arXiv:quant-ph/0202133v2 9 Apr 2002 pndnwpsiiiis npriua,Jacobson particular, approach in dual-Fock-state possibilities; This new opened [8]. sensitivity limited iiiycnb mrvdb atrof factor a by improved be sen can inteferometer sitivity the quantum particles, prepare the carefully between we correlations when However, in- [1,2]. single-particle with puts performed of is number in- experiment the the the times or through time, passing measurement during particles terferometer of number average the hs estvt f∆ of sensitivity phase ulFc states Fock dual o l ausof values all for ieadadc-okr rtsoe httebs pos- best the that showed first co-workers and Wineland two interferometer. the Mach-Zhender Ram- formally optical Here, using the stan- is clock to frequency technique equivalent atomic of separated-oscillatory-fields an determination sey’s spectroscopy; the and in dards achieved be sensitivity Heisenberg-limited dual interferometry. yield the atom that also in showed can [10] state Kasevich Fock and Bouyer and [9], tdsniiiyta clswith scales that sensitivity ited curdfrseilvle of values special for occurred eoee hr aldthe inter- called Mach-Zehnder (here the ferometer for input Fock-state correlated Yurke Also, estvt o clswt 1 with scales now sensitivity n ugnrlzdti ehdt ban1 obtain to method this generalized Ou and mtrwt qezdlgtyed htnielwrthan lower shot-noise interfer- a Mach-Zehnder N yields light optical squeezed an with of ometer port input magnitude. ondary of orders eight of ratio noise sig- to enhanced nal an quantum to the corresponds improvement milliwatts, inter- sensitivity several optical at For operating ferometers principle. uncertainty Heisenberg the − usqety oln n unt rpsdteueof use the proposed Burnett and Holland Subsequently, iia mrvmn nmaueetsniiiycan sensitivity measurement in improvement similar A eei lsia nefrmtrhsaso-os lim- shot-noise a has interferometer classical generic A serya 91 ae hwdta edn h sec- the feeding that showed Caves 1981, as early As 2 1 (where ASnmes 36.d 25.v 36.a 22.z 85. 42.25.Hz, 03.67.-a, 42.50.Dv, 03.65.Ud, numbers: PACS spectroscopy. atom Ro in The work sensitivity. should Heisenberg-limited achieve to correlation order desired in Rosetta the “quantum generating the for scheme introduce measurement we equivalence spectroscope this addr Ramsey on We Based the i interferometer, spectroscopes. be Mach-Zehnder can Ramsey the measurements and Such interferometers Zehnder protocols. classical over cision tal et N π/ esneglmtdmaueetpoooscnb sdt ga to used be can protocols measurement Heisenberg-limited ϕ snwteaeaepoo ubr [3]. number) photon average the now is o h form the (of 45 hwdi 96ta properly a that 1986 in showed [4,5] . -ussrpeetteba splitters. beam the represent 2-pulses unu optn ehooisGop xlrto System Exploration Group, Technologies Computing Quantum [6,7]. ϕ ≃ uk state Yurke unu oet tn o Interferometry for Stone Rosetta Quantum A O /N S1637 80OkGoeDie aaea A91109-8099 CA Pasadena, Drive, Grove Oak 4800 126-347, MS | (1 ,N N, e rplinLbrtr,Clfri nttt fTechno of Institute California Laboratory, Propulsion Jet ϕ hslmti moe by imposed is limit This . /N adr n Milburn, and Sanders . N wn e,Pee o,adJnta .Dowling P. Jonathan and Kok, Pieter Lee, Hwang − .Ti improvement This ). i ogi Heisenberg gain to ) 1 2 Here, . √ ol edt a to lead could ) N /N hti,the is, That . N sensitivity seither is Mrh3,2002) 31, (March tal et - . 1 efrmti ihgah n h eie unu state quantum in- desired quantum the of and Heisen- description lithography the brief terferometric achieving A sensitivity. of limit ways III berg different section In three circuit. discuss logic we quantum equiva- generic formal atomic a and interferometer, the clocks, Mach-Zehnder on the describe based between we stone, lence (II), Rosetta section quantum next a the In quan- with entanglement. estimation phase tum introduce previous then Following we [19], limit. work shot-noise standard the and ity paper follows: The as states. organized quantum is desired for the method of another generation inter- describe the in we entanglement and devices, quantum ferometeric by enhancement the of in discussed state entangled 11. maximally Ref. the of of resolution form a same de- yielding The lithography interferomet- light. quantum ric substan- for the state features of quantum wavelength entangled the sired the image than to smaller the entangled-photons possible tially than same is smaller the it much using technique, be cannot wavelength. resolved optical be can that photons entangled of overcome nature be [13–18]. quantum can the limit classical exploiting this by that shown been has u etr iei eemndb h alihdiffrac- Rayleigh the by mini- of of determined limit the tion wavelength is lithography of size optical the classical feature resolution by mum In limited the used. is is light which light the size, with feature bottleneck writing the the and practice, reading In for sensitivity. enhanced this to be can [12]. sensitivity gyroscopes improved atom-laser this in that (JPD) exploited shown us was of it one Similarly, by [11]. using by states obtained entangled is maximally standard frequency in precision sible nscinI edsusteHiebr-iie sensitiv- Heisenberg-limited the discuss we I, section In aspects some of overview an present we paper this In detail finest the too, microscopy optical classical In unu ihgah n irsoyi lsl related closely is microscopy and lithography Quantum et tn hntlsu h aemethod same the us tells then stone setta ewe h nefrmti nu states input interferometric the between s 40.Hp peetduig .. pia Mach- optical e.g., using, mplemented n eei unu oi circuit. logic quantum generic a and , tn” n edsrb projective- a describe we and stone”, s h omleuvlnebetween equivalence formal the ess na nraei esrmn pre- measurement in increase an in λ/ ,where 4, uooy eto 367 Section Autonomy, s logy λ stewvlnt ftelgt It light. the of wavelength the is λ/ (4 N a the has ) of light is given in section IV. In section V we discuss quantum state preparation with projective measurements 2 N and its application to Heisenberg-limited interferometry. (∆Jz) = , (4) 4 which, then, corresponds to the usual shot-noise limit. I. THE HEISENBERG UNCERTAINTY This result gives the spread of measurement outcomes PRINCIPLE AND PARAMETER ESTIMATION of an observable Qˆ in a (N + 1)-level system. However, it is not yet cast in the language of standard parame- ter estimation. The next question is therefore how to In this section we briefly describe the measurement- achieve this Heisenberg-limited sensitivity when we wish sensitivity limits due to Heisenberg’s uncertainty princi- to estimate the value of a parameter ϕ in N trials. ple, and how, in general, quantum entanglement can be The shot-noise limit (SL), according to estimation the- used to achieve this sensitivity. There are several stages 1 ory, is given by ∆ϕ = N − 2 . We give a short derivation in the procedure where quantum entanglement can be of this value and generalize it to the quantum mechanical exploited, both in the state preparation and in the de- case. Consider an ensemble of N two-level systems in the tection. First, we derive the Heisenberg limit, then we state ϕ = ( 0 + eiϕ 1 )/√2, where 0 and 1 denote give the usual shot-noise limit, and we conclude this sec- the two| i levels.| Ifi Aˆ = |0i 1 + 1 0 , then| i the expectation| i tion with entanglement enhanced parameter estimation. value of Aˆ is given by| ih | | ih | Suppose we have a (N + 1)-level system. Furthermore, we use the angular momentum representation to find the minimum uncertainty ∆Q in an observable Qˆ that is a ϕ Aˆ ϕ = cos ϕ . (5) h | | i dual to the angular momentum operator Jˆz. That is, Qˆ and Jˆz obey a Heisenberg uncertainty relation (~ = 1): When we repeat this experiment N times, we obtain

N ˆ(k) 1 N ϕ . . . 1 ϕ A ϕ 1 . . . ϕ N = N cos ϕ . (6) ∆Q ∆J . (1) h | h | k⊕=1 ! | i | i z ≥ 2 Furthermore, it follows from the definition of Aˆ that ˆ When we want minimum uncertainty in Q (minimal Aˆ2 = 11 on the relevant subspace, and the variance of ˆ ∆Q), we need to maximize the uncertainty in Jz (max- Aˆ given N samples is readily computed to be (∆A)2 = +N/2 2 2 imal ∆Jz). Given the eigenstates m of Jˆz: N(1 cos ϕ)= N sin ϕ. According to estimation theory {| i}m=−N/2 − Jˆz m = m m , let us, for example, take the state of the [2], we have | i | i 1 +N/2 form ψ = (N + 1)− 2 eiφm m , which consists | i m=−N/2 | i of equally distributed eigenstates of Jˆ . The variance in ∆A 1 P z ∆ϕSL = = . (7) Jˆ is then given by d Aˆ /dϕ √N z | h i | This is the standard variance in the parameter ϕ after 1 N 2 N N trials. In other words, the uncertainty in the phase is (∆J )2 = ψ Jˆ2 ψ ψ Jˆ ψ 2 = + . (2) z h | z | i − h | z| i 3 4 2 inversely proportional to the square root of the number   of trials. This is, then, the shot-noise limit.

It immediately follows that the leading term in ∆Jz scales Quantum entanglement can improve the sensitivity of with N. this procedure by a factor of √N. We will first employ 1 If we choose the quantum state ψ =2− 2 ( m = N + an entangled state | i | 2 i eiφ m = N ), we obtain | − 2 i iNϕ ϕN N, 0 + e 0,N , (8) 2 | i ≡ | i | i 2 N (∆Jz) = . (3) where N, 0 and 0,N are collective states of N qubits, 4 defined,| fori example,| i in the computational basis 0 , 1 , as {| i Using the equality sign in Eq. (1), i.e., minimum uncer- | i} tainty, and the expression for ∆J , we find that ∆Q = z N, 0 = 0 0 0 1/N. | i | i1| i2 ···| iN By contrast consider a quantum state of the form 0,N = 1 1 1 2 1 N . (9) N 1 | i | i | i ···| i − +N/2 iφm ψ = 2 2 e ( C ) 2 m , which is a iNϕ | i m=−N/2 N m+N/2 | i The relative phase e can be obtained by a unitary ˆ binomial distributionP of the eigenstates of Jz. The vari- evolution of one of the modes of ϕN . Here, in the ance in Jˆ in this case is given by case of Eq. (8), each qubit in state | 1 iacquires a phase z | i

2 iϕ − 1 shift of e . When we measure the observable AˆN = N ), the phase sensitivity is still given by N 2 , indicat- 0,N N, 0 + N, 0 0,N we have |ingi that the photon counting noise does not originate | ih | | ih | from the intensity fluctuations of the input beam [12,25]. In the next section we will see how we can improve this ϕ Aˆ ϕ = cos Nϕ . (10) sensitivity. h N | N | N i In a Ramsey spectroscope, atoms are put in a superpo- ˆ2 2 Again, AN = 11 on the relevant subspace, and (∆AN ) = sition of the ground state and an excited state with a π/2- 1 cos2 Nϕ = sin2 Nϕ. Using Eq. (7) again, we ob- pulse (Fig. 1b). After a time interval of free evolution, a − tain the so-called Heisenberg limit (HL) to the minimal second π/2-pulse is applied to the atom, and, depending detectable phase: on the relative phase shift obtained by the excited state in the free evolution, the outgoing atom is measured ei- ther in the ground or the excited state. Repeating this ∆A 1 N procedure N times, the phase ϕ is determined with pre- ∆ϕHL = = . (11) 1 d Aˆ /dϕ N − 2 | h N i | cision N . This is essentially an atomic clock. Both procedures are methods to measure the phase shift, ei- √ The precision in ϕ is increased by a factor N over the ther due to the path difference in the interferometer, or standard noise limit. Detailed descriptions of phase esti- to the free-evolution time in the spectroscope. When we mation have been given by Hradil [20], and Lane, Braun- use entangled atoms in the spectroscope, we can again stein, and Caves [21]. increase the sensitivity of the apparatus. As shown in Bollinger et al. [11], Eq. (11) is the optimal accuracy permitted by the Heisenberg uncer- (a)



tainty principle. In quantum lithography, one exploits A beam splitter the cos(Nϕ) behaviour, exhibited by Eq. (10), to print ϕ closely spaced lines on a suitable substrate [14]. Gy- roscopy and entanglement-enhanced frequency measure- ments [12] exploit the √N increased precision. The phys- B ical interpretations of AN and the phase ϕ will differ



in the different protocols. In the following section we present three distinct physical representations of this con- (b) π/2 pulse struction. We call this the quantum Rosetta stone. Of particular importance is the interplay between the cre- A ϕ ated states ϕN and the measured observable AˆN in B Eq. (10). In| sectioni III we present three types of quan- tum states and observables that yield Heisenberg-limited sensitivity. (c) Hadamard gate A ϕ II. QUANTUM ROSETTA STONE HH B In this section we discuss the formal equivalence be- tween the Mach-Zehnder interferometer, the Ramsey FIG. 1. The quantum Rosetta stone. (a) An optical spectroscope, and a logical quantum gate. Mach-Zehnder interferometer. (b) Ramsey atomic clock. (c) In a Mach-Zehnder interferometer [22], the input light A quantum logic gate representing the equivalent physical field is divided into two different paths by a beam split- proccess. ter, and recombined by another beam splitter. The phase difference between the two paths is then measured by bal- A third system is given by a qubit that undergoes a anced detection of the two output modes (see Fig. 1a). Hadamard transform H, then picks up a relative phase A similar situation, which we will omit in our discus- and is then transformed back with a second Hadamard sion, can be found in Stern-Gerlach filters in series [23], transformation (Fig. 1c). This representation is more and the technical limitations of such a device has been mathematical than the previous two, and it allows us discussed by Englert, Schwinger, and Scully [24]. to extract the unifying mathematical principle that con- With a coherent laser field as the input the phase sen- nects the three systems. In all protocols, the initial state 1 sitivity is given by the shot noise limit N − 2 , where N is is transformed by a discrete Fourier transform (beam the average number of photons passing though the inter- splitter, π/2-pulse or Hadamard), then picks up a relative ferometer during measurement time. When the number phase, and is transformed back again. This is the stan- of photons is exactly known (i.e., the input is a Fock state dard “quantum” finite Fourier transform, such as used

3 in the implementation of Shor’s algorithm [26]. It is not III. QUANTUM ENHANCEMENT IN PHASE the same as the classical algorithm in engineering. MEASUREMENTS As a consequence, the phase shift (which is hard to measure directly) is applied to the transformed basis. There have been various proposals for achieving The result is a bit flip in the initial, computational, ba- Heisenberg-limited sensitivity, corresponding to different sis 0 , 1 , and this is readily measured. We call the physical realizations of the state ϕ and observable Aˆ {| i | i} N N formal analogy between these three systems the quantum in Eq. (10). Here, we discuss three| differenti approaches, Rosetta stone. categorized according to the different quantum states. The importance of the Rosetta stone was that, by giv- We distinguish Yurke states, dual Fock states, and max- ing an example of writing in three different languages: imally entangled states. Greek, Demotic, and Hieroglyphics, it enabled the French scholar Jean Fran¸cois Champollion to crack the code of the hieroglyphs, which was not understood until his work A. Yurke states in 1822 [27]. In discussing quantum computer circuits with researchers from the fields of quantum optics or By utilizing the su(2) algebra of spin angular momen- atomic clocks, we find the “quantum Rosetta stone” a tum, Yurke et al. have shown that, with a suitably corre- useful tool to connect those areas of research to the lan- lated input state, the phase sensitivity can be improved guage of quantum computing – which can be as indeci- to 1/N [4,5]. Similarly, Hillery and Mlodinow, and Brif pherable as hieroglyphs to such researchers. and Mann have proposed using the so-called minimum These schemes can be generalized from measuring a uncertainty state or the “intelligent state”. The mini- simple phase shift to evaluating the action of a unitary mum uncertainty state is defined as ∆J ∆J = Jˆ /2, x y |h zi| transformation Uf associated with a complicated func- and such a state with ∆Jy 0 can yield the Heisenberg tion f on multiple qubits. Such an evaluation is also limited sensitivity under certain→ conditions [38,39]. known as a quantum computation. Quantum computing Leta ˆ†, ˆb† denote the creation operators for the two with generalized Mach-Zehnder interferometers has been input modes in Fig. 1a. In the Schwinger representation, 2 proposed by Milburn [28], and later linear optics quan- the common eigenstates of Jˆ and Jˆz are the two-mode tum computation has been suggested by Cerf, Adami, Fock states j, m = j + m A j m B , where and Kwiat [29]. Recently, Knill, Laflamme and Milburn | i | i | − i demonstrated an efficient linear optics quantum compu- Jˆ = (ˆa†ˆb + ˆbaˆ†)/2 ; Jˆ = i(ˆa†ˆb ˆbaˆ†)/2 ; x y − − tation [30]. Implementations of quantum logic gates have Jˆ = (ˆa†aˆ ˆb†ˆb)/2 ; Jˆ2 = Jˆ2 + Jˆ2 + Jˆ2 . (12) also been proposed by using atoms and ion traps [31–34]. z − x y z According to our Rosetta stone, the concept of quantum The interferometer can be described by the rotation of computers is therefore to exploit quantum interference the angular momemtum vector, wherea ˆ†aˆ + ˆb†ˆb = N = in obtaining the outcome of a computation of f. In this 2j, and the 50/50 beam splitters and the phaser shift are light, a quantum computer is nothing but a (complicated) corresponding to the operators eiπJˆx/2 and eiϕJˆz , respec- multiparticle quantum interferometer [35,36]. tively. This logic may also be reversed: a quantum interfer- For spin-1/2 fermions, the entangled input state (which ometer is therefore a (simple) quantum computer. Often, we call the ‘Yurke state’) ϕN Y is given by when discussing, Heisenberg-limited interferometry with | i complicated entangled states, one encounters the critique 1 N 1 N 1 ϕN Y = j = ,m = + j = ,m = that such states are highly susceptible to noise and even | i √2 2 2 2 − 2 one or a few uncontrolled interactions with the environ- 1   = N+1 , N−1 + N−1 , N+1 , (13) ment will cause sufficient degradation of the device and √2 2 2 AB 2 2 AB recover only the shot-noise limit [37]. Apply the quantum   Rosetta stone by replacing the term “quantum interfer- where the notion of j, m follows the definition given ometer” with “quantum computer” and we recognize the in Eq. (12) and the subscripts| i AB denote the two in- exact same critique that has been leveled against quan- put modes. For bosons, a similar input state, namely tum computers for years. However, for quantum com- j = N/2,m =0 + j = N/2,m =1 , has been proposed | i | i puters, we know the response–to apply quantum error- [5]. The measured observable AˆN is given by Jˆz. After correcting techniques and encode in decoherence free sub- evolving the state ϕN Y (in the Ramsey spectroscope spaces. Quantum interferometry is just as hard as quan- or the Mach-Zehnder| interferometeri with phase shift ϕ), tum computing! The same error-correcting tools that we the phase sensitivity ∆ϕ can be determined to be pro- believe will make quantum computing a reality, will also portional to 1/N for special values of ϕ (Fig. 2a). Al- be enabling for quantum interferometry. though the input state of Eq. (13) was proposed for spin- 1/2 fermions, the same state with bosons also yields the phase sensitivity of the order of 1/N [12].

4 B. Dual Fock states processing method based on Bayesian analysis was pro- posed by Kim et al. [40]. Recently, Wisemen and co- In order to achieve Heisenberg-limited sensitivity, Hol- workers have proposed an adaptive measurement scheme, land and Burnett proposed the use of so-called dual Fock where the optimal input state, however, is different from states N A N B for two input modes A and B of the the dual Fock states [41]. Mach-Zehnder| i ⊗ | interferometeri [8]. Such a state can be For atom interferometers, a quantum nondemolition generated, for example, by degenerate parametric down measurement is required to give the total number of conversion, or by optical parametric oscillation [40]. atoms [10]. In a similar context, Yamamoto and co- To obtain increased sensitivity with dual Fock states, workers devised an atom interferometry scheme that uses some special detection scheme is needed. In a conven- a squeezed π/2 pulse for the readout of the input state tional Mach-Zehnder interferometer only the difference of correlation [9]. the number of photons at the output is measured. Sim- Due to its simple form, the dual Fock-state approach ilarly, in atom interferometers, measurements are per- may shed a new light on Heisenberg-limited interferom- formed by counting the number of atoms in a specific etry. In particular, exploiting the fact that atoms in internal state using fluorescence. For the schemes using a Bose-Einstein condensate can be represented by Fock dual Fock-state input, an additional measurement is re- states, Bouyer and Kasevich, as well as Dowling, have quired since the average in the intensity difference of the shown that the quantum noise in atom interferometry us- two output ports does not contain information about the ing dual Bose-Einstein consensates can also be reduced phase shift (Fig. 2b). to the Heisenberg limit [10,12].

(a)



A ϕ C. Maximally entangled states

magic Ψ The third, and last, category of states is given by the B maximally entangled states. There have been proposals



to overcome the standard shot noise limit by using spin- squeezed states [42–46]. However, it was demonstrated (b)



by Wineland and co-workers [11] that the optimal fre- ϕ quency measurement can be achieved by using maximally entangled states [47], which have the following form: N N magic D 1 ϕN = ( N, 0 AB + 0,N AB) . (14)



| i √2 | i | i

(c) This state has an immediate resemblence with the state



in Eq. (10) after acquiring a phase shift of eiNϕ. Al- magic BS ϕ though partially entangled states with a high degree of symmetry were later shown to have a better resolution N 0 N 0 + 0 N in the presence of decoherence [48], these maximally cor- related states are of great interest for optical interferom-

etry.

FIG. 2. Three catagories for achieving the Heisenberg lim- Hadamard gate ited phase sensitivity. Emphasis on the distinctive features H are termed as “magic”. (a) Correlated input state (magic state), (b) dual Fock-state input (magic detectors), (c) max- imally correlated state (magic beam splitter).

One measures both the sum and the difference of the photon number in the two output modes [8]. The sum contains information about the total photon number, and the difference contains information about the phase shift. Information about the total photon number then allows FIG. 3. Quantum logic gates for generation of the maxi- for post-processing the information about the photon- mally entangled states (representing the “magic” beam split- number difference. A combination of a direct measure- ter in Fig. 2c). N qubits become maximally entangled via a ment of the variance of the difference current and a data- Hadamard gate followed by N C-NOT gates.

5 In terms of quantum logic gates, the maximally cor- As depicted in Fig. 3, the maximally correlated state related state of the form of Eq. (14) can be made using of the form of Eq. (14) can be made using a Hadamard a Hadamard gate and a sequence of C-NOT gates (see gate and a sequence of N C-NOT gates. However, build- Fig. 3). Note that one distinctive feature, compared to ing optical C-NOT gates normally requires large optical the other schemes described above, is that the state of nonlinearities. Consequently, in generating such states the form Eq. (14) is the desired quantum state after the it is commonly assumed that large χ(3) nonlinear optical first beam splitter in the Mach-Zehnder interferometer, components are needed for N > 2. not the input state (Fig. 2c). In that the desired input Knill, Laflamme, and Milburn proposed a method state is described as the inverse beam-splitter operation for creating probabilistic single-photon quantum logic to the state of Eq. (14). gates based on teleportation. The only resources for All the interferometric schemes using entangled or this method are linear optics and projective measure- dual-Fock input states show a sensitivity approaching ments [30]. Probabilistic quantum logic gates using po- 1/N only asymptotically. However, using the maximally larization degrees of freedom have been proposed by correlated states of Eq. (14), the phase sensitivity is equal Imoto and co-workers, and Franson’s team [51,52]. In to 1/N, even for a small N. particular, Pittman, Jacobs, and Franson have exper- imentally demonstrated polarization-based C-NOT im- plementations [53]. IV. QUANTUM LITHOGRAPHY AND Using the concept of projective measurements, we have MAXIMALLY PATH-ENTANGLED STATES previously demonstrated that the desired path-entangled states can be created when conditioned on the measure- ment outcome [19,54]. This way, one can avoid the use Quantum correlations can also be applied to optical of large χ(3) nonlinearities [55]. In the next section we lithography. In recent work it has been shown that the discuss the path-entanglement generation based on pro- Rayleigh diffraction limit in optical lithography can be jective measurements, and its application to Heisenberg- circumvented by the use of path-entangled photon num- limited interferometry. ber states [13,14]. The desired N-photon path-entangled state, for N-fold resolution enhancement, is again of the form1 given in Eqs. (8) and (14). V. PROJECTIVE MEASUREMENTS AND Consider the simple case of a two-photon Fock state HEISENBERG-LIMITED INTERFEROMETRY 1 1 , which is a natural component of a sponta- | iA| iB neous parametric down-conversion event. After passing As is discussed in the previous sections, the Yurke state through a 50/50 beam splitter, it becomes an entangled approach has the same measurement scheme as the con- number state of the form 2 A 0 B + 0 A 2 B. Interfer- ventional Mach-Zehnder interferometer (direct detection | i | i | i | i ence suppresses the probability amplitude of 1 A 1 B. of the difference current), but it is not easy to generate | i | i According to quantum mechanics, it is not possible to the desired correlation in the input state [12]. On the tell whether both photons took path A or B after the other hand, the dual Fock-state approach finds a rather beam splitter. simple input state, but requires a complicated mea- When parametrizing the position x on the surface by suremet schemes. In this section we describe a method ϕ = πx/λ, the deposition rate of the two photons onto for creating a desired correlation in the Yurke state ap- the substrate becomes 1 + cos2ϕ, which has twice bet- proach directly from the dual Fock state by using the ter resolution λ/8 than that of single-photon absorption, projective measurements. 1 + cos ϕ, or that of uncorrelated two-photon absorp- Consider the scheme depicted in Fig. 4. The input tion, (1 + cos ϕ)2. For N-photon path-entangled state modes are transformed by the beam splitters as follows: of Eq. (14), we obtain the deposition rate 1 + cos Nϕ, corresponding to a resolution enhancement of λ/(4N). aˆ† itaˆ′† + ruˆ′†, → It is well known that the two-photon path-entangled ˆb† itˆb′† + rvˆ′†, (15) state of Eq. (14) can be generated using a Hong-Ou- → Mandel (HOM) interferometer [49] and two single-photon where i = √ 1, and it and r are the transmission and input states. A 50/50 beam splitter, however, is not suf- reflection coefficients− given by t2 + r2 = 1. In our con- ficient for producing path-entangled states with a photon vention, a 50/50 beam splitter, for example, is identified number larger than two [50]. as t =1/√2 and r = 1/√2. −

1 We call the state of the form |N, 0i + |0, Ni as the NOON state, and the High NOON state for a large N.

6 ′ ′ N−1 The modesu ˆ andv ˆ further pass through an addi- N . The output state of the form Eq. (20) shows the tional 50/50 beam splitter, which is characterized by the exact correlation for the Yurke state [12]. The genera- transformations: tion of the desired correlations described in section IIIA can thus be obtained from the dual Fock state N,N ′† ˆ† † √ uˆ (id cˆ )/ 2, with probability A 2. Note that this probability| has itsi → − | | vˆ′† (icˆ† dˆ†)/√2. (16) asymptotic value of 1/(2e2), independent of N. → − A similar correlation can also be ontained by post se- lecting the outcome, conditioned upon only one photon u detection by either one of the two detectors. The anal- ysis becomes simpler than the one decribed above: one a a' photon either from mode a or from b is to be detected. u' In this case, however, one needs to allow a non-detection c protocol [19] since one of the two detectors should not be fired. d In this section we described generation of quantum v' correlation by using projective measurments, where the fundamental lack of which-path information provides the b b' entanglement between the two output modes. Using v a stack of such devices with appropriate phase shifters FIG. 4. If one and only one photon is detected at each de- has been proposed for generating maximmally path- tector, two photons are “peeled off” from either input mode entangled state of the form Eq. (14) [19]. In atom in- a or b. terferometry a similar technique for the the generation of such a correlation has been proposed [12] with selec- Suppose the input state is given by the dual Fock state tive measurements on two interfering Bose condensates [56]. (ˆa†)N (ˆb†)N Ψ = N,N = 0 . (17) | ini | i N! | i Accordingly, the beam splitter transformation, e.g., for the mode a, can be written as VI. SUMMARY

(ˆa†)N C (ˆa′†)N−k(ˆu′†)krk(it)N−k . (18) → N k Xk In this paper we discuss the equivalence between the Let both modes pass through a beam splitter, and re- optical Mach-Zehnder interferometer, the Ramsey spec- combine the reflected modes u′ and v′ in a 50/50 beam troscope, and the quantum Fourier transform in the form splitter as depicted in Fig. 4. We post-select events on of a specific quantum logic gate. Based on this equiva- a two-fold detector coincidence at the output modes of lence, we introduce the so-called quantum Rosetta stone. this beam splitter. Assuming ideal detectors, this will The interferometers and spectroscopes can yield restrict the final state of the outgoing modes a′ and b′. Heisenberg-limited phase sensitivity when suitable in- Then, from Eq. (16), we have put states are chosen. We presented three types of such states and showed that they need non-local correlations ′† ′† † 2 † 2 uˆ vˆ i (ˆc ) + (dˆ ) /2 , of some sort. Path-entangled multi-photon states can be →− h i generated using projective measurements. These states (ˆu′†)2 (ˆc†)2 2icˆ†dˆ† (dˆ†)2 /2 , → − − yield 1/N sensitivity also for small N, and are essen- h i tial for applications such as quantum lithography. The (ˆv′†)2 (ˆc†)2 2icˆ†dˆ† + (dˆ†)2 /2 . (19) → − − dual Fock-state approach to Heisenberg-limited interfer- h i ometry must be accompanied by “non-local” detection We note that only (ˆu′†)2 and (ˆv′†)2 terms are selected by schemes. The generation of a suitable correlation from the detection of a single photon at each detector. The the dual Fock state via projective measurements may output state conditioned on a two-fold coincident count be useful in that it avoids complicated signal processing is given by or quantum nondemolition measurements. One can also A construct a single-photon quantum nondemolition device N,N [ N,N 2 + N 2,N ] , (20) | i → √2 | − i | − i and a quantum repeater in this paradigm [57,58]. Many more important insights are expected by the application where A = N(N 1)/2 r2t2N−2, and the maximum of the quantum Rosetta stone to quantum metrology and probability success is− obtained when r2 = 1 and, t2 = quantum information processing. p N

7 ACKNOWLEDGEMENT B.-G. Englert, and J. Schwinger, Phys. Rev. A 40, 1775 (1989). This work was carried out by the Jet Propulsion Lab- [25] M.O. Scully and J.P. Dowling, Phys. Rev. A 48, 3186 oratory, California Institute of Technology, under a con- (1993). [26] A. Ekert and R. Jozsa, Rev. Mod. Phys. 68, 733 (1996). tract with the National Aeronautics and Space Admin- [27] J.F. Champollion, Pr´ecis du syst`eme hi´eroglyphique des istration. We wish to thank D.S. Abrams, C. Adami, anciens Egyptiens (Paris, 1824). N. J. Cerf, J. D. Franson, P. G. Kwiat, T. B. Pittman, [28] G.J. Milburn, Phys. Rev. Lett. 62, 2124 (1989). Y. H. Shih, D. V. Strekalov, C. P. Williams, and D. J. [29] N.J. Cerf, C. Adami, and P.G. Kwiat, Phys. Rev. A 57, Wineland for helpful discussions. We would also like to R1477 (1998). acknowledge support from the ONR, ARDA, NSA, and [30] E. Knill, R. Laflamme, and G.J. Milburn, Nature 409, DARPA. P.K. and H.L. would like to acknowledge the 46 (2001). National Research Council. [31] P. Domokos et al., Phys. Rev. A 52, 3554 (1995). [32] J.I. Cirac and P. Zoller, Phys. Rev. Lett. 74, 4091 (1995). [33] Q. A. Turchette et al., Phys. Rev. Lett. 75, 4710 (1995). [34] C. Monroe et al., Phys. Rev. Lett. 75, 4714 (1995). [35] A. Ekert, Physica Scripta T76, 218 (1998). [36] R. Cleve, et al., Proc. R. Soc. Lond. A 454, 339 (1998). [37] H. J. Kimble, private communication. [1] For example, see, M.O. Scully and M.S. Zubairy, Quan- [38] M. Hillery and L. Mlodinow, Phys. Rev. A 48, 1548 tum Optics (Cambridge University Press, Cambridge, (1993). UK, 1997). [39] C. Brif and A. Mann, Phys. Rev. A 54, 4505 (1996). [2] C.W. Helstrom, Quantum detection and estimation the- [40] T. Kim et al., Phys. Rev. A 57, 4004 (1998). ory, Mathematics in Science and Engineering 123 (Aca- [41] D.W. Berry and H.M. Wiseman, Phys. Rev. Lett. 85, demic Press, New York, 1976). 5098 (2000); D.W. Berry, H.M. Wiseman, and J.K. Bres- [3] C.M. Caves, Phys. Rev. D 23, 1693 (1981). lin, Phys. Rev. A 63, 053804 (2001). [4] B. Yurke, Phys. Rev. Lett. 56, 1515 (1985). [42] M. Kitagawa and M. Ueda, Phys. Rev. Lett. 67, 1852 [5] B. Yurke, S.L. McCall, and J.R. Klauder, Phys. Rev. A (1993). 33, 4033 (1986). [43] D.J. Wineland et al., Phys. Rev. A 46, R6797 (1992). [6] B.C. Sanders and G.J. Milburn, Phys. Rev. Lett. 75, [44] M. Kitagawa and M. Ueda, Phys. Rev. A 47, 5138 (1993). 2944 (1995). [45] W.M. Itano et al., Phys. Rev. A 47, 3554 (1993). [7] Z.Y. Ou, Phys. Rev. Lett. 77, 2352 (1996). [46] G.S. Agarwal and R.R. Puri, Phys. Rev. A 49, 4968 [8] M.J. Holland and K. Burnett, Phys. Rev. Lett. 71, 1355 (1994). (1993). [47] N.D. Mermin, Phys. Rev. Lett. 65, 1838 (1990). [9] J. Jacobson, G. Bj¨ork, Y. Yamamoto, Appl. Phys. B 60, [48] S.F. Huelga et al., Phys. Rev. Lett. 79, 3865 (1997); S.F. 187 (1995). Huelga et al., Appl. Phys. B 67, 723 (1998). [10] P. Bouyer and M.A. Kasevich, Phys. Rev. A 56, R1083 [49] C.K. Hong, Z.Y. Ou, and L. Mandel, Phys. Rev. Lett. (1997). 59, 2044 (1987). [11] J.J. Bollinger et al., Phys. Rev. A 54, R4649 (1996). [50] R.A. Campos, B.E.A. Saleh, and M.C. Teich, Phys. Rev. [12] J.P. Dowling, Phys. Rev. A 57, 4736 (1998). A 40, 1371 (1989). [13] E. Yablonovich and R.B Vrijen, Opt. Eng. 38, 334 (1999). [51] M. Koashi, T. Yamamoto, and N. Imoto, Phys. Rev. A [14] A.N. Boto et al., Phys. Rev. Lett. 85, 2733 (2000). 63, 030301 (2001). [15] P. Kok et al., Phys. Rev. A 63, 063407 (2001). [52] T.B. Pittman, B.C. Jacobs, and J.D. Franson, Phys. Rev. [16] G.S. Agarwal et al., Phys. Rev. Lett. 86, 1389 (2001). A 64, 062311 (2001). [17] M. D’Angelo et al., Phys. Rev. Lett. 87, 013602 (2001). [53] T.B. Pittman, B.C. Jacobs, and J.D. Franson, “Demon- [18] E.M. Nagasako et al., Phys. Rev. A 64, 043802 (2001). stration of Non-Deterministic Quantum Logic Opera- [19] P. Kok, H. Lee, and J.P. Dowling, “The creation of large tions using Linear Optical Elements,” quant-ph/0109128. photon-number path entanglement conditioned on pho- [54] H. Lee, P. Kok, N.J. Cerf, and J.P. Dowling, Phys. Rev. todetection,” Phys. Rev. A (in press), quant-ph/0112002. A 65, 030101 (R) (2002). [20] Z. Hradil, Quant. Opt. 4, 93 (1992). [55] C. Gerry and R.A. Campos, Phys. Rev. A 64, 063814 [21] A.S. Lane, S.L. Braunstein, and C.M. Caves, Phys. Rev. (2001). A 47, 1667 (1993). [56] Y. Castin and J. Dalibard, Phys. Rev. A 55, 4330 (1997). [22] The discussion about the Mach-Zehnder is equally appli- [57] P. Kok, H. Lee, and J.P. Dowling, “Implementing cable to the Fabry-Perot interferometer. a single-photon quantum nondemolition device with [23] R.P. Feynman, R.B. Leighton, and M. Sands, The Feyn- only linear optics and projective measurements,” quant- man Lectures on Phyisics III (Addison-Wesley, Reading, ph/0202046. MA, 1979). [58] P. Kok, C.P. Williams, and J.P. Dowling, “Practical [24] B.-G. Englert, J. Schwinger, and M.O. Scully, Found. quantum repeaters with linear optics and double-photon Phys. 18, 1045 (1988); J. Schwinger, M.O. Scully, and guns,” quant-ph/0203134. B.-G. Englert, Z. Phys. D 10, 135 (1988); M.O. Scully,

8