Azerbaijan Journal of Mathematics

V. 10, No 1, 2020, January

ISSN 2218-6816

Local Generalization of Pauli’s Theorem

N.G. Marchuk, D.S. Shirokov∗

Abstract. Generalized Pauli’s theorem, proved by D. S. Shirokov for two sets of anti-
commuting elements of a real or complexified Clifford algebra of dimension 2n , is extended
to the case, where both sets of elements depend smoothly on points of Euclidean space
of dimension r. We prove that in the case of even n there exists a smooth function such
that two sets of Clifford algebra elements are connected by a similarity transformation.
All cases of connection between two sets are considered in the case of odd n. Using the
equation for the spin connection of general form, it is shown that the problem of the local
Pauli’s theorem is equivalent to the problem of existence of a solution of some special
system of partial differential equations. The special cases n = 2, r ≥ 1 and n ≥ 2, r = 1
with simpler solution of the problem are considered in detail.
Key Words and Phrases: local Pauli’s theorem, Clifford algebra, field equation, spin
connection
2010 Mathematics Subject Classifications: 15A66, 53C05, 70S15

1. Introduction

In [12], one of the authors presented statements describing the connection

between two sets g a , ha , a = 1, . . . , n of elements of a real or complexified Clifford
algebra C`F (p, q), p + q = n, that satisfy the relations

g a g b + g b g a = 2η ab e, ha hb + hb ha = 2η ab e, a, b = 1, . . . , n,

where η ab is a diagonal matrix with 1 and −1 appearing on the diagonal p and

q times, respectively. These statements generalize Pauli’s theorem proven for
n = 4 [10]. They have been called algebraic generalized Pauli’s theorem and
significantly used in the study of spin groups [13] and n-dimensional spinors [14].

∗
Corresponding author.

http://www.azjm.org 38
c 2010 AZJM All rights reserved.
 Local Generalization of Pauli’s Theorem 39

In this paper, we generalize those statements to the case where both sets
of elements depend smoothly on the points of Euclidean space. We call these
generalized statements local generalized Pauli’s theorem.
First we show that the connection between two sets is local, in the neigh-
borhood of the point of Euclidean space. Then we generalize this statement to
the case of the entire Euclidean space under certain assumptions. We call the
obtained statement local, as is customary in differential geometry (statement is
called global if it holds for the entire non-trivial manifold, while Euclidean space
is a trivial manifold because it is covered by one chart).

2. Local generalized Pauli’s theorem in the neighborhood of the

point of Euclidean space

Let V be an r-dimensional Euclidean space with a scalar product (x, y),

∀x, y ∈ V and the norm
p
kxk = (x, x), ∀x ∈ V.

Let Ω be an open domain in V and let ε > 0 be a positive real number.

ε-neighborhood of a point x0 ∈ V is the domain

Oε (x0 ) = {x ∈ V : kx − x0 k < ε}.

Let us consider the real Clifford algebra C`R (p, q) := C`(p, q), n = p + q with
the generators ea , a = 1, . . . , n, and the basis of 2n elements

e, ea , eab , . . . , e1...n , (1)

enumerated by the ordered multi-indices of the length between 0 and n. The

identity element of C`R (p, q) is denoted by e. The generators satisfy the anticom-
mutative relations

ea eb + eb ea = 2η ab e, a, b = 1, . . . , n,

where η ab are the elements of diagonal matrix η of order n with 1 and −1 ap-
pearing on the diagonal p and q times, respectively.
Let us consider the complexified Clifford algebra C`C (p, q) := C ⊗ C`R (p, q) [3],
[8]. These two cases C`F (p, q), F = R, C are important for various problems of
mathematical physics, in particular, in the study of the Dirac equation and the
Yang-Mills equations [9], [7], [5].
40 N.G. Marchuk, D.S. Shirokov

The subspace of C`F (p, q) spanned by the basis elements enumerated by the
ordered multi-indices of length k is denoted by C`Fk (p, q) and is called the subspace
of grade k. We have
n
M
C`F (p, q) = C`Fk (p, q).
k=0

Note that we have the important special case r = n, where the basis of V is
the set of generators ea . In this case, V can be considered as pseudo-Euclidean
space with two metrics (Euclidean and pseudo-Euclidean). We use Euclidean
metric to determine a neighborhood of a point.
Consider a function
f : Ω → C`F (p, q)
with values in C`F (p, q). The function f = f (x) can be written in the form

f = ue + ua ea + . . . + u1...n e1...n ,

where u = u(x), ua = ua (x), . . . are the functions Ω → F and the basis elements
(1) do not depend on x ∈ V .
If (real or complex) functions u, ua , . . . , u1...n have continuous derivatives up
to order k in Ω, then we say that the functions u, ua , . . . , u1...n and f belong to
the class C k (Ω) (C 0 (Ω) is the class of continuous functions in Ω).

Theorem 1. (The case of even n).

Let n be an even positive number and ha = ha (x), g a = g a (x), a = 1, . . . , n
be functions Ω → C`F (p, q) of the class C k (Ω) such that

ha (x)hb (x) + hb (x)ha (x) = 2η ab e, a, b = 1, . . . , n, ∀x ∈ Ω,

a b b a ab
g (x)g (x) + g (x)g (x) = 2η e, a, b = 1, . . . , n, ∀x ∈ Ω.

Then for any point x0 ∈ Ω there exist ε > 0 and T = T (x) : Oε (x0 ) → C`F (p, q)
such that

1. T (x) is a function of the class C k (Oε (x0 ));

2. T (x) is an invertible element of C`F (p, q) for any point x ∈ Oε (x0 );

3. g a (x) = T −1 (x)ha (x)T (x), a = 1, . . . , n, ∀x ∈ Oε (x0 );

4. The function T (x) is defined up to multiplication by a (real in the case

F = R and complex in the case F = C) function of the class C k (Oε (x0 ))
that is not equal to zero for any point of Oε (x0 ).
 Local Generalization of Pauli’s Theorem 41

Proof. Let us consider the special case where the elements g a = g a (x) are
equal to the generators ea of C`F (p, q), which do not depend on x. First we prove
the theorem for this case.
We denote the elements of the basis (1) by eA , where A are the ordered
multi-indices of the length between 0 and n. Denote eA = (eA )−1 .
Denote the elements

e, hab := ha hb , 1 ≤ a < b ≤ n; ..., h1...n := h1 · · · hn

by hA , where A is an arbitrary ordered multi-index of the length between 0 and

n. Let us consider the following sums over all such 2n multi-indices A:
X
hA (x)F eA (2)
A

where F is an arbitrary element of the basis (1).

Let us consider an arbitrary point x0 ∈ Ω. By generalized Pauli’s theorem
[12], we have at least one basis element from (1) (we denote it by Fh ) such that
X
Th := hA (x0 )Fh eA 6= 0. (3)
A

We define the norm of Clifford algebra elements by

q
|U | = Tr(U † U ),

where Tr : C`F (p, q) → C`F0 (p, q) is the projection operation onto the subspace
C`F0 (p, q) and the operation of Hermitian conjugation † is defined in [4]. Using
(3), we get X
|Th | = | hA (x0 )Fh eA | = δh > 0.
A

Since a linear combination of functions k

k
P A of the class C (Ω) is a function of the
class C (Ω), we conclude that | A h (x)Fh eA |, x ∈ Ω, is a continuous function.
Thus there exists a real number εh > 0 such that
X
| hA (x)Fh eA | > δh /2, ∀x ∈ Oεh (x0 ).
A

Consequently, we construct a function

X
Th (x) = hA (x)Fh eA 6= 0, ∀x ∈ Oεh (x0 )
A
42 N.G. Marchuk, D.S. Shirokov

of the class C k (Oεh (x0 )). By generalized Pauli’s theorem [12] we have

ea = Th−1 (x)ha (x)Th (x), a = 1, . . . , n, ∀x ∈ Oεh (x0 )).

Similarly (replacing the symbol h by g), we can obtain the connection between
the elements ea and g a using the element Tg (x):

ea = Tg−1 (x)g a (x)Tg (x), a = 1, . . . , n, ∀x ∈ Oεg (x0 )).

Choosing ε = min(εh , εg ), we get

g a (x) = T −1 (x)ha (x)T (x), a = 1, . . . , n, ∀x ∈ Oε (x0 )),

where T (x) = Th (x)Tg−1 (x). The theorem is proved. J

Note that a key role in the proof of the local Pauli’s theorem is played by an
algorithm [12] for computing the element, which connects two sets of anticom-
muting elements by similarity transformation.
Let us formulate and prove the corresponding theorem for the case of odd n.

Lemma 1. Let n be a positive odd number and ha = ha (x), g a = g a (x), a =

1, . . . , n be functions Ω → C`R (p, q) of the class C k (Ω) such that

ha (x)hb (x) + hb (x)ha (x) = 2η ab e, a, b = 1, . . . , n, ∀x ∈ Ω,

a b b a ab
g (x)g (x) + g (x)g (x) = 2η e, a, b = 1, . . . , n, ∀x ∈ Ω.

Then the products

h1...n := h1 (x)h2 (x) . . . hn (x), g 1...n := g 1 (x)g 2 (x) . . . g n (x)

do not depend on x and are equal to ±e1...n or ±e (last case is possible only for
p − q = 1 mod 4).

Proof. Lemma 1 follows from the algebraic Pauli’s theorem (see [12]). J

Theorem 2. (The case of odd n and real Clifford algebra).

Under the assumptions of Lemma 1, for any point x0 ∈ Ω there exist ε > 0
and T = T (x) : Oε (x0 ) → C`R (p, q) such that

1. T (x) is a function of the class C k (Oε (x0 ));

2. T (x) is an invertible element of C`R (p, q) for any point x ∈ Oε (x0 );

3. (a) g a (x) = T −1 (x)ha (x)T (x) ⇔ h1...n = g 1...n ,

 Local Generalization of Pauli’s Theorem 43

(b) g a (x) = −T −1 (x)ha (x)T (x) ⇔ h1...n = −g 1...n ,

(c) g a (x) = e1...n T −1 (x)ha (x)T (x) ⇔ h1...n = e1...n g 1...n ,
(d) g a (x) = −e1...n T −1 (x)ha (x)T (x) ⇔ h1...n = −e1...n g 1...n ,
where equalities hold for a = 1, . . . , n and ∀x ∈ Oε (x0 );

4. The function T (x) is defined up to multiplication by the elements λ(x)e +

ν(x)e1...n , where λ(x) and ν(x) are real functions of the class C k (Oε (x0 ))
such that λ(x)e + ν(x)e1...n is an invertible element for any point of the
domain Oε (x0 ).

Lemma 2. Let n be a positive odd number and ha = ha (x), g a = g a (x), a =

1, . . . , n be functions Ω → C`C (p, q) of the class C k (Ω) such that

ha (x)hb (x) + hb (x)ha (x) = 2η ab e, a, b = 1, . . . , n, ∀x ∈ Ω,

a b b a ab
g (x)g (x) + g (x)g (x) = 2η e, a, b = 1, . . . , n, ∀x ∈ Ω.

Then the products

h1...n := h1 (x)h2 (x) . . . hn (x), g 1...n := g 1 (x)g 2 (x) . . . g n (x)

do not depend on x and are equal to ±e (in the case p − q = 1 mod 4), ±ie (in
the case p − q = 3 mod 4), or ±e1...n (in both cases).

Proof. Lemma 2 follows from the algebraic Pauli’s theorem (see [12]). J

Theorem 3. (The case of odd n and complexified Clifford algebra).

Under the assumptions of Lemma 2, for any point x0 ∈ Ω there exist ε > 0
and T = T (x) : Oε (x0 ) → C`C (p, q) such that

1. T (x) is a function of the class C k (Oε (x0 ));

2. T (x) is an invertible element of C`C (p, q) for any point x ∈ Oε (x0 );

3. (a) g a (x) = T −1 (x)ha (x)T (x) ⇔ h1...n = g 1...n ,

(b) g a (x) = −T −1 (x)ha (x)T (x) ⇔ h1...n = −g 1...n ,
(c) g a (x) = e1...n T −1 (x)ha (x)T (x) ⇔ h1...n = e1...n g 1...n ,
(d) g a (x) = −e1...n T −1 (x)ha (x)T (x) ⇔ h1...n = −e1...n g 1...n ,
(e) g a (x) = ie1...n T −1 (x)ha (x)T (x) ⇔ h1...n = ie1...n g 1...n ,
(f ) g a (x) = −ie1...n T −1 (x)ha (x)T (x) ⇔ h1...n = −ie1...n g 1...n ,
where equalities hold for a = 1, . . . , n and ∀x ∈ Oε (x0 );
44 N.G. Marchuk, D.S. Shirokov

4. The function T (x) is defined up to multiplication by λ(x)e + ν(x)e1...n ,

where λ(x) and ν(x) are complex functions of the class C k (Oε (x0 )) such
that λ(x)e + ν(x)e1...n is an invertible element for any point of the domain
Oε (x0 ).

Proof. Proofs of Theorems 2 and 3 are similar to the proof of Theorem 1 and
we must use generalized Pauli’s theorems [12] for the Clifford algebra with odd
n.
First we prove the connection between the set ha = ha (x) and the set ea ,
a = 1, . . . , n, which does not depend on x. Instead of (2) we consider the following
expressions: X
hA (x)Fh eA
A:|A|=0 mod 2

where the sum is taken over the ordered multi-indices of even length |A|.
The element Fh does not depend on x because it is always among the basis
elements {eB } or among the expressions {eB + eC } (see [12]). All other consid-
erations are similar to the considerations for the case of even n.
Using the connection between the sets ha (x) and ea , the sets g a (x) and ea , we
obtain the connection between the sets g a (x) and ha (x) in some neighborhood of
the point x0 ∈ Ω. The theorem is proved. J
Note that the connection between the sets in the case of odd n (see Theorems
2 and 3) can be written for all cases in the following form:

g a (x) = h1...n g1...n T −1 (x)ha (x)T (x), a = 1, . . . , n, ∀x ∈ Oε (x0 ),

where g1...n := (g 1...n )−1 .

3. Local generalized Pauli’s theorem in the entire Euclidean

space and the connection with one field equation

As was shown above, the connection between two sets of elements satisfying
the defining anticommutative relations of Clifford algebra is realized in the form
of similarity transformation (or in other similar forms in the case of odd n) locally,
in the neighborhood of the corresponding point of Euclidean space. Does local
generalized Pauli’s theorem hold in the entire Euclidean space? Namely, does
there exist a function T = T (x) from Theorems 1, 2, and 3, which is invertible,
continuous, and connects two sets of elements ha (x), g a (x), a = 1, . . . , n for any
point x ∈ V ? In this section we prove the corresponding theorem under additional
assumptions (see conditions (5)).
 Local Generalization of Pauli’s Theorem 45

The Cartesian coordinates of Euclidean space V , dim V = r, are denoted by

xµ , µ = 1, . . . , r and the partial derivatives are denoted by ∂µ = ∂x∂ µ , µ = 1, . . . , r.
All the given functions in this section are smooth to simplify the presentation.
Let a set of smooth functions ha : V → C`F (p, q), a = 1, . . . , n, satisfy
ha (x)hb (x) + hb (x)ha (x) = 2η ab e, a, b = 1, . . . , n, ∀x ∈ V. (4)
In the case of odd n, we also require the additional condition
Tr(h1...n ) = 0, (5)
where Tr : C`F (p, q) → C`F0 (p, q) is the projection operation onto the subspace
C`F0 (p, q). We need this condition (5) to obtain independent elements ha (x), a =
1, . . . , n, which generate the basis of the Clifford algebra C`F (p, q). Otherwise,
they can generate a basis of Clifford algebra of lower dimension (see [12]). Under
this additional condition, we have only two cases of connection between two sets
of Clifford algebra elements in the case of odd n: g a = ±T −1 ha T (instead of four
and six cases in Theorems 2 and 3).
Note that the problem of connection between the sets ha (x) and g a (x), a =
1, . . . , n is equivalent to the problem of connection between the set ha (x), a =
1, . . . , n and the set of generators ea , a = 1, . . . , n, that do not depend on x ∈ V .
Using the connection between the sets ha (x) and ea , the sets g a (x) and ea , we
obtain the connection between the sets ha (x) and g a (x). Therefore, in what
follows we consider the problem of connection between the set ha (x), a = 1, . . . , n
with conditions (4), (5) and the set ea , a = 1, . . . , n.
In [7], [5], [11], one primitive field equation (the system of partial differential
equations) for the spin connection of the general form was considered. In [7] and
[5], the expressions hµ were considered as vector expressions with respect to the
orthogonal transformations of coordinates. In this work, as well as in [11], we
consider instead of them expressions ha , which do not change under orthogonal
transformations of coordinates.
Consider the following field equation (system of equations)
∂µ ha − [Cµ , ha ] = 0, a = 1, . . . , n, µ = 1, . . . , r, (6)
where the components Cµ : V → C`F (p, q) of covector field with values in the
Clifford algebra are considered as unknowns. Thus, we have n × r equations for
r unknown functions.
Since the expressions Cµ are inside the commutator in (6), it is convenient to
consider these expressions up to element of the Clifford algebra center: Cµ : V →
C`F (p, q) \ Cen(C`F (p, q)), where
 F
F C`0 (p, q), in the case of even n;
Cen(C` (p, q)) =
C`0 (p, q) ⊕ C`n (p, q), in the case of odd n.
F F
46 N.G. Marchuk, D.S. Shirokov

The equation (6) is gauge invariant. Namely, the expressions

h́a = S −1 ha S, Ƶ = S −1 Cµ S − S −1 ∂µ S (7)

for a smooth invertible function S : V → C`F (p, q) such that S −1 ∂µ S : V →

C`F (p, q) \ Cen(C`F (p, q)), satisfy the system of equations of the same form

∂µ h́a − [Ƶ , h́a ] = 0, a = 1, . . . , n, µ = 1, . . . , r.

In [5], [11], it is proved that the system (6) has a unique solution Cµ : V →
C`F (p, q) \ Cen(C`F (p, q)) of the following form:

2[ n
2
]
X 1
Cµ = µk π[h]k ((∂µ ha )ha ), µk = , ha := (ha )−1 ,
n − (−1)k (n − 2k)
k=1
(8)
where π[h]k : C`F (p, q)
→ C`F [h]k (p, q)
is the projection operation onto the sub-
space C`F [h]k (p, q) spanned over the basis elements ha1 ...ak with ordered multi-
indices of length k. Explicit expressions for the elements Cµ in the cases of
small n are given in [5] and [11]. We say that the solution (8) describes the spin
connection of general form.
If the generators ea of Clifford algebra, which do not depend on x, are con-
sidered as the expressions h́a (7), then the corresponding connection for these
elements given by the equations (6) is equal to zero: Ƶ = 0. Using (7), we get

∂µ S(x) = Cµ (x)S(x), µ = 1, . . . , r. (9)

The system of partial differential equations (9) with known Cµ (x) is considered
as a system for finding a function S = S(x), which is invertible in the entire
Euclidean space and connects two sets ha (x) and ea , a = 1, . . . , n. We have the
following theorem.

Theorem 4. (Local Pauli’s theorem in the entire Euclidean space).

Let us consider functions ha : V → C`F (p, q), a = 1, . . . , n, that satisfy

ha (x)hb (x) + hb (x)ha (x) = 2η ab e, a, b = 1, . . . , n, ∀x ∈ V

and, in the case of odd n, additional condition Tr(h1...n ) = 0.

Then there exists a function S = S(x) : V → C`F (p, q), ∃ S −1 (x) ∀x ∈ V ,
satisfying the following system of equations:

∂µ S(x) = Cµ (x)S(x), µ = 1, . . . , r (10)

 Local Generalization of Pauli’s Theorem 47

for all x ∈ V , where Cµ : V → C`F (p, q) \ Cen(C`F (p, q)) is a unique solution of
the system of equations:

∂µ ha − [Cµ , ha ] = 0, a = 1, . . . , n, µ = 1, . . . , r, (11)

and there exists a function T (x) = S(x)K (for some invertible element K ∈
C`F (p, q), which does not depend on x), which is also a solution of the system
(10) invertible in the entire Euclidean space and connects two sets of elements:

ea = T −1 (x)ha (x)T (x), a = 1, . . . , n, ∀x ∈ V (12)

in the case of even n and

ea = h1...n e1...n T −1 (x)ha (x)T (x), a = 1, . . . , n, ∀x ∈ V (13)

in the case of odd n, where h1...n e1...n = ±e.

Proof. Using methods of differential geometry, it can be proved that the

function S(x) from the statement of the theorem always exists.
In [5], [11], it is proved that from the system of equations:

∂µ ha − [Cµ , ha ] = 0, a = 1, . . . , n, µ = 1, . . . , r (14)

the following relation for the functions Cµ follows

∂µ Cν − ∂ν Cµ − [Cµ , Cν ] = 0, µ, ν = 1, . . . , r, ∀x ∈ V. (15)

In terms of differential geometry this means that the curvature

Rµν := ∂µ Cν − ∂ν Cµ − [Cµ , Cν ]

is equal to zero, i.e. the connection Cµ is flat. It is known that every flat
connection on a simply-connected manifold is trivial, i.e. it can be represented
in the form Cµ = (∂µ S)S −1 for some function S = S(x) (see, for example, [15]).
Since Euclidean space is simply-connected, there exists a function S(t) invertible
in the entire Euclidean space which satisfies (10). Substituting the expression
Cµ = (∂µ S)S −1 into the equation (11), we get

∂µ ha − (∂µ S)S −1 ha + ha (∂µ S)S −1 = 0. (16)

Using SS −1 = e, we obtain

(∂µ S)S −1 + S∂µ (S −1 ) = 0. (17)

48 N.G. Marchuk, D.S. Shirokov

Multiplying both sides of the equation (16) from the left by S −1 , from the right
by S, and using (17), we get

∂µ (S −1 ha S) = 0, µ = 1, . . . , r. (18)

Since (18), it follows that the set

f a := S −1 ha S, a = 1, . . . , n

does not depend on x ∈ V and satisfies the defining anticommutative conditions

of Clifford algebra:

f a f b + f b f a = S −1 ha SS −1 hb S + S −1 hb SS −1 ha S = ha hb + hb ha = 2η ab e.

If the function S(x) is a solution of the system (10), then any function of the
form S(x)K, where the element K does not depend on x, is also a solution of
the system (10). By the algebraic generalized Pauli’s theorem [12], there is an
invertible element K ∈ C`F (p, q) such that

ea = K −1 f a K, a = 1, . . . , n

in the case of even n and

ea = f 1...n e1...n K −1 f a K, a = 1, . . . , n

in the case of odd n. We conclude that the element

T (x) = S(x)K

connects two sets of elements ea and ha (x), a = 1, . . . , n in the forms (12) and
(13). The theorem is proved. J
Note that Theorem 4 gives us an algorithm for computing the function S =
S(x). Using this algorithm and algorithm for computing the element K provided
by the algebraic Pauli’s theorem (see [12]), we obtain an algorithm for computing
the function T (x) = S(x)K, which connects two sets of elements ha (x), ea ,
a = 1, . . . , n.
Below we give two particular cases (Theorems 5 and 6) of the statement of
Theorem 4, in which the function T (x) has a simpler form. We give different
proofs of these theorems which do not use the fact that any flat connection on a
simply-connected manifold is trivial. In these particular cases, it is sufficient to
use the theory of matrix differential equations or the Poincare lemma.
The following theorem describes the local Pauli’s theorem in the case of Eu-
clidean space V = R1 of dimension r = dim V = 1.
 Local Generalization of Pauli’s Theorem 49

Theorem 5. (Particular case: r = 1, n ≥ 2).

Let us consider smooth functions ha : R → C`F (p, q), a = 1, . . . , n, that satisfy

ha (x)hb (x) + hb (x)ha (x) = 2η ab e, a, b = 1, . . . , n, ∀x ∈ R

and additional condition Tr(h1...n ) = 0 in the case of odd n.

Then there exists a function T = T (x) : R → C`F (p, q) such that

ea = T −1 (x)ha (x)T (x), a = 1, . . . , n, ∀x ∈ R (19)

in the case of even n and

ea = h1...n e1...n T −1 (x)ha (x)T (x), a = 1, . . . , n, ∀x ∈ R (20)

in the case of odd n, where h1...n e1...n = ±e.

Moreover, T (x) = S(x)K, where S(x) is any invertible in the entire Euclidean
space solution of the equation
dS(x)
= C1 (x)S(x), (21)
dx
C1 : R → C`F (p, q)\Cen(C`F (p, q)) is a unique solution of the system of differential
equations
dha
− [C1 , ha ] = 0, a = 1, . . . , n, (22)
dx
and K is an invertible element of the Clifford algebra C`F (p, q).

Proof. In the case r = dim V = 1, the system of partial differential equations

(10) becomes the ordinary differential equation (21) and the system of partial
differential equations (11) becomes the system of ordinary differential equations
(22).
According to the theory of matrix differential equations (see, for example,
[1], Section 14), the equation (21) for a continuous function C1 (x) has always a
solution S(x) which is invertible in the entire Euclidean space, and the general
solution of the equation (21) has the form

T (x) = S(x)K,

where K ∈ C`F (p, q) is any element that does not depend on x. The function
T (x) = S(x)K for some invertible element K ∈ C`F (p, q) connects two sets of
elements in the form (19) in the case of even n and in the form (20) in the case
of odd n.
50 N.G. Marchuk, D.S. Shirokov

Note that the solution of the system (21) in the general case can be written in
the form of multiplicative integral (see [1]). In the case of additional conditions
[C1 (x1 ), C1 (x2 )] = 0 for any x1 , x2 ∈ R, the solution has a simpler form
Z x
S(x) = exp( C1 (x)dx)
x0

for some point x0 ∈ R. J

Now let us consider for arbitrary r ≥ 1 the case where the functions (4) take
values in the subspace C`F1 (p, q) spanned over the generators ea , a = 1, . . . , n:

ha (x) : V → C`F1 (p, q) (23)

In this case we have

ha (x) = yba (x)eb
for some smooth functions yba (x) : V → F. The conditions

ha (x)hb (x) + hb (x)ha (x) = 2η ab e, a, b = 1, . . . , n, ∀x ∈ V

on the functions ha (x), a = 1, . . . , n are equivalent to the following conditions on

the functions yba (x), a, b = 1, . . . , n:

yba (x)ydc (x)η bd = η ac , ∀x ∈ V. (24)

Note that (24) is the orthogonality condition for the matrix

Y = ||yba || ∈ O(p, q, F) = {Y ∈ Mat(n, F), Y T ηY = η},

where Y = ||yba || means that the entry in the a-th row and b-th column of the
matrix Y is denoted by yba .
In the case ha : V → C`F1 (p, q), the unique solution (8) of the system of
equations:

∂µ ha − [Cµ , ha ] = 0, a = 1, . . . , n, µ = 1, . . . , r

has the form (see [7])

1
Cµ = (∂µ ha )ha , ha := (ha )−1 , (25)
4
which is known as the spin connection (see, for example, [2]). In [7], it is shown
that in this case the function (25) takes values in the subspace of grade 2, i.e.
Cµ : V → C`F2 (p, q).
 Local Generalization of Pauli’s Theorem 51

Note that the subspace C`R2 (p, q) is the Lie algebra (with respect to the commu-
tator) of the spin group Spin+ (p, q) and the exponents of the elements of C`R 2 (p, q)
−1 a a b a
are elements of Spin+ (p, q) [3], [6]. The formula S e S = yb e =: h describes
two-sheeted covering of the orthogonal group SO+ (p, q) by the corresponding spin
group Spin+ (p, q): for each orthogonal matrix Y = ||yba || ∈ SO+ (p, q) there exist
two elements ±S ∈ Spin+ (p, q) of the corresponding spin group.
The following theorem describes the local Pauli’s theorem in the entire Eu-
clidean space in the case n = 2 for an arbitrary r with the additional assumption
ha : V → C`F1 (p, q), a = 1, 2.

Theorem 6. (Particular case: n = 2, r ≥ 1).

Let us consider smooth functions ha : V → C`F1 (p, q), a = 1, 2 with values in
the subspace of grade 1 that satisfy

ha (x)hb (x) + hb (x)ha (x) = 2η ab e, a, b = 1, 2, ∀x ∈ V. (26)

Then there exists a function C(x) : V → C`F2 (p, q) such that

1
dC(x) = C1 (x)dx1 + · · · + Cr (x)dxr , Cµ (x) = (∂µ ha )ha .
4
Moreover, the function
T (x) = exp(C(x))K
satisfies
ea = T −1 (x)ha (x)T (x), a = 1, 2, ∀x ∈ V,
for some invertible element of Clifford algebra K ∈ C`F (p, q).

Proof. As mentioned before the theorem, if ha : V → C`F1 (p, q), then the
functions Cµ , µ = 1, . . . , r take values in the subspace C`F2 (p, q) (see [7]).
In the case n = 2, all elements of the subspace C`F2 (p, q) have the form λe12 ,
λ ∈ F. Hence all the functions Cµ : V → C`F2 (p, q) commute with each other:

[Cµ (x), Cν (x)] = 0, µ, ν = 1, . . . , r, ∀x ∈ V. (27)

As mentioned above, in [5], [11], it is proved that the system of equations (14)
implies (15). Thus, by (15), the conditions (27) are equivalent to

∂ µ Cν = ∂ ν Cµ , µ, ν = 1, . . . , r, ∀x ∈ V. (28)

Let us consider the following 1-form L(x) = Cµ (x)dxµ . By the Poincare

lemma, if this form is closed, i.e. dL(x) = 0, then it is exact, i.e. there exists
52 N.G. Marchuk, D.S. Shirokov

a 0-form C(x) such that dC(x) = L(x). By (28), it follows that there exists a
function C(x) : V → C`F2 (p, q) such that

dC(x) = C1 (x)dx1 + · · · + Cr (x)dxr .

Thus, under the conditions (28), the system of equations (10) can be written in
the form

∂µ S(x) = ∂µ (C(x))S(x), µ = 1, . . . , r. (29)

Let us consider the expression

∞
1 2 1 X 1
exp(C(x)) = e + C(x) + C (x) + C 3 (x) + · · · = C k (x).
2! 3! k!
k=0

This series is always convergent and invertible (see, for example, [1]):

(exp(C(x)))−1 = exp(−C(x)). (30)

Since Cµ (x) : V → C`F2 (p, q), we conclude that C(x) also takes values in C`F2 (p, q).
The functions C k (x), k = 1, 2, . . ., and exp(C(x)) take values in C`F0 (p, q) ⊕
C`F2 (p, q). In the Clifford algebra of dimension n = 2 we have [C`F0 (p, q) ⊕ C`F2 (p, q),
C`F2 (p, q)] = 0, and hence

[C k (x), ∂µ C(x)] = 0, ∀k = 1, 2, . . . ,

which is a sufficient condition for the validity of for the following formula (see
[1]):

∂µ (exp(C(x))) = (∂µ C(x)) exp(C(x)). (31)

From (30) and (31), it follows that the function

S(x) = exp(C(x))K

is a solution of the system (29) invertible in the entire Euclidean space for any
invertible element K ∈ C`F (p, q) that does not depend on x:

∂µ (exp(C(x))K) = ∂µ (exp(C(x)))K = (∂µ C(x)) exp(C(x))K.

Using the algebraic Pauli’s theorem, we conclude that for some invertible K ∈
C`F (p, q) the function T (x) = exp(C(x))K connects two sets:

ea = T −1 (x)ha (x)T (x), a = 1, 2, ∀x ∈ V.

 Local Generalization of Pauli’s Theorem 53

The theorem is proved. J

Let us give some examples illustrating the statement of Theorem 6.
1) Let us consider the real Clifford algebra C`R (2, 0) and the functions ha :
V → C`F1 (2, 0) satisfying the relations (26), ha = yba eb . We can parameterize the
elements of the matrix Y = ||yba || ∈ O(2) by the function ϕ = ϕ(x) : V → R. We
have two cases (det Y = ±1):

h1 = cos ϕ e1 + sin ϕ e2 , h2 = − sin ϕ e1 + cos ϕ e2 ; (32)

1 1 2 2 1 2
h = cos ϕ e + sin ϕ e , h = sin ϕ e − cos ϕ e . (33)

In both cases, after direct calculations, we obtain

1 ∂µ ϕ 12
Cµ = (∂µ ha )ha = − e .
4 2

We conclude that there exists a function C(x) = − ϕ(x)

2 e
12 such that C (x)dxµ =
µ
dC(x). We obtain the following solution of the equation ∂µ S(x) = Cµ (x)S(x):

ϕ(x) ϕ(x) 12
S(x) = exp(C(x))K = (cos( ) e − sin( ) e )K,
2 2
where K is an arbitrary element of Clifford algebra that does not depend on x.
We take T (x) = exp(C(x))K for some invertible element K. In the first case (32),
we take K = e, in the second case (33), we take K = e12 . The element T (x) is
invertible for any x ∈ V and connects two sets of elements ea = T −1 (x)ha (x)T (x),
∀x ∈ V . Indeed, it is easy to verify that
ϕ ϕ ϕ ϕ
(cos e + sin e12 )(cos ϕ e1 + sin ϕ e2 )(cos e − sin e12 ) = e1 ,
2 2 2 2
ϕ ϕ 12 ϕ ϕ
(cos e + sin e )(− sin ϕ e + cos ϕ e )(cos e − sin e12 ) = e2 .
1 2
2 2 2 2

2) The case of real Clifford algebra C`R (0, 2) is considered similarly. We have
two cases: (32) and (33). In both cases, we get

1 ∂µ ϕ 12
Cµ = (∂µ ha )ha = e
4 2
and
ϕ(x) ϕ(x) 12
T (x) = exp(C(x))K = (cos( ) e + sin( ) e )K,
2 2
which satisfies ea = T −1 (x)ha (x)T (x), ∀x ∈ V for some invertible K.
54 N.G. Marchuk, D.S. Shirokov

3) In the case of real Clifford algebra C`R (1, 1) we have four cases (since the
group O(1, 1) has four connected components):
h1 = cosh ϕ e1 + sinh ϕ e2 , h2 = sinh ϕ e1 + cosh ϕ e2 ;
h1 = cosh ϕ e1 + sinh ϕ e2 , h2 = − sinh ϕ e1 − cosh ϕ e2 ;
h1 = − cosh ϕ e1 − sinh ϕ e2 , h2 = sinh ϕ e1 + cosh ϕ e2 ;
h1 = − cosh ϕ e1 − sinh ϕ e2 , h2 = − sinh ϕ e1 − cosh ϕ e2 .
In all cases, we have
1 ∂µ ϕ 12
Cµ = (∂µ ha )ha = − e
4 2
and
ϕ(x) ϕ(x) 12
T (x) = exp(C(x))K = (cosh( ) e − sinh( ) e )K,
2 2
which satisfies ea = T −1 (x)ha (x)T (x), ∀x ∈ V for some invertible K.
4) Let us consider the case of real Clifford algebra C`R (3, 0) and functions
ha : V → C`F1 (3, 0) satisfying the relations (4), ha = yba eb . For simplicity, we
consider the case of matrix Y = ||yba || ∈ O(3) with the determinant det Y = 1.
In this case, the matrix Y ∈ SO(3) can be parameterized by three Euler angles
ϕ(x), ψ(x), θ(x), depending on x. We have 0 ≤ ϕ, ψ < 2π, 0 ≤ θ < π.
The matrix Y has the form
 
cos ϕ cos ψ cos θ − sin ϕ sin ψ − cos ϕ sin ψ cos θ − sin ϕ cos ψ cos ϕ sin θ
 sin ϕ cos ψ cos θ + cos θ sin ψ sin ϕ sin ψ cos θ + cos ϕ cos ψ sin ϕ sin θ  .
− cos ψ sin θ sin ψ sin θ cos θ
We get
1 1
Cµ = (∂µ ha )ha = ((cos θ ∂µ ϕ + ∂µ ψ)e12 +
4 2
+(− sin ψ sin θ ∂µ ϕ − cos ψ ∂µ θ)e13 + (cos ψ sin θ ∂µ ϕ + sin ψ ∂µ θ)e23 ).
In this example, we obtain [Cµ , Cν ] 6= 0, and therefore ∂µ Cν 6= ∂ν Cµ and the
Poincare lemma is not applicable, unlike the case n = 2.
Problems related to the local Pauli’s theorem are useful in field theory, in the
study of the Dirac equation [9] and the Yang-Mills equations [7], [11]. In [11], a
class of covariantly constant solutions of the Yang-Mills equations is proposed.
The connection between these solutions and constant solutions is described by
the local Pauli’s theorem.
An interesting question is whether the Pauli’s theorem is valid on the curved
manifolds. Our hypothesis is that the Pauli’s theorem will be true only in some
particular cases (see Theorems 5 and 6) depending on n, dimension of the mani-
fold r, or topological properties of the manifold (simply connectedness).
 Local Generalization of Pauli’s Theorem 55

Acknowledgements

The authors are grateful to N. Hitchin and M. Katanaev for useful discussions.
The authors are, also grateful to the reviewers for carefully reading this paper and
helpful remarks. This work was supported by the Russian Science Foundation
(project 18-71-00010).

References

[1] F.R. Gantmacher, The theory of matrices, GITTL, Moscow, 1953. (in Rus-
sian)

[2] M. Green, J. Schwarz, E. Witten, Superstring theory, 1, Mir, Moscow, 1990.

(in Russian)

[3] P. Lounesto, Clifford Algebras and Spinors, Cambridge Univ. Press, Cam-
bridge, 1997.

[4] N.G. Marchuk, D.S. Shirokov, Unitary spaces on Clifford algebras, Advances
in Applied Clifford Algebras, 18(2), 2008.

[5] N.G. Marchuk, D.S. Shirokov, General solutions of one class of field equa-
tions, Rep. Math. Phys., 78(3), 2016, 305-326.

[6] N.G. Marchuk, D.S. Shirokov, Introduction to the theory of Clifford algebras,
Phasis, Moscow, 2012. (in Russian)

[7] N.G. Marchuk, On a field equation generating a new class of particular so-
lutions to the Yang-Mills equations, Proceedings of the Steklov Institute of
Mathematics, 2014, Vol. 285, 197-210.

[8] N.G. Marchuk, Demonstration Representation and Tensor Products of Clif-

ford Algebras, Proc. Steklov Inst. Math., 290, 2015, 143-154.

[9] N. Marchuk, Field theory equations, Amazon, CreateSpace, 2012.

[10] W. Pauli, Contributions Mathematiques a la Theorie des Matrices de Dirac,

Ann. Inst. Henri Poincare 6, 1936, 109-136.

[11] D.S. Shirokov, Covariantly constant solutions of the Yang-Mills equations,

Advances in Applied Clifford Algebras, 28, 2018, 53.

[12] D.S. Shirokov, Extension of Pauli’s theorem to Clifford algebras, Dokl.

Math., 84(2), 2011, 699-701.
56 N.G. Marchuk, D.S. Shirokov

[13] D.S. Shirokov, Calculations of elements of spin groups using generalized

Pauli’s theorem, Advances in Applied Clifford Algebras, 25(1), 2015, 227-
244.

[14] D.S. Shirokov, Pauli theorem in the description of n-dimensional spinors

in the Clifford algebra formalism, Theoret. and Math. Phys., 175(1), 2013,
454-474.

[15] G. Walschap, Metric structures in differential geometry, Series: Graduate

Texts in Mathematics, 224, Springer, 2004.

Nikolay G. Marchuk
Steklov Mathematical Institute of Russian Academy of Sciences, 119991, Moscow, Russia
E-mail: [email protected]

Dmitry S. Shirokov
National Research University Higher School of Economics, 101000, Moscow, Russia
Kharkevich Institute for Information Transmission Problems of Russian Academy of Sciences,
127051, Moscow, Russia
E-mail: [email protected]

Received 24 October 2018

Accepted 01 April 2019