to appear in

Rep. Math. Phys.

vol. 42 (1998) hep-th/xxxxxxx

DE DONDER–WEYL THEORY AND A HYPERCOMPLEX

EXTENSION OF QUANTUM MECHANICS TO FIELD THEORY

Igor V. Kanatchikov∗
arXiv:hep-th/9810165v1 21 Oct 1998

Laboratory of Analytical Mechanics and Field Theory

Institute of Fundamental Technological Research
Polish Academy of Sciences
Świȩtokrzyska 21, Warszawa PL-00-049, Poland

(Submitted May 1998 —— Accepted September 1998)

A quantization of field theory based on the De Donder-Weyl (DW) covariant Hamiltonian

formulation is discussed. A hypercomplex extension of quantum mechanics, in which
the space-time Clifford algebra replaces that of the complex numbers, appears as a
result of quantization of Poisson brackets of differential forms put forward for the DW
formulation earlier. The proposed covariant hypercomplex Schrödinger equation is shown
to lead in the classical limit to the DW Hamilton-Jacobi equation and to obey the
Ehrenfest principle in the sense that the DW canonical field equations are satisfied for
the expectation values of properly chosen operators.

1 . Introduction

It is commonly believed in theoretical physics that a generalization of the Hamiltonian

formalism to field theory requires a distinction between the space and time variables and
implies the treatment of fields as infinite dimensional mechanical systems. However,
another approach is possible. It treats the space and time coordinates on equal footing
(as analogues of a single time parameter in mechanics) and does not explicitly refer to
an idea of a field as a mechanical system evolving in time by treating the field rather
as a system varying both in space and in time. The approach has been known as
the De Donder–Weyl (DW) theory in the calculus of variations since the thirties [1]
although its applications in physics have been rather rare. For recent discussions of
mathematical issues of DW theory and further references see [2, 3, 4].
Usually the Hamiltonian formalism serves as a basis for the canonical quantization.
It is quite natural, therefore, to ask whether the DW formulation, viewed as a field
theoretic generalization of the Hamiltonian formalism in mechanics, can lead to a cor-
responding quantization procedure in field theory. In the present paper we discuss an
approach to such a quantization (for earlier discussions see [5, 6]). It is our hope that
∗
e-mail: ikanat@[Link], kai@[Link]

1
2 I.V. KANATCHIKOV

the study of quantization based on the DW theory can contribute to our understand-
ing of the fundamental issues of quantum field theory and to provide us with a new
framework of quantization which could be useful in situations where the applicability
of the conventional canonical quantization in field theory can be in doubt. Note also
that the manifest covariance of the approach can make it especially appealing in the
context of quantization of gravity and extended objects.

2 . De Donder-Weyl theory: a reminder

Let us recall the essence of the DW formulation. Given a Lagrangian density L =

L(y a , ∂µ y a , xν ), where {y a } are field variables, {∂µ y a } denote their space-time deriva-
tives and {xµ }, µ = 1, ..., n, are space-time coordinates, we can define the new set
of Hamiltonian-like variables: pµa := ∂L/∂(∂µ y a), called polymomenta, and H :=
∂µ y a pµa − L, called the DW Hamiltonian function, which allow us to write the Euler-
Lagrange field equation in an appealing, manifestly covariant first order form

∂µ y a = ∂H/∂pµa , ∂µ pµa = −∂H/∂y a (2.1)

referred to as the DW Hamiltonian field equations.

Similar to the Hamiltonian formulation in mechanics an analogue of the Hamilton-
Jacobi (HJ) theory can be developed for the DW Hamiltonian field equations. It is
formulated in terms of n HJ functions on the field configuration space S µ = S µ (y a, xµ )
which fulfill the DW HJ equation

∂µ S µ + H(xµ , y a, pµa = ∂S µ /∂y a ) = 0. (2.2)

The quest of a formulation of a quantum field theory which in the classical limit would
give rise to the DW HJ equation has been one of the motivations of the present study.
Let us consider an example of interacting scalar fields y a described by the Lagrangian
density
1
L = ∂µ y a ∂ µ ya − V (y). (2.3)
2
Then the polymomenta and the DW Hamiltonian function are given by

1
paµ = ∂µ y a , H = paµ pµa + V (y), (2.4)
2
the DW Hamiltonian field equations take the form

∂µ y a = paµ , ∂µ pµa = −∂V /∂y a , (2.5)

and the DWHJ equation reads

1
∂µ S µ + ∂a S µ ∂ a Sµ + V (y) = 0. (2.6)
2
HYPERCOMPLEX EXTENSION OF QUANTUM MECHANICS TO FIELD THEORY 3

3 . Poisson bracket of forms: properties and the equations of motion

To develop an analogue of the canonical quantization procedure we need the Poisson

bracket possessing appropriate algebraic properties, a notion of the canonically con-
jugate variables, and a representation of the field equations in terms of the Poisson
bracket.
In previous papers [4, 5, 7] we have shown that the proper analogue of the Poisson
bracket for the DW Hamiltonian formulation can be defined on horizontal differential
p-forms
p 1
F := F µ1 ... µn−p (z M )∂µ1 ... µn−p , ω
(n − p)!
p = 0, 1, ..., (n − 1), which naturally play a role of dynamical variables. The following
notations are used throughout

ω := dx1 ∧ ... ∧ dxn ,

∂µ1 ... µq := ∂µ1 ∧ ... ∧ ∂µq ,
ωµ1 ... µq := ∂µ1 ... µq ω,
{z M } := {y a, pµa , xµ }. (3.1)

The sign denotes the inner product of a multivector field with a form, such that e.g.
∂µ dx = δµν . The same symbol ∂M denotes either the partial derivative with respect
ν

to the variable z M or a tangent vector ∂z∂M according to the context. For details of the
construction of the Poisson bracket and its properties we refer to [4, 7]. For us it is
most important here that the bracket defined on forms leads to several generalizations
of the Poisson algebra of functions in mechanics, and that it also enables us to represent
the equations of motion of dynamical variables in terms of the bracket with the DW
Hamiltonian function.
In particular, on the class of specific forms, called in [4, 5] Hamiltonian, the bracket
determines the structure of the so-called Gerstenhaber algebra, a specific graded gen-
eralization of the Poisson algebra. By definition, it is a graded commutative algebra
equipped with a graded Lie bracket operation which fulfills the graded Leibniz rule with
respect to the graded commutative product in the algebra. The grade of an element of
the algebra with respect to the product differs by one from its grade with respect to
the bracket operation.
The graded commutative (associative) product on Hamiltonian forms is what we
called the co-exterior product and denoted •. It is defined as follows1 :
p q p q
F • F := ∗−1 (∗F ∧ ∗F ), (3.2)

where ∗ denotes the Hodge duality operator acting on horizontal forms and ∗−1 is its
p q q p p q
inverse. As a consequence, F • F = (−1)(n−p)(n−q) F • F , deg(F • F ) = p + q − n, and
a form of degree p has a grade (n − p) with respect to the co-exterior product.
1
Prof. Z. Oziewicz pointed out to the author that this product was introduced much earlier by
Plebanski, see [9].
4 I.V. KANATCHIKOV

The bracket operation on Hamiltonian forms is graded Lie, with the grade of a
bracket with a p-form being (n − p − 1), so that the bracket of a p-form with a q-form
is a form of degree q − (n − p − 1). The bracket also fulfills the graded Leibniz rule
with respect to the •-product
p q r p q r q p r
{[F , F • F ]} = {[F , F ]} • F + (−)(n−q)(n−p−1) F • {[F , F ]}. (3.3)

All these properties characterize the space of Hamiltonian forms as a Gerstenhaber

algebra.
Hamiltonian forms of non-zero degree are polynomials of (n − 1)-forms pia ωi with
respect to the •-product, with the coefficients being arbitrary functions of the field and
space-time variables (cf. eq. (2.4) in [7]). Note that the variables pia ωi can be viewed
as canonically conjugate to the field variables since their Poisson bracket is

{[pµa ωµ , y b]} = δab .

In fact, owing to the implicit graded canonical symmetry in the theory there are other
canonical pairs of forms of various degrees corresponding to the field variables and
polymomenta (cf. sect. 4.1). The corresponding canonical brackets are of particular
interest from the point of view of the canonical quantization.
Note, that a bracket of any two Hamiltonian forms can be calculated using the
canonical brackets and the graded Leibniz property of the bracket, independently of
the construction in our previous papers which uses the notion of the polysymplectic
form and the related map from forms to multivector fields. However, still it is not
clear how the co-exterior product, the space of Hamiltonian forms, and the canonical
brackets could be invented or motivated independently of the construction in [4].
The equations of motion can be written in terms of the Poisson bracket of forms.
An analogy with mechanics suggests that they are given by the bracket with the DW
Hamiltonian function. However, the degree counting shows that the bracket with H
exists only for Hamiltonian forms of degree (n − 1): F := F µ ωµ . For these forms the
equations of motion can be written in the form

d•F = −σ(−1)n {[H, F ]} + dh •F (3.4)

where d• denotes the operation of the “total co-exterior differential”

p 1
d• F := ∂M F µ1 ... µn−p ∂µ z M dxµ • ∂µ1 ... µn−p ω, (3.5)
(n − p)!

dh is a “horizontal co-exterior differential”:

p 1
dh • F := ∂µ F µ1 ... µn−p dxµ • ∂µ1 ... µn−p ω,
(n − p)!
and σ = +1 (−1) for the Euclidean (Minkowskian) signature of the space-time metric.
It is evident that co-exterior differentials identically vanish on forms of degree lower
than (n − 1) sharing the property with the operation of the bracket with H.
HYPERCOMPLEX EXTENSION OF QUANTUM MECHANICS TO FIELD THEORY 5

Note, that the form of the equations of motion in (3.4) is different from that pre-
sented in our previous papers in which the left hand side has been written in terms of
the operator ∗−1 d, where d is a total exterior differential defined as follows
p 1
dF := ∂M F µ1 ... µn−p ∂µ z M dxµ ∧ ∂µ1 ... µn−p ω.
(n − p)!
In fact, the action of −σ(−1)n d• on (n − 1)-forms coincides with that of ∗−1 d, so
that the essence of the equations of motion in both representations remains the same.
However, the use of d• better conforms with the natural product operation • of
Hamiltonian forms and also with the fact that the bracket with H exists only for
forms of degree ≥ (n − 1).
Note also, that the Poisson bracket formulation of the equations of motion can be
extended (in a weaker sense) to arbitrary horizontal forms. For this purpose one have
to make sense of the bracket with the DW Hamiltonian n-form Hω. The result is that
the bracket with Hω corresponds to the total exterior differential of a form [4].

4 . Elements of the canonical quantization

4.1. Quantization of the canonical brackets

The problem of quantization of the Gerstenhaber algebra of Hamiltonian forms is by
itself, independently of its application to field theory, an interesting mathematical prob-
lem, which could be approached by different mathematical techniques of quantization,
such as a deformation quantization or a geometric quantization. However, in this paper
we shall follow a more naive approach based on extending the rules of the canonical
quantization to the present framework.
Let us recall that in quantum mechanics it is sufficient to quantize only a small
part of the Poisson algebra given by the canonical brackets. Moreover, it is known to
be impossible to quantize the whole Poisson algebra due to the limits imposed by the
Groenewold-van Hove theorem (see e.g. [10]). Therefore, to begin with let us confine
ourselves to an appropriate small subalgebra in the algebra of Hamiltonian forms.
From the properties of the graded Poisson bracket discussed in the previous section
it follows that the subspace of (n−1)-forms and 0-forms constitutes a Lie subalgebra in
the Gerstenhaber algebra of Hamiltonian forms. Let us quantize the canonical brackets
in this subalgebra. Nonvanishing brackets are given by [4]

{[pµa ωµ , y b]} = δab , {[pµa ωµ , y b ων ]} = δab ων , {[pµa , y bων ]} = δab δνµ . (4.1a, b, c)

As usual, we associate Poisson brackets to commutators divided by ih̄ and find the
operator realizations of the quantities involved on an appropriate Hilbert space. In the
Schrödinger y-representation from quantization of (4.1a) it follows that the operator
corresponding to the (n − 1)-form pia ωi can be represented by the partial derivative
with respect to the field variables:

pd
µ
a ωµ = ih̄∂a . (4.2)
6 I.V. KANATCHIKOV

Quantization of the bracket in (4.1b) does not add anything new. However, quantiza-
tion of (4.1c) is nontrivial.
Let us write p̂µa in the form ih̄p̂µ ∂a , where the operator p̂µ has to be found. Then,
from the commutator corresponding to (4.1c) we obtain

[p̂µa , yd
b ω ] = ih̄p̂µ ∂ ◦ yd
ν a
b ω − yd
ν
b ω ◦ ih̄p̂µ ∂
ν a
b µ µ b
= ih̄δa p̂ ◦ ων + p̂ ◦ ων ŷ ih̄∂a − ω
b b b ν ◦ p̂µ ŷ b ih̄∂a , (4.3)

where ◦ denotes a composition law of operators which implies some not known in
advance multiplication law of horizontal operators ω b i and p̂i . The right hand side of
(4.3) will be equal to ih̄δνµ δab , as it is required by (4.1c), if the following two conditions
are fulfilled:
p̂µ ◦ ω b ν = δνµ , p̂µ ◦ ω b ν ◦ p̂µ = 0.
bν − ω (4.4)
Hence, the composition law ◦ is a symmetric operation, i.e. p̂µ ◦ ωb ν = 21 (p̂µ ◦ ω b ν ◦ p̂µ ).
bν +ω
These properties can be satisfied quite naturally by the hypercomplex imaginary units
of the Clifford algebra of the space-time. These hypercomplex imaginary units γµ
(which in four-dimensional Minkowski space-time can be represented by Dirac matrices)
are defined by the relation γµ γν +γν γµ = ηµν , where ηµν is the space-time metric tensor.
Then the operators above can be realized as follows

p̂ν = −κγ ν , b ν = −κ−1 γν ,

ω (4.5)

where the quantity κ of the dimension [length−(n−1) ] appears in order to account for
the physical dimensions of pν and ων . Since ων is essentially an infinitesimal volume
element the absolute value of κ−1 can be expected to be very small. As a result,
the theory under consideration requires the introduction of a certain analogue of the
fundamental length from the elementary requirement of matching of the dimensions.
Note that the realization (4.5) in terms of Dirac matrices is not uniquely determined
by (4.4). In sect. 5.3 we show that this choice is consistent with the Ehrenfest theo-
rem. Still, an open question to be investigated is whether or not other hypercomplex
systems can be useful for the realization of the commutation relations following from
quantization of the Poisson brackets of forms.

4.2. DW Hamiltonian operator

In order to quantize a simple field theoretic model given by (2.3) we have to construct
the operator corresponding to the DW Hamiltonian in (2.4). Note that we cannot
just naively multiply operators. For example, from (4.4) for the operator pd µ
a ω µ one
µ b
would obtain p̂a · ωµ = −inh̄∂a , whereas the correct answer consistent with the bracket
in (4.1a) is pd
µ
a ω µ = ih̄∂a . To find the operator of pµa paµ in (2.4) let us quantize the
bracket
1
{[ pµa paµ , y bων ]} = pbν . (4.6)
2
HYPERCOMPLEX EXTENSION OF QUANTUM MECHANICS TO FIELD THEORY 7

From the corresponding commutator

1 µ a d 1 d
[ pd b µ a d
b d
b d
µ a
bbν
a pµ , y ων ] = (pa pµ ◦ y ων − y ων ◦ pa pµ ) = ih̄p (4.7)
2 2
using the representations of ωb µ and p̂µ a a b
a , and the commutator [△, y ] = 2δb ∂ , where
△ := ∂a ∂ a is the Laplacian operator in the field space, we obtain

pd
µ a 2 2
a pµ = −h̄ κ △ .

Hence the DW Hamiltonian operator for the system of interacting scalar fields takes
the form
c = − 1 h̄2 κ2 △ +V (y).
H (4.8)
2
For a free scalar field V (y) = 21 m2 y 2 /h̄2 , and the above expression is similar to the
Hamiltonian of the harmonic oscillator in the field space.

5 . Generalized Schrödinger equation

Our next step is to formulate a dynamical law. An analogue of the Schrödinger equation
here has to fulfill the following natural requirements:

• the familiar quantum mechanical Schrödinger equation should be reproduced if

the number of space-time dimensions n = 1;

• the DW HJ equation should arise in the classical limit;

• the classical field equations in the DW canonical form should be fulfilled for the
expectation values of the corresponding operators.

We also imply that basic principles of quantum theory such as the superposition princi-
ple and the probabilistic interpretation should be inbuilt in the desired generalization.
Additional hint comes from the bracket form of the equations of motion (3.4) and an
analogy with quantum mechanics. They suggest that the sought-for Schrödinger equa-
ˆ ∼ HΨ,
tion has the symbolic form ı̂dΨ c where ı̂ and dˆ denote appropriate analogues of
the imaginary unit and the exterior differentiation respectively.
The above considerations have led us to the following generalization of the Schrödin-
ger equation
c
ih̄κγ µ ∂µ Ψ = HΨ, (5.1)
where H c is the operator corresponding to the DW Hamiltonian function, the constant
κ of the dimension [length]−(n−1) appears again on dimensional grounds, and Ψ =
Ψ(y a , xµ ) is a wave function over the configuration space of the field and space-time
variables. Equation (5.1) leaves us with two options as to the nature of the wave
function Ψ. The latter can be either a hypercomplex number

Ψ = ψI + ψµ γ µ + ψµν γ µν + ψµ1 ...µn γ µ1 ...µn , (5.2)

8 I.V. KANATCHIKOV

where γ µ1 ...µp := γ [µ1 ...γ µp ] , or a Dirac spinor (the choice of the Dirac spinors is based
on the fact that they exist in arbitrary space-time dimensions and signatures) which
actually can be understood as an element of a minimal left ideal in the Clifford al-
gebra [8, 12]. The choice in favor of spinors is made in sect. 5.2 on the basis of the
consideration of the scalar products.

5.1. Quasiclassical limit and DW HJ equation

Let us show that (5.1) leads to the DW HJ equation in the quasiclassical limit. It is
natural to consider the following generalization of the quasi-classical ansatz

Ψ = R exp(iS µ γµ /h̄κ), (5.3)

where R and S µ are functions of both the field and space-time variables. The exponent
in (5.3) is understood as a series expansion so that one has the analogue of the Euler
formula
Sµ
exp(iαS µ γµ ) = cos α|S| + iγµ sin α|S|, (5.4)
|S|
q
where |S| := S µ Sµ can be both real and imaginary, and α := 1/h̄κ. Thus the
nonvanishing components of the quasiclassical wave function (5.3) are as follows

µ Sµ
ψ = R cos α|S|, ψ = iR sin α|S|. (5.5)
|S|

The wave function of the form

Ψ = ψ + ψµ γ µ

is sufficient to close the system of equations which follows from (5.1). Indeed, in this
case (5.1) reduces to
c
ih̄κ∂µ ψ µ = Hψ, (5.6)
c ,
ih̄κ∂µ ψ = Hψ (5.7)
µ

and the remaining equation ∂[µ ψν] = 0, which follows from the γ µν -component, is
c is assumed to be independent of
equivalent to the integrability condition of (5.7) if H
x-s.
Now, let us substitute (5.5) to (5.6) and (5.7) with the DW Hamiltonian operator
given by (4.8) and collect together the terms appearing with the cos and sin functions
respectively. Then from (5.6) we obtain

S µ ∂µ |S| 1 m2
R = − (R ∂a |S|∂a |S| − α−2 △ R + 2 y 2R) (5.8)
|S| 2 h̄
R 1 RS µ ∂µ |S| − |S|S µ ∂µ R
∂µ S µ = − (R △ |S| + 2∂a R ∂a |S|) + , (5.9)
|S| 2 |S|2
HYPERCOMPLEX EXTENSION OF QUANTUM MECHANICS TO FIELD THEORY 9

and (5.7) yields

1 Sµ 1 Sµ 1 Sµ
∂µ R = − ∂a R ∂a |S| − R △ |S| − R ∂a ∂a |S|, (5.10)
2 |S| 2 |S| 2 |S|
1 m2 S µ 1 −2 S µ 1 Sµ
R ∂µ |S| = − 2 y 2 R + α △ R + α−2 ∂a R ∂a
2 h̄ |S| 2 |S| 2 |S|
µ µ
1 S 1 S
+ α−2 R △ + R ∂a |S| ∂ a |S|. (5.11)
2 |S| 2 |S|
µ
By contracting (5.10) with 2R S|S| and using (5.9) we obtain

Sµ
∂µ S µ = ∂µ |S|. (5.12)
|S|
Sµ
Similarly, eq. (5.8) and eq. (5.11) contracted with |S|
yield

∂a S µ ∂ a Sµ = ∂a |S|∂ a |S|. (5.13)

With the aid of (5.12) and (5.13) equation (5.8) can be written in the form
1 1 m2 2 1 2 2 △R
∂µ S µ = − ∂a S µ ∂a Sµ − y + h̄ κ . (5.14)
2 2 h̄2 2 R
Obviously, in the classical limit h̄ → 0 (5.14) reduces to the DW HJ equation (2.6).
However, besides (5.14) the quasiclassical ansatz (5.3) leads to two supplementary
conditions (5.12) and (5.13) on HJ functions S µ . These conditions are just trivial
identities at n = 1. At n > 1 they represent a kind of duality between the field
theoretical Hamilton-Jacobi formulation in terms of n functions S µ and the mechanical-
like Hamilton-Jacobi equation (in the space of field variables) for the eikonal function
µ
|S|, with the analogue of the time derivative given by the directional derivative S|S| ∂µ .
On the one hand, a possible speculation could be that this is just another manifestation
of a quantum duality between the particle and the field (wave) aspects of a quantum
field. On the other hand the appearance of the supplementary conditions alien to the
DW HJ theory can be related to the fact that the ansatz (5.3) does not represent the
most general hypercomplex number: instead of 2n components we have in (5.3) only
(n + 1) independent functions. The most general ansatz would be
Ψ = R exp{i(S µ γµ + S µν γµν + ... + S µ1 ... µn γµ1 ... µn )/h̄κ}.
However, its substitution to (5.1) leads to cumbersome expressions which we have
been unable to analyze and, moreover, it is not clear whether the antisymmetric quan-
tities S µν , S µνα etc. can be interpreted within some generalized (maybe Lepagean?)
Hamilton-Jacobi theory for fields.
Note, that all the above conclusions can be extended to the case when the wave
function in (5.1) is a spinor. For this purpose the quasiclassical ansatz for the spinor
wave function can be taken in the form
Ψ = R exp(iS µ γµ /h̄κ)η, (5.15)
10 I.V. KANATCHIKOV

where η denotes a constant reference spinor, e.g. η = ||1, 0, ..., 0||T , which allows us to
convert a Clifford number to an element of an ideal of the Clifford algebra, i.e. to a
spinor. The same extends to the presented above more general ansatz.

5.2. Scalar products: hypercomplex vs. spinor wave functions

Let us return now to the issue of the probabilistic interpretation of the wave function.
Note first that if we restrict ourselves to the hypercomplex wave functions of the form
Ψ = ψ + ψµ γ µ , then equations (5.6), (5.7) and their complex conjugates lead to the
conservation law Z
∂µ dy [ψψ µ + ψψ µ ] = 0 (5.16)
under the assumption that the wave function sufficiently rapidly decays at |y| → ∞
and that the operator of the DW Hamiltonian is Hermitian with respect to the L2
scalar product of functions in y-space. From (5.16) it follows that the spatial integral
over a space-like hypersurface Σ
Z Z
ωµ dy[ψψ µ + ψψ µ ] (5.17)
Σ

is preserved in time (or, equivalently, does not depend on the variation of the hypersur-
face Σ) and, therefore, could be viewed as a norm of the hypercomplex wave function.
As this norm involves the integration over a space-like hypersurface it could be useful
for the calculation of the expectation values of global observables. However, its sig-
nificant drawback is that it is not necessarily positive definite as a consequence of the
similarity of (5.17) with the scalar product in the Klein-Gordon theory. The similarity
is evident from (5.7) which essentially states that ψµ ∼ i∂µ ψ.
As a matter of fact, for the purposes of the present theory we need rather a scalar
product for the calculation of the expectation values of operators representing local
quantities. This scalar product should be scalar (to not change the tensor behavior
of operators under averaging) and involve only the integration over the field space
dimensions (to keep the local character of the quantities under averaging). For the
hypercomplex wave function of the type Ψ = ψ + ψµ γ µ the scalar product could be
chosen in the form Z
dy[ψψ + ψ µ ψµ ]
which is, however, not positive definite in general. In fact, non-existence of the ap-
propriate scalar product for wave functions taking values in algebras different from
the real, complex, quaternion and octonion numbers follows from the natural axioms
ensuring the availability of the probabilistic interpretation and general algebraic con-
siderations (see e.g. [11]). Moreover, our attempts to use the just mentioned scalar
product in order to obtain an analogue of the Ehrenfest theorem have failed.
To avoid, at least partially, the difficulties above, we assume that the wave function
in (5.1) is a Dirac spinor. Indeed, in this case the analogue of the global scalar product
(5.17) Z Z
ωµ dyΨγ µ Ψ, (5.18)
Σ
HYPERCOMPLEX EXTENSION OF QUANTUM MECHANICS TO FIELD THEORY 11

where Ψ denotes the Dirac conjugate of Ψ, is positive definite. We also can write a
scalar product for the averaging of local quantities
Z
< Φ|Ψ >:= dy ΦΨ. (5.19)
b are calculated according to the formula
Then the expectation values of operators O
Z
< b
O >:= b
dy ΨOΨ. (5.20)

However, since the scalar product ΨΨ is not positive definite the validity of the aver-
aging in (5.20) is questionable. Moreover, from the generalized Schrödinger equation
(5.1) written in the form
c
ih̄κ∂µ Ψ = −ih̄κγµν ∂ ν Ψ + γµ HΨ (5.21)
and its conjugate we derive
Z Z
∂µ dy ΨΨ = −i dy [Ψγµν ∂ ν Ψ − ∂ ν Ψγµν Ψ] 6= 0, (5.22)

so that the scalar product in (5.19) is space and time dependent, the property which
makes it unsatisfactory analogue of the scalar product of wave functions in quantum
mechanics. Nevertheless, we show in what follows that the use of the formula (5.20)
for the expectation values allows us to obtain an analogue of the Ehrenfest theorem.

5.3. The Ehrenfest theorem

Let us consider the evolution of the expectation values of operators calculated according
to (5.20). Using the generalized Schrödinger equation (5.1) for the evolution of the
expectation value of the polymomentum operator pbµa = −ih̄κγ µ ∂a we obtain
Z
∂µ < pbµa > = −∂µ dyΨih̄κγ µ ∂a Ψ
Z
= −ih̄κ dy[∂µ Ψγ µ ∂a Ψ + Ψ∂a γ µ ∂µ Ψ]
Z
= c
dy[(HΨ)∂ c
a Ψ − Ψ∂a (HΨ)]
Z
= − c
dyΨ(∂a H)Ψ
c >,
= − < ∂a H (5.23)
c with respect to the scalar product of functions in y-space is
where the Hermicity of H
used. Similarly,
Z
−1
∂µ < yd µ
a ω > = −κ ∂µ dyΨγ µ ya Ψ
Z
i c c
= dy[(HΨ)y a Ψ − Ψya (HΨ)]
h̄κZ2
= ih̄ dyΨ∂a Ψ

= < pd
µ
a ωµ > . (5.24)
12 I.V. KANATCHIKOV

By comparing these results with the DW Hamiltonian equations (2.1) we conclude that
the latter are fulfilled in average as a consequence of (i) the generalized Schrödinger
equation (5.1), (ii) the rules of quantization leading to the realization (4.5) of the op-
erators, and (iii) the prescription (5.19) for the averaging of the operators representing
local dynamical variables.
Thus we have arrived at the field theoretic counterpart (within the approach under
discussion) of the Ehrenfest theorem known in quantum mechanics. Its validity could be
seen as a justification and consistency check of the whole approach of the present paper.
However, the situation is not that perfect because the norm used for the calculation
of the expectation values is neither positive definite nor constant over the space-time.
This brings about potential problems with the probabilistic interpretation and points
to the need for improvement of the current formulation or for a better understanding of
its physical content. Moreover, the analogue of the Ehrenfest theorem can be obtained
only for specially chosen operators and the principle behind this choice is not clear.
Note also, that the presented proof of the Ehrenfest theorem is not sensible to the
identification of γ-s with the Dirac matrices. In principle, the use of other hyper-
complex units (which appear in various first-order relativistic wave equations) in the
realization of operators pbµa and ω
b µ and in the generalized Schrödinger equation can also
be consistent with the Ehrenfest theorem, but we lack an appropriate interpretation of
this observation.

6 . Conclusion

The De Donder-Weyl formulation provides us with the alternative covariant canoni-

cal framework for quantization of field theory. On the classical level it possesses the
analogues of the appropriate geometric and algebraic structures, such as the Poisson
bracket with corresponding Lie and Poisson algebraic properties, the notion of the
canonically conjugate variables, and the Poisson bracket formulation of the equations
of motion. Within the DW formulation field theory is treated essentially as a gen-
eralized multi-parameter, or “multi-time” generalized Hamiltonian system with the
space-time variables entering on equal footing as generalizations of the time parameter
in mechanics. The configuration space is a finite dimensional bundle of field variables
over the space-time, of which the field configurations are the sections, instead of the
usual infinite dimensional space of the field configurations on a hypersurface of the
constant time. The analogue of the canonical formalism for the DW formulation [4]
arises as a graded version of the canonical formalism in mechanics, with the role of
dynamical variables played by differential forms.
Quantization of the canonical brackets leads us to a hypercomplex extension of the
quantum mechanical formalism, with the usual complex quantum mechanics recovered
in the limiting case of a one-dimensional “field theory”, that is in mechanics. In higher
dimensions the Clifford algebra of the corresponding space-time manifold plays the
central role. Namely, in the Schrödinger picture considered here the quantum operators
are realized as the differential operators with hypercomplex coefficients, and the wave
functions take values in the spinor space, which is known to be the minimal left ideal
HYPERCOMPLEX EXTENSION OF QUANTUM MECHANICS TO FIELD THEORY 13

in the Clifford algebra [12]. The generalized Schrödinger equation formulated in sect.
5 can also be viewed as a multi-parameter hypercomplex extension of the quantum
mechanical Schrödinger equation: the right hand side of the latter, i∂t , is generalized
to the Dirac operator iγ µ ∂µ . This equation, with some reservations, is shown to lead in
the classical limit to the field theoretic DW Hamilton-Jacobi equation and to give rise
to the analogue of the Ehrenfest theorem for the evolution of the expectation values
of operators corresponding to field variables and polymomenta. However, a potential
problem with the proposed generalized Schrödinger equation is that the scalar product
involved in the proof of the Ehrenfest theorem is not positive definite and not constant
over the space-time.
Although we have entertained here a point of view that the space-time Clifford al-
gebras play the central role, some of the results, except the derivation in sect. 5.1 of the
DW HJ equation in the quasiclassical limit and quantization of the canonical brackets
in sect. 4.1, seem to hold true if γ µ -s are generating elements of other hypercomplex
systems used in the first order relativistic wave equations, such as the Duffin-Kemmer
ring. How the corresponding non-Clifford hypercomplex extensions of quantum me-
chanics can be reconciled with quantization based on the DW theory, whether they
follow from quantization similar to that in sect. 4.1, and which extension (Clifford or
non-Clifford) can be suitable in physics are the questions we hope to address in our
further research.
Though the proposed quantization scheme reproduces essential formal ingredients
of quantum theory, the prospects of its physical applications remain obscure. The
obstacle is that the conceptual framework of the present approach is different from the
usual one, that makes the translation to the conventional language of quantum field
theory a difficult task.
A possible link could be established with the functional Schrödinger picture in quan-
tum field theory [13]. On the one hand, the Schrödinger wave functional Ψ([y(x)], t) is a
probability amplitude of finding the field in the configuration y(x) on the hypersurface
with the time label t. On the other hand, it is natural to interpret our wave function
Ψ(x, t, y) as a probability amplitude of finding the field value y in the space-time point
(x, t). Hence, the Schrödinger wave functional could appear as a kind of composition
of amplitudes given by our wave functions taken at all points x of the space. A more
technical discussion of this issue in [6] points to a relation between both at least in the
ultra-local approximation of vanishing wave vectors. However, beyond this unphysical
approximation the relation remains conjectural and requires further study.
Note added: In the recent preprint [14] M. Navarro considered an approach to
quantization in field theory which is similar to the approach of the present paper.

Acknowledgments
The work has been partially supported by the Stefan Banach International Mathematical
Center to which I express my gratitude. I wish to thank M. Flato, Z. Oziewicz, J.
Slawianowski and C. Lopez Lacasta for discussions, J. Stasheff for his useful comments
on the author’s paper [7], and S. Pukas for his helpful suggestions.
14 I.V. KANATCHIKOV

REFERENCES
[1] Th. De Donder, “Theorie Invariantive du Calcul des Variations” , Nuov. éd., Gauthier-Villars,
Paris (1935); H. Weyl, Ann. Math. (2) 36, 607 (1935); for review see: H. Rund “The Hamilton-
Jacobi Theory in the Calculus of Variations”, D. van Nostrand, Toronto (1966)

[2] M.J. Gotay, J. Isenberg and J. Marsden “Momentum Maps and Classical Relativistic Fields”
(various versions are known since 1985 as the GIMMSY paper), Part I: Covariant Field Theory,
physics/9801019

[3] G. Giachetta, L. Mangiarotti and G. Sardanashvily, “New Lagrangian and Hamiltonian Methods
in Field Theory”, World Scientific, Singapore (1997).

[4] I.V. Kanatchikov, Rep. Math. Phys. 41, 49 (1998); hep-th/9709229

[5] I.V. Kanatchikov, From the Poincaré-Cartan form to a Gerstenhaber algebra of Poisson brackets
in field theory, in: “Coherent States, and Complex Structures”, J.-P. Antoine e.a. eds., Plenum
Press, N.Y. (1995), p. 173, hep-th/9511039

[6] I.V. Kanatchikov, Int. J. Theor. Phys. 37, 333 (1998), quant-ph/9712058; see also I.V.
Kanatchikov, Hypercomplex wave functions and the Born-Weyl quantization in field theory,
preprint March 1996, 16pp, (presented as a part of the poster at the GROUP21 Colloquium,
Goslar, Germany 1996), unpublished

[7] I.V. Kanatchikov, Rep. Math. Phys. 40, 225 (1997); hep-th/9710069

[8] D. Hestenes, “Space-time algebra,” Gordon and Breach, New York (1966); see also: B. A.
Rosenfeld, ”A History of Non-Euclidean Geometry: Evolution of the Concept of a Geometric
Space,” Springer-Verlag, New York (1988)

[9] J.F. Plebanski, J. Math. Phys. 20, 1415 (1979)

[10] G.G. Emch, “Mathematical and Conceptual Foundations of 20th-Century Physics”, North-
Holland, Amsterdam, (1984), Ch. 8.1

[11] S.L. Adler, “Quaternionic Quantum Mechanics and Quantum Fields”, Oxford Univ. Press, New
York (1995)

[12] I. M. Benn and R. W. Tucker, “An Introduction to Spinors and Geometry with Applications in
Physics,”, Adam Hilger, Bristol (1987); see also: P. Becher and H. Joos, Z. Phys. C 15, 343
(1982); I.M. Benn and R.W. Tucker, Comm. Math. Phys. 89, 341 (1983); N.A. Salingaros, G. P.
Wene, Acta Appl. Math. 4, 271 (1985)

[13] B. Hatfield, “Quantum Field Theory of Point Particles and Strings”, Addison-Wesley, Redwood
City (1992); see also: S.S. Schweber, “An Introduction to Relativistic Quantum Field Theory,”
Harper and Row, New York (1961); P. Mansfield, M. Sampaio and J. Pachos, Short distance
properties from large distance behavior, hep-th/9702072 (and references cited therein)

[14] M. Navarro, Toward a finite dimensional formulation of quantum field theory,

quant-ph/9805011.