Professional Documents
Culture Documents
Weyl, Robertson - The Theory of Groups & Quantum Mechanics
Weyl, Robertson - The Theory of Groups & Quantum Mechanics
,
s(t) = Ulr which the state vector z undergoes in time 1
Since |x,{8)|tis constant, the probabilities for the various energy
rabies do nat change in the course of time. The finite law
X() = UQXO-NY (64)
for the dependence of the statistical state X(i) on the time ¢
is fully equivalent to the diferential lav (8.2).
The mean value g— q(t) of the physical quantity represented
by the fixed Hermitian operator Q :
alt) = te [XQ -0}
can, on taking into account the symmetry properties of the
trace, be written also in the form
a = 6 EX 90]
2) = U-ynQUI. (@)
Consequently the situation can be described either by con-
sidering Q as fixed for all time and the statistical state X(é) a3
‘varying with the time in accordance with the law (8.4)—and
this is the fundamentai stand taken by quantum mechanics—
or we can take the initial state X as representing the state of
the system for all time and allow the operator Q(f} representing
the quantity Q to vary with time in accordance with the law
(8.5). This latter interpretation lends itself to comparisan with
classical mechanics, (86) is equivalent to the differential law
where
ha
342 _ uo on, (28)
for in virtue of (8.9) and (8.6)
a aX a
dae (G0) = (x-2)#2 QUANTUM THEORY
In particular, the quantity Q is constant in time, i.e. the proB-
abilities associated with it do not change in course of time, if the
Hermitian form Q which represents it commutes with H.
Tn Heizenberg's co-nrdinate system equation (8.8) becomes
downll) — dean = & Sono, (8.7)
The matrix O(t) is thus expressed in terms of components per
forming simple oscillations with frequencies ¥mq— vy, Ihe
corresponding amplitude is gq. On going over from the me'*
ta the n® stationary state the system loses an amount h{vq, — Yn)
of enerey; if this energy is radiated as light, its frequency
is given by
Chassival anech 6
a fixed level m to all possible levele = 1, 2, +
State of motion, the motion of the system in the mt quantum.
state, whose harmonic components have the corresponding
ramsition frequencies Yq, Yeu) "~~~ For any quantity A it
Urcrefore assuciates a constant amplitude dp with the transition
m-—>n. But int classical mechanics (for systems with one degree
of freedom) we have
Yan = holt), k=m—2,
instead of equation (8.8). On multiplying the two Fourier
series A, B
Foxe and Eby
we ublain the Fourier series C with coefficients
a= Sab +s=h).
Accordingly classical mechanics associates with the quantit:
C= AD the amplitudes * ”
Con = Em mr Bm ane (7 +S = m—n), (8.9)
whereas quantum mechanies assigns to it the amplitudes
Fun = Eat bin = Lib, mor" bars ns (8.10)
‘The difference between these two results lies ‘in the fact that in.
(8.9) both factors a, b have the first index m in common, whereas.
in (8.10) che first index of 0 1s the same as the last index of a.
This is in exact analogy with the difference between the "classical **
and the correct Rite-Rydberg combination principle. This was
Heisenberg's starting-point; the correct combination principle
indicates the pertinent-fact that the rule (8,9) for the multi -
lication of amplitudes must be replaced by (8.10). Admittedly
THE DYNAMICAL LAW 83
such multiplication is not commutative, and st collects together
amplitudes which the older model assigned to different orbits.
We donate [anq|? 28 the intensity nf the quantity A in the
transition m=». When multiple energy lovele occur ("de
‘n for which Eq ~ E" and all dices m for which Ey = 2",
has an invariantive significance; in such a case this sum is
taken as the intensity of A in the transition B’> 2". It Ay
is that portion of A in which §R(E’ intersects §(E”) the sum
Ucfined above is the trace of yd
Consider an atom with one ot more electrons and let t be
the vector from the nucleus to a representative electron. ‘Then
q=et, or in case there is more than one electron the sum
= Sr, extended over the varias electrons, is the electric
dipole moment of the atom. In classical electrodynamics the
intensity of the light of frequency v emitted by the atom is calculated
from the amplitude q() of the harmonic components of q with
the same frequency y in the following manner. The rate at
which energy flows through a surface element do at the point P,
whose distance from the atom at 0 is large compared with the
wave-length, is given by
oye
Ite) gear
where 4! i the component of 4 perpendicular to OF and de is
the solid angle subtended at 0 by'do, We have turther assumed
that the wave-length under consideration is large compared with
the radius of the atom. Since each photon of frequency
carries with it energy hy, we postulate that this law ie to be
taken over into quantum theory as follows: the probability
that an atom in state n goes over into slate n’ in unit Gwe and
emits 2 photon of frequency », whose direction lies within the
solid angle du, 3s given by
Val 5
We thus arvive at a definite rule for the calculation of the intensities
af the tines omiited hy the atnon The fact that wa can naw mabe
such a prediction indicates distinct superiority of the new
theory over the old. In particular, the transition n—r n' does
not occur if the corresponding coeficient in the Hermitian form
(nn)
By fn ea ta ete yoo =A ona In be basen
ads ae86 QUANTUM THEORY
Jor q is sero, This constitutes the general selection rule, The
connection between the state of polarization of the emitted
light and the direction of oscillation of the electric moment is
also carried over into quantum theory. But a real derivation.
of our intensity rule can naturally only be obtained by con-
sidering the question of interaction between the atom and the
ether ; see §13.
Examples: 1. The Oscillator.
‘The Hermitian form
Vek) ea,
representing the co-ordinate x of the oscillating particle has,
as we have already found [(8.11)], the coefficients
aw =0 if wnt; }
aa stl
dana, tna HED } (3.12
with respect to Heisenberg's co-ordinate system, in which the
energy is referred to its principal axes, We thus obtain the
slection rule n-> 7 +1; the quantum number n can only change
bby 4 1, the oscillator then absorbing or emitting a photon of fre~
quency'v = w and energy hes, in accordance with (3.10). The
Selection rule makes it clear why no higher harmonics arc ex-
cited in the simple oscillator. We have also found that the
matrix ([Paq{|, which represents the linear momentum in Heisen=
berg's co-ordinate system, is given by (3.12)
—_ 1 fime: 1 limo FT
Pee Poot VESEED | Han
Pay =0 for wen kl
2. Electron in spherically symmetric ld.
‘The result (4.7) for surface harmonics yields the selection rule.
Irth (8.14)
for the asimuthal quantum number t; for} = 0 only the transition.
O— Lis possible, On introducing the magnetic quantum numbex
m as in §4, the characteristic functions yl) depend on the:
meridian angle & shant the veciuenveel
factor im; here il
ahiy=rsin de, g—7 008 8,
THE DYNAMICAL LAW 85
In order to cbtain the dependence of the matrices 9. + iz»
q2 — gy, ge on the transition m > m’ we must evaluate the
Integral
Jelag) e(— md) elm’d) a4,
where a=1, —1, 0, respectively. The integral vanishes
unless m!-+ a= m. The only components of 9. + ig, which do
not nanish ave thove corresponding to the transitions m—>m — 1
in which the magnetic quantum number decreases by 1; for
Yx— iy, m—r m+ 1, for yy m—> me.
This last selection rule cannot be obtained from the spectra
themselves as long as the terms corresponding to dilterent
values of m (|m| <0) coincide, But these terms are broken
up into their various components by a homogeneous magnetic
field in the direction of the z-axis (Zeeman effect]. On “ longi-
tudinal " vbscivativu uf the light cuiliod ia the edievtiua we
find instead of tite one line (1, 1) > (w’,[) several left- and right-
circularly polarized components, the former of which arise from
the transitions m—>m—1 and the latter from m—> m-+ 1.
On “ transverse” observation, eg. along the y-axis, we find
trvo tranoveree linearly polarized lince arising from m > m 1
and in addition a longitudinally (i.e. along the z-axis) polarized
line corresponding to the transition m—> m, (Polarization as
here used means the direction of oscillation of the electric dipole,
and therefore the direction of the electric field strength.)
In the term spectrum of the alkali elements, which is, however,
typical in this respect, even for the more complicated spectra
of the other elements, ‘we distinguish between several series by
means of the letters §, p, 4, f, g) + **. Each series consists of
infinitely many terms Which We number in the direction of
increasing frequency by the integer m. It is found convenient
to let m run from T on in the s-series, from 2 on in the fs
from $ in the d-eeries, etc. The valuee of the terms ns, np,
nd,~ + + ave then given by the “ hydrogen-like " formula
wre
in which = ry, xp mg, + is a correction tecm depending but
slightly on m, the numerical value of which but rarely exceeds
1/2 and is very close to 0 for high series (f.g. .. J. Omly terms
lying in neighbouring series combine to produce a line, i.e. an
s-term combines only with a p-term, p only with s and d, d with
p and f, ete. In particular, the transitions np > Is give rise86 QUANTUM THEORY
the principal series, which also appears in absorption, nd -» 9
ee ee Ge dice sorts, me Op to the sharp sree,
and nf-r 3 to the Bergmar series 3
‘The alkalies are univalent, ie. ia chemical reactions only
‘one electron, the valence electron, plays @ role; the others,
together with the nucleus, constitute an inert closed shell. It
ie therefore reasonable to’ assume that the optical spectra of
tthe alkalies are caused by quantum jumps involving only this
Valence electron, while the core A* remains in ite normal state,
We have seen above that hydrogen in the normal state ie re
presented by a sphetically symmetric wave function ¥; we
Eheretore assume, disregarding the reaction of the valence
electron on the core, that this feature of the core being " closed "*
is to be expressed by ascribing spherical symmelry to it." We
hhave then to deal with the problem of an electron in a spherically
aymmetsc feld, which we have already diseussed ahove. Tn
accordance with the empirical combination principle and the
theoretical selection rule for the azimuthal quantum number J,
the 5, p, d, f, + + terms are to be taken as having I= 0, 1, 2, 3,
++ respectively. then runs from + 1 on in the series with
azimuthal quantum number /, as in hydrogel
§ 9. Perturbation Theory
‘The problem with which perturbation theory is concerned is
the following : Let the energy H consist of two terms H=H-+eW",
Uke second of which, Use pertw bation erm el, iy sinall compared
with the first; this we express by the " infinitesimal ” numerical
constant €, of which powers higher than the first are to be
neglected, "Assume that the quantum problem for the "un
pperturhed system with energy W has already heen solved, 50
that the Hermitian form H has already been brought into
normal (diagonal) form, and let $¥, 8, ~ - - be the chavacter-
istie spaces of 27 with chavacteristic nuutbers £", 2", The
problem is to find the soiution of the equations for the “ per-
turbed system" with energy H.
In order to illustrate the typical difference between degenerate
and non-degenerate systems we first consider the system epace a
2. inatead af en-dimencinnal ; then i
Bi FA)
i | O £.
sng WAY He and not H is the rt closed atom is only to be understood as
| +e.
PERTURBATION THEORY 87
Hf By ++ Dy the unitary transformation which brings H into
diagonal form differs from the identity only by terms of order &.
Consequently the probabilities jr,[4, jza[* that in the pure state
H has the values &,, 2, will change only by amounts of the
same relative order €; they remain constant to the same ap-
proximation with which sl” may he neglected in camparisnn
with H. But the situation is quite different for degenerate
systems, for which E, = Ey —E, for the principal axes of H
are Ulen indeterminate aud this’ arbitrariness. is expressed ia
the “instability " of the system under the influence of a per-
turbation, We set up that normal co-ordinate system e', ey’
in which W assumes the diagonal form; the co-ordinate vectors
are then also characteristic vectors of 'H, since Ey =, But
these vectors can obviously differ arbitrarily from the sriginal
co-ordinate vectors €, €, whereas the encegics ly’, hr’ can oalyy
differ from E by a termof order ¢. On returning to the original
co-ordinate system we have
y= aye — ryt) + aaa ef — veld),
Fa = ag * e(— vy") + dans e(— ve't),
where a = (a, 4), = (4 @n) are two mutually per-
pendicular vectors whose directions coincide with those of e,t.
he probabilities for the tun states e,, e, vary periodically in
time with the emall beat frequency »y' — vy) (resonance between
slates €, e). Quantum states with the came energy are therefore
in resonance withone unother. The wagnivudes of the vampencats
of in the characteristic spaces W, 9”, * - +, ie. the probabilities
for the various numerically ditferent values of remain ap-
proximately constant under a small perturbation, but this is
not the case for the absolute values |2,| of the individual com-
ponente x, resolved along the axes of an arbitrary Heisenberg
Tn accordance with the foregoing we can formulate the
perturbation problem in two forms: I. Detetniae the change,
due to the perturbation, in those states in which the energy
H of the unperturbed system is determinate, ‘This formulation
hhas a sound physical interpretation if we consider the perturba-
tion as acting during a time interval f,, t, We then find how the
probabilities for the various quantum states change wader the
influence of the perturbation lL Determine the quantum
states and energy levels of the perturbed system, i.c. the chac-
acteristic values and characteristic spaces of H. We ask in
particular how the terms are broken up and displaced under the
perturbation. We consider TI frstcy QUANTUM THEORY
‘We first decompose the Hermitian form W into two parts:
W,+V. To the first belong those portions of W in wi
a characteristic space ®Y, M”, «+ = of H intersects itself, and to
¥ those in which two different characteristic spaces intersect.
If the characteristic values of H have but finite multiplicity
tthe problem of bringing W", that part of W in which W intersects
iteclf, into diagonal form deals only with the space ® of a fnite
number of dimensions. If 8 is not simply a one-dimensional
space, the resonance phenomena mentioned above will appear,
‘The co-ordinate system, consisting of characteristic vectors of
H, is now more precisely specified, for now W, also appears as
a éiagonal matrix; let E, be the characteristic values of the
H+ eW) = Hp so obtained. The single term value E’ asso-
ciated with has in general been resolved into as many different
characteristic values E, of H, as there are dimensions in the
sub-space ®
‘The remainder V = |loms|]| of the matrix is such that tag = 0
if the characteristic values Bq Ey of HY are equal. The in-
finitesimal unitary rotation
Seme-Cx, C= |lemnll
of order # transforms H into H+ BH where
3H = e(HC ~ CH) ~ (HC — CH.
‘On choosing this transformation in such a way that 6H = — eV,
H = Ha-+ eV gocs over into He: this can be accomplished by
choosing Cam = 0 if Em = Eq anc
otherwise. The characteristic values Eq of Hy are therefore the
energy levels ofthe perturbed system of energy H if we neglect terms
of order 8.
W, can be considered as the time mean of the perturbation
wagsd over the motion of the unpeituibed ayate
For by (87) the mean value of the element aeq() of the matrix
A(d), which represents an arbitrary physical quantity of the
system, is aq, oF 0, according as vq =, or not. In statistics
ongilar bracbets ate
a quantity ; we may therefore write
We= W>, Ho= HD.
The solution of TI naturally provides an answer to. the
question i. But it is more convenient to employ the method of
variation of constants for the éalculation of the effect of the
imited time interval—the smaller the
THE PROBLEM OF SEVERAL BODIES 89
constant ¢, the longer we may take this time interval to be,
Asoume that at time ¢ = 0 the system is in the quantum state
0 and that the perturbation bogins to act at this time; “we ask
for the probability that the system will be found in the state
nat time £ That is, we seck that solution of the equations
which reduces to
Pree
attimet=0. Writing
the eyuations for €, are
Lent eekncinn rs
for ¢=0, &
the initial conditions
f=1, Asha =0
as the 0! approximation; on substituting these values in the
equation we obtain as the first approximation
Neglecting terme of order <#, we can take
Wy = me
a=52 #y).
On setting » — ¥4 =», the desired probability is
fi — cos i]
— See nal @)
It is to be noted that in accordance with this result the probability
of transition from state 0 to state m is determined by (H,,|2. Th
the case of resomance (vq — vq) the transition probability: in
‘creases at first with the square uf 1;
fralt = (j) = [PPaelt#.
§ 10. The Problem of Several Bodies. Product Space
A physical system consisting of two particles of masses m, m',
co-ordinates xy; x's’ and linear momenta p, p’, has as
its Hamiltonian function
= helt eaten + ert ete)
+Peys x72), (104)90 QUANTUM THEORY
where Vis the potential energy. We assume, as in the older
physics of central forces, that we are here dealing with an action
at a distance so that the potential energy depends only on the
imultaneous positions of the two particles. This assumption
rurally break down when, in accordance with the theory of
relativity, we take into account the finite velocity of propaga
tion of the disturbance, which requires the introduction of a
field. ‘The wave function of the system will depend on all
six co-ordinates xy; 2’/'a' in addition to t: the operators
corresponding to these functions in the domain of such functions
4 are multiplication by #, +++; 7, +++ and to the linear
A
i
momenta cmap he deta #2. oo, BS
From (10.1) we then obtain the wave equation
Ao ® A,
Fit mbt ge HAV t=. (10.2)
We must ask for the probability that the one particle is wo be found
at a point P and, simultaneously, the other is to be found at a
point P', The probability density is accordingly to be computed
for a G-dimensional space with co-ordinates xy 2; x’ y’ 2
Tadeed, the wave Bald fe not to rep i
taking place in physical space, but is to determine the appear
ance at definite positions of with definite energies and momenta ;
there is consequently nothing absurd in the fact that its medium
4s this abstract 6-dimensional configuration space.
In order to be independent of the special procedure by which
the scalar wave mechanics puts together two systems @, B to
form a single system ¢, as suggested by this example involving
the Hermitian forms representing the co-ordinates and momenta
of the two systems, We must first discuss the multiplication of
spaces from a purely mathematical standpoint.
‘With each vector z= (x) in a space i of m dimensions and
each vector = (94) in a space © of w dimensions there is
associated a vector j= ¢ x 9 with components
a= 20s (10.3)
in an m: ndimensional space E— R x G; the product space.
The components are here numbered by means of the index
pair (ik) = 1. The totality of vectors 3= 2 x 9 do not them-
SG conatitue a linear manifold, but thee linear combinations
entire product space £, With the linear correspondenc
Aio Rand Bin S: ia
we Favete ve = Zhan ye
THE PROBLEM OF SEVERAL BODIES a1
is associated a linear correspondence C= 4 X Bin:
Ip = Edeuben tie
th
ae Devity cor arbyn (I (ib), = 72)
t
Naturally, to this multiplication corresponds the law of com
position
(4 x B)lay x By) = (Ady x BB),
where A, Ay are correspondences of on itself and B, By are
correspondences in GA coordinate system in §R and one in
G together determine a co-ordinate system in T; if the co
ordinate system in §R is subjected to the traneformation 4 and
that in G to the transformation B, thea the co-ordinate system
associated with them in £ undergoes the transformation A x B.
In accordance with the equation
dla) = dev ye + mee dyn,
to the infinitesimal correspondence IT in , 7 in © corresponds
the infinitesimal correspondence i
(Hx A)+L x 7) (10.4)
in &, where 1,, 1, denote the unit matrices in , G, respectively.
All of the foregoing 1s applicable to arbitrary vector spaces.
‘When & and © are both unitary spaces, then & is also, for by
(10.3)
Sin = Shae Dh
is an invariant if Diy, Day, are; ci x Bis unitary if a aud
Bare
‘Accordingly, two physical systems @ and b are compounded
to form a total system ¢ as follows. ‘The system space & of ¢
is Bt x G, where W is the system space of @ and © of B. Let
the arbitrary physical quantity x in § be represented by
the Hermitian form A; on replacing all these forms A by
Ax 1,, where 1, is the unit form in an arbitrary space ©, there
exist between these latter exactly the same relations as between
the A, s0 that from the solution of a quantum problem in
there arises a solution for the corresponding problem in R x S,
but there oxists no real distinction between the two. In the
system ¢ obtained by composition we have therefore to as-
soviate the Hermitian form A x 1, with a quantity a of a and
1, x B with f of 8 where 4, B are the forms associated with
a, B in ®, G, respectively. The totality of quantities of the
composite system ¢ is obtained by starting from the quantitiesee QUANTUM THEORY
Uelongiog to the component systems a and Wand multiplying
and aidfag them together in all posible ways, "The quantities
a of a commute with the quantities p of 6, for
(Ax Ljlly x B) =A x B= (hy x BMA A
We refer to the content of these last two sentences when we
Bay that consists of two kinematically independent parts and 6.
The wo systess are dynamically independent if the energy
H of the composite system is the sum of the energies H), #®)
of the partial pysteme =
H= (HO x1) + x HO,
ann
The infinitesimal unitary ooerespondence 4.17 in the total
system space is then that one which is due to the infinitesimal
unitary correspondences 2-H, 2-H in the two original
system spaces [(10.4)]. If HO and H are both in diagonal
form, then H is also, and the characteristic numbers are given
by
Bem BP + BP or yao + oP hm (my)
If we have a pure state for the total system which is repre-
seated by the vector ¢ of absolute value 1 and components
Gq» and if Q = [gi is an arbitrary quantity in a, then the ex-
pectation of Qin the pure state ¢
Q) = Lqebwtacer.
‘This has the form (7.2) with
A= lll = 1ZFacrl
Alt) is the Hermitian form
Zizek?
in But we se from this that we are not dealing with a pure
state in a, for ay will not in general have the form agiy. Con
ditions which insure a maximum of homogeneity within ¢ need
not require a maximum in this respect within the partial system @.
Furthermore: if the state of a and the state of b are known, the
state of ¢ is in general not uniquely specified, for a positive definite
Hermitian form ||@j,, || in the product space, which describes
4 statistical aggregate of states ¢, is not uniquely determined by
the Hermitian torms
Fain Bam we
COMMUTATION RULES 98
to which i gives rise in the spaces 8, ©. In this significant
ence quantum theory subscribes to the view that “the whole
is greater than the sum of its parts," which has recently been
raised to the status of a philosophical ereed by the Vitalists
and the Gestalt Psychologists.
‘The kinematically independent parts into which a system
can be resolved need not be spatially separated, nor need they
even refer to different particles. We can, for example, resolve
a single particle, whose physical quantities can all be expressed
in terms of %, y, £; Pz, Py, Pa into three partial systems with
fundamental ‘quantities x, ps, py|o, pb» For quantitics
which belong to different partial cystems, for example a quaulily
which can be expressed in terms of x, p. alone and one which
is in torms of y, py alone, commute with each other in the sense
of maltix multiplitation.
In the perturbation theory we are usually concerned with a
system which consists of two kinematically independent parts
and which are almost dynamically independent. Disregarding
the interaction eH for the moment, let J. and hp. be the enerey
levels of the two parts, so that lve -- p.) are the energy levels
of the unperturbed total system. Gn writing in equation (9.1)
= (n, *} in place of 0 and s!— (w', r') in place of m, whence
we find as the probability that the total system goes over from
the state s to the state s’ during time t:
SiR ea UR
af) = fest pelt art, wry (008)
‘The probability that the first system will be found in the state
n’ after time ¢, the total system having been in the state s = (nr)
originally, is obtained from (10.5) by summation with respect
tor’.
§ 41. Commutation Rules. Canonical Transformations
‘The development of wave mechanics in §§ 1-3 went beyond
the general scheme of §§ 7 and 8 in that it employed certain
specific Hermitian forms to represent the co-ordinates and
momenta of the particle. We are now interested in seeing how
this can be formulated in an invariant manner, without recourse
to any special co-ordinate system in system space.
For the Hermitian forms 4, p representing a rectangular4 QUANTUM THEORY
co-ordinate and its associated momentum we postulate the
commutation rule
Ifthe ayatem has only one degree of freedom, these two quantiti
appear ts canonical oriaes in classical mechanics. All physical
quantities of the system are then functions of p and q; in order
to avoid complications we restrict ourselves to polynomials f in.
p and q, and assume, in particular, that the Hamiltonian function
H has this form, What are we to understand by the derivatives
Jp and fy of f with respect to p and q in this domain in which
$ and q are not commutative in multiplication? We should
in any case require that differentiation with respect to g should
obey the following postulates:
() =O =;
(2) F+ee g and (af),
8) Gdemhee thee
We see immediately that these conditions uniquely determine
the derivative of a polynomial f, unless they happen to lead to
contradictions. But that they do not lead to contradictions
ccan be seen from the fact that they are obeyed by the definition,
th f= fp Pf. (11.2)
(2) follows immediately from the enmmutation rule (11.1), and
the Linearity (0) of the process is evident, (8) is proved by the
formula
“Jo. where «is anumber ;
(gle — pl fe) = flep — pe) + (fp — pfle
‘which involves only the dis
of matrix multiplication. Si
~theh=A—-y. a2)
f ipa
—which obviously suffices to establish the corresponding result
for any polynomial f of p and g—and comparing it with the
formule (11.2) applied to the particular function H, we are led
to the familiar Hamiltonian equations of rlasciral markawice »
aaa taaiel ana)
C4
at
COMMUTATION RULES 9
It is a wniversal trait of quantum theory to retain all the relations
of classical physics; but whereas the latter interpreted these re
lations as conditions to which the values of physical quantities were
subject in all individual cases, the former interprets them as con
ditions on the quantities themselves, or rather on the Hermitian
matrices which represent them. This is the more significant
formulation which the new quantum theory has given Bohr’s
correspondence principle.
The commutation rule (11.1) is of a rather remarkable
nature, it 18 entirely impossible tor matrices in a space of a
finite number of dimensions, and it alone precludes the possi-
bility that in an co-dimensional space q (or p) have only a discrete
spectrum of characteristic numbers. For on referring g to its
principal axes
4= |\Qnn||) gon— au dan = (m+); P= |lPmnlly
the left side of the commutation rule has the components
Pala — gu), hence the main diagonal consists of nothing
hut zeros!” The question arises as to whether it can be con:
cluded from (11.1) alone that the forms representing q and p
can always be given the form
fever oe) ae, oe vhthew
for an arbitrary vector ¥ with components $x) on employing
‘an appropriate co-ordinate system in system space. We shall
see in Chap. IV, § 18, that, on introducing a certain irreducibility
condition, this is in fact the case.
On taking into account the three space co-ordinates g, and
their associated linear momenta p. (e= 1, 2, 3), we have in
place of the one commutation rule (11.1) the following :
Pabp— Babe = dude ~ Sofa — 0 for all «, Bs
h fl (@=—) (4)
Pate a= 78-1, Bo =l0 (eee. |
‘The same commutation rules apply to the case in which we have
several particles, the only difference being that then « runs
through 6, 9, + «'+ values, according to the number of particles,
instead of 8. These commutation rules are the necessary and
suficient condition that the dynamical law, which yovemns the
time rate of change of the state vector £ in system space, leads
tu the Hamiltonian equations for the “canonical variables
Ye, Po Tepresenting the co-ordinates and associated momenta of
the various particies composing the physical system—whatever06 QUANTUM THEORY
the dependence of the Hamiltonian function Hon these quantities
may be.
To classical mechanics the Hamilionian equations are invariant
with respect 10 canonical transformations." In a system of
Ff degrees of freedom the transition from a set of variables g,, Pa,
describing the state to a set gl, 92 (a= 1,2 +++, fis a
canonical transformation if the difference
Epa, — Epitas (11.5)
is a total differential. If, for example, the gs are subjected to
a transformation
sai ++ a)
among themselves, the ?. must transform as the components
of a “ covariant vector” in g-space in order that the whole be
& canonical transformation (“extended point transformation”)
4,
= rth.
Pit Pe
Perhaps the simplest canonical transformation is that in which
the roles of y and p are interchanged =
P= tn te= Pe
The canonical transformations constitute a group [cf. II, §1}. For
the identity, ie. the transition from (p, 4) t0 (Pq is & canonical
transformation; the inverse (p’, q')— (p, 9) of a canonical
transformation (p, 9)—» (p', 7) is also canonical ; and from the
canonical transformations (p, 2) (p', @’), (®, a) > (P" 4")
iE follows that the tesllant (lanafonaation (9, > (@", £°}
is also canonical, for if
Epads— Etattn Evade — Eade
ae total differentials their sum
Zpulds — Epallte
a=
is also.
‘An infinitesimal canonical transformation is one in which
qf siifer infiively iitite from p,q. We can consider it as
‘an infinitesimal deformation of the 2f-dimensional (p, ¢)-space
which takes place in the infinitesimal time interval e= 5. We
introduce the componente Bp, 8g of the displacement vector by
COMMUTATION RULES 7
Since (11.5) must be a total differential,
ZPde + Eyip, = aT (12.6)
must also; in our case T must differ only infinitesimelly from
Zpate We may therefore write
T= pg eS;
‘considering Sas a f
with (11.6),
ction of pa and gf we have, in accordance
as
or
a7)
Be
Since we may legitimately neglect terme of order of, we may
identify ¢{ with g, on the right-hand side of these cyuations.
We ca S the generating furition of the infinitesimal canonical
transformation.
In accordance with the Hamiltonian equations, the state
of a system, represented by a point (p, ¢) in (p, ¢)-space, goes
over into a state (p + de, g+ da) during time dt. If we follow
this transition for all possible initial states (p, 9) we otain an
infinitesimal deformation of the space whose points represent
the state of the system. The Hamiltonian equations ascort that
this deformation is an infinilesimal canonical transformation with
generating function I1-dt, Tt follows from this without any
calculation that these equations have a significance which is
independent of any particular choice of canonical variables.
Now in quantum theory the Hamiltonian equations (11.3)
assert that the state vector 5 in system space undergoes the
infinitesimal unitary rotation
a
ia
+H, (8)
s0 the infinitesimal canonical transformation of the quanti
Pq is here obtained by cubjecting the argument 5 in the Hi
‘uilian forms representing them to the infinitesimal rotation6 QUANTUM THEORY
land, in virtue of the commutation relations (11.4), this agrees
exactly with (117). On generating a finite canonical trans
formation by the successive application of an infinity of
finitesimal ones we arrive at the result that the unitary corre-
‘spondences of system space on itself in quantum theory .
v=Ut
correspond to the canonical transformations of classical mechanics §
more precisely, only those for which the matrix U is expressible
in terms of the matrices p, g, but we may for the present pass
‘aver the question as to whether every matrix J can be obtained,
for at least arbitrarily closely approximated, in this way. Since
the commutation rules (11.4) remain unchanged under rotations
of the normal co-ordinate system, they are valid for an arbitrary
set of canonical veriables. This is also evident from the fact
that they are the conditions that the dynamical law (8.1) lead
to the Hamiltonian equations
a
a
The general procedure for the quantum mechanical treat
ment of a physical system suffers from the disagreeable fact
that the expression for the energy in terms of the canonical
variables must be taken from the classical model, and in ad-
dition the transition to quantum mechanics is even then not
unique, for the model offers no means of telling whether a
monomital such as ply is to be interpreted as p%, pap, gp* or
a linear combination of all three [cf. IV, 5 14]. The provisional
character of such a procedure is clear, but the results so far ob-
tained seem to justify the hope that the path we have entered
upon will lead to a unique formulation of the laws governing.
the actual physical phenomena We need then concern our=
selves longer with the general mechanical scheme.
(11.3)
{12 Motion of a Particle in an Electro-magnetic
Field. Zeeman Effect and Stark Effect
Let the spatial co-ordinates x2 now be denoted by #2 5
andthe time! by %. If gis the scalar and cl the vector potential
of the electro-magnetic field, then in the theory of relativity
(= $, We, Hy, Ba) = (Ba, bs. be $5)
are the components of a vector in the space dual to the 4-di-
mensional world. Let
2p _ ae,
aaa Bagh
PARTICLE IN ELECTRO-MAGNETIC FIELD 99
Puy Fy Fy are the components of the electric field strength ©,
(Fes, Fay, Fyq) the components of the magnetic field strength 9.
Denoting’ the components ot the velocity of a particle by
U4, Up, Us, its proper time is
ds = Vd — (at | da} dade
av ect (ot = vt + + Ot
With the world vector w=? is associated the dual u. with
components
tau =1,2,3), w= Aut
‘The invariant equations of motion for a particle of mass m and
charge — e are
se)
a 1,28) (121)
The right-hand side is in fact the ponderomotive force
(E+ 2v6))
These equations arise from the Hamiltonian function
T= ety + eal + Zot obit, (12.2)
in which %,%2%); Ps P2Ds are the canonical variables, In tact,
the Hamiltonian equations
yield
Pit ey = muy |
in the remaining equations
dpe
UL fib, 2 Ub dates
ana g Sele aca ees
the left-hand side i ) 24, 2
domi (2b, de,
Oa et Ze 8)
But this is the desired equation (12.1)
Ma Ee wk z Ges _ 28)o,},
a Nb Oty) 72 Nae, ae100 QUANTUM THEORY
‘The negative energy — H is the time component py of the dual .
‘veetor whose space components are the components of linear
momentum p = (Ps, Ps, Ps), 80 the equation (12.2) can be written
in the rational form
Hat cba? — 3 oct ohh = mie,
From this we obtain the simple rule: Tike influence af am slectra-
magnetic field on a particle of charge — 1 can he exppresced by ve-
placing p. by ?. + eb. in the equations of mation for a free particle
On going over to quantum theory p. becomes the operator
4 2a is eamtagrin tthe cinenoalciglacemet
de, a8 is seen from the equation
HNN sAeae
urls iow On inedsing lf oi
a 2
aay Mt be replaced by Yh (22.8)
in the wave equation of the particle. Only $y has a simple physical
Significance; itis therefore to be assumed that the laws Which
govern ¥ remain invariant on replacing y by e+, where A is
any real function of position in space-time. On the other hand,
in the ciassicai theory of tne electro-magnetic held only the
field strengths, and. not the potentials, have an objective signif-
ie. the laws are invariant on replacing gy by ge —
where ys is also an arbitrary function of the x, On examining
our wave equation for these invariantive properties we find
that it is not invariant under each of them separately, but that
‘there must exist a certain relation between A and wu. The field
equations for the potentials } and ¢ of the material and electro-
‘magnetic waves are invariant under the simullaneous replacement
of
ka
here ) is an arbitrary function of the space-time co-ordinates.
‘This * principle of gauge invariance” is quite analogous to that
previously set up by the author, on speculative grounds, in
order to attive at a unified theory of gravitation and electricity #8
But I now believe that this gange invariance does not tie to-
B by ep and $. by be—
PARTICLE IN RLECTRO-MAGNETIC FIFLD 101
gether electricity and gravitation, but rather electricity and
matter in the manner described above. We shall discuss this
principle more thoroughly in Chap. IV; its significance and
its interpretation will then be more apparent.
On passing to the limit c—> oo in (12.2), after separating
out the factor mc, we return to ordinary mechanics :
Hx eb + A ylp, + of)
ot a
On neglecting torms which ars quadratic in the $4 we find, in
addition to the kinetic energy Epf/2n, the potential
Va—d +S im. (2.4)
We have already made use of the first part, that due to the
lectric field, in § 5, If we have, in addition to the field originat-
ing in the nucleus, a homogeneous electro-static field in the
direction of the 2-axis and of strength F, for which $= — F +2,
it adds the perturbation term
Waekon
to the energy. A homogeneous static magnetic field § is
obtained from the vector potential «= Fig}, ©= ( », 4);
this adds to the energy the perturhation term
sitter) =
ae thtoe)
de
68 |. (2.8)
Zeeman Bypecl—It the homogeneous magnetic hela strength,
ot magnitude jgj, 1s in the direction ot the s-axis, the per-
turbation term 1s
lo
ue
On choosing the characteristic functions yig? as our co-ordinate
system in the system space of the functions J, W7, as well as
the energy of the unperturbed atom, is in diagonal form ;
the state defined by wi, m it has the value
hoo m. (22.7)
he,
Who-Ly
(12.9)102 QUANTUM THEORY
The components (nl, m) —r (n,m), consistent with the selection:
rule for m, into which the line with frequency ¥ = 5 (Eq: — Ew’)
is broken up give rise to but three lines: one corresponding to
fll the transitions mm, which is linearly polarized in the
Sircetion of the # asds and is undisplaced ; one which is circularly
polarized perpendicular to the s-axis, the frequency v of which
Frdisplaced by -+o (me > m — 1); and one which is cireularly
polarized in the opposite sense, with frequency » —o instead
bev (m-—> m+ 1), This normal Zeeman effect is found only
in the so-called singlet lines, i i
‘Stark Effect—In accordance with the general perturbation
theory, the displacement and resolution of terms in tae presence
of a homogeneois electric field is determined, to terms of first
order, by the matrix
oF (s).
In consequence of the selection rule J+ +t 1, <#) = 0, unless
‘accidentally all energy levels whose azimuthal quantum numbers
differ by Leoincide, Ignoring this exceptional case, we should
expect to find no VW order perturbation effect increasing linearly
with the field strength F (linear Stark effet), but only a quadratic
effect, which is much smaller. This is in agreement with the
experimental data on alkali atoms. Hydrogen is, however,
degenerate, since for it energy levels with the same principal
quantum number and J=0, 1, *--, 2—1 coincide. The
caleulations for this case have been carried out by Schridinger
and compared with experiment.”
§ 13, Atom in Interaction with Radiation
Following Jeans, black body radiation is mathematically
equivalent to a system of infinitely many oscillators. Maxwell's
equations for the free ether are
divG=0, ci
ivE=0, cul G—
In order to simplify the relations, we assume that the walls of
the radiation cavity of volume '¥ are reflecting; then © is
perpendicular to the walls at the boundaries of the cavity.
Since the black body 15 at rest it is of no particular advantage
to carry through the calculation in a relativistically invariant
manner; we may therefore normalize the vector potential
ATOM IN INTERACTION WITH RADIATION 103
9{ in such @ way that the scalar potential vanishes. We then
igi
On the boundary 9 is normal to the walls, Let the characteristic
numbers and cliaracteristic functions of the equations
AM+5%=0, divA =o,
with the boundary condition that Wis there normal, be denoted
vo pe (EO Me [x-1,98,---},
normalized in accordance with
$(2.24)a" = tnd
7
On setting
Mare %,
‘where the coefficients g” depend on time but not on position, we
find for them the equations
a 2
ge
Inibacing Em pin ana the gs eguton i
that for an oscillator with Hamiltonian function
betters
we readily find on appl
Ea Trew, SacTor-cul
that the energy of the radiation field is in fact given by
1
~L[e.gur- on
n ale FOV = THe
with this we have proved the theorem due to Jeans. For high
frequencies p there are approximately
as.)104 QUANTUM THEORY
modes of oscillation in the frequency interval p, p+ dp.” We
are interested above all in the limiting case of an infinitely large
cavity ; the spectrum then becomes continuous and our formula
for the'density of frequencies becomes exact.
On quantising this mechanical system of infinitely many
oscillators # in accordance with the theory of the oscillator (§ 3)
and the process of composition (§ 10—but ef. remark on p. 108),
wwe find as possible quantum states s, each of which ischaracterized
by the fact that in it there is associated with each index @ an
integer m, 20. in this quantum state
1
Hy = healt 3)s
or, on choosing the additive constant in the energy in auch a
way that the lowest energy value which the black body radi
is capable of assuming is 0,
Ham tap Ho Dia how
Jn the language of photons this means that when the cavity
is in the state s it contains x, photons of each kind a The
matrix element
Gor [S= May May Mey
wnishes unless all the equations
Sm My My]
MS Mm, mm, mM, +
hold with the exception of n= n., which is to be replaced by
Nes mepl or nan
In the fret case we have, by eg. (2.13),
Gov = ED
and in the second
“ Ve (Absorption) (13)
The first transition s—> s consists in a photon of kind « springing
into being, the second in the disappearance of one such photon,
It follows from the above that in a transition for which qi ++ 0
all other gf, must vanish.
Let an'alom with fixed nucleus and electric dipole moment q
interact with the radiation field. Differentiate the quantum
states of tho atom from one another by means of the index »
and denote the corresponding energies by hv, ; then g
+1) (Emission), (13.2)
ie
ATOM IN INTERACTION WITH RADIATION 105
‘A quantum state of the total system consisting of both atom
‘and radiation is characterized by the quantum numbers
“The effect of the radiation on the atom is, in accordance with
eq. (12.4) of the preceding paragraph, given to a first approxima-
tion by the perturbation term
2 — Ga.
Jt can be shown that the addition of such a term to the
Hamiltonian function of the total system will, according to
classical theory, not only indicate an influence exerted on the
atom by the radiation field, but will also modify the equations
of Maxwell in 2 way which indicates that the motion of the
clectrons in the atom affects the radiation field. The per
‘turbation term will accordingly call forth emission as well as
absorption, To a sulficient approximation we may take for 2%
its value at the point occupied by the nucleus, provided we restrict
ourselves to radiation whase wave-length 15 large compared with
the dimensions of the atom, We now have
W = EGG) ¢. (13.3)
Prom this it follows than an clement ¢- Wy, gw can only differ
from 0 if s aud s’ are such that all nj = np with the exception
ot a single one nj, which must equal n, ++ 1. hen only the
a term contributes to the sum (13.3), and we have
Fg = (ban Me) in (3.4)
Bohr’s frequency condition, which asserts that the emission or
absorption of a photon in state a with energy fp, is associated
with a quantom jump of the atom in which an amount
“L hiva —v_) — Ap, of energy ic lost or won, need by no means
be satis hore. ‘The finite cavity has ite own frequencios pa,
aad may therefore be in no position to take up the frequencies
associated with the quantum jumps of the atom, This is true
in principle, but as a matter of fact, as we shall see, Bohr's
frequency condition is fulfilled to a very close approximation in
the overwhelming majority of all transitions ; and this is more
land more the ease the larger the cavity is
TLet the atom be in the state and the radiation in the
Ehrape = V+ Ulplde, (15.0)
where the sum on the left is to be extended over those indices
for which py lies between p snd p + dp; hence Uip)ép i the
energy density of the radiation contained in the frequency106 QUANTUM THEORY
range p, p+ dp.. In accordance with (10.8), the probability
that the atoin will Sind iteclf in the state »’ after time
by
03 Ya + pret
Se
‘The contribution to this sum due to the cases in which a photon
js emitted is, in accordance with equations (13.2), (18.4), given
bs
7 a 1 cot Cou = et Mire +1) |
Bz Gan =P pe
and that for absorption by
Dg L— 608 Pane + pall ite) og
Be Tew Fad Hye | Ew
Consider first the case in which the term level vq is higher than.
vai Yost =a Yay = —v is then negative. We now collect
together all those terms a in the sum (13+6,) for which p. les
between p and p-+ dp. Since the position of the atom is not
exactly fixed—even in consequence of the variations caused by
the emission of photons—we may, for small wave-lengths,
replace YE by its mean value xj as given by the normalizing
equation f%jd7 = 4m, and we may also assume that all
directions are equally probable for Of, The square |(.4)|* of
the scalar product of M with e Sxod vector g hes 4!
value 4%. gl, (19.6,) then becomes
sv
1 0s (p — v)t te line? Eh ta Pa
Coy 8
On introducing (13.8) the sum (
tion, be replaced by the integral
Ae [few l? i 1 = 008 (p
oe eo
Essentially the only elements which contribute to the value of
this integral, for a time £ large in comparison with the duration I/y
of an oscillation, are those or which plies near tov. On developing
Wa) (13)
i. (13.6,)
may, toa good approxima
in powers of p —», the first term in the expansion contributes
te
eh PF ae a at OY) (is.7)
ATOM IN INTERACIIUN WITH RADIATION 107
to the integral; all others are to Le neglected. Similarly the
entire amount (13.0,) due to emission is negligible, for its de-
nominator (p+ v)* vanishes nowhere. This means that the
transition is almost invariably associated with the absorption of,
‘a photon whose frequency lies very close to v, The probability.
that the atom will appear in the higher state n' after lapse of
time f increases in proportion with ¢; the factor
4a* UY) gt dat
3 (nls Sow? = ape
is the probability that the transition n> n' take place in unit time,
‘This formula was obtained for the case in which the state
n’ possessed a higher energy level than », In the reverse case
only the sum (13.6,) due to emissions contributes an appreciable
amount. We now put My — % — Me » and obtain the same
formula with this difference : in place of ny we now have s+ 1,
Ub) Gnwl?
or in place of the sum (18.5) the sam
Thon, + Up. = Thn.p. + The.
The first is I” Ulp}ap, and we denote the second by V+ u(phdp,
This latter is equal to (hp) times the number of modes of vibra
tion of the cavity within the frequency interval p, p + dp; hence
by (13.1)
telde
=
V -ulp)dp =V
The probabitity that the atom drop from state n into the lower state
nin unit time is given by
oY
SEL) + ull tow
‘The additional term u(r) is characteristic for spontaneous
emission. When the radiation is not enclosed in a black
body, i.e, when there is no radiation density Ul), the prob
Bility’ that the atom drop from the state n to the lower siate n’
in unit time, emitting thereby a photon wnose trequency lies
in the immediate neighbourhood of v = vx — Yn 18
4
Tae!
This agrees with the formula obtained by integrating (8.11)
over all direviiuus. The probability that the
the level m into a higher level n’ (rq > vq) under thes
conditions 1s zero.108 QUANTUM THEORY,
Tn the energy field of the black hady radiation wo fad mot
only absorption, but aleo “stimulated emission,” both of which
are proportional to the energy density U(y). Ou setting
Aan = 3 [awl*| (13.8)
ie
tthe probability for a jump from state m to a higher state n’ in
wunit time is
‘Wan = Aan UW) | =r —»,), (18.9)
and the probability for the inverse jump, the drop from #2”
toms
n= Ava) £ uO] (13.9)
Since [|4,e{] is an Hermitian matrix,
[4av = Ane (18,10)
If there are a number af atoms
whale eyetem ie in worage 33 many
tome must make the jump nn’ in unit time as make the
inverse jump n'->n, On denoting the number of atoms in
the state m by WV,, these considerations are expressed in the
condition
the radiation field and the
Ml Aan NaU) = Aura NolO + wb)
N, ls)
m+ 98. (13.11)
The probability coefficients Aww = Avy have entirely
appeared—or rather, almost entirely, for the equation is valid,
only under the assumption that yy £0 Of Guy #0, 1.6.
tise icausicion nz n° is not to De torbidden by the selection
rrules. But for such a system in thermal equilibrium NV, must
as shown by Boltamann, be proportional to i
CF hl
where 9 is the temperature and A the Boltzmann constan
‘Equation (13.11) chen becomes i
eoermen at
or the Planck radiation formula: te
we)
vo)
IN INTERAL H KAVIALIUN 109
this formula is valid for all frequencies » whuse energies can be
exchanged by the absorbing and emitting atoms in accordance
with Bohr's frequency condition.
‘We have thns finally retuned to the historical origin of the
oe sy. We 3
this treatment, duc to Dirac, of encrgy exchange between matter
and radiation.’ In the first’ place, it is able to explain the fact
that the spectral lines are not sharp, but possess a natural breadth.®
Secondly, we must inquire what causes this difference between
absorption and emission, procescee which are transformed into
each other on changing the direction of time. Indeed, the
fundamental mechanical and field laws ate invariant under the
transformation t+ —1! The answer is that this difference is
due to the preferential direction in time involued in the application
of tha theory of probability » we assume a fixed initial state and
caloulate, with # of transition probabilities, the distribu
tion over the various states at a Jaler time, not the distribution
which would result from the equations for an earlier time. If
no assumption is made concerning this preferential direction,
# should be replaced by {t] in (18-7). And finally, the fact that
‘we have here treated Maxwell's equations as classieal equations
‘of motion, and as cuch have oubjected them to the process of
‘quantization, may give tise to serious doubts—for in our general
formulation ‘Maxwell's equations are already the quantum-
theoretic wave equations for the photon! But we shall see
in Chap. TV, §11, that this method is in fart the correct one
ploy in order to go from one corpusele to an indefinite
number of corpuscles, For since the number of photons must
remain indefinite—as a photon can, in contrast Wo an electron,
spring into being or disappear—the method of composition
described in § 10 is not applicable to them,CHAPTER U1
GROUPS AND THEIR REPRESENTATIONS
§1. Transformation Groups
HE concept of a group, one of the oldest and most
profound of mathematical concepts, was obtained by
abstraction from that of a group of transformations.
A point-field, 2 domain of elements which we call points,
on which the transformations operate, underlies the trans-
formations, This point-field may be either the totality of a
finite number of individually exhibited elements or an infinite
set. in particular a continuum such as space or time. A
mapping or correspondence S of the point-field on itself is
determined by a law which associates with each point p of the
field a point p’ as image: p—> p’—= Sp; two correspondences
Sp and Tp are identical if for all points p the two image
points Sp and Tp coincide. If the point-ficld contains a finite
number of elements the correspondence S can be defined by
giving explicitly the image point for each point p; for infinite
sets, however, the association is only possible by giving the
law of the function S.
Among such correspondences there is a particular one which
associates with each point p the point p itself: pp; it is
called the identity I. Two correspondences can be applied
successively :, if the first sends the arbitrary point p into p’ = Sp,
the second p’ into p” = Tp’, then the correspondence resulting
from the composition of the’ two is defined by the association
2 > p" = T(Sp) and is denoted by TS (read from right to left !).
‘The resultant correspondence depends on the order of the two
factors § and T. In order that composition be possible it is
essential that the correspondences are ones which map the
point-feeld on iteelf, and not on another point-field,
We shall restrict ourselves to one-to-one correspondences S:
the image points p’ = Sp associated with p shail always be
distinct, and each given point p’ shall appear as the image of
one (and only one) of the points p. Consequently such a one-to-
110
TRANSFORMATION GROUPS um
ane correspandence S: pp" determines a second, the innerse
Sts pls p of S, which just cancels it:
Ss
The inverse of $-1 is again S and the identity Tis its own inverse.
The resultant TS of two one-to-one correspondences S, T is
itself une-to-one, and its inverse is (73)-*= SOT! —for
fon inverting the correspondences p> p'—> p” there resuits
p> p>. Henceforth we shall consider only those corre-
spondences, also called transformations or substitutions, which
are one-to-one. In this domain we have, in accordance with
what has been said, the two fundamental operations of inversion
and eamposition
Examples.—1. Let the point-feld consist of elements
exhibited individually; bring them into a particular order by
numbering them with the integers
AS nm ay
This numbering consists in a one-to-one reciprocal relation
between the elements of the point-field and the integers or
possible “positions” g in the series (1.1). A permutation con-
sists in the transition from one such arrangement to another.
Tf wie wich to operate in epare we may thinl: of the positions as
fixed compartments into which the movable elements can be
we may think of the clements as xed and
shift the movable numbers about. With each permutation is
ed a one-to-one correspondence p> p’ which tells
which element p’ occupies, after the exchange, the position
previously held by p. Insotar as the method of numbering is
considered as left to convention, the permutation is nothing
more than this one-to-one correspondence. The concept is to
he understood in this way when we are concerned with the
composition oF sureessive application af permutations
2. A kinematical example of 2 group is offered by the motions
of a space-filling substance, in particular those of a rigid body.
The positions ot numbers’ of the preceding example are here
represented by he miaterial points and the puint-field is the
space itself. The one-to-one correspondence p> p’ connects
the mitial with the final state: that material point which origin-
ally covered the spatial point is taken to the point 9’ by the
motion, Congruent correspondences of space on to itself will
also be briefly referred to as “motions” in the geometrical
The concept of a group of transformations is now readily2 GROUPS AND THEIR REPRESENTATIONS
formulated. We understand by it any system @ of transforma
tions of a given point-field, which is closed in the sense of the
following condition:
1. Terrontains the identity ;
2. If Shetongs to G}, then its inverse St does als
8. The resultant T'S of any two transformations S, T of @
is also a transformation of @.
examples we name the group of all »! permutations of
things, the congruent mappings or '' motions" of 3-dimensional
Euelidean space, all homogencous linear transformations ia
‘2 variables with non-vanishing determinants (affine correspond-
tence of an n-dimensional vector space) and the group of unitary
transformations in n dimensions,
If the point p goes over into p' by means of a transformation
uf the group ©, thea p” is said to be equivalent to p (with veapact
to the group @). ‘The same concept is applied when we arc
considering instead of a point p a figure consisting of points.
Expressed in these terms, the Ue requirements for @ group
are nothing other than the three axioms of equality =
1. pis equivalent to p;
2. If p” is equivalent to p, then p is equivalent to ";
3. If pis equivalent to p and p” to p’, then p” is equivalent
top.
‘Recording to Klein's Kslanger Lrogram * any geometry of
a point-field is based on a particular transformation group ©
of the field; figures which are equivalent with respect to ®,
and which can therefore be carried into one another by a trans:
formation of G, are to be considered as the same, In Euclidean
geometry this réle is played by the group of congruency trans:
formations, consisting of the motions referred to above, and
in affine geometry by the group of affine transformations, etc.
‘The group expresses the specific isotropy or homogeneity of the
space ; it consists ofall one-to-one “ isomorphic correspondences "
of the space on itself, i.e. those transformations which leave
undisturbed all objective relations between points of the space
which can be expressed geometrically. The symmetry of a
particular figure in such a space is described by a sub-group of
consisting of all transformations of @ which carry the figure
‘over into itself. The art of ornamental tiling, which was per-
fected by the Egyptians, contains implicitly considerable know-
ledge of a group-theoretic nature; we here find, perhaps, the
oldest fragment of mathematics in human culture. But only
recently have we been able to formulate clearly the formal
principles of this art; attempts in this direction were already
made by Leonardo da Vinci, who sought to give a general and
ARSTRACT GROUPS AND THRIR REALIZATION 119
systematic account of the various types of symmetry possible
in a building. But the most wonderful symmetrical structures
are exhibited in crystals, the symmetry of which is described
hy those congriency transformations of Fnelidean space which
bring the atomic lattices of the erystal into coincidence with
themsclves. The moot important application of group theory
to natural science heretofore has been in this field.
‘The following cousiderativus fie waiuraily iniv the present
discussion, Let the point-field M on which the transformations
'S of the group G operate be mapped on the point-field N by
‘means of the one-to-one correspondence A: p—>q; the case
in which the correspondence serves to introduce new numbering
or new co-ordinates is of particular importance. Through this
correspondence A of Man N the transfarmatinn S af M heroes,
a transformation T of N; in the particular case mentioned above
T ie simply description of the transformation $ in the new
coordinates. It is evident that to the composition of trans
formaiivus 3 wirespouds the Lumpusitiva uf the weespomdiag
transformations T of N and that a group © of transformations 3
goes over into a group 9 of transformations T, The relation
between these two transformations is
T= ASA, (1.2)
for if we denate the transformation S by p> 9’ and if 9. 9! are
the points af N associated with pp! hy 4, then the transforma.
tion ¢—> 4 of N is effected by
nee l Pie Pie
‘We may aisu write § = AGA, In particular, these considera:
tions apply when Nand M are the same point-field.
§ 2. Abstract Groups and their Realization
‘An arbitrary number of transformations of a given point-field
on to itself can be applied successively; we are of course not
restricted to merely two. But when we perform this process
step by step it ie automatically reduced to a succession of com-
ne talzen tan af a Hime»
[B(C +
‘This possibility of performing an extended composition in steps
involving but two Uansformations at a time shows that the
associative law
(AB)C = A(BC)
holds for any three transformations 4, B, C.114 GROUPS AND THEIR REPRESENTATIONS
The structure of a transformation group is obtained from it
by abstraction when we allow the transformations themselves
to degencrate into cements of an immaterial nature, retaining
two given transformations are compored, in a given order, tO
forts third, Ta accordance with what as Been said such.
Composition necesarly obeys the asyociative law. Perhaps 3€
iso obeys other universal laws, Dut since we have at presonst
no indication of this ve attempt a formulation of the abstract
Structure of the group by means of the following definitions:
"An abstract group ts asjsem of elements within which « laze
of composition 1s tlven such that by means 0 i there arises frorre
any ito (the same or diferent) elements a,b ofthe Rrowp, taken 172
{his order, on element ba. The follwing conditions shall thereb'y
he atta
T Phe associatoe la c(ba) = (8a:
2, Thee shall sett aw sleet 1, the omit somont,o
an arbitrary dlament a tmaltered on composition with it:
la a
8. To each element a shall exist an inverse a! which yields ov
composition with it the unit element 1:
ich lentie s
Such an abstract group is not to be confused with its reala'-
zation by transformations, i.e. by one-to-one correspondences of
a given point-feld. 4 realization consists in ascociating witFe
tach element a of the abstract group a transformation Ta) of the
point-feld in such a way that to the composition of elements of
the group corresponds composition of the associated transforma
Hons:
oe
T(ba) — T(e)T(a). (22)
At follows from this that to the unit element | corresponds the
identity J and to inverse elements a, a-* correspond inverse
transformations
Tort) = THe), (2.2)
‘The first assertion follows from the particular case
T(a)T(1) = T(a)
of (2:1) by left handed composition with the seciproval of Ue
transformation 7a); (2.2) is then contained in (2) a the
particular case )— a", “The sealization is said to be falchfur
when to distinct clements of the group correspond distinct
transformations :
T(@) = T(b) when a +b.
ABSTRACT GROUPS AND THEIR REALIZALION 115
In accordance with the fundamcatal equation (2.1) the nevessaty
ud sufficient wundition for “faiti(ulucss is Uhat T(a) shall be
the identity only if-a is the unit element. For if a, } are two
elements of the group it then follows from £(a) = Z(b), ue.
T(a)T-4) — T(a)\TW) = Tab) —
that under these conditions ab’ = | ue. a=b, If the abstract
group is obtained from a transformation group (by abstraetinn,
then conversely (is a faithful realization of it.
In the study of transformation groups we always deal with
two manifolds, the structureless point field and the manifold of
givup cicmenis, the siuciure uf wiki is expressed by the iaw
of composition, The original problem thus resolves itself into
two; the examination of the various group structures possible
and ‘the examination of the possibility of obtaining realizations
of the given abstract group by transformations of a given point-
field. The historical development of the subject has shawn that
it is advantageous to affect this division into two problem:
they are of fundamentally different charactor and require
fundamentally different mathematical equipment for their
discussion.
Tn accordance with our method of introducing the abstract
group, which we henceforth refer to simply as the group, it
serves merely to give the structure of the group; the nature of
its elements is immaterial. ‘This abstraction from the nature
of the elements is expressed mathematically hy the concent of
isomorphism Vf we have two groups g, gf and there is ase
sociated with each element a of gan clement a’ of g’ in a one
to-one way : a za’, such that
(ba)’ = Ba’, (2.3)
then the two groups are said to be simply isomorphic. Simply
isomorphic abstract groups offer no means of distinguishing one
from the other. The concept of isomorphism can, of course, be
applied to transformation groups. Two isomorphic transforma-
tion groups can be considered as faithful representations of
one and the same abstract group. A group may be isomorphic
with iteclf; it is then said to be automorphic. Such an auto:
morphism occurs when g and g’ coincide, ic. when a one-to-one
reciprocai association aga’ satisfying’ the condition (2.3) is
established between the elements ot the group g,
The question arises whether or not every abstract group
possesses a faithful realization. If this were not the case the
concept of an abstract grap as developed above would be tno
Ibroad—there would exist, in addition to the ascoriative law,116 GROUPS AND THEIR REPRESENTATIONS
other purely formal laws for the composition of transformations
‘hich are satisfied by every transformation group. Conversely,
2 proof vf the realizability of any abstract group would tell us
that all that can be said about the formal laws for the com-
position uf transformations is contained in our conditions (1)
fo (8). We can, in fact, construct a faithful realization of any
abstract group g by taking as the point field the group manifold
itself und letting correspond to cach clement @ of the group
the trausformation
of the group manifold on to itself. This “‘left-translation ”
4, is obviously a one-to-one reciprocal transformation which
hhas as inverse the transformation s= ants’. If a and b are
distinct elements the corresponding transformations t,, ty are
distinct, for they allow the unit element I to correspond to the
distinct elements a, respectively. If we perform in succession
two left-translations
sadma, Pa abt
the resulting transformation is, in consequence of the associative
law,
5+ 8" = Ofas) = (ba)s.
Consequently the left-translations constitute in fact a faithful
realization of the abstract group. However, the right-trans-
lations behave otherwise, for if we denote the mapping
$+ 8’ = sa of the group manifold on itself by #*(a), we find
instead of (2.1) the equation
(ba) = t*(a}t"(b).
§3, Sub-groups and Conjugate Classes
A sub-group gf of a given abstract group gis a sat of elements
contained in g which itself fulfils the characteristic group con-
ditions: the unit element 1 belongs to g’, with a belongs also
2-1 and with a, b also ba. ‘These three conditions can be reduced
to the one: if a, b are any two elements of g', then ba? also
belongs to 4’. We assume, of course, that the partial system
consists not merely of the element I, but the other limiting
case, in which g’ coincides with g, shall be included under the
concept of a sub-group.
Examples are readily found. In the group of Euclidean
motions are contained, for example, the group of rotations
(which leaves one point, the centre, fixed) and the group of
translations. The unitary transformations constitute a sub-
SUB-GROUPS AND CONJUGATE CLASSES 117
group of the complete group of all homogeneous linear transforma-
ions; the even permutations a sub-group of the eroup of all
permutations. If we are dealing with a transformation group
©, all those transformations of @ which leave a particular
point p fixed (j.c. which carry p over into itself) constitute a
sub-group Gp. Instead of a point p the fixed element may be
any figure composed of points; the transformations of the sub-
gronp must either leave the figure as a whole fixed (ie. they must
carry each point of the figure over into another such) or the
more restrictive condition that they leave cach point of the
figure fixed. We can also obtain sub-groups of © by employing
invariant functions instead of invariant figures. Jf y(p) 1s any
function of position on the point-feld with elements we
sociate with the transformation S:p— p' the function yf
defined by y/(p") — $(p) and say that it is obtained from ¢ by
the transformation S. If p’ = Sp, p” = Tp’, the equations
He) — ¥(e) — ¥"P")
show that the composition of the transitions ¥—>y' and
W > W" associated with S and T result in the transition > B”
associated with TS. Now consider all transformations S of G
which carry ${p) over into itself, ic. for which (Sp) — $(p) is
an identity in p; they constitute a sub-group © of G, and
¥(p) is an invariant of ©. In this way we can separate out
the rotations from the homogenenns linear transformations by
requiring the invariance of the unit quadratic form. ‘The sub
groups contained in a finite group g, which is described by
exhibiting each of its elements and giving explicitly the result
of composition of each two, can be obtained by inspection,
‘There is assnciated with each element a of the group g a
eytlie sub-group denoted by (8)
Tenet Peale
a ae, (31)
lve elemets a* uf which are defined inductively by the equations
al, ota ata
These elements constitute in fact a group, for m and m being
any integral exponents we have
are common to all sub-groups of g which contain a. The
elements of the set (3.1) can either be distinct or—and this
latter must be the case if @ is a finite group—they must r
themselves after a cycle of k terms: I, a, a?, =~, ab! are
distinct but at = I. his called the order of the element a.118 GROUPS AND THEIR REPRESENTATIONS
The order of @ finite group is the number of its clements;
accordingly, tre order of an elemeut o agrees with the urder ut
the cyclic sub-group (a) generated by 2. A group is said to be
commutative or Abelian if composition ofits elements obeys
reese
Ifa rune through the sub-group iy (left)
translations t, constitute a group of transformations which is
simply isomorphic with b, the point-field of which is the group
manifold. We say that two elements s, s’ which are equivalent
with respect to this transformation group are (left-Jequivalent
with respect to and express this situation by the notation
‘§ = 5 with respect to 9"; the condition for it is that s’ = as
where a is an element of B. In this way the elements of q are
divided into sate of elements which ace equivalent to 5. Tf
Ug is a finite group the number of elements in each of these
sets is given by the order of b, for ditterent translations f, send
s into different elements : as “+ bs if ab. The order of b is
‘accordingly a divisor of the order of g, and the quotient of these two
is the index of 6.
The considerations at the end of §2 above, which were
developed for groups of transformations, suggest a second
realization of the abstract group g. We associate with the
clement a the correspondence
sos = aset (3.2)
of the group manifold on itself. This correspondence, which
we call the " conjugation ” f,, 18 reciprocal one-to-one, and has
as inverse s=a-1¥a. ‘The law of composition is obeyed, for
from
Se ase, soe a bo
wwe obtain the product
"= basa? = (ba}stba)*
Two elements $, 5’ of g ate sald to be conjugate if they are
equivalent with’ respect to the group of all conjugations. Ac-
cordingly, the whole group is divided into classes, any element
of one of which is conjugate to any other element of the same
class. When we speak of classes within a group without a
more explicit description we mean these conjugate classes.
The realization of g by the group of conjugations is in general
a“ contracted " rather than a faithful realization. In particular,
the conjugation % coincides with the identity if a commutes
‘with all elements 5 of the group. The totality of all such ele-
ments a is called the central of the group; it is obviously
SUBGROUPS AND CONJUGATE CLASSES 10
an Abelian sub-group of g. But this disadvantage of con-
jugation over translation is offset by an advantage ; conjugation
is an isomorphic correspondence within the group itself which
leaves the unit element invariant and which associates with
each sub-group } of q another such, the conjugate sub-group
aba. These facts, which are expressed by the equation
a(st)a* = (asa™)(ata~*),
wete alucady contained implicitly in the considerations at the
end of § 1. fis said to be a self-conjugate or iavariant
sub-group if it coincides with all its conjugate sub-groups.
The importance of this last concept is best seen in the
following :
Theorem. If) is an invariant sub-group and = denotes equivar
ienve wits reopesi iu it, nen it fottows from
s'as,U'st that sif'=st (3.3)
To prove this we note that of — as, — bt (a, b in §) yield
st = asbt = (ac) (st). (4)
for ¢— bs" belongs to with b. Since ac lies in h our assertion
is proven. It is readily seen that the invariantive nature of fis
necessary ac well as sufficient for the validity of (3.8). Tn deal-
ing with an invariant sub-group h we need not distinguish
’ with ceepect to Heindeed,
the above proof was based on this fact
We may, if we like, consider equivalent elements as not
differing from one another (by application of the principle of
definition by abstraction) ; but by thus allowing equivalent
clements to fall together the group property of g is, in general
forfeited. In accordance with the above theorem it still remains,
however, if is an invariant sub-group. The group obtained
from g by identifying all clements which are equivalent with
respect to § is called the factor group g/); its order is the
index of the invariant sub group § of
These concepts are of assistance in examining the way in
which a group may be “contracted " on setting up a realization,
Let the transformation T(a) of a given point field on iteelf
‘correspond to the clement a of the abstract group g i the real
iow under cousideration. Then T(a) = Tia’) if and only if a
is obtained from a by comprsition with an clement ¢ (i.e. a’ — a)
for which Tle) is the identity. Such elements ¢ obviously con-
stitute a sub-group § of g, for it follows from
Tle) =I, Tee’)=L that Tle’) = Tle’) = 1.320 GROUPS AND THEIR REPRESENTATIONS
sup, for if T(e) is the identity.
the same is true of
Taea) = Tla)Tle)T-Ma) = Ta)T-*0).
In any realization of an abstract group g by « group uf iransfurree-
tinns the elements of a certain invariant sub-group }) of g correspond
Le the identical transformation; two different elements will be
ie ‘ad only af thoy ave
equivalent with respect to, The group of transformations is
consequently a faithful realisation of the factor group gi.
§ 4, Representation of Groups by Linear
‘Tran ions.
(On requiring that the transformations which are to serve
as a realization of a given abstract group g be linear and homo-
geneous we arrive at'a problem which is most fruitful from the
fhathematical standposnt and which is at the same time of
greatest importance for quantum mechanics ; we then speak of
a representation, instead of a realization, of the group.* An
editnensional representation of g, or a representation of degree 72,
Consists in asvocialing with each clement of the group att
affine transformation U(s) of the n-dimensional vector space
‘R=, in such a way that these transformations obey the:
law of composition
Ug = UIs). (41)
We thon say What » indies the transformation U(s) in the
representation space On choosing a definite co-ordinate
system in each transformation U( is represented by a square
matrix of » rows and columns, the determinant of which does
fot vanish. On replacing the original co-ordinate system by
another, obtained from it by the transformation A, the corre-
spondence which was formerly represented by the matrix U(s)
is now represented by the matrix AU(s)A-}. Consequently if.
the association s— (0) i a representation, the aseociation,
SAU A*
is obviously also one; this latter representation is said to be
equivalent to the former. ‘They are essentially the same,
differing only in the choice of the co-ordinate system in terms
of which they are described.
Examples-—A representation in one dimension consists in
signing to cach element s of the group a non-vanishing number
xis) in such a way that
ale = x0) x10. (42)
REPRESENTATION OF GROUPS 121
In particular, x(1) = 1. A most trivial L-dimensional repre-
sentation is obtained by assigning to each s the number 1:
xi) = 1. This special case is ealled the identical representation
Consider next the co-called symmetric group, the group
ny — ny of all f! permutations of f things. The assuciation
soh= sl)
according as s is an even or an odd permutation, defines a
L-dimensional representation, the “ alternating” representation
of the group 7. For the ‘character 5, which distinguishes
hetween the even and the odd permutations, satisfies the
equation
Bu = Bee
Let g be a finite cyclical group of order hj the elements
sare then
Lats aht
and a = 1. Consider tne L-dimensional representation $—> x(s)
in which via) =. The condition (4,2) for a representation
then tells ne that ta the elements « nf this arias enrrespond
4, 6,640 04,
Conversely, when e is an arbitrary
fe* root ot unity ‘this assocration dehnes a i-dimensional ré-
presentation of g. We have thus obtained a complete survey
of all paseible Iedimensional representations af a eyelieal group.
The only example of a multi dimensional representation
which we offer at this tine is the following tivial one, If
is itself a group of linear transformations of an n-dimensional
vector space'i then the association $—> s defines an n-dimensional
representation of ¢. This example implies more than one might
at first sight imagine. We have in fact to do the following:
wwe firet cbtain the otructure of the group g by abstraction from
the gioup of lineat Wausformations aod then return to the
original ealization by means of Ue correspondence ¥—> s
between an element s of the abstract group on the one hand
and the linear transformation s on the other,
The concept of equivalence has a more general significance
than that discussed above, It may refer to an arbitrary system
E of linear conteapondences U of the dimensional vector
space We need not assume Uat Uses cuscespondeuces
Possess an inverse (ie. that they have a non-vanishing deter-
inant), nor need we assume that they are associated with123 GROUPS AND THEIR REPRESENTATIONS
the elements s of a group, as is the case with representations.
On expressing the set of correspondences U in terms of a new
co-ordinate system each matrix U goce over into the matrix
U! = AUA; the system Fis transformed into the equivalent
system 2 consisting of the U', a is here a fixed non-singular
matrix.
Consider @ correspondence U of & on to itself. A linear
subspace W of R is said lo be invariant under U if the vectors
of BY are transformed into vectors of BY by U. Ie BY is invariant
then the space B (mod. H) obtained by projecting w with
spect to R’ is also invariant (cf. 1, § 2, in particular Fig. 1).
being invariant, U gives rise to a correspondence U' ot
on to itself; we say that U induces U' in ¥. Similarly for
the space obtained by projection, We now pass from a single
correspondence U to a system Z of correspondences. is
said to be invariant under Z if it is invariant under each corre:
spondence U of Z. Describing ® in terms of a co-ordinate
system which is adapted to the invariant sub-space , all
matrices U of the system 2 reduce simultaneously to the form
illustrated in Fig. 1, p. 8. 2 is called irreducible if 8 con:
tains no sub-space, other than & itself and the space 0 consisting
only of the vector 0, which is invariant under Z. We shall
have occasion to reduce f in such a way that each constituent
separated off is irreducible under a given system J. This
requires the construction of a series of sub-spaces
8, Diy By + +, = (83)
beginning with 0 and ending with ®, in which each member
is contained in the preceding one and is such that 8; (mod. 8.)
is irreducible, Naturally 9, shall actually be larger than Riau,
not merely coincide with it. The implications of thie reduction
fare most readily seen in terms of the matrices U of the corre-
spondences of the system 2 on adapting the co-ordinate system
to the “composition series” (4.8), i.c. by choosing first a co:
ordinate system in GR, then supplementing it with additional
fundamental vectors in order to obtain a co-ordinate system
for My, Ry, > «in turn,
|__ 2s said to be completely reducible if R can be decomposed
into two sub-spaces X-+ ®’, each of which are invariant under Z
and such that ucither of them consists merely of the vector 0.
This woucepi of compiese reducibitity is more exacting than that
of mere reducibitity, On describing 9 in terms of a co-ordinate
system which is adapted to this decomposition, each matrix
U of E assumes the form illustrated in Fig. 2, p. 9. We are
then faced with the problem of decomposing’ % (or Z) into
FORMAL PROCESSES 123
constituents, none of which is completely reducible, ie. of
decomposing = ¥y ++ "+e into invariant. sub-spaces,
none of which is completely reducible.
We often find that reducibility implies complete reduciility,
i.e, that in many cases we have the theorem: If § is an in-
variant sub-space of R, a second invariant sub-space 9” can
tbe found such that s is completely reducible (with respect to
B) into RL, We shall soon see that this is actually the
case when § is a unitary space and 3’ is a system of unitary
transformations.
Tt was shown in Chap. I, § 3, that if the oystem J is re
ducible, then the system J* of “transposed ” correspondences
of the dual space un itself is also reducible. If $37 Uis)
is an n-dimensional representation of Ue group g te transposed
U*(s) do not constitute a representation ; it is readily seen,
however, that on employing instead the contragredient corre-
spondences
“
we do oblain a representation s—- Bia) of the dual vector space.
This we call the contragredient representation §,
§5. Formal Processes. Clebsch-Gordan Series
Continnous groups offer what are perhaps the simplest
examples of the theory of representations. We consider in
particular the group ¢— ¢, of all linear and hamogeneaus trans-
formations ¢ ia n variables xy, %y * * %; # with non-vanishing
determinants; we consider cach sct of values x, as a vector
in.an n-dimensional vector space t= ty. The classical theory
of invariants, first develuped in England about the middle of
the last century, concerned itself in particular with the repre-
sentations of ¢ induced on the coefficients of arbitrary forms
an the variables x;. A quadratic form in these variables is a
linear combination of the n(n + 1)/2 linearly independent
products x;,; under the influence of a linear transformation
of the 2, these products undergo a linear transformation [5],
and the correspondence $—> {S}» is obviously a representation
{63 in (y+ 1)/2 dimensions of the group ¢, The transformation
of the wariables +, sends the arbitrary quadratic form
Za tty
inta a quadratic farm
Zain He424 GROUPS AND THEIR REPRESENTATIONS
in the new variables, where the coefficients aj, are obtained
from the ay, by a certain linear transformation s, associated with
S; sy is obviously contragredient to [sly The quadratic form
Characterized by a fixed act of (+ 1)/2 coefficients aj, may
therefore be considered as a vector in a space of this number of
dimensions, and the transformation s of the variables #, induces
the transformation sy in this space. The space thus defined by
ithe totality of mary quadratic forms is thus tie puini~feid for
a group of linear homogeneous transformations which constitute
a representation of the group ¢.
‘We may in the same way deal with cubic, quartic, +
fic forms. The totality of monomials of order f are contained
in the formula
afiatts +s als (6.)
‘here the j; are non-negative integers whose sum
fthte that
They constitute the substratum of a representation {¢}/ in
G]-t
mee
dimensions.
But we can exhibit representations of ¢ which are formally
yet simpler than these arising from the theury of forms, Let
(@) and (4) be two arbitrary vectors in our n-dimensional
space t and consider the products x;y, On subjecting the x
and te y, to the same transformation s of c (transition to a
new co-ordinate system) the n* products undergo a certain
linear transformation 5X s associated with s and the corre-
spondence s—> s X sis an n-dimensional representation (¢)¥ of ¢
Now a system of numbers F(i, 4), depending on two indices i, &
which run through the values 1, 9, - - -, m, is said to be a tensor
of second order if under the influence of a transformation s of
t the F(i, &) undergo the same transformation as the products
4.x 0f the components of two arbitrary vectors z, of t. Hence
the tensors of order 2 arr the substratum of the represent:
(0 of & (6)8 contains the reprecontation {e]8 whi 4
in the sub-space of symmetric tensors of order 2; tho tensor
with components F(i, #) being symmetric if F(iR) = F(ki).
In geometry the azti-symmetric tensors, i.e. tensors whose
components satisfy the condition Pik) = — F(ki), play a more
important réle than the symmetric ones.” In particular, two
arbitrary. vectors (x), (y) define a surface element’ with
‘component
{ik} = XO. — Ys
FORMAL PROCESSES 10
of these quantities but n(n —1)/2 are fincarly independent,
say those tor which ¢ <2. On subjecting the components 2,
of the vector £ and the components y, of the vector to the
same linear transformation 5, the components of the surface
clement defined by them undergo an n(n — 1)/2-dimensional
Tinear transformation {5}. s—> {sj is a representation {Q?
whose substratum is the totality of antisymmetric tensors of
order 2, Hence the representation (c)2 is reduced into the
reptescntations {* and {6}%, for any tensor F(ik) can obviously
bbe waitten
ik)
y
BURGH) + PUb)) + BFR) — FC)
ue. in a unique manner as the sum of its symmetric and anti-
symmetric parts. That this reduction is correct is further borne
ut by the fact that the dimensionalities satisfy
met), mln — 1)
wo MED mn 1
‘Similarly three arbitrary vectors g, y, g determine a S-dimen-
sional element of volume with components
eat
These clements constitute the substratum of a representation
{tin
in} _ nin — In — 2}
Sa a a ae
dimensions. Coatinning in this way we can construct 4-,
y+ ++, nedimensional elements ; this process must cease with
m-rowed determinants, for a determinant of the form (6.2) with
more than » rows must necessarily vanish identically.
We shall sce that the representations of ¢ whose substrata,
are the symmetric und anti-symmetric tensors of order f are
irreducible, and shall in fact solve the general problem of eifect-
ing the complete reductions of (¢)/, the representation induced
by ¢ in the space of all tensors of order f, into its irreducible
constituents (Chap. V).
The tencar concept really depends on the X-multiplication
introduced in TT, § 10. Tf the m variables x, undergo a trans-
formation A and the » variables y, a transformation B, then
the mn products ag’, undergo a transformation A x B.’ Con-
sidering the xq as the components of an arbitrary vector z in
an m-dimensioual space Ry, and the yy a9 the components of
ied} I (62)126 GROUPS AND THEIR REPRESENTATIONS
‘yin B, the products x, yx may be considered as the components
of a vector F X 1) in an mn-dimensional vector space Bm Ry.
Hence two representations
Sr 08), Gre 0) (5.3)
of g in m, n-dimensions, respectively, give rise to a new mn-
dimensional representation which we denote by x =
& X G1: 5-+ Us) x U's), (5.4)
‘This presente a goneral method of obtaining a new representa.
tion § x §' from two given representations 9, $.
Dencting the reprezentation s->s of the lincar group ¢ for
the moment by (6), the representations of ¢ whose substrata
fare the tenors of order 2, 3, - - + aro then (6) x (¢) — (9%
x0 x () = 0% :
We should, perhaps, have discussed the addition | of two
representations before discussing their multiplication X. Con.
sider the vatiables x and y, as the components of a single vector
4 in an (m-+n)-dimensional vector space; when the #% are
subjected to the transformation A and the yy to the trans-
formation B these m+n variables undergo a certain trans-
formation (4, B). Heuce we obtain from (6.8) the representation
Gx H':5> (U9, U'S)]
in m-+ m dimensions. The inverse of this process is complete
reduction, as discussed above: + §! is completely reducible
into the components & and §.
Another important forwal metiod is the following: Any
representation T in N-dimensions of the linear group tq in
n-dimensions may be used to construct an N-dimensional
representation of any abstract group g from an n-dimensional
representation § of the same. I associates with the linear
transformation u in t-dimensional space @ linear transformation
U in W dimensions, so if :5— u is an n-dimensional repre-
sentation of the group g with elements s, then
canst
is an N-dimensional representation s+ U of g which we may
denote by I). To this is due the importance of the repre-
sentations of the linear group for the general theory of repre-
sentations. For example, take I’ to be the representation of
¢ whose substratum is the dual space, the space of all tensors
of order 2, of the symmetric or anti-symmetric tensors of order 2,
ete. ; we then obtain from the representation § of the abstract
group g the representation & § x G16 X $1, (@ X Gh ete.
KURMAL PROURSSES: 127
‘The three most important formal processes are (1) addition,
(2) X-multiplication, and (3) the I” process. The first two
generate @ New representation from one or two given repre-
pentations, the third a new one from a given representation.
The first two are completely circumscribed, but the third
contains generat weihod, fur Pinay be ay vepreseutation of
the linear group Cy
If gis a ‘sub-group of g, then any representation
G1 s- [T(8) of g contains a representation of g’; we need only
Ict the element ¢ run through the sub group g’!' Thic too may
te considered as a formal process (#) which gonecates a repre~
sentation of g’ from a given representation of g
‘The X-multipication occurs in yet another connection.
Given two groups g, g’, we can consider the pairs (s, s'), the
first member s of which is an element of g and the second
waa clement of g', as the elements of a new group 3 x g’, the
direct product vf g aud g’, obeying the multiplication law
) = (ot, o'¥).
wel
‘The order of g x g’ is the product of the orders of g and g’. If
H:s—U() is an rdiniensional representation of g and
9:5’ > U'(e) an n-dimensional representation of g’, then
(5!) + Ul) x Ue) (65)
ie obviously a representation in nn’ dimensions of the group
3X9); we denote it by § x G (with a boldface x). This
construction may be broken up into two steps. First introduce
the representation
(5, <)> U8)
of g XQ’; there is no reason why we should not designate it
by the same letter § as the representation s—> Us} of g—we are
accustomed Wo calling Use function f(x), considered as a function
of the two variables x, y, Dy the same letter ws the function
fx) of the single variable x. U(s) and Us) are thus to be
considered as finnetions of the same variable pair (s, s), and then
the representation § % § of g x’ may he obtained by ordinary
X-multiplivation froin © and §'. The differentiation between
boldtace x and ordinary x is accordingly puscly pedantic,
Examples. Unimodular Group in Two Dimensions
Let g a consist of all linear transformations + of two
variables 1, 9 :
wv maxtby, roel dy, (66)128 GROUPS AND THEIR REPRESENTATIONS
whose determinant ad — be = 1 (“unimodular linear trans-
formations *). A homogeneous polynomial in x, y of order f is
a linear combination of the f + 1 monomials
MINNA Ae (6.7)
Under the influence of $ they undergo a lincar transformation
which we denoted above by |sj/; they constitute the substratum
of a representation [elf :s—> (s]y in f + 1 dimensions which we
now denote by Gy. 6, is, although we have yet to prove it,
irreducible,
‘We can restrict ourselves within ¢ to the sub-group ¢ of
principal” transformations which transform each of ‘the
variables separately :
¥ mar, y=1y, (5.8)
where a +0 is an arbitrary constant, G is Abelian, This
transformation multiplies tte monomials of the set (5.7) by
POPU actor
On associating the number at with the element (5.8) of ¢ we
obtain a 1dimensional representation which we denote for the
moment by Gl"; here r can be any fixed integral exponent.
We have just seen that the irreducible reprecentation Gy of &
is completely reduced on zestreting oursclres to the sub-group
& into f-+ 1 onc dimensional representations €) with r= f,
f-3,—h Th
‘AS an caanmipte of uu
the problem of reducing the product 6, x G, of the two repre-
sentations Gy, , of ¢ into its irreducible ‘components. The
result is contained in the formula
Gx 6,= 26, (6.9)
where v runs through the series
va ftefts \=el (6.10)
without repetition, decreasing by 2 from term to term. This
equation is essentially identical with the Clebsch-Gordan series
which plays such an important réle in the theory of invariants
of binary forms, We shall zee in the succeeding chapters that
it may justly be considered as the fundamental mathematical
lenate the group of all non-singular linear transformations
Wy de
‘wil however ectasionally be used fo denote the more
lar group, in which vase dhe seatsction sil be explicitly
FORMAL PROCESSES 129
formula for the classification of atomic spectra and for the theory
of the valence bond,
The proof consists in showing that
GX C= Cay + (pa X Gps) (o.21)
for (6.9) then follows by mathematical induction and the fact
Gx G=ep
A new co-ordinate system for the representation space of Ey
is obtained by replacing the basis (5.7) of homogeneous poly
nomials of order / by another basis. In this sense we can say
that the polynomials of order f constitute the substratum of
the representation G,. The substratum of the representation
G, ».G, is then the totality of polynomials
= Oxy: En)
depending on the components of two arbitrary vectors (x3),
(7), homogeneous and ot order f in the nrst, and homogeneous
and of order g in the second ; we write the total order f+ z= h.
The ® are thus linear combinations of the (f+ 1) +1)
‘monomials
Ky Gop where i+ k= fet
Doth vectors ate Lrausformed cogiediently under the same t
formation s, (5.6). The problem consists in completely reducing
the space of the polynomials into two sub-spaces (P), and
(®) which are the substrata of the representations @, and
GX 6, , respectively We first disuse the structure of
these twa sub-spaces
(@). Expand
(2 + PM ok + Bole ot do (Peed bo EB Be
(5.13)
in powers of the undetermined coefficients a, . The
4, = dey: fq) are special polynomials of the type ® and span
the subsspace (O)y. We must now show that this subsspace is
invariant under ‘the transformation (5.6) of the variables ;
ie. that $f = di(v'y j27!) is a linear combination of the
$i = deny; G7). Ie is clear that if this is the case then ¢ in-
duces the fepresentation Gy in (®)g, for on identifying the two
vectors
& (6.12)
PINT (6.14)
diley a9) argh
; becomes130 GROUPS AND THEIR REPRESENTATIONS
Hence we are certain @ priori that the # + 1 functions 4, are
linearly independent. i .
In order to arrive at the desired proof we replace x, y in
(8.18) by
dmarthy, yoatdy,
and in the same way & 9 by
Fae thy amt dy
Now note that az’ + fy’ is the linear form
(oa Pele + (ub + Rd)y = Ax + By
in z and y; hence
(ae? + Ay’ Wlat’ + fr!)
and by (8.18)
DY sigh gt WY pint Bi «
EG)MP Ha 2 (arate
Su replacing 4, 3 vu the sighi-haud side of this equation by
A=aa+ fe, B= ab + a,
and equating coefficients of aiip', we obtain $j as a linear
combination of the dy.
(@)'. The substratum of the representation Gj, x 6-1
ists of the polynomials
P= Ploy; &)
of order f—1in (x, 9) and of order g —1in (£9). They are not
polynomials of type ®; in order to increase the order in the
components of each vector by 1 we replace each such Y by
Om (oy — 9)
The factor thus introduced in no way affects the representation.
‘The last step in the proof consiete in showing that the total
space of polynomials @ is completely reducible into these two
subspaces; “ie. in showing that any polynomial @ can be
‘ritten in the form > pe
= (dub + ards ++ «+ aga) + (2m — v8) (6.15)
with unique constant coefficients a; (The develo
: levelopment in
terms of powers of the determinant 4 — y€ obtained from this
by induction is the Clebsch-Gordan series.) First, the dimen-
sionalities are correct, for
(f+ Ve+N=F+24+1) +4.
rufficce to show thal the various terms in (9.10) are
fependent, i.e. that an expression of the form (6.16),
(Ae-+ PoM(AE + Bye,
co
Hence
linearly i
THE JORDAN-HOLDER THEOREM 131
jn which Y iy a polynomial of order f— Lin (2, 9) and of order
g—1in 7), can vanish only if ¥ vanishes identically and if
all the coefficients a, are zero. ‘The proot is extremely simple.
We first let (fn) = (zy) as in (6.19), then the equation ® = 0
becomes
gh baat y +--+ anyh
identically in x and y; hence a,—0. Having established this
we return lo the two sets of variables ay; fy aud oblain the
equation
(ey — 9e)¥
trom which it follows that Y= (—in an algebraic identity
for polynomials we may always remove a factor, such as
= 98, which does not vanish identically
Our formula (5.9) also holds for the group ¢ of all linear
transformations of x, y with non-vanishing determinant. We
must then interpret ©,, v= h — 21 in (5.9) as that representation
whose substratum is ‘the totality of homogeneous polynomials
of order v in % and y multiplied by (am — y€). In other words,
the new 6, differs from the old in that the transformation of
the (v+ 1)-dimensional representation space corresponding to
5 in the representation €, is to be multiplied by the ?* power
of the determinant ad — bc.
&, XG, is a representation of ¢y Xt, the group consisting
of pairs (s,'s') whose members s and #’ run independently through
the entics groupe, On intendscing the restriction that cis
‘the element F obtained from s by replacing the coefficients of
the linear uausforimation s by Useir conjugate complex, 6, x Cy
becomes a representation ©,» of cy, the substratum of which
may be taken as the monomuals
syhaye (itk-f obe-e)
of order fin (x, 9) and order g in (Z, g). Tt can be shown that
G,,, is also irreducible.
§ 6, The Jordan-Hélder Theorem and its Analogues
Perlaps the most fundamental theuren of mathematics is
that on which the concept of cardinal numbers depends. Let
the members of a finite set of objects distinguished by marks
4,5, c+ + + be exhibited individually in this order and associated
‘with the symbols 1, 2, +++ m. The theorem then states that
the “number” x is independent of the order in which the
objects ure exhibited. The proof of this theorem is of con:
siderable mathematical interest and offers the simplest example182. GROUPS AND THEIR REPRESENTATIONS
of the type of proof employed in establishing the Jordan-
Holder theorem. A new enumeration consists in associating
the symbol 1 with any one of the objects, the symbol 2 with
any one of the remaining objects, etc., until the entire set is
exliausted, the last object receiving the symbol n’, We assert
that n’ =n
‘The proof is divided into two steps. (1) I in the new enumer-
ation the symbol 1 is associated with the same object a as in
the old, our theorem for the series from 1 to m is reduced to that
for the series from 1 to n— 1. This is immediately evident on
discarding the object @ and reducing by one the symbols as-
sociated with the objects 8, c,- - - in the new as well as in the
‘old enumeration. (2) If, on the other hand, the symbol 1 is
associated with one of the other objects 8, c, - + then in the
new enumeration the object a is associated with some symbol
4 contained in the series 2, 3,---, n', We now introduce a
third enumeration which enables us to make the transition
between the first and the second by interchanging the symbols
1 and i in the second enumeration. The number n’ is obviously
unaltered by this process. But we have now introduced an
equivalent enumeration in which the object a is associated with
the same symbol 1 as in the original and have reduced the
general case to the one considered in (1) above. The proof of
the theorem then follows immediately by the method of
mathematical induction,
As an auxiliary result of these fundamental considerations
wwe have the theorem that any permutation can be obtained by
the successive application of transpositions.
The Fordan-Hélder theorem is concerned with an abstract
group ¢. An invariant sub-group 9! of g which does not coincide
with @ itself is said to be maximal if there exists no invariant
sub-group of g—except. g’ and g—containing g’. The factor
group gla’ is then simple, i.e. it contains no invariant sub-group
with the exception of itself and that consisting only of the
unit element I. As was recognized hy Galvis, the so-called
composition series
$= 8 Bi Be *s Brau Be E (6a)
is of fundamental importance for the solution of algebraic
equations, This series begins with g and ends with I, and each
member is a maximal invariant sub-group of the’ preceding
member. We assume that the composition series terminates ;
this is naturally the case for finite groups, as the order necessarily
decreases trom term to term, The successive factor groups
UG, Gales * + *s BraalBr — Bran (6.2)
THE JORDAN-HOLDER THEOREM 153
are simple. The Jordan-Tolder theorem asserts thet he
structure of these factor groups, except for the order in which they
‘appear, ts untguely determined by g.
Consider, theretore, a second composition series
% = 8 Gr Si
of the same group g; it is to be compared with the “ standard
series” (8.1). The proof of the fact that this new series also
contains exactly r+ 1 terms and that the corresponding factor
groups are, except for the order in which they occur, isomorphic
with the factor groups (8.2) is again accomplished in two steps
(1) If the two second members 9’, 9, coincide, the theorem
for the group g, whose standard series contains r-} 1 members,
duced to the corresponding theurem for the group g,, whose
dard series evtains but r members,
(2) If g, and g's do not coincide we construct the inter-
section ) of g; and gy, 1€. the set consisting ot all elements
common to the two. } is then an invariant sub-group af ¢’,
and, as we shall prove, a’x/) is isomorphic with a/g,. That
two elements s, ¢ of g are equivalent with respect to gy, ie. that
they belong ta the same set,” is expressed by the equation
fas where a, ising, I sand tare at the same time elements
of the sub-group g's, then a, is aleo in g', and consequently it
is an clement of §. We may therefore consider as the clements
of gly those sets in g which contain an clanent of g',. The
elements contained in these classes then constitute an invariant
sub-group 9 of g containing both g, and gy, and g°/9 is simply
isomorphic with 9/g. But since g’, is maximal either
or 5 ‘The second case implies that a, is contained in a’y,
and since it is maximal it must coincide with g',, contrary to
assumption. Hence § coincides with g and our assertion is
proved The intoreection & af g, and g', dapende cymmetrically
on both, whence g/g’, and g4/6 are also simply isomorphic.
We now proceed as follows. We construct a composition
secea for hy which we denote simply by Baad Compare
the flowing four cousposition sees of
Ane
8 Bb os
Aa
a a
‘The comparison of the first and sezond series is reduced to case (1).
The second and third series agree from the member } on, and
the two foregoing faster groups
6/e, oui18¢ GROUPS AND THEIX REPRESENTATIONS
are, as we have seen, simply isomorphic with
Biss aa}
‘on interchanging their order, The comparison betweea the
third and fourth scrics is again reduced to the case (1). The
proof of the theorem for composition series containing r+ 1
‘members is thus reduced to the proof of the corresponding
theorem for series with but r members, and since it obviously
holds for r-—2 (ke. for simple groups) the method of mathe-
siatical induction establishes its general validity.
The close methodological agreement between the construction
involved in the proof of this theorem and that involved in the
proof of the independence of the cardinal number of a sct of
the order in which the objects are enuncrated is immediately
evident,
E. Noether has given a generalization af the Jordan-Holder
theorem which is of importance for us. A correspondence
3-3! = As of the group on itself is said Wo be automorphic it
‘multiplication is invariant under it, Le. if (61)" = s—we here
neither assume that different elements s generate different
elements s' nor that for a given element s! there existe an element
5 such that s-> s! in virtue of the automorphism, Let 2 be
& system of such automorphic correspondences of g. We now
admit only sub-groups of @ which afe invariant under 5, ie.
sub-groups whose elements are carried aver by all operations
of the system J into elements of the same cub group, We say
that two such “allowed” sub-groups g) and gy have the same
structure if we can set up a one-to-one simple isomorphic
correspondence between the elements of the one and the cle.
ments of the other in such a way that every operation il of
ystems D Slenuenis uf the two sub-
reaps ove ta corapondng elements, ”‘The Jordan older
heorer sti ‘under this modification ; its ean
aken over unaltered. saad
The vectors of an a-dimensional vestor space $ constitute
grap ‘whose sauliipiicaiion is che adaition + of
vectors, We must for the moment supplement addition by
the operation of multiplication of a veetor by an arbitrary
number; hence the concepts and theorems applying to vectcr
Space are not truly specializations of the concepts and theorems
of Abelian groups, but Ulere exists a thorough-going analogy
Yetweeu the two. ’ Indicating this analogy hetween a group (ex
the let) and vector space (an the right) by — we have he om
ample, ‘sub-eroup m~ inear sub-epace, automorphism ~ linear
pondence, “Indeed, a linear sub-space is a system BR of
THE JORDAN HOLDER THEOREM 135
vectors such that with g and ip their sum x49 and the product
Ag by an arbitrary number 2 also belong to W’, and a corre-
spondence g > 2” — Ag is lincar if it sends z +. 9 and Ag over into
Pepi aud Ae, acspoctively. Every “subgroup is inere
invariant, as We are dealing with Abelian groups. If ® is
a sub-space of 8 the space Uf (mod. Ht) odtained by projecting
A with respect to fi’ is the exact analogue of a factor group.
‘A composition series consists of a sequence of spaces each
member of which is a linear subspace of the preceding one
‘and has one less dimension. The last member is the space 0,
consisting of the vector 0 alone, and the number of members in
the series is 1 greater than the dimensionality n. The Jordan:
Holder theorem is here vaiid but trivial,
Ua the other hand, this theorem is of considerable importance
on going over to Noether's generalization. Consider 2 system
Z of linear correspondences of the vector space on itself ; the
terms invariant, equivalent, reduction shall in the follwing refer
to this system, | Two invariant sub-spaces WR, and fR, are similar
or equivalent if a one-to-one lincar correspondence fy ZZ, can
be set up between the vectors of the one and the vectors of the
other in such a way that any operation 4 uf the syste sends
corresponding vectors over into corresponding vectors. On
reading the series (4.3) established in §4 backwards, we have
the exact analogue of the composition series: each member of
the cories ie followed hy a maximal sub-spare which is invariant
under E. (The possibility of constructing the composition
series in increasing as well as decreasing order is due to the
fact that the addition of vectors is commutative.) Furthermore,
wwe can obtain the concepts and theorems relating lv a system
2 of correspondences as genuine special cases of those of group
theory, and not merely as analogues, by supplementing the
system Z with all similarity transformations, i.e, by all corre-
spondences of the form ¥—> 2" = Mg representing multiplication
by aa arbitrary number The Jordan-Holder-Noether theorem
0, Ry, Ry +R (6.8)
the corresponding projection spaces
MW, (mod. HK), Hy (mod. HH, + =»
are equivalent to the projection spaces (4.3)
Hy, Ry (mod. W,), Me (mod. RW), + + =186 GROUPS AND THEIR REPRESENTATIONS
of the original series, taken in a suitable order. The number
of members is, of course, the same in both. The reader is
advised to reconstruct the proof of this theorem by carrying
through the proof of the Jordan-Holder theorem step by step
for this case
In particular, if the eystom Z consists of the transformations
UG} ssuciaied with the various clamenis s of a group in a
representation 9:5 —> U(g), our result yields the
Uniqueness theorem : The irreducible representations separated
off from & by successive reduction are completely determined by ©,
except for the order in which they occur, considering equivalent
representations as the same. In particular, the complete reduction
of into srreducible components is unique, always considering
equivalent representations as the same.
§7. Unitary Representations
For the case in which the representation space 9 is unitary
coreespandencee Tile) af an itself,
the clement s of the group under consideration, are also unitary,
certain of the concepts introduced above are to be modified
accordingly. Iwo representations
SUG), s+ Us) = AU)A,
are to be considered as equivalent only if is unitary, i.e. it it
is a transformation from one normal co-ordinate system in
® to another such. If 8 is a sub-space of a unitary-orthog-
onal co-ordinate systom can be set up in fR and supplemented
by additional fundamental vectors to form a complete unitary
orthogonal co-ordinate system for the entire space ®: every
sub-space of a unitary space is fer se unitary. Tnvariance and
reduction remain as before, but we allow only those decom:
positions of 9 into two sub-spaces 8 +f, in which OR,
are perpendicular. For a system of unitary correspondences
reducibility implies complete reducibility and we have the theorem :
IR’ is invariant with respect to E then ® may be broken up into,
RW" in such a way that H" és also invariant under ES. We
need merely to define "as the space defined by all vectors per-
pendicular to 8’. The theorem naturally holds for the case in
which Z is a system of infinitesimal unitary correspondences or,
what amounts to the same, a system of Hermitian forms. The
theorem developed in the ‘preceding section proves that these
irreducible components are uniquely determined, in the sense
of (unitary) equivalence, to within a permutation,
UNITARY REPRESENTATIONS. 137
Examples
(1) The Unitary Group in Two Dimensions
‘The group ¢ ~ cy of linear transformations in Lwo dimeusivus
coutains the sub-group uu, of unilary transformations.
Hence the representation Gy of c obtained in § 9 is also a repre
sentation of #, 1mis representation 13 not unitary as it stands,
but it can readily be made unitary by a slight change. The
transformation of Cy corresponding to the unitary transiorma-
tion s of the co-ordinates x, y is that induced by son the monomials
waxy (ETR=S/) (1)
of order f. For purposes of symmetry we label these ro-ardinates
with the index w—i—B which mis through the values
Ff —9,+++,—f. This is also desirable because on restricting
‘urselves to the cub group of principal trancformations ”
xsex, pana
EY
x, is multiplied by the factor en. We now employ, instead of
(7.1), the variables
ne (ra)
Vilk!
obtained from them by multiplication with a constant, Ihe
representation G, of tt will then be unitary, as follows from the
equation
1 igh sigh
Fiset yp = LATTE = Late
‘We call 6 even or odd according as fis even or odd. The even
representations associate the identity 1 with the reflection
Homa Yann
and the odd associate with it the transformation —1. G, is
also irreducible when considered as a representation of u, and
fan letting f accnme the values 0, 1, 9, « + + they form a complete
system of inequivalent irreducible representations of t. The proof
of these assertions, which we employ heuristically in the follow.
ing, will be given in Chapter V. On writing a homogeneous
polynomial of order f in the variables 2, y in the form
Zita
he coefficients ag Wausform under the influence of @ unitary
transformation 5 like the components of « vector iu the sepre-
sentation space of Uy.188 GROUPS AND THEIR REPRESENTATIONS
‘The complete reduction
Wy X Oe) = Cpe + Cpr X Coad
‘was accomplished by breaking up the space of thet polynomials
” into two invariant sub-spaces (0), and ()'. We must
now verify that these two sub-spaces are mutually orthogonal
in the unitary sense. A general polynomial ® may he written
‘where the x, are given by (7.2) and the , are the corresponding
monomials
‘Two such polynomials ® with coefficients ayy, Dy, are orthogonal
if
Thar be = 0.
The polynomial Ey, whose highest coefficients ay = 1 while
all others vanish, is to within a constant factor #-£¢ and is,
obviously perpendicular to all polynomials (@)', for in all these
jatter the coefficient of + 4) the sum
Z lFlaia s+ + ial (1.6)
Gap
is a unitary invariant, On restricting ourselves to the
{7}-dimensional linear manifold of antisymmetric tensors we
fake as the variables in tensor space those components
Fohig = ++ i) for whien iy > + Sa
‘The snm (7.8) hecomes in this case
fl MO Hut
Fail fs Sail
“KTR
extended over all integral f, 2 0 for which f; + f+
The coefficient indicates how often the term |Fligis - - = i|*
occurs in the sum in consequence of the fact that its value is
unchanged on permuting the indices. We must therefore
consider the quaatities
Pfy fy 2 fe)
VATAT fal
as independent components of an arbitrary symmetric tensor
of order f in order to obtain a unitary representation (uf140 GROUPS AND THEIR REPRESENTATIONS
‘The truth of this assertion follows from the fact that the special
tensor (7.7) satisfies the equation
cafes af «cafe
tn
(7.8)
hat @ normal co-ordinate
im as that a. commutativ n of
unitary correspondences is completely reduced to a cet of
T-dimensional systems, The only irreducible unitary ropre-
sentations of an Abelian group are accordingly \-dimensional.
For it follows from
UE)UY = Vos) (42)
and the Abelian character of the group that the unitary matrices
U(e) associated with the elements s are commutative
Ef @ and §! are unitary representations, then § + 9,
© XH! are also.—The first fundamental problem for a given
group g is to find a complete system of inequivalent irreducible
unitary representations of g, for then any unitary representa-
tion of g can be obtained by the addition of these irreducible
representations, The second fundamental problem ic to reduce
the product § X §! of two irreducible representations %, O° of §
into is irreducible components ; or better (after having solved the
first problem), 19 determine how often each of the irreducible
representations occurs in this product.
We illustrate these problems on the cxample offered by
sotation groups, which are of particular importance in quantum
physics.
§ 8. Rotation and Lorentz Groups
(a) The Group of Rotations in the Plane
We describe the 2-dimensional plane by a complex co-
ordinate x. The rotations of the plane are then given by
zoe mon, (8.1)
where = e* is a constant with unit modulus. (The rotations
of the real Z-dimensional plane thus coincide with the unitary
transformations of a single complex variable.) ‘Ihe angle of
rotation ¢ determines the rotation completely, but it is of course
only determined mod. 2 by the rotation. The angle of rotation
behaves additively on composition: the rotation 4 followed by
the rotation ¢’ results in the rotation ¢ + ¢’. This rotation
ROTATION AND LORENTZ GROUPS 4a
group is accordingly one parameter continuous Abstian group.
We obtain a I-dimensioual representation D') of our rotation
group b = by by ussuciating with the element ¢, (8.1), the linear
correspondence
noe Ser em, (8.2)
where m is any fixed integer. Tassert that the D), m running
through all integral values, constitute a complete system of
irreducible unitary representations of Dy, This can be seen as
follows.
Any irreducible representation is necessarily I-dimensional :
it associates with the rotation ¢ a number x(¢) of absolute value
Tench that
xlé |!) = xld) x8)
We aseume that our representation is continuous; then x(¢)
is a continuous function of ¢ with period 2x. First, x(0) — 1
‘We write x(f) — e# and determine Aig) uniquely by the require.
ments that A(0) = 0 and that Ag) shall be a continuous function
of g. We then have
NG + #) =A) +B), (3)
for the right- and left-hand sides of this equation could at most
‘by an integral multiple of 9m, but as it is written both
tides agree for 4’ —0 and vary cantinnmsly with 4’ (8.3)
satisfies the condition (0) = 0 and we obtain from it the further
equations
A= §) = AS), Ali) = ANP), (6.4)
where h is any integer. On replacing ¢ in the second of these
equations by ¢/h we obtain
ae) = pr. (8.5)
It follows immediately from (8.4), (8.5) that for every rational
number kj (e, h integers)
ahd) = fae. (86)
In accordance with our assumptions X(2n) is an integral multiple
Qme of 2m. On setting y — Bn ia (8.6) we obtain the equation
AGG) = mp for all which ate tational fractions of 27; the
continuity requirement then allows us to assert its validity
for all real values of the argument ¢.M2 GROUPS AND THRIR REPRESENTATIONS
‘The simple equation
Din) x Dla!) — Din
is here valid.
Consider the function j(p) on the unit circle in the complex
x plane. JE the point p goes over into the point p” under the
rotation , the function f goes into a function f” which is defined
by the equation
fe) = fe).
‘The transition f—> f' is a linear correspondence in the co-dimen-
sional space of functions f(p) and is associated with the rotation
t; this obviously defines an co-dimensional representation of
the rotation group bs which we denote by & @ is unitary if
wwe take a5 the square of the absolute value of a “vector” f
the integral of [/(p)|! with respect to the element of are dp on
the unit circle. The fact that any function (satisfying suitable
Series meaus that in the seduction of & into its irreducible com-
ponents each of the I-dimensional representations Dim) occurs
bce and only once. Moré precisely, Uis reduction is to be inter-
preted with regard to the completeness relation,
(®) The Group of Rotations in 3-dimensional Space
We consider the functions f= f(P) on the unit sphere as
the vectors of an co-dimensional unitary space whose metric
is given by {/UP)|!o; do is the eurface element of the ephere
over which the integration is to be extended. It the point
goes over into ’=sP under the rotation s, the function f
oes over into the function j” defined by f(P')=f(P). The
surface harmonics ¥; of degree 1 (ef, II, § 4] obviously span a
(21+ 1)-dimensional sub-space ft, which is invariant under the
totality of transitions f+’ induced in function space by the
various elements 5 of the rotation group b = bj—here again we
speak of this representation as ®. They are consequently the
substratum of a certain representation ®, of 6 which is induced
in ®, by b. On choosing a definite direction ae that of the
svaxit we may, a8 in TT, §4, take the eet
YY (m= 1,1 —
as a basis for the surface harmonics of degree J. We then have
a unitary representation, and the sub-spaces ®t corresponding
to the various values 0, 1, 2,- - - of J are mutually perpendicular
ROTATION AND LORENTZ GROUPS 49
in the unitary sense (arthagonality properties of surface har-
monies). b contains the 2-dimensional rotation group by—c-
a5 the sub-group of rotations about the saxis, The stiucture of
Yfp! shows that on restricting , to this sub-group by the
representation ; is reduced into the I-dimensional representa-
tions D" for which m=i,1—1,+++,—L The fact that
any function on the unit sphere possesses a unique expansion
jn terms of surface harmonies means that on reducing # into
its irreducible components cach of the representations D,, 1 — 0,
4, 2, *- +, occurs exactly once. This reveals the true signif
cance of Surface harmonies; tey are characterized by the
fundamental symmetry properties here developed, and the
solution of the potential equation in polar co-ordinates is merely
an accidental approach to their theory
Rotations are orthogonal transformations of three variables
4,9, % Iwo wish to include with the proper rotations with
dcicrminant -+ 1 also the improper once with determinant — 1
~* auguiened rotation group °"—this can be done by intro-
ducing the reftection
ives, yary va—
67
in the origin, Its reiteration if is the identity, and it commutes,
swith ail rotations. The matrix corresponding to it in the
representation defined by the surface harmonics of degree 1 is
the (21 + 1)-dimensional matrix (— 4)}, for the surface harmonics
of degree J are homogencous polynomials of degree I in x, 7,
We can thus obtain wu representations Dj, Dp of the aug
‘mented rotation group from the representation ®; of proper
rotations ; these two coincide with D, for proper rotations,
but in the first the matrix associated with the reflection iis +
whereas in the second it is — 1. We call this + 1 the signature
of the representation, Hence in the cm-dimensional repre:
sentation § of the augmented group b” cach D, occurs once
with signature (— IJ), but not with the opposite signature,
Although we are not’ as yet in a position. to prove it, the
Dy (L=0, 1, % ++) constitute a complete system of in-
equivalent irreducible (single-valued) representations of the
rotation group b, and the Dir, D> together constitute such a
ystem for the augmented rotation group 0
Now consider the unitary function space of ll functions
JP) in 8-dimensional space for which the integral |f|* over all
space is finite, Let the representation induced in this space
by rotations s, in which the transition from to the transformed
function f' ='sf is associated with s, be denoted by @. Each144 GROUPS AND THEIR REPRESENTATIONS
function /(P) can be expanded in a series of terms of the form
(0) * Vi Choose a complete ortiogonal system di(r), da(r), + = =
Jn the domain of functions (7) of the radius r, in the eense of
the equations
JrBalriba(ridr = 8,
‘The functions of the form ¢a(r) - Ys then constitute a (+ 1)-
dimensional sub-space qx which is invariant under rotations
‘and in which © induces ‘the representation ,. Different Rae
‘are mutually unitary-orthogonal. Each ®, then appears in
© infinitely often, its various occurrences being distinguished by
the ‘‘radial quantum number” m. Consider the analysis of
single electron spectra given in Chap. TI, §, in the light of these
mathematical developments, We then see that the azimuthal
quantum number 1 is of purely group-theoretic significance,
‘whereas the radial quantum number m refers to the dynamical
situation, for the manner in which the orthogonal system 4.(7)
is to be chosen is determined by the dynamical differential
equation.
‘The proper rotations of 3-dimensional Euclidean space about
tthe origin of Cartesian co-ordinates x, 9, 2, i. the real orthog-
onal transformations with determinant +1, are most easily
represented by a stereographic projection of the unit ephere
bout ths origin on to the equatorial plane » — 6, the wut pole
of the sphere being the centre of projection.” If the point
(#,, ¥, 0} be the image on the plane of the point (x. ». 2) on the
sphere and we write {= 2 + iy’, the formule for the projection
stipe th, ey iat
GTN Eee INTEC MT te
But it is preferable to introduce the two homageneons complexe
co-ordinates £, 9 in place of { by means of the equation f= y/f 5
the south pole #071 is thea included, We then have"
xhiyeiy: oe 1
an 2 265 2 ae +
Accordingly each unitary transformation
ok abt iy, af = He + oq
‘of the co-ordinates g, 9 corresponds to a rotation $ of the sphere,
the points of which are represented by the rays € 9 of Dedimen’
sional unitary space, Since, as is readily seen, any point and
ROTATION AND LORENTZ GROUPS 145,
tangential direction through it on the sphere can be carried
over into any other such configuration on the sphere by means
ot such rotations, we obtain in this way all rotations. Since
we are only concerned with the ratios of the coefficients 2, 8,
y. 8, the arbitrary factor of proportionality may be chosen in
such a way that the determinant of the transformation is 1
Nevertheless this normalization is somewhat artificial as the
correspondence is still double valued, for on multiplying thé
cocfiicicnts of the unitary transformation by — 1, ie. ou going
over from v ty —y, the normalizatiun is unaffected, Heuce to
cach element @, (T-4), of the unimodular univary group u corre-
sponds a rotation S:0->s under which the co-ordinates
+ ty, x — ty, s transtorm like
nf, 2H, GE — a5, (8.8)
1 a 7
a~akt hi y~ cobb), s~ ke (8.9)
(The eymbol ~, which we occasionally employ, means that the
expression on the left transforms like the onc on the right.)
We obtain in this way all rotations, each one exactly bwice.
The rotations about the saxis are oblaiued from Ue “principal
transformations"
faagr-b
of u In fact, on setting € e(w) the angle of rot:
about the z-axis is $= — 20. In virtue of the correspond:
os the sutations in S-dimensions constitute a represcntation
of the group w; and, conversely, the assuviation 9—> uv is a
representation of the group 0= by uf S-dimensional rotations
by t, although this representation is double-valued, In virtue
of this correspondence $—> o any representation U(o} of tt yields
a representation of by ("'I’ process,” § 8) ; , may thus be thought
of 5 a representation of D,, in which case we write it 9, where
faye. ine ("even") with inf are sgvae,
those with half-integral (ic, half an odd integer} j are double
valued. On restticting Use gioup by to the sub-group dy of
rotations about the s-axis Dy is reduced uly the 2/-+1 onc:
dimensional representations Di"! (m= j,j—1,**%,—3). To
show this we first note that the substratum ot our representation
D, consists of the monomials (7.2)
G+k=%, i-k=2m),146 GROUPS AND THEIR REPRESENTATIONS
where m runs through the values j, j—1, +++, —j. The
transformation induced on these variables’ by a rotation
about the saxis is arcordingly
(mm) > e(— mg) + x(m).
‘The representation o> 3 of i is itself contained among the
representations D, of ut constructed above; it is, in fact, Dy.
To show this we note that if (6, 9), (f°, 9') be subjected to the
same transformation o of u, then the determinant £9’ — mf’, as
well as €€-+ 7, is invariant. Consequently (£, 4) transform'co-
grediently to (7', — £'), or as (y, — ); hence
Btn ay ewe, 0)
The representations D, with integral j are identical with those.
obtained above as the representations induced on surface ha
monies af order j, for each polynomial in x, 9, ¢ of degree j is,
in virtue of (8.16), equivalent to a form of order 23 in €, 9.
If we wish to augment = ty in a manner paralleling the
augmentation of b=b, by the improper rotation (reflection
in the origin) we must consider it as an abstract group rather
than a group of linear transformations in two variables, Denote
the element corresponding to ¢ by « and the elements of the
original 1 by @ as before. We deine the augmented u' as the
totality of elements of the types o and ie; + must naturally
obey the multiplication laws
=o, uw
Gy and Gare then those representations of u' which coincide
‘with ©, for elements of the restricted group u and which as-
sociate with the element « the unit matrix +1 and its negative
= 1, respectively. ‘The sign +t is again called the signature
The representation G7 astotates the augmented rotation group
with w.
(©) The Lorente Group
Let the 3-dimensional Euclidean space be referred to homo-
geneous projective co-ordinates 2, (x = 0, 1, 2, 8) defined by
(8.11)
ROTATION AND LORENTZ GROUPS uy
and the formnle for the sterengraphie projection considered
above become
a=itm a
\n-18, a=
On subjecting £, to an arbityary linear transformation @ the
xq undergo a corresponding zeal linear transformation s which
Teaves the equation (R11) invariant Tf the ahsalute value of
the determinant of o is 1, we can readily chow that the form
e+ at atm (8.13)
is itself invariant under tive correspouding s, and that the deter-
minant of $38 + 1.
We now consider = ch, x, 2, 3 as the co-ordinates of
space-time; (8.11) is then the equation of the light-cone, the
generators of which are the possible paths for a beam of light.
In the restricted thehry of relativity normal co-ordinate systems
for space-time are connected with each other by arbitrary
Lorenta transformations, i.e. by any real linear transformation
which leaves the form (8.13) invariant and which docs not
interchange past and future. Lorentz trausfonwations con-
stitute a group, tne “ complete Lorentz group,” and this group
describes the homogeneity of the 4-dimensional world, ‘This
group consists of “ positive” and “ negative” transformations,
ice. transformations with determinants + 1 and — 1, respectively.
‘The first constitute the “ restricted Lorentz gromp,” fram which
the complete group is obtained by inteadusing in addition the
spatial reflection
Kyo hy hy
(8.19)
ty
(a=1,2, 3). (B14)
Under the restricted group right and left, as well as past and
future, are fundamentally different. Since the expression for
‘zg in (8.12) is positive definite, we may state the result vbtained
above in the form: any linear transformation of &, 9, with deter
‘minant of absolute value 1, induces a positive Lorents transforma
tion s in the x. Transformations o which differ only by a factor
#4 of absolute value | give rise to thesames. The correspondence
a» ¢is naturally a representation.
‘The question of whether every positive Lorents transformation
g can be obtained in thie way arieos immediatoly. That thie
is in fact the case can be seen frum general continuity con
siderations, for the positive Lorentz transformations constitute
a single connected continuum. But it is also easily proved by
elementary methods. Since we have seen in (0) above that the148 GROUPS ANT THEIR REPRESENTATIONS
rotations of space $ are obtained from the unitary transforma-
tions o, we need only to examine the Lorentz. transformation
tH) > eet A), an) Ha,
affecting the time axis, where a is a real non-vanishing constant,
But this transformation is obtained from the unimodular o:
frag, node
Returning to the general case, the correspondence s—>o is @
dimensional representation of the restricted Lorentz group.
But o is determined by s only to within the arbitrary “ gauge
factor" e*; we may therefore normalize it by the condition
that the determinant of o shail itself be unity, not merely its
absolute value. Even so, @ remains double-valued, for ~o
satisfies the normalizing condition as well aso, This repre-
sentation $->@ contains the representation of the rotation
‘group considered in (6) on allowing s to run through the sub-
group of spatial rotations contained in the restricted Lorentz
group.
‘The expressions (8.12) are Hermitian forms with matrices
to} you po —4| allllo
heise | :
Lo af lo} li 0 o-1
Hence if x denotes the one-columned matrix with elements
equations (8.12) may be written
y= ESE (8.16)
On replacing £, y by i, — € the x, undergo the spatial re-
flection (8.14). That is one way of including the negative
Amirenite transformations. But if we require that the corre-
npunsling transformation of , 7 be linear, we must introduce in
uldition tog» (, 9) a second pair g/ = (¢, n') which undergoes
the transformation af contragredient to 3. ‘Then
(3, - H(G, 7!) to within the factor d,
ln, ~~ ~(2 3) to within the factor d,
. , Se
(8.18)
| =
the quantities
ROTATION AND LORENTZ GROUPS 149
undergo the same transformation s as (8.16), provided the
absolute value of the determinant of o is 1. The same is true
for any linear combination of the two, eg. x. + x1. Henre the
quantities
te ES ET ESE (8.17)
undergo the given positive Lorentz transformation s when &, 7
ace subjected to a cortain transformation ¢ and simultaneously
£, W[ to the transformation o! contragredicnt to . Furthermore,
they undergo the transformation (8.14) om interchanging the two
pairs z, F, ie. on subjecting the four variables to the crans-
formation
Tibsk, non: Psk on (B18)
The expression
a+ a
is invariant in virtue of the transformation law of {', 1 defined
above. To obtain an expression which is also invariant under
the interchange (8.18) we must aud to the above the expression
obtained from it by this interchange :
(+ in!) + GE + an) (A.19)
It will be found advantageous to denote the column con-
sisting of the four elements (€, 9: @. 1') by a single letter z.
Let that linear transformation’ of these four variables which
transforms £, 1 in accordance with S, and €', 9’ in accordance
Si be denoted simply by S,: (8-17) then becomes
4 = EST. (8.16/)
‘We must now ask to what extent the lincar transformation o
of the four variables x is determined by the requirement that
it induce a given (positive or negative) Lorentz transforuration
$ of the Hermitian torms x,. It suffices for this purpose co
inquire what transformations ot the £ induce the identity on
the variables x... The only transformations of this latter kind
are those which multiply &, » with a common factor e# of absolute
vahie | and at the same time €', 9' with any factor e (inde.
pendent of the first) of absolute value 1. Rut @ can be more
precisely apecified by the requirement that (8.19), ie. FT'z, he
also invariant. The two arbitrary “ gauge factors"