,
140: but propagate in a different (not necessarily orthogonal) direction.
141: In any case, the Hilbert space $H_{A,B}$
142: associated with the
143: two modes is the tensor product $H_{A,B}=H_A\otimes H_B$ of the Hilbert spaces
144: of each of the modes.
145:
146: Let us look at a simple example of a state in $H_{A,B}$:
147: \begin{equation}\label{ex1}
148: |\Psi\rangle_{A,B}=[|0\rangle_A|1\rangle_B+|1\rangle_A|0\rangle_B]/\sqrt{2}.
149: \end{equation}
150: Although properly we should have written $|0\rangle_A\otimes |1\rangle_B$ etc., we used here the
151: physicists' convention to be lazy.
152: This state is entangled because it cannot be written as a
153: product of two states of the modes
154: $A$ and $B$. Thus neither mode is in a well-defined pure state of itself.
155: Loosely speaking, we could say there is one photon, and it is either in mode $A$ or in mode $B$.
156: This way of speaking misses some essential points though. The state (\ref{ex1})
157: expresses more than just a classical joint probability distribution to find the photon
158: in modes $A$ and $B$. In particular, if modes $A$ and $B$ are spatially separated, then
159: the statistics of measurement outcomes on the state (\ref{ex1}) can in general
160: not be obtained from any classical
161: joint probability distribution without superluminal communication.
162: Since physicists don't like superluminal communication, this expresses a serious difference
163: between quantum and classical mechanics.
164: [Note that this does not imply that entanglement can be used for
165: superluminal communication, it is just that a classical local model
166: cannot explain quantum mechanics.]
167: This basically is the content of Bell's theorem.
168: I refer to Bell's book \cite{bell} for the details.
169: \subsection{What is entangled?}
170: We found that the state (\ref{ex1}) is entangled. But what is entangled
171: with what?
172: One might think photons are entangled with each other.
173: That is indeed not the most stupid answer one can give, as after all light
174: consists of photons. Nevertheless,
175: it is not the photons
176: that are entangled. In fact, there is only one photon in the state
177: (\ref{ex1}).
178: Instead, modes $A$ and $B$ are entangled with each other.
179: The number of photons just determines the
180: states of the modes.
181:
182: Now, however, this leads to a slight problem.
183: After all, nobody tells us how to define our modes. We can take any
184: complete set of functions to expand our electric field in, as in Eq.~(\ref{sol}),
185: and to each member of that set corresponds a mode.
186: So unlike in the case of, say, atoms in a trap, we are more or less free to define what we mean by our systems.
187: With atoms in a trap, the atoms are clearly the systems, but for light modes it is a lot less clear.
188: For instance, if modes $A$ and $B$ correspond to horizontal and vertical polarization, respectively
189: (with all other mode variables, the wave vector and the frequency, the same), then the entangled state (\ref{ex1})
190: turns out to be equal to a state with one diagonally polarized photon.
191: Formally, by defining new annihilation operators
192: $a'_{\pm}=(a_A\pm a_B)/\sqrt{2}$, we find that there is one photon in mode $'+'$ and no photons in mode $'-'$.
193: However, that state we would never call entangled.
194:
195: On the other hand, suppose modes $A$ and $B$ correspond to spatially separated modes, mode $A$ is here,
196: mode $B$ is there. The $\pm$ modes can theoretically still be defined. However, in practice the modes
197: ``here plus there''
198: and ``here minus there'' are useless.
199: In particular, noone would know how to directly measure a photon in such a mode.
200: (Of course, one can always indirectly measure such modes by recombining the modes so that
201: they end up in the same location. But that's cheating as one explicitly changes the modes and thereby the
202: entanglement.) Thus, when the modes $A$ and $B$ are spatially separated no redefinitions of those modes are useful.
203: So we decree that it makes sense to talk unambiguously about entanglement between two
204: modes $A$ and $B$ only if the modes are spatially
205: separated. If the modes are in the same location we do allow redefinitions of the modes and
206: thereby of the entanglement.
207: \subsection{Some remarks}
208: The importance attached to nonlocality for the issue of defining entanglement is consistent with
209: several other results: first of all,
210: there is above-mentioned Bell's theorem. Second,
211: there has been a lot of interest in relativistic aspects of entanglement of both light and particles,
212: such as electrons. One conclusion \cite{scudo},
213: in particular, is that the entanglement between different degrees of
214: freedom of one particle (namely, the spin and the momentum) is not Lorentz-invariant. That is,
215: the amount of entanglement present within that one particle depends on the velocity of the observer. The reason is that different observers don't agree on what
216: the spin and momentum degrees are.
217: This is similar to having different observers disagreeing
218: on how the modes of the electromagnetic field are defined.
219: When talking about two particles in different positions, no such problems arise.
220: Third, it has been shown \cite{spreeuw} that many quantum properties of light, even including
221: entanglement,
222: can be simulated by using just classical light beams. However,
223: this is only so if the entanglement is not nonlocal.
224:
225: The point that entanglement is between modes and not between photons was made in several different papers \cite{pz,em},
226: all coming out at roughly the same time. Apart from making that same point,
227: the papers do have different ideas otherwise.
228: In particular, my own attempt \cite{em} at a paper discusses by how much
229: redefinitions of modes can change the entanglement. For example, in the simple
230: state (\ref{ex1}), the smallest and largest possible amounts of entanglement one can get
231: are just zero, and one unit of entanglement, respectively.
232: There are other simple examples where the entanglement cannot be
233: transformed away completely, and also examples where the maximum
234: entanglement possible is probably more than what one would have expected at first and even second glance.
235:
236: Finally, I also note that the importance of what can actually be measured
237: for the definition of entanglement was discussed in a more general
238: and more precise context in \cite{barnum}. This was also discussed at the
239: Workshop by Lorenza Viola.
240: \section{Entangled coherent states}
241: This Section serves a number of purposes. First, it shows an explicit
242: nontrivial (as opposed to (\ref{ex1})) example of an entangled state.
243: Second, that example will display some curious behavior that is known to
244: occur in principle
245: in infinite-dimensional Hilbert spaces, but for which there was no more or less
246: practical example known.
247:
248: The easiest way to generate a light beam with nice quantum properties is to switch on a laser.
249: The quantum state of the light beam is a so-called coherent state
250: [disregarding some subtleties that are of no interest here].
251: For any complex number $\alpha$ it is defined as
252: \begin{equation}\label{coh}
253: |\alpha\rangle:=\exp(-|\alpha|^2/2)
254: \sum_{n=0}^\infty \frac{\alpha^n}{\sqrt{n!}}|n\rangle,
255: \end{equation}
256: in terms of the number states $|n\rangle$. The easiest way to produce a nontrivial state of two modes, is to take
257: a laser beam and split it on a beam splitter.
258: Annoyingly, though, one can never get an entangled state out of these two easy operations.
259: If we split two coherent states on a beam splitter we just get two new coherent states with different amplitudes,
260: but nothing more.
261: In fact, more generally speaking, with linear optics one cannot create entanglement from coherent states.
262: The reason is that all linear optics elements just lead to linear transformations of operators $a_i$ of modes $i$.
263: Namely, the general transformation is of the form
264: \begin{equation}\label{trans}
265: a'_j=\sum_i U_{ij}a_j,
266: \end{equation}
267: with $U_{ij}$ a unitary matrix. Since a coherent state (\ref{coh}) is an eigenstate of $a$, and since
268: the transformations
269: (\ref{trans}) can only transform eigenstates of the operators $a_i$ into eigenstates of new mode operators
270: $a'_j$, coherent states are always transformed into coherent states.
271:
272: Thus, in order to create entanglement from a coherent state one needs nonlinear optics.
273: One particular type of nonlinearity I consider here is the Kerr nonlinearity. It is characterized by a
274: Hamiltonian of the form
275: \begin{equation}\label{H}
276: H=\hbar \chi a^{\dagger2}a^2,
277: \end{equation}
278: where $\hbar$ is Planck's constant.
279: The Hamiltonian determines the evolution in time of any state, with
280: the evolution operator as a function of time $t$ given by
281: \begin{equation}
282: U(t)=\exp(-it H/\hbar).
283: \end{equation}
284: Starting from a coherent state $|\alpha\rangle$ at time $t=0$
285: the state evolves thus as
286: \begin{equation}
287: |\psi(t)\rangle=\exp(-it H/\hbar)|\alpha\rangle.
288: \end{equation}
289: The evolution operator can be rewritten in terms of the number operator
290: defined in (\ref{N}),
291: \begin{equation}
292: U(\tau)=\exp(-i \tau a^{\dagger2}a^2)=
293: \exp(-i \tau N(N-1)),
294: \end{equation}
295: where $\tau=\chi t$ is a dimensionless time.
296: The operator $U$ becomes periodic in $N$ with period $M$
297: (that is, it becomes invariant under $N\rightarrow N+M$)
298: at times $\tau=\pi/M$ if $M$ is an odd integer.
299: This implies one can write down Fourier series as follows
300: [following Ref.~\cite{tara}]
301: \begin{equation}\label{odd}
302: \exp\big( \frac{-i\pi}{M}N(N-1) \big)=\sum_{q=0}^{M-1} f_q^{(o)}\exp
303: \big( \frac{-2i\pi q}{M}N\big).
304: \end{equation}
305: Similarly, for even values of $M$ one has
306: \begin{equation}\label{even}
307: \exp\big( \frac{-i\pi}{M}(N+M)^2 \big)=
308: \exp\big( \frac{-i\pi}{M}N^2 \big),
309: \end{equation}
310: so that we can expand
311: \begin{equation}\label{Feven}
312: \exp\big( \frac{-i\pi}{M}N^2 \big)=\sum_{q=0}^{M-1} f_q^{(e)}\exp
313: \big( \frac{-2i\pi q}{M}N\big).
314: \end{equation}
315: The coefficients $f_q$ are not explicitly evaluated in
316: Ref.~\cite{tara}, but one can
317: actually derive them (see \cite{multi} for more details),
318: \begin{eqnarray}\label{MM}
319: f_q^{(o)}&=&\frac{1}{\sqrt{M}}\exp\big(\frac{\pi iq(q+1)}{M} \big)
320: \exp\big(\frac{-\pi iK(K+1)}{M} \big),\nonumber\\
321: f_q^{(e)}&=&\frac{1}{\sqrt{M}}\exp\big(\frac{\pi iq^2}{M} \big) \exp(-\pi i/4),
322: \end{eqnarray}
323: where in the first line $K$ is such that $M=2K+1$ for odd $M$.
324: The expansions (\ref{odd}) and (\ref{even}) are
325: useful as it becomes easy to calculate the effect of the
326: evolution opertor on a coherent state. The reason is the simple relation
327: \begin{equation}
328: \exp(i\phi N)|\alpha\rangle=|\alpha \exp(i\phi)\rangle.
329: \end{equation}
330: Thus, if one starts with a coherent state
331: at time $\tau=0$,
332: this then immediately leads to the following time evolution under $U$:
333: \begin{eqnarray}
334: U(\pi/M)|\alpha\rangle
335: &=&\sum_{q=0}^{M-1}f_q^{(o)}|\alpha\exp(-2\pi iq/M)\rangle\nonumber\\
336: {\rm for}\,\,M\,\,{\rm odd},\nonumber\\
337: U_A(\pi/M)|\alpha\rangle
338: &=&\sum_{q=0}^{M-1}f_q^{(e)}|\alpha\exp(\pi i(1-2q)/M)\rangle\nonumber\\
339: {\rm for}
340: \,\,M\,\,{\rm even}\end{eqnarray}
341: If one subsequently takes these states and splits them on a 50/50 beamsplitter with the vacuum, the output state is an entangled state of the form
342: \begin{equation}\label{ent2}
343: |\Phi_M\rangle=\sum_{q=0}^{M-1}f_q^{(o)} |\beta\exp(-2\pi iq/M)\rangle
344: |\beta\exp(-2\pi iq/M)\rangle,
345: \end{equation}
346: for $M$ odd
347: with $\beta=\alpha/\sqrt{2}$, and something similar for even $M$.
348:
349: The state (\ref{ent2}) is entangled. A measure for how much
350: entanglement there
351: is in a state of two modes can be obtained after first
352: writing the state in a standard form,
353: the so-called Schmidt-decomposition form.
354: There is a unique
355: (up to some trivial transformations) way to write a bipartite state as
356: \begin{equation}\label{S}
357: |\Psi\rangle=\sum_{q=0}^{M-1} \sqrt{p}_q |\psi_q\rangle|\phi_q\rangle,
358: \end{equation}
359: where the $p_q$ are positive real numbers with $\sum_q p_q=1$
360: (so they can be interpreted as probabilities), and with
361: the states $|\psi_q\rangle$ for $q=0\ldots M-1$ all orthogonal to each other,
362: and the same for $|\phi_q\rangle$.
363: The entanglement is then given by
364: \begin{equation}
365: E=-\sum p_q\log_2p_q,
366: \end{equation}
367: which some may recognize as the Shannon entropy of the probability distribution $\{p_q\}$.
368: If we have $M$ terms in (\ref{S}) then the entanglement can at most be $\log_2 M$.
369: Now the state (\ref{ent2}) is almost written in the Schmidt form (\ref{S}). It is only ``almost'', as
370: I had not told you yet
371: that the coherent states are not orthogonal, as can be checked directly
372: from the definition.
373: However, for large values of $|\alpha|$ the different
374: coherent states do become orthogonal,
375: so in that limit the entanglement in the state (\ref{ent2}) is
376: $E=\log_2 M$.
377: \begin{figure}
378: \includegraphics[scale=0.4]{eeeob}
379: \caption{Entanglement for the states (\ref{ent2}) as a function of $\tau$, for
380: various values of $|\alpha|^2$: $|\alpha|^2=1$ for the bottom curve (circles), $|\alpha|^2=10$
381: for the middle curve (crosses), and $|\alpha|^2\rightarrow\infty$ for the top curve (solid line).}
382: \end{figure}
383:
384: Now here is a peculiar effect: we get more entanglement with increasing $M$, that is, if we remember we
385: generated this state after a time $\tau=\pi/M$, with {\em decreasing} time.
386: So the shorter the nonlinear Hamiltonian (\ref{H}) acts,
387: the more entanglement we get (and we started with no entanglement).
388: How can this possibly be right?
389: The answer lies hidden in the assumption of large $\alpha$:
390: for increasing $M$ the different coherent states are only orthogonal when
391: $|\alpha/M|\gg 1$. So, with increasing $M$ one needs ever larger values of $\alpha$.
392: Now the energy in a coherent state
393: is proportional to $|\alpha|^2$, so the energy needed to make $\log_2 M$ units of entanglement grows as $M^2$.
394: Now it is known \cite{wehrl}
395: that entanglement in infinite dimensions is not a continuous function of the state.
396: But, if one restricts oneself to states with an upper bound on the total energy, then that function is continuous.
397: Here too, the peculiar behavior of entanglement disappears
398: when we {\em fix} the value of $\alpha$ and then calculate the entanglement as a function of time $\tau$.
399: For small $\tau$ the entanglement will be small: the above-mentioned argument that
400: the entanglement should be large does not hold for fixed $\alpha$ as the coherent states
401: become very non-orthogonal for large $M$.
402: This is illustrated in Fig.~1, where the entanglement as a function of $\tau$ is plotted
403: for 2 different values of $\alpha$ together with the maximum possible amount of entanglement
404: (only reached for $|\alpha|\rightarrow\infty$).
405:
406: A final question to be answered is the following.
407: Even though we know now that there is no real contradiction arising
408: from the fact that in principle a large amount of entanglement can
409: be created after a short interaction time, one may wonder whether that effect can be useful for creating lots of entanglement without being bothered too much by decoherence.
410: After all, a shorter interaction time implies in general
411: less decoherence.
412: If you fund my research, please stop reading now.
413: Unfortunately, there is a catch.
414: One only creates those large amounts of entanglement in short times
415: if the amplitude $\alpha$ is large. That in turn leads to more decoherence.
416: The more photons there are in a state, the easier they are lost.
417: A more precise analysis shows that the increasing decoherence indeed
418: seriously counteracts
419: the potential increase of entanglement. Thus,
420: it seems one is not better off using very short interaction times to create entanglement with the Kerr interaction.
421: On that happy note, I end.
422: \section{Acknowledgments}
423: I thank Olivier Pfister for having been a great host and Karen Klintworth for
424: having taken care of all the annoying but important details.
425:
426:
427: \begin{thebibliography}{99}
428: \bibitem{T}{\tt www.twobarkingdogs.com}.
429: And God said, Let there be Light.
430:
431:
432: \bibitem{bell}J.S.~Bell, {\em Speakable and unspeakable in quantum mechanics},
433: Cambridge University Press, 1993.
434:
435: \bibitem{scudo}A. Peres, P.F. Scudo, and D.R. Terno,
436: Phys. Rev. Lett. {\bf 88}, 230402 (2002).
437:
438: \bibitem{spreeuw}R.J.C. Spreeuw, Found. Phys. {\bf 28}, 361 (1998);
439: Phys. Rev. A {\bf 63}, 062302 (2001).
440:
441:
442: \bibitem{pz}P. Zanardi, Phys. Rev. A {\bf 65}, 042101 (2002);
443: Y. Shi, Phys. Rev. A {\bf 67}, 024301 (2003);
444: V. Vedral, {\tt quant-ph/0302040}.
445:
446: \bibitem{em}S.J. van Enk, Phys. Rev. A {\bf 67}, 022303 (2003).
447:
448: \bibitem{barnum}H. Barnum {\em et al.}, {\tt quant-ph/0305023}.
449:
450:
451: \bibitem{tara}K.~Tara, G.S.~Agarwal, and S.~Chaturvedi,
452: Phys. Rev. A {\bf 47}, 5024 (1993).
453:
454: \bibitem{multi}S.J. van Enk, Phys. Rev. Lett. {\bf 91}, 017902 (2003).
455:
456: \bibitem{wehrl}A.~Wehrl, Rev. Mod. Phys. {\bf 50}, 221 (1978);
457: see also J.~Eisert, C.~Simon, and M.B.~Plenio,
458: J. Phys. A {\bf 35}, 3911 (2002).
459: \end{thebibliography}
460: \end{document}
461:
462:
463:
464: