Mathematical Proceedings of

the Cambridge Philosophical

Society
http://journals.cambridge.org/PSP

Additional services for Mathematical

Proceedings of the Cambridge
Philosophical Society:

Email alerts: Click here

Subscriptions: Click here
Commercial reprints: Click here
Terms of use : Click here

On Quantum Algebra

Dr P. A. M. Dirac

Mathematical Proceedings of the Cambridge Philosophical Society / Volume 23 /

Issue 04 / October 1926, pp 412 - 418
DOI: 10.1017/S0305004100015231, Published online: 24 October 2008

Link to this article: http://journals.cambridge.org/

abstract_S0305004100015231

How to cite this article:

Dr P. A. M. Dirac (1926). On Quantum Algebra. Mathematical
Proceedings of the Cambridge Philosophical Society, 23, pp 412-418
doi:10.1017/S0305004100015231

Request Permissions : Click here

Downloaded from http://journals.cambridge.org/PSP, IP address: 203.64.11.45 on 16 Mar 2015

412 Dr Dirac, On quantum algebra

On Quantum Algebra. By Dr P. A. M. DIRAC, St John's College.

(Communicated by Mr R. H. FOWLER.)
[Received 17 July, read 26 July 1926.]
For the purposes of atomic physics it has been found convenient
to introduce the idea of quantities that do not in general satisfy
the commutative law of multiplication, but satisfy all the other
laws of ordinary algebra*. These quantities are called ^-numbers,
and the numbers of ordinary mathematics c-numbers, while the
word number alone is used to denote either a g'-number or a
c-number. Both ^-numbers and c-numbers can occur together in
the same piece of analysis, and even in the same equation, as
numbers of the two kinds can be added together or multiplied. A
c-number may, in fact, be regarded as a special case of the more
general ^-number. In the particular case when two numbers a; and
y satisfy xy = yx, we shall say that x commutes with y. A c-number
is assumed to commute with every number.
The development of the algebra of g-numbers has up to the
present been greatly hampered by the fact that there has been no
general definition of a function of a g-number variable. One has
been obliged to restrict the word function to apply only to analytic
functions specified by algebraic equations involving the operations
of addition, multiplication and reciprocation (which last applied to
^-numbers can easily be given a definite meaning), and one has
usually contented oneself with taking a theorem involving arbitrary
functions to be true generally when one has proved it true for
power series.
In the present paper a general definition of a function is pro-
posed which appears to enable one to establish theorems involving
arbitrary functions with rigour. The differential coefficient of a
function of a single variable can also be conveniently defined, its
most important properties being contained in, or easily deducible
from, its definition.
Functions of a q-number variable.
Consider a simple algebraic function of a single g-number
variable x,' such as x2. Evidently every number that commutes
with x must also commute with a?. This property provides us with
the general definition of a function of a g-number variable.
Let x and X be two ^-numbers such that any number that
commutes with x also commutes with X. We then say that X is
a function of the single variable x, and consider the two numbers
• Roy. Soc. Proc. A, vol. 110, p. 561 (1926).
 Dr Dirac, On quantum algebra 413
X and x as specifying a function. A function of a g-number
variable is thus specified by a pair of g-numbers that satisfy a
certain condition and is quite different from a function of a real
c-number variable, for the specification of which an infinite number
of pairs of numbers are required, which need not satisfy any
conditions.
In order that we may consider the two g-numbers x and X as
really specifying a function, we must be able to say what is the
value of this function for any particular value of the variable, y say.
We take the rule for this to be that if b satisfies the equation
y=bxb~> (1),
then Ydefined by Y=bXb~1 (2)
is the same function of y that X is of x. This rule completes the
definition of a function of a single ^-number variable. (It is taken
as an axiom that there always exists a ^-number b satisfying (1)
when x and y are ^-numbers.)
In order that our definition of a function may be tenable, one
must be able to deduce the following theorems:
(i) All values of b that satisfy (1) must give the same
value for Y defined by (2).
(ii) Y defined by (2) must be a function of the single
variable y, i.e. any number that commutes with y must also
commute with Y.
(iii) Functions of a single variable can be formed by the
addition or multiplication of functions of that variable.
(iv) A function of a function of a certain variable must be a
function of that variable.
To prove (i) we observe that if b' is another root of (1), we
have
bxb~* = y = b'xb'~\
whence, dividing by b in front and multiplying by b' behind, we
get
xb~1b' = b-1b'x.
Thus b~*b' commutes with x, and therefore also with X, since by
hypothesis every number that commutes with x also commutes
with X. Hence

so that, multiplying bv b in front and dividing by b' behind, we

find

from (2), which proves the theorem.

414 Dr Dirac, On quantum algebra

To prove (ii), let a be any number that commutes with y, so

that
ay = ya,
or abxb~* = bxb'1 a,
from (1). This gives b~iabx = xb~1ab,
so that b~*ab commutes with x, and hence also with X. Thus

which gives abXb~l = bXb~la,

or aY= Ya,
which proves the theorem.
To prove (iii), we observe that if Xt and X, are two functions
of the single ^-number variable x, then Xj + X2 and X-LX2 are also
functions of the single variable x, since any number that commutes
with x must commute with Xj and X2, and therefore commutes
with Xj + X 2 and Xt X2. To complete the proof that functions of
a single variable may be formed by the addition or multiplication
of functions of that variable, we must show that if Yx and F2 are
the same functions of y that Xx and X2 are of x, then Fa + Fa and
Fi Y2 are the same functions of y that Xt + X2 and XXX2 are of x.
By hypothesis we have

and hence Y1+Y2 = b (Xj + X2) 6"1,

F, F2 = 6Z.6-'. bX.b-1 = bXlXtb~\
which proves the theorem.
Finally, to prove (iv), we see that if X is a function of the
single variable x and £ is a function of the single variable X, then
any number that commutes with x must commute with X, and
hence must also commute with f, so that f is a function of the
single variable x. Also, if F is the same function of y that X is of
x, we have
y = bxb-\ Y = bXb~\
and if in addition 17 is the same function of Y that f is of X, we
have
V = &?&-'.
It follows that T) is the same function of y that £ is of x.
Theorem (iii) shows that algebraic functions denned by the
operations of addition and multiplication are still functions in the
new sense of the word. Also it is easily verified that or1 (defined
by x~*x = 1 or xx~x = 1) is a function of x, and hence from (iv) the
reciprocal of any function of a; is a function of x. Any function of
 Dr Dirac, On quantum algebra 415

a single variable must commute with any other function of that

variable. A c-number may be regarded as a particular case of
a function of any ^-number variable.
The extension of the theory to functions of more than one
variable is rather restricted. We define X to be a function of the
variables xlt x2, ... xn only, if X commutes with every number
that commutes with xu x2, ... xn. If we now have
yx = bxj-1, y2 = bxjr1, ... yn = bxnb~1 (3),
and put Y=bXb-\
we can define Y to be the same function of the variables y,, y 2 ,... yn
that X is of the variables xiy x2, ... xn, and can easily deduce the
theorems corresponding to (i)...(iv) for functions of several
variables. It is, however, only when the y's satisfy the conditions
(3) that one can give a meaning to the expression "the same
function of the y's that X is of the x's," owing to the fact that one
cannot consider the x's to vary independently, as one could if they
were c-numbers. Suppose, for instance, that xY and a?2 satisfy
XiX, — X,Xt = 1.
Then one would have to count x^x2 and (1 + x2x^) as two ways of
writing the same function of xY and x2; but one would not in
general have y^y% equal to (1 + y2yi), except when the y's satisfy (3),
so that it is only in this case that one can hope to give a meaning
to the same function of the y's.

Differential coefficients.
Let Q be a function of the single variable q, and let p be a
number that satisfies
qp-pq = l (4).
Put Qp-pQ^Q' (5).
Then we define Q' to be the differential coefficient of Q with
respect to q, and write it dQ/dq.
In order that our definition may be tenable, one must be able
to deduce the following theorems:
(i) All values of p that satisfy (4) must give the same
value for Q defined by (5).
(ii) Q' defined by (5) must be a function of the single
variable q.
(iii) The ordinary rules for the differential coefficient of a
sum and a product must still hold.
(iv) The ordinary rule for the differential coefficient of a
function of a function must still hold.
416 Dr Dirac, On quantum algebra
To prove (i), we observe that if p is another root of (4), we
have
qp-pq = l=qp'-p'q,
or q(p-p') = (p-p')q-
Thus (p —p') commutes with q, and therefore also with Q. Hence

or QP'-P'Q = QP-PQ = Q,
which proves the theorem.
To prove (ii), let a be any number that commutes with q. It
must therefore commute with Q. We have to prove that it com-
mutes with Q'. From (4), since (qp — pq) is a c-number, we have
0 = a(qp-pq) - (qp -pq) a
= aqp — apq — qpa + pqa
= qap — apq — qpa + paq
= q (ap — pa) — (ap — pa) q.
Hence (ap — pa) commutes with q, and therefore also with Q, so
that
0 = Q (ap -pa) — (ap -pa) Q
= Qap - Qpa - apQ + paQ
= aQp — Qpa - apQ + pQa
= a(Qp-pQ)-(Qp-pQ)a
= aQ' - Q'a.
Thus a commutes with Q'. To complete the proof of (ii), we must
show that if Qx is the same function of qx that Q2 is of q2, then Q/
is the same function of qx that Q2' is of q2, i.e. if

and Q, p! -piQi= Qi, Qtp2 -PiQi = QJ,

then Q2' = 6Q/6-1.
We have 1 = qz p2 - p2 ft = &?! b~lpt - p2 bq1 b~\
from which, by multiplying by b behind and b-1 in front, we
obtain
1 = qjr^pj) — b~lp2bqx.
 Dr Dirac, On quantum algebra 417
This shows that b~*p2b may be taken to be p,. Hence
Q,'=Q.fr-Ij».&-&-1ft&Qi,
or bQ.'b-1 = bQ1b~ip2 -

which proves the theorem.

To prove (iii) let & and Q2 be two functions of the single
variable q. It is obvious that
(Qi + ft)' = Q»' + ft'-
Further (Q,Q3)' = QiQ*P~ pQiQ,
+ Q. (ftj> -

The ordinary laws for the differentiation of a sum and a product

thus still hold.
To prove (iv), let Q be a function of the single variable q and £
a function of the single variable Q, and let
qp-pq = 1,
and QP-PQ = 1 (6).
We then have by definition
Qp-pQ = dQ/dq (7),
fr-pS=dE/dq (8),
and ZP-P£=dZ/dQ (9).
Now dQ/dq must commute with Q, since they are both functions
of the single variable q, and hence, dividing (7) by dQ/dq in front,
we obtain

This shows that (dQ/dq)'1 p may be taken to be the P of equation

(6). Hence, using this value for P in equation (9), we obtain

using (8) and the fact that {dQ/dq)~l commutes with f. Hence
dl = dQ d |
dq dq dQ'
which is the ordinary rule for the differentiation of a function of
a function.
From (iii) and (iv) it follows that the differential coeflScients of
simple algebraic functions have their -ordinary values. In order
VOL. XXIII. PART IV. 28
418 Dr Dirac, On quantum algebra
that the converse process of integration applied to simple algebraic
functions may also lead to the usual results, it is necessary and
sufficient that the only functions whose differential coefficients are
zero shall be constants (i.e. c-numbers). This requires that the
only numbers that commute with p and also with every number
that commutes with q, when p and q satisfy (4), shall be c-numbers.
We take this to be an axiom in the new algebra. It is a particular
case of a more general axiom which may be enunciated thus: If a
number commutes with every number that commutes with the
^-numbers q, qu q2, ... qn and with one number that does not
commute with q but commutes with qlt qz>... qn, then it commutes
with every number that commutes with qlt qiy ... qn*. The axiom
may also be stated geometrically thus: Each 5-number may be
represented by a point in space of an infinite number of dimen-
sions in such a way that the condition that a g-number shall
commute with certain given ^-numbers is that it shall He in a
certain flat space.
The definition of a differential coefficient may be extended to
the partial differential coefficients of a function of several variables
that commute with one another. If Q is a function of the
variables q, qlt qit ... 'qn that commute with one another, to
define dQ/dq one requires a p that satisfies (4) and that commutes
with qu qit ... qn. One can then define dQ/dq by the left-hand side
of (5), and can show that it satisfies theorems corresponding to
(i)...(iv). It is impossible, however, to extend the definition to the
partial differential coefficients of functions of several arbitrary
variables. Suppose, for example, that #, and x2 are ^-numbers such
that

where c is a c-nuraber; then xtx2 and x2 (xl + c) must be two ways

of writing the same function of xx and x2, and it would obviously
be impossible to give a rational definition of the partial differential
coefficient of this function with respect to x2.
* Note that a number x that commutes with every number must be a c-number,
on account of the axiom that if x is a {-number, there must be a {-number b that
makes bxb~l equal an arbitrary ^-number.