All you ever wanted to know about the quantum Zeno effect in 70 minutes
AAll you ever wanted to know about the quantum Zeno effectin 70 minutes
Saverio Pascazio
Dipartimento di Fisica and MECENAS, Universit`a di Bari, I-70126 Bari, Italy& INFN, Sezione di Bari, I-70126 Bari, Italyemail: [email protected] (Processed: November 27, 2013; Received: .)
Abstract.
This is a primer on the quantum Zeno effect, addressed to students and re-searchers with no previous knowledge on the subject. The prerequisites are the Schr¨odingerequation and the von Neumann notion of projective measurement.
1. Introduction and motivation
The evolution of an unstable quantum system is characterized by threedistinct regimes [1, 2]: a short-time region, where the decay is quadratic, anintermediate region, during which the exponential law sets in, and a long-time region, governed by a power law. A sketch (not in scale!) of such anevolution is given in Fig. 1.Unlike in classical (statistical) mechanics, where a decaying system istreated heuristically and the exponential decay law is easily obtained, thequantum analysis turns out to be involved and sometimes difficult to follow,even for experienced physicists. Scrutiny of the quantum evolution, gov-erned by the Schr¨odinger equation, unveils the presence of an unavoidablequadratic region at short (sometimes very short) times. This region was bap-tized “Zeno” by Misra and Sudarshan [3] in 1977. The classical allusion tothe sophist philosopher is due to an intriguing application: if one frequentlyinterrogates the system, checking whether it is still in its initial state, onecan slow down (and eventually stop) its evolution [1, 2, 4]. This is similar toZeno’s arrow, that would not reach its target if observed at a given position[5]. The purpose of this note is to give an introduction to this topic, addressedto students, young researchers and physicists with no previous knowledge onthe subject. This is the summary of a 70 minute lecture [6] delivered in Toru´n,Poland, on June 21th, 2012, during the 44th Symposium on MathematicalPhysics on “New Developments in the Theory of Open Quantum Systems”.
Exemplary OSID style a r X i v : . [ qu a n t - ph ] N ov Author and title] quadratic (Zeno)exponential power survivalprobability time Z − t τ Ze − γt t − α t Fig. 1: Survival probability of a decaying quantum system. The initial Zenoregion is followed by an exponential decay and finally superseded by a powerlaw. Notice that the extrapolation of the exponential law back to t = 0 yieldsa value Z that is in general (cid:54) = 1.The audience provided an excellent arena to test the pedagogical aspects ofthe lecture and helped me understand which facets of the problem are moredifficult to grasp. I can only hope that I succeeded in making my presentationas palatable as possible. Some of the examples investigated here have beenpresented elsewhere [2, 7]. I do not aim at novelty, but rather at clarity,sometimes at the expenses of rigor.These notes are a somewhat more detailed version of the lecture. Thelevel of the presentation will be kept as elementary as possible. The readeris invited to perform all calculations.In Sec. 2. we review the main features of the quantum evolution law.These are straightforward consequences of the Schr¨odinger equation. Weintroduce the quantum Zeno effect in Sec. 3.. As anticipated, it is a verygeneral, unavoidable by-product of the quantal dynamics. We then clarifythese general aspects by looking at the simplest non-trivial quantum me-chanical example (a two-level system) in Sec. 4. We briefly comment on thephysical and mathematical origin of the quantum Zeno region in Sec. 5. andon the “meaning” of a von Neumann projective measurement in Sec. 6.The general analysis of the Zeno effect is disguisingly simple. In Sec. 7. weturn to genuinely unstable systems, that require a quantum field theoreticaldescription, and derive a closed expression for the survival amplitude. Theanalysis makes use of an analytic continuation in the complex energy plane.Before embarking in this adventure, we remind in Sec. 8. how to perform Author and title]
2. The quantum mechanical evolution
We start off by scrutinizing the quantum-mechanical evolution law, fo-cusing on its short-time features. Let H be the Hamiltonian of a quantumsystem and | ψ (cid:105) = | ψ ( t = 0) (cid:105) its initial state. We shall set henceforth (cid:126) = 1and assume that all functions to be dealt with are sufficiently regular to ad-mit series expansions. We shall focus on the “survival” amplitude A andprobability p that the system has survived in its initial state | ψ (cid:105) at time t : A ( t ) = (cid:104) ψ | ψ ( t ) (cid:105) = (cid:104) ψ | e − iHt | ψ (cid:105) , (1) p ( t ) = |A ( t ) | = |(cid:104) ψ | e − iHt | ψ (cid:105)| . (2)Let the system evolve for a short time δt . The Schr¨odinger equation yields | ψ ( δt ) (cid:105) = e − iHδt | ψ (cid:105) = | ψ (cid:105) − iH | ψ (cid:105) δt − H | ψ (cid:105) ( δt ) + O(( δt ) ) ≡ | ψ (cid:105) + | δψ (cid:105) . (3)The short-time expansion (3) yields A ( δt ) = 1 − i (cid:104) H (cid:105) δt − (cid:104) H (cid:105) ( δt ) , (4) p ( δt ) = 1 − ( δt ) τ + O(( δt ) ) , (5)where (cid:104)· · ·(cid:105) ≡ (cid:104) ψ | · · · | ψ (cid:105) and τ − ≡ (cid:104) H (cid:105) − (cid:104) H (cid:105) , (6)is the Zeno time [2]. In deriving (5) from (4) the Hermitianity of H , ensuringthe reality of (cid:104) H (cid:105) , played a primary role. Notice that according to (4) thewave function evolves linearly away from the initial state, but the survivalprobability (of remaining in the initial state) evolves quadratically away from1, due to (5). Recall that due to the unitarity of the evolution, wave functionsare always normalized to unity: || ψ ( t ) || = || ψ (0) || = 1 , ∀ t : the tip of the statevector never leaves the unit sphere. The features of the short time evolutionare pictorially displayed in Fig. 2(a). Author and title] ψ (0) ψ ( δt ) δψ (a) unitary (b) non-unitary ψ (0) ψ ( δt ) δψ unit sphere Fig. 2: (a) Unitary evolution engendered by a Hermitian Hamiltonian. Theevolution takes place on the unit sphere: || ψ ( δt ) || = || ψ (0) || = 1. (b) Non-unitary evolution engendered by a non-Hermitian Hamiltonian. The tip ofthe state vector can leave the unit sphere (and enter the unit ball): || ψ ( δt ) || ≤|| ψ (0) || = 1. In both cases, δψ is linear in δt . Let us add a non-Hermitian part to the Hamiltonian: H (cid:48) = H − iV, (7)where V > .The new survival amplitude and probability read A (cid:48) ( t ) = (cid:104) ψ | ψ ( t ) (cid:105) = e − V t (cid:104) ψ | e − iHt | ψ (cid:105) , (8) p (cid:48) ( t ) = e − V t |(cid:104) ψ | e − iHt | ψ (cid:105)| . (9)A short-time expansion yields a linear behavior both for amplitude and prob-ability A (cid:48) ( δt ) = 1 − ( V + i (cid:104) H (cid:105) ) δt −
12 ( (cid:104) H (cid:105) − V − iV (cid:104) H (cid:105) )( δt ) + O(( δt ) ) , The term “optical” is due to the analogy with the interaction of light with a mediumthat is both refractive and absorptive. Such an interaction can be analyzed by introducinga complex refractive index. Analogously, the scattering and absorption of nucleons bynuclei can be treated by introducing effective neutron-nucleus interaction potentials andby averaging such effective potentials over many nuclei in order to obtains the neutron-matter (complex) optical potential. A consistent expression of V was first derived byFermi and Zinn [9]. Author and title] p (cid:48) ( δt ) = 1 − V δt + O(( δt ) ) . (11)Optical potentials “eat up” probability and account for decay channels. SeeFig. 2(b). The tip of the state vector can leave the unit sphere and enter theunit ball: || ψ ( t ) || ≤ || ψ (0) || = 1. [It would leave the unit ball if the opticalpotential − iV in (7) had the opposite sign.]In physics, one tends to regards property (5) as more “fundamental”, asit ensues from the Hermitianity of the Hamiltonian and the unitarity of theevolution, that are regarded as very general principles. Yet optical potentialshave their own charm and play an important role in effective descriptionsof decaying and dissipative systems. Nowadays they have been supersededby the rigorous mathematical framework of Gorini, Kossakowski, Sudarshanand Lindblad [10] that describes the physics of quantum dissipative systems[11, 12, 13].It is also worth noticing that the exponential law in a quantum contextis always the consequence of approximations of some sort. Examples of suchapproximations can be a macroscopic limit [14] or the intervention of anexternal apparatus, governed by classical laws, that interacts with the systeminvestigated [15]. If the Hamiltonian is composed of a free and an interaction parts H = H + H int (12)we can obtain an interesting expression, that sheds light on the meaning ofthe Zeno time. Let | ψ n (cid:105) be the eigenstates of the free Hamiltonian, that forma complete set H | ψ n (cid:105) = ω n | ψ n (cid:105) . (13)We require that the initial state be an eigenstate of the free Hamiltonianand (as it is customary in quantum field theory) that the interaction beoff-diagonal: H | ψ (cid:105) = ω | ψ (cid:105) , (cid:104) H int (cid:105) = 0 . (14)In this interesting case the Zeno time reads τ − = (cid:104) H (cid:105) = (cid:88) n (cid:104) ψ | H int | ψ n (cid:105)(cid:104) ψ n | H int | ψ (cid:105) (15)and depends only on the interaction Hamiltonian. Author and title] (a) lifetime (b) Zeno time ψ ψ n ψ n ψ f ψ ψ Fig. 3: (a) The lifetime γ in Eq. (16) contains only “on-shell” contributions:the delta function entails energy conservation ω f = ω ; ψ f is in general (very)degenerate (think of an atom in an S -wave emitting a photon: there is a 4 π degeneracy in the direction of emission). (b) The Zeno time τ Z in Eq. (15)explores the whole Hilbert space.Formula (15) should be compared to the Fermi “golden rule” [16] , yield-ing the inverse lifetime γ of a decaying quantum system: γ = 2 π (cid:88) f |(cid:104) ψ f | H int | ψ (cid:105)| δ ( ω f − ω ) , (16)where the summation (integral) is over the final states and the continuumlimit is implied.One comment. While (16) contains only “on-shell” contributions (becausethe delta function ensures energy conservation), the expression (15) exploresthe whole Hilbert space. See Fig. 3.
3. Quantum Zeno effect
The most familiar formulation of the QZE makes use of Von Neumannmeasurements, represented by one-dimensional projectors. Perform N mea-surements at time intervals τ = t/N , in order to check whether the systemis still in its initial state | ψ (cid:105) . After each measurement the system’s stateis “projected” back onto its initial state | ψ (cid:105) and the evolution starts anewaccording to Schr¨odinger’s equation with initial condition | ψ (cid:105) . [The systemcan also be projected onto an orthogonal state | ψ ⊥ (cid:105) , with (quadratic) proba-bility 1 − p ( τ ) = τ /τ , according to Eq. (5). As τ = O(1 /N ), such an eventbecomes increasingly unlikely as N increases.]The survival probability p ( N ) ( t ) at the final time t = N τ reads p ( N ) ( t ) = p ( τ ) N = p ( t/N ) N Fermi considered expression (16) the second golden rule. If you are curious about thefirst one, see pages 136 and 148 of
Nuclear Physics [16].
Author and title] (cid:28) tP (t) Fig. 4: Quantum Zeno effect for N = 5 “pulsed” Von Neumann measure-ments. The dashed (full) line is the survival probability without (with)measurements. The gray line is the interpolating exponential (18). As N increases, p ( N ) ( t ) → , t ]. The units on the abscissae arearbitrarily chosen for illustrative purposes. (cid:39) (cid:2) − ( t/N τ Z ) (cid:3) N N large −→ exp( − t /N τ ) N →∞ −→ , (17)where we made use of Eq. (5). For large N the quantum mechanical evolutionis slowed down and in the N → ∞ limit (infinitely frequent measurements) itis halted, so that the state of the system is “frozen” in its initial state. Thisis the QZE. It is a consequence of the short-time behavior (5).Observe that the survival probability after N pulsed measurements ( t = N τ ) is interpolated by an exponential law [17] p ( N ) ( t ) = p ( τ ) N = exp( N log p ( τ )) = exp( − γ eff ( τ ) t ) , (18)with an effective decay rate γ eff ( τ ) ≡ − τ log p ( τ ) . (19)For τ → N → ∞ ) one gets from (5) p ( τ ) (cid:39) exp( − τ /τ ), so that γ eff ( τ ) (cid:39) τ /τ , τ → . (20)The Zeno evolution for “pulsed” Von Neumann measurements is pictoriallyrepresented in Figure 4. Author and title]
4. The simplest non-trivial quantum mechanical example: thetwo-level system
Consider a two-level system undergoing Rabi oscillations. This is thesimplest nontrivial quantum mechanical example, for it involves 2 × H = H int = Ω σ = Ω( | + (cid:105)(cid:104)−| + |−(cid:105)(cid:104) + | ) = (cid:18) (cid:19) , (21)where Ω is a real number, σ j ( j = 1 , ,
3) the Pauli matrices and | + (cid:105) = (cid:18) (cid:19) , |−(cid:105) = (cid:18) (cid:19) (22)are eigenstates of σ . We are neglecting the energy difference between thetwo states |±(cid:105) . Let the initial state be | ψ (cid:105) = | + (cid:105) = (cid:18) (cid:19) , (23)so that the evolution yields | ψ ( t ) (cid:105) = e − iH int t | ψ (cid:105) = cos(Ω t ) | + (cid:105) − i sin(Ω t ) |−(cid:105) = (cid:18) cos Ω t − i sin Ω t (cid:19) . (24)The survival amplitude (1) and probability (2) and the Zeno time (6) or (15)read A ( t ) = cos Ω t, (25) p ( t ) = cos Ω t, (26) τ Z = Ω − , (27)respectively. The effective decay rate (19) if N measurements are performedin time t reads γ eff ( τ ) = τ Ω . (28)In this simple case, Eq. (20) is exact (and not simply an approximation forshort τ ). Look again at Figure 4. Author and title] Ο(δτ ) x z ψ(0) ψ(δτ) Ο(δτ) y Fig. 5: Short-time evolution of phase and probability: δτ ∼ /N .
5. Comments.
At the end of the day, the QZE is ascribable to the following mathematicalproperties of the Schr¨odinger equation: in a short time δτ ( ∼ /N ), the phaseof the wave function evolves like O( δτ ), while the probability changes byO( δτ ), so that P ( N ) ( t ) (cid:39) (cid:2) − O(1 /N ) (cid:3) N N →∞ −→ . (29)Stated differently, the projection onto the inital state “slowly” evolves awayfrom unity. This is sketched in Fig. 5 and is a very general feature of theSchr¨odinger equation, as well as of other “fundamental” evolution equationsin physics . Equations that do not have this feature (e.g. dissipative equa-tions) tend to be regarded as less fundamental, the consequence of approxi-mations of some sort.
6. Unraveling a von Neumann measurement
What is a (von Neumann [18]) measurement? This is a difficult question,that has been debated for decades and is still a subject of controversy [19].The mathematical answer is clear, the physical one is not. The evolution dueto a measurement process is non-unitary and many reasons lead many physi-cists (including myself) to think that a von Neumann projection is but aneffective description of a quantum measurement process. Stated differently,von Neumann’s projectors are a short-hand notation: they summarize the Such as the Maxwell equations and (super-)renormalizable quantum field theories.
Author and title] . This dynamics is neglected in most analyses of the QZE: in the elapseof time between two subsequent projections, the system evolves under theaction of a Hamiltonian H that does not account for its movement from theregion of space where one projection occurs to the (macroscopically) differentregion of space where the following projection will take place. Think of theexample in Sec. 4.: everything was neglected but the two-level structure of thesystem. The physics behind the measurement process is dismissed altogetherin a single sentence after Eq. (27)! We are so accustomed at computingprojections that we do not even think about the underlying physical processesanymore.We shall henceforth neglect all these problems and act pragmatically. Inthis section we forget philosophical standpoints and personal taste, and en-deavor to give a heuristic description of a quantum measurement, by propos-ing an effective model for the measuring “apparatus”. Clearly, we are noteven hoping of contributing to solving the mistery behind a quantum mea-surement. Let us show that the action of a measuring apparatus (performing the VonNeumann measurement) can be mimicked by a non-Hermitian Hamiltonian.Consider the Hamiltonian (notation as in Sec. 4.) H int = (cid:18) − i V (cid:19) = − iV + h · σ , h = (Ω , , iV ) T , (30)that yields Rabi oscillations of frequency Ω, but at the same time absorbsaway the |−(cid:105) component of the state vector, performing in this way a “mea-surement.” H is non-Hermitian, therefore probabilities are not conserved:we are focusing our attention only on the | + (cid:105) component. State |−(cid:105) can beviewed as a “decay channel”, according to the discussion in Sec.2.2.. There are situations where the system need not move between measurements, butthey are rare, and presuppose the existence of a control mechanism that keeps at a givenplace the physical system undergoing the measurement. An example is an atom in a givenposition that is shined by a laser: by observing the photons that are scattered/emitted,one can infer which atomic level is populated.
Author and title] =4 V Γ |+>|-> Ω (cid:10)t=(cid:25)P(t) 0:4 (cid:10)2 (cid:10)V = 10 (cid:10) Fig. 6: Survival probability for a system undergoing Rabi oscillations inpresence of absorption ( V = 0 . , , V = 0).Elementary algebra [and properties of SU(2)] yields e − iH int t = e − V t (cid:20) cosh( ht ) − i h · σ h sinh( ht ) (cid:21) , (31)where h = √ V − Ω and we supposed V (cid:29) Ω (this hypothesis is not vital,but makes the measurement “fast” and therefore effective). The survivalamplitude in the initial state (23) reads A ( t ) = (cid:104) ψ | e − iH int t | ψ (cid:105) = e − V t (cid:20) cosh( ht ) + Vh sinh( ht ) (cid:21) = 12 (cid:18) Vh (cid:19) e − ( V − h ) t + 12 (cid:18) − Vh (cid:19) e − ( V + h ) t . (32)Notice the presence of a slow and a fast decay. The survival probability P ( t ) = |A ( t ) | is shown in Fig. 6 for V = 0 . , , t → ∞ .Moreover, for large V , by expanding in the small parameter Ω /V , one finds P ( t ) (cid:39) (cid:18) V (cid:19) exp (cid:18) − Ω V t (cid:19) , (33)where the wrong normalization at t = 0 is an artifact of the approximation(the decay is always quadratic at short times and the above expansion be-comes accurate very quickly, on a time scale of order V − ). The effectivedecay rate γ eff ( V ) = Ω /V is counterintuitive. Try and show the left panelin Fig. 6 to a friend or a colleague of yours, who has no familiarity with theQZE, and ask the following question: what happens if one initially populates Author and title] | + (cid:105) and increases the decay rate V out of state |−(cid:105) ? Chances are thatyour friend/colleague will reply: state | + (cid:105) will be depleted faster. Not so: V appears in the denominator of the exponent in Eq. (33). Now show yourfriend the right panel in Fig. 6. The effective lifetime becomes larger as V increases, eventually halting the “decay” (absorption) of the initial state inthe V → ∞ limit. A larger V entails a more “effective” measurement of theinitial state. This is an interesting example of QZE.The global process described here can be viewed as a “continuous” (neg-ative result) measurement performed on the initial state | + (cid:105) . State |−(cid:105) iscontinuously monitored with a response time 1 /V : as soon as it becomespopulated, it is detected within a time 1 /V . The “strength” V of the obser-vation can be compared to the frequency τ − = ( t/N ) − of measurements inthe “pulsed” formulation of Sec. 3.. Indeed, for large values of V one getsfrom Eq. (33) γ eff ( V ) = Ω V = 1 τ V , (34)which, compared with Eq. (20), yields a cute relation between continuousand pulsed measurements [22] V (cid:39) /τ. (35) We now show that the non-Hermitian Hamiltonian (30) can be obtainedby considering the evolution engendered by a Hermitian Hamiltonian actingon a larger Hilbert space and then restricting the attention to the subspacespanned by {| + (cid:105) , |−(cid:105)} . Let H = Ω( | + (cid:105)(cid:104)−| + |−(cid:105)(cid:104) + | )+ (cid:90) dω ω | ω (cid:105)(cid:104) ω | + (cid:114) Γ2 π (cid:90) dω ( |−(cid:105)(cid:104) ω | + | ω (cid:105)(cid:104)−| ) , (36)that describes a two-level system coupled to a one-dimensional massless bosonfield in the rotating-wave approximation. Notice that the coupling is “flat”:the two-level system couples to all frequencies in the same way: this enablesus to pull out of the last integral a coupling constant √ Γ that is equal for allfrequencies. The state of the system at time t can be written as | ψ ( t ) (cid:105) = x ( t ) | + (cid:105) + y ( t ) |−(cid:105) + (cid:90) dω z ( ω, t ) | ω (cid:105) (37)and the Schr¨odinger equation reads i ˙ x ( t ) = Ω y ( t ) , Author and title] i ˙ y ( t ) = Ω x ( t ) + (cid:114) Γ2 π (cid:90) dω z ( ω, t ) , (38) i ˙ z ( ω, t ) = ωz ( ω, t ) + (cid:114) Γ2 π y ( t ) . By using the initial condition x (0) = 1 and y (0) = z ( ω,
0) = 0 one obtains z ( ω, t ) = − i (cid:114) Γ2 π (cid:90) t dτ e − iω ( t − τ ) y ( τ ) (39)and i ˙ y ( t ) = Ω x ( t ) − i Γ2 π (cid:90) dω (cid:90) t dτ e − iω ( t − τ ) y ( τ ) = Ω x ( t ) − i Γ2 y ( t ) . (40)Observe that in order to obtain this result the integral over ω has to beextended over the whole real line (from −∞ to + ∞ ). Also, (cid:82) t δ ( t − τ ) dτ =1 / − i Γ /
2. This is ascribable tothe afore-mentioned “flatness” of the continuum [there is no form factor orfrequency cutoff in the interaction term of Eq. (36)], which yields a purelyexponential (Markovian) decay of y ( t ).In conclusion, z ( ω, t ) drops out of the first two equations (38), that nowdescribe the (reduced) dynamics in the subspace spanned by | + (cid:105) and |−(cid:105) : i ˙ x ( t ) = Ω y ( t ) ,i ˙ y ( t ) = − i Γ2 y + Ω x ( t ) . (41)Of course, this dynamics is not unitary, for probability flows out of the sub-space, and is generated by the non-Hermitian Hamiltonian H = Ω( | + (cid:105)(cid:104)−| + |−(cid:105)(cid:104) + | ) − i Γ2 |−(cid:105)(cid:104)−| = (cid:18) − i Γ / (cid:19) . (42)This Hamiltonian is the same as (30) when one sets Γ = 4 V . QZE is obtainedby increasing Γ: a larger coupling to the environment leads to a more effective“continuous” observation on the system (quicker response of the measuringapparatus), and as a consequence to slower decay (QZE). Try and ask thesame tricky question mentioned after Eq. (33) to another friend/colleague.Rather than the left panel in Fig. 6, draw a figure in which level |−(cid:105) decaysto a photon field, and increase the coupling Γ between them.We leave it to the reader to judge whether the analysis of the measurementprocess proposed in this section is more satisfactory than that outlined in Sec. Author and title]
7. Genuine unstable systems and field theory
We shall now forget about quantum measurements and QZE and focus onthe non-exponential features of decay. The arguments given in Sec. 2.1. arevery general and cannot be rejected: decay cannot be exponential at shorttimes. However, it is of great interest to discuss this problem in a quantumfield theoretical framework. This will help us focus on the important roleplayed by the form factors of the interaction.We start by generalizing the two-level Hamiltonian (21) to N states | j (cid:105) ( j = 1 , . . . , N ) with different energies H = ω | + (cid:105)(cid:104) + | + N (cid:88) j =1 ω j | j (cid:105)(cid:104) j | = ω . . . ω . . . . . . ω N . (43)and (real) couplings H int = N (cid:88) j =1 g j ( | + (cid:105)(cid:104) j | + | j (cid:105)(cid:104) + | ) = g . . . g N g . . . g N . . . (44)In order to obtain a truly unstable system we need a continuous spectrum,so we consider the continuum limit ω j → ω, | j (cid:105) → √ δω | ω (cid:105) , g j → √ δωg ( ω ),with δω → H = H + H int = ω | + (cid:105)(cid:104) + | + (cid:90) dω ω | ω (cid:105)(cid:104) ω | + (cid:90) dω g ( ω )( | + (cid:105)(cid:104) ω | + | ω (cid:105)(cid:104) + | ) . (45)State | + (cid:105) is normalizable, but states | ω (cid:105) are not: (cid:104) + | + (cid:105) = 1 , (cid:104) ω | ω (cid:48) (cid:105) = δ ( ω − ω (cid:48) ) , (cid:104) + | ω (cid:105) = 0 . (46) {| + (cid:105) , | ω (cid:105)} is the eigenbasis of H and is a resolution of the identity | + (cid:105)(cid:104) + | + (cid:90) dω | ω (cid:105)(cid:104) ω | = 1 . (47)As before, we take as initial state | ψ (cid:105) = | + (cid:105) . The interaction of this statewith the continuum of states | ω (cid:105) is responsible for its decay and depends on Author and title] form factor g ( ω ). We assumed (with no loss of generality) g ( ω ) to bereal.It is worth stressing that the purpose of studying model (45) is verydifferent from the motivations that led us to analyze model (36). In Sec.6.2. we were interested in the QZE on level | + (cid:105) that arises when level |−(cid:105) is“measured”, while in this section we focus on the deviations from exponentialwhen level | + (cid:105) is coupled to a continuum. There is no level |−(cid:105) here .The Fourier-Laplace transform of the survival amplitude (1) for thismodel can be given a convenient analytic expression. The transform of thesurvival amplitude is the expectation value of the resolvent A ( E ) = (cid:90) ∞ dt e iEt A ( t ) = (cid:104) + | (cid:90) ∞ dt e iEt e − iHt | + (cid:105) = (cid:104) + | iE − H | + (cid:105) (48)and is defined for Im E > E − H = 1 E − H + 1 E − H H int E − H (49)one obtains A ( E ) = (cid:104) + | (cid:20) iE − H + 1 E − H H int iE − H ++ 1 E − H H int E − H H int iE − H (cid:21) | + (cid:105) = iE − ω + 1 E − ω (cid:90) dω |(cid:104) + | H int | ω (cid:105)| E − ω A ( E ) . (50)In the above derivation we used the resolution (47) of the identity and thefact that H int is completely off-diagonal in the eigenbasis of H [compare Eq.(14)]. The advantage of looking at the Fourier-Laplace transform (48) lies inthe fact that Eq. (50) is algebraic and can be solved to yield A ( E ) = iE − ω − Σ( E ) , (51)where the self-energy function Σ( E ) is related to the form factor g ( ω ) by asimple integrationΣ( E ) = (cid:90) dω |(cid:104) + | H int | ω (cid:105)| E − ω = (cid:90) dω g ( ω ) E − ω . (52) Although it would not be difficult to introduce it.
Author and title] | + (cid:105) and the continuum. Byinverting Eq. (48) we finally get A ( t ) = (cid:90) B dE π e − iEt A ( E ) = i π (cid:90) B dE e − iEt E − ω − Σ( E ) , (53)the Bromwich path B being a horizontal line Im E =constant >
8. Intermezzo: Analytic continuation on the second Riemannsheet
Consider the function F ( z ) = (cid:90) ∞ dE f ( E ) E − z (54)where z = x + iy ∈ C , f is a smooth function and E a real variable. F isan analytic function in the complex z plane, but has a (logarithmic) cut forpositive real z . We obtain F ( x ± i + ) = (cid:90) ∞ dE f ( E ) E − x ∓ i + = (cid:90) ∞ dEf ( E ) (cid:18) P E − x ± iπδ ( E − x ) (cid:19) , (55)where P denotes principal value. The discontinuity across the cut is therefore F ( x + i + ) − F ( x − i + ) = 2 πi (cid:90) ∞ dEf ( E ) δ ( E − x ) = 2 πif ( x )( x > . (56)Clearly, in Eq. (56) the function F ( x ± i + ) is evaluated on the first Riemannsheet, immediately above and below the cut on the positive real half-line.Let’s now smoothly cross the positive real half-line, going from the firstto the second Riemann sheet. The value of F ( z ) above the real axis, on thefirst sheet, and below it, on the second sheet, is the same by definition: F ( x + i + ) = F II ( x − i + ) ( x > , (57) Author and title] !!!!"""" ! !!!!"""" %%%%$$$$ !!!!"""" ! &&&&$$$$ !!!!"""" ’’’’$$$$ ! !!!!"""" (((($$$$ Fig. 7: Analytic continuation across the cut in the complex E -plane. a) Eq.(54); b) Eq. (60); c)-e) Eq. (61).where F II is the function evaluated on the second Riemann sheet. By usingEqs. (56)-(57), one gets F II ( x − i + ) = F ( x − i + ) + 2 πif ( x ) ( x > . (58)Therefore the “jump” (56) of F ( z ) evaluated on the two edges of the cut(on the first Riemann sheet) is equal to the difference of the values of thefunction evaluated on the second and first sheet. By analytically extendingformula (58) one obtains F II ( z ) = F ( z ) + 2 πif ( z ) . ∀ z ∈ C . (59)It is obvious that in the above considerations we are implicitly assumingthat analytic continuation is licit. Assume now that F ( z ) in Eq. (54) bedefined for Im z > z <
0. It iseasy to see that the definition F ( z ) = (cid:90) Γ dE f ( E ) E − z , for Im z > Author and title] ∞ by remaining below z . Notice that E in Eq. (60) takes complex valuesand Γ can be arbitrarily deformed, as far as its configuration with respect tothe singularity z is respected. See Fig. 7b.The extension to the case Im z < z smoothlycrosses the positive real axis, going to the second Riemann sheet, the contourintegration in the complex E plane remains below z , respecting the positionof the singularity. This yields again the result (59): the contour is firstdeformed in order to remain below z , then deformed into a small circle, thatruns counterclockwise ariound z , plus the original contour F ( x + iy ) y< −→ F II ( x + iy ) = (cid:90) Γ dE f ( E ) E − x − iy = (cid:90) ∞ dE f ( E ) E − x − iy + 2 πif ( x + iy )= F ( x + iy ) + 2 πif ( x + iy ) (61)This is identical to (59). In this case the difference between F and F II isgiven by the pole. See Fig. 7c-e. These beautiful mathematical ideas will bevery useful to analyze the behavior of the propagator (53).
9. Analytic continuation of the propagator
The function A ( E ) in Eqs. (51), (53) has a branching point at E = ω g ,the lower bound of the continuous spectrum of the Hamiltonian H , a cutthat extends to E = + ∞ and no additional sigularities on the first Riemannsheet, while singularities can appear on the second sheet. These importantfeatures were studied by Araki et al. [23] and Schwinger [24] in the 50’s .Indeed, A ( E ) is defined for Im E >
0, so that its Fourier transform, thesurvival amplitude (53), converges for t >
0. When the self-energy functionis analytically continued to the second Riemann sheet, the contour must bemodified so that its position with respect to the singularity is mantained.The initial state has energy ω > ω g and is therefore embedded in thecontinuous spectrum of H . If | Σ( ω g ) | < ω (which happens for sufficientlysmooth form factors and small coupling), the resolvent is analytic in the wholecomplex plane cut along the real axis (continuous spectrum of H ) [23, 24].On the other hand, there exists a pole E pole located just below the branchcut in the second Riemann sheet, solution of the equation E pole − ω − Σ II ( E pole ) = 0 , (62) Those were the golden years of renormalization in quantum field theory.
Author and title] I pole E ω II E Fig. 8: The pole E pole on the second Riemann sheet is (coupling constant) -close to ω : see Eqs. (53) and (62). We drew the circle of convergence of anasymptotic expansion around ω . The derivation of Eq. (51) from Eq. (48)requires the definition of the self-energy function (52). Try and understandwhich mathematical hypotheses are needed.Σ II being the determination of the self-energy function in the second sheet.Remember that the self-energy function is a “small” quantity, being propor-tional to the square of the coupling between level | + (cid:105) and the continuum:the pole E pole is therefore very close to ω . See Figure 8.The pole has a real and imaginary part E pole = ω + δω − iγ/ , (63)that can be easily computed by following the mathematical technique out-lined in the previous section δω = Re Σ II ( E pole ) (cid:39) Re Σ( ω + i + ) = P (cid:90) dω g ( ω ) ω − ω , (64) γ = − II ( E pole ) (cid:39) − ω + i + ) = 2 πg ( ω ) . (65)In the above formulas, δω is the energy shift and γ the inverse lifetime,according to the Fermi “golden” rule [16]. Both quantities are written atsecond order in the coupling constant. Check that γ is the same quantity Author and title] .In conclusion, the survival amplitude (53) has the general form A ( t ) = A pole ( t ) + A cut ( t ) , (67)where A pole ( t ) = e − i ( ω + δω ) t − γt/ − Σ (cid:48) II ( E pole ) , (68)is due to the pole contribution (62) and A cut ( t ) = i π (cid:90) cut dE e − iEt E − ω − Σ( E ) , (69)is the branch-cut contribution, as explained in the previous section: see Fig.7e).It is not difficult to see that, if the coupling is small, at intermediate timesthe pole contribution dominates the evolution and P ( t ) (cid:39) |A pole ( t ) | = Ze − γt , Z = (cid:12)(cid:12) − Σ (cid:48) II ( E pole ) (cid:12)(cid:12) − , (70)where Z , the intersection of the asymptotic exponential with the t = 0 axis,is the so-called wave-function renormalization. This explains the behaviorsketched in Fig. 1. It would be interesting to see [1] that the cut contribution(69) cannot be neglected at short and long times, where it yields the quadraticZeno behavior and the power tail, respectively. In order to obtain a purely exponential decay, one can simply neglect thebranch cut contribution altogether and retain only the dominant contributionof the pole singularity. An interesting way to obtain the desired result is toreplace the self-energy function with a constant (equal to its value at thepole) in Eq. (51): A ( E ) −→ iE − ω − Σ II ( E pole ) = iE − E pole ≡ A W ( E ) , (71) The derivation of Eqs. (64)-(65) is left as an exercise (a very useful one). Be careful inderiving γ in (65), you might miss a factor 2. Modern literature (unlike classic literature)is plagued by missing factors 2. The correct solution is obtained by using the formulalim γ → γE + γ = 2 πδ ( E ) , (66)that is valid because γ is a small quantity (second order in the coupling constant), andneglecting fourth-order terms in the coupling constant. Author and title] A ( t ) = exp( − iE pole t ), without short- and long-time corrections .Another nice way to obtain a purely exponential decay is to replace theform factor g in Eq. (45) by a constant value, say (cid:112) γ/ π . This is a usefulexercise. (Hint: follow the same strategy as in Sec. 6.2..)Another important problem is the duration of the non-exponential Zenoregion and the onset to the power law. The answer to these questions requirescareful evaluation of the cut contribution (69). One finds that the Zeno regionis superseded by the exponential decay after a time of the order of the inversefrequency cutoff of the form factor g in the interaction Hamiltonian (45) andthe exponential is superseded by a power law after a time of the order ofa significant number (say 10 ) of lifetimes. However, these conclusions aremodel-dependent and neglect important numerical factors. As a general rule,time evolutions in quantum field theory are a complex problem [26] and leadto the inverse Zeno effect [27, 28, 17]
10. Conclusions and apologies.
The title of these notes is “All you ever wanted to know about the quan-tum Zeno effect in 70 minutes”. Admittedly, I lied: my lecture would havelasted 90 minutes, if my chairman had not (very politely) stopped me. How-ever, the title contains a second, more deceitful lie: these notes are by nomeans all you ever wanted to know about the QZE. However, I dont feelguilty about the second (white) lie. The main purpose of a lecture is not toexplain everything; rather, it is to make students curious, so that they cango and deepen the subject. This contains, in embryo, what we nowadays callcuriosity-driven research. If I managed to get my students interested, mylecture was successful.
Acknowledgments
I would like to thank D. Chru´sci´nski, A. Jamio(cid:32)lkowski and M. Michalskifor the kind invitation to lecture at the 44th Symposium on MathematicalPhysics “New Developments in the Theory of Open Quantum Systems”, heldin Toru´n, Poland, in June 20-24, 2012. Special thanks to B. Bylicka and F.Pepe for suggestions and comments and to P. Facchi and H. Nakazato formany early conversations on the quantum Zeno phenomenon. Many thanks My former teacher M. Namiki used to tell me that great physicists know in advancethe result they want to get and use mathematics in a “creative” way to obtain what theyneed. The older I get, the more I agree.
Author and title]
Bibliography
1. H. Nakazato, M. Namiki and S. Pascazio, Int. J. Mod. Phys. B , 247 (1996).2. P. Facchi and S. Pascazio, Progress in Optics, edited by E. Wolf (Elsevier, Amsterdam) , 147 (2001); J. Phys. A: Math. Theor. , 493001 (2008).3. B. Misra and E. C. G. Sudarshan, J. Math. Phys. , 756 (1977).4. K. Koshino and A. Shimizu, Phys. Rep. , 191 (2005).5. Aristotle, Physics 4 239b10.6. http://tv.umk.pl/
7. P. Facchi and S. Pascazio, “Unstable systems and quantum Zeno phenomena in quan-tum field theory” Quantum Probability and White Noise Analysis
XVII , 222 (2003).[quant-ph/0202127]8. E. Fermi, Ricerca Scientifica , 13 (1936); E. Fermi, Nuovo Cimento , 407 (1954);H. Feshbach, C.E. Porter and V.F. Weisskopf, Phys. Rev. , 448 (1954).9. E. Fermi and W. H. Zinn, Phys. Soc. Cambridge Conf. Rep. 92, Chicago, 1947. [SeeEnrico Fermi, Collected Papers, ed. E. Segr`e (University of Chicago Press, 1962),Paper no. 220.]10. V. Gorini, A. Kossakowski and E. C. G. Sudarshan , J. Math. Phys. , 821 (1976);G. Lindblad, Comm. Math. Phys.
119 (1976)11. R. Alicki and K. Lendi,
Quantum Dynamical Semigroups and Applications (Springer,Berlin, 1987).12. U. Weiss,
Quantum Dissipative Systems , (World Scientific, Singapore, 2000).13. H.-P. Breuer and F. Petruccione,
The Theory of Open Quantum Systems (OxfordUniv. Press, Oxford, 2007).14. H. Nakazato and S. Pascazio, Phys. Rev. A , 1066 (1993).15. P. Facchi, S. Pascazio and A. Scardicchio, Phys. Rev. Lett. , 61 (1999).16. E. Fermi, Rev. Mod. Phys.
87 (1932);
Nuclear Physics (University of Chicago,Chicago, 1950) p. 136, 142, 148;
Notes on Quantum Mechanics. A Course Given atthe University of Chicago in 1954 , edited by E. Segr`e (University of Chicago, Chicago,1954) Lec. 23.17. P. Facchi, H. Nakazato and S. Pascazio, Phys. Rev. Lett. , 2699 (2001).18. J. von Neumann, Die Mathematische Grundlagen der Quantenmechanik (Springer,Berlin, 1932). [English translation by E. T. Beyer,
Mathematical Foundation of Quan-tum Mechanics (Princeton University Press, Princeton, 1955)].19. M. Schlosshauer, J. Kofler and A. Zeilinger, “A Snapshot of Foundational AttitudesToward Quantum Mechanics”, Stud. Hist. Phil. Mod. Phys. , 222 (2013) ([quant-ph]arXiv:1301.1069).20. J.A. Wheeler and W.H. Zurek, eds, Quantum Theory and Measurement (PrincetonUniversity Press, 1983).
Author and title]
21. M. Namiki and S. Pascazio, Phys. Rept. , 301 (1993).22. L.S. Schulman, Phys. Rev. A , 1509 (1998).23. H. Araki, Y. Munakata, M. Kawaguchi and T. Goto, Progr. Theor. Phys. (1957)419.24. J. Schwinger, Ann. Phys. (1960) 169.25. G. Gamow, Z. Phys. , 204 (1928); V. Weisskopf and E.P. Wigner, Z. Phys. , 54(1930); , 18 (1930); G. Breit and E.P. Wigner, Phys. Rev. , 519 (1936).26. C. Bernardini, L. Maiani and M. Testa, Phys. Rev. Lett. , 2687 (1993); L. Maianiand M. Testa, Ann. Phys. (NY) , 353 (1998); P. Facchi and S. Pascazio, Phys.Lett. A , 139 (1998); Physica A , 133 (1999); I. Joichi, Sh. Matsumoto, and M.Yoshimura, Phys. Rev. D , 045004 (1998); R.F. Alvarez-Estrada and J.L. S´anchez-G´omez, Phys. Lett. A , 252 (1999); A.D. Panov, Physica A , 193 (2000); I.Antoniou, E. Karpov, G. Pronko and E. Yarevsky, Phys. Rev. A , 795 (2011)27. S.R. Wilkinson, C.F. Bharucha, M.C. Fischer, K.W. Madison, P.R. Morrow, Q. Niu,B. Sundaramand M.G. Raizen, Nature , 575 (1997); M.C. Fischer, B. Guti´errez-Medina and M.G. Raizen Phys. Rev. Lett. , 040402 (2001)28. A.M. Lane, Phys. Lett. A
359 (1983); W.C. Schieve, L.P. Horwitz and J. Levitan,Phys. Lett. A , 264 (1989) A.G. Kofman and G Kurizki, Nature , 546 (2000);J. ˇReh´aˇcek, J. Peˇrina, P. Facchi, S. Pascazio, and L. Miˇsta, Phys. Rev. A62