Conservation laws in quantum noninvasive measurements

Stanisław Sołtan,1 Mateusz Frączak,1 Wolfgang Belzig,2, ∗ and Adam Bednorz1 1Faculty of Physics, University of Warsaw, ul. Pasteura 5, PL02-093 Warsaw, Poland 2Fachbereich Physik, Universität Konstanz, D-78457 Konstanz, Germany (Dated: March 22, 2021) Conservation principles are essential to describe and quantify dynamical processes in all areas of physics. Classically, a conservation law holds because the description of reality can be considered independent of an observation (measurement). In quantum mechanics, however, invasive observa- tions change quantities drastically, irrespective of any classical conservation law. One may hope to overcome this nonconservation by performing a weak, almost noninvasive measurement. Interest- ingly, we find that the nonconservation is manifest even in weakly measured correlations if some of the other observables do not commute with the conserved quantity. Our observations show that conservation laws in quantum mechanics should be considered in their specific measurement context. We provide experimentally feasible examples to observe the apparent nonconservation of energy and angular momentum.

I. INTRODUCTION has been noticed long ago by Wigner, Araki and Yanase (WAY)[4–6], later been discussed in the context of con- sistent histories [7], modular values [8], and the quan- Conserved quantities play an important role in both tum clock [9]. The generation, measurement, and con- classical and quantum mechanics. According to the clas- trol of quantum conserved quantities, in particular, an- sical Noether theorem, the invariance of the dynamics of gular momentum, has become interesting recently, both a system under specific transformations [1] implies the experimentally and theoretically [10–12]. Measurements conservation of certain quantities: translation symme- incompatible with energy lead to thermodynamic cost try in time and space results in energy and momentum [13–15]. conservation, respectively, rotational symmetry in angu- The quantum objectivity is one aspect of the general lar momentum conservation and gauge invariance in a concern of Einstein [16] and Mermin [17] if the (quan- conserved charge. In quantum mechanics, the observ- tum) Moon exists when nobody looks. The randomness ables (in the Heisenberg picture) are time-independent of quantum mechanics does not exclude objective reality when they commute with the Hamiltonian. Furthermore, [18]. Here, we assume that objective observations should some conserved quantities, like the total charge, commute be noninvasive i.e., leaving the probed system unchanged. also with all observables. We shall call them supercon- Unfortunately, unlike the classical case, the fundamental served. Classically all conserved quantities are also super- uncertainty prevents a completely noninvasive measure- conserved. In high energy nomenclature the former are ment in quantum mechanics [19]. Hence, the only re- known as on-shell conserved whereas the latter are called maining possibility seems to be to consider the limit of off-shell conserved [2]. The concept of superconservation weak measurements, which are almost noninvasive. The is closely related to the superselection rule which consti- objectivity based on weak measurements can lead to un- tutes an additional postulate that the set of observables expected results such as weak values [20] or the violation is restricted to those commuting with the superconserved of the Leggett-Garg inequality [21–23]. Unlike the stan- operators [3]. dard projection, which is highly invasive, the extraction Conservation principles become less obvious when one of objective values from weak measurements requires a tries to verify them experimentally. While an ideal clas- special protocol involving the subtraction of a large de- sical measurement will keep the relevant quantities un- tection noise. Therefore, such objectivity is debatable

arXiv:1907.06354v6 [quant-ph] 19 Mar 2021 changed, neither a nonideal classical nor any quantum [24, 25]. In our opinion, weak quantum measurements are measurement will necessarily reflect the conservation ex- the closest counterparts of classical measurements [26], actly. Even the smallest interaction between the system so they are prime candidates to define objective reality and the measuring device (detector) may involve a trans- and, consequently, conservation principles are expected fer of the conserved quantity. The system might become to hold in systems with an appropriate invariance. a coherent superposition of states with different values In this paper, we will show that for quantum mea- of a conserved quantity (e.g. energy), or in the case of surements in the weak limit superconservation holds but a superconserved quantity an incoherent mixture (e.g. quantities such as energy, momentum, and angular mo- charge). The problem of proper modeling of the mea- mentum apparently violate conservation even if an ap- surement of quantities incompatible with conserved ones propriate symmetry results in classical conservation law. The violation of conservation appears in third-order time correlations as we illustrate in simple model systems (Fig. 1). The violation is caused by (at least two) other ∗Electronic address: [email protected] observables that are not commuting with the conserved 2

commuting) projections onto the eigenspace of the value time q (i.e. Qˆ|ψqi = q|ψqi). Now, the superselection postulate t' 1 says that the state of the system ρˆ is always an incoherent 1 t 1 P ˆ ˆ ˆ t1 t'1 mixture q PqρˆPq, if Q is superconserved. Then the pro- 2 jective measurement of Aˆ will not alter the q−eigenspace ˆ P ˆ ˆ t as there exists a decomposition A = aPqa with Pqa 2 t q,a 3 3 being the projection onto the joint eigenspace of Qˆ and Aˆ with respective eigenvalues q and a. For instance, if the initial state is already a q−eigenstate then it will remain Δq Δq g→0 such an eigenstate after the projection. For general mea- SYSTEM DETECTOR surements, positive operator-valued measures (POVM), represented by Kraus operators Kˆc (the index c can rep- FIG. 1: (top) Three weak measurements. The detectors are resent an eigenvalue of Aˆ, Qˆ or both but in general it can P † initially independent and couple instantly to the system at t1 be arbitrary) such that Kˆ Kˆc = 1ˆ, the state ρˆ will 0 c c (or t1), t2 and t3. The conserved quantity is measured (empty ˆ ˆ † 0 collapse to ρˆc = KcρˆKc , normalized by the probability circles) either at time t1 < t2 or t1 > t2. The outcomes 1 and 0 Trˆρc. In principle Kˆc can act within q-eigenspaces, i.e. 1 inferred from the three-point correlator might differ, even ˆ ˆ ˆ ˆ for a conserved quantity in the quantum case. (bottom) Fail- c = qa and Kqa = PqKaPq. In the most general case, the ure of conservation in the weak measurement. The transfer superconserving Kraus operator reads ∆q of the conserved quantity q between the system and de- ˆ ˆ ˆ ˆ tector does not scale with the measurement strength g. Kq0aq = Pq0 KaPq. (1) It means that the superconserved value can change but the system remains an incoherent mixture of q- one. We write down an operational criterion to witness eigenstates. This applies e.g. to a charge measurement the violation of a conservation principle and discuss when in a quantum dot (which is superconserved), where the it is satisfied. Then we propose a feasible experiment charge can leak out into an incoherent bath. The (nor- probing position and magnetic moment of a charge in a mally) conserved quantities do not impose any additional circular trap. Last, considering an imperfect conserva- postulates so the state can be a coherent superposition of tion or measurement of the quantity we develop then a the states of different values of energy, angular momen- Leggett-Garg-type test of objective realism. tum, etc. A projective measurement of Aˆ which does not commute with Q is enough to turn a q-eigenstate into a superposition. Now, if we try to postulate a II. SUPERCONSERVATION POVM with superconserving Kraus operators then the actually measured operator involves a linear combina- ˆ P ˆ ˆ † ˆ ˆ ˆ ˆ The physical Hermitian quantity Q, defined within tion of qq0 PqKaPq0 KaPq so it must commute with Q the system is conserved when [H,ˆ Qˆ] = 0 for the sys- which would become superconserved. This is a modern tem’s Hamiltonian Hˆ . The Qˆ can be superconserved if version of the WAY theorem [4–6], that the measurement ˆ P ˆ † ˆ ˆ there exists a set A of allowed Hermitian observables of A = a aKaKa not commuting with Q, cannot con- such [A,ˆ Qˆ] = 0 for every Aˆ ∈ A. In principle, one sist of only Kˆq0aq defined above with q−eigenspaces of could make every conserved quantity superconserved by Qˆ. Such a formulation is simpler than the original WAY a proper choice of the set A. However, for instance, for a theorem, as is does not need the dicussion of an auxil- component of angular momentum we would have to ex- iary detector. On the other hand both approaches are clude position and momentum or even other components equivalent due to the Naimark theorem [27]. of angular momentum. Instead, we will distinguish quan- The unavoidability of coherent superpositions of only tities that are conserved but not superconserved by al- conserved values is the key problem considered here. lowing measurement of observables not commuting with them. An example of a superconserved quantity is the total electric charge, while the set of observables and pos- III. WEAK MEASUREMENTS AND sible initial state density matrices is restricted to those OBJECTIVE REALISM that do not change the charge. This is also known as superselection rule. Whether this rule is an axiom or a Strong projections are highly invasive, i.e. ρˆ → practical assumption, depending in the considered Hamil- P ˆ ˆ P P ρˆP changes the state ρ very much. On the other tonian, is a matter of debate [3], because one can in hand, unlike classical physics, quantum mechanics does principle imagine dynamics without e.g. charge super- not offer completely noninvasive measurements. The selection. Nevertheless, here we treat superselection and only possibility is weak measurement [20] where we apply superconservation as an axiom for certain quantities, like Kraus operators charge. Let us assume the decomposition of a super- ˆ P ˆ ˆ ˆ 1/4 ˆ 2 conserved quantity Q = q qPq where Pq are (mutually Kg(a) = (2g/π) exp(−g(A − a) ) , (2) 3 with the measurement strength g → 0+ so that the state IV. CONSERVATION IN WEAK almost does not change, MEASUREMENTS Z ˆ ˆ ˆ ˆ ρˆ → daK(a)ˆρK(a) ≈ ρˆ − g[A, [A, ρˆ]]/2. (3) The conservation means that the measurable corre- lations (7) involving the conserved quantity q(t) cor- 0 The actually measured probability p (a) = responding to Qˆ(t) = Qˆ will not depend on t. It is ˆ ˆ † TrK(a)ˆρK (a) of the outcome a at the state ρˆ has true at the single average, where hq(t)i = hQˆi. In- a form of convolution terestingly, also for second order correlations the order Z p0(a) = D(a − A)p(A)dA, (4) of measurements has no influence on the result, since hq(t1)a(t2)i = h{Q, A(t2)}i/2 is independent of t1. How- p(A) = hδ(A − Aˆ)i = Trδ(A − Aˆ)ˆρ, ever, the situation changes for three consecutive mea- surements (see Fig. 1), since in the last line of (7) the with the dominating detection noise D(a) = time order of operators matters, which has been demon- p −2ga2 2 2g/πe , with ha iD = 1/4g, diverging for strated also experimentally [30]. Considering the differ- g → 0. The quantity p(A) is the probability of the ence of two measurement sequences Q → A → B and outcome A in the case of a strong, projective mea- A → Q → B, we obtain the jump (which is absent in 0 surement g → ∞ (p(A) = limg→∞ p (A)), to which perfectly noninvasive classical measurements [26]) the noise D is added. Therefore we can expect that ˆ ˆ ˆ ˆ ˆ ˆ p(A) is in fact the probability that the quantity Aˆ has h{Q, {A(t2), B(t3)}} − {A(t2), {Q, B(t3)}}i = (9) objectively the value A. Unfortunately such an idea fails h[[Q,ˆ Aˆ(t2)], Bˆ(t3)]i ≡ 4h∆qa(t2)b(t3)i . in sequential measurements, as shown already in [20], This quantity will show up as jump ∆q = q(t ) − q(t ) because p(A, B) can be negative when measuring first Aˆ 1 2 ˆ ˆ ˆ at t1 = t2, when measuring hq(t1)a(t2)b(t3)i. The jump and then B, such that [A, B] 6= 0. The original concept will be non-zero for Q not commuting with A and B. [20] involved postselection, i.e. the last measurement Obviously, for superconserved quantities Q (commuting is strong, not weak, and the conditional probability with every measurable observable) the jump is absent. p(A, B)/p(B) is considered. However, the strength of The violation of the conservation principle is caused by the last measurement is irrelevant, as the system is not the measurement of Aˆ – not commuting with Qˆ – which touched any more. In our approach, all measurements, allows transitions between spaces of different q with the including the last one, can be assumed weak. jump size ∆q not scaled by the measurement strength g, We shall discuss the problem of a negative p in Sec- see Fig. 1. This difference is transferred to the detec- tion VI. Nevertheless, p(A, B, ...) is well defined in the tor, assuming that the total quantity (of the system and limit g → 0 and we can probe it, hence, assuming that it detector) is conserved regardless of the system-detector reflects a property of the system. Note that this construc- ˆ interaction. This observation can be compared to the tion is still correct in the superconserved case because A, WAY theorem, which applies to projective or general Kˆa and the state ρˆ is commuting with Qˆ so Kˆa splits ˆ ˆ measurements. Here, we have shown that even taking into a simple sum of Kqa. The actual form of Ka can be the special limit of noninvasive measurement, the non- different but the outcome is almost independent in the commuting quantity causes a jump in third (and higher) limit g → 0. In the lowest order we can also neglect all ˆ th correlations. We can call it weak-WAY theorem, as both Kq0aq. In the g → 0 limit, the n -order correlation of a input (the special construction of g−dependent measure- ˆ ˆ ˆ ˆ sequence of measurements A, B, C, D with respect to p ments) and the output (correlations) are based on weak 0 (or also p if the quantities are different) reads [26, 28, 29] measurements. Note that imposing the condition that ˆ ha(t)i = hAˆ(t)i, (5) the jump (9) vanishes, equivalent conservation of Q at the level of third-order correlation for an arbitrary state ˆ ˆ ha(t1)b(t2)i = h{A(t1), B(t2)}i/2, (6) ρˆ (allowed by superselection rules if any apply), namely ha(t )b(t )c(t )i = h{Aˆ(t ), {Bˆ(t ), Cˆ(t )}}i/4, (7) 1 2 3 1 2 3 [[Q,ˆ Aˆ], Bˆ] = 0 (10) ha(t1)b(t2)c(t3)d(t4)i ˆ ˆ = h{Aˆ(t ), {Bˆ(t ), {Cˆ(t ), Dˆ(t )}}}i/8 (8) for all allowed observables A and B suffices to keep con- 1 2 3 4 servation also at all higher order correlations. Then Qˆ is for t1 < t2 < t3 < t4 with the anticommutator {A,ˆ Bˆ} = not necessarily superconserved, it can commute with ob- AˆBˆ + BˆAˆ and quantum averages hXˆi = TrXˆρˆ. servables to identity, like momentum and position. This At this stage, we would like to note that the classi- subtle difference between the weak-WAY and the tradi- cal counterpart of this protocol replaces the anticommu- tional WAY theorem in sketched in Fig. 2 tators like {A,ˆ Bˆ} by simple products of a phase space As an example we can take the basic two-level system functions A(q, p) and B(q, p). The invasiveness (3) can (|±i basis) with the Hamiltonian Hˆ = Qˆ = ~ω|+ih+| be reduced to zero and the time order of observables is and Aˆ = Bˆ = Xˆ = |+ih−| + |−ih+|. Then, with irrelevant. For a more detailed analysis of classical-to- ω > 0 the ground state is |−i and the third order corre- quantum correspondence we refer the reader to [26]. lation hh(t1)x(t2)x(t3)i for the ground state for t3 > t1,2 4

general weak DETECTOR v g CLOCK

WAY SYSTEM weak-WAY

FIG. 4: Detection model based on a clock. The clock is a localized particle traveling with a constant speed v. The in- teraction between the detector and system takes place only when the clock is passing the interaction point.

√ less position 2Xˆ =a ˆ† +a ˆ = Aˆ = Bˆ, we find for the jump h∆hx(0)x(t)i = − ω cos(ωt)/4 independent of the FIG. 2: The difference between the WAY and the weak- ~ WAY theorem. The former applies to general measurement state of the system (see Appendix A). As illustrated in and shows that lack of coherence between eigenstates of Qˆ in Fig. 3, the jump becomes unobservable at high tempera- Kraus operators (1) leads to superconservation of the mea- tures since the average energy hhi = ~ω/[exp(~ω/kT )−1] sured quantity. The latter applies to weak measurements (2), increases with temperature. leading to a weaker condition forthe observed conservation. The previously discussed very simple examples illus- trate the fundamental finding of our manuscript. If one tries to verify the conservation of energy while measur- 0.8 ing an other observable that is not commuting with the

i 0.7 Hamiltonian it is possible to find a violation of the en- h

h ergy conservation. It constitutes a pure quantum effect , 0.6 i

) since it vanishes at high temperature, where the classi- + 0.5 cally expected conservation holds. One could object that (0

x 0.4 performing a series of measurements already breaks time-

(0) 0.3 translational symmetry and therefore the total energy is

hx 0.2 not conserved. However, one can keep the time sym-

∆ metry by replacing the detector-system interaction by a h 0.1 clock-based detection scheme [9], see Fig. 4. The total 0.0 0.0 0.5 1.0 1.5 2.0 Hamiltonian reads kT/¯hω Hˆ + Hˆx + Hˆz + HˆI (12)

where Hˆ is the system’s part, Hˆx – the detector’s part, FIG. 3: The non-conserving jump for τ = 0 (thick lines) + Hˆz - the clock’s part, and finally HˆI is the interaction compared to the average energy (thin lines) for the two-level between the clock, the system and detector. Each part system with level spacing ~ω (red) and the harmonic oscilla- is time-independent so the time translation symmetry is tor with eigen frequency ω (blue). At high temperatures the jump becomes unobservable and the classical conservation is preserved. Both the detector and the clock can be repre- sented by single real variables, x and z. Now, to measure restored. All quantities are normalized to ~ω. the system’s Aˆ at time t1 we set Hˆx = 0 and

Hˆz = vpˆz, HˆI = gAδˆ (ˆz)ˆpx (13) reads ~ω(1 − θ(t2 − t1)) cos(ω(t2 − t3))/2. The jump is h∆hx(0)x(τ)i = ~ω cos(ωτ)/2 for ∆h = h(0−) − h(0+). where pˆx,z are conjugate (momenta), i.e. pˆx = −i~∂/∂x The result can be generalized to a thermodynamical en- and g → 0 is a weak coupling constant. The initial state semble with a finite temperature T and reads (see Ap- (at t = 0) reads ρˆρˆxρˆz, where both ρˆx,z = |ψx,zihψx,z| pendix A) are taken as Gaussian states

h∆hx(0)x(τ)i = ~ω cos(ωτ) tanh(~ω/2kT )/2 . (11) −1/4 2 ψz(z) = (2πσ) exp(−(z + vt1) /4σ), For increasing temperature the jump diminishes as illus- −1/4 2 ψx(x) = (π/2) exp(−x ) , (14) trated in Fig. 3. Another basic example is the harmonic oscillator with respectively. For small g and σ the interaction effectively ˆ ˆ † † H = Q = ~ωaˆ aˆ with [ˆa, aˆ ] = 1. Taking the dimension- occurs at time t = t1 and, in the end (after the clock 5 decouples the system and the detector again) to lowest order we find (see details in Appendix B)

hxi ' ghAˆi = gha(t1)i . (15) For sequential measurements one simply adds more in- dependent detectors and clocks, obtaining in the lowest ~X order of g

2 hxAxBi ' g ha(t1)b(t2)i, V 3 hxAxBxC i ' g ha(t1)b(t2)c(t3)i, (16) 4 hxAxBxC xDi ' g ha(t1)b(t2)c(t3)d(t4)i, Φ=Bz A with the right hand sides are given by the quantum ex- pressions (7). Although the above detection model is based on time- invariant dynamics, the initial state of the clock spoils the symmetry. The time-invariant state would require a ~Y constant flow of particles or field a constant velocity, so V that the position on the tape imprints time of measure- ment, see [31] for detailed construction. However, such a constant interaction between the detector (clock) and FIG. 5: A trap for a charged particle invariant under rotation the system leads to a backaction and makes the measure- about z axis with a detector weakly coupled to angular mo- ment invasiveness growing with time, which needs to be mentum Lz and the position in the x − y plane detected by reduced by additional resources, e.g. additional coupling appropriate capacitors. The magnetic moment of the particle to a heat bath. can be measured by a sensitive magnetometer, e.g. a SQUID In order to show that the nonconservation can also oc- cur independent from the time asymmetry present either ˆ ˆ ˆ intrinsically or induced by a quantum clock one can look and hX(t2)Lz(t1)Y (t3)i contribute in (7). These terms at other quantities that are conserved, e.g., due to spatial can appear only when t2 < t1 or t3 < t1. We find symmetries. As example, we will use one component of the angular momentum in a rotationally invariant system hlz(t1)x(t2)y(t3)i = (1 − θ(t2 − t1)θ(t3 − t1)) ×(17) in the following. hXˆ(t2)Lˆz(t1)Yˆ (t3) + Yˆ (t3)Lˆz(t1)Xˆ(t2)i/4 = (1 − θ(t2 − t1)θ(t3 − t1))~ sin ω(t2 − t3)/4. V. ANGULAR MOMENTUM CONSERVATION The jump is therefore given by

We propose an experiment to demonstrate the failure h∆lzx(t2)y(t3)i0 = ~ sin [ω(t2 − t3)] /4 (18) of the conservation principle for angular momentum in and is again state-independent as in the case of the har- third-order correlations in weak measurements. Instead monic oscillator. It illustrates that the angular mo- of energy we consider one component of angular momen- mentum conservation is violated by this experiment. tum, say Lˆz = XˆPˆy − Yˆ Pˆx which can be measured in At finite temperature T for t1 < t2,3, the correlator principle by a sensitive magnetometer (e.g. a supercon- hl x(t )y(t )i = sin[ω(t − t )]/4 sinh2( ω/2k T ) in- ducting quantum interferometer device). The other two z 2 3 ~ 2 3 ~ B creases with temperature and makes the (temperature- ˆ ˆ observables will be the particle’s positions X and Y , with independent) jump unobservable. the readouts x and y respectively, which can be measured Since in this setup the detectors are coupled perma- e.g. by the voltage of a capacitor depending linearly on x nently, a frequency-domain measurement might be more and y for small changes in position, see the setup sketch appropriate. In the frequency domain, the observables in Fig. 5. The two positions x and y will be measured at are A(α) = R dteiαtA(t). Taking all our previous argu- times t2 and t3, respectively. ments to frequency domain, the conservation of a quan- The quantity of matter is hlz(t1)x(t2)y(t3)i. Suppose tity Qˆ(α) means that correlators vanish for α 6= 0. In- the particle is in a harmonic trap rotationally invari- terestingly, transforming to frequency domain we find at ant about z axis. The xy part of the trap Hamilto- zero temperature and for γ, α, β 6= 0 that nian reads Hˆ = ω(ˆa† aˆ +a ˆ† aˆ ), with [ˆa , aˆ† ] = 1, ⊥ ~ x x y y√ x,y x,y † † iπ ω(β − α)δ(γ + α + β) [ˆax, aˆy] = [ˆa , aˆy] = 0. Then 2Xˆ =a ˆ +a ˆx and ~ √ x x hlz(γ)x(α)y(β)i = (19) ˆ † 2(α2 − ω2)(β2 − ω2) 2Px/i~ =a ˆx − aˆx (rescaled by a length unit), simi- ˆ † † larly for y, and Lz = i~(ˆaxaˆy − aˆyaˆx). In the ground The conservation principle for angular momentum is vio- state |0i we have Lˆz|0i = 0 so only hYˆ (t3)Lˆz(t1)Xˆ(t2)i lated by (19) because it is non-zero. Hence, either by 6

nore decoherence effects, e.g. due to coupling to a ther- I xI' y z mal environment. The decoherence could be modeled e by Lindblad-type terms added to the Hamiltonian dy- namics of the density matrix. The test of conservation makes sense only for times/frequencies within the co- herence timescale. Any observable roughly tracking the FIG. 6: A charged particle (e.g. electron) goes along the charge in two perpendicular directions will suffice. The tube with angular momentum measured by current in either 0 tube may be not perfectly harmonic or not homogeneous of two coils, I or I while the position in x and y direction is ˆ measured by two perpendicular capacitors. in the z-direction, and Lz can be only approximately conserved or imprecisely measured. To quantify these considerations, we will now develop a Leggett-Garg-type time- or frequency-resolved measurements, one should test [21]. Let us consider the measurement of four ob- 0 0 see experimentally the nonconservation of angular mo- servables: q = q(t1), q = q(t1), x = x(t2) and y = y(t3) mentum. with q being an approximate value of the conserved quan- 0 0 To realize a time-resolved measurement, we suggest to tity and t1 < t2 < t1 < t3. Here q, q can correspond to test the angular momentum conservation with a charge angular momentum lz, while x, y are the lateral positions moving inside a round tube along z direction, similar to in the test presented in the previous section. Note that the recent test of the order of measurements [30]. In the LG-type tests for angular momentum were discussed in simplest model take Hˆ = Hˆz +Hˆ⊥ and we keep the same a different context in [33, 34]. According to the objective harmonic potential in the xy plane as above and add realism assumption, the values of (q, q0, x, y) exist inde- some Hˆz = vpˆz with velocity v (like the clock form the pendent of the measurement. If there is a corresponding previous section). Preparing a wavepacket as a product joint positive probability p(q, q0, x, y), then correlations of the ground state of Hˆ⊥ and ψ(z) of sufficiently short with respect to p must satisfy the following two Cauchy- width, we can measure essentially the same quantity (17) Bunyakovsky-Schwarz inequalities by putting a sequence of weak detectors along the tube, 0 2 2 2 0 2 see Fig. 6. The angular momentum can be measured by h(q − q ) iρhx y iρ ≥ h(q − q )xyiρ, 0 2 2 2 0 2 the current signature in the coil, like in the recent experi- h(q − q ) y iρhx iρ ≥ h(q − q )xyiρ . (23) ment [10]. We simplify the coil-electron beam interaction to λ(z)IˆLˆz where λ is only non-zero inside the coil. Sim- However if we test these inequalities using p defined in (4 ilarly, the measurement of x and y can be modeled by and quantum correlations (7) then they could be violated. local capacitive couplings. In this way, the measurement Classically, the measurements of the conserved quantity times are translated into position according to t = z/v, at two different times should not depend on whether an- just like in the time-invariant energy detection in the pre- other observable is measured in between and both sides vious section. The jump (18) can then be detected by of Eqs. (23) vanish. Using Eqs. (7), the left hand sides placing the coil at two different positions, see Fig. (6). of Eqs. (23) vanish for a perfectly conserved quantity. 0 2 As regards the rotational invariance of the system, de- First, h(q − q ) ip = 0 because Qˆ(t) = Qˆ is independent 0 2 2 tection of X and Y can be performed by the detector- of time. Second, h(q − q ) y ip = 0 because in addition system coupling y is measured after both q and q0. On the other hand, the right hand side of (23) exactly corresponds to the Hˆ = gδ(ˆz)(Xˆpˆ + Yˆ pˆ ) (20) I xD yD quantum mechanical jump in the third-order correlator D D with the initial state ρˆρˆx,y and the detector’s state ρˆx,y = as defined in (18). Hence, even if q is not exactly con- |ψihψ|, served then the left hand sides can be small enough to −1/2 2 2 violate the inequalities. These violations can be readily ψ(xD, yD) = (π/2) exp(−(xD + yD)) (21) tested in the setup suggested in Fig. 6. Note that the Then both the interaction and the initial state of the inequalities must involve fourth moments because of the system and detector are rotationally invariant so the to- so-called weak positivity [32] stating that lower moments tal angular momentum is conserved. Only the readout, are insufficient to violate realism for continuous variables. either x or y of the detector, prefers one direction, i.e.

hxDi ' ghxi (22) VII. CONCLUSION with straightforward generalization to sequential mea- surements like (16). We have shown that conservation laws in quantum mechanics need to be considered with care since their experimental verification might depend on the measure- VI. LEGGETT-GARG INEQUALITIES ment context even in the limit of weak measurements. The conservation is violated if extracting objective real- The above proposals face some practical challenges. ity from the weak measurements. It means that either (i) The velocity v should be sufficiently high in order to ig- weak measurements cannot by considered noninvasive, or 7

(ii) the conservation laws do not hold in quantum objec- with Z = (1 − e−~ω/kT )−1 and tive realism. Exceptions are superconserved observables, √ which will be conserved whatever measurement will be Aˆ = Bˆ = Xˆ(t) = (eiωtaˆ + e−iωtaˆ†)/ 2 (A4) performed, and more generally observables, that satisfy the weak-WAY condition (10). The nonconservation can The independence of the jump of the state follows from also be formulated as Leggett-Garg-type test showing the the fact that connection to the absence of objective realism in quan- tum mechanics. In the future, it might be interesting on [[H,ˆ Xˆ(t )], Xˆ(t )] ∝ 1ˆ (A5) one hand to study more realistic scenarios for quantum 2 3 measurements taking into account decoherence or more † general detectors [35]. Furthermore, one might generalize because [H,ˆ Xˆ] is linear (momentum) in aˆ and aˆ and, these findings to more fundamental relativistic field the- hence, the outer commutator becomes a c-number. ories [36, 37], testing correlations involving components Appendix B: Weak correlations with a quantum of stress-energy-momentum tensor. clock

A single detector and a single clock are defined by (13) Acknowledgements and (14), respectively. The detector position x is mea- sured (projectively) at some time later than the interac- We thank E. Karimi for fruitful discussion and N. tion moment t1. The average Gisin for bringing the quantum clock to our attention. W.B. gratefully acknowledge the support from Deutsche Z hxi ' Tr dt[ˆx, Hˆ (t)]ˆρ/i (B1) Forschungsgemeinschaft (DFG, German Research Foun- I ~ dation) through Project-ID 32152442 - SFB 767 and Project-ID 425217212- SFB 1432. in the interaction picture, in the lowest order, according to the decomposition (12). With

Appendix A: Derivation of correlation jumps HˆI (t) = gAˆ(t)δ(ˆz − vt)ˆpx (B2)

To derive (11) one needs to take the thermal state for the initial initial states (14) we have used the identity ρˆ = Z−1(e−~ω/kT |+ih+| + |−ih−|) (A1) [CˆD,ˆ ρˆ] = {C, [D,ˆ ρˆ]} + {D,ˆ [C,ˆ ρˆ]} (B3) with Z = 1 + e−~ω/kT and for CˆDˆ = DˆCˆ, hψx|{x,ˆ pˆx}|ψxi = 0 (anticommutator), ˆ −iωt iωt 2 X(t) = e |+ih−| + e |−ih+| (A2) [ˆx, pˆ] = i~, hψz|δ(ˆz − vt)|ψzi = |ψz(vt)| to get (15). To extend it to the sequential case (16) we do not apply the plugged into (9) with Qˆ = Hˆ = ~ω|+ih+| and Aˆ = Bˆ = trace over the system space in (B1) (only over x and z), Xˆ with t1 = 0±, t2 = 0 and t3 = τ. getting the matrix The case of harmonic oscillator can√ be written in Fock basis |ni, n = 0, 1, 2, ... with aˆ|ni = n|n−1i, [ˆa, aˆ†] = 1, {Aˆ(t), ρˆ}/2 (B4) nˆ =a ˆ†aˆ = P n|nihn|, nˆ|ni = n|ni and Hˆ = Qˆ = ωnˆ, n√ ~ Xˆ = (ˆa +a ˆ†)/ 2. The thermal state Note that it is Hermitian but not positive definite. Nev- ertheless we can apply the above scheme iteratively, re- ρˆ = Z−1e−~ωn/kTˆ (A3) placing the initial system’s state by (B4) to get (16).

[1] E. Noether, Invariante Variationsprobleme, Nachrichten Mechanical Operators, Phys. Rev. 120,622 (1960). von derGesellschaft der Wissenschaften zu Göttingen, [6] M. M. Yanase, Optimal Measuring Apparatus, Phys. Rev. Mathematisch-Physikalische Klasse 1918, 235 (1918). 123, 666 (1961). [2] M. Peskin, D. Schroeder, An Introduction to Quantum [7] J. B. Hartle, R. Laflamme, D. Marolf, Conservation laws Field Theory, CRC Press (2018). in the quantum mechanics of closed systems, Phys. Rev. [3] S.D. Bartlett, T. Rudolph, R.W. Spekkens, Reference D 51, 7007 (1995). frames, superselection rules, and quantum information, [8] Y. Aharonov, S. Popescu, D. Rohrlich, On Conservation- Rev. Mod. Phys. 79, 555 (2007). Laws in Quantum Mechanics, Proc National Acad Sci [4] E. P. Wigner, Die Messung Quantenmechanischer Oper- 118, e1921529118 (2021). atoren, Z. Phys. 131, 101 (1952). [9] N. Gisin and E.Z. Cruzeiro, Quantum Measurements, En- [5] H. Araki and M. M. Yanase, Measurement of Quantum ergy Conservation and Quantum Clocks, Ann. der Phys. 8

530, 1700388 (2018). [24] A.J. Leggett, Comment on "How the Result of a Mea- [10] H. Larocque, F. Bouchard, V. Grillo, A. Sit, S. Frabboni, surement of a Component of the Spin of a Spin-1/2 Par- R. E. Dunin-Borkowski, M. J. Padgett, R. W. Boyd, and ticle Can Turn Out to be 100", Phys. Rev. Lett. 62, 2325 E. Karimi, Nondestructive Measurement of Orbital An- (1989) gular Momentum for an Electron Beam Phys. Rev. Lett. [25] A. Matzkin, Weak Values and Quantum Properties, 117, 154801 (2016). Found. Phys. 49, 298 (2019) [11] S.M. Lloyd, M. Babiker, G. Thirunavukkarasu, and J. [26] A. Bednorz, K. Franke, W. Belzig, Noninvasiveness and Yuan, Electron vortices: Beams with orbital angular mo- time symmetry of weak measurements, New J. Phys. 15, mentum, Rev. Mod. Phys. 89, 035004 (2017). 023043 (2013). [12] K.Y.Bliokh et al. Theory and applications of free-electron [27] M. Neumark, Spectral functions of a symmetric opera- vortex states, Phys. Rep. 690, 1 (2017). tor,Izv.Akad. Nauk SSSR Ser. Mat.4, 277 (1940). [13] J.V. Koski,V.F. Maisi; T. Sagava; J.P. Pekola, Experi- [28] L. Plimak and S. Stenholm, Causal signal transmission mental Observation of the Role of Mutual Information by quantum fields. V: Generalised Keldysh rotations and in the Nonequilibrium Dynamics of a Maxwell Demon, electromagnetic response of the Dirac sea, Ann. Phys. Phys. Rev. Lett. 113, 030601 (2014). (N.Y.) 327, 2691 (2012). [14] C. Elouard, D. Herrera-Marti, B. Huard, and A. Auf- [29] A. Bednorz and W. Belzig, Quasiprobabilistic Interpre- feves, Extracting Work from Quantum Measurement in tation of Weak Measurements in Mesoscopic Junctions, Maxwell’s Demon Engines Phys. Rev. Lett. 118, 260603 Phys. Rev. Lett. 105, 106803 (2010). (2017). [30] D. Curic, M. C. Richardson, G. S. Thekkadath, J. Florez, [15] M. Naghiloo, J.J. Alonso, A. Romito, E. Lutz, and L. Giner, and J. S. Lundeen, Experimental investigation K.W. Murch, Information Gain and Loss for a Quantum of measurement-induced disturbance and time symmetry Maxwell’s Demon Phys. Rev. Lett. 121, 030604 (2018). in quantum physics, Phys. Rev. A 97, 042128 (2018). [16] A. Pais, Einstein and the quantum theory, Rev. Mod. [31] A. Bednorz and W. Belzig, Models of mesoscopic time- Phys. 51, 863 (1979). resolved current detection, Phys. Rev. B 81, 125112 [17] N.D. Mermin, Is the Moon There When Nobody Looks? (2010). Reality and the Quantum Theory, Phys. Today 4, 38 [32] A. Bednorz, W. Bednorz, W. Belzig, Testing locality and (1985). noncontextuality with the lowest moments, Phys. Rev. A [18] P. Grangier, A. Auffeves, What is quantum in quan- 89, 022125 (2014). tum randomness?, Phil. Trans. R. Soc. A 376, 20170322 [33] S. Mal, D. Das, and D. Home, Quantum Mechanical Vio- (2018). lation of Macrorealism for Large Spin and Its Robustness [19] A. Peres, Quantum Theory – Concept and Methods, against Coarse-Grained Measurements, Phys. Rev. A 94, (Kluwer, Dordrecht, 2002) 062117 (2016). [20] Y. Aharonov, D.Z. Albert, and L. Vaidman, How the [34] S. Kumari and A. K. Pan, Probing Various Formulations result of a measurement of a component of the spin of a of Macrorealism for Unsharp Quantum Measurements, spin-1/2 particle can turn out to be 100, Phys. Rev. Lett. Phys. Rev. A 96, 042107 (2017). 60, 1351 (1988). [35] J. Bülte, A. Bednorz, C. Bruder, and W. Belzig, Nonin- [21] A. J. Leggett and A. Garg, Quantum mechanics ver- vasive Quantum Measurement of Arbitrary Operator Or- sus macroscopic realism: Is the flux there when nobody der by Engineered Non-Markovian Detectors, Phys. Rev. looks?, Phys. Rev. Lett. 54, 857 (1985). Lett. 120, 140407 (2018). [22] C. Emary, N. Lambert, F. Nori, Leggett-Garg inequalities, [36] R. F. Streater and A. S. Wightman, PCT, Spin and Rep. Prog. Phys. 77, 016001 (2014). Statistics, and All that Princeton University Press (2000), [23] A. Palacios-Laloy, F. Mallet, F. Nguyen, P. Bertet, D. ISBN 9780691070629 Vion, D. Esteve, and A.N. Korotkov, Experimental viola- [37] A. Bednorz, Relativistic invariance of the vacuum, Eur. tion of a Bell’s inequality in time with weak measurement, Phys. J. C 73, 2654 (2013). Nat. Phys. 6, 442 (2010).