Sie sind auf Seite 1von 8

ISSN 1063-7788, Physics of Atomic Nuclei, 2008, Vol. 71, No. 5, pp. 884–891.


c Pleiades Publishing, Ltd., 2008.

ELEMENTARY PARTICLES AND FIELDS

Nonlinear Quantum Mechanics: Results and Open Questions*


G. A. Goldin∗∗
Departments of Mathematics and Physics, Rutgers University, Busch Campus, NJ, USA
Received August 10, 2007

Abstract—About 15 years ago, we (Heinz-Dietrich Doebner and I) proposed a special type of nonlinear
modification of the usual Schrödinger time-evolution equation in quantum mechanics. Our equation was
motivated by certain unitary representations of the group of diffeomorphisms of physical space, in the
framework of either nonrelativistic local current algebra or quantum Borel kinematics. Subsequently,
we developed this and related approaches to nonlinearity in quantum mechanics considerably further, to
incorporate theories of measurement, groups of nonlinear gauge transformations, symmetry and invariance
properties, unification of a large family of nonlinear perturbations, and possible physical contexts for
quantum nonlinearity. Some of our results and highlights of some open questions are summarized.
PACS numbers: 03.65.-w
DOI: 10.1134/S1063778808050177

I would like to dedicate this paper to my friend


and collaborator, Professor H.-D. Doebner,
on the occasion of his 75th birthday

1. INTRODUCTION [Jop (g1 ), Jop (g2 )] = −iJop ([g1 , g2 ]).


Beginning in the late 1980s, H.-D. Doebner and Here, ρop and Jop are self-adjoint operator-valued
I sought to interpret quantum-mechanically a cer- distributions, indexed, respectively, by functions and
tain class of unitary representations of the diffeomor- vector fields. The arguments f , f1 , f2 are real-valued,
phism group of physical space. We did not expect then compactly supported C ∞ functions on Rn , the phys-
to be drawn into a nonlinear modification of quan- ical space of the theory, while g, g1 , g2 denote com-
tum mechanics. On the contrary, our tacit assump- pactly supported C ∞ vector fields on Rn , and the
tion that quantum theory must be linear—which, in bracket [g1 , g2 ] = g1 · ∇g2 − g2 · ∇g1 is the usual
line with conventional thinking, we abandoned only Lie bracket of vector fields [4–7]. The correspond-
reluctantly—proved to be the main obstacle to finding ing group (obtained by exponentiating the current
the interpretation we sought. When we had the idea of commutators) is the natural semidirect product of
relinquishing that assumption, we found that the rep- the additive group of scalar functions and the group
resentations led immediately to a family of derivative of compactly supported diffeomorphisms of Rn . This
nonlinear Schrödinger equations and a natural inter- group is easily understood as describing the local
pretation of these equations as describing quantum symmetry of physical space.
systems with dissipation [1–3]. Making use of the usual notation for operator-
Let us review these early ideas briefly. The infinite- valued distributions, one may also write
dimensional Lie algebra of density and current oper- 
ators, also known as nonrelativistic, equal-time local ρop (f ) = ρop (x)f (x)dnx, (2)
current algebra, is specified at arbitrary time t by the Rn
brackets 
[ρop (f1 ), ρop (f2 )] = 0, (1) Jop (g) = Jop (x) · g(x)dn x.
[ρop (f ), Jop (g)] = iρop (g · ∇f ), Rn


The text was submitted by the authors in English. Nonrelativistic local current algebra is an oper-
∗∗
E-mail: geraldgoldin@dimacs.rutgers.edu ator theoretic approach to the understanding and

884
NONLINEAR QUANTUM MECHANICS 885

classification of quantum-mechanical systems, based one has the standard expressions for time-dependent
on self-adjoint representations satisfying the com- quantum probability density and flux in the
mutator algebra of Eqs. (1), or (alternatively) based Schrödinger representation,
on the unitary representations of the correspond- 
ing group. The inequivalent representations describe ρ = ψψ, j := jD=0 = [ψ∇ψ − (∇ψ)ψ]. (5)
a wide variety of distinct quantum systems; for a 2mi
self-contained introduction and recent review, see [8]. But for D arbitrary, one has jD = j − D∇ ∇(ψψ).
Quantum Borel kinematics, on the other hand, is an
The continuity equation may be understood as
approach to quantization that is based on an algebra
a kinematical constraint on the dynamics of a sys-
of scalar functions and vector fields on the config-
uration space of a system, rather than the physical tem whose kinematical properties are described by
space within which the system moves [7]. In the means of representation of Eqs. (1). Imposing it on ρ
case of a single point particle in Rn , the physical and jD , i.e., requiring ∂t ρ = −∇ · jD , we obtain a
space (Rn ) and the configuration space (one-point Fokker–Planck-type equation,
subsets of Rn ) can be identified, so that both perspec- ∂t ρ = −∇ · j + D∇2 ρ, (6)
tives lead to the study of the 1-particle representa-
tions of Eqs. (1). where ρ and j are given by Eq. (5). But unless D = 0,
The usual 1-particle representation at time t may there is no linear time-evolution equation for ψ that
be written satisfies this constraint.
ρop (f )ψ(x1 , t) = mf (x1 )ψ(x1 , t), (3) Allowing nonlinearity, H.-D. Doebner and I ob-
tained (easily) from Eq. (6) a family of nonlinear
 Schrödinger equations with an extra term, in which
Jop (g)ψ(x1 , t) = {g(x1 ) · ∇x1 ψ(x1 , t) the purely imaginary functional i(D/2)∇2 ρ/ρ mul-
2i
+ ∇x1 · [g(x1 )ψ(x1 , t)]}, tiplies ψ on the right-hand side. The form of the non-
linearity given in this term is required by the current
where x1 is the particle coordinate, ψ(x1 , t) is a algebra representation. Then, having been forced to
square-integrable function of x1 that may also be abandon the assumption of linearity in the quantum
written as ψt (x1 ), and m is the mass of the parti- time evolution, there remained no principle ruling out
cle. Note that, if the test function f (x1 ) comes to the presence of additional, real nonlinear functionals
approximate the Dirac distribution δ(n) (x1 − x) multiplying ψ. We limited these functionals to homo-
for a fixed point x ∈ Rn , then the expectation value geneous rational expressions with no more than two
(ψt , ρop (f )ψt ) approaches m|ψ(x, t)|2 , the mass derivatives in the numerator only. To express them in
terms of ψ, it is useful to define a current without 
times the usual spatial probability density; i.e.,
(ψt , ρop (x)ψt ) = m|ψ(x, t)|2 . And similarly, if g(x1 ) or m. Setting ĵ = (m/)j, so that
comes to approximate a constant vector field in the 1
ρ = ψψ, ĵ = [ψ∇ψ − (∇ψ)ψ], (7)
direction of the jth spatial coordinate xj , then Jop (g) 2i
approaches the differential operator −i∂/∂xj , which we define the real functionals Rj [ψ], j = 1, . . . , 5, by
is just the jth component of the usual momentum
operator acting on a domain in the Hilbert space ∇ · ĵ ∇2 ρ ĵ2
R1 := , R2 := , R3 := , (8)
L2dn x (Rn ). Thus we recover standard 1-particle ρ ρ ρ2
quantum mechanics.
But there exists a parametrized family of unitarily ĵ · ∇ρ (∇ρ)2
R4 := , R5 := .
inequivalent representations of Eqs. (1), describing ρ2 ρ2
(in principle) physically distinct quantum theories of
the particle [9, 10]. For D ∈ R, consider The relevant family of nonlinear Schrödinger equa-
tions, at this stage of the development, becomes
D
Jop (g) = Jop (g) + Dρop (∇ · g). (4)
∂ψ i
i = H0 ψ + DR2 [ψ]ψ (9)
Here, D has the dimensionality of a diffusion co- ∂t 2
efficient, while the distributions ρop and Jop D also

satisfy Eqs. (1). To interpret these representations, 


5
+ Dj Rj [ψ]ψ,
let us take expectation values at time t, defin-
j=1
ing the mass-density expectation value mρ(x, t) =
(ψt , ρop (x)ψt ) and the momentum-density expecta- where D and Dj (j = 1, . . . , 5) are diffusion coeffi-
tion value mjD (x, t) = (ψt , JD
op (x)ψt ). When D = 0, cients whose values may be chosen independently.

PHYSICS OF ATOMIC NUCLEI Vol. 71 No. 5 2008


886 GOLDIN

Here, H0 is the usual linear Hamiltonian operator, interval of the outcome of the measurement (“projec-
1 tion postulate”); (v) the assumption that mixed states
H0 ψ = [−i∇ − qA(x, t)]2 ψ (10) are described by density matrices, which are formed
2m
from linear combinations of linear projectors onto
+ [V + qΦ(x, t)]ψ, pure states; and (vi) the assumption that the space of
for a particle of charge q in the presence of the external pure states for a composite system is the tensor prod-
electromagnetic potential (Φ, A), where we have in- uct of the spaces of pure states for its components,
cluded for generality any nonelectromagnetic poten- and the operators for the corresponding observables
tial V (x, t). are obtained by linear extension to the tensor product
Special cases of Eq. (9) include many different space.
nonlinear modifications of quantum mechanics that Each of these assumptions could be the topic of
have been suggested independently by other re- a lengthy discourse. We shall not discuss all of them
searchers [11–17]. Typically, these have been “put here, but focus now on how it is necessary to modify
in by hand,” without any special, fundamental mo- the third one. Many theorists have argued that any
tivation for the nonlinearity, apart from the desirable quantum-mechanical observation of a particle sys-
properties of homogeneity and/or separability. Such tem reduces ultimately to a sequence of positional
modifications generally do not included the local, measurements at different times, with the possible
purely imaginary nonlinear functional multiplying ψ imposition of external force fields between positional
in Eq. (9). measurements [19]. In 1974, Mielnik [20] used this
The form of Eq. (9) seemed to suggest that the idea to propose a theory of measurement for nonlinear
nonlinear terms should be regarded as (presum- variations of quantum mechanics. This view means
ably quite small) perturbations of the usual, linear that the physics can be invariant under transforma-
Schrödinger equation. Each nonlinear term would tions of the wave function that leave the probability
then have its own physical meaning, and the mag- density ψψ in configuration space unchanged (for all
nitude of each diffusion coefficient—in each case, a values of t).
new physical constant—would govern the size of the For simplicity consider the case of a single quan-
effect described by the corresponding term. This was tum particle and set ψ = R exp[iS], where the mod-
the initial point of view that H.-D. Doebner and I ulus R ≥ 0 and the argument S are real-valued
took, but it proved to be incorrect. Nonlinear time (measurable) functions. Then ρ = R2 and
evolution requires a modified theory of measurement, j = (/m)R2 ∇S. The usual, unitary gauge trans-
which entails modification of other ways in which the
formations of quantum mechanics transform ψ to
assumption of linearity enters conventional quantum
ψ  = R exp[iS  ], where R = R, but S  = S + θ(x, t).
mechanics.
The transformations not only leave the outcomes
of positional measurements invariant, but also are
2. NONLINEAR GAUGE strictly local in space and time. Then ρ = ρ, while
TRANSFORMATIONS j = j + (/m)R2 ∇θ. That is, while ρ is gauge
An earlier article characterized at least six dis- invariant, j is not. If ψ satisfies the usual linear
tinct, partially independent assumptions of linearity in Schrödinger equation in the absence of a vector
conventional quantum mechanics [18]. In abbreviated potential, i.e., i∂t ψ = −(2 /2m)∇2 ψ + V ψ, then ψ 
form, these are: (i) the assumption that the space of satisfies i∂t ψ  = (2 /2m)[−i∇ − gradθ]2 ψ  + [V −
pure states forms a linear space H equipped with an θ̇]ψ  . This fact is sometimes used to motivate the
inner product and the time-evolution operator acts “minimal coupling” of the external electromagnetic
linearly on that space; (ii) the assumption that ob- potentials A and Φ and with the Schrödinger equation
servables are described by linear, self-adjoint op- given by Eq. (10). And if we begin with Eq. (10),
erators on H, with the spectrum of such an operator we have that ψ  satisfies the transformed equation
being the set of possible real-valued outcomes of obtained by substituting the gauge-transformed po-
the corresponding observation; (iii) the assumption
that gauge transformations (which leave invariant the tentials: A = A + (/q)gradθ, and Φ = Φ − (/q)θ̇.
outcomes of measurements) act on the space of states The gauge-invariant current is Jgi = j − (q/m)ρA,
as linear, unitary operators; (iv) the assumption that with ∂t ρ = −∇ · Jgi . Likewise the physical fields
immediately upon the performance of a measurement, B = ∇ × A and E = −∇Φ − ∂t A are gauge in-
the initial state for the subsequent time evolution variant. These standard facts establish the pattern
is obtained by applying to the state vector a linear for introducing nonlinear gauge transformations in
orthogonal projection operator onto the subspace the context of our class of nonlinear Schrödinger
of states corresponding to the observed value or value equations.

PHYSICS OF ATOMIC NUCLEI Vol. 71 No. 5 2008


NONLINEAR QUANTUM MECHANICS 887

Abandoning the usual assumption that gauge A2 (x, t) · ĵ


transformations act linearly, Doebner and I intro- + + α1 ln ρ + α2 S,
ρ
duced a group of nonlinear transformations that are
strictly local in space and time, that obey a certain where as above S is arg[ψ] and where ν1 , ν2 , µ1 , . . . ,
separation property for product states (see also [21]), µ5 , α1 , and α2 are real coefficients (that may be, in
and that leave our class of equations invariant (as a general, time-dependent). Using Eq. (13), one recov-
class) [22–24]. Let γ and Λ be continuously differ- ers Eq. (9) from Eq. (14) by making the substitutions
entiable, real-valued functions of t, Λ = 0, and let θ  1 q
be a continuously differentiable, real-valued function ν1 = − , ν2 = D, A= A, (15)
2m 2 2m
of x and t. Define N(γ,Λ,θ) to act on the space of
single-particle wave functions by ψ  = N(γ,Λ,θ) [ψ],  
µ1 = D1 , µ2 = − + D2 , µ3 = + D3 ,
with ψ  given by 4m 2m

R = R, S  = ΛS + γ ln R + θ. (11) µ4 = D4 , µ5 = + D5 ,
8m
Notice that if R = 0 then R = 0; at such points, 1 q2
both S and S  are undefined. But notice also that U = [V + qΦ] + A2 ,
 2m
if S is modified by adding 2πn, where n is an integer, q
then ψ is unchanged at that point, while ψ  is not A1 = 0, A2 = − A, α1 = α2 = 0.
m
uniquely defined (since Λ need not be an integer).
Now let us consider what happens under nonlinear
Thus Eqs. (11) do not specify ψ  uniquely, even up
gauge transformation. We have
to an overall phase, unless (for example) some addi-
tional continuity property is assumed. Nevertheless, ρ = ψ  ψ  = ρ, (16)
we shall see shortly that as a transformation on our
class of nonlinear Schrödinger equations, N(γ,Λ,θ) is 1  γ
ĵ = [ψ ∇ψ  − (∇ψ  )ψ  ] = Λĵ + ∇ρ + ρ∇θ.
in fact well-defined. 2i 2
From Eq. (11), we obtain the semidirect product If ψ solves Eq. (14) with certain coefficients and ex-
group law for nonlinear gauge transformations, ternal fields, then ψ  obtained from (11) solves the
N(Λ1 ,γ1 ,θ1 ) N(Λ2 ,γ2 ,θ2 ) (12) same equation with transformed coefficients and ex-
ternal fields. The new coefficients are
= N(Λ1 Λ2 ,γ1 +Λ1 γ2 ,θ1 +Λ1 θ2 ) . ν1 γ
ν1 = , ν2 = − ν1 + ν2 , (17)
In order to determine how N(γ,Λ,θ) transforms nonlin- Λ 2Λ
γ
ear Schrödinger equations in our class, it is helpful to µ1 = − ν1 + µ1 ,
expand Eq. (9)—including the linear term—entirely Λ
γ 2 γ
in terms of the basis functions Rj . We make use of µ2 = ν1 − γν2 − µ1 + Λµ2 ,
the identity 2Λ 2
 µ3
∇2 ψ 1 1 µ3 = ,
= iR1 [ψ] + R2 [ψ] − R3 [ψ] − R5 [ψ] (13) Λ
ψ 2 4 γ γ2 γ
µ4 = − µ3 + µ4 , µ5 = µ3 − µ4 + Λµ5 ,
and work within the following general class of non- Λ  4Λ  2
linear Schrödinger equations, which includes some γ 1 Λ̇ Λ̇
additional terms that are quite natural [23]. Let the α1 = Λα1 − α2 + γ − γ̇ , α2 = α2 − ,
2 2 Λ Λ
real-valued external field U (x, t) be a (sufficiently
smooth) scalar function, and let A(x, t), A1 (x, t), so that we have a well-defined representation of the
and A2 (x, t) be distinct (sufficiently smooth) real- group of nonlinear gauge transformations on the lin-
valued vector fields. We consider the family of non- ear space of coefficients. The transformed vector and
linear Schrödinger equations, scalar fields are
⎡ ⎤ ν1
ψ̇ 2
∇ · (A(x, t)ρ) ⎦ A = A − ∇θ, (18)
i = i⎣ νj Rj [ψ] + (14) Λ
ψ ρ
j=1 γ
A1 = ΛA1 − γA − A2

5
∇ · (A1 (x, t)ρ) γ 2

+ µj Rj [ψ] + U (x, t) + γ
ρ + ν1 − µ1 + µ3 − µ4 ∇θ,
j=1 Λ Λ

PHYSICS OF ATOMIC NUCLEI Vol. 71 No. 5 2008


888 GOLDIN
2µ3
A2 = A2 − ∇θ, together with the group of nonlinear gauge transfor-
Λ mations we introduced, provide an important unifica-
  tion and generalization of many versions of nonlinear
 Λ̇ µ3 quantum mechanics that have been considered over
U = ΛU − θ̇ + − α2 θ + [∇θ]2 the years.
Λ Λ
γ
γ
+ µ4 − µ3 ∇2 θ + ∇ · A2 − A2 · ∇θ. 3. GAUGE-INVARIANT NONLINEAR
Λ 2
QUANTUM MECHANICS
We remark that permitting γ and Λ in (11) to
be time-dependent has implied now that νj , µj , As we have seen, neither the coefficients nor the
and αj in Eq. (14) should likewise be allowed to external fields nor many of the nonlinear functionals
be time-dependent, so that the class of nonlin- that occur in Eq. (14) are gauge invariant. We next
ear Schrödinger equations remains invariant under summarize a complete set of gauge-invariant quanti-
Eqs. (17). We remark also that the αj must be allowed ties and describe the nonlinear quantum mechanics
of a single particle in terms of these. The resulting
to assume nonzero values, since even if we begin with
equations of motion generalize the usual “hydrody-
α1 = 0 and α2 = 0, Eqs. (17) can result in nonzero
namical” version of quantum mechanics and allow us
values for the αj . Similarly, we note that in Eq. (14) a to see where the dissipative effects enter.
nonzero value for the field A1 , and a general value for We already have the gauge invariance of ρ. The
the field A2 in relation to A, are required by (18).
gauge-invariant current Jgi , entering the continuity
Some of these new features come about solely equation ρ̇ = −∇ · Jgi , is given by
from the allowed nonlinearity of the gauge trans-
formations. If we begin with the linear Schrödinger Jgi = −2ν1 ĵ − 2ν2 ∇ρ − 2ρA. (19)
equation, where from (15) we have ν2 = µ1 = µ4 =
Of course this reduces in the linear case to the usual
α1 = α2 = 0, µ3 = −ν1 , A1 = 0, and A2 = −2A, gauge-invariant current j − (q/m)ρA. The gauge in-
these conditions are preserved by the subgroup
of linear gauge transformations (where γ = 0 and variance of Jgi also means that assuming all mea-
Λ = 1). But allowing general, time-dependent non- surements to be reducible to a succession of posi-
linear gauge transformations to act on the linear tional measurements is unnecessarily restrictive—it
Schrödinger equation introduces nonzero values is sufficient that all measurements be expressible us-
for ν2 , µ1 , µ4 , α1 , α2 , and A1 . Relaxation of the ing gauge-invariant quantities, including ρ, Jgi , and
condition µ3 = −ν1 is not required by gauge trans- the gauge-invariant fields of force defined below. We
formations; it originates with the construction of the also have gauge-invariant parameters
general nonlinear Schrödinger equation; then gauge 1
transformation leads to relaxation of the condition τ 1 = ν2 − µ 1 , τ 2 = ν1 µ 2 − ν2 µ 1 , (20)
2
A2 = −2A. But of course, nonlinear gauge trans- µ3 µ3
formations do not change the physics! The subclass τ3 = , τ 4 = µ4 − µ1 ,
ν1 ν1
of Eq. (14) that is linearizable by means of nonlinear
2 µ3
gauge transformations is physically equivalent to τ 5 = ν1 µ 5 − ν2 µ 4 + ν2 ,
ν1
ordinary quantum mechanics. This fact alone demon-
strates that our descriptions of quantum mechanics ν̇1 ν̇1
β1 = ν1 α1 − ν2 α2 + ν2 − ν̇2 , β2 = α2 − .
need not be linear, and we are provided with a class of ν1 ν1
counterexamples to the more sweeping claims to the Doebner and I discussed the physical meaning of
contrary. some of these parameters [22]; in particular,
As noted above, many of the terms in Eq. (9), τ1 = 0, τ4 = 0, or β2 = 0 violates time-reversal
which is generalized by Eq. (14), have been previ- invariance, while τ3 = − 1 or τ4 = 0 breaks Galilean
ously introduced. In addition, the term with coefficient invariance. The linearizable case (equivalent to or-
α1 was proposed by Bialynicki-Birula and Myciels- dinary quantum mechanics) occurs when τ1 = τ4 =
ki [25]. The term with coefficient α2 was considered by β1 = β2 = 0, τ3 = −1, and τ5 = −τ2 /2. The gauge-
Kostin [26]. Haag and Bannier [27] considered non- invariant parameter β2 , which is most closely related
linear Schrödinger equations with arbitrary values of to Kostin’s nonlinear term, has a natural interpreta-
A2 ; but to the best of my knowledge the external field tion as a coefficient of friction, as we shall see in one
A1 was first introduced in my 1997 article [23]. In of the equations of motion below [28].
short Eq. (14), the general version of the equation Now it is no longer permissible simply to iden-
that Doebner and I derived from representations of tify the gauge-dependent coefficient −ν1 with the
local current algebra and quantum Borel kinematics, gauge-independent, observable constant /2m.

PHYSICS OF ATOMIC NUCLEI Vol. 71 No. 5 2008


NONLINEAR QUANTUM MECHANICS 889

Rather, we should write /m in terms of gauge- equation. It is useful that under nonlinear gauge
invariant parameters. In [22], we noted that in the transformation, we have
linearizable case, τ2 corresponds to the observed
ν1 ν1 Λ̇
value of 2 /8m2 . Then one can take the classical Û  = Û + θ̇ + α2 θ − ν1 2 θ. (25)
limit of our nonlinear quantum mechanics in a gauge- Λ Λ Λ
invariant way, by letting τ2 → 0. But this still leaves We also have two new gauge-invariant vector fields,
open the question of what formula to use for /m
in the general case. In a more recent work, Doebner Agi
1 = ν1 A1 (26)

and I considered the relationship between the energy 2ν2 µ3
levels of stationary states of our nonlinear equation + − µ 1 − µ 4 A − ν2 A 2 ,
ν1
with a spherically symmetric potential and the energy ν1
levels of the corresponding linear Schrödinger equa- Agi
2 = A2 − A.
tion [29]. We proposed the formula 2µ3
2 τ22 An important consequence of this analysis is that
= − , (21)
4m2 τ3 (τ2 + τ5 ) the formula for E in terms of Φ and A has been
modified from its usual form to include an extra term,
which reduces to 2τ2 when (as in the linearizable whose source is in Kostin’s nonlinearity:
case) τ3 = −1 and τ5 = −τ2 /2; the classical limit still
corresponds to τ2 → 0. Noting that for a stationary ∂A
E = −∇Φ − − β2 A. (27)
state having angular frequency ω the combination ∂t
ν1 ω is gauge invariant, we further proposed that its This extra term is unavoidable. If we omit it from
energy be given by the formula Eq. (23), we lose the gauge invariance of E. But now,
2mν1 ω Maxwell’s equations must also be modified. Combin-
E= , (22) ing B = ∇ × A with Eq. (27), we have
τ3
which reduces to ω when (as in the linear case) ∇×E=−
∂B
− β2 B, ∇ · B = 0. (28)
ν1 = −/2m and τ3 = −1. ∂t
Our motivation for these choices is the surprising Note that the second of these equations is still con-
fact that when we make them, the energy levels of sistent with the first. Suppose β2 is a constant, inde-
the stationary states of our nonlinear equation with a pendent of t. If E0 (x, t), B0 (x, t) satisfy the original
spherically symmetric potential, as well as the energy Maxwell equations (with β2 = 0), then fields satisfy-
of plane-wave solutions to the free equation, are un- ing (28) are given by
changed from the corresponding energy levels of the
corresponding linear Schrödinger equation. Sensitive E = E0 e−β2 t , B = B0 e−β2 t . (29)
spectral measurements might thus not be able to
detect even a large nonlinearity. However, we do not But nonlinear quantum mechanics alone does not
at present have an independent justification for these specify the remaining two Maxwell equations. A more
choices, as opposed to other possible gauge-invariant extensive discussion of possibilities associated with
expressions that reduce to the correct answer in linear this modification is included in [30].
quantum mechanics. This is one of the “open ques- Finally, we shall write the fully gauge-invariant
tions” to which the title refers. dynamical equations for the gauge-invariant (hydro-
As described in [28], we also write gauge-invariant dynamical) variables, describing the entire class of
fields; see also [30]. We have the external magnetic nonlinear quantum theories embodied in Eq. (14).
induction B and electric field E, given by The gauge-invariant variables of interest are ρ and
q Jgi /ρ, where the latter has the interpretation of a
B = ∇×A = B, (23) gauge-invariant velocity field. The gauge-invariant
2m
∂A 1 q magnetic induction may then be interpreted as a vor-
E = −∇Û − − β2 A = − ∇V + E, ticity field, as
∂t 2m 2m gi
where J q
∇× = −2B = B. (30)
Û = −ν1 U − τ3 A2 (24) ρ m
− (τ4 − 2τ1 τ3 )∇ · A + A · A2 − ν2 ∇ · A2 . The dynamical equations are the continuity equation,
Thus Û is identified with (1/2m)(V + qΦ), which ∂ρgi
= −∇ · Jgi , (31)
may be verified directly for the linear Schrödinger ∂t

PHYSICS OF ATOMIC NUCLEI Vol. 71 No. 5 2008


890 GOLDIN

together with the equation for the time derivative of 4. OPEN QUESTIONS
the velocity field, Nonlinear quantum mechanics remains a fasci-
gi
∂ Jgi J nating if highly speculative endeavor, of philosophical
= ∇ 2τ1 ∇ · (32) as well as physical interest. We have demonstrated
∂t ρ ρ
that the conventional, axiomatic assumption of lin-
gi 2  earity is too strong, as there exists a class of nonlinear
∇2 ρ 1 J
+ 2τ2 + τ3 quantum theories physically equivalent to ordinary
ρ 2 ρ quantum mechanics via nonlinear gauge transforma-
tion. Even if the quantum mechanics of nature turns
+ ∇ (2τ1 [1 + τ3 ] − τ4 ) out to belong to this “linearizable” class, there is value
in the construction of essentially nonlinear quantum
 theories—they are needed so that we can meaning-
Jgi ∇ρ (∇ρ)2 fully ask the question of why quantum mechanics
× · + 2τ5
ρ ρ ρ2 should be linear, seeking a principle more fundamen-
tal than linearity that distinguishes quantum theories
∇ · (Agi1 ρ) that are linearizable from those that are not.
+∇ 2
ρ In the preceding discussion, we have made refer-
gi  ence to some specific open questions. One of these
gi J is the matter of writing a general, gauge-invariant
− 2τ3 A2 · + 2β1 ln ρ expression for the energy of a quantum state that
ρ
gi applies for arbitrary values of the gauge-invariant
J 1 q parameters and potentials. A related question is that
− β2 − ∇V + E.
ρ m m of justifying the general, gauge-invariant expression
for the observed value of /m. Some progress has
In nonlinear quantum mechanics, we should prop-
been made in addressing the initial-value problem for
erly discuss the “expected value” of an observation
our general nonlinear Schrödinger equation [31] and
(a probabilistic average over positional space) rather
in writing global solutions for particular potentials;
than the “expectation value” of an observable (an
but the full generalization to nonlinear time-evolution
inner product of a self-adjoint operator in a linear
operators of the theory of densely defined unbounded
space). Let us write expressions for the (gauge-
self-adjoint operators in Hilbert space remains, as
invariant) expected values of the position, velocity,
far as I know, to be accomplished. We have also
and acceleration of the quantum particle. Each is, of
alluded to the question of the most general nonlin-
course, a function of the time t. We obtain
 ear Schrödinger–Maxwell electrodynamics. Another
x = xρ(x)dx, (33) avenue of ongoing research by Doebner and his col-
laborators is the generalization to particles with spin
 gi  and to relativistic systems.
dx J
v = = ρ dx = Jgi (x)dx, But the most fundamental open question, in my
dt ρ
 opinion, pertains to the relation between nonlinear
dv quantum time evolutions and local causality. We have
a = = ρ mentioned that the extension of a set of 1-particle
dt
gi 2 gi  nonlinear time evolutions to a many-particle system
1 J J q ∂ Jgi is not unique [21]. Some time ago, Lücke [32] demon-
× ∇ + × B+ dx.
2 ρ ρ m ∂t ρ strated that if we assume the most straightforward
such extension for particles obeying Eq. (9), local
In Eqs. (32), (33), the laws of force describing the causality is violated except in the linearizable case—
interaction of the charged particle with the external in the sense that the positional probability distribution
E and B fields are unchanged from those in linear of a distant particle can change instantaneously as
quantum mechanics. a result of a nearby change in the applied potential.
Taking β2 > 0 in Eq. (32), we understand its in- But to my knowledge, the generality of this result
terpretation as a gauge-invariant coefficient of fric- is not yet established—can it be proven to apply to
tion, since it governs the magnitude of the term in all extensions of nonlinear time evolutions to mul-
∂t (Jgi /ρ) that is proportional to Jgi /ρ. This also ex- tiparticle systems (in which case we would arrive at
plains why the modified Maxwell equations are no a deeper understanding of the logical link between
longer relativistic—a preferred frame of reference has local causality and linearity), or does it allow for the
been established (perhaps, the frame of the universe) existence of a method of extension that respects lo-
by introducing a velocity-dependent frictional force. cal causality? It is also unclear as to whether this

PHYSICS OF ATOMIC NUCLEI Vol. 71 No. 5 2008


NONLINEAR QUANTUM MECHANICS 891

constitutes a fundamental objection to nonlinearity in 14. L. Stenflo, M. Y. Yu, and P. K. Shukla, Phys. Scr. 40,
quantum mechanics, or whether, as Svetlichny [33] 257 (1989).
has suggested, it might (for example) point us in the 15. P. C. Sabatier, Inverse Probl. 6, L47 (1990).
direction of nonlinearities that become important at 16. B. A. Malomed and L. Stenflo, J. Phys. A 24, L1149
Planck scales. (1991).
Answers to these questions could help point the 17. G. Auberson and P. C. Sabatier, J. Math. Phys. 35,
way to a new set of experiments measuring devia- 4028 (1994).
tions, if any, from linearity in quantum mechanics, or 18. G. A. Goldin, in Trends in Quantum Mecha-
should be establishing upper bounds to their magni- nics: Proceedings of the International Sympo-
sium, Goslar, Germany, 1998, Ed. by H.-D. Doeb-
tudes.
ner, S. T. Ali, M. Keyl, and R. F. Werner (World Sci.,
I would like to thank the organizers of the Singapore, 2000), p. 227.
SYMPHYS XII Conference in Yerevan, July 2006, 19. R. P. Feynman and A. R. Hibbs, Quantum Mecha-
for the opportunity to present this work. nics and Path Integrals (McGraw-Hill, New York,
1965).
20. B. Mielnik, Commun. Math. Phys. 37, 221 (1974).
REFERENCES
21. G. A. Goldin and G. Svetlichny, J. Math. Phys. 35,
1. H.-D. Doebner and G. A. Goldin, Phys. Lett. A 162, 3322 (1994).
397 (1992). 22. H.-D. Doebner and G. A. Goldin, Phys. Rev. A 54,
2. G. A. Goldin, Int. J. Mod. Phys. B 6, 1905 (1992). 3764 (1996).
3. H.-D. Doebner and G. A. Goldin, J. Phys. A 27, 1771 23. G. A. Goldin, Nonlinear Math. Phys. 4, 6 (1997).
(1994).
4. R. Dashen and D. H. Sharp, Phys. Rev. 165, 1867 24. H.-D. Doebner, G. A. Goldin, and P. Nattermann,
J. Math. Phys. 40, 49 (1999).
(1968).
5. G. A. Goldin and D. H. Sharp, Lect. Notes Phys. 6, 25. I. Bialynicki-Birula and J. Mycielski, Ann. Phys.
300 (1970). (N.Y.) 100, 62 (1976).
6. G. A. Goldin, J. Math. Phys. 12, 462 (1971). 26. M. D. Kostin, J. Chem. Phys. 57, 3589 (1972).
7. H.-D. Doebner and J. Tolar, in Symposium on Sym- 27. R. Haag and U. Bannier, Commun. Math. Phys. 60,
metries in Science, Ed. by B. Gruber and R. S. Mill- 1 (1978).
man (Plenum, New York, 1980), p. 475. 28. G. A. Goldin, in Quantum Theory and Symme-
8. G. A. Goldin, in Contemporary Problems in Mathe- tries, Proceedings of the International Sympo-
matical Physics: Proceedings of the Third Inter- sium, Goslar, Germany, 1999, Ed. by H.-D. Doeb-
national Conference, Cotonou, Benin, 2003, Ed. by ner, V. K. Dobrev, J.-D. Hennig, and W. Lücke (World
J. Govaerts, M. N. Hounkonnou, and A. Z. Msezane, Sci., Singapore, 2000), p. 111.
(World Sci., Singapore, 2004), p. 3. 29. H.-D. Doebner and G. A. Goldin, in Group Theore-
9. G. A. Goldin, R. Menikoff, and D. H. Sharp, in tical Methods in Physics, Proceedings of the XXV
Measure Theory and Its Applications, Ed. by International Colloquium, Cocoyoc, Mexico, 2004,
G. A. Goldin and R. F. Wheeler (DeKalb, IL, Northern Ed. by G. S. Pogosyan, L. E. Vicent, and K. B. Wolf
Illinois Univ. Dept. of Mathematical Sciences, 1981), (Institute of Physics, Bristol, 2005).
p. 207. 30. G. A. Ascoli and G. A. Goldin, phys/0610020v2.
10. B. Angermann, H.-D. Doebner, and J. Tolar, Lecture
Notes Math. 1037, 171 (1983). 31. H. Teismann, in Physical Applications and Math-
11. T. Kibble, Commun. Math. Phys. 64, 73 (1978). ematical Aspects of Geometry, Groups and Al-
12. F. Guerra and M. Pusterla, Lett. Nuovo Cimento. 34, gebras, Ed. by H.-D. Doebner, P. Nattermann, and
351 (1982). W. Scherer (World Sci., Singapore, 1997), p. 433.
13. D. Schuch, K.-M. Chung, and H. Hartmann, 32. W. Lücke, quant-ph/9710033v2.
J. Math. Phys. 24, 1652 (1983). 33. G. Svetlichny, Int. J. Theor. Phys. 44, 2051 (2005).

PHYSICS OF ATOMIC NUCLEI Vol. 71 No. 5 2008

Das könnte Ihnen auch gefallen