Abstract
It is known that non-commuting observables in quantum mechanics do not have joint probability. This statement refers to the precise (additive) probability model. I show that the joint distribution of any non-commuting pair of variables can be quantified via upper and lower probabilities, i.e. the joint probability is described by an interval instead of a number (imprecise probability). I propose transparent axioms from which the upper and lower probability operators follow. The imprecise probability depend on the non-commuting observables, is linear over the state (density matrix) and reverts to the usual expression for commuting observables.
Export citation and abstract BibTeX RIS
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
1. Introduction
Non-commuting observables in quantum mechanics do not have a joint probability [1–5] (see appendix
As a possible alternative to quasi-probabilities, one can relax the requirement that the sought joint probability correctly reproduces the marginals3
. This is done when studying joint measurements of non-commuting variables [4, 5, 21–26]. Such measurements have to be approximate, since they operate on an arbitrary initial state [4, 5]. They produce positive probabilities for the measurement results, but it is not clear to which extent these probabilities are intrinsic [27], i.e. to which extent they characterize the system itself, and not approximate measurements employed. Alternatively, one can consider two consecutive measurements of the non-commuting observables [28, 29]. These two-time probabilities do not (generally) qualify for the joint probability of the non-commuting observables; see appendix
It is assumed that the sought joint probability is linear over the state (density matrix). If this condition is skipped, there are positive probabilities that correctly reproduce marginals for non-commuting observables [20, 24], e.g. simply the product of two marginals [13]. However, they do not reduce to the usual form of the joint quantum probability for commuting observables4 ; hence their physical meaning is unclear [13].
The statement on the non-existence of joint probability concern the usual precise and additive probability. This is not the only model of uncertainty. It was recognized since early days of probability theory [49] that the probability need not be precise: instead of being a definite number, it can be a definite interval [51–54]; see [55] for an elementary introduction5 .
Instead of a precise probability for an event E, the measure of uncertainty is now an interval , where are called lower and upper probabilities, respectively. Qualitatively, () is a measure of a sure evidence in favor (against) of E. The event E is surely more probable than E', if . The usual probability is recovered for . Two different pairs and can hold simultaneously (i.e. they are consistent), provided that and for all E. In particular, every imprecise probability is consistent with6 ,
It is not assumed that for all E there is a true (precise, but unknown) probability that lies in . This assumption is frequently (but not always [34]) made in applications7
[52, 53], and it did motivate the generalized Kolmogorovian axiomatics of imprecise probability [54]; see appendix
My purpose here is to propose a transparent set of conditions (axioms) that lead to quantum lower and upper joint probabilities. They depend only on the involved non-commuting observables (and on the quantum state).
The next section discusses previous attempts to introduce imprecise probability in quantum physics. Section 3 recalls standard linear algebra notations employed in this work. Section 4 describes physical conditions that are imposed on the sought imprecise probability. Section 5 outlines the main linear-algebra tool (CS-representation for projection operators) that is employed for finding the imprecise probability operators. Details of this representation are outlined in appendix
2. Previous work
In 1967 Prugovecki tried to describe the joint probability of two non-commuting observables in a way that resembles imprecise probabilities [30]. But his expression was not correct, since it still can be negative [13]; cf footnote 1 and see also [18] in this context.
In 1991 Suppes and Zanotti proposed a local upper probability model for the standard setup of Bell inequalities (two entangled spins) [31]; see also [32, 33]. The formulation was given in the classical event space of hidden variables, and it is not unique even for the particular case considered. It violates classical observability conditions for the imprecise probability [31, 34, 54]. In particular, no lower probability exists in this scheme. Despite of such drawbacks, the pertinent message of [31] is that one should attempt at quantum applications of the upper probabilities that go beyond its classical axioms.
More recently, Galvan attempted to empoy (classical) imprecise probabilities for describing quantum dynamics in configuration space [37]. For a general discussion on quantum versus classical probabilities see [38].
3. Notations
All operators (matrices) live in a finite-dimensional Hilbert space . For two Hermitean operators Y and Z, means that all eigenvalues of are non-negative, i.e. for any . The direct sum of two operators refers to the following block-diagonal matrix:
is the range of Y (set of vectors , where ). is the unity operator of . is the subspace of vectors with .
and 0n are the n × n unity and zero matrices, respectively.
In the direct sum of two sub-spaces, it is always understood that and are orthogonal. The vector sum of (not necessarily orthogonal) sub-spaces and will be denoted as . This space is formed by all vectors , where and .
4. Axioms for quantum imprecise probability
Existing axioms for imprecise probability are formulated on a classical event space with usual notions of con- and disjunction and complemention [51, 51–54]; see appendix
The usual quantum probability can be defined over (Hermitean) projectors [39, 40]. A projector generalizes the classical notion of characteristic function. Each uniquely relates to its eigenspace . refers to a set of Hermitean operators :
is a projector to an eigenspace of or to a direct sum of such eigenspaces, i.e. refers to an eigenvalue of or to a union of several eigenvalues. The quantum (precise and additive) probability to observe is , where the density matrix defines the quantum state [4, 5, 39, 40].
Let be another projector which refers to the set of observables. Generally, . Given the density matrix ρ, we seek upper and lower joint probabilities of and (i.e. of the corresponding eigenvalues of and ):
where and are Hermitean operators. Note that the upper and lower probabilities in (2) are assumed to have the usual Born's form, as far as their dependence on ρ is concerned.
We impose the following conditions (axioms):
Equation (2) implies that and depend on and only through and . This non-contextuality feature holds also for the ordinary (one-variable) quantum probability [43, 44]. Provided that the operators and are found, and can be determined in the usual way of quantum averages.
Conditions (3) stem from that are demanded for all density matrices ρ. Equation (4) is the symmetry condition necessary for the joint probability. Equation (5) is reversion to the commuting case. In particular, (5) ensures and
Since means that is anywhere, (8) is the reproduction of the marginal probability. The latter cannot be recovered by summation, since the very probability model is not additive.
For the joint probability is . This expression is well-defined (i.e. positive, symmetric and additive) also for or (but not necessarily ). If , one obtains by measuring (ρ is not disturbed) and then . Alternatively, one can obtain it by measuring the average of an Hermitean observable . Thus (5), (6) demands that and are consistent with the joint probability , whenever the latter is well-defined.
Finally, equation (7) means that () can be measured simultaneously and precisely with or with (on any quantum state), a natural condition for the joint probability (operators)8 .
If there are several candidates satisfying (3)–(7) we shall naturally select the ones providing the largest lower probability and the smallest upper probability.
5. CS-representation
This representation will be our main tool. Given the projectors and Q, Hilbert space can be represented as a direct sum [45–47] (see appendix
where the sub-space of dimension is formed by common eigenvectors of and having eigenvalue α (for ) and β (for Q). Depending on and every sub-space can be absent; all of them can be present only for . Now is the intersection of the ranges of and . has even dimension 2 [46, 47], this is the only sub-space in (9) that is not formed by common eigenvectors of and . There exists a unitary transformation
so that and get the following block-diagonal form related to (9) [46]:
where C and S are invertible square matrices of the same size holding
Now and are sub-spaces of . One has and , where T is the operator analogue of the angle between two spaces. are absent, if P and Q do not have any common eigenvector. This, in particular, happens in .
6. The main result
Note that if (3)–(7) holds for and Q, they hold as well for and , because for . Appendix
Let be the projector onto intersection of and . We now return from (10), (14) and (15) to original projectors and (see appendix
For , , and we revert to . Note that .
7. Physical meaning of upper and lower probability operators
When looking for a joint probability defined over two projectors and one wonders whether it is just not some (operator) mean of and . For ordinary numbers and there are three means: arithmetic , geometric and harmonic . Now (16) is precisely the operator harmonic mean of and [57]
where is the inverse of A if it exists, otherwise it is the pseudo-inverse; see appendix
The intersection projector appears in [39–43]. It was stressed that cannot be a joint probability for non-commutative and [21]. Its meaning is clear by now: it is the lower probability for and . Note that
since two different rays ( and ) cross only at zero. Thus, is non-zero for , only if (or ). I consider this as a natural features of the quantum lower probability, because the classical case—where the lower probability is expected to be non-zero and close to the upper probability—can be generically reached due to the coarse-graining, i.e. due to (or , or both) being sufficiently larger than 1.
Let us now turn to . The transition probability between two pure states is determined by the squared cosine of the angle between them: . Equation (14) shows that depends on , where T is the operator angle between and . Note from (11), (12) that the eigenvalues λ of , which hold are the eigenvalues of C2, and as seen from (14)—they are also (doubly-degenerate) eigenvalues of . Thus we have a physical interpretation not only for (transition probability), but also for eigenvalues of ( and have the same eigenvalues).
Equations (10), (14) and (15) imply that the upper and lower probability operators can be measured simultaneously on any state (cf (7)):
The operator quantifies the uncertainty for joint probability, the physical meaning of this characteristics of non-commutativity is new.
Appendix
Note that the conditional (upper and lower) probabilities are straighforward to define, e.g. (cf (2)): .
The distance between two probability intervals and can be calculated via the Haussdorff metric [56]
which nullifies if and only if and , and which reduces to the ordinary distance for usual (precise) probabilities.
Let us see when we can use the notion of 'surely more probable'. Now
means that the pair of projectors is surely more probable (on ρ) than ; see appendix
holds for , then it also hods for (and vice versa). Though in a weaker sense than (23), (24) means that and together is more probable than neither of them together (which is the pair ). equations (23), (24) are examples of comparative (modal) probability statements; see [36] in this context.
Further features of and are uncovered when looking at a monotonic change of their arguments; see appendix
8. Coordinate and momentum
Coordinate x and momentum p operators, () is the most known example of non-commutativity in quantum mechanics. Hence I shall illustrate the upper and lower probability for this example. In the (one-dimensional) x-representation, x-operator amounts to multiplication, while . For intervals and the corresponding projectors read in the coordinate representation
where is the characteristic function of interval X. Recall that QY and PY are linked via the Fourier transform:
The first thing to note is that if X and Y are finite intervals, then , i.e. and lead to . This is a well-known result in the Fourier analysis; see, e.g. [61–63]. The simplest way to show it is to note (from (26)) that has a finite support, hence is analytic. On the other hand, ϕ should have a finite support. Thus . This argument extends to the case, where (say) Y is semi-infinite, e.g. , while X differs from by a finite (or semi-infinite) interval [63]. Indeed, now is analytic for , while from it follows that is zero in a finite interval at least.
Thus, for finite (or at least one semi-infinite) intervals X and Y the lower probability for the joint distribution of the coordinate and momentum is zero (cf (16) )
However, if both X and Y are finite intervals, , e.g. the above analiticity argument does not work. Moreover, has a discrete spectrum, and its range is infinite-dimensional [61, 62]. We shall avoid this complication by looking at those case, where (at least) one of X and Y is semi-infinite. Then (27) still holds, while the upper probability operator reduces to (cf (17))
and is straightforward to calculate via (25). Several examples of the upper probability calculated from (28) are presented in figure 1.
9. Summary and open problems
The main message of this work is that while joint precise probability for non-commuting observables does not exist, there are well-defined operator expressions for upper and lower imprecise probabilities. They are not additive, but otherwise they do satisfy a number of reasonable conditions: positivity, reproduction of correct marginals, direct observability via quantum averages, consistency with the (effectively) commuting case, where the joint probability is well-defined etc.
Several open questions are suggested by this research. First of all, it is not clear what is the suitable way of defining averages over the imprecise probability. This would be necessary for defining various correlation functions. Recall that the average of a random variable X that has a precise probability is defined via two conditions (see e.g. [60]): linearity, , and monotonicity, implies . Presumably, these conditions are to be modified for imprecise probability; in addition the sought average should reduce to the usual one when averaging over a single observable. This question should be clarified before the imprecise probability can be efficiently applied in quantum statistical mechanics.
Another open issue relates to the point (see (6)) that whenever the joint probability for non-commuting projectors and is well defined due to e.g. (see the discussion after (8)), the upper and lower probabilities and are merely consistent with the exact probability , but are not equal to it, which would be a more desired outcome. It is thus not completely clear whether the found imprecise probabilities cannot be made more precise by looking at more general conditions (axioms), e.g. those that involve a nonlinear dependence on the density matrix ρ; cf (2). Such a dependence might however impede the direct observability of imprecise probabilities; the resulting issues need further investigations.
In a more remote perspective, one can ask about the joint imprecise probability of 3 (and more) non-commuting observables. In contrast to the previous two open problems, where the progress looks to be feasible, this is a difficult problem, because no analogue of the CS-representation for 3 (or more) non-commuting projectors seems to exists; see however [64].
Acknowledgments
I thank KV Hovhannisyan for discussions. I was supported by COST network MP1209.
Appendices
All 6 appendices can be read independently from each other.
Appendices A–C recall, respectively, the no-go statements for the joint quantum probability, generalized axiomatics for the imprecise probability and the CS-representation. This material is not new, but is presented in a focused form, adapted from several different sources.
Appendix D contains the derivation of the main result, while appendices E and F demonstrate various feature of quantum imprecise probability.
Appendix G illustrates it with simple physical examples.
Appendix A.: Non-existence of (precise) joint probability for non-commuting observables
A.1. The basic argument
Given two sets of non-commuting Hermitean projectors:
we are looking for non-negative operators such that for an arbitrary density matrix ρ
These relations imply
Now the second (third) relation in (A4) implies (). Hence .
Thus, if (e.g. when Pk and Qi are one-dimensional (1D)), then , which means that the sought joint probability does not exist.
If , then the largest that holds the second and third relation in (A4) is the projection on . However, the first relation in (A4) is still impossible to satisfy (for ), as seen from the superadditivity feature (F1):
A.2. Two-time probability (as a candidate for the joint probability)
Given (A1), (A2), we can carry out two successive measurements. First (second) we measure a quantity, whose eigen-projections are (). This results to the following joint probability for the measurement results (ρ is the density matrix)
Likewise, if we first measure and then , we obtain a quantity that generally differs from (A6):
If we attempt to consider (A7) (or (A6)) as a joint additive probability for Pi and Qk, we note that (A7) (and likewise (A6)) reproduces correctly only one marginal:
One can attempt to interpret the mean of (A6), (A7)
as a non-additive probability. This object is linear over ρ, symmetric (with respect to interchanging Pk and Qi), non-negative, and reduces to the additive joint probability for . The relation can be interpreted as consistency with the correct marginals (once is regarded as a non-additive probability, there is no point in insisting that the marginals are obtained in the additive way).
However, the additive joint probability is well-defined also for (or for ). If holds, is not consistent with , i.e. depending on ρ, Pk and Qi both
are possible.
To summarize, the two-time measurement results do not qualify as the additive joint probability, first because they are not unique (two different expressions (A6) and (A7) are possible), and second because they do not reproduce the correct marginals. If we take the mean of two expressions (A6) and (A7) and attempt to interpret it as a non-additive probability, it is not compatible with the joint probability, whenever the latter is well-defined.
Appendix B.: Axioms for classical imprecise probability
B.1. Generalized Kolmogorov's axioms
Given the full set of events Ω, and defined over sub-sets of Ω (including the empty set ) satisfy [52–54]:
where includes all elements of Ω that are not in A, and where means intersection of two sets; holds for elementary events.
Here are some direct implications of (B1)–(B5)
Equation (B7) follows directly from (B4). Equation (B6) follows from (B4), (B3). Next relation:
which, in particular, implies
To derive (B8), note that (B4), (B3) imply or , which is the first inequality in (B8). The second inequality is derived via (B5), (B3).
The following inequality generalizes the known relation of the additive probability theory
To prove (B10), we denote , which means . Now
where in (B11) (resp. in (B12)) we applied the first (resp. the second) inequality in (B8).
Note that the (non-negative) difference between the upper and lower probabilities also holds the super-additivity feature (cd (B5))
Employing (B8) one can derive [58] for arbitrary A1 and A2:
B.2. Joint probability
The joint probabilities of A and B are now defined as
Employing the distributivity feature
which holds for any triple , we obtain from (B4), (B5) for
B.3. Dominated upper and lower probability
The origin of (B1)–(B5) can be related to the simplest scheme of hidden variable(s) [52]. One imagines that there exists a precise probability , where the parameter θ is not known. Only the extremal values over the parameter are known:
However, it is generally not true that (B1)–(B5) imply the existence of a precise probability that holds (B21) [53].
Appendix C.: Derivation of the CS-representation
C.1. The main theorem
Let and are two subspaces of Hilbert space that hold ( is the orthogonal complement of )
The simplest example realizing (C1) is when and are 1D subspaces of a two-dimensional (2D) .
Let and be projectors onto and respectively. Now is the projector of , and let be the projector . Employing the known formulas (see e.g. [47])
we get from (C1)
which means that should be even for (C1) to hold9 .
Here is the statement of the CS-representation [46]: after a unitary transformation and can be presented as
where all blocks in (C4) have the same dimension m.
To prove (C4), note that and can be written as (cf (C3))
Next, let us show that
Since , we need to show that for any , means . Indeed, we have , which together with (see (C1)) leads to . Equation (C6) implies that there is the well-defined polar decomposition ( is Hermitean, while V is unitary)
We transform as
where . We shall now employ the fact that the last matrix in (C8) is a projector:
The first and second relations in (C9) show that . Then the third relations produces . Since (due to ), we conclude that . The rest is obvious.
C.2. Joint commutant for two projectors
Given (C4), we want to find matrices that commute both with and [46]. Matrices that commute with read
Employing (C4), we get that (C10) commutes with if
Since C and S are invertible, (C11), (C12) imply that . And then (C13) implies that X = Y. Hence
C.3. General form of the CS representation
The above derivation of (C4) assumed conditions (C1). More generally, the Hilbert space can be represented as a direct sum [45–47]
where the sub-space of dimension is formed by common eigenvectors of and having eigenvalue α (for ) and β (for Q). Depending on and every sub-space can be absent; all of them can be present only for . Now . has even dimension [46, 47], this is the only sub-space that is not formed by common eigenvectors of and .
After a unitary transformation
and get the following block-diagonal form that is related to (C15) [46] and that generalizes (C4):
where C and S are invertible square matrices of the same size holding
refers to P' and Q'. If and do not have any common eigenvector, and .
Appendix D.: Derivation of the main result (equations (16), (17) of the main text)
We start with representation (C17), (C18) and axioms (3)–(7) of the main text. These axioms hold for , and (see (C16)) instead of , and ρ, because for (recall that and are Taylor expandable). Hence we now search for and .
The block-diagonal form (C17), (C15) remains intact under addition and multiplication of and . Hence and have the block-diagonal form similar to (C17), where the diagonal blocks are to be determined. Let now be the projector on . We get []
Hence condition (5) of the main text implies [for and ]
Aiming to apply (6) of the main text, we write down (C17) explicitly as
The most general density matrix that commutes with reads (in the same block-diagonal form)
Now is seen from the fact that after permutations of rows and columns, and become and , respectively. Note that and .
Condition (7) (of the main text) and (C14) imply
Recall condition (6) of the main text. It amounts to (D5) (D6) that should hold for arbitrary aik and bik. Hence we deduce: and hence . Likewise, (D7) (D5) leads to , ; recall that we want the smallest upper probability. Now (4) (of the main text) holds, since
Thus (cf (C15))
Now is the projector onto . To return from (D10), (D11) to original projectors and Q, we note via (C17), (C18) (recall that is the projector onto ):
We act back by U, e.g. , and get finally
Appendix E.: Various representations of upper and lower probability operators
E.1. Representations for the upper probability operator
Let us turn into a more detailed investigation of (D18). Note from (C17), (C18) and (D12) that is the projector to , where is the vector sum of two sub-spaces. Note the following representation [48]:
where is the pseudo-inverse of Hermitean A, i.e. if (where V is unitary: ), then .
The third equality in (E1) is the obvious feature of the pseudo-inverse. The first equality in (E1) follows from the fact that is the projector on and the known relation [48]:
Employing , can be presented as a function of (cf (E1)):
Note another representation for the projector to [40]
where equals to the minimal Hermitean operator A that holds two conditions after .
E.2. Representations for the lower probability operator
Let us first show that the projector into holds [57]
where is the pseudo-inverse of A (cf (E1)).
The last equality in (E5) follows from the fact that is the projector to (see (E1), E2)), which then leads to .
The first equality in (E5) is shown as follows. Let . Then using (E5):
Thus, . On the other hand, and , where the first relation follows from the implication: if , then .
There are two other (more familiar) representations of (see e.g. [40, 48]):
Equation (E6) can be interpreted as a result of (infinitely many) successive measurements of and . Equation (E7) should be compared to (E4).
Yet another representation is useful in calculations, since it explicitly involves a 2 × 2 block-diagonal representation [57]:
where , , and are, respectively, , , , matrices.
E.3. Direct relation between the eigenvalues of and
We can now prove directly (i.e. without employing the CS representation) that there is a direct relation between the eigenvalues of and . Let be the eigenvector of Hermitean operator :
where is the eigenvalue. Multiplying both sides of (E10) by (by Q) and using () we get
which then implies
Thus () is an eigenvector of () with eigenvalue .
As seen from (E11), the 2D linear space formed by all superpositions of and remains invariant under action of both and . Together with this means that if (E10) holds, then has eigenvalue with the eigen-vector living in .
Further details on the relation between and can be looked up in [59].
Appendix F.: Additivity and monotonicity
We discuss here the behavior of and (given by (D17), (D18)) with respect to a monotonic change of their arguments. For two projectors and Q, means , where and . Now (D17), (D18) and (E7) imply that is operator superadditive
Likewise, is operator subadditive, but under an additional condition:
They are the analogues of classical features (B4) and (B5), respectively. Note that (F1) and (B4) are valid under the same conditions, since is the analogue of . In that sense the correspondence between (F2) and (B5) is more limited, since is more restrictive than .
We focus on deriving (F1), since (F2) is derived in the same way. Note from (E7) that and imply . Since , . Using (E7) for we obtain (F1).
Note as well that both and are monotonous []:
Equation (F3) for follows from (F1). For it is deduced as follows (cf (E4), (E7)):
Let us now discuss whether (F2) can hold under the same condition as (F1). Now
amounts to
First of all note that for and we get and (F6) does hold for the same reason as (F1).
For , equation (F6) is invalid in 3D space (as well as for larger dimensional Hilbert spaces). Indeed, let us assume that and are 1D:
Given as
we get (cf (E9))
where is the pseudo-inverse of a33.
Likewise
Now , since is a 1d projector. We can now establish that generically
(let alone (F6)), because the difference has both positive and negative eigenvalues.
The message (F9) is that the function is not sub-additive.
Now consider (F5), (F6), but under additional condition that . Now (F6) amounts to
which holds as equality since .
Appendix G.: Upper and lower probabilities for simple examples
G.1. 2D Hilbert space
It should be clear from (D10), (D11) that in 2D Hilbert space, any lower probability operator is zero (since two rays overlap only at zero), while the upper probability operator just reduces to the transition probability (i.e. to a number) . Thus for the present case both and do not depend on ρ.
G.2. Spin 1
G.2.1. Projectors
The 3 × 3 matrices for the spin components read
Now for are the 1D projectors to the eigenspace with eigenvalues ± 1 or 0 of La:
where the zero components are orthogonal to each other:
Other overlaps are simple as well
Given two projectors P and Q, we defined as the projector on . For calculating we employ (E5).
G.2.2. Fine-grained joint probabilities for and
Here are upper probability operators for joint values of and :
Since and are 1D projectors, all the lower probability operators nullify. Equation (G7) means that the precise probability of Pz0 and Px0 is zero; cf (G5).
This matrix is larger than , since its eigenvalues are , and .
Note from (C17), (C18) that for 3 × 3 matrices , while (if this sub-space is present at all). Hence the eigenvalues of relate to transition probabilities (G6). Indeed, the eigenvalues of matrices in (G8), (G9) (resp. in (G10), (G11)) is (resp. ). Hence the maximal probability interval that can be generated by (G8), (G9) is smaller than the maximal interval generated by (G10), (G11). As an example, let us take the upper probabilities generated on eigenstates of Ly ( ):
G.2.3. Coarse-grained joint probabilities for and
Let us now turn to joint probabilities, where the lower probability is non-zero
Now for (G14), (G15) has eigenvalues , while for for (G16), (G17) this matrix has eigenvalues (the last case (G18) refers to the commutative situation). Hence the probabilities for (G16), (G17) are more uncertain.
Next, let us establish whether certain combinations can be (surely) more probable than others. Note that
Once there is (one) positive eigenvalue, there is a class of states ρ for which
i.e. and is more probable than and . Note that
Such examples can be easily continued, e.g.
Footnotes
- 1
Negative probabilities were not found to admit a direct physical meaning [14] (what can be less possible, than the impossible?). In certain cases what seemed to be a negative probability was later on found to be a local value of a physical quantity, i.e. physically meaningful, but not a probability [14]. Mathematical meaning of negative probability is discussed in [18, 19].
- 2
One should stress here that the usage of quasi-probabilities in statistical mechanics is frequently implicit, but is nevertheless essential. For instance, the routine introduction of symmetrized correlators of non-commuting variables [17] implies an implicit choice of the underlying Terletsky–Margenau–Hill quasi-probability, because the symmetrized correlators are the 'real' correlators with respect to this quasi-probability. This point is seen in the standard quantum fluctuation–dissipation theorem [17].
- 3
Employing instead the unbiasedness: the averages of the non-commuting quantities are reproduced correctly [22].
- 4
Given two projectors and and state , this product is , while the correct form for is .
- 5
Ellsberg's paradox is an example in psychology, where the ordinary probability theory does not apply, while imprecise probabilities can be used fruitfully for explaining experimental results on human decision making [50].
- 6
This 'nothing is known' situation cannot be represented by usual probabilities, the simplest example showing that imprecise probabilities can model types of uncertainty that are not captured by the precise model.
- 7
- 8
- 9