VOLUME 74, NUMBER 24 PH YSICAL REVIEW LETTERS 12 JUNE 1995

Photonic de Broglie Waves

Joseph Jacobson, * Gunnar Bjork, Isaac Chuang, and Yoshihisa Yamamoto~ERATO Quantum Fluctuation Project, Edward L Ginzton Laboratory, Stanford University, Stanford, California 94305

(Received 25 January 1995; revised manuscript received 10 April 1995)

In quantum optics an ensemble of photons is treated as a Bose condensate which has a de Brogliewavelength given by Ao/N where Ao and N are the wavelength and average number of constituentphotons, respectively. We describe an interferometer which is capable of measuring this de Brogliewavelength. We show specifically that a coherent state input has a de Broglie coherence lengthproportional to I/~N. Finally we show that the interferometer is Heisenberg limited in sensitivity(AP;„= 1/N) and thus forms a new class of interferometers operating at the fundamental quantumlimit.

PACS numbers: 42.25.Hz, 42.50.Ar

The advent of quantum mechanics requires us to take anoperational view of nature in which we describe propertiesof a system in terms of measurable observables. In thispaper we show that the measured wavelength of an objectis, fundamentally, dependent on the internal structure ofthe object being measured. In general, the wavelength ismeasured by splitting an incident object into a coherentsuperposition of states and interfering those states witheach other as a function of phase delay. In order to treatthe case of incident objects with internal structure weintroduce the concept of "effective" beam splitters whichimbue an effective internal structure such as bindingbetween constituent parts of the object. In a simple casewe show that such an effective internal structure mayhave dramatic effects, for instance, that incident statesof definite energy do not necessarily have a well-definedde Broglie wavelength. We then extend this conceptof effective beam splitters to construct an interferometerwhich is capable of measuring the de Broglie wavelengthof an incident state of photons (light packet) as a whole.

Consider an interferometer for molecules as has beenrecently demonstrated [1]. The output of the inter-ferometer oscillates as a function of the path lengthdifference. The oscillation period is a measure of themolecule's de Broglie wavelength. In this paper we weremotivated by considering what happens to the measuredwavelength as a function of the binding energy of themolecule as follows: Assume an iodine molecule I2 withmass =2mt (ignoring the mass contribution of binding).The de Broglie wavelength of the molecule, which weexpect to measure, is given by h/2mtv where v is themolecular velocity. However, in the limit that the bind-ing energy between the two I atoms goes to zero themolecule's constituent atoms become disassociated andwe effectively have two separate I atoms each with deBroglie wavelength h/mt v incident on the interferometer.In this case we expect the measured wavelength to indi-cate the de Broglie wavelength of the individual atoms.At what point does the molecular object make a transfor-mation from a wavelength of h/2miv to a wavelength ofh/mi u?

We may construct a simplified model of the molecularinterferometer by considering a modified Mach-Zehnderinterferometer with effective beam splitters [denoted byE in Fig. 1(inset)] which imbue an effective binding tothe constituent particles of the incident state. The inputstate is written as ii/to) = i2, 0). The first label of the ketdenotes the number of constituent particles incident on thehorizontal port and the second label indicates the numberof constituent particles incident on the vertical port. In ourmodel the interaction Hamiltonian for the effective beamsplitter is given by

Hb, =ih y[a b —b a]4

+ ih (1 —g) [(atb) —(bta) ],8

where a (at) and b (bt) are the lowering (raising) bosonoperators corresponding to the different paths of the inter-ferometer. The unitary time propagation operator for thebeam splitter is given by Ub, = e 'H"'/~. A 50/50 beamsplitting is achieved for a propagation from t = 0 to t = 1.

2.0

1.5A

~ 1 ~ 0V

0.5

0 ~ 00 ' 0 0 ' 5

WA

l2& jy E

10& 4r1 ' 0 1 ' 5 2 ' 0 2.5 3.0Phase t'ai (units of z)

3 ' 5

FIG. 1. (Inset) A modified Mach-Zehnder interferometer witheffective beam splitters. (Plot) The expectation number (ata)as a function of the phase P for various values of the parameter

(g = 1, no binding) corresponds to a beam splitter whichsplits incident composite objects into constituent parts. (y = 0,tight binding) corresponds to a beam splitter which does notsplit incident composite objects into constituent parts. Theincident state is i2, 0).

0031-9007/95/74(24)/4835(4)$06. 00 1995 The American Physical Society 4835

VOLUME 74, NUMBER 24 PH YS ICAL REVIEW LETTERS 12 JUNE 1995

A

le& =

z/2 z/2

,JC 3Ce g

z/2

t0&

FIG. 2. Schematic of a Mach-Zehnder interferometer formeasuring the de Broglie wavelength of an optical pulse.

4836

g is a parameter on [0, 1] which changes the beam splittercontinuously from one which imbues an effective binding(i.e., does not split composite objects into constituent parts;y = 0) to one which does not imbue an effective binding(i.e., normal beam splitter which does split composite ob-jects into constituent parts; ~ = 1).

The phase shift operator is written simply as the freespace propagator for one arm of the interferometer and isgiven by

—iCuOa at —i Pa aUy

where coo is the frequency of the constituent particles(= 2rrc/Ao) and Al is the path length difference.cook/ /c.

Figure 1 shows the output (expectation value of theparticle number (at a&) as a function of the phase @ forseveral different values of the functional splitting param-eter ~. It may be discerned from the figure that as thebeam splitter is transformed from a normal beam splitterwhich splits incident composite objects into constituentparts (~ = 1, no binding) to an effective beam splitterwhich does not split incident composite objects into con-stituent parts (y = 0, tight binding) the measured wave-length is halved (e.g. , the measured frequency is doubled).We note that at intermediate "binding energies, " despitethe fact that the incident object has a well-defined energy,the wavelength is nonetheless ill defined.

Recently, Davidovich er al. [2] described an ingeniousquantum switch for creating a superposition between anarbitrary incident state of light (e.g. , coherent state lightpacket) and the vacuum by means of a microwave cavitywhose resonance is dependent upon the state of a singleatom in the cavity. In this paper we use two such switchesto construct an interferometer for states as a whole.

Consider the schematic setup of Fig. 2. Here an atomincident on two cavities controls the optical properties ofthese cavities depending on whether or not the atom is inthe ground or upper state due to a state dependent indexof refraction. In our setup we consider a situation wherean atom in the upper state causes the incident photonfield to be rejected, whereas an atom in the ground stateallows the incident field to be transmitted. Consider anupper-state atom incident on the first cavity. The atom issubjected to a vr/2 pulse which transforms an excited stateatomic wave function according to le) ~z (le) + lg))and a ground state atomic wave function according to

lg) ~ (lg) —le)). Simultaneously, a coherent state1

la) = e ~ g„(n"/n!'~ )In) and the vacuum state I0)are incident on the two optical input ports of the cavity.

We may write our incident wave function as I Po) =In, O, e) in the basis la, b, x), where a and b are thetwo photon modes and x denotes the state of the atom(e excited, g ground). Therefore after both the initialrr/2 pulse and the first cavity we have the state IP& =~ (IO, a, e) + In, O, g)), where the labels in the first keton the right side have been exchanged due to the action ofthe excited state atom in the cavity in which the presenceof an excited state atom in the cavity causes the incidentphoton state to be exchanged from mode a to mode b andvice versa (i.e., reflected), whereas the presence of a groundstate atom causes no mode exchange (i.e., transmitted).

As before the phase shift operator is given by Eq. (2).Thus after the initial m. /2 pulse, the first cavity and aphase shift P along mode a we have that the state ifgiven by IP) = ~2(IO, ae& + Ine '~, O, g&), where we

have used e '&' 'Ia) = Ine '~).Using the rules above we have that after traversing the

rest of the apparatus (Fig. 2), consisting of another vr/2pulse followed by the second cavity and finally followedby a finishing vr/2 pulse, the final state is given by

(I ~, o, e& +I ~, o, g & +

I o, ~, g&—Io, ~, e&

2 2+ Ine '~, O, g) —Ine '@, O, e&

—IO, ne '~, e) —IO, ne '~, g&). (3)

Consider the case where we measure the final stateof the atom to be in the upper state le). We have theprojected photon field to be

1Iyf&, = —(I~, O) —IO, ~& —l~e '~, 0) —IO, ~e '~&),

and the expected output at port A for our interferometer is

e(l/Jf la aI Pf ), = —[1 —e '" ' cos(P + N sing)],

where here the average photon number N = Inl2. As anaside we note that if we were simply to trace over the finalatomic state, instead of conditioning the photon outputat port A on a specific atom state as we have done, theoutput of the interferometer would be, (t/lflatalff&, +g(Pf Ia aloof)g = N yielding no interference.

The upper-state conditioned interferometer output ofEq. (5) is plotted as a function of @ in Figs. 3(a, solid line)and 3(b). In these plots the average photon number is N =100. From Fig. 3(a), solid line, we see that the output hasan oscillation period of 2'/N indicating a measurement ofthe de Broglie wavelength (Ao/N) of the light packet as awhole. We may understand the decay of the interferencefor long phase delays [large P —see Fig. 3(a), solid line]by recalling that the de Broglie wavelength is given by themean number of photons, W, and therefore the "coherence

VOLUME 74, NUMBER 24 PH YS ICAL REVIEW LETTERS 12 JUNE 1995

80A6$65 40V

I

(a)

0 I I I

0.00 0.02 0.04 0.06 0.08 0.10 0.12Phase 1i1 (units of z)

80—A

~ 40-V

(b)

0-0

I I I I

2 3 4 5

Phase P (units of ~)

80AC5

05 40V

~ ~ 4~ ~

00 ~ 00 0 ~ 02 0.04 0 ~ 06 0 ~ 08

Phase 1i1 (units of0 ~ 10

z)0.12

length" of the light packet, Al„ is a function of thefluctuations in photon number, AN, which for a coherentstate is given by ~N. The dashed line in Fig. 3(a) is aplot of the function z (1 —e ~~& cosN@) and indicatesthat the coherence length of the coherent state light packetis given approximately by Al, = c/cuo~N

Figure 3(b) shows that all of the wavelength compo-nents of a coherent state light packet are rephased at peri-ods of 2~. This is to be contrasted with the revivals in theJaynes-Cummings system, where the revival time there isdependent on the square root of the photon number, ~N[3]. We further note that, whereas there are also oscilla-

FIG. 3. (a, solid line), (b) The expectation number (ata) out-put of the de Broglie wavelength Mach-Zehnder interferome-ter for an incident coherent state with average photon numberN = 100. (a, dashed line) An approximate model of the outputindicating the coherence length of the coherent state; see text.(c, dashed) The expectation number output of the de Brogliewavelength Mach-Zehnder interferometer for several differentincident Fock states. (c, solid) The ensemble average outputfor 500 incident Pock states with photon numbers distributedaccording to a Poisson distribution centered at W = 100.

tions as a function of time delay in the Jaynes-Cummingsexperiment, that experiment does not represent a measureof the de Broglie wavelength, since the oscillation periodis proportional to the square root of the photon numberJN whereas in our case it is proportional to the pho-ton number N directly. We do note that for less than orequal to two photons three-level systems, such as the twophoton micromaser [4], may be treated as effective two-level systems with a Rabi frequency proportional to N [5].The effective Hamiltonian which governs these systems issimilar to the second part of our Eq. (1).

It is also interesting to consider the (normal beam split-ter) Mach-Zehnder interferometer with Fock states (in, n))incident on the two input ports [6,7]. An exact solu-tion of that interferometer generalized for arbitrary n [8]finds that the probability that the difference in photo-count between the two output ports is zero (i.e., the per-fect coincidence rate) is given by Po($) = [L„(cos@)]2,where L„are the Legendre polynomials defined byL„(x) = (1/2" n!) (d" /dx") (x —1)". While the Legendrefunctions contain many frequency components and thuscannot be considered a pure measure of the de Brogliewavelength of the incident Fock states (recall that theFock state has only a single de Broglie frequency compo-nent at ncuo), such dual Fock input interferometers do fea-ture components of their coincidence rate at the de Brogliefrequency. For the special case of single photon inputsn = 1 the Legendre function becomes a pure cosine at thede Broglie frequency.

Referring back to Fig. 3(a) it is interesting to note thatthe decayed interference at large P may be revived fromdata which have already been recorded. This may beaccomplished by correlating the list of measured valuesof photon number at exit port A with the total photonnumber measured after the interferometer as the sum ofcounts at A and B. We recall that the Fock state hasan infinitely long de Broglie coherence length, and thusif we condition the recorded interferometer data on thetotal photon count we may again recover interference eventhough the conglomerate set of data shows a decay of theinterference. Figure 3(c), dashed lines, shows the outputof the interferometer for several incident Fock states eachof different photon number. Figure 3(c), solid line, isthe average interferometer output for a statistical mixtureof incident Fock states whose numbers are distributedaccording to a Poisson distribution of average photonnumber N = 100. We note the similarity with Fig. 3(a).

In order to calculate sensitivity we must calculatefluctuations in the output signal. We have that theexpectation value for the number operator squared is

,(Pf iataataigf), = —((N + 1)+ e

X [N sin@ sin(P + N sing) —(1 + N cos@)cos(@ + N sing)]). (6)

4837

VOLUME 74, NUMBER 24 PH YS ICAL REVIEW LETTERS 12 JUNE 1995

0.08—td

.~ 0.04-

'U-

(a)

0.00 ~

0.00I I I

0.02 0.04

Phase P (units of tt)

0.06

GO

4

CDO 0

-4E

-8CDO

-4 -2 0 2

L.og[N]4 6 8

FIG. 4. The minimum phase 5@;„measurable by the deBroglie wavelength Mach-Zehnder interferometer for incidentcoherent state pulses: (a) As a function of the phase P for anaverage photon number of N = 100. (b) For the optimal phasesetting as a function of average photon number N.

Defining the interferometer signal as(ata)~+s~ —(ata)~ and the interferometer noise as

(ataatct)~ —(ata)~ we may solve for the mini-mum detectable phase shift 6P;„by requiring thatsignal/noise ~ 1.

Figure 4(a) plots 5@;„asa function of the phase Pfor an incident coherent state with average photon numberN = 100. For a phase setting of @ close to zero theinterferometer is Heisenberg limited in sensitivity, suchthat 6@;„=1/N for N ) 1 and b, @;„=I /~N forN ~ 1. Figure 4(b) shows that the ideal performanceis maintained from the region of much less than oneaverage photon to much more than one average photon.This establishes that the device forms a new class ofinterferometers operating at the fundamental quantum limitof sensitivity [9].

The above result may be understood when recallingthat a normal interferometer when fed a single photonin one input port and the vacuum in the other port isboth Heisenberg limited and standard quantum limited as1/N = I /~N for N = 1. Therefore the incident coherentstate photon packet is in essence a single de Broglieparticle when incident on the special beam splitters ofthe de Broglie wave interferometer. It is interesting tonote that in the case of the dual Fock state interferometer[7], as we have mentioned, the coincidence rate there isgiven by the Legendre polynomials. While, as noted,these functions contain many frequency components, ifone chooses the operating point near @ = 0 the dominantcomponent becomes the de Broglie frequency and thus theHeisenberg limited operation of that interferometer can al-

so be understood in terms of the de Broglie waveinterferometer.

In conclusion, we have shown that wavelength is aninternal structure dependent quantity. Specifically, theoutcome of a wavelength measurement is dependent onthe interaction which takes place at the beam splitterspresent inside the wavelength measurement apparatus.The interaction at the beam splitters may be a property ofthe interaction between constituent particles in an incidentcomposite object as in an interferometer for molecules orin other bound boson or bound photon systems such as inoptical solitons or diphoton systems [10]or the interactionmay be a function of the beam splitters themselves as inthe case of the effective beam splitters introduced in thispaper.

We show that such effective beam splitters may be usedto construct a de Broglie wave interferometer for opti-cal states as a whole. In the specific case of an coherentstate optical pulse we show that there is a finite coherencelength proportional to I /~N due to the uncertainty in pho-ton number. Furthermore, there is a revival of the inter-ference at phase periods of 2' due to the discrete natureof the constituent number states. A calculation of noiseproperties establishes that the interferometer is Heisenberglimited in sensitivity and thus forms a new class of inter-ferometers operating at the fundamental quantum limit.

*Electronic address: jacobson@sierra. stanford. edu~Also affiliated with NTT Basic Research Laboratories,Atsugishi, Kanagawa, Japan.

[1] Ch. J. Borde, N. Courtier, F. du Burck, A. N. Goncharov,and M. Gorlicki, Phys. Lett. A 1SS, 187 (1994).

[2] L. Davidovich, A. Maali, M. Brune, J.M. Raimond, andS. Haroche, Phys. Rev. Lett. 71, 2360 (1993).

[3] P. Meystre and M. Sargent, III, Elements of QuantumOptics (Springer Verlag, Berlin, 1991), 2nd ed. , p. 356.

[4] L. Davidovich, J.M. Raimond, M. Brune, and S. Haroche,Phys. Rev. A 36, 3771 (1987).

[5] M. Brune, J.M. Raimond, and S. Haroche, Phys. Rev. A35, 154 (1987); H. I. Yoo and J. H. Eberly, Phys. Rep.118, 239 (1985).

[6] J.G. Rarity, P. R. Tapster, E. Jakeman, T. Larchuck, R. A.Campos, M. C. Teich, and B.E.A. Saleh, Phys. Rev. Lett.65, 1348 (1990).

[7] M. J. Holland and K. Burnett, Phys. Rev. Lett. 71, 1355(1993).

[8] J.M. Jacobson and G. Bjork (unpublished).[9] A number of other classes of interferometers operating at

the Heisenberg limit of sensitivity are known. See C. M.Caves, Phys. Rev. D 23, 1693 (1981);M. O. Scully, Phys.Rev. Lett. 55, 2802 (1985); and Ref. [7].

[10] R. Y. Chiao and I. H. Deutsch, Phys. Rev. Lett. 67, 1399(1991).

4838