Sei sulla pagina 1di 26

PROCEEDINGS OF SPIE

SPIEDigitalLibrary.org/conference-proceedings-of-spie

Emergent mechanics, quantum and


un-quantum

John P. Ralston

John P. Ralston, "Emergent mechanics, quantum and un-quantum," Proc.


SPIE 8832, The Nature of Light: What are Photons? V, 88320W (1 October
2013); doi: 10.1117/12.2025000

Event: SPIE Optical Engineering + Applications, 2013, San Diego, California,


United States

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use


Emergent Mechanics,
Quantum and un-Quantum
John P. Ralston
Department of Physics & Astronomy, University of Kansas, Lawrence KS 66045 USA

ABSTRACT
There is great interest in quantum mechanics as an ”emergent” phenomenon. The program holds that non-
obvious patterns and laws can emerge from complicated physical systems operating by more fundamental rules.
We find a new approach where quantum mechanics itself should be viewed as an emergent information man-
agement tool not derived from physics nor depending on physics. The main accomplishment of quantum-style
theory comes in expanding the notion of probability. We construct a quantum probability map from macroscopic
information as “data” to quantum information. The map permits a hidden variable description for quantum
states, and efficient use of the helpful tools of quantum mechanics in unlimited circumstances. We construct a
dynamical quantum map from generalized Hamiltonian dynamics to quantum dynamics. We show that under
wide circumstances Hamiltonian dynamics emerges from structureless dynamical systems by embedding them in
an appropriate phase space. The uses of the quantum information management tools are illustrated by numerical
experiments and practical applications.
Keywords: Quantum mechanics, photons, entanglement, quantum probability, density matrix theory

1. PHYSICS EVOLVES
The concept of photons was developed more than a century ago, yet engineers and physicists are still asking
“what are photons?” We have a concise proposal: Photons are a theory.
Quantum theory has evolved. Photons no longer mean the “light quantum” Einstein proposed. Yet historical
holdovers impede the development and communication of physics. We will explore an approach breaking with
the historical development. We view quantum theory as emergent, a word that has two meanings. The word
“emergent” means discovering non-obvious characteristics or orderliness in macro-systems that are not put in by
hand in the micro-system. Some look to deeper physics beyond the Standard Model to provide this.1, 2 By clearing
away historical prejudice we have developed a different view. We find that the framework of quantum mechanics
itself lacks fundamental information about micro-systems. We find it is descriptive for when customized to
systems desired to be described, not predictive. Many redundant postulates of quantum mechanics also emerge
as outcomes, in that they can be derived, by simply dealing with the topic from a modern perspective and
minimal mathematics.
We believe that adding new assumptions to explain old ones would be the wrong road. It appears that
“axioms” seldom matter for what physicists and engineers do in practice. Unlike mathematics, there is no
conservation law on the number of assumptions needed to set up physics. What was minimal in the mid-1920’s
has evolved because concepts and the larger theory has evolved. It was once a great accomplishment to set up
a new quantum principle and defend it. We say it is a great accomplishment to get rid of redundant principles,
one by one. The exaggeration of mystery found in early quantum mechanics has probably run its course.
We maintain that much of quantum theory is practical and procedural, and that the traditional restriction of
quantum methods to describing micro-physical objects of fundamental physical character will soon be obsolete.
But rather than dispute with those who might be offended about what constitutes quantum mechanics, we will
say that in our approach of un-Quantum systems there are no deep postulates.
E-mail: ralston@ku.edu; telephone: 1 785 864 4626

The Nature of Light: What are Photons? V, edited by Chandrasekhar Roychoudhuri,


Al F. Kracklauer, Hans De Raedt, Proc. of SPIE Vol. 8832, 88320W · © 2013 SPIE
CCC code: 0277-786X/13/$18 · doi: 10.1117/12.2025000

Proc. of SPIE Vol. 8832 88320W-1


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
1.0.1 Example: Compton Scattering
For reasons we can’t explain, communication about theory has not kept up with the changes of theory. When
the theory changes the meaning of the words changes, and concepts need to be revised.
Many mixups were caused by accepting the word “particle” to mean a point-like entity with three canonical
coordinates (q1 (t), q2 (t), q3 (t)). In 1923 Compton used that kind of particle framework, with relativistic energy
and momentum conservation, to interpret his data on x-ray scattering. Nearly a century later the argument is
still presented as proof that photons are “the particles of light,” especially the particle aspect held over from
the old quantum theory. Yet in 1926 everything about the old quantum theory disappeared. By 1928 Klein
and Nishina3 had calculated the differential cross section of Compton scattering. The calculation describes
both electrons and photons with plane waves. Translationally invariant correlations (Section 4) obey laws of
conservation of frequency and wavenumber whether or not a field theory or quantum mechanics is involved.
Compton’s experimental analysis, indirectly assuming translational symmetry, gives evidence for translational
symmetry. By 1928 the philosophical quandaries of waves versus particles should have already faded away.
Unfortunately the development of quantum mechanics was less than ideal. We can’t explain why 43 years
passed between the Schrödinger equation and Lamb and Scully4 noticing the quantum photoelectric effect is a
resonant transition of electron waves, which does not need photons, hence not by itself a proof of photons. In the
interim most of the Standard Model was discovered, without anywhere needing to dwell on the Correspondence
Principle, the Complementarity Principle, the Bohr-Sommerfeld Quantization Principle, or any other Principle
about the quantum world expressed before 1926. None were generally true.
We propose that quantum theory has evolved so much that retaining the concepts and methods of its founda-
tion is not just obsolete, it is disabling. To deal with history some of our discussion is pedagogical. Yet our main
thrust is to generate a reasonable and scientifically conservative framework, which is at the same time entirely
new. There are several novel results. We propose that quantum theory is a bookkeeping tool for managing the
information in correlations of experimental data. There is no physical information in the humanly-constructed
organizational structure and nothing to test scientifically. This is new, inasmuch as it has usually been assumed
that the consistency of quantum theory hinges on faithful representation of new and bizarre quantum objects the
human mind cannot fully comprehend. It has usually been assumed that the framework of quantum mechanics
was the major physical discovery of the 20th century. Within that framework come particular and lesser details
which are models. The quantum Hydrogen atom is a model, the two-state, spin-1/2 system is a model, the
photon is a model.
We maintain the opposite of the old view: The actual physical accomplishments lie entirely in the model. We
have nothing but admiration for a good model. How do we find good models? We will show that good models
can inevitably be found, given data. The key is minimal (not maximal) entropy.

2. THINGS THAT EMERGE


We will summarize results developed in more detail elsewhere.5–7 It may appear at first that topics are out
of order, in the sense that the order does not go from “elementary” to “advanced”. In some cases elementary
concepts only came to be addressed at a late point in history using advanced methods.

2.1 The Path Integral


Opening any textbook on quantum field theory,8 one will find impressive formulas for correlations C(y1 ...yp ),
such as
Z
C(y1 ...yp ) = d[φ] e−S(φ) φ(y1 )...φ(yp ). (1)

Textbooks will present a derivation from the path-integral approach to quantum mechanics, following a long road
needing many independent physical assumptions in advance, one would think.
Find another path. Much of 21st century physics is about correlations. The concept is more general than
quantum theory. A correlation is a function, or a collection of numbers, obtained on some statistical basis,

Proc. of SPIE Vol. 8832 88320W-2


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
and including a set of labels. Let x1 ...xN be the labels, and for discussion imagine they represent space-time
coordinates. Other labels are allowed, of course. The correlation C(x1 ...xN ) can be real-valued or complex: if
complex, it is a pair of reals, using some humanly-defined map to make it complex. A device called a generating
functional Z(J) organizes correlations using the definition
Z Z
Z(J) = dx1 J(x1 )C(x1 ) + dx1 dx2 J(x1 )J(x2 )C(x1 , x2 )
Z
+ ... dx1 dx2 ...dxN J(x1 )J(x2 )...J(xN )C(x1 , x2 ...xN );
δ δ
C(x1 ...xN ) = ... Z(J) |J=0 .
δJ(x1 ) δJ(xN )

Symbol δ/δJ(x) is a functional derivative, needed when label x is continuous, otherwise a partial derivative. We
are expressing nothing profound here beyond the fact that given the correlations we can define Z(J) as a Taylor
series; the symbol |J=0 evaluates all J → 0 after the derivatives are computed. Define a Fourier transform Z̃(φ)
from J(x) → φ(x):
Z
Z̃(φ) = d[φ] Z(J)eiJx φx .

Symbol d[φ] is a functional measure, and Jx φx stands for dN x J(x)φ(x), which is the continuous sum when label
R

x is continuous. The functions for which the calculations can be done at arbitrary dimensionality are Gaussians,
so we will be describing coupled Gaussian correlations, and related tricks, in most applications. Replace symbol
Z̃(φ) → e−S(φ) , which defines S. By definition
Z
δ δ
iN
d[φ] e−S(φ) φ(y1 )...φ(yp ) = ... Z(J)|J=0 = C(y1 ...yp ). (2)
δJ(y1 ) δJ(yp )

All features of the correlations, and in particular their symmetries, so important to physics, must be encoded
in S. If the measure and the fields φ are invariant under a symmetry, and S is invariant, then the correlation
will be invariant. It’s no wonder that S was noticed and given the name “the action.”
In the way we obtained it, Eq. 2 is so general that it is empty mathematical formalism. We don’t advertise it
as our accomplishment, because it actually is mere formalism and well-known in several contexts. The formula
is not specific about how correlations are found or used. Our contribution is to find it unhelpful to say the
path integral of Eq. 1 comes from quantum physics, depends on many quantum mechanical principles and
assumptions, or is the culmination of modern theory. That confuses the long road to an elementary result with
the result. We are drawing a line between the physical model, wherein lies all the information, and the machinery
organizing it, which has none.

2.2 Entanglement
The hyperbolic promotion of “entanglement” as a uniquely quantum mechanical effect is breathtaking. At the
level of calculus I, a student will challenge a presentation based on asserting a general function f = f (x, y, z) =
X(x)Y (y)Z(z), because it is not general. Yet if a finite dimensional quantum wave function ψab 6= ψa ψb its
entanglement is taken to be something serious. Does that make sense?.
We prefer to think in terms of division.7 Division is the inverse of making direct products. Consider a
list of data DI , where I = 1...N . Partition the list into adjacent bins of length n labeled by J = 1...Jmax . If
Jmax = N/n is not an integer, discard data to make all bins the same length. Within each bin label the points
j = 1...n, that is, j = (modJ, n). The result is a map from one to two indices,

DI → DiJ = ΓIiJ DI .

The symbol ΓIiJ represents a linear transformation with an array of numbers (1’s and 0’s) that reproduce what
we just described with words. The transformation is invertible by flattening out DiJ → DI .

Proc. of SPIE Vol. 8832 88320W-3


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
The partition illustrates division of a vector space into direct products of two spaces. Defining two spaces is
an agreement that notation DiJ will represent the same space using the i indices for all the labels J. By this
act of notation, we take a simple pure state on dimension N , and create an entangled state on dimensions n and
N/n. There is nothing invariant about the degree of entanglement. That is largely controlled by our choice of
division, i.e. our coordinate system. The different partitions of the illustration are a tiny subset of the infinite
linear transformations we could make with other coordinate conventions.
We reversed the usual ordering of logic where “small spaces are put into interaction” to make tensor products
on bigger spaces. That approach conceals the arbitrariness of entanglement. In our approach entanglement
depends on the division of data, from which correlations are eventually drawn. It is generally agreed that Nature
does not depend on our coordinate systems. There is a clue in entanglement having few if any invariant properties.
Why bother? We suggest the freedom of division creates organizational opportunities. Any array of number
DiJ can be expressed as a diagonal sum:
X
DiJ = eα
i Λ
(α) α
fJ . (3)
α

This is the singular value decomposition, svd. The factors are orthonormal:
β

i ei = δ
αβ
; fJα fJβ = δ αβ .

(The usual conjugation of inner produces is used if data is complex.) The sum over α extends up to the dimension
of the smaller space, or less. That forces the “coupling” to the larger space to be restricted to a subspace of the
same dimension. This has unexpected consequences whether or not one is considering fundamental microscopic
physics.

2.2.1 The Schrödinger Life Insurance Company


Suppose an insurance company monitors 11,293 variables (label i) toward optimizing its profits. The company
pays off life insurance benefits in a discrete 2-dimensional space labeled J = |1 > (client alive) and J = |0 >
(client dead). The actuaries can linearize the profits and loss around a model with an 11, 293 × 2 array of cost
coefficients $ CiJ . Strangely enough, only two (2) linear combinations |Xα > of the 11,293 input variables are
coupled. The remaining 11,291 variables are orthogonal to the input-output response, and decouple. For business
purposes they can be ignored. The two crucial combinations |Xα > are not coupled simply to the original |0 >,
|1 > basis. They are coupled to two linear combinations αα |0 > +βα |1 >. Thus

C = |X1 > (α1 |0 > +β1 |1 >) + |X2 > (α2 |0 > +β2 |1 >).

Given that actuaries are an intelligent and well-paid group, we suspect they will find a business opportunity in
using linear combinations of alive and dead people.
In the past, however, engineers and physicists were not given the same freedom. They used classical statistical
methods for classical systems, and quantum statistical methods for quantum systems. The use of linear combi-
nations was approved after a system had been set up as a microphysical theory with numerous assumptions and
much reference to Planck’s constant. It may seem reckless, but we propose that the helpful organizational math-
ematics developed for quantum systems has many more uses than being restricted to fundamental microphysics.
If there is a way for an insurance company to make money from it, they will not hesitate to use it.

2.2.2 Physical Repercussions


The repercussions of svd cannot be physically profound, because it is a fact of math, but they do spread deeply
across quantum mechanics as we see it. Each term in the sum of Eq. 3 is decoupled from the others. Discovering
subsystems that decouple nicely is what physicists call progress. Suppose one has a correlation Ca, t , where
a = 1, 2 and t is continuous. The t-dependence is infinite dimensional. Yet on the infinite dimensional space,
only two (2) normalized patterns, f1 (t) and f2 (t), suffice to reproduce the correlations. That might easily
masquerade as a significant physics discovery.

Proc. of SPIE Vol. 8832 88320W-4


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
How do “photons,” “electrons,” etc come in here? Consider the paradox of the quantum interferometer. If
there is one particle of light found in one arm or the other, how does it go both ways to interfere with itself to
make the observed patterns? We don’t buy an empty phrase that “light is both a wave and a particle”. Both the
classical particle picture and the classical wave picture are inadequate because they are obsolete models. The
description of the quantum interferometer needs less of a model than better notation to discuss it.
Let |k1 >, (|k2 >) be a photon in arm 1 or two, respectively. Let |01 >, (|02 >) be no photon in the respective
arms. The interferometer is at least as complicated to need a state of the type
1
|Ψ >= √ (|k1 > |02 > +|k2 > |01 >). (4)
2
We made the ansatz to represent the observation that if |k1 > is detected on one arm, then |02 > is found on
the other, and vice-versa. Notice this is a descriptive statement that does not come from quantum mechanics.
The correlations are those of rocks going one way or the other through the device. The states |kj > produce
translationally invariant spatial dependence, which is not expected of rocks. But (as we will see) one cannot-not
expect translationally invariant correlations over a reasonable length of space in the arm of an interferometer.
It is conventional to designate a quantum interferometer a triumph of quantum theory. The other approach
was used by Schrödinger’s Life, Ltd., and seeks merely to be descriptive. When theory is descriptive, it does not
claim to know in advance what the experimentalist measures, but helps organize data. A real interferometer is
certainly more complicated than Eq.4. We have written Eq. 4 simply to get started on matching an idealized
description with mathematical tools. If the correlations are more complicated: then they will be different.
In Dirac’s textbook on quantum mechanics,9 with remarkably tentative and vague reasoning (for Dirac), a
rule of composing subsystems by direct products is motivated on the basis that probabilities multiply: at least
sometimes. The rule to build up interacting systems from microscopic subsystems needs to know the subsystems
to begin with. We do not know the fundamental subsystems of the Universe to this day. We propose a simpler
explanation of how correlations are classified and manipulated. Given a correlation, it is probably worth some
effort to find a division which makes it appear simple. The successful divisions are nice coordinate conventions
physicists remember and teach each other how to use. Like the people described by αi |0 > +βi |1 >, the
subsystems may or may not partake of “physical reality”. And it is a distraction to dwell on what that term
means.
The facts are that a great deal of experimental data is neatly expressed with the theoretical division into
“photons.” Using photons represents the discovery of a good coordinate system and a good model. Photons are
the factors in division of certain information that tends to be helpful, sometimes.

2.2.3 Strict Correlation: Calcite


Optics was already highly developed by the 1840’s. The early workers in optics picked up the clues from double
refraction and came to correct conclusions. In the 21st century we ought to understand the behavior in ways
that are even more simple, not more complicated.
~ x, t). Let Aax represent the field with compo-
The data for polarization indicates light is a vector field A(~
nent (polarization direction) α at point ~x. The notation treats the space and polarization indices as matrix
components. The matrix Aax is not square, but has dimensions 2 × ∞. Apply svd:
X
Aax = χαaΛ
(α) α
φx .
α

The sum over α runs over α = 1, 2 for transverse spin-1 vectors,


(1) (1)
Use the information of svd to revisit double refraction. The first term in the sum is χa Λ(1) φx . Suppose
(1)
χa points in the x direction. Selecting this polarization with a filter will find the space wave function φ(1) and
no other. Conversely, selecting the wave function φ(1) with a spatial filter will find the polarization wave function
(1)
χa and no other. The components of space and polarization are strictly correlated.

Proc. of SPIE Vol. 8832 88320W-5


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
There is nothing mysterious about this, and Fresnel explained the double-refraction of light in calcite before
1820. The facts of svd do not predict what spatial wave functions are separated in the crystal. By a near
translational symmetry (see the next Section) they will be close to plane waves. We need a more complete
theory to know the polarizations are transverse in free space, and related to a dielectric tensor in the crystal.
That confirms our rule: when there is more information from a model, there is more information in physics. In
comparison, the separation of the polarization states of light into strictly correlated components has no physics
in it: it is a general mathematical fact.
Compare this picture with the presentation of the Stern-Gerlach experiment in textbooks.10 Instead of being
trivial, it is presented as the prototype of a mysterious and inexplicable micro-world. While new quantum
postulates are necessary, in that older approach, they are not needed in ours.
The discussion highlights more than two quantum principles that can be dropped, at least in our approach.
One postulate says that if you measure a quantity, you will get an eigenvalue of the operator. Another part of
the postulate says the resulting state is an eigenvector of the operator. By putting a paper card in front of the
top beam in a Stern-Gerlach machine, the experimenter selects a spatial state |φ(1) >. Strict correlation gives
him a polarization state |χ(1) >. Without a theory and other experiments, the experimenter does not know in
advance what |χ(1) > will be. The information he is measuring a “spin operator” is not given in advance. (In
fact, the two states separated in calcite don’t happen to measure “spin”.) We need an operator; we notice |χ(1) >
is trivially the eigenvector of the operator |χ(1) >< χ(1) |. The other beam’s polarization is an eigenvector of
another operator |χ(2) >< χ(2) |. These operators are Hermitian, and by construction span the possibilities.
A relation of svd to quantum measurement has been recognized on and off for decades.11 If the model of
waves with polarization is given, the time evolution can be solved.12 Things that are solved with equations
should not be profound. We conclude that the physical information (if any) is in the model, not in general
mathematical facts.

2.3 The Born Rule


2.3.1 Quantum Probability From a Hidden-Variable Map
Since the work of Bell13 it has been popular to say that “quantum probability is incompatible with hidden
variables.” That is not what Bell actually showed. Bell’s theorem showed that a certain correlation calculated
using quantum probability could not be reproduced by probability defined with distributions. This should not
have been a surprise.
It is certainly possible for quantum probability to sometimes be equivalent to probability defined by distri-
butions. The breakdown of the converse should not have been a surprise, because the quantum calculations
depend on the ordering of operators used in the calculations. In 1989 Werner14 finally confronted the question
of exactly when a quantum system might be described with distributions. Just in case the reader is thinking
that ψ(x)∗ ψ(x) is the very definition of the quantum “probability density to find the particle at x,” that is a
false piece of beginning pedagogy. From the ψ ∗ ψ rule one cannot derive the Born rule. From the Born rule one
derives the ψ ∗ ψ rule for a hypothetical normalized wave function isolated in a domain dx → 0. The hypothetical
localized state has never been observed, would expand explosively if prepared for an instant, and it is not a basis
of modern discussion.
Quantum mechanics actually has hidden variables, and we suggest they cannot but contribute to quantum
probability. A simple map from classical data to quantum probability illustrates the ideas. Let |D > be a big
vector we call “data”. By means which are quite arbitrary (as discussed) it is partitioned in a collection of smaller
vectors DiJ =< iJ|D > with names J = 1...Jmax and components i = 1...imax . But in this Section we tentatively
interpret J as labeling the sample history taken from some (deterministic or random) process. The other index
i is interpreted as describing “objects”. The sample space and object space are tentative. Different divisions
of data can mix them. For convenience the record is normalized < D|D >= 1 in the usual way, removing one
number set aside.
Now we seek a notion of orderliness or physical regularity. We will expand the vectors in an orthonormal
basis set {|eα >}, where eα
i =< i|α >, and seek some form of statistical repetition. The basis matters, so which

Proc. of SPIE Vol. 8832 88320W-6


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
basis is used? There are two preferred bases from svd:

|Di = |eα i Λα |sα i . (5)

Here
X
DD† = |eα > (Λα )2 < eα |; (6)
α
X
D† D = |sα > (Λα )2 < sα |.
α

The singular values Λα are invariant when the data is transformed by independent orthogonal (complex unitary
) transformations on the object and sample spaces.
We turn to an instructive illustration of mapping data into probabilities.
2.3.2 The Probabilistic Quantum Map
Example: Suppose the data consists of integer numbers of “objects” called |zebra >, |giraf f e > or |other >.
This is classical mutually exclusive (me , alive or dead) type of data: by existing in different spaces, <
zebra|giraf f e >= 0. Sufficiently fine sampling will produce samples which are either 1 or 0. Typical data
is then

|D >= (|zebra >, |other >, |giraf f e >, |zebra >, |giraf f e >,
. . . |zebra >).

The expression only makes sense if these me objects are normalized, < zebra|zebra >= 1, and so on, else the
normalization would conflict with the number of zebras. Expand in the natural basis we started with,

|D >=|zebra > (1, 0, 0, 1, 0, . . . 1)


+ |giraf f e > (0, 0, 1, 0, 1, . . . 0) + .... (7)

The diagonal form of svd has appeared, up to a normalization. Whenever data consists of disjoint me objects,
one can show those same objects are automatically the svd factors. The fact of strict correlation comes with
projecting onto one object such as |zebra > and producing its sample vector, which is automatically orthogonal
to all the other sample vectors:

< zebra|D >= (1, 0, 0, 1, 0, . . . 1).

Conversely, selecting one of the me sampling histories automatically selects a unique object. These are features
of classical “events.”


To reach the svd form implies samples that are normalized: sα | sβ = δ αβ . Let Nzebra be the total number
of zebras observed. Let Ntot be the total of zebras and giraffes. Remember that we normalized our data. Then

(1, 0, 0, 1, 0, . . . 1) →
p p
Nzebra /Ntot (1, 0, 0, 1, 0, . . . 1)/ Ntot /Nzebra
p
= Nzebra /Ntot |szebra > . (8)

Once normalized we can read off the singular values:


p
|D >= Nzebra /Ntot |szebra > |zebra >
q
+ Ngiraf f e /Ntot |sgiraf f e > |giraf f e > .

Next suppose there is a unitary transformation of our data on either object of sample or both spaces - but
not mixing them. The result will involve linear combinations of the form α|zebra > +β|giraf f e >, which is

Proc. of SPIE Vol. 8832 88320W-7


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
classically “taboo.” For better or worse, we cannot stop linear transformations from being used or being useful.
They represent a notational freedom we can’t stop people from using. Whatever the coordinate system, we can
construct the svd factors and singular values as invariants under rotations of the coordinates. Anything that is
measured has to be related via invariants.
Not all data is mutually exclusive. It is standard in physics to project data into a basis, invariably defined
by representations of some group. Repeating the exercise above with projections finds the sample space vectors
cease to have integer components. They are not expected to be orthogonal, and even with “classical” data don’t
need to have an easy interpretation as actual events. That makes no difference to the procedure.
In many cases (and always in physics) we are actually forced to suppress some detail by summing over
unwanted or unrecorded details of the sampling history. We use that observation in constructing the density
matrix ρobject of the object system:
X
ρobject = trs (|D >< D|) = |eα > (Λα )2 < eα |.
α

The form ρobject we summed over the sample history (symbol trs ). This identity now defines the me |objectα >=
|eα >. By construction, whenever me data is used, we have a convenient invariant formula for the probability of
finding such an object:

P (|eα > |ρobject ) =< eα |ρobject |eα >= tr(ρobject |eα >< eα |).

The symbol P (|eα > |ρobject ) is read as the probability of |eα > given ρobject , and exactly coincides with counting
numbers: thus
p
P (|zebra > |ρ) = ( Nzebra /Ntot )2 = Nzebra /Ntot .

Notice the right hand side defines probability by frequencies, just as experiments do. In general form the
probability P to get an observable < Â > is

P = tr(ρÂ)/tr(ρ). (9)

This is the Born rule in its general form.7

2.3.3 Wave Functions


The general Born rule is the heart of quantum probability. From it comes dependence of outcomes on operator
ordering, the rules for correlations of EPR type, etc. We find that expressing the rules of quantum probability
with wave functions is not general. When the density matrix is rank-1, then ρ = |ψ >< ψ|, and a description by
a wave function |ψ > is equivalent: otherwise not.
In Section 5 we characterize systems by their entropy. The most orderly systems have zero entropy and
are described by wave functions. This is convenient when it can be arranged. But in just the same way that
the “data” of a particular system will determine whether a simple description is appropriate, we don’t find it
helpful to build a quantum description on the basis of wave functions. Density matrices are more general and
the appropriate tool.

2.3.4 Classical versus Non-Classical Probability


The discussion shows that quantum probability needs no cooperation from special attributes of “quantum ob-
jects” to be used consistently. It is un-Quantum. What is called quantum probability uses non-me equivalences
classes to create an extension of classical probability – not a contradiction with classical probability. The quan-
tum equivalence classes are defined by orthonormal subspaces. This is much different from putting every number
in a list in its own me class, and subdividing every slot into a large number of me boxes, towards making dis-
tributions from data. As discussed in Ref.,7 the actual process of sampling distributions on large spaces is not
practical. Large space are exponentially large. We believe the impracticality of classical probability definitions
must force a description based on quantum probability regardless of the physical world.

Proc. of SPIE Vol. 8832 88320W-8


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
The discussion was tentative about which spaces in a quantum description should be “objects” or “sample
history.” When counting zebras and giraffes there exists complete information about the space. When it comes to
the physical world we never have complete information. This is important. In early days photons and electrons
were deemed fundamental entities of the universe. Free quantum field theory emulates their properties well,
up to a certain approximation. For more fundamental purposes theory of the physical photon or electron as
eigenstates of free field theory is naive and obsolete. By now it is known that Nature is at least as complicated
as interacting quantum field theory. Going from an interacting theory to a description on a subspace requires
“tracing out” (summing over) the diagonal undetected elements of the full density matrix. The full density matrix
of modern interacting theories has every field interacting with every other...including the three generations of
quarks, leptons, and gauge bosons. That is a very noisy process at the microscopic level. The objects divided out
as “photons” come from tracing out every field except the electromagnetic field. That process involves a huge
sampling history. We suggest that the Born rule originates in summing correlations over the undetected hidden
quantum variables modern quantum physics is known to have. That is the sense in which we earlier described
quantum mechanics as a “macro-system.”
Some may find this provocative. Shift the discussion from photons to the sort of industrially-applied spin
-1/2 systems observed daily in hospital magnetic resonance imaging. Can one describe the observed electrons
spinning in a magnetic field with a 2-component wave function, and the famous Bloch equations? No, their
degree of polarization is not conserved. The technicians use a 2 × 2 density matrix. Does that density matrix
describe a fundamental electron found in the problem? The formalism might, but not in this application. Every
cubic millimeter measured for making images sums over a vast horde of electrons. The hidden variables are the
degrees of freedom in the molecules that are not directly detected, but which are indirectly detected by their
effects on the density matrix. And more. Whether or not the quantum dynamics of spin-1/2 electrons were
perfectly classical it would make no difference. Exactly the same correlations would lead to just the same density
matrix.

2.4 Quantum Spin


We come to model features. Physical models are described by continuum Lagrangian or Hamiltonian densities.
The old approach contains ad-hoc postulates before even getting to a model. Here we discuss how the La-
grangian and transformation properties provide full detail about every dynamical fact without needing further
information.5
Quantum Spin is a misidentified feature of classical physics. Given a continuum field φ(x, t), and a Lagrangian
density L(φ, ∂µ φ), where ∂µ = ∂/∂xµ , the formula π(x) = ∂L/∂(∂φ/∂t) defines the conjugate momenta. Suppose
we make a particular change of variables, φ(x) → φa (x), where a are continuous parameters. The infinitesimal
transformation near a = 0 is φa → φ+a∂φ/∂a|a=0 . The projection of all the field momenta onto the infinitesimal
rate is
Z
Pa = d3 x π(x)∂φa (x)/∂a.

There is a sum over all the fields that transform. This is Noether’s Theorem, which finds the total, particular
collective momentum of the field, conjugate to the particular transformation specified, as a basic element of
classical Lagrangian mechanics. Noether’s Theorem is usually developed as a local conservation law ∂µ J µ = 0
with prior assumptions of relativistic Lorentz invariant field theory. The steps we are using coincide while being
slightly more general.
~ x), which existed in electrodynamics long before quantum mechanics.
Consider the classical vector potential A(~
Under a rotation represented by a 3 × 3 matrix R, (RRT = 1), the field rotates by the rule

~ x) → A(~
A(~ ~ x)0 = R · A(R
~ −1 · ~x). (10)

~ transform by basically the same relation: verifying that R−1


The spatial label ~x and the orientation labels of A
operates on the argument is a matter of drawing pictures. Every rotation can be represented by Cartesian angles
~ whereby R = exp(i P θk jk ). An infinitesimal 3 × 3 rotation of the vector part is A0 ∼ A + θ~ × A.
θ, ~ Here × is
k

Proc. of SPIE Vol. 8832 88320W-9


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
the cross product, so that (θ~ × A)
~ a = abc θb Ac . From this the rotation generators jk have the matrix elements
(jk )ij = −iijk . These are the outcomes of classical geometry.
The Lagrangian density for the electromagnetic field is

~ 2 /2 − B
L=E ~˙ 2 /2 − (∇
~ 2 = (A) ~ × A)
~ 2 /2. (11)

The canonical momenta πj = Ȧj = −Ej , where E stands for the electric field. This is found in many textbooks.
Noether’s classical theorem predicts two parts for the angular momentum of the electromagnetic field, represented
by vector potential A.~ The two parts come because Eq. 10 has two infinitesimal changes from the rotation of
the polarization (just reviewed) and the related infinitesimal shift of the position, Ai → A0i = Ai − (θ~ · ~x × ∇)A
~ i.
The result for the total angular momentum L ~ is
Z
~ j = d3 x Ai jik Ek + Ei xa abj ∂Ai /∂xb .
(L) (12)

The first term is the spin angular momentum in the gauge A0 = 0. It has the form ~q × ~π where A ~ → ~q and
~π = −E, ~ point by point in x. This term uses no spatial gradients. The angular momentum in the tumbling
polarization variables has a perfectly Newtonian form (point by point) because the kinetic energy is quadratic in
~˙ like Newton’s. The other term involves the operator ~x × ∇,
A, ~ and represents the orbital angular momentum.
The linear momentum operator ∇ ~ comes from classical physics. The orbital angular momentum operator ~x × ∇ ~
comes from classical physics of fields sloshing around in space. Both spin and orbital angular momentum are
basic feature of classical physics.
How does this approach relate to the old quantum tradition? The old tradition makes physical postulates to
explain things that can be derived by modern theory. In Eq. 12 the spin-part is a “sandwich” of some matrix
elements abk between the momentum −Ei and the field Ai . So long as a general change δA ∼ A under the
variable transformation, there will always be one or other matrix in a sandwich. The matrix is also the generator
of the transformation under discussion. For rotations of 3-vectors our calculations have found the generators are

(jk )ij = −iijk , (13)

which have the algebra

[ji , jj ] = iijk jk . (14)

We don’t need to make this a postulate because we’ve just obtained it.
Some will be expecting to see a “correct” commutation relation involving Planck’s constant. Define a symbol
Jk ≡ ~jk . Then

~2 [ji , jj ] = i~2 ijk jk ; (15)


[~ji , ~jj ] = i~ijk ~jk ;
[Ji , Jj ] = i~ijk Jk . (16)

We are not going to make this a postulate because we’ve just obtained it from our definitions. Obviously, ~2
cancels out on both sides of Eq. 15, which does not depend on ~ overall. If one were given Eq. 16 as a inexplicable
postulate, it would naturally be much harder to discover that ~ cancels out.

2.4.1 Discusson
The fact that spin is a derived attribute of classical polarization ought to be better known. What explains the
gaps that keeps it a secret? Lie groups were known for decades before quantum theory, but not generally known
by physicists. In the first years of quantum mechanics physicists misidentified many consequences of mathematics
as physics discoveries. The momenta of the electromagnetic field are also somewhat subtle. The canonical result
disagrees with the expected angular momentum density going like ~x × (E ~ × B).
~ When ~x × (E~ × B)~ is asserted

Proc. of SPIE Vol. 8832 88320W-10


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
as “the final answer,” the angular momentum density has no spin part, leading to great confusion, and many
contradictions. The upshot was that spin was misunderstood. Every form of spin got blamed on quantum
mechanics. .
The problem began with asserting Poynting’s formula E ~ ×B~ is the translational momentum density. That is
essentially a guess, based on consistent divergence identities. In comparison Lagrangian physics defines momen-
tum, and definitely does not give Poynting’s result for the density. Poynting’s methods cannot actually determine
the correct formula, because they allow arbitrary curls to be added to divergence identities. An inexplicable pref-
erence (tradition?) for Poynting’s expression seems to persist because it is gauge invariant. Actually the vector
potential is completely well defined when the gauge is fixed. We have never seen cogent logic for replacing the
Noether currents by Poynting’s guesses, which (as remarked) cause concept difficulties.
Since the definitions are gauge-dependent, there is no such thing as the momentum or angular momentum
density of the electromagnetic field. Lagrangian physics will give you the momentum or angular momentum
density of your own coordinates of the field. No new information from quantum mechanics is involved.

2.4.2 Quantum Substitution Rules


Some might be looking for the “weirdness” of quantum mechanics to enter with spin-1/2. No. All quantum
mechanical spin formulas present facts a classical field theory in an operator-based notation.
The Lagrangian of Schrödinger’s theory of a scalar function, no polarization, is

LSE = iψ ∗ (x)ψ̇(x) − ψ(x)Ĥψ(x).

Symbol Ĥ is the Hamiltonian operator.The canonical momentum πψ (x) = ∂LSE /∂ ψ̇(x) = iψ ∗ (x). (Later we
will see how the complex numbers themselves originate in coordinate conventions.) Let us derive the collective
(total) momentum conjugate to spatial translations. Under a translation of coordinates

~x → ~x + ~a;
ψ(~x) → ψa (~x) = ψ(~x − ~a).

The rate of change with respect to parameters ~a is ∂ψa (~x)|~a=0 = −∂ψ/∂~x. Noether predicts the classical
translational momentum is
Z
∂ψ ~ >.
p~a = d3 x πψ (x)(− ) =< ψ| − i∇|ψ
∂~x

Here is the momentum operator∗ −i∇.~ Once we derive this on our own, it is not helpful to call it an independent
postulate.
Under a rotation by Cartesian angles θ~ a scalar function transforms ψ(~x) → ψθ (~x) = ψ(~x − θ~ × ~x). The
classical angular momentum is
Z
~ ∂ψ ~ >.
L = d3 x πψ (x)( ) =< ψ| − i~x × ∇|ψ
∂ θ~
~ “is postulated to be the angular momentum operator.”
There is nothing to gain by declaring that “−i~x × ∇
~ → ψ+θ× ψ.
A matter wave function transforming like a 3-vector, spin-1, will have additional transformation ψ ~
~ ~ ~
That will lead to an addition angular momentum from the polarization Lspin−1 =< ψ|j|ψ >, where j is given
by Eq. 13 and the inner product sums over the polarization components, just as with the vector field.
Elsewhere5 we have considered these facts using Poisson bracket notation. Any sandwich of the form <
ψ|Q̂|ψ >→ Q, where Q̂ is Hermitian operator, is a map from canonical coordinates πψ , ψ into a number. Poisson
brackets are an efficient way to determine transformation properties of such maps. If the Poisson brackets of

An earlier convention multiplied the Lagrangian by ~. Since it cancels out, there’s no point in doing it, explaining
why ~ is absent.

Proc. of SPIE Vol. 8832 88320W-11


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
two quantities P, Q equals 1, they are candidates for being canonically conjugate, subject to checking the rest
of the brackets. Asking for the numbers Q to represent a classical collective coordinate or momentum of the
system puts consistency relations on the operators Q̂ used in the sandwich. The operator substitution rules
replacing Poisson brackets with commutators follows.5 Since we find them by doing algebra, there is no logic to
calling those rules foundation postulates. They are unrelated to the uses of quantum probability, or the value of
Planck’s constant: None of that entered our discussion† .
Hopefully this puts the proper perspective on spin-1/2. It was a great discovery about the physical world to
find matter requires a model of waves. Once we have that information, classifying the transformation properties
will follow. Angular momentum is particularly important, because symmetry under rotations of a Lagrangian
theory will not permit one class (representation of the rotation group) to transform to another class. The fact
that complex spin-1/2 representations of the rotation group (technically SU (2)) exist comes from math. One
needs to do experiments to discover whether Nature uses those fields. Experiments led to a model, containing
the information that fields with spinor polarization were observed. After that, the formulas for spin-1/2 and all
other spins represent the facts of classical physics of polarized waves.

3. QUANTUM DYNAMICS: FLOWS


We come to dynamics. By dynamics we mean equations for time evolution. Except for classifying systems under
symmetries, physics actually predicts very little but what’s found in models of time evolution.
Making no postulates but the proposal that little is known, lets us choose to describe a system that three
time-dependent position variables transforming like a 3-vector ~q(t). We might be considering a “classical particle”
emerging from a propagating quantum system. We might be considering something more fundamental: we leave
the discussion general.
The velocity ~q˙ = d~q/dt is something that can be observed when the description applies. Dynamical system
theory has a very general description of the velocity as a flow, namely a set of ordinary differential equations:
q̇i = fi (q(t), t). (17)
Differential equations of any order fit this pattern for an appropriate number of variables qi .
We cannot assume the dynamics will be Newtonian, which was the primary bias at the dawn of quantum
theory. It would be biased even to assume the dynamics are Hamiltonian. Hamiltonian equations are extremely
special. For one thing, in assuming a flow we have not developed a notion of conjugate momenta pi . We treat
Eq. 17 only as a description of particular situations, obtained from “data.”
Yet many situations in physics assume an underlying phase space exists. We approach the question by using
what is observable to discover whether there may exist an appropriate phase space. Every three-dimensional
vector field can be decomposed into a gradient and a remainder:
~f = (∇Ω
~ − A).
~ (18)
By using 4 quantities to represent 3, the ansatz is not unique until resolved by one constraint. The usual Helmholz
theorem‡ assumes A ~=∇ ~ ×W ~ , so that ∇~ ·A
~ = 0, or A~→∇ ~ ×W ~ . We leave the convention open, as explained
further below.
Towards finding a notion of momentum p, recall that momenta are usually eliminated to make autonomous
equations which are second-order in time. We calculate the acceleration ~a, which is:
d~v ∂~v ~ v;
~a = = + (~v · ∇)~
dt ∂t
∂fi ∂fi
ai = + fj . (19)
∂t ∂qj
Now there are two distinct cases:

History went otherwise. Heisenberg did not realize that repeating the algebra of Poisson brackets in the notation of
commutators was bound to reproduce classical Hamiltonian dynamics

For n arbitrary number of variables there is the Hodge decomposition, expressed with differential forms.

Proc. of SPIE Vol. 8832 88320W-12


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
~ → 0. By algebra
• Assume gradient flow, A

1 ∂2Ω 1 ∂Ω ∂ 2 Ω
ai = +( 2) .
m ∂qi ∂t m ∂qj ∂qi ∂qj
The right hand side is an exact derivative:
~ q ),
m~a = −∇φ(~ (20)
" #
∂Ω ~ 2
(∇Ω)
where φ = − + . (21)
∂t 2m

Given φ from data will predict Ω from solving Eq. 21 as a partial differential equation:
X 1 ∂Ω 2 ∂Ω
( ) + φ(q) + = 0. (22)
j
2 ∂q j ∂t

This is the Hamilton-Jacobi equation of a Newtonian system, with Hamiltonian

~2 /2m + φ(~
Hef f = p q , t).

~=
• Gradient flow is exceptional. When A 6 0 the acceleration is

∂2Ω ∂Ai ∂ ∂Ω ∂Ai


q¨i = q̇j − q̇j + − . (23)
∂qi ∂qj ∂qj ∂t ∂qi ∂t
Write
∂2Ω ∂ ∂Ω ∂Aj
= ( − Aj ) + ,
∂qi ∂qj ∂qi ∂qj ∂qi
to find
∂φ ∂Ai ∂Aj ∂Ai
q̈i = − − + q̇j ( − ), (24)
∂qi ∂t ∂qj ∂qj
where
1 X ∂Ω ∂Ω
φ=− ( − A j )2 + . (25)
2 j ∂qj ∂t

This is a different Hamilton-Jacobi equation. In three dimensions Eq. 24 becomes

~a = E q˙ × B,
~ +~ ~ (26)
~
where E ~ − ∂A ;
~ = −∇φ ~ =∇
B ~ × A.
~
∂t
~ and B
Eq. 26 is called the Lorentz force, where E ~ are prescribed electric and magnetic fields, respectively. Evaluating
Eq. 25 as a Hamilton-Jacobi equation, the Hamiltonian coming from Eq. 25 is
X
H= (pi − Ai )2 /2 + φ. (27)
i

It is the generalized Hamiltonian of a charged system on an arbitrary number of dimensions. For 3-vector ~
q the
Hamiltonian is

H = (~ ~ 2 /2 + φ.
p − A) (28)

Such systems are classical but not Newtonian. The momentum conjugate to ~
q from Eq. 28 is

q˙ = ∂H/∂~
~ p;
p
~=~ ˙ ~
q + A. (29)

We are not including a mass parameter. It simply sets the scale of the Hamiltonian, i.e. the units of time.

Proc. of SPIE Vol. 8832 88320W-13


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Remarkably we have discovered Hamiltonian behavior on a very general basis. There was a twist to the
method. If we had assumed a number of coordinates qi , i = 1...N , and particular pre-determined momenta
pi ...pN , the flow of a generic system η̇ = (q̇1 ...q̇N , ṗ1 ...ṗN ) could not generally be Hamiltonian. Hamilton’s
equations are
∂H ∂H
q̇i = ; ṗi = − . (30)
∂pi ∂qi
Differentiation gives

∂ q̇i /∂qi + ∂ ṗi /∂pi = 0, i = 1...N. (31)

There is no sum over i. Eq. 31 shows that Hamiltonian systems with 2N q’s and p’s have N conservation
laws, which is one of Poincarè’s theorems. It is extraordinary for a dynamical system to have those conservation
laws, which (one can show) are equivalent to requiring Hamilton’s equations. Yet we see that any flow can be
embedded in the extremely special Hamiltonian systems. Discovering this degree of order hidden in generic flows
is very unexpected.
Note we have not shown that one Hamiltonian function governs all flows, always. The flow of one physics
experiment may lead to a different Hamiltonian than another experiment. If the description happens to be
second order in the time derivatives of q, (or if that approximation suffices), we have found emergent classical
Hamiltonian behavior.
Pause to Assess: Our perspective has shifted again. Recall the quantum Eherenfest relations for < ~x >
(t) =< ψ(t)i|~x|ψ(t) >. When first discovered physicists assumed particles existed and assumed Newton’s laws
in Hamiltonian form. Eherenfest showed a way quantum waves might be compatible. While welcome early, the
arguments lack controlled error estimates and do not pass the quality-control standard of 21st century physics.
Our result suggests a way that the flow of qi (t) might be such that classical Hamiltonian physics cannot-not be
obtained. There are consequences. The parameters of the macroscopic theory do not need to be related to those
of the microscopic theory in any naive way. In another paper6 we have recommended abandoning macroscopic
definitions and measurements of physical constants, in particular “mass,” when determining the constants of
more fundamental quantum physics. Use quantum definitions for quantum theory.

3.0.3 Emergent un-Quantum Dynamics


The last Section seems to create a contradiction. We are finding classical Hamiltonian dynamics. What about
quantum systems? Classical Hamiltonian dynamics seems to contradict quantum time evolution. Textbooks
say quantum theory has a “new and uniquely quantum mechanical” structure replacing the previous dynamical
framework.
Consider a finite-dimensional quantum system with wave-function ψi (t) = (ψ1 (t), ψ2 (t)...ψN (t)). The real
part of the wave function defines the flow of a set of coordinates qi ∝ Re(ψi ). Having shown that all flows
have an underlying Hamiltonian representation, the argument of the previous section applies. There must be a
Hamiltonian. Much of the following material comes from Refs.5, 7

3.1 The Dynamical Quantum Map


Hamilton’s Equations (Eq.30) couple the time evolution of q’s and p’s. The coupling is conveniently represented
by a 2N × 2N matrix of 1’s and 0’s denoted J:
 
0N ×N 1N ×N
J= . (32)
−1N ×N 0N ×N

Defining the 2N dimensional multiplet η = (q1 ...qN , p1 ...pN ), Hamilton’s equations are

∂H
η̇i = Jij . (33)
∂ηj

Proc. of SPIE Vol. 8832 88320W-14


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
The matrix is invariant under transformations J → SJS T = J where S ∈ Sp(2N ), the symplectic group of 2N
dimensions. Hence Hamilton’s equations are invariant; it is the group of canonical transformations. J is often
called the “symplectic metric.”
To decouple the system go to coordinates where J is diagonal. Since J is antisymmetric, the transformation
goes from real to complex numbers:
   
q1 ψ1
 q2   ψ2 
   
 ...   ... 
η=  → Ψ =  ψ1∗ 
   
 p1   ∗ 
 p2   ψ2 
... ...

Calculation gives

Ψ = Uη; UU † = 1.

Then
 
i 1N ×N 0
UJU † = − .
0 −i 1N ×N .

The diagonalizing transformation is


 
1 1N ×N i 1N ×N
U=√ ,
2 1N ×N −i 1N ×N

The transformation to complex variables is summarized by



ψi = (qi + ipi )/ 2. (34)

We call Eq. 34 the “quantum map”.


In the new coordinates H(qi , pi ) → H(ψ, ψ ∗ ). Consistency will find that H is real. Diagonalization simplifies
Hamilton’s equations (Eq. 30), which become

∂Ψ ∂H(Ψ, Ψ∗ )
i = . (35)
∂t ∂Ψ∗
Since it is an important point we show the algebra for one degree of freedom and its conjugate momentum. We
are given
∂H ∂H
q̇ = ; ṗ = − .
∂p ∂q
Combine two real numbers into one complex one:

ψ = (q + ip)/ 2. (36)

It explains how complex numbers came to be essential in quantum theory. Compute


√ ∂H ∂H √
ψ̇ = (q̇ + iṗ)/ 2 = ( −i )/ 2. (37)
∂p ∂q
The chain rule gives
∂ ∂ √ ∂
−i = 2 ∗,
∂p ∂q ∂ψ

Proc. of SPIE Vol. 8832 88320W-15


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
and then Hamilton’s equations are
∂H
iψ̇ = .
∂ψ ∗

Continuing: When H(q, p) is bilinear then H(Ψ, Ψ∗ ) is bilinear. Quantum mechanics always uses such Hamil-
tonian functions or models. It is not always recognized, due to language developed for Hamiltonian operators.
The bilinear form is simply parameterized by certain operators matrix elements. The general form is

H(Ψ, Ψ∗ ) = Ψ∗ · Ω̂ · Ψ. (38)

Here Ω̂ is the coordinate-free notation for the operator, more commonly designated a symbol Ĥ. Eq. 35 and 38
give
∂Ψ
i = Ω̂ · Ψ. (39)
∂t
This is Schrödinger’s equation Ĥ|ψ >= i|ψ˙ >, which is nothing more than Hamilton’s equation in complex
notation. We mix the symbols deliberately, while we prefer symbol Ω̂ to Ĥ. In the form above Ω̂ is called the
frequency operator. Its eigenvalues are frequencies. The eigenvalues equation is

Ω̂ · ψn = ωn ψn .

The solutions ψn are known as normal modes. The expansion in normal modes solves the dynamics:
X
ψ(t) = ψn e−iωn t . (40)
n

In our approach this does not come from any new principles: None are needed.
3.1.1 Linear Theories
Linear equations of motion are generated by Hamiltonians that are bilinear in qi , pi :
1
H(q, p) = η T Hη η;
 2 
1 hqq hqp
Hη = (41)
2 hTqp hpp

Matrix multiplication is implied, and hqq , hqp ...etc are N × N arrays of constant parameters. Any linear terms
like αq + βp have already been removed by translating coordinates. We have no commitment to the bilinear
form, which is cited to reproduce what is familiar. It is not our responsibility to defend the model.
More familiar matrix notation is§

hqq = K; hpp = M −1 hqp = −ΓT M.

Complete the square:


1 1
H(q, p) = pM −1 p + qKq + qM Γp + pΓT M q,
2 2
1
= (p − A(q))M −1 (p − A(q)) + V;
2
1
where A(q) = Γq; V = q(K − ΓT M −1 Γ)q.
2
Here is the same general form as Eq. 27, discovered to be hidden in first-order flows.
§
There’s no worry of an inverse existing in symbol M −1 . It implies the inverse on the space where an inverse exists,
i.e. the pseudoinverse.

Proc. of SPIE Vol. 8832 88320W-16


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
3.1.2 Discussion
We collect several topics for discussion:

• Unitarity: Eq. 40 is the solution to time evolution. In coordinate-free notation it is ψ(t) = e−iΩ̂t ·ψ(0). The
textbook expression writes e−iΩ̂t = e−iĤt/~ , where Ĥ = ~Ω, and ~ cancels out. Unitarity of time evolution
is automatic when Ω̂ has real eigenvalues. Otherwise a more general real dynamical system will have
eigenvalues which are complex conjugate pairs. Hermiticity of the Hamiltonian is equivalent to choosing
models that do not have exponentially growing and decreasing oscillations. This seems to be good judgment
more than a principle. As noticed in Ref.15 , it is better to classify Hermitian systems by their eigenvalues
than by the self-adjoint matrix rule Ĥ = Ĥ † . If the self-adjoint rule is satisfied, then Hermeticity follows,
but if general coordinate transformations are applied to a Hermitian operator, the self-adjoint property can
easily fail. One of the main reasons for insisting that quantum coordinate transformations be unitary is to
maintain the notation of Eq. 39 form-unchanged. Now that we realize complex notation is optional, that is
not compelling. When more general symplectic transformations are applied, Hamilton’s equations are form-
unchanged, showing that the full symmetries of quantum dynamics are Sp(2N ). Unitary transformations
are also the largest group maintaining the values of quantum probability: this is compelling.
• “Non-Linear Operators” : There is confusion about quantum Hamiltonians, which tends to draw questions
whether the step ∂H/∂ψ ∗ → Ω̂ψ imposes some kind of restriction on our part. The step is as general as
quantum dynamics. That is because conventional quantum dynamics uses a linear model. It is not our
burden of proof here to justify the model, or explain why it might work. Our dynamical formalism has the
generality to explore genuinely non-linear quantum dynamics, if desired.
Some physicists will cite quantum Hamiltonians as being “non-linear.” A common misunderstanding says
that “only quantum harmonic oscillators have quadratic Hamiltonians. The interesting systems have non-
linear Hamiltonians.” Mixups come from different uses of the word “Hamiltonian.” In physics and in
our usage H(η) → H(Ψ, Ψ∗ ) is that function on the phase space variables η by which η̇ = ∂H/∂η. To
repeat, H(ψ, ψ ∗ ) =< ψ|Ĥ|ψ > is a number with a classical dynamical meaning. In the other usage, Ĥ
refers to the operator, namely a matrix of parameters, used to compute the number. In wave theories the
matrix is infinite dimensional, and conventionally built out of powers of −i∇~ and ~x. We are not concerned
with guessing the Hamiltonian’s parameters. Our results are the same whether or not the operator Ĥ
is a complicated function of other operators. No matter how many operators are piled together in the
formula for Ĥ, it is still a linear operator acting in the linear differential equations iψ̇ = Ĥψ. (And
this remains true in relativistic quantum field theory. A Schrödinger picture exists, satisfying the linear
Hamilton-Schrödinger Eq. 39.)
• Hamiltonian Prediction Recipes: The historical recipes to “predict” a system Hamiltonians are quaint, but
irrelevant. Physics has evolved to ignore the classical limit and select Hamiltonian parameters by fitting
data. For some reason this is kept secret. It is also irrelevant that the Heisenberg equations of motion
appear very challenging when operators are composed into non-linear combinations of operators. The linear
system’s dynamics is trivial in the mathematical sense, solved by Eq. 40.
• Counting Degrees of Freedom: Confusion exists in counting the number of dynamical degrees of freedom,
dof . We count each generalized coordinate as one dof . The dimension of ηi labels (twice the number)
of dof . Degrees of freedom are real numbers, put into pairs to make complex numbers by an act of
notation. The number of dof depends on the system. The possibility to describe an arbitrary number of
dof always existed in Hamiltonian physics, but only became important with quantum Hamiltonian physics.
The dynamical discovery of the complex Schrödinger notation is good, but not breakthrough. However the
discovery that electrons are waves, requiring infinitely many dof for their description, is a breakthrough
model that violates every conception of pre-quantum physics.
Unfortunately there is a different quantum usage that identifies operators in the Heisenberg picture as
dynamical “operator degrees of freedom”. That is a disabling choice of words. In that usage three infinite-
dimensional operators ~x would be called “one quantum particle.” In that usage the word “particle” is
abused, and the word “one” means three sets of continuous infinity. We suggest not using those words.

Proc. of SPIE Vol. 8832 88320W-17


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
• The Heisenberg Picture: The Heisenberg picture discusses time-dependent Hermitian operators which have
N 2 parameters. There is an immediate conflict between N 2 dynamical variables in operators and N in the
wave functions. If there actually were N 2 independent dynamical variables, it would not be possible to
get them from N equations. Schrödinger noticed this, and it was Schrödinger who derived the Heisenberg
picture from a unitary transformation. Schrödinger’s language in calling the two pictures “dynamically
equivalent” was too subtle. It means that N 2 operator equations are not independent, that N 2 independent
operator degrees of freedom don’t exist, and that the operator formalism adds no new information to what
is found in the wave function.
Indeed the Heisenberg formalism has less information that the Schrödinger (ordinary Hamiltonian) for-
malism. The Heisenberg operators themselves lack wave functions to read out, or project the information.
The mixup seems to originate in excessive focus on operator eigenvalues - which cannot fully characterize a
system - that were given emphasis early in history. We have nothing against helpful mathematical methods,
such as the interaction picture. Meanwhile it is unhelpful to say that “operators are the physical variables”,
due to the prior definitions of independent dynamical degrees of freedom. We suggest not using the words
in quotes, because they are false.

4. CONSEQUENCES OF SYMMETRY
So far we have left the details of models open. Symmetries can restrict models enough so there is a position that
little but the consequences of symmetries has been discovered so far in physics.
Let D(~x, t) be a list of data measuring “something about the photons” in an experiment. Perhaps we have a
number of hits in a particle detector. Perhaps we have a sequence of field amplitudes measured in a radio beam.
We choose not to be more specific, and treat the data as very general.
Let e(~x) be a time-independent “space pattern” occurring in the data. We cannot know the data in advance,
and must find the pattern from the data. By definition the pattern is normalized: we treat the normalization of
the data separately. The overlap at time t, < e|d(t) >, measures how well the pattern represents the data. We
define an optimal pattern as having the maximum time-averaged overlap-squared:
Z
dt | < e|d(t) > |2 → max.

Replace the labels ~x, t by subscripts-xt, and re-copy:


e∗x Cxx0 ex0 → max; (42)
Z
Cxx0 = dt Dxt Dx∗0 t .

By the Rayleigh-Ritz variational theorem, Eq. 42 is extremized at the eigenvectors of the matrix Cxx0 :
Cxx0 eα
x0 = λ
(α) α
ex .
The optimal patterns are automatically orthogonal. In its diagonal frame C is written
X
C= |eα > λ(α) < eα |, (43)
α

Since the basis guarantees the maximal-squared projections for each state α, truncating the expansion at any
given number of terms α = 1..p will give the best possible p− dimensional approximation. Ref.16 seems to be
the origin of these facts.
Similarly, let f (t) be a space-independent “time pattern” occurring in the data. Define the optimal time
pattern to have the maximum overlap-squared, averaged over space. Repeating the argument above gives
C̃tt0 ftα0 = λ̃(α) ftα ;
Z
C̃tt0 = d3 x Dxt Dx∗0 t .

Proc. of SPIE Vol. 8832 88320W-18


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Expand the data in both optimal bases:
X
D(~x, t) = eα α
x Λα ft .
α

The expansion coefficients Dαβ =< eα | < f β |D >= δ αβ (λα )2 , thus Λα = (λα )2 . Remarkably, the joint ex-
pansion is diagonal, and the correlation coefficients λ(α) = λ̃(α) . The joint expansion is just the singular value
decomposition. The sum over α can only range over the dimension of the smaller space, here the number of
points on the time axis.
Notice that Cxx0 is a standard spatial correlation obtained from the data. C̃tt0 is a standard time correlation.
This is very general, and so far we have used no physical information. Information comes from a model.
Translational symmetry of correlations is a fine model. Translations (T ) are described by

~x → ~xT = ~x + ~x0 ;
~x0 → ~x0T = ~x0 + ~x0 ;
t → tT = t + t0 ;
t0 → t0T = t0 + t0 . (44)

Symmetry under Eq. 44 specifies four relations. Then C and C̃ depends only on differences:
Z
~ 0
Cxx0 = C(~x − ~x0 ) = d3 k eik·(~x−~x ) C(~k)
Z
0
C̃tt = C̃(t − t ) = dω e−iω(t−t ) C̃(ω)
0
0
(45)

We reverted to function notation, and used the same letters C for the Fourier transforms as the correlation,
because the arguments ~k, ω are sufficient to label them. Each of the Fourier transforms happen to be in diagonal
form (Eq. 43):
X
C̃(t − t0 ) = < t|ω > C̃(ω) < ω|t0 >;
ω
X
C̃ = |ω > C̃(ω) < ω|.
ω

The joint expansion of the data in the optimal bases is diagonal:


Z
~
D(~x, t) = d3 kdω eik·~x−iωt δ(ω − ω(~k))D(~k).

This result is remarkable. It has much more information than a generic expansion in Fourier modes. The double
Fourier expansion is doubly-optimal, and controlled by one (1) set of singular values, whose dimension cannot
exceed the dimension of the smaller space. The smaller space here is labeled by ω, representing one infinity. The
larger space is labeled by ~k, representing three infinities. The matched singular values imply a dispersion relation
ω = ω(~k) generated by the diagonal character of the sum¶ . (Otherwise, D → D(~k, ω) would be a function of
four variables.) We emphasize that no physical assumptions – other than translational symmetry – has been
put in. Yet when we seek to reproduce physical correlations, we be led to a “wave equation” ωD = ω(~k)D to
reproduce the data.
Suppose data correlations are rotationally invariant. That is, the correlations may depend on angles, but
not on the angular origin of coordinates. Let n̂, n̂0 be unit vectors on the sphere. Rotational symmetry implies
The eigenvalues λ̃α (ω) = λα (~k). By the implicit function theorem and with mild assumptions of differentiability,

one set of variables ω = ω(~k) can be eliminated as a local function in some domain. This is also written ~k = ~k(ω), a
2-dimensional surface in the space of ~k.

Proc. of SPIE Vol. 8832 88320W-19


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
correlations C(n̂, n̂0 ) = C(n̂ · n̂0 ). (The dot product is the only scalar one can make from (n̂, n̂0 ).) The addition
theorem for spherical harmonics says
X 4π
C(n̂ · n̂0 ) = ∗
Y`m (θ, φ)C` Y`m (θ0 , φ0 ).
2` + 1
m`

Here θ, φ are the angular coordinates of n̂, with similar notation for n̂0 . The correlation has been factored into
diagonal form, showing the spherical harmonics are the optimal basis for angular correlations. Suppose the time
correlations have no preferred origin. Repeating the singular value decomposition for data that depends on polar
coordinates gives
X
D(θ, φ, ω) = f`m (ω`m )Y`m (θ, φ);
`m
f`m (ω) = δ(ω − ω`m )f`m (ω)

These expressions are “trying to look like” quantum mechanics. The dispersion relation and wave equation
emerges whether or not there is an underlying “wave character” producing the data.
4.0.3 Numerical Experiment One
As an experiment we took a mundane 48 × 54 matrix M sampled at 8-bit accuracy. We cyclically permuted the
rows (label x) 48 times, and added (M · M T )xx0 from each permutation. We cyclically permuted the columns
(label t) 54 times, adding (M T ·M )tt0 each time. The sums give two correlations, Cxx0 and Ctt00 . As predicted the
eigenvectors are pure Fourier modes, degenerate under k → −k, ω → −ω (Figure 1). The optimal expansion of
the total data is a joint Fourier expansion in discrete Fourier series eikx−iωt , where ω = ω(k) is a 1-1 relation on
the smaller space of dimension 48. This dispersion relation is shown in Figure 2. The relation is approximately
ω ∼ k, mainly because low Fourier modes tend to coincide and dominate, but differs somewhat because essential
information is encoded by svd in the singular values. One might say the result is something like a photon
dispersion relation.
The interpretation of free-space dispersion relations has evolved greatly. When Einstein propose E 2 =
p~ c + m2 c4 it was bold and heuristic. We now believe the dispersion relation ω 2 = ~k 2 c2 + µ2 is an immediate
2 2

consequence of translational and Lorentz symmetry of a low-order correlation.

5. THE PRINCIPLE OF MINIMAL ENTROPY


Our primary goals are practical and we believe there is great promise in the practical application of quantum
(or un-quantum) tools to numerous situations having no microphysical component. We mentioned earlier that
division of data sets has an arbitrary element, and that some divisions simplify analysis more than others. We
conclude our discussion with a helpful discovery we call the Principle of Minimal Entropy.
Given data D, we have shown how to construct density matrices ρ, whose statistical features coincide with the
Born rule of quantum probability. In order not to offend anyone seeking Planck’s constant and deep mysteries
in quantum mechanics, our density matrices are “un-Quantum.” From our density matrices we an entropy S .
The entropy is a measure of how the data fills the space from which the density matrix is derived. The definition
of this entropy is

S = −tr(ρ log(ρ)). (46)

One cannot learn much from the single-number summary S. However it is a measure that is unique in being
extensive, additive for non-interactive subsystems, etc. as Von Neumann17 showed. From its definition 0 ≤
S ≤ log(N ) on an N -dimensional complex (2N dimensional real) system. The maximum is reached when
ρ = 1N ×N /N , the unit matrix, which has no preferred subspaces. The minimum is reached when ρ = |ψ >< ψ|,
namely when ρ is rank-1. Then (as we mentioned) description by a wave function happens to be consistent.
None of this procedure assumes probability defined by distributions (informally) or defined by Kolmogorov-
style axioms (very formally). Instead we use our method as a generalization of probability that can be applied to

Proc. of SPIE Vol. 8832 88320W-20


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
\t/ 4A7
40'
Figure 1. “Quantization” by symmetry. Rows and columns of generic finite matrix M were cyclically permuted to make
two correlations, M M T and M T M . The eigenvectors are optimal bases for the original matrix, and by translational
symmetry, Fourier modes.

30

ω 25
20

15

10

k
0 5 10 15 20

Figure 2. The dispersion relation ω = ω(k) obtained from Fig. 1.

any system, regardless of its origin. We propose that physicists associate “order” and “simplicity” with relatively
low entropy. This habit, which actually originates in the physics of classical probability, carries straight over to
un-Quantum systems. A numerical experiment provides an illustration.

5.0.4 Numerical Experiment Two


The data 48 × 54 dimensional data set discussed in Section 4, and generating Fig. 1, was divided by partition
into a number of M × N matrices, for N = 2, 3, 4....120. The entropy of the N × N density matrices (same as
the M × M entropy) was computed. The entropy is shown as a function of the division dimension N in Figure
3. Also shown is the maximum entropy log(N ), and the entropy of an array of 48 × 54 random real numbers
from the interval -1/2 to 1/2. The data’s entropy on every division is much smaller than the random numbers:
it is rather orderly data. Something special happens at periodic values of N . The entropy dips at these values,
indicating a special orderliness and simplicity.
Figure 4 shows a more clear picture of the dips in the plot of the data entropy alone. The first dip is at

Proc. of SPIE Vol. 8832 88320W-21


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
N = 54, the next is at N = 108 = 2 × 54, and so on. The exercise takes the data as string of 48 ∗ 54 = 2592
numbers, and by the principle of minimal entropy, discovers the 48 × 54 division into a special matrix that is
exceptionally orderly. The matrix is shown as a two-dimensional density plot in Figure 5.

4
S
3

0
0 20 40 60 80 100 120
N
Figure 3. The bottom curve shows the quantum entropy S of the data used in Section 4 divided on dimension N as a
function of N . The entropy is considerably smaller than the maximum possible entropy (dashed line) and the entropy of
a matrix of random numbers (middle curve). Dips occur in the entropy at divisions which are especially orderly. the first
dip is indicated by the arrow at N = 54.

0.8

S 0.6
0.4

0.2

0.0
0 20 40 60 80 100 120
N
Figure 4. Dips in the entropy versus dimension N of Figure 3 on a smaller scale. The dips occur at N =54, 108,... which
identifies particularly orderly representation of the data. The first dip is indicated by an arrow.

Proc. of SPIE Vol. 8832 88320W-22


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Figure 5. A density plot of the 48 × 54 matrix of data discussed in Section 4. It has been partitioned at the favored
division of the first dip in Figure 4, also used to make Figures 1- 4. Each pixel is an 8-bit integer sampled from 0-255,
originating in a JPG image.

6. OUTLOOK
Our view of the topic known as “quantum theory” has evolved to draw a sharp distinction between successful
physical models and claims of a profound framework. The first breakthrough was discovering the electron and
photon have infinite dimensional phase spaces. Since then, successful models come from painstaking experimen-
tation and meticulous theoretical consistency checks, rather than a predictive framework. We do not find the
framework of linear algebra, transformation theory, and quantum probabilities customized to mesh with it to
have much physics in it. However the methodology of quantum probability for data representation and analysis is
a breakthrough compared to the older definitions of probability based on mutually exclusive equivalence classes.
As Feynman must have said, sometimes it is a new breakthrough to use an old breakthrough on a new class
of problems. We mentioned in Section 1 there are no reasons to restrict un-Quantum methods to micro-physical
objects. We have been acting on this opportunity for some time. Several examples show that the methods are
useful and powerful:

• The radio frequency emissions of relativistic protons were used to construct density matrices in Ref.18 By
distinguishing signal-based and noised-based density matrices, the signal to noise ratio was improved by a
factor of more than 100, producing the first detection of virtual Cherenkov radiation from protons.

• In Refs.19 observational data from the cosmic microwave background was analyzed to test the “cosmological
principle” based on isotropy. The alignment of spherical harmonic multipoles and the entropy of their power
distribution contradicts isotropy at a high degree of statistical significance. The observations cannot be
explained by galactic foreground subtractions.20
• In Ref.21 density matrices were constructed from high-dimensional spectroscopic data of a pharmaceutical
protein. The principal values were sorted to make projections onto certain subspaces from which the phases
of the protein could be determined by inspection. Ref.22 reviews subsequent progress. The method has
been used to make empirical phase diagrams towards characterizing the active states, phase transitions
and shelf-life of about 100 pharmaceutical compounds and vaccines

Proc. of SPIE Vol. 8832 88320W-23


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Acknowledgements
I thank Hans De Raedt and Kristel Michielsen for correcting errors, for providing comments on the manuscript,
and for helpful discussions.

REFERENCES
[1] There are too many references to cite. A good introduction is the book by S. L. Adler, Quantum theory
as an emergent phenomenon: The statistical mechanics of matrix models as the precursor of quantum field
theory, Cambridge, UK: Univ. Pr. (2004) 225 p. See also G. ’t Hooft, “Emergent Quantum Mechanics and
Emergent Symmetries,” AIP Conf. Proc. 957, 154 (2007) [arXiv:0707.4568 [hep-th]]
[2] An extensive blbliography with wide discussion is given by the Archiv preprint of B. Hu, Emergence: Key
Physical Issues for Deeper Philosphical Enquiries, J. Phys. Conf. Ser. 361 (2012) 012003, arXiv:1204.1077
[physics.hist-ph]. See also H. -T. Elze, “Does quantum mechanics tell an atomistic spacetime?,” J. Phys.
Conf. Ser. 174, 012009 (2009) [arXiv:0906.1101 [quant-ph]].; “Deterministic models of quantum fields,” J.
Phys. Conf. Ser. 33, 399 (2006) [gr-qc/0512016]; “Symmetry aspects in emergent quantum mechanics,” J.
Phys. Conf. Ser. 171, 012034 (2009); M. Blasone, P. Jizba and F. Scardigli, “Can quantum mechanics be an
emergent phenomenon?,” J. Phys. Conf. Ser. 174, 012034 (2009) [arXiv:0901.3907 [quant-ph]]; G. ’t Hooft,
“Classical cellular automata and quantum field theory,” Int. J. Mod. Phys. A 25, 4385 (2010).
[3] Klein, O. and Nishina, Y. ”ber die Streuung von Strahlung durch freie Elektronen nach der neuen relativis-
tischen Quantendynamik von Dirac”. Z. Phys. 52 (11-12): 853; ibid 869, (1929).
[4] Lamb, W., and Scully, M. ”The photoelectric effect without photons,” in [Polarization, Matire et Rayon-
nement,] Volume in Honour of A. Kastler (Presses Universitaires de France, Paris, 1969)
[5] Ralston, J. P., “Embedding P T Quantum Mechanics in Symplectic Quantum Mechanics,” J. Phys. A: Math.
Theor. 40, 9883 (2007).
[6] Ralston, J. P., “Quantum Mechanics without Planck’s Constant,” arXiv:1203.5557 [hep-ph].
[7] Ralston, J. P., “Emergent un-Quantum Mechanics,” in Advances in Quantum Mechanics, edited by Paul
Bracken, (2013), ISBN 978-953-51-1089-7.
[8] A typical text is Kaku, M., [Quantum Field Theory: A Modern Introduction], (Oxford University Press,
1993).
[9] Dirac, P.A.M., [The Principles of Quantum Mechanics], (Cambridge University Press, 1967).
[10] Feynman, R.P., [The Feynman Lectures on Physics], Vol.III (Addison Wesley, 1964); Sakurai, J.J., [Modern
Quantum Mechanics], (Addison Wesley, 1998).
[11] d’ Espangat, B. [Conceptual Foundations of Quantum Mechanics], (2nd ed. Addison Wesley, 1979).
[12] Dewdney, C., Holland, P.R., Kyprianidis, A., “What happens in a spin measurement?,” Phys. Lett. A119,
259 (1986).
[13] Bell, J. S. ”On the Einstein Podolsky Rosen paradox”, Physics 1,195,( 1964). See also [Speakable and
Unspeakable in Quantum Mechanics], Cambridge Univ. Press (1987).
[14] Werner, R. S. “Quantum states with Einstein-Podolsky-Rosen correlations admitting a hidden-variable
model,” Phys. Rev. A 40, 4277, (1989); Peres, A.“Separability Criterion for Density Matrices,” Phys. Rev.
Lett. 77, 1413 (1996).
[15] Bender, C. M., and Boettcher, S. “Real Spectra in Non-Hermitian Hamiltonians Having PT Symmetry,”
Phys. Rev. Lett. 80, 5243 (1998) [arXiv:physics/9712001]; Bender, C. M., Boettcher, S. and Meisinger, P.
“PT-Symmetric Quantum Mechanics,” J. Math. Phys. 40, 2201 (1999) [arXiv:quant-ph/9809072].
[16] Karhunen, H. Ann. Acad. Science. Fenn, Ser A. I. 37, (1947); Loeve, M. supplement to P. Levy, P. [Processes
Stochastic et Mouvement Brownien], Paris, Gauthier Villars, (1948); Hotelling, H. J. Educ. Psychology 24,
417, (1933); ibid 24, 448, (1933).
[17] Von Neumann, J., [Mathematische Grundlagen der Quantenmechanik] Springer, Berlin; English translation
in [The Mathematical Foundations of Quantum Mechanics], (Princeton University Press, Princeton, 1971).
[18] Bean, A., Ralston, J. P. and Snow, J. “Evidence for Observation of Virtual Radio Cherenkov Fields,” Nucl.
Instrum. Meth. A 596, 172 (2008) [arXiv:1008.0029 [physics.ins-det]].

Proc. of SPIE Vol. 8832 88320W-24


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
[19] Samal, P. K., Saha, R., Jain, P. and Ralston,J. P. , “Testing Isotropy of Cosmic Microwave Background
Radiation,” Mon. Not. Roy. Astron. Soc. 385, 1718 (2008); [arXiv:0708.2816 [astro-ph]]; “Signals of Sta-
tistical Anisotropy in WMAP Foreground-Cleaned Maps,” Mon. Not. Roy. Astron. Soc. 396, 511 (2009)
[arXiv:0811.1639 [astro-ph]].
[20] Aluri, P. K., Samal, P. K, Jain, P. and Ralston, J. P. “Effect of Foregrounds on the CMBR Multipole Align-
ment,” Mon. Not. Roy. Astron. Soc. 414, 1032 (2011) [arXiv:1007.1827 [astro-ph.CO]]. See also Ralston,
J. P, J. P. Ralston,“Question isotropy,” in Tanner, D. Ed. [Axions 2010], (University of Florida, Gainsville
2010), AIP Conf. Proc. 1274, 72 (2010)
[21] Kueltzo, L. Fan, J., Ralston, J. P., DiBiase, M., Faulkner, E., and Middaugh, C. R., “The solution behavior
of IFN-beta-1a: An empirical phase diagram based approach, ” J. Pharm. Sci. 94(9), 1893 (2005).
[22] Maddux, N. Joshi, R., Middaugh, C. R., Ralston, J. P., and Volkin, R. “Multidimensional Methods for the
Formulation of Biopharmaceuticals and Vaccines, ” J. Pharm. Sci. 100 4171 (2011).

Proc. of SPIE Vol. 8832 88320W-25


Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Mar 2019
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

Potrebbero piacerti anche