See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/351047225

Schrödinger's equation as a diffusion equation

Article in American Journal of Physics · May 2021

DOI: 10.1119/10.0002765

CITATIONS READS

2 676

1 author:

Katsunori Mita
St. Mary's College of Maryland
16 PUBLICATIONS 129 CITATIONS

SEE PROFILE

All content following this page was uploaded by Katsunori Mita on 26 April 2021.

The user has requested enhancement of the downloaded file.

Schrödinger's equation as a diffusion equation
Katsunori Mita

Citation: American Journal of Physics 89, 500 (2021); doi: 10.1119/10.0002765

View online: https://doi.org/10.1119/10.0002765
View Table of Contents: https://aapt.scitation.org/toc/ajp/89/5
Published by the American Association of Physics Teachers

ARTICLES YOU MAY BE INTERESTED IN

Particle velocity = group velocity: A common assumption in the different theories of Louis de Broglie and Erwin
Schrödinger
American Journal of Physics 89, 521 (2021); https://doi.org/10.1119/10.0003165

Space-time computation and visualization of the electromagnetic fields and potentials generated by moving point
charges
American Journal of Physics 89, 482 (2021); https://doi.org/10.1119/10.0003207

Two-body bound states through Yukawa forces and perspectives on hydrogen and deuterium
American Journal of Physics 89, 511 (2021); https://doi.org/10.1119/10.0002998

Simple derivation of the explicit forms of quantum-mechanical fundamental representations

American Journal of Physics 89, 535 (2021); https://doi.org/10.1119/10.0003900

Using mobile-device sensors to teach students error analysis

American Journal of Physics 89, 477 (2021); https://doi.org/10.1119/10.0002906

A guide to the literature of the finite rectangular well

American Journal of Physics 89, 529 (2021); https://doi.org/10.1119/10.0003327
 € dinger’s equation as a diffusion equation
Schro
Katsunori Mitaa)
Department of Physics, St. Mary’s College of Maryland, St. Mary’s City, Maryland 20686
(Received 25 July 2020; accepted 11 November 2020)
Schr€odinger’s equation can be considered as a diffusion equation with a diffusion coefficient
b2 ¼ h=2m. In this article, we explore the implications of this view. Rewriting the wave function in a
polar form and transforming Schr€odinger’s equation into two real equations, we show that one of them
reduces to the continuity equation, and the other, a nonlinear dynamical equation for the probability
density. Considering these two equations as if they were the basic equations of quantum mechanics, we
apply them to several one-dimensional quantum systems. We show the dispersive properties in
the probability densities of stationary states of a particle in a rigid box and in harmonic potential;
quasi-classical Gaussian probability densities of a free particle; and coherent states and squeezed states
of the harmonic oscillator. We also present the soliton as a quantum mechanical representation of a free
particle. We discuss the meaning of the diffusion coefficient b2 for each quantum system using a
density plot. VC 2021 Published under an exclusive license by American Association of Physics Teachers.
https://doi.org/10.1119/10.0002765

I. INTRODUCTION justified in Sec. II (Eq. (9)). The appearance of the imaginary

unit i in the right-hand side of Eq. (4) reflects the fact that
In the course of their studies, physics students encounter the wave function w is complex,2 while u and T are real
two important partial differential equations.1 One of them is functions of x and t.
the one-dimensional wave equation Of course, propagating waves are solutions of Eq. (3).
Then, how can we show the diffusive nature of Eq. (3)? Let
@2u @2u us note that a plane wave cannot represent a free particle for
2
¼ v2 2 ; (1)
@t @x well-known reasons:3 It is not normalizable and its phase
velocity is half the velocity of the corresponding classical
where u is a scalar function representing the physical quan-
particle. Thus, in order to obtain a proper quantum mechani-
tity that depends on space x and time t, and v is the phase
cal representation of a free particle, we take a linear combi-
speed of the wave. The other is the diffusion equation
nation (the Fourier transform) of plane waves to construct a
@T @2T wave packet, usually in the Gaussian form. Then, the wave
¼c 2: (2) packet can be normalized and the particle is well localized,
@t @x satisfying the uncertainty principle. But the dispersion of the
Random physical processes resulting in the diffusion of a wave packet comes into play, as the component plane waves
given quantity T may be described in Eq. (2), where the con- disperse at different phase velocities. This results in the dis-
stant c is the diffusion coefficient. If T is the temperature, persion of the probability density with the coefficient b2 , as
Eq. (2) is called the heat equation. We here consider one- we will see. In this article, we examine the dispersion of the
dimensional cases. probability densities for a Gaussian wave packet as well as
What about Schr€odinger’s equation? Is it a wave equation for other quantum systems.
or a diffusion equation? To answer this question, let us look Dispersion processes of probability densities were previ-
at the free particle Schr€odinger’s equation for the wave func- ously examined by this author through the integrand of the
tion w, expectation value of the momentum operator.4 In the present
work, we go further and initiate the discussion starting from
@w h2 @ 2 w Schr€odinger’s equation. We do so by expressing the wave
i
h ¼ : (3) function in a polar form. Then Schr€odinger’s equation, being
@t 2m @x2
a complex equation, gets split into two real equations. One
Comparing Eq. (3) with Eq. (1) or Eq. (2), we are compelled of them is the continuity equation of the probability density
to say that it looks more like a diffusion equation than a and the other a nonlinear dynamical equation of “the proba-
wave equation, contrary to our common understanding. bility fluid,” a concept that we will explain in detail through-
Let us rewrite Eq. (3) in the form out this paper. The derivation of these equations is done in
Sec. II.
@w @2w In Secs. III–V, we consider these two equations as if they
¼ ib2 2 ; (4)
@t @x were the basic equations of quantum mechanics. We solve
the nonlinear dynamical equation for various quantum sys-
where tems with the continuity equation as a subsidiary condition.

h We also give our probability fluid interpretation for each
b2 ¼ : (5) quantum system, paying special attention to the role the
2m coefficient b2 plays. We will examine stationary states for a
Comparison of Eq. (4) with Eq. (2) suggests that b2 might be particle in the rigid one-dimensional box (Sec. III) and in the
regarded as the diffusion coefficient. This point will be harmonic potential. We then discuss the general properties

500 Am. J. Phys. 89 (5), May 2021 http://aapt.org/ajp C 2021 Published under an exclusive license by AAPT
V 500
 ð
of quasi-classical states represented by the Gaussian proba- ~ ¼ 0:
Dds (11)
bility densities (Sec. IV). As concrete applications of V0
Gaussian probability densities, we analyze the free particle
and the coherent and squeezed states of the harmonic oscilla- Equation (11) states that, at any time t, the total dispersion of
tor. These Gaussian quasi-classical states exhibit varied dis- the probability density always vanishes.
persive behaviors depending on the potential in which the The kinetic energy of a particle can be decomposed into
particle is confined, or the lack of it. two parts: A part due to the current density J~ and the other
Since our dynamic equation is nonlinear, we will examine part due to the dispersion density D~ of the probability density
if this nonlinearity can be counter-balanced by the dispersion distribution. It can be shown that
of the probability density for a given potential. We will see  2 ð
that such a potential can be found and the soliton appears as P
the solution of the equation. This soliton probability density, ¼ eK ds; (12)
2m V0
maintaining a constant degree of the particle’s localization,
would be the quantum mechanical representation closest to a where
classical free particle. A discussion of the free particle soli-
ton is presented in Sec. V. Concluding remarks are given in 1 J 2 þ D2
Sec. VI. eK ¼ m : (13)
2 q
II. DISPERSIVE PROPERTIES OF THE The dispersive properties of usual quantum systems were
PROBABILITY DENSITY IMPLICIT IN discussed in Ref. 4.
€
SCHROINGER’S EQUATION To go beyond this discussion, let us transform the wave
function according to
We will first summarize the dispersive properties of the
quantum mechanical probability densities. A more detailed w ¼ eRþiS ; (14)
derivation of these can be found in Ref. 4.
Consider the expectation value of the momentum operator where R and S are scalar functions of position ~ r and time t.
~ ¼ i
P ~
hr, Let us call this transformation the polar transformation.
ð There are two reasons for this transformation. First, we can
~ ¼
hPi w Pwds;
~ (6) rewrite the kinematical variables in a manner more amenable
V0 to physical interpretation. Substituting Eq. (14) into Eq. (8),
the probability current density can be written as
where wð~r ; tÞ is the wave function of a quantum system,
ds ¼ dxdydz is the infinitesimal volume, and V0 is the entire J~ ¼ q~
g; (15)
~ can be written as
space. The integrand of hPi
  with
w Pw
~ ¼ m J~ þ iD ~; (7)
~ ~ ð2SÞ:
g ¼ b2 r (16)
where
The quantity ~ g has the unit of velocity, and for this reason,
  ~ 
h let us call it the current velocity. Similarly, the dispersive6
J~ ¼ Im w rw (8) ~ can be written as
m density D
is the probability current density, and ~ ¼ q~
D n; (17)

D ~
~ ¼ b2 rq; (9) with

with q ¼ w w being the probability density of a quantum ~ ~ ð2RÞ:

n ¼ b2 r (18)
system. The current density J~ is a familiar quantity appear-
ing in textbooks of quantum mechanics. But the quantity D ~ We will call ~ g and ~
n the dispersive velocity. Note that both ~ n
is seldom discussed. Being the negative gradient of a density, depend on ~ r and t. These are the kinematical parameters of
~ has the form of a diffusion current. Depending on the
D the probability density distribution, the probability fluid.7
quantum system under consideration, we will see that D ~ in The constant b2 then is the dispersion coefficient of this
fact is a diffusion current density, a part of J~ that satisfies the probability fluid. We will see the significance of this coeffi-
continuity equation. cient in various examples.
Equation (9) has a form of Fick’s first law of diffusion, Thanks to Eqs. (15) and (17), Eqs. (10) and (11) can be
where b2 is recognized as the diffusion coefficient. Equation rewritten as
(2) is known as Fick’s second law of diffusion.5 ð
Substituting Eq. (7) into Eq. (6) and noting that hPi ~ is ~
hPi ¼ mq~ gds (19)
real, we have V0
ð and
~ ¼ m Jds
hPi ~ (10) ð
V0
mq~
nds ¼ 0: (20)
and v0

501 Am. J. Phys., Vol. 89, No. 5, May 2021 Katsunori Mita 501
The quantity mq may be thought of as the mass density of Taking the divergence of Eq. (29) and using Eq. (28), we
the probability fluid. It is as if the mass of the particle m have
were distributed over this fictitious fluid. Then the quan-
~ Þ ¼ 1 1r~  J~  1 r2 S:
*
tity mqg in Eq. (19) appears as the momentum density of ð rR
~ Þ  ðrS (30)
the transport process of the probability fluid. Furthermore, 2q 2
* 4b
the quantity mqn may be considered as the momentum
density of the dispersive process. Equation (20) states that Substituting Eqs. (27) and (30) into Eq. (25), we obtain
the sum of the dispersive momentum always vanishes.
Also, the kinetic energy density in Eq. (13) can be rewrit- @q ~ ~
þ r  J ¼ 0: (31)
ten as @t
1   Equation (31) is the continuity equation stating local conser-
eK ¼ mq g2 þ n2 ; (21) vation of probability. Again, this shows that the distribution
2
of probability density behaves as it were a fluid.8
which shows that eK does not diverge as q ! 0, contrary to g and ~
From the definitions of ~ n (Eqs. (16) and (18)), we
what may be inferred from Eq. (13). Later, we will see that have
the first term on the right hand side of Eq. (21) will give rise
to the kinetic energy of the corresponding classical particle. ~ ¼ 1 ~
rS g (32)
The second term yields the purely quantum mechanical 2b2
kinetic energy due to the dispersive process of the probabil-
ity fluid. and
The other more important reason for the transformation in
Eq. (14) is that we can derive two basic equations that ~ ¼ 1 ~
rR n: (33)
describe the behavior of the probability fluid. To this end, let 2b2
us rewrite Schr€odinger’s equation for a particle in an external
potential V as The divergence of Eq. (33), then is
@w V 1 ~ ~
¼ ib2 r2 w þ w: (22) r2 R ¼  r  n: (34)
@t ih 2b2
Using Eq. (14), we have Taking the gradient of Eq. (26), in which Eqs. (32)–(34) are
  inserted, we obtain
@w @R @S
¼ þi w; (23)   1
@t @t @t g 1 ~  2 1 ~  2
@~ ~ r~ ~ ~
þ r g ¼ r n  b2 r n  rV:
@t 2 2 m
and (35)
h i
2 2
r w ¼ ð rR
2 ~ Þ  ðrS
~ Þ þr R w 2
To simplify Eq. (35), let us take the curl of Eq. (33),

þ i 2ð rR
~ Þ  ðrS
~ Þ þ r2 S w: (24) ~ ~ ~  ð rR
~ Þ ¼ 0:
r n ¼ 2b2 r (36)
Substituting Eqs. (23) and (24) into Eq. (22), the real part of
the resulting equation is Hence

@R    *
¼ b2 2ð rR~ Þ  ðrS
~ Þ þ r2 S ; (25) ~ r
r ~ n ¼r ~  n  r2~
~ r n ¼ 0; (37)
@t
and the imaginary part for any vector function ~
n. Therefore, it follows that
h i  
@S V ~ r
r n ¼ r2~
~ ~ n: (38)
2 ð ~ Þ2 ð ~ Þ 2 2
¼ b rR  rS þ r R  : (26)
@t h
Using Eq. (38) in Eq. (35), we finally obtain
Since q ¼ e2R , we have
g 1 ~  2 1 ~  2
@~ 1~
@R 1 @q þ r g ¼ r n  b2 r2~
n  rV: (39)
¼ ; (27) @t 2 2 m
@t 2q @t
Equation (39) describes the balance between the accelera-
and tion of the probability current and the one due to the diffu-
sion current. We view this equation as the equation of
~ ¼ 1 rq:
rR ~ (28) motion for the probability fluid.9 While Schr€odinger’s equa-
2q tion is linear, this equation of motion is nonlinear.
Mathematically, this is due to the fact that this equation
Also, from Eqs. (15) and (16), is derived from Schr€odinger’s equation through the polar
transformation (Eq. (14)). In the rest of this paper, we
~ ð2SÞ:
J~ ¼ qb2 r (29) will apply Eq. (39) to various quantum systems, using

502 Am. J. Phys., Vol. 89, No. 5, May 2021 Katsunori Mita 502
the continuity equation as the subsidiary condition. dn 1 2
Despite the somewhat involved appearance of Eq. (39), it ¼ n þ c0 ; (47)
dx 2b2
reduces to simple, integrable nonlinear differential equa-
tions when it is applied to the concrete systems that we where c0 is an arbitrary integration constant. Next rewrite
consider, thanks to the continuity equation. We will see Eq. (47) as
that the dispersive velocity ~
n plays a significant role in
each of these cases. dn
¼ dx; (48)
c0 þ n2 =2b2
III. ONE-DIMENSIONAL STATIONARY STATES
and integrate once more, we have
For a stationary state, the probability density does not
 
depend on time c1
n ¼ c1 b tan x þ c2 ; (49)
2b
@q
¼ 0: (40) pffiffiffiffiffiffiffi
@t where c1 ¼ 2c0 and c2 is another arbitrary constant.
Substituting Eq. (49) into Eq. (43), we have
Thus, according to the continuity equation (Eq. (31)),
@J=@x ¼ 0. This means that J is constant along the x-axis.  
2 c1
Furthermore, when a particle is confined, J ¼ 0. Hence q ¼ Ncos x þ c2 : (50)
2b
g ¼ 0. Though the probability current does not exist, the dis-
persive process of the probability fluid does take place.10
We next impose the boundary conditions: q ¼ 0 at x ¼ 0 and
Since J ¼ 0, for a particle confined in a one-dimensional
x ¼ a. These give us
potential, the equation of motion for the probability fluid
(Eq. (39)) simplifies to 2b p
c1 ¼ np and c2 ¼ ; (51)
dn 2
d n 1 dV a 2
n  b2 2  ¼ 0: (41)
dx dx m dx and we obtain
 
This is our basic dynamical equation for one-dimensional 2 2 npx
qn ¼ sin : (52)
stationary states. a a
From Eqs. (9) and (17), it follows that
This is the familiar result found in any textbook of quantum
1 1 mechanics.
dq ¼  2 ndx: (42)
q b One can say, that this is a complicated way to obtain this
simple result. We certainly do not need a nonlinear differential
Integrating Eq. (42), we have equation to solve this quantum system. But the point here is
ð that the physical picture has been changed. When we teach this
1 particular example, we usually remind students of the analogy
q ¼ Nexp  2 ndx ; (43)
b of the standing waves of a string of length a, reflecting the
wave-particle duality in the framework of the standard quan-
where N is the normalization constant. This is our basic tum mechanics. In the view presented here, however, an anal-
equation that relates the dispersive velocity to the probability ogy would be a one-dimensional box filled with this fictitious
density. probability fluid. The fluid particles, that have nothing to do
As the first example of application of Eq. (41), let us take with the particle confined in the box, are moving randomly
the case of the stationary state of a particle in the one- without changing the density distribution, satisfying Eq. (20)
dimensional infinite square potential well. Suppose a particle Using our fluid modeling, how the system gains energy
of mass m is confined in the potential can be easily explained. If we look at Eq. (21), the kinetic
( energy density of the system is given by ek ¼ mqn2 =2. It is
0; 0  x  a; as if, at position x, an infinitesimal volume of fluid having
V ð xÞ ¼ (44) mass mqdx moves with velocity n, possessing the kinetic
1; otherwise: energy of ðmqn2 =2Þdx. The energy of the system is found by
summing all of these kinetic energies in the box.
The particle in this potential is free, and Eq. (41) reduces to We next find the energy of the system. From Eqs. (49) and
(51), we have
d2 n dn
b2 ¼n : (45)  
dx2 dx 2npb2 npx
nn ¼  cot : (53)
a a
This simple nonlinear differential equation can be solved in
the following way. First, rewrite Eq. (45) as Thus
ða
d2 n 1 d  2 1 2n2 p2 mb4
¼ n : (46) En ¼ m qn n2n dx ¼ ; (54)
dx2 2b2 dx 2 0 a2

Integrating Eq. (46), we have which is equal to n2 p2 h2 =2ma2 , as it should be.

503 Am. J. Phys., Vol. 89, No. 5, May 2021 Katsunori Mita 503
 Our method can be applied to the stationary states of a From the continuity equation (Eq. (31)), it follows that
particle in the harmonic potential well. As we expect, the ð
mathematics of this case get more demanding than the case J ¼  qt dx; (59)
of the rigid box. But this is also true in the standard way to
solve Schr€ odinger’s equation. However, in our method,
unlike the standard analysis, there is no need to separate out where the subscript t implies the partial derivative with
the asymptotic part of the solution thanks to Eq. (43). This respect t. We will use the subscript notation in the rest of this
equation gives the complete probability density of the har- article. Substituting Eq. (55) into Eq. (59) and using Eqs.
monic oscillator, including the asymptotic part. (17) and (58), we obtain
In our method, the physics of a particle in the one- rrt
dimensional harmonic potential well is very similar to that of a J ¼ qlt þ D: (60)
particle in the rigid box. There is no probability current. But b2
the probability fluid possesses a dispersive momentum and a
In Sec. III, we mentioned that the dispersive density D is not
kinetic energy, as well as a potential energy in this case. We
a current for one-dimensional stationary states. Looking at
can thus calculate the total energy of this fictitious fluid.
Eq. (60) though, this quantity appears as a part of the proba-
We present this case as a problem with a hint: Substituting
bility current J, which satisfies the continuity equation. This
the harmonic potential into Eq. (41), we obtain a nonlinear means that D contributes to a transport of probability fluid.
differential equation, called the Riccati equation.11 A nonlin- Thus, for quasi-classical states, D can be recognized as a dif-
ear Riccati equation can always be reduced to a second order fusion current. From Eq. (60), it immediately follows that
linear differential equation. Using some appropriate paramet-
ric substitutions, we then obtain the differential equation of rrt
g ¼ lt þ n: (61)
which Hermite polynomials are solutions. b2

IV. GAUSSIAN PROBABILITY DENSITIES: From Eqs. (21), (58), and (61), the kinetic energy density
QUASI-CLASSICAL STATES of the system is given by
!
Next, we examine time-dependent quantum states of a 1 1 r2 r2t 2 rrt
2
one-dimensional particle, taking the Gaussian probability ek ¼ mqlt þ mq 1 þ 4 n þ mq 2 lt n; (62)
densities. Let us assume the probability density of the form12 2 2 b b
"  2 #
1 1 x  lðtÞ and integrating ek over the entire space, we obtain the kinetic
qð x; tÞ ¼ pffiffiffiffiffiffi exp  : (55) energy
2prðtÞ 2 rðtÞ
 
1 2 1 2 b4
This probability density contains two time-dependent param- Ek ¼ mlt þ m rt þ 2 ; (63)
2 2 r
eters lðtÞ and rðtÞ. The meaning of these two parameters can
be seen in the following way. First, we find the expectation where Eq. (20) is used. Note that the first term in the right-
value of the position of the particle hand side of Eq. (63) is the kinetic energy of the correspond-
ð1 ing classical particle. The second term is due to the disper-
hxi ¼ qxdx ¼ lðtÞ: (56) sive process of the probability fluid, and its origin is purely
1 quantum mechanical (or due to “quantum fluctuations”).
Next, we evaluate the momentum uncertainty for the
We may regard lðtÞ as the trajectory of the corresponding Gaussian probability densities and see the significance that this
classical particle. The classical trajectories, of course, do not kinetic energy has for the uncertainty principle, due to the dis-
have any position ambiguities. But at a given time, the prob- persive process. The momentum uncertainty is given by
ability density distribution in Eq. (55) has the uncertainty of qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi DP ¼ hP2 i  hPi2 : (64)
Dx ¼ hx2 i  hxi2 ¼ rðtÞ; (57)
From Eqs. (19) and (61), we have
which specifies the dispersive properties of this probability ð1 ð1 ð
density. Thus, as the corresponding classical particle moves rrt 1
along the trajectory lðtÞ, the probability density is spread hPi ¼ mqgdx ¼ mlt qdxþ 2 mqndx ¼ mlt ;
1 1 b 1
around this trajectory according to the Gaussian form. Let us
(65)
call quantum systems corresponding to such probability den-
sity the quasi-classical states.13 since q is normalized and the second integral vanishes
Equation (42) still holds valid and using Eq. (55), we have according to Eq. (20). To evaluate hP2 i, we note that
b2  
nð x; tÞ ¼ ½ x  lðtÞ: (58) 2 2 2 2 2 b4
r2 ðtÞ hP i ¼ 2mEk ¼ m lt þ m rt þ 2 ; (66)
r
Equation (58) indicates that the dispersion of the probability where Ek is given in Eq. (63). Thus substituting Eqs. (65)
density stems from the location of the corresponding classi- and (66) into Eq. (64), we obtain
cal particle at a given time t.

504 Am. J. Phys., Vol. 89, No. 5, May 2021 Katsunori Mita 504
 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the Gaussian form.14 We now show how we can present a free
b4 particle in the context of our probability fluid view.
DP ¼ m r2t þ 2 : (67)
r We saw that the dispersion of the Gaussian probability
density for a free particle is given in Eq. (70). In order to
If the second term in Eq. (63) did not exist, DP would have solve this nonlinear differential equation, let
been zero and this would violate the uncertainty principle.
Therefore, we see that the kinetic energy due to the disper- rt ¼ q; (74)
sive process is an essential feature of quantum mechanics. In
Secs. IV A and IV B, we will use Eq. (67) to evaluate DP for then
concrete quantum systems.
Let us next rewrite the equation of motion (Eq. (39)) in rtt ¼ qqr ; (75)
one-dimension as
and Eq. (70) becomes
12
gt þ ggx ¼ nnx  b nxx  Vx : (68) dr
m qdq ¼ b4 : (76)
r3
Substituting Eqs. (58) and (61) into Eq. (68), we then obtain
Integrating Eq. (76), we have
 
rtt b4 1
 4 ðx  lÞ ¼  Vx  ltt : (69) b4
r r m q 2 ¼ c1  ; (77)
r2
The parameter lðtÞ is the trajectory of the corresponding
classical particle to be specified. Equation (69) is a differen- where c1 is an integration constant. Since r ¼ Dx is the posi-
tial equation for the dispersion parameter rðtÞ. We now tion uncertainty of the particle, r and q are positive. Then
examine the possible potentials that Eq. (69) allows. sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Case 1: Vanishing right-hand side of Eq. (69) b4
q ¼ c1  2 : (78)
(a) V ¼ 0 and l ¼ v0 t: Corresponding to a free classical r
particle of velocity v0 .
(b) V ¼ ax and l ¼ at2 =2m: Corresponding to an accel- Integrating q with respect to t,
erating classical particle of acceleration a. ð
dr
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ t þ c2 ; (79)
In both cases, Eq. (69) reduces to
c1  b4 =r2
r3 rtt  b4 ¼ 0; (70)
where c2 is another integration constant. Let us specify that
because x  l cannot be zero for all x. This is the differential the time starts at t ¼ 0, then c2 ¼ 0 and we denote
equation for the position uncertainty rðtÞ for case 1. rð0Þ ¼ r0 , the initial position uncertainty. Then after the
Case 2: The right-hand side of Eq. (69) is proportional to integration in Eq. (79), we obtain
xl sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2ffi
This is the case if t
r ¼ r0 1 þ ; (80)
1 T
V ð xÞ ¼ mx2 x2 ; (71)
2
where
with  2
r0
lðtÞ ¼ Acosðxt  dÞ; (72) T¼ : (81)
b
where x and d are real constants. Equations (71) and (72) Thus, according to Eq. (55), the probability density for the
represent the classical harmonic oscillator. Substituting Eqs. free particle is given by
(71) and (72) into Eq. (69), we obtain the differential equa-
tion for the dispersive process in this case as "  #
1 1 x  v0 t 2
r3 rtt þ x2 r4  b4 ¼ 0: (73) qðx; tÞ ¼ pffiffiffiffiffiffi exp  ; (82)
2prðtÞ 2 rðtÞ
We will examine the quasi-classical particles for all these
cases in the following. We will see that, depending on the where rðtÞ is given in Eqs. (80) and (81).
potential under which the particle is subject, the probability Recall that b2 ¼ h=2m is the diffusion coefficient. Let us
fluid displays various dispersive behaviors. see the effect of this coefficient on the probability density.
Figures 1(a) and 1(b) show the density plots for b ¼ 0:10
A. Free particle and b ¼ 0:15, respectively. The horizontal axis specifies the
time and the vertical, the position of the particle. The grey
As mentioned in Sec. I, in quantum mechanics, to describe a level indicates the density, brighter white being higher den-
free particle with a sense of localization, we construct a linear sity. The classical trajectory would be a white straight line.
superposition of plane waves to form a wave packet, usually in As time elapses, the probability density disperses in both

505 Am. J. Phys., Vol. 89, No. 5, May 2021 Katsunori Mita 505
 The first term is the classical kinetic energy of a free particle.
The second term depends on the initial position uncertainty
r0 . Thus, we may call this kinetic energy the localization
energy, or the zero-point energy of the free particle. When
the particle is very well localized, r0 is small and the zero-
point energy may become very large.
From Eqs. (67), (80), and (81), we have

DP ¼ mb2 =r0 ; (84)

and the minimum uncertainty product is

h
DxDP ¼ mb2 ¼ ; (85)
2
occurring at t ¼ 0. The uncertainty principle is satisfied.
For a particle under a constant acceleration, the differen-
tial equation for the dispersion rðtÞ is the same as for a free
particle, as discussed at the beginning of this section. Thus,
the same discussion holds valid for this case as for the free
particle, except for the classical path, which is parabolic in
this case.

B. Quasi-classical states of the harmonic oscillator

For the harmonic oscillator, the differential equation for
rðtÞ is given in Equation (73). Looking at this equation, we
see that there are two possible cases. First, if r is constant,
rtt ¼ 0, and we have
b
rc ¼ pffiffiffiffi ; (86)
x

where the subscript “c” refers to the coherent state. In this

case, the probability density is
"  2 #
1 1 x  lðtÞ
qc ð x; tÞ ¼ pffiffiffiffiffiffi exp  ; (87)
2prc 2 rc

where the classical trajectory lðtÞ is given in Eq. (72). This

is the coherent state of the harmonic oscillator.15 Figures
2(a) and 2(b) show two cycles of density plots of the coher-
ent states for b ¼ 0:3 and b ¼ 0:7, respectively. In each of
Fig. 1. Plots of one-dimensional Gaussian probability densities: (a) b ¼ 0:10
and (b) b ¼ 0:15. Other parameters are chosen as r0 ¼ 0:05 and v0 ¼ 1:0
these density plots, the degree of dispersion from the classi-
for both cases. cal trajectory does not fluctuate as time elapses since rc is
constant. But in Fig. 2(b), the oscillating strip is thicker
because the diffusion coefficient is higher.
Figs. 1(a) and 1(b). But the dispersion occurs much quicker The other possibility is that the degree of dispersion fluc-
and over a larger distance when the particle has a larger dif- tuates as time elapses and r is a function of time t. The
fusion coefficient. Note that the higher mass particle has a resulting quantum states are the squeezed-states of the har-
lower diffusion coefficient. Thus, as the mass of the particle monic oscillator.16 In this case, if rt ¼ q, then Eq. (73)
increases, it behaves more like a classical particle displaying becomes
less dispersion. For a free particle, the dispersion of the prob-
ability density continues indefinitely. 1 d 2 b4
q ¼ 3  x2 r: (88)
Next, we consider the kinetic energy for the free particle. 2 dr r
The velocity of the corresponding classical particle is
Integrating Eq. (88) with respect to r, we have
lt ¼ v0 ; and the dispersion rðtÞ is given in Eqs. (80) and
(81). Substituting these into Eq. (63), we obtain
b4
q2 ¼ c   x2 r2 ; (89)
1 1 b4 r2
Ek ¼ mv20 þ m 2 : (83)
2 2 r0 where c is an integration constant. Thus

506 Am. J. Phys., Vol. 89, No. 5, May 2021 Katsunori Mita 506
 The imaginary part of Eq. (92) is
c
 2xr2 ¼ 6sin½2ðxt  h0 Þ; (93)
x
where h0 ¼ xt0 . Because c is an arbitrary constant, Eq. (93)
gives the dispersion rðtÞ of the form
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rs ðtÞ ¼ rs0 1 þ c sin½2ðxt  h0 Þ; (94)

where the subscript s refers to the squeezed state and

0  c < 1, so that r is a real and positive constant. The con-
stant rs0 can be determined by substituting Eq. (94) into Eq.
(73). The result is
b
rs0 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : (95)
x ð1  c2 Þ4
2

The constant c specifies the degree of squeezing. When

c ¼ 0, rs ! rc , and the squeezed state reduces to the coher-
ent state. As c approaches 1, the squeezing becomes distinct.
The squeezing occurs when

sin½2ðxt  h0 Þ ¼ 1; (96)

at which times rðtÞ becomes minimum.

Figures 3(a)–3(c) show the probability densities of the
squeezed-states, for h0 ¼ 0, h0 ¼ p=4, and h0 ¼ 3p=4. The
squeezing appears as bright white spots, satisfying the condi-
tion given in Eq. (96). By changing the parameter h0 , we can
obtain the entire spectrum of the squeezing phenomena.17
The words “coherent state” and “squeezed state” are used
in quantum mechanics and quantum optics, and they stem
from the concept of waves. Since our “modeling” is not of
waves, but rather of fluid and diffusion, the words lose their
original meaning.
Before closing this section, we evaluate the uncertainty
product DxDP for the squeezed states. Substituting Eqs. (94)
and (95) into Eq. (67), we obtain
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
DP ¼ mxrs 1  c sin½2ðxt  h0 Þ: (97)

Since Dx ¼ rðtÞ, we have

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
DxDP ¼ mxr2s 1  c2 sin2 ½2ðxt  h0 Þ: (98)
Fig. 2. Coherent states of the harmonic oscillator: (a) b ¼ 0:3 and (b)
b ¼ 0:7. Other parameters are set as x ¼ 1, d ¼ p=2, and A ¼ 1:
When squeezing occurs, we have
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b4 h
ð DxDPÞmin ¼ mb2 ¼ ; (99)
rt ¼ 6 c  2  x2 r2 ; (90)
r 2
and according to Eq. (96). This is the minimum uncertainty prod-
ð uct, and the uncertainty principle is satisfied.
dr
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 6ðt  t0 Þ; (91)
4
c  b =r2  x2 r2 V. FREE PARTICLE SOLITONS
In Secs. III–IV, we have examined dispersive properties
where t0 is an arbitrary constant with the unit of time. After of the probability densities of various one-particle quantum
carrying out the integration in Eq. (91), we obtain systems, applying our dynamic equation and the continuity
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   equation to these systems. Unlike Schr€odinger’s equation,
2 2 4 4 c 2 Eq. (39) is nonlinear. It is well known that, when effects of
2 cr  x r  b þ i  2xr
x nonlinearity are balanced by dispersion, solitons
result.18,19 It might happen that the cancellation of nonline-
¼ exp½6i2xðt  t0 Þ: (92) arity and dispersion takes place in our probability fluid

507 Am. J. Phys., Vol. 89, No. 5, May 2021 Katsunori Mita 507
Fig. 3. Squeezed-states of the harmonic oscillator: (a) h0 ¼ 0, (b) h0 ¼ p=4, and (c) h0 ¼ 3p=4. Other parameters are set as A ¼ 1, x ¼ 2p, d ¼ p=2, b ¼ 1,
and c ¼ 0:9.

 
method and the soliton emerges as a solution of our 1 2 x  v0 t
dynamical equation. In this section, we show that this is in qðx; tÞ ¼ sech ; (100)
2r0 r0
fact the case.
The soliton would be the ultimate quasi-classical represen- where v0 is the velocity of the corresponding classical free
tation of a free particle, in that its dispersion remains con- particle and r0 , a constant with the unit of length. Figure 4
stant, as in the coherent state of the harmonic oscillator, and shows the density plot of Eq. (100). Comparing Fig. 4 with
the particle remains more like a classical particle maintain- Figs. 1(a) and 1(b), we see that, unlike the Gaussian proba-
ing its initial localization. However, this localization is never bility density, there is no change in dispersion as time elap-
point-like. ses for the soliton probability density.
In Sec. IV, we examined the Gaussian representation of a From the continuity equation (Eq. (31)), it follows that
free particle. We saw that, in the absence of a potential, the ð
probability density disperses indefinitely. We need some
kind of “confining potential” to prevent an indefinite disper- J ¼  qt dx ¼ qv0 ; (101)
sion. But this potential cannot be stationary. Otherwise, the
particle is not free. Thus, such a potential must be a function and thus
of space and time. The question is whether or not such a
potential exists. g ¼ v0 : (102)
To find the potential that will give rise to the soliton, let us
assume a probability density The dynamic equation (Eq. (39)) in one-dimension is

508 Am. J. Phys., Vol. 89, No. 5, May 2021 Katsunori Mita 508
 this phenomenon is a manifestation of combined effects of
dispersion and nonlinearity. This issue deserves a further
examination. Such a study, however, goes beyond the scope
of this article.
In order to justify this point, let us look at Eq. (104) again.
As mentioned before, this equation is identical to that used
in the one-dimensional stationary states. However, the physi-
cal system considered here is dynamical, and the process
develops as time elapses. Thus, we next rewrite Eq. (104) as
a dynamical equation. From the one-dimensional continuity
equation, and Eqs. (15) and (102), it follows that

qt ¼ v0 qx : (108)

From Eqs. (104), (107), and (108), we obtain

r 0 v0  
qt ¼ 4
nnx  b2 nxx : (109)
8b

Equation (109) shows that the time-development of the sol-

Fig. 4. Free particle soliton. v0 ¼ 1 and r0 ¼ 0:09. iton probability density is due to its own dispersive
velocity.
Next, we show that the energy of the soliton can be nega-
1 tive, if it is moving slowly and thus is self-confined.
gt þ ggx ¼ nnx  b2 nxx  Vx : (103) Substituting Eqs. (100), (102), and (106) into Eq. (21), and
m
integrating eK over the probability fluid, we obtain the
Due to Eq. (102), Eq. (103) reduces to kinetic energy of the soliton
 2
1 P 1 2 b4
nnx  b2 nxx  Vx ¼ 0: (104) ¼ mv20 þ m 2 : (110)
m 2m 2 3 r0
Note Eq. (104) is identical to Eq. (41), which appeared in the The potential energy of the soliton can be obtained as
discussion of the one-dimensional stationary states in Sec.
III. From Eq. (104), it follows that ð1
8 b4
ð  hVi ¼ qVdx ¼  m 2 : (111)
 1 3 r0
V ð x; tÞ ¼ m nnx  b2 nxx dx: (105)
Thus, the energy of the soliton is
Next, we find the dispersive velocity n. Substituting Eq.  2
P 1 b4
(100) into Eq. (42), we have E¼ þ hVi ¼ mv20  2m 2 : (112)
2m 2 r0
 
2b2 x  v0 t
nð x; tÞ ¼ tanh : (106) When E ¼ 0, the velocity of the soliton is
r0 r0
b2
Substituting Eq. (106) into Eq. (105), we obtain the desired v0c ¼ 2 : (113)
r0
potential for the soliton
Looking at Eq. (113), we see that, when the velocity of the
8mb4 particle v0 is below this critical velocity v0c , the energy of
V ð x; tÞ ¼  qð x; tÞ: (107)
r0 the soliton becomes negative, and the “free particle” is self-
confined in its own potential. Though this situation appears
The potential given in Eq. (107) is proportional to the strange, we have to remember that the confining potential
negative of the probability density of the soliton. The appear- moves with the soliton.
ance of a potential proportional to the probability density is We next evaluate the uncertainty product for this soliton.
not unique to our discussion here. In the analysis of Bose- As shown in Fig. 4, the dispersion of the probability density
Einstein condensates, the interaction between particles is does not depend on time. Hence, we can evaluate the posi-
represented as a potential proportional to the probability den- tion uncertainty Dx at any time. It is easiest to do this at time
sity of the atoms in the nonlinear Schr€odinger equation.20 t ¼ 0, and we obtain
But in our case, we are not assuming any external influ-
ence. An external object, whatever it is, cannot move with pr0
Dx ¼ pffiffiffi : (114)
the soliton. The particle is assumed to be free. A possible 2 3
interpretation is that the soliton, if it exists in this manner,
generates its own potential and the soliton is self-sustaining, The momentum uncertainty can be evaluated from Eq. (110)
just like the solitary waves of water or light. We suspect that and hPi ¼ mv0 . The result is

509 Am. J. Phys., Vol. 89, No. 5, May 2021 Katsunori Mita 509
 2 b2 principle does not apply. Also, not having tried ourselves, it
DP ¼ pffiffiffi m : (115) is not clear if we can solve problems in dimensions higher
3 r0
than 1. In this context, it will be interesting to see if we can
Thus, the desired uncertainty product is analyze the H-atom in terms of the probability fluid, starting
from Eq. (39).
p
h On the other hand, however, if we look at fluid dynamics,
DxDP ¼ : (116) its fundamental equation, Euler’s equation, is intrinsically
32
nonlinear, even though its dynamics simply reflects
The factor p=3 is 1:05, and Newton’s second law. Looking at superb richness of fluid
dynamics, we suspect that the future development of our

h approach may not be so prohibiting.
DxDP > : (117)
2
ACKNOWLEDGMENTS
The product is slightly above h=2, satisfying the uncertainty
principle. The author would like to thank B. Tennyson, T. Darkhosh,
For the Gaussian free particle representation, Eq. (85) and C. Lobb for enlightening conversations.
gives the minimum uncertainty product of h=2, occurring at
time t ¼ 0. For t > 0, the product is larger than h=2, mono- a)
Electronic mail: [email protected]
1
tonically increasing. There initially was no guarantee that the See, for example, S. M. Lea, Mathematics for Physicists (Brooks/Cole,
above-described solitonic free particle state satisfied the Belmont, CA, 2004), p. 173.
2
uncertainty principle, so that this should be checked. R. Karam, “Schr€ odinger’s original struggles with a complex wave
function,” Am. J. Phys. 88, 433–438 (2020).
However, Eq. (117) shows that the uncertainty product is 3
See, for example, D. J. Griffiths and D. F. Schroeter, Introduction to
constant and slightly above h=2 at all times. Quantum Mechanics (Cambridge U. P., Cambridge, UK, 2018), pp. 55–57.
4
K. Mita, “Dispersive properties of probability densities in quantum
VI. CONCLUSION mechanics,” Am. J. Phys. 71, 894–902 (2003).
5
J. Crank, The Mathematics of Diffusion (Clarendon Press, Oxford, 1975),
Rewriting the wave function in a polar form and trans- p. 4.
6
forming Schr€ odinger’s equation into two real equations, we We use two words “diffusion” and “dispersion” interchangeably; which-
showed that one of them reduces to the continuity equation, ever appears more appropriate in the context of the discussion.
7
C. Cohen-Tannoudji, B. Diu, and F. Lalo€e, Quantum Mechanics (Wiley,
and the other, to a nonlinear dynamical equation for the New York, 1977), p. 238.
probability fluid. We consider these as if they were the 8
K. Mita, “Fluid-like properties of the probability densities in quantum
basic equations of quantum mechanics and apply them to mechanics,” Am. J. Phys. 69, 470–475 (2001).
9
several one-dimensional quantum systems, obtaining famil- E. Nelson, “Derivation of the Schr€ odinger Equation from Newtonian
iar results using this unfamiliar method. We give physical Mechanics,” Phys. Rev. 150, 1079–1085 (1966). In this paper, Nelson
interpretation to each of these examples employing the lan- used the polar transformation, as we do in Eq. (15). As a consequence, our
Eq. (39) is identical to Nelson’s Eq. (48b). The physical interpretation,
guage of fluid mechanics and diffusion theory. Also however, is different. Nelson attempted to show that his “stochastic
included is a discussion of the free particle soliton as a mechanics” yields the basic equations equivalent to the Schr€ odinger equa-
quantum state. tion. And Eq. (48b) appears as a bridge between the two. Our method,
One might wonder if there is a fluid dynamics analogue of however, is to consider Eq. (39) as if it were the fundamental equation of
Eq. (39). Indeed there exists such an analogue with some motion for the probability fluid. Our intention here is to explore the proba-
caveats. Let us compare Eq. (39) with a version of Euler’s bility density focusing on its dispersive behavior. We do not introduce any
physical hypothesis and try to delve further into a deeper layer of quantum
equation of fluid dynamics21 mechanics, as Nelson attempted to do.
10
The quantity D is not a current in this case since it is not a part of J, which
@~
v 1 ~ ð 2Þ 1~
þ r v ~ v  ðr v Þ ¼ fapp  rp;
~ ~ (118) satisfies the continuity equation. In other words, D does not contribute to a
@t 2 q transport of the probability density in this case. We will see later that, in
some cases, D becomes a part of J (see, for example, Eq. (60)). In such
where ~ v is the velocity and p is the pressure at each point of case, we recognize D as a current.
11
See, for example, G. B. Arfken and H. J. Weber, Mathematical Methods
the fluid, at time t. If we assume that the fluid is irrotational, for Physicists (Harcourt/Academic Press, San Diego, CA, 2001), p. 1070.
then r~ ~ v ¼ 0. The quantity fapp denotes the applied force 12
R. W. Robinett, Quantum Mechanics (Oxford U. P., New York, 2006), p. 646.
13
per unit volume. Let us assume this to be zero as well. The It appears that the word “quasi-classical states” is used only for the coher-
~ ~ ent states of the harmonic oscillator. In this paper, however, we loosely
term rp=q corresponds to rV=m in Eq. (39), since the expand the use of this word to include the states, whose position expecta-
potential V confines the probability fluid. Then we see that tion values are identical to the corresponding classical particles.
the current velocity ~ g corresponds to the velocity ~ v of the 14
K. Mita, “Dispersion of non-Gaussian free particle wave packets,” Am. J.
classical fluid. On the other hand, dispersive velocity ~ n, as 15
Phys. 75, 950–953 (2007).
we defined it, is a purely quantum mechanical quantity as we See, for example, Ref. 9, p. 360, Q 12.6.
16
R. W. Henry and S. C. Glotzer, “A squeezed-state primer,” Am. J. Phys.
saw throughout this paper. Consequently, the dispersive 56, 318–328 (1988).
velocity ~n is absent in Eq. (118). Thus, it appears that Eq. 17
G. Breitenbach, S. Schiller, and J. Mlynek, “Measurement of the quantum
(39) is a quantum mechanical version of Euler’s equation, states of squeezed light,” Nature 387, 471–475 (1997).
18
containing both fluid dynamics and diffusion process. B. A. P. Taylor, “What is a solitary wave,” Am. J. Phys. 47, 847–850 (1979).
19
M. Remoissenet, Waves Called Solitons (Springer-Verlag, Berlin, 1994), p. 10.
Since our method relies on a nonlinear dynamical equa- 20
J. Rogel-Salazar, “The Gross-Pitaevskii equation and Bose-Einstein con-
tion, there may be restrictions as to how much further it can densates,” Eur. J. Phys. 34, 247–257 (2013).
be developed. For example, probability density does not 21
See, for example, Eq. (48.19) of A. L. Fetter and J. D. Walecka, Theoretical
allow simple linear combinations, and the superposition Mechanics of Particles and Continua (McGraw Hill, NY, 1980), p. 294.

510 Am. J. Phys., Vol. 89, No. 5, May 2021 Katsunori Mita 510

View publication stats