You are on page 1of 6

Home Search Collections Journals About Contact us My IOPscience

The mathematics of the electron's spin

This content has been downloaded from IOPscience. Please scroll down to see the full text.

1997 Eur. J. Phys. 18 164

(http://iopscience.iop.org/0143-0807/18/3/006)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 142.3.100.23
This content was downloaded on 07/09/2015 at 23:43

Please note that terms and conditions apply.


Eur. J. Phys. 18 (1997) 164–168. Printed in the UK PII: S0143-0807(97)80250-4

The mathematics of the electron’s


spin

Roger Penrose
Mathematical Institute, 24–29 St Giles’, Oxford OX1 3LB, UK

Received 11 December 1996

Abstract. The Dirac equation for the electron is central to a Résumé. L’équation de Dirac pour l’électron est au centre
considerable body of mathematics. Its essential mathematical d’une partie importante de mathématiques. Ses ingrédients
ingredients, namely Clifford algebras and the theory of mathématiques essentiels, c’est à dire les algèbres de Clifford
spinors in n dimensions have wide applications in numerous et la théorie de spineurs en dimension n, ont de larges
fields. A taste of all this is supplied here. applications dans nombre de disciplines. Ici l’on tâche de les
faire goûter.

1. Introduction somewhat earlier, but he had not published his result.)


The algebra of quaternions consists of objects built up
One of the most impressive and original uses of from—i.e. generated by—ordinary real-number algebra
mathematics in modern theoretical physics was Paul and the additional three basic elements i, j, and k, these
Dirac’s discovery of the wave equation for the electron being subject to the laws
by taking of a ‘square root’ of the D’Alembertian
operator. In this way, he was able to accommodate the i2 = −I, j2 = −I, k2 = −I,
electron’s spin, with its correct value of 12 in units of ij = −ji = k, jk = −kj = i, ki = −ik = j,
h̄, in a natural and almost automatic way (Dirac 1928). where I is the identity element of the algebra. The
The Dirac equation is, indeed, one of the cornerstones elements i, j, and k are said to anti-commute. The
of modern theoretical physics. Its properties led Dirac general quaternion is an expression of the form
to predict the existence of anti-particles. When taken
together with the Maxwell electromagnetic equations q = sI + xi + yj + zk,
and the principles of quantum field theory it provides
a scheme of extraordinary power and accuracy— where s, x, y and z are ordinary real numbers. This
quantum electrodynamics—having a precision of some is a division algebra, in the sense that every non-zero
11 decimal places. quaternion q has an inverse q −1 satisfying q −1 q =
I shall not be concerned with such physical qq −1 = I. The ordinary algebra of three dimensional
implications of Dirac’s theory here. Instead, I vectors can be expressed very easily in terms of
shall touch upon some of the important mathematical quaternions.
connotations of the Dirac equation, the significance of There is, unfortunately, the drawback that quaternions
which may not be widely appreciated by physicists or are specific to the geometry of three-dimensional
physics teachers. Euclidean space. However, a generalization of the
quaternion algebra was introduced in 1878 by the
brilliant but short-lived English geometer William
2. Clifford algebras Kingdon Clifford. A Clifford algebra is generated by
n anti-commuting elements
To start, let us turn back to the 19th century. In 1843, γ1, γ2 . . . , γn, (γ i )2 = −I
in his attempts to generalize the notion of a complex γ i γj = −γj γ i ,
number, so that it applied to three-dimensional space
rather than just to the two dimensions of the complex for each i, j = 1, 2, . . . , n, which extend the list
plane, the Irish mathematician and physicist William of three quantities i, j, k generating the system of
Rowan Hamilton discovered what he called the algebra quaternions (see Brauer and Weyl 1935, Budinich and
of quaternions. (Apparently, the German mathematician Trautman 1988 and Penrose 1986, Appendix, for further
Carl Friedrich Gauss had already found these quantities information). Now, however, the product of two of

0143-0807/97/030164+05$19.50
c 1997 IOP Publishing Ltd & The European Physical Society 164
The mathematics of the electron’s spin 165

the γs is not another quantity of the same kind, which so there is now no advantage in requiring that our metric
relates to the fact that, in four or higher dimensions, tensor gij be positive definite, i.e. that the squares of
there is no invariant notion of ‘cross product’ of a pair the γs all be equal to −1, with respect to the original
of vectors to produce another vector. Instead, it is more choice gij = diag(1, 1, . . . , 1). In our Minkowskian
appropriate to think of this ‘cross product’ as providing case, labelling the γs as γ 0 , γ 1 , γ 2 , γ 3 , we shall now
a mathematical representation of the (oriented) plane have
element spanned by the two vectors (which is, in a
sense, the more ‘correct’ way to think of a cross product (γ 0 )2 = −I, (γ 1 )2 = +I,
in three dimensions in any case). (γ 2 )2 = +I, (γ 3 )2 = +I.
In fact, even earlier than Clifford, in 1844, the We can also generalize this, without difficulty, to curved
German high school teacher Hermann Grassmann found spacetime.
an algebra which generalizes to n dimensions the normal
vector algebra of three dimensions. Grassman’s algebra
differs from Clifford’s only in that the relation (γ i )2 = 3. The Dirac equation
−1, given above, is replaced by (γ i )2 = 0. Whereas
linear combinations of the γs give objects with the what does this have to do with electrons? The key
geometrical interpretation of vectors (quantities with property of Clifford algebras that Dirac used—in fact,
direction and magnitude), the product of two γs is he rediscovered it, being unaware of Clifford’s earlier
an object with the interpretation of a plane element work on this
(oriented, and with the magnitude of an area). Likewise,  very thing—is that if we define the ‘Dirac
the product of three γs has the interpretation of a volume operator’ ∇ by

element, and so on. ∇ = γ a ∇a ,
The Clifford algebra is somewhat richer than the
Grassmann algebra, and in fact one can extract then we have
the Grassmann algebra from the Clifford algebra by 2  
considering only the anti-symmetrical elements of the ∇ = ∇∇ = γ a γ b ∇a ∇b = −,
algebra where  is the D’Alembertian operator, defined by
γ [i γj . . . γ m] ,  = ∇a ∇ a = gab ∇ a ∇ b ,
where the square brackets around the indices denote the usual gradient operator ∇a being defined by
anti-symmetrization, e.g.

γ [i γj ] = 12 (γ i γj − γj γ i ), ∇a = .
∂x a
γ [i γj γ k] = 16 (γ i γj γ k − γj γ i γ k − γ i γ k γj 
+γj γ k γ i + γ k γ i γj − γ k γj γ i ). Thus, the Dirac operator ∇ is indeed the square root
of the D’Alembertian  (apart from the sign, which is
In fact, all the elements of the Clifford algebra are merely conventional, and is a result of my choice of
linear combinations of such anti-symmetrical elements, negative signature for gij ). 
of which there are precisely 2n independent ones all We must ask what the operator ∇ actually operates
together (since there are just n!/r!(n − r)! independent on. To understand this, we need to know what sort
such objects with r anti-symmetric indices). This comes of ‘things’ the elements of the Clifford algebra act
about because of identities of the form on. In Clifford’s scheme, these things would have
γ i γj = γ [i γj ] − gij I been other elements of the Clifford algebra, but in
γ i γj γ k = γ [i γj γ k] − gij γ k + gik γj − gj k γ i , the context of quantum mechanics, which was Dirac’s
specific interest, we think of these things as quantum
etc, where gij = diag(1, 1, . . . , 1). state-vectors or, in Dirac’s terminology, ket vectors
Employing the ‘metric tensor’ gij , we can, in fact, |ψi. Dirac looked upon the elements of the Clifford
write the entire set of Clifford relations succinctly as algebra as quantum operators acting on such kets. In
some explicit coordinate frame, these operators would
γ i γj + γj γ i = −2gij I.
be represented by matrices, here called ‘Dirac matrices’
This has the advantage that we can generalize the notion and the kets would be ‘column vectors’. How big would
of Clifford algebra to non-positive-definite metrics these matrices have to be? The smallest size they can
such as that appropriate to ordinary four-dimensional be, in the case of ordinary spacetime, is 4 × 4. Recall
spacetime, for which I shall adopt the signature −2, so that the dimension of the Clifford algebra, in the case
that of an n-dimensional space, is 2n . Here, n = 4, so
2n = 16. The number of independent (4 × 4) matrices
gij = diag(+1, −1, −1, −1)
(i.e. the number of elements of such a matrix is indeed
in flat spacetime, with respect to ordinary Minkowskian 4 × 4 = 16, so the numbers check).
coordinates. Unlike the quanternion algebra, the Note that the ‘4’ that is the dimension of the
Clifford algebras (for n > 2) are not division algebras, spacetime is a ‘different 4’ from that which occurs in
166 R Penrose

the size of the Dirac matrices. In the case of a six- but we now find that η anti-commutes with each of
dimensional spacetime, for example, the size of the γ 0 , γ 1 , . . . , γ 2n−1 . Thus, algebraically, η is on a similar
Dirac matrices would be 8 × 8, since 8 × 8 = 26 . For a footing to each of the γs. Indeed, people might
2n-dimensional spacetime, it would be 2n × 2n . sometimes write γ 2n for η and, in the spacetime case,
What happens if the dimension of the spacetime is γ 5 is the usual physicist’s notation (the ‘γ 0 ’ being not
odd, say (2n + 1)-dimensional? Consider the quantity always used). We have
η = γ 0 γ 1 . . . γ 2n . η 2 = ±1,
It is easy to see that η, actually commutes with the plus sign holding when the signature is a multiple of
every other element of the Clifford algebra, so we can 4 and the minus sign when it is not. In the first case, η
represent it by a scalar (in fact, by one of the numbers has eigenvalues ±1, and in the second case, ±i. Thus,
±1 or ±i, the first pair of possibilities being relevant (idempotent) operators
when the signature of the spacetime is congruent to 1
mod 4 and the second, when it is congruent to 3 mod 4). 1 ∓ η,
This implies that any element α of the Clifford algebra for signature a multiple of 4, or
will be represented by the same matrix as ηα (up to
a possible sign or factor i), so the number of linearly 1 ± iη
independent matrices in the representation need only be for signature not a multiple of 4, serve to reduce
1
2
× 22n+1 = 2n × 2n , whence the Dirac matrices actually the space of spinors—called spin-space—to half the
turn out to be of size 2n × 2n . dimensionality that it had before. The eigenvectors
What then, are these ‘things’ that the Dirac of these operators are called reduced spinors. The
matrices—or, more abstractly, the Clifford algebra dimension of the reduced spin-space for a spacetime
elements—actually operate on? They are called spinors, of dimension 2n is 2n−1 , where we recall that the
for the good reason that in the Dirac equation, for dimension of unreduced spin-space is 2n . In the case
the normal spacetime case, the Clifford elements act when the number of spacetime dimensions is the odd
upon the ket vector |ψi representing the electron’s state. number 2n + 1, the spin-space does not reduce, and has
The four components of |ψi incorporate the degrees of dimension 2n , as described above.
freedom involved in the electron’s spin.
The free Dirac equation can be written
 5. Calculus of 2-spinors
(∇ + m)|ψi = 0,
where I have chosen units so that the speed of light In the case of ordinary four-dimensional spacetime, with
c = 1 and Dirac’s form of Planck’s constant h̄ = signature −2, we have a two-dimensional reduced spin-
1. Moreover, as is the usual practice, I have also space. Such reduced spinors are referred to as 2-spinors
suppressed the appearance of the identity Clifford (or sometimes Weyl spinors, the unreduced 4-spinors
element I (which should strictly multiply the electron being frequently called Dirac spinors). In fact, 2-spinors
mass m), and I shall continue to suppress it henceforth. can be used to provide a powerful calculus for the study
Note that of spacetime geometry. This comes about because a 2-
  2 spinor may be regarded as, in some sense, a square-root
0 = (∇ − m)(∇ + m)|ψi = (∇ − m2 )|ψi
= − ( + m2 )|ψi. of a spacetime vector. More correctly, it is the future-
pointing null vectors which admit a kind of square root
The vanishing of ( + m2 )|ψi is the assertion that |ψi in this sense. Let us see how this comes about.
indeed describes a free particle of mass m. Let V be a spacetime 4-vector, with components
(T , X, Y, Z) in a standard coordinate system (with c =
1). Consider the matrix
4. Reduced spinors
1  T + Z X + iY 
√ .
The four-component entity |ψi is usually referred to as 2 X − iY T − Z
a Dirac spinor. Why does it have four components, This is Hermitian, and it is positive definite if V is
when the spin of the electron, having the value 12 , has future-timelike. If the index notation V a is used for
only two components? An answer often given to this the components of V , then the corresponding notation
question is that this doubling arises because the Dirac 0
V AA would be used for the components of the matrix.
equation describes both the electron and the positron at Here the index A, referring to the rows of the matrix,
the same time. However, I think this is a somewhat labels the components for the reduced spin-space, and
misleading way of stating things, and I prefer looking the primed index A0 , referring to the columns of the
at the Dirac equation in a slightly different way. matrix, labels the components of the complex conjugate
Recall the quantity η that commutes with all the reduced spin-space. (The ‘reduced spin-space’ and
Clifford elements when the spacetime dimension is odd. its ‘complex conjugate’ here refer to the two possible
In the case of 2n dimensions, we can still define choices of eigenvalue ±i for the quantity η, i.e. to
η = γ 0 γ 1 . . . γ 2n−1 , the two different types of 2-spinor.) The translation of
The mathematics of the electron’s spin 167

spacetime vector or tensor quantities into 2-spinor form asymmetric particle was known at the time; all these
amounts to replacing each tensor index a, b, c, etc by its properties are now thought to apply to the neutrino.) We
corresponding pair of 2-spinor indices AA0 , BB 0 , CC 0 , now see that the two parts of the Dirac 4-spinor each
etc. behave like ‘neutrinos’, but with the other one as a kind
Although this looks at first like a complication, of ‘source’. These are the left-handed and right-handed
great simplifications can arise because two-dimensional components of the electron’s wavefunction.
things are simpler to handle than four-dimensional ones; The 2-spinor calculus is valuable also in other
moreover, the spin-spaces are complex, and complex contexts, one of these being Einstein’s general relativity.
spaces have a simpler algebra than do real ones. The Here, the ‘graviton’ is a spin-2 object. Indeed, 2-spinors
power of the 2-spinor algebra frequently stems from turn out to be useful for describing objects of higher spin
the fact that the spinor indices can be manipulated (Dirac 1936). Instead of having a quantity with just one
separately in a simple fashion, and the corresponding 2-spinor index, for spin 12 n, one can use an n-index
tensor operations that these manipulations encode are spinor quantity ψAB...N with n symmetrical indices. A
often exceedingly obscure and complicated (see van der useful decomposition now applies:
Waerden 1929, Penrose and Rindler 1984, 1986).
Now consider the case when V is a future-pointing ψ AB...N = α(A β B . . . ν N ) ,
null vector, so it points along the future light cone in where the parentheses around the indices denote sym-
spacetime. Then we have metrization. Since each of the 2-spinors αA , β B . . . , ν N
T 2 − X 2 − Y 2 − Z 2 = 0, T > 0. determines a direction along the light cone, we have an
elegant geometrical interpretation of ψ AB...N in terms of
This tells us that the corresponding matrix has rank 1 an unordered set of n null directions. This is useful in
and is non-negative definite, whence it is the product many contexts. (In particular, it gives a direct way of
of some column vector with the row vector which is deriving an ingenious, but not very well known, rep-
its conjugate transpose. In index-notation form, we can resentation of higher spin states obtained by Majorana
write this (1932).)
0 0
V AA = λA λ̄A ,
0
for some 2-spinor λA , where λ̄A is its complex
conjugate. This is the ‘square-root’ operation referred 6. Further remarks about spinors
to above. It also tells us that 2-spinors can be directly
use to label the points of future light cones. There is Higher-dimensional spinors also turn out to have
an additional phase contained in the 2-spinor λ, whose numerous applications. One of these is to the twistor
components are λA , because the phase freedom theory of flat Minkowski 4-space, which can be viewed
as the theory of reduced spinors for a ‘spacetime’ of
λA → eiθ λA dimension 6 and signature −2. Here, the ordinary four-
does not affect V . In fact, λ can be represented, up dimensional spacetime is the space of ‘light rays’ along
to a sign, as a flag on the light cone, and this phase a ‘light cone’, in the six-dimensional space. The reduced
transformation corresponds to the rotation of the flag spinors for this 6-space have four complex components,
plane about the (null) flagpole through 2θ. The details and are referred to as twistors. Spinors of much higher
of this will not be entered into here. numbers of dimensions are also used in other physical
In 2-spinor notation, the Dirac equation for the theories, such as in string theory.
electron can now be written as the pair of 2-spinor There are also important applications in pure
equations mathematics. Perhaps the most noteworthy is in
0 0 the Atiyah–Singer index theorem. This is something
∇AA0 αA = −µβA0 and ∇ AA βA = µαA , that applies to differential equations and various
where the Dirac 4-spinor is now split into√ the two generalizations in a very comprehensive manner. Very
2-spinors αA and βA0 and where m = µ 2. The roughly speaking, what Atiyah and Singer studied was
2-spinor notation enables us to write, very simply, the what happens when differential equations are varied
Weyl equation for the neutrino: continuously, and they showed that certain things remain
unchanged under such variation. As part of their
∇AA0 αA = 0. general theory, the higher-dimensional Dirac equations
As a consequence, αA satisfies (or, rather, what the physicists call ‘Weyl neutrino
equations’, as described above) play a particular role—
αA = 0, in the sense that all other equations can be continuously
showing that the equation describes a massless particle. broken down into combinations of them.
(Mathematicians sometimes refer to this equation as All these things spring from the remarkable equation
the ‘Dirac equation’—which is, in a certain sense, satisfied by the humble electron (and its massless cousin,
correct, because Dirac found this equation first (cf the neutrino—if indeed the actual neutrino has no mass).
Dirac 1982), before discarding it in favour of his This is an excellent example of the great richness so
electron equation because no spin- 12 massless reflection often found in the mathematics of the laws of Nature.
168 R Penrose

References Dirac P A M 1982 Pretty mathematics Int. J. Theor. Phys. 21


603–5
Brauer R and Weyl H 1935 Spinors in n dimensions Am. Majorana E 1932 Atomi orientati in campo magnetico
J. Math. 57 425–49 variabile Nuovo Cimento 9 43–50
Budinich P and Trautman A 1988 The Spinorial Chessboard Penrose R and Rindler W 1984 Spinors and Space-Time,
(Trieste Notes in Physics) (Berlin: Springer) vol 1: Two-Spinor Calculus and Relativistic Field
Dirac P A M 1928 The quantum theory of the electron, Part I (Cambridge: Cambridge University Press)
Proc. R. Soc. A 117 610–24 —— 1986 Spinors and Space-Time, vol 2: Spinor and Twistor
—— 1928 The quantum theory of the electron, Part II Proc. Methods in Space-Time Geometry (Cambridge: Cambridge
R. Soc. A 118 361 University Press)
—— 1936 Relativistic wave equations Proc. R. Soc. A 115 van der Waerden B L 1929 Spinoranalyse Nachr. Akad. Wiss.
447–59 Götting. Math.-Physik Kl. 100–9

You might also like