Wehrl entropy, entropic uncertainty relations and entanglement

Stefan Floerchinger,∗ Tobias Haas,† and Henrik Müller-Groeling‡ Institut für Theoretische Physik, Universität Heidelberg, Philosophenweg 16, 69120 Heidelberg, Germany

The Wehrl entropy is an entropy associated to the Husimi quasi-probability distribution. We discuss how it can be used to formulate entropic uncertainty relations and for a quantification of entanglement in continuous variables. We show that the Wehrl-Lieb inequality is closer to equality than the usual Białynicki-Birula and Mycielski entropic uncertainty relation almost everywhere. Furthermore, we show how a Wehrl mutual information can be used to obtain a measurable perfect witness for pure state bipartite entanglement, which additionally provides a lower bound on the entanglement entropy.

I. INTRODUCTION a more regular use of the Wehrl entropy SW(ρ) when ana- lyzing tabletop experiments from an information theoretic Concepts of quantum information theory play an in- perspective. creasingly important role for describing and understand- In this work, we focus on two major branches of modern ing complex quantum systems. Starting point is usually quantum information theory, where the Wehrl entropy the von Neumann or quantum entropy associated to a reveals its strength. state or density matrix ρ [1–4], First, we discuss the role of Wehrl entropy in the context of entropic uncertainty relations (EURs). The first en- S(ρ) = −Tr {ρ ln ρ} . (1) tropic uncertainty relation was introduced by Białynicki- Birula and Mycielski [13] based on previous works by It quantifies missing information or uncertainty encoded Everett [14], Hirschmann [15] and Beckner [16]. It is an in the state ρ [5, 6]. Despite its importance in quantum uncertainty relation for a single pair of conjugate vari- information theory, the von Neumann entropy S(ρ) has a ables, namely position x and momentum p, which may be major drawback: it can typically not be accessed directly distributed according to the probability density functions in experiments. A computation of the von Neumann en- f(x) = hx|ρ|xi and g(p) = hp|ρ|pi. This corresponds to tropy requires detailed knowledge of the quantum state ρ, the so-called homodyne detection protocol, where both which is oftentimes out of reach. While there are experi- variables are recorded separately by measuring in their mental approaches to reconstruct the full state ρ through corresponding eigenbases. Then, the Białynicki-Birula quantum state tomography [2], the effort for these tech- and Mycielski (BBM) entropic uncertainty relation reads niques scales unfavorably with system size. Alternatively, [17–19] quantum entropies can be measured in (relatively small) quantum systems with finite dimensional Hilbert space h(f) + h(g) ≥ ln eπ, (2) dimH < ∞ by considering multiple copies of the quantum system and using quantum many-body interference [7, 8]. where h(f) is the classical differential entropy associated For infinite dimensional quantum states as they arise to the distribution f(x) [20] for example in continuous variable systems or quantum Z field theory it is substantially more difficult to access or h(f) = − dx f(x) ln f(x). (3) constrain quantum entropies experimentally. In this work, we investigate a somewhat different ap- It should be noted that the bound on the right-hand side proach and consider the Wehrl entropy SW(ρ), which is an of the inequality (2) is state independent whereas the entropy of the Husimi Q-distribution, as a quasi-classical left-hand side increases for mixed states due to concavity. arXiv:2103.07229v2 [quant-ph] 28 Jun 2021 measure for missing information in quantum mechanical Thus, one can expect the bound to be rather loose for phase space [9, 10]. While the density operator ρ and mixed states. Nevertheless, the relation (2) turns out to the Husimi Q-distribution contain both the full informa- be stronger than and implies the well-known formulation tion about the state of the quantum system (see e. g. ref. of an uncertainty relation in terms of variances [21–24], [11]), the latter has the advantage that it can be accessed which is discussed in detail for example in ref. [19]. experimentally (see e. g. [12] for recent developments in Another entropic uncertainty relation for continuous spinor Bose-Einstein condensates). Therefore, we suggest variables was found by Frank and Lieb (FL) [25] based on previous work by Rumin [26]

h(f) + h(g) ≥ ln 2π + S(ρ). (4) ∗ stefan.fl[email protected] Therein the von Neumann entropy S(ρ) accounts for the † [email protected] mixedness of the system’s state ρ. Obviously ln 2π alone ‡ [email protected] is a less tight bound than the one in (2), but for highly 2 mixed states one expects the bound in (4) to be closer to and (4) in section III. In particular, we consider common equality. classes of states, such as number eigenstates or thermal Also for the Wehrl entropy there exists an entropic states. This analysis allows us to show that the Wehrl- uncertainty relation, which is known as the Wehrl-Lieb Lieb inequality is stronger than the relation by Białynicki- inequality (WL) [27–31] Birula and Mycielski almost everywhere in the examined regimes. In section IV we consider a bipartite system of SW(ρ) ≥ 1, (5) N + M modes and construct a Wehrl mutual information from a Wehrl relative entropy, which witnesses every pure where equality only holds for pure coherent state projec- entangled bipartite state. The concept is applied to some tors ρα = |αi hα|. well-known examples like the two-mode squeezed state It was shown that the latter relation does also imply or the class of N00N-states. Additionally, we discuss the the variance based uncertainty relation by Heisenberg and Wehrl conditional entropy. Finally, we summarize our an entropic uncertainty relation of the form (2) based on results and provide an outlook in section V. differential entropies of smeared-out distributions (the Notation. In this paper we work with natural units, marginals of the Husimi Q-distribution, which are not the i.e. ~ = kB = 1 and we drop operator hats. Rather, we true marginals f and g)[30]. employ capital letters for observables X and small letters One main concern of this work will be to investigate for their eigenvalues x and eigenvectors |xi. Also, we use in which regimes the Wehrl-Lieb entropic uncertainty the symbol S for the von Neumann entropy, h for classical relation (5) is tighter (closer to equality) than the relations differential entropies and SW for Wehrl entropies with the in terms of differential entropies (2) and (4); this will be underlying probability distribution f(x) (or state ρ) as addressed in section III. argument h = h(f) (or S = S(ρ)). We denote conditional The second question we study is how the Wehrl entropy entropy by S(A|B) and mutual information by I(A : B). SW(ρ) can be used to make statements about entangle- ment in continuous variable systems. In general, detecting bipartite entanglement for a state ρ which lives in an infi- II. COHERENT STATES AND WEHRL nite dimensional Hilbert space H can be quite challenging ENTROPY [32]. Thus, many criteria are concerned with analyzing the second-order moments [33–36], which is a powerful We start from a set of N continuous quantum vari- approach if the state ρ is of Gaussian form (see e. g. refs. ables Xj and conjugate momenta Pj and the (bosonic) [37–39] for a general discussion of Gaussian states). In- commutation relation corporating also higher-order moments allows to define a sufficient condition for bipartite entanglement, but can [Xj,Pk] = iδjk. (6) be unmanageable in practice [40]. While we use a notation as for positions of particles, the Another approach is to witness entanglement by vio- variables X could also denote other quantities such as lating conditions based on differential entropies [41] or j amplitudes of field modes. Note that we make no detailed differential conditional entropies [42–44], which are closely assumptions about the Lagrangian or Hamiltonian at this related to entropic uncertainty relations (see also ref. [45– point. 47]). They turn out to be successful in detecting many We also introduce the annihilation operators classes of entangled states, especially in the pure state case. 1 aj = √ (Xj + iPj), (7) What these approaches have in common is that they 2 rely on a homodyne detection protocol. In section IV † † we will present an approach based on a Wehrl mutual and creation operators aj = (aj) such that eq. (6) reads information1, which presumes a heterodyne detection † [aj, ak] = δjk. Coherent fields are defined as usual as scheme and turns out to witness every pure entangled state eigenstates of the annihilation operators (see e. g. [11, 49]), and is there furthermore able to bound the entanglement entropy from below. Moreover we discuss the role of the aj |αi = αj |αi , (8) Wehrl conditional entropy, which differs from its fully quantum analog. with eigenvalues αj ∈ C. They can be expressed in terms This paper is organized as follows. We begin intro- of number eigenstates (with ground state |0i) ducing coherent states and the Wehrl entropy of an N N ( ) 1 mode system in section II. Then, we discuss the relations Y † nj |ni = p (aj) |0i, (9) between the three entropic uncertainty relations (5), (2) j=1 nj! through   N ∞ nj  1 2 α  1 Y X − |αj | j An approach solely based on Wehrl entropy is discussed in ref. |αi = e 2 p |ni . (10) [48]. nj! j=1 nj =0  3

For our purposes, it will often be convenient to param- What we have described here is known as the hetero- eterize the complex vector of eigenvalues α in terms of dyne detection protocol [11]. Intuitively, it corresponds position x and momentum p, to jointly measuring position x and momentum p with the highest precision allowed by Heisenberg’s uncertainty 1 α = √ (x + ip) . (11) relation. 2 In contrast to the Wigner W -distribution, the Husimi Q-distribution is non-negative and bounded [56, 57] Coherent states have many interesting properties. They are not orthogonal, 0 ≤ Qρ(α) ≤ 1, (18) 2 −|α−β|2 |hα|βi| = e , (12) which allows to associate a (differential) entropy to it. This is known as the Wehrl entropy [9, 10] and span an (overcomplete) basis, 2N 2N Z d α Z d α S (ρ) = − Qρ(α) ln Qρ(α) 1 = |αi hα| . (13) W πN πN (19) Z dN x dN p = − Qρ(x, p) ln Qρ(x, p). Moreover, they minimize the uncertainty principle, which (2π)N suggests that they can be regarded as a favorable mea- surement basis for simultaneous measurements of x and In the second equation we used the parametrization in- p. Note that there are other states minimizing the un- troduced in eq. (11). certainty principle (squeezed states). Coherent states are The Wehrl entropy can be seen as a quasi-classical the only such states symmetric in the quadratures. In differential entropy in quantum mechanical phase space. It fact, we can construct a positive operator-valued measure is bounded from below by the true von Neumann entropy (POVM), see e. g. ref. [50], based on pure coherent state (as expected for a coarse-grained notion of an entropy), projectors but also bounded from above by the von Neumann entropy of the measured state3 Eα = |αi hα| , (14) S(ρ) ≤ S (ρ) ≤ S(ρ ). (20) which resolve the identity according to (13). The proba- W α bility density for the outcome α is given by the Husimi Furthermore, it is uniquely minimized by any pure coher- Q-distribution [11, 49, 51], ent state projector ρ = |αi hα|, in which case its value Qρ(α) = Tr {ρ E } = hα|ρ|αi , (15) is SW = N. This result is known as the Wehrl-Lieb α inequality [27, 28, 59], such that the probability to find α in some interval is given by Qρ(α)d2N α/πN . SW(ρ) ≥ SW (|αi hα|) = N. (21) Based on the Husimi Q-distribution one can formally define a state after measurement, without recording the Rather recently, improvements of this bound were dis- outcome, cussed in refs. [60, 61]. Another property of the Wehrl entropy SW(ρ) is im- Z 2N portant when it comes to entanglement. In contrast to d α ρ ρα = Q (α) |αi hα| , (16) the von Neumann entropy S(ρ), it is monotonous under πN partial trace, i.e. if we consider a bipartite quantum sys- and it is of Glauber P -form. tem H = HA ⊗ HB with the reduced density operator The Husimi Q-distribution is a quasi-probability distri- ρA = TrB{ρ}, we have bution2 due to the non-orthogonality of coherent states, which causes the Husimi Q-distribution to violate Kol- SW(ρA) ≤ SW(ρ). (22) mogorov’s third axiom, namely σ-additivity. In other words, two values Qρ(α) and Qρ(α0) for α 6= α0 do not This highlights the classical nature of Wehrl entropy. The represent probabilities for mutually exclusive events. Nev- analogous statement is oftentimes not true for the von ertheless, Qρ(α) is normalized to unity in the sense Neumann entropy S(ρ) due to entanglement [9, 10]. This already indicates that capturing entanglement with a Z d2N α Wehrl entropy may require a different line of reasoning Qρ(α) = Tr{ρ} = 1. (17) πN compared to a von Neumann entropy.

2 For general literature on quasi-probability distributions see e.g. 3 The second statement can be proven using the well-known Berezin- refs. [49, 52–56]. Lieb inequality [58], see e.g. ref. [10]. 4

III. MONOPARTITE QUANTUM SYSTEM Furthermore, the probability density functions f(x) and g(p) are of equal shape, At first we consider a monopartite quantum system. 1 2 In the following, we want to compare the differential en- 2 √ 2 −x fn(x) = |ψn(x)| = n Hn(x) e , (26) tropies of the two phase space distributions Qρ(x, p)/(2π) π 2 n! and f(x)g(p) associated with the heterodyne and homo- where H (x) are the commonly-known physicist’s Hermite dyne measurement protocol respectively. Let us first state n polynomials. The Husimi Q-distribution is given by the following relations 2 Qρ  Z Qρ(x, p) Qρ(x, p) Qn(x, p) = |hn|αi| h = − dx dp ln (27) 1 2 2
n − 1 (x2+p2) 2π 2π 2π = x + p e 2 . 2nn! = SW(ρ) + ln(2π), Z (23) From these expressions all appearing entropies can be h(fg) = dx dpf(x)g(p) ln(f(x)g(p)) calculated in the following. The differential entropies h(f) and h(g) appearing in = h(f) + h(g). the Białynicki-Birula and Mycielski as well as the Frank- The two differential entropies are related up to additive Lieb uncertainty relation were investigated in refs. [62–65]. The main challenge in this computation is due to the constants with SW(ρ) and h(f) + h(g), for which we know entropic uncertainty relations. Hermite polynomials Hn(x). In principle the differential Henceforth, our goal is to compare the uncertainty entropy is given by deficit, i.e. the distance to the bound of an entropic un- √ n
1 1 certainty relation, of these relations as an estimator for h(fn) = ln π 2 n! + n + + √ E(Hn), (28) 2 π 2nn! the strength of the statement. We refer to a lower uncer- tainty deficit as the tightness of an uncertainty relation. where E(Hn) is the so-called entropy of Hermite polyno- It is clear that being close to equality is desirable for in- mial equalities, however it should be noted here that this does Z +∞ not immediately lead to a formal implication between 2 2
−x2 E(Hn) = − dx Hn(x) ln Hn(x) e . (29) different uncertainty relations. In fact, the Wehrl-Lieb −∞ inequality is tight on a different set of states than the inequality by Białynicki-Birula and Mycielski. After a tedious and lengthy calculation one finds [64, 65] To properly compare them, we rearrange the three in- √ equalities such that their right-hand sides coincide. Thus, n
1 h(fn) = ln π 2 n! + n + + nγ we have the following three relations 2 n    X 2 3 2 WL S (ρ) + ln π − 2 xn,i 2F2 1, 1; , 2; −xn,i W  2 BBM h(f) + h(g) ≥ ln eπ. (24) i=1 n n  X n(−1)k2k X  1  FL h(f) + h(g) − S(ρ) + ln e/2 + F k, , −x2 . k k 1 1 2 n,i k=1 i=1 We begin our analysis with pure number eigenstates, then (30) we investigate a mixture of the lowest two number eigen- states and finally consider thermal states. This analysis Therein 2F2(a, b; c, d; z) denotes a generalized hypergeo- will allow us to state a conjecture about their relations metric function, 1F1(a, b; z) is Kummer’s confluent hy- for a general state ρ. If possible, results will be given pergeometric function for z ∈ C and xn,i are the roots analytically. Furthermore, the Husimi Q-distribution will 2 of the n-th Hermite polynomial Hn(x ). Also, we have be written down in terms of position x and momentum p, introduced the Euler-Mascheroni constant γ ≈ 0.577. i.e. using the parametrization (11), in order to make the The latter expression can be evaluated numerically for different (quasi)-probability distributions comparable. all n ∈ N. Furthermore, for n  1 one can give a simple approximate formula for the differential entropy [64] A. Number eigenstates 1 h(f ) ≈ −2 + ln 2π2n
. (31) n 2 The density matrix of a pure number eigenstate In contrast, the Wehrl entropy can be computed analyti- cally due to the simple form of the Husimi Q-distribution ρn = |ni hn| , (25) for number eigenstates (27). Although this is a rather has vanishing quantum entropy, S(ρn) = 0. As a conse- technical issue, one might appreciate the fact that com- quence, the Frank-Lieb bound will be less tight than the puting a Wehrl entropy is much easier than computing a Białynicki-Birula and Mycielski bound. differential entropy in many cases. 5

EURs EURs 7 4 WL WL 6 3.5 BBM BBM ��� 5 3 FL WL n≫1 4 BBM n≫1 2.5 Bound 3 Bound 2 n n 1 2 3 4 5 6 7 8 9 10 100 200 300 400

Figure 1. Left panel: Values of the three entropic uncertainty relations (EURs, cf. (24)) as a function of the energy level n. One can see that the Białynicki-Birula and Mycielski (BBM) as well as the Wehrl-Lieb (WL) inequality are equivalent and tight for the ground state n = 0. For n = 1 we have SW(ρ1) + ln π ≈ 2.72 > h(f1) + h(g1) ≈ 2.69, whereas for n = 2 we get SW(ρ2) + ln π ≈ 2.992 < h(f2) + h(g2) ≈ 2.997. Thus, the Wehrl-Lieb uncertainty relation is the strongest for n ∈ N \{1}, i.e. almost everywhere. For pure states the bound of the Frank-Lieb (FL) relation becomes state-independent. In that case it is less tight than the relation by Białynicki-Birula and Mycielski. Right panel: Behavior of the Wehrl-Lieb and the Białynicki-Birula and Mycielski uncertainty relations for large n (the latter was investigated numerically in [63]). The result for the Wehrl entropy is already valid for small n, whereas the result for the differential entropies is only valid for n  1. This is again due to the fact that the Wehrl entropy is of such simple form.

Starting from (27), one can compute the Wehrl entropy Białynicki-Birula and Mycielski relation. Also, we observe for number eigenstates. After a straight forward exercise that the Stirling approximation (35) is close to the true in Gaussian integration one finds (see also ref. [66]) result (32) also for small n ∼ 1, while (31) becomes a reasonable approximation only for n  1. SW(ρn) = ln n! + n + 1 + nγ − nηn, (32) where ηn is the n-th harmonic number B. Mixture of number eigenstates n X 1 η = , (33) n k k=1 Next, we analyze some particular mixtures of number eigenstates. For simplicity we concentrate on a mixed with η0 = 0. state of the first two number eigenstates, as well as a The result can be further simplified when considering thermal state. We will again compare different entropic large n  1 and using the Stirling approximation formula uncertainty relations and investigate how close they are √ nn to being violated. n! ≈ 2πn , (34) e which leads to 1. Mixture of first two number eigenstates 1 SW(ρn) ≈ (1 + ln 2πn) , (35) 2 A relatively simple mixed state is the following super- for n  1. position of the first two number eigenstates, For the sake of a visual comparison we present all three bounds as a function of n in the left panel of Figure 1. The ρ01 = q |0i h0| + (1 − q) |1i h1| . (36) main result is that the Wehrl-Lieb relation is tightest for n > 1. Moreover, we obtain the expected result for n = 0, The resulting entropies involve now a sum inside a loga- i.e. the Wehrl-Lieb as well as the Białynicki-Birula and rithm. We use numerical methods to solve the integrals. Mycielski relations are tight and equivalent. In general, The resulting curves for the entropic uncertainty rela- the Frank-Lieb relation is worst since we consider pure tions are shown in Figure 2 (left panel). Therein, one states. can see that the Wehrl-Lieb relation is tighter than the Furthermore, it is interesting to compare the Wehrl- Białynicki-Birula and Mycielski relation except in a small Lieb relation with the Białynicki-Birula and Mycielski region around q = 0, where (36) corresponds to the first relation for large n  1 (cf. right panel of Figure 1). excited state. Thus we observe a similar situation as for From the asymptotic expressions (31) and (35) we can pure states. Also, the Wehrl-Lieb relation turns out to be read off the scaling behaviors for large n  1, which less concave than that of Białynicki-Birula and Mycielski. gives h(fn) + h(gn) ∝ ln n and SW(ρn) ∝ 1/2 ln n. There- In contrast, the Frank-Lieb relation becomes tightest for fore, the Wehrl-Lieb relation grows much slower than the more mixed states. 6

EURs EURs 3 2.8 2.8 WL WL 2.6 2.6 BBM BBM ��� ������ 2.4 FL 2.4 FL Bound 2.2 2.2 Bound q βω 0.2 0.4 0.6 0.8 1 2 4 6 8 10

Figure 2. Left panel: Entropic uncertainty relations (EURs, cf. (24)) for a mixture of the ground state with probability q and the first excited state with probability 1 − q. The Wehrl entropy decreases monotonically for larger q, i.e. if the ground state contributes more to the total state. Right panel: Entropic uncertainty relations for the thermal state. The Wehrl-Lieb relations is again tighter than the Białynicki-Birula and Mycielski relation and both become tight for zero temperature β → ∞. In contrast the Frank-Lieb (FL) relation becomes tight in the infinite temperature limit β → 0, but is less tight for lower temperatures.

2. Thermal state We see that for β → ∞, i. e. in the zero temperature limit, the inequality becomes tight, as expected based on the discussion in sectionIIIA. At last, we consider a thermal state ρT under the as- sumption that the Hamiltonian is given by H = ω(n+1/2). Since thermal states are highly mixed we should have The thermal state is then the state with maximal entropy a closer look at the Frank-Lieb relation. Additionally to for a fixed energy expectation value [67–70]. It reads the computation of the differential entropies, we need to calculate the von Neumann entropy of the thermal state ∞ 1 X ρ . It is given by ρ = e−βH = p |ni hn| , (37) T T Z n n=0 −βω βω S(ρT ) = − ln(1 − e ) + . (44) where pn is the Boltzmann distribution eβω − 1

1 −βω(n+ 1 ) pn = e 2 , (38) Thus, the left-hand side of the Frank-Lieb relation reads Z (again we rearrange everything such that the right-hand and side equals ln eπ) ∞ −βω/2 X −βω(n+ 1 ) e 1 Z = e 2 = = , (39) h(f ) + h(g ) − S(ρ ) + 1 − ln 2 1 − e−βω 2 sinh(βω/2) T T T n=0 π 1 − e−βω  βω (45) = 2 + ln − . is the canonical partition function. To calculate the prob- 2 tanh(βω/2) eβω − 1 ability density fT (x) we make use of Mehler’s formula [71], which leads to Also, we need to calculate the Husimi Q-distribution, ∞ 2 which yields 1 X 2 −βω(n+ 1 ) 1 − x fT (x) = |ψn(x)| e 2 = √ e 2σ2 , Z 2 ∞ n=0 2πσ 1 X Q (x, p) = Q (x, p)e−βEn (40) T Z n where we used the abbreviation n=0 ∞ 1 βω  1 − 1 (x2+p2)− βω X 1 2 2 n −βωn = tanh . (41) = e 2 2 (x + p ) e 2 Z 2nn! 2σ 2 n=0 Thus, the distribution f (x) is a Gaussian centered 1 − 1 (x2+p2) 1−e−βω − βω T = e 2 ( ) 2 . around x = 0 with variance σ2, which depends non- Z linearly on the inverse temperature β. The associated (46) differential entropy is For the left-hand side of the Wehrl-Lieb uncertainty rela- √  h(f) = ln 2πeσ2 . (42) tion we then find A similar consideration for the momentum distribution βω π  SW(ρT ) + ln π = 1 + + ln csch(βω/2) . (47) leads to the following expression for the uncertainty rela- 2 2 tion of Białynicki-Birula and Mycielski We show a comparison of the three entropic uncertainty  βω  relations in Figure 2 (right panel). The Wehrl-Lieb re- h(f ) + h(g ) = 1 + ln π − ln tanh . (43) T T 2 lation is again stronger than the Białynicki-Birula and 7

Mycielski relation. This is due to the fact that for a A. Wehrl relative entropy thermal state the ground state is populated most and higher energy eigenstates are suppressed exponentially. In the following we will consider different quantities Moreover, both become tight for the temperature tending used in classical and quantum information theory for to zero, which corresponds to the system approaching Husimi Q-distributions defined in analogy to the Wehrl the ground state. Interestingly, the relation by Frank entropy SW(ρ). We begin with defining the Wehrl relative and Lieb behaves differently, which was already noted by entropy, which will also be helpful to define a Wehrl the authors themselves [18, 25]. The uncertainty relation conditional entropy and Wehrl mutual information. For becomes tight in the infinite temperature limit in contrast any state ρ and some model state σ on an N-mode system to the other two inequalities and it is more loose for low it reads (in complete analogy to the standard definition temperatures. by Kullback and Leibler [72, 73])

Z d2N α 3. General comment S (ρkσ) = Qρ(α)(ln Qρ(α) − ln Qσ(α)). (50) W πN

What we have observed is that the Frank-Lieb relation The usual support condition for the states supp(ρ) ⊆ (4) is tightest for highly mixed states, while the Wehrl-Lieb supp(σ) then translates into the support condition for relation (21) is tightest in the number eigenstate regime the Husimi Q-distributions supp(Qρ) ⊆ supp(Qσ). If (except for n = 1). Furthermore, we have seen that the this condition is violated we set SW(ρkσ) = +∞. In- Wehrl-Lieb relation is less concave under mixing than the tuitively, (50) can be interpreted as a measure for the Białynicki-Birula and Mycielski relation (cf. Figure 2). distinguishability of the true Husimi Q-distribution Qρ Thus, we can conjecture that the Wehrl-Lieb entropic and some model distribution Qσ. Furthermore, it is a uncertainty relation is tighter than the Białynicki-Birula non-negative quantity being zero if and only if the two and Mycielski entropic uncertainty relation in most situa- distributions agree4. However, it should be emphasized tions. Furthermore, we have seen that the Wehrl entropy that the most important property of the quantum rel- is very often easy to compute, especially compared to ative entropy, namely its monotonicity under quantum differential entropies of marginal distributions. channels, does not hold for the Wehrl relative entropy. This is due to the fact the Wehrl entropy is not invariant under general unitary transformations. In fact, the Wehrl IV. BIPARTITE QUANTUM SYSTEM entropy is only invariant for transformations which map coherent states to coherent states [10]. In the second part of this work we consider a bipartite quantum system AB with partitions A and B of N and M modes, respectively. Moreover, we associate the sets of B. Wehrl conditional entropy coherent states |αi to A, |βi to B and |αβi = |αi ⊗ |βi ρ to AB. Then, the local Husimi Q-distribution QA(α) can Let us recall first some properties of quantum con- be obtained from the global one Qρ(α, β) by integrating ditional entropies. In the finite dimensional case there out system B exist several ways of defining this quantity, which are all Z 2M equivalent. For an infinite dimensional Hilbert space this ρ d β ρ is not necessarily the case. An adequate definition for QA(α) = Q (α, β). (48) πM both cases comprises the quantum relative entropy and Note that this agrees with the Husimi Q-distribution requires S(ρA) < ∞ (see ref. [74]). Then, we can define the quantum conditional entropy as obtained from the reduced density matrix ρA = TrB{ρ}. Similarly, one can define a local Husimi Q-distribution ρ ρ S (A|B) = S(ρA) − S(ρkρA ⊗ ρB), (51) QB(β) for subsystem B. It is also useful to consider a situation where a hetero- which reduces to dyne measurement is performed first only on subsystem B. Assuming that the outcome is β leads to a Husimi Q- ρ S (A|B) = S(ρ) − S(ρB), (52) distribution for subsystem A that we call the conditional Husimi Q-distribution, provided that the von Neumann entropy of subsystem B  hβ|ρ|βi  hαβ|ρ|αβi is also finite S(ρB) < ∞. Qρ(α|β) =Tr E = The quantum conditional entropy Sρ(A|B) allows for a hβ|ρ |βi α hβ|ρ |βi B B (49) simple separability criterion. In particular, one can easily Qρ(α, β) = ρ . QB(β) Note that these relations are in complete analogy to clas- sical probability densities. 4 See appendixA for a simple proof. 8 prove that all separable states5 C. Wehrl mutual information

X i X i i
ρ = pi ρ = pi ρA ⊗ ρB , (53) To quantify the correlations between the two subsys- i i tems A and B we introduce the Wehrl mutual information P Iρ (A : B) = Sρ (ρkρ ⊗ ρ ). (60) where pi ≥ 0 and i pi = 1 is a probability distribution W W A B i i and ρA and ρB can be taken to be pure, fulfill Most importantly, its non-negativity is inherited from the Wehrl relative entropy (50). Also, it is zero if and only Sρ(A|B) ≥ 0. (54) if the state ρ is a product state. We can express (60) in terms of other Wehrl entropies provided that they are Thus, the violation of the latter inequality is a sufficient finite condition for entanglement [75, 76]. One may easily con- ρ struct examples where this bound is violated, e. g. a pure IW(A : B) = SW(ρA) + SW(ρB) − SW(ρ) ρ (61) entangled state. = SW(ρA) − SW(A|B). In complete analogy to (51) we define the Wehrl condi- tional entropy for S (ρ ) < ∞ as In general, a mutual information can not distinguish W A between classical and quantum correlations. Thus, let Sρ (A|B) = S (ρ ) − S (ρkρ ⊗ ρ ). (55) us now consider the case where some state ρ is known W W A W A B to be pure. In this case the Wehrl mutual information ρ Analogously, we get IW(A : B) quantifies only the quantum correlations be- tween A and B. Moreover, it is zero if and only if the ρ state is a product state ρ = ρ ⊗ ρ , which follows di- S (A|B) = SW(ρ) − SW(ρB) (56) A B W rectly from the according property of the Wehrl relative entropy. Therefore, we can conclude that the Wehrl mu- if SW(ρB) < ∞. ρ Note that one may also write this with the conditional tual information IW(A : B) is a perfect witness for pure Husimi Q-distribution in (49) as state entanglement. It should be noted that the positive partial transpose Z 2M (PPT) criterion already provides a necessary and sufficient ρ d β ρ S (A|B) = − Q (β) condition for pure state entanglement (see e.g. [32]), but W πM B (57) requires the knowledge of the state ρ, which is typically Z 2N d α ρ ρ inaccessible in experiments. Therefore, the advantage × N Q (α|β) ln Q (α|β), ρ π of the Wehrl mutual information IW(A : B) lies in its measurability and in this sense it can be regarded as a similar as for a classical conditional entropy. Because measurable equivalent to the PPT criterion for pure states. the conditional Husimi Q-distribution is itself a Husimi Also, one should note that not every classical mutual in- Q-distribution, the Wehrl-Lieb inequality implies formation is a perfect witness for pure state entanglement. ρ For example, any marginal mutual information does not SW(A|B) ≥ N, (58) contain information about correlations in or with other variables. which is a refined version of the monotonocity property ρ Additionally, it is interesting to check whether IW(A : (22) since it can be reformulated as B) fulfills the requirements for an entanglement measure or an entanglement monotone. Since the Wehrl mutual SW(ρB) + N ≤ SW(ρ). (59) information does not reduce to the entanglement entropy, it is not an entanglement measure. Also, the Wehrl mu- Also, the relation (58) can be interpreted as an entropic tual information does not fulfill a minimum requirement, uncertainty relation in the presence of classical memory namely invariance under local unitaries (see [77] for an entropic uncertainty relation with quantum memory). In contrast to the relation discussed in ref. ρ0 ρ IW(A : B) 6= IW(A : B), (62) [77], where the analysis was restricted to quasi-classical 0 † † states and corresponding entropies, (58) shows that a with ρ = (UA ⊗ UB) ρ (UA ⊗ UB) and UA,B unitary. A fully classical Wehrl conditional entropy can not serve as simple counterexample is given by UA = 1 and UB = an entanglement witness analogous to eq. (54). In this S(κ), where S(κ) is the squeezing operator on subsystem sense, it behaves like a classical conditional entropy. B. Therefore, the Wehrl mutual information should be considered as a witness for pure state entanglement and not as an entanglement monotone. Nevertheless, there is a non-trivial relation between the Wehrl mutual information and the quantum mutual 5 Note that for an infinite dimensional Hilbert space a separable information. Following [29], we have the relation state may also be written as a convex integral over pure product states. This is particularly interesting for mixed Gaussian states. ρ ρ IW(A : B) ≤ I (A : B). (63) 9

For globally pure states the latter relation reduces to and the symmetric matrix C has the form   1 ρ CA CM IW(A : B) ≤ S(ρA) = S(ρB), (64) C = T . (68) 2 CM CB which is an interesting and useful relation when it comes to Note that C is related to the covariance V in symplectic quantifying entanglement in an actual experiment. If the phase space, Husimi Q-distribution of a pure state ρ is measured, one 1 can not only decide if the state ρ is entangled or not, one Vij = Tr {ρ (rirj + rjri)} , (69) can also give a lower bound on the entanglement entropy 2 S(ρA). Thus, the measurable Wehrl mutual information by Iρ (A : B) allows to make a statement about how much W  1 −1 two subsystems are entangled at least. If the global state C = V + 1 . (70) ρ is not pure, one can still make a statement about how 2 much the two subsystems are correlated at least. This follows from convolving the corresponding Gaussian It should be noted that the left-hand side of the relation Wigner W -distribution with a Gaussian distribution of (63) is not restricted to the Wehrl mutual information, but covariance (1/2)1. holds true for all classical mutual informations constructed Also, Z is a normalization constant, such that (17) is from positive operator-valued measures in subsystems A fulfilled. Then, the expression (66) can be rewritten as and B (cf. corollary 2 in ref. [29]). ρ Also, we want to point out that the relation (63) is Q (rA, rB) similar to an open conjecture stated in ref. [78], which is 1  1 1  = exp − rT C r − rT C r − rT C r , based on a homodyne detection protocol. It reads Z 2 A A A 2 B B B A M B ? (71) ρ ρ ρ I (fA : fB) + I (gA : gB) ≤ I (A : B), (65) where the correlations are encoded in the mixing term where f = f(x ), f = f(x ) are probability densities T A A B B rACM rB. If this term would be absent, the Husimi Q- of the positions and gA = g(pA), gB = g(pB) are the mo- distribution for the joint system AB would factorize into mentum distributions measured in A or B, respectively. the individual Husimi Q-distributions. Note that the measured mutual informations in the latter Performing a Gaussian integration, we arrive at the relation do not contain any information about correla- joint Wehrl entropy tions between positions and momenta, which are naturally 1 included in the Wehrl mutual information Iρ (A : B). S (ρ) = − ln det C + N + M. (72) W W 2 Note that the Wehrl-Lieb inequality for N + M modes D. Entanglement of common pure states then implies that  1  In the following, we compute the Wehrl mutual infor- det C ≤ 1 ⇔ det V + 1 ≥ 1, (73) ρ 2 mation IW(A : B) and the Wehrl conditional entropy ρ which is an uncertainty relation in terms of the covariance SW(A|B) for some well-known pure states. To that end we use the parametrization (11) for each subsystem. matrix V for the special case of Gaussian states. Unlike the Robertson-Schrödinger uncertainty relation for Gaussian states det V ≥ 1/2N+M (see for example 1. Gaussian states [19, 39]), (73) is not invariant under similarity transfor- mations of V like symplectic transformations (e.g. squeez- We begin with a general Gaussian state ρ. Such a state ing). This shows a difference between the Robertson- is characterized by a global Husimi Q-distribution of the Schrödinger uncertainty relation and the Wehrl-Lieb in- form6 equality. The former introduces a minimum for the uncer- tainty product while the latter is only tight for coherent 1  1  Qρ(r) = exp − rT · C · r , (66) states, i.e. equal uncertainty in position and momentum. Z 2 The next step is to compute the reduced Husimi Q- where the bilinear form in the exponent contains all four distribution for one subsystem. Without loss of generality phase space vectors xA, pA, xB, pB combined into a single we consider subsystem B and end up with vector ρ QB(rB) r = (r , r )T = (x , p , x , p )T , (67) 1 Z A B A A B B = d2N r Qρ(r) (2π)N A r   det C 1 T T −1
6 = exp − rB CB − CM CA CM rB . Without loss of generality we assume that the expectation values det CA 2 of the quadratures vanish. (74) 10

2 1.5 ρ ρ 1.5 IW (A:B) IW (A:B) 1 ρ ρ 1 SW (A|B) SW (A|B) Vacuum EPR Iρ(A:B) 0.5 Iρ(A:B) 0.5

λ N 0.2 0.4 0.6 0.8 1 2 4 6 8 10

ρ ρ Figure 3. Left panel: Wehrl mutual information IW(A : B) and Wehrl conditional entropy SW(A|B) together with the quantum mutual information Iρ(A : B) of the two-mode squeezed state as a function of the squeezing parameter λ. For λ = 0 one recovers the product state of two vacua, while for λ → 1 the state approaches the fully correlated Einstein-Podolsky-Rosen (EPR) state. ρ ρ Right panel: Wehrl mutual information IW(A : B), Wehrl conditional entropy SW(A|B) and quantum mutual information Iρ(A : B) for the N00N states up to N = 10 (the curves represent interpolations). Higher N00N states have larger Wehrl mutual information, although the value for N = 1 is slightly smaller than for N = 2.

Calculating the local Wehrl entropy gives 2. Two-mode squeezed state 1 1 SW(ρB) = − ln det C + ln det CA + M, (75) As a more concrete example, we consider the two-mode 2 2 squeezed state for N = M = 1, which is defined by [37] where we used the identity ∞ p 2 X n T −1
|ψi = 1 − λ (−λ) |nA, nBi , (80) det C = det CA det CB − CM CA CM , (76) n=0 which holds provided that det C 6= 0. Here, the Wehrl- A where the squeezing parameter λ ∈ [0, 1] is related to the Lieb inequality (21) requires squeezing parameter r ∈ [0, ∞) by det C ≤ det C . (77) A λ = tanh r. (81) Assuming that all Wehrl entropies are finite allows to Since the two-mode squeezed state is a pure Gaussian calculate the two quantities of our interest. Both appear state, the corresponding Husimi Q-distribution is of Gaus- to have a simple form sian form (66). Determining the matrix C yields

ρ 1 SW(A|B) = N − ln det CA, CA = CB = 12 and CM = diag(λ, −λ). (82) 2 (78) 1 det C det C ρ A B As expected, the correlation matrix C becomes zero for IW(A : B) = ln . M 2 det C λ = 0, in which case the state (80) is separable, such that It can be seen easily that in the pure state case that the global Husimi Q-distribution factorizes. For the Wehrl ρ conditional entropy and the Wehrl mutual information IW(A : B) = 0 is a necessary and sufficient separabil- ity condition for Gaussian states as it corresponds to (cf. Eq. (78)) we find the Wehrl relative entropy between the full Husimi Q- Sρ (A|B) = 1, distribution and its factorization. The latter is equivalent W ρ 2
(83) to CM = 0, which itself is equivalent to VM = 0. This IW(A : B) = − ln 1 − λ . already implies separability via the positive partial trans- pose (PPT) criterion for pure Gaussian states [38, 79]. Both quantities allow for a straight-forward classification For the special case N = M = 1, which is the setup of in an actual experiment: since the Wehrl conditional en- the Simon criterion [33], we show that the Simon crite- tropy is independent of quantum correlations encoded rion indeed implies CM = 0 for pure separable Gaussian in the squeezing parameter λ, any measured deviation states in appendixC. At least in this special case, the from 1 is sufficient to show that the state exhibits addi- criterion of ref. [33] and our condition for separability, tional classical correlations. For example, the squeezing ρ parameter λ may fluctuate in subsequent realizations of IW(A : B) = 0, are therefore equivalent. Interestingly, we find also that the Wehrl conditional an experiment, which increases the Wehrl conditional en- ρ tropy due to its concavity. Note that this statement is a entropy SW(A|B) is independent of the correlations en- coded in CM . Furthermore, applying (58) to (78) leads consequence of the Wehrl conditional entropy (78) being to local uncertainty relations of the form independent of the correlations encoded in CM . Also, it is interesting to note that the lower bound (58) is attained det CA ≤ 1 and det CB ≤ 1. (79) for the two-mode squeezed state (see Figure 3, left panel, 11 red dotted curve), to indicate the amount of quantum V. CONCLUSION AND OUTLOOK correlations captured by the Wehrl mutual information. Furthermore, the Wehrl mutual information allows to In summary, we have investigated here the relation be- set a lower bound on the amount of bipartite entanglement tween the Wehrl-Lieb entropic uncertainty relation for the between A and B. The latter is monotonically increasing Husimi Q-distribution and other relations for marginal for an increasing squeezing parameter (see Figure 3, left distributions. Moreover, we have discussed to which ex- panel, blue straight curve). In particular, it vanishes if and tent the Wehrl entropy and related quantities are able to only if the state is separable (λ → 0) and tends to infinity witness entanglement of continuous variables. if the state approaches the Einstein-Podolsky-Rosen state We found that the Wehrl-Lieb inequality is tighter (λ → 1) representing perfect correlations xA = xB and than the Białynicki-Birula and Mycielski relation almost pA = −pB. Furthermore, we see how the bound (63) everywhere. In particular, we have shown its superior behaves as we have also plotted the quantum mutual tightness for number eigenstates ρn = |ni hn|, where n = 1 information (see Figure 3, left panel, green dotted curve), turned out to be an exception. Also, the calculation which can be calculated easily following refs. [37, 80]. showed that Wehrl entropies are technically simpler to handle compared to differential entropies of marginal distributions. In the second half, we have introduced analogs of well- 3. N00N states known quantum information theoretic quantities in terms of Wehrl entropies. It turned out that especially the Another interesting class of pure entangled states are Wehrl mutual information is of great use when it comes the N00N states, which are of the form (the normalization to entanglement witnessing. Although it is neither an constant is chosen such that the following expression is entanglement measure, nor an entanglement monotone, normalized to unity also for N = 0) it is a perfect witness for pure state entanglement, i.e. it detects all pure entangled states. Furthermore, it gives a 1 lower bound on the entanglement entropy, which allows |ψN i = p (|N, 0i + |0,Ni) , (84) to quantify to which extent some pure state ρ = |ψi hψ| 2(1 + δ0N ) is entangled at least. for 1 + 1 modes (note that here N denotes the Nth To exemplify the potential of the Wehrl mutual infor- excitation instead of the Nth mode). These states are mation we discussed several examples. For the Gaussian separable if and only if N = 0, in which case we end state case, after a straight forward exercise in Gaussian up with two vacua again. Furthermore, for N = 1 one integration, we found a simple formula allowing for a recovers one of the four Bell states as a special case. direct classification. Here it would be of further interest to relate the Wehrl mutual information as perfect witness The N00N states are known to be hard to detect if con- for pure Gaussian states to existing criteria like for exam- sidered in the context of a continuous variables quantum ple the Simon criterion [33]. Additionally, we have seen system [37]. For example, the strong entropic separability that the Wehrl-Lieb inequality for N + M mode Gaussian criterion in ref. [41] only detects N00N states up to N = 5. states implies a novel uncertainty relation in terms of the Since the Wehrl mutual information is a perfect witness, covariance matrix, which may be studied further in the all N00N states are witnessed. future. Another interesting example we considered was Calculating the global and one local Husimi Q- the class of N00N states, which are particularly hard distribution gives to detect, especially for large N. Also in this case the Wehrl mutual information was a reasonable entanglement 1 2 2 ρ −(rA+rB )/2 witness, in fact it is a witness for all N ∈ N. Q (r) = N+1 e 2 N! (1 + δ0N ) Besides the Wehrl mutual information we also consid- × (x − ip )N + (x − ip )N
ered the Wehrl conditional entropy, which turned out A A B B (85) N N
to be bounded from below similar to the Wehrl entropy × (xA + ipA) + (xB + ipB) , itself. This statement can be interpreted as an entropic ρ 1 −r2 /2 2N N uncertainty relation in the presence of classical memory Q (r ) = e B r + 2 N!
, B B 2N+1 N! B as well as a refined version of the monotonicity property of the Wehrl entropy under partial trace. Furthermore, 2 2 2 2 where we used the abbreviations ri = |ri| = xi + pi for the special case of Gaussian states, we found that the for i ∈ {A, B}. The Wehrl conditional entropy and the Wehrl conditional entropy is independent of the correla- Wehrl mutual information can be computed numerically tions encoded in the off-diagonal blocks of the covariance (see Figure 3, right panel, same color scheme as before). matrix. We observe that the Wehrl mutual information increases Overall, we want to emphasize again that the signifi- monotonically for N ≥ 2. It can be seen that it approaches cance of the Wehrl entropy and all related quantities is the quantum mutual information, which is simply Iρ(A : that they are measurable when making use of the hetero- B) = 2 ln 2 for all N ∈ N. dyne detection scheme. Thus, the Wehrl mutual informa- 12 tion and the Wehrl conditional entropy can be applied Moreover, equality holds iff the two Husimi Q- immediately to actual experiments, where they serve as distributions agree since − ln x is strictly convex for x > 0. a perfect witness for pure state entanglement and as an indicator for non-Gaussianity, respectively. One question left over for future work concerns a more Appendix B: Concavity of Wehrl conditional entropy general notion of coherent states, namely SU(N) coher- ent states. It should be checked whether the presented Writing out the Wehrl conditional entropy approach to witness pure state entanglement can be gen- ρ eralized to SU(N) coherent states. This question is of SW(A|B) particular interest for e.g. experiments with spinor Bose- Z d2N α d2M β = − Qρ(α, β) (ln Qρ(α, β) − ln Qρ (β)) Einstein condensates [46, 47, 81]. πN+M B It is noteworthy, that the Wehrl-Lieb inequality holds (B1) also for generalized coherent states (see refs. [30, 82]), which may be another direction for future investigations. reveals that the integrand is given by the simple function Other directions for subsequent research are the gener- f(x, y) = −x ln x + x ln y for real positive x and y. This alization of the presented quantities to quantify entangle- function is jointly concave in x and y, which can easily ment also in the multipartite case and to other interesting be proven using Lieb’s concavity theorem [2]. Then, the states, e. g. Schrödinger cat states (see [83] for results on monotonicity of the integral implies concavity for the the Wehrl entropy in this case). Wehrl conditional entropy. Furthermore, it would be of great interest to generalize the concept of an entropy in quantum mechanical phase space to a quantum field theory. In this way, it should Appendix C: Wehrl mutual information and Simon be possible to formulate entropic uncertainty relations criterion for conjugate quantum fields with a Wehrl (relative) en- tropy and to constrain measurable (quantum) correlations We wish to show that the Simon criterion implies in the analogous to what we have discussed here. notation of eq. (68) CM = 0 for pure separable Gaussian states. To that end, let V be the covariance matrix of a (1 + 1)-mode pure Gaussian state, such that we can ACKNOWLEDGEMENTS write V = (1/2)ST S with det V = 1/16 and where S is a symplectic matrix S ∈ Sp(4, R). V can be represented 0T 0 0 The authors thank Martin Gaerttner and Markus as S V0S with S = diag(S1,S2) being local symplectic Oberthaler for valuable discussions. This work is sup- transformations and V0 denoting the normal form of V , ported by the Deutsche Forschungsgemeinschaft (DFG,  a 0 c 0  German Research Foundation) under Germany’s Excel- 1 0 a 0 c lence Strategy EXC 2181/1 - 390900948 (the Heidelberg V =  2 . (C1) 0 c 0 b 0  STRUCTURES Excellence Cluster), SFB 1225 (ISO- 1 0 c 0 b QUANT) as well as FL 736/3-1. 2 Using that symplectic transformation have unit determi- nant, det S = det ST = 1, the purity condition can be Appendix A: Non-negativity of Wehrl relative re-expressed as entropy 2 2 1 (ab − c1)(ab − c2) = . (C2) Although the Husimi Q-distribution is only a quasi- 16 probability distribution, it is non-negative and normalized, The parameters a, b, c1 and c2 can be further constrained which suffices to apply Jensen’s inequality [3]. Because from the symplectic eigenvalues λi of V0, which have to − ln x is a convex function for real positive x we get fulfill λi = 1/2 for pure Gaussian states. The symplectic eigenvalues also appear in pairs (±λi) when considering Z 2N σ d α ρ Q (α) the eigenvalues of iΩV0, where Ω = 1 ⊗ J with SW(ρkσ) = − N Q (α) ln ρ π Q (α)   ! 0 1 Z d2N α Qσ(α) J = , (C3) ≥ − ln Qρ(α) −1 0 πN Qρ(α) (A1) ! is the symplectic form, leading to another purity condition Z d2N α = − ln Qσ(α) = 0, 1 πN a2 + b2 + 2c c = det V + det V + 2 det V = . (C4) 1 2 A B M 2 where the validity of the support condition was assumed Note that the latter conditions contains the invariants in the third step. I1 = det VA,I2 = det VB,I3 = det VM and I4 = 13

T Tr(JVAJVM JVBJVM ) under the local symplectic trans- criterion implies det VM = c1c2 = 0. Without loss of 0 2 formations S and that det V = I1I2 + I3 − I4 [33]. generality we proceed with assuming that c2 = 0. Then, Applying the PPT criterion, which corresponds to a the two purity conditions together with the non-negativity mirror reflection in one of the two subsystems, to the of ab imply covariance matrix V0 of a pure Gaussian state leads to   ˜ 2 1 2 a new covariance matrix V0. Assuming that the PPT f(a) = a − a ≥ 16. (C6) condition holds, V˜0 also needs to fulfill the latter purity 2 condition. We find Since f(a) attains a global maximum for a = 1/2 with f(a = 1/2) = 1/16, we can conclude that a = b = 1/2 2 2 1 a + b − 2c1c2 = det VA + det VB − 2 det VM = , (C5) and therefore c = 0, implying V = C = 0 for every 2 1 M M pure separable Gaussian state. as the mirror reflection only flips the sign of det V˜M = − det VM and thus det V = det V˜ . Therefore, the PPT

[1] J. von Neumann, Mathematical Foundations of Quantum (Springer, Dordrecht, 2011). Mechanics (Princeton University Press, 1955). [18] P. J. Coles, M. Berta, M. Tomamichel, and S. Wehner, [2] M. A. Nielsen and I. L. Chuang, Quantum Computation Entropic uncertainty relations and their applications, Re- and Quantum Information: 10th Anniversary Edition views of Modern Physics 89, 015002 (2017). (Cambridge University Press, 2010). [19] A. Hertz and N. J. Cerf, Continuous-variable entropic [3] M. M. Wilde, Quantum Information Theory (Cambridge uncertainty relations, Journal of Physics A: Mathematical University Press, 2013). and Theoretical 52, 173001 (2019). [4] I. Bengtsson and K. Zyczkowski, Geometry of Quantum [20] T. M. Cover and J. A. Thomas, Elements of Information States, 2nd ed. (John Wiley and Sons, 2017). Theory, Second Edition (John Wiley and Sons, 2006). [5] C. E. Shannon, A mathematical theory of communication, [21] H. P. Robertson, The Uncertainty Principle, Physical The Bell System Technical Journal 27, 379 (1948). Review 34, 163 (1929). [6] E. T. Jaynes, Information Theory and Statistical Mechan- [22] E. H. Kennard, Zur Quantenmechanik einfacher Bewe- ics (Notes by the lecturer), in Statistical Physics 3 (1963) gungstypen, Zeitschrift für Physik 44, 326 (1927). pp. 181–218. [23] W. Heisenberg, Über den anschaulichen Inhalt der quan- [7] D. A. Abanin and E. Demler, Measuring Entanglement tentheoretischen Kinematik und Mechanik, Zeitschrift für Entropy of a Generic Many-Body System with a Quantum Physik 43, 172 (1927). Switch, Physical Review Letters 109, 020504 (2012). [24] E. Schrödinger, Zum Heisenbergschen Unschärfeprinzip, [8] R. Islam, R. Ma, P. M. Preiss, M. E. Tai, A. Lukinand, Sitzungsberichte der Preußischen Akademie der Wis- M. Rispoli, and M. Greiner, Measuring entanglement senschaften. Physikalisch-mathematische Klasse 14, 296 entropy in a quantum many-body system, Nature 528, (1930). 77 (2015). [25] R. L. Frank and E. H. Lieb, Entropy and the Uncertainty [9] A. Wehrl, General properties of entropy, Reviews of Mod- Principle, Annales Henri Poincaré 13, 1711 (2012). ern Physics 50, 221 (1978). [26] M. Rumin, Balanced distribution-energy inequalities and [10] A. Wehrl, On the relation between classical and quantum- related entropy bounds, Duke Mathematical Journal 160, mechanical entropy, Reports on Mathematical Physics 567 (2011). 16, 353 (1979). [27] E. H. Lieb, Proof of an entropy conjecture of Wehrl, [11] W. P. Schleich, Quantum Optics in Phase Space (Wiley- Communications in Mathematical Physics 62, 35 (1978). VCH Verlag Berlin, 2001). [28] E. A. Carlen, Some integral identities and inequalities [12] P. Kunkel, M. Prüfer, S. Lannig, R. Rosa-Medina, A. Bon- for entire functions and their application to the coherent nin, M. Gärttner, H. Strobel, and M. K. Oberthaler, Si- state transform, Journal of Functional Analysis 97, 231 multaneous Readout of Noncommuting Collective Spin (1991). Observables beyond the Standard Quantum Limit, Physi- [29] E. H. Lieb and R. Seiringer, Stronger subadditivity of cal Review Letters 123, 063603 (2019). entropy, Physical Review A 71, 062329 (2005). [13] I. Białynicki-Birula and J. Mycielski, Uncertainty rela- [30] M. Grabowski, Wehrl-Lieb’s inequality for entropy and the tions for information entropy in wave mechanics, Commu- uncertainty relation, Reports on Mathematical Physics nications in Mathematical Physics 44, 129 (1975). 20, 153 (1984). [14] H. Everett, "Relative State" Formulation of Quantum [31] M. Ohya and D. Petz, Quantum Entropy and Its Use Mechanics, Reviews of Modern Physics 29, 454 (1957). (Springer-Verlag Berlin Heidelberg, 1993). [15] I. I. Hirschman, A Note on Entropy, American Journal [32] R. Horodecki, P. Horodecki, M. Horodecki, and of Mathematics 79, 152 (1957). K. Horodecki, Quantum entanglement, Reviews of Mod- [16] W. Beckner, Inequalities in Fourier Analysis, Annals of ern Physics 81, 865 (2009). Mathematics 102, 159 (1975). [33] R. Simon, Peres-Horodecki Separability Criterion for Con- [17] I. Białynicki-Birula and Ł. Rudnicki, Entropic Uncertainty tinuous Variable Systems, Physical Review Letters 84, Relations in Quantum Physics, in Statistical Complexity 2726 (2000). 14

[34] V. Giovannetti, S. Mancini, D. Vitali, and P. Tombesi, Quantum Mechanical Descriptions of Statistical Light Characterizing the entanglement of bipartite quantum Beams, Physical Review Letters 10, 277 (1963). systems, Physical Review A 67, 022320 (2003). [56] H.-W. Lee, Theory and application of the quantum phase- [35] O. Guehne, Characterizing Entanglement via Uncertainty space distribution functions, Physics Reports 259, 147 Relations, Physical Review Letters 92, 117903 (2004). (1995). [36] P. Hyllus and J. Eisert, Optimal entanglement witnesses [57] N. D. Cartwright, Some Formal Properties of the Density for continuous-variable systems, New Journal of Physics Matrix, Physica A: Statistical Mechanics and its Applica- 8, 51 (2006). tions 83, 210 (1976). [37] C. Weedbrook, S. Pirandola, R. García-Patrón, N. J. [58] F. A. Berezin, Covariant and Contravariant Symbols of Cerf, T. C. Ralph, J. H. Shapiro, and S. Lloyd, Gaussian Operators, Mathematics of the USSR-Izvestiya 6, 1117 quantum information, Reviews of Modern Physics 84, 612 (1972). (2012). [59] E. H. Lieb and J. P. Solovej, Proof of an entropy conjecture [38] A. Serafini, Quantum Continuous Variables (CRC Press, for Bloch coherent spin states and its generalizations, Acta 2017). Math. 212, 79 (2014). [39] J. Berges, S. Floerchinger, and R. Venugopalan, Dynamics [60] G. D. Palma and D. Trevisan, The Conditional Entropy of entanglement in expanding quantum fields, Journal of Power Inequality for Bosonic Quantum Systems, Commu- High Energy Physics 04, 145 (2018). nications in Mathematical Physics 360, 639 (2018). [40] E. Shchukin and W. Vogel, Inseparability Criteria for [61] G. D. Palma, The Wehrl entropy has Gaussian optimizers, Continuous Bipartite Quantum States, Physical Review Letters in Mathematical Physics 108, 97 (2018). Letters 95, 230502 (2005). [62] R. J. Yáñez, W. V. Assche, and J. S. Dehesa, Position and [41] S. P. Walborn, B. G. Taketani, A. Salles, F. Toscano, momentum information entropies of the D-dimensional and R. L. de Matos Filho, Entropic entanglement criteria harmonic oscillator and hydrogen atom, Physical Review for continuous variables, Physical Review Letters 103, A 50, 3065 (1994). 160505 (2009). [63] V. Majerník and T. Opatrný, Entropic uncertainty re- [42] S. P. Walborn, A. Salles, R. M. Gomes, F. Toscano, and lations for a quantum oscillator, Journal of Physics A: P. H. S. Ribeiro, Revealing Hidden Einstein-Podolsky- Mathematical and General 29, 2187 (1996). Rosen Nonlocality, Physical Review Letters 106, 130402 [64] J. Sánchez-Ruiz, Logarithmic potential of Hermite polyno- (2011). mials and information entropies of the harmonic oscillator [43] R. L. Frank and E. H. Lieb, Extended Quantum Con- eigenstates, Journal of Mathematical Physics 38, 5031 ditional Entropy and Quantum Uncertainty Inequali- (1997). ties, Communications in Mathematical Physics 323, 487 [65] I. Toranzo and J. Dehesa, Exact Shannon entropies for the (2013). multidimensional harmonic states, Physica A: Statistical [44] F. Furrer, M. Berta, M. Tomamichel, V. B. Scholz, and Mechanics and its Applications 516, 273 (2019). M. Christandl, Position-momentum uncertainty relations [66] A. Orłowski, Wehrl’s entropy and classification of states, in the presence of quantum memory, Journal of Mathe- Reports on Mathematical Physics 43, 283 (1999). matical Physics 55, 122205 (2014). [67] E. T. Jaynes, Information Theory and Statistical Mechan- [45] Y. Huang, Entanglement Detection: Complexity and ics, Physical Review 106, 620 (1957). Shannon Entropic Criteria, IEEE Transactions on In- [68] E. T. Jaynes, Information Theory and Statistical Mechan- formation Theory 59, 6774 (2013). ics II, Physical Review 108, 171 (1957). [46] B. Bergh and M. Gärttner, Entanglement detection in [69] E. T. Jaynes, Prior Probabilities, IEEE Transactions on quantum many-body systems using entropic uncertainty Systems Science and Cybernetics 4, 227 (1968). relations (2021). [70] L. D. Landau and J. M. Lifschitz, Vol. 5 - Statistical [47] B. Bergh and M. Gärttner, Experimentally accessible Physics, Part I, in Course of Theoretical Physics (Perga- bounds on distillable entanglement from entropic uncer- mon Press Ltd., 1980) 3rd ed. tainty relations (2021). [71] F. Mehler, Ueber die Entwicklung einer Function von [48] F. Mintert and K. Życzkowski, Wehrl entropy, Lieb con- beliebig vielen Variablen nach Laplaceschen Functionen jecture, and entanglement monotones, Physical Review A hoeherer Ordnung, Journal fuer die reine und angewandte 69, 022317 (2004). Mathematik 66, 161 (1866). [49] L. Mandel and E. Wolf, Optical Coherence and Quantum [72] S. Kullback and R. A. Leibler, On Information and Suffi- Optics (Cambridge University Press, 2013). ciency, The Annals of Mathematical Statisticals 22, 79 [50] A. S. Holevo, Probabilistic and Statistical Aspects of Quan- (1951). tum Theory; 2nd ed., Publications of the Scuola Normale [73] S. Kullback, Information Theory and Statistics (Dover Superiore. Monographs (Springer, Dordrecht, 2011). Publications, 1968). [51] K. Husimi, Some Formal Properties of the Density Ma- [74] A. A. Kuznetsova, Conditional Entropy for Infinite- trix, Proceedings of the Physico-Mathematical Society of Dimensional Quantum Systems, Theory of Probability Japan. 3rd Series 22, 264 (1940). and its Applications 55, 709 (2011). [52] W.-M. Zhang, D. H. Feng, and R. Gilmore, Coherent [75] N. J. Cerf and C. Adami, Negative Entropy and Informa- states: Theory and some applications, Reviews of Modern tion in Quantum Mechanics, Physical Review Letters 79, Physics 62, 867 (1990). 5194 (1997). [53] E. P. Wigner, On the Quantum Correction For Thermo- [76] N. J. Cerf and C. Adami, Quantum extension of condi- dynamic Equilibrium, Physical Review 40, 749 (1932). tional probability, Physical Review A 60, 893 (1999). [54] R. J. Glauber, Coherent and Incoherent States of the [77] G. D. Palma, Uncertainty relations with quantum memory Radiation Field, Physical Review 131, 2766 (1963). for the Wehrl entropy, Letters in Mathematical Physics [55] E. C. G. Sudarshan, Equivalence of Semiclassical and 108, 2139 (2018). 15

[78] J. Schneeloch, C. J. Broadbent, and J. C. Howell, Uncer- A. Frölian, T. Gasenzer, M. Gärttner, and M. K. tainty relation for mutual information, Physical Review Oberthaler, Spatially distributed multipartite entangle- A 90, 062119 (2014). ment enables EPR steering of atomic clouds, Science 360, [79] L. Lami, A. Serafini, and G. Adesso, Gaussian entan- 413 (2018). glement revisited, New Journal of Physics 20, 023030 [82] E. B. Davies, Quantum Theory of Open Systems (Aca- (2018). demic Press, London, 1976). [80] A. Serafini, F. Illuminati, and S. D. Siena, Symplectic [83] A. Miranowicz, J. Bajer, M. R. B. Wahiddin, and invariants, entropic measures and correlations of Gaus- N. Imoto, Wehrl information entropy and phase distri- sian states, Journal of Physics B: Atomic, Molecular and butions of schrödinger cat and cat-like states, Journal of Optical Physics 37, L21 (2003). Physics A: Mathematical and General 34, 3887 (2001). [81] P. Kunkel, M. Prüfer, H. Strobel, D. Linnemann,