The London, Edinburgh, and Dublin Philosophical

Magazine and Journal of Science

Series 7

ISSN: 1941-5982 (Print) 1941-5990 (Online) Journal homepage: https://www.tandfonline.com/loi/tphm18

CIII. Communication theory and physics

D. Gabor

To cite this article: D. Gabor (1950) CIII. Communication theory and physics, The London,
Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 41:322, 1161-1187, DOI:
10.1080/14786445008561157

To link to this article: https://doi.org/10.1080/14786445008561157

Published online: 29 Jul 2010.

Submit your article to this journal

Article views: 45

View related articles

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tphm19
 [ 116~ ]

CIII. Communication Theory and Physics.

B y D. GABOR,
Imperial College, London*.
[Received August 10, 1950.]

SUMMARY.
The electromagnetic signals used in communication are subject to the
general laws of radiation. One obtains a complete representation of a
signal b y dividing the time-frequency plane into cells of unit area and
associating with every cell a " ladder " of distinguishable steps in signal
intensity. The steps are determined b y Einstein's law of energy
fluctuation, involving both waves and photons.
This representation, however, gives only one datum per cell, viz.,
the energy, while in the classical description one has two d a t a : an
amplitude and a phase. It is shown in the second part of the paper that
both descriptions are practically equivalent in the long-wave region, or
for strong signals, as they contain approximately the same number of
independent, distinguishable data, but the classical description is always
a little less complete than the quantum description. In the best possible
experimental analysis by an electronic device the number of distinguishable
steps in the measurement of amplitude and phase is only the fourth
root of the number of photons. Thus it takes a hundred million photons
per cell in order to define amplitude and phase to one per cent each.

COM~U~ICATIO~ theory has up to now developed mainly on

mathematical lines, taking for granted the physical significance of the
quantities which figure in its formalism. But communication is the
transmission of physical effects from one system to another, hence
communication theory should be considered as a branch of physics.
Thus, it is necessary to embody in its foundations such fundamental
physical data as the quantum of action, and the discreteness of electric
charges. This is not only of theoretical interest. With the progress of
electrical communications towards higher and higher frequencies, we are
approaching a region in which quantum effects become all-important.
Nor must one forget that vision, one of the most important paths of
communication, is based essentially on quantum phenomena.
Some years ago, I have proposed a mathematical framework for the
representation of signals (Gabor 1946). I have been rightly criticized
for having lef~ out noise, which is an essential feature of all communica-
tions. This will be remedied here, and at the same time the description
will be brought in line with modern physics. But as the mathematical
frame will serve as a useful foundation, it will be necessary to give first
a short review of it.
* Communicated by the Author.
1162 D. Gabor on the

§ 1. CLASSICAL I%EPRESEIqTATION OF SIGNALS.

The previous work started from the observation that the description
of a signal in the conventional way, as a continuous function of time, is
redundant and non-physical. A continuous function contains in any
interval, however small, an infinity of data, corresponding to an infinite
range of frequencies. A similar objection can be raised against the
Fourier representation, which involves infinite time. I n the new
description one considers the signal simultaneously as a function of
frequency a n d time. I t is convenient to use only positive frequencies in
the description. This can be done b y introducing a certain complex
function, whose real part is the physical signal. (The theory of these
complex or " analytical " signals has in the meantime received interesting
additions b y the work of J. A. Ville (1948)). I f the time-frequency
half-plane, fig. 1, is divided, by any network, into cells of unit area~

Fig. 1.

C~÷I,K
ci,K c4K+J
W
Cz'_l.K _At
oJ dv
L
q,_
--F
II

Information diagram. The time-frequency haff-plane is divided up into cells

of lmit area and an elemeatary signal is asspciated with each, witl~
coefficient cik.

A t A v = 1, one finds that the signal in any domain contMning a sufficiently

large number of cells is fully described b y associating two real data, or
one complex datum, with every cell. In other words, each cell has two
degrees of freedom.
One can represent an arbitrary signal in an infinity of ways as a linear
function of certain " elementary " signals, associated with the individual
cells. There is, however, one description of particular interest, in which
the elementary signals are harmonic functions, modulated with a
" gaussian " signal, i.e., they have envelopes of probability shape.
 C o m m u n i c a t i o n Theory and P h y s i c s 1163,

(Fig. 2.) These share with other functions the p r o p e r t y t h a t their Fourier
t r a n s f o r m s h a v e the same shape, b u t t h e y are unique in t h a t respect
t h a t the p r o d u c t of their " effective " duration a n d of their effective
f r e q u e n c y w i d t h is the smallest possible for a n y function. Thus, it can
be said t h a t these e l e m e n t a r y functions overlap as little as possible.
T h e y h a v e also the a d v a n t a g e t h a t the familiar concepts " amplitude "
and " p h a s e " can be used in connection with t h e m in the same w a y as
with infinite harmonic functions. The complex e l e m e n t a r y function is
Cos Jr jSin, if we denote b y Cos the even, and b y Sin the odd t y p e of
real e l e m e n t a r y signal. Multiplying these with suitable complex
coefficients ci~, as indicated in fig. 1, we can represent a n y a r b i t r a r y
signal. Time-description and frequency-description (Fourier integral}

Fig. 2.

Cosine t'y/z~

[ " _.

L. 5/he t y p e l

Elementary signals of the " cosine type " (even), and of the " sine type " (odd).

can be considered as e x t r e m e special cases of this representation. I n t h e

first case the even e l e m e n t a r y signal degenerates to a delta-function,
and the odd one to its derivative, in the second case b o t h become o r d i n a r y
harmonic functions.
The " m a t r i x " representation, illustrated in fig. 1, is p r o o f against the
objections raised against the pure " time " or " f r e q u e n c y " descriptions,
b u t it does not go far enough. The infinity of d a t a has been reduced t o
a finite number, b u t these data, i.e., the coefficients c/k, are still supposed
to be exactly defined. B u t a single e x a c t d a t u m still contains an infinite
a m o u n t of information, i.e. an infinite n u m b e r of " yeses or noes "
I n reality, of course, these amplitudes, like e v e r y physical d a t u m , h a v e
a certain a m o u n t o f u n c e r t a i n t y or " noise." This has been t a k e n into
 1164 D. G a b o r on the

consideration in the m a t h e m a t i c a l theories of S h a n n o n (1948) a n d

Tuller (1949), where the noise amplitudes are assumed as known. B u t
we c a n n o t be satisfied w i t h this in a physical theory. E v e n if all
accidental imperfections o f the i n s t r u m e n t s are eliminated, there remain
certain basic uncertainties, which we are now going to investigate.

§2. STATISTICAL PROPERTIES OF THE INFORMATION CELL.

I n order to connect the m a t h e m a t i c a l scheme with physical reality,
w e must first observe t h a t e v e r y physical signal has a certain energy
associated with it. W e can also associate a certain e n e r g y with e v e r y cell*,
a n d , for simplicity, we will say t h a t it is " contained " in the cell.
W e will limit the discussion to electric communications, t h o u g h most
o f our results will a p p l y equally well to sound, or, in short, to com-
munication b y a n y q u a n t i t y which is considered as continuous in classical
physics. We observe first t h a t all electric signals are conveyed b y
radiation. E v e n if lines or cables are used in the transmission, b y the
M a x w e l l - P o y n t i n g theory, the e n e r g y can be located in e m p t y space.
Hence we can a p p l y to our problem the well-known results of the t h e o r y
o f radiation.
F o r simplicity, we consider our communication system as having the
uniform t e m p e r a t u r e T. T h e u n c e r t a i n t y connected with the concept
o f t e m p e r a t u r e will produce certain fluctuations in the e n e r g y of the
cells, which we can calculate b y the rules of statistical t h e r m o d y n a m i c s
once we know the law for the m e a n t h e r m a l e n e r g y ~T of a cell, in function
o f the t e m p e r a t u r e . This we obtain at once, if we observe t h a t e v e r y cell
in the signal has two degrees of freedom, of which only one c o u n t s for the
purpose of statistics, as the other is o f the n a t u r e of a phase. Thus, b~"
Planck's law
hv
~= exp (hv/kT)--l' " . . . . . . (1)
where r is the m e a n f r e q u e n c y of the cell. I n other words, we identify
e v e r y information cell with a "' P l a n c k o s c i l l a t o r "
This requires perhaps a little more explanation, as physic~ists are less
familiar with a discussion of radiation in terms of f r e q u e n c y and time
t h a n in terms of f r e q u e n c y and space. B u t the first case is i m m e d i a t e l y
reduced to the second if we imagine the signal propagating with the
velocity c, and plot the i n f o r m a t i o n diagram against the length ct instead
o f against the time t. The state of the field in such a linear system

* The energies of the elementary signals of the type discussed will not add up
exactly to the energy of the whole signal, because they are not quite orthogonal,
but the error will vanish in the mean over large number of cells. One can,
however, to meet objections, consider instead an orthogonal set of elementary
signals, such as the signals with " limited spectrum ", introduced by Shannon
(1949) and by Oswald (1949, 1950) which have uniform spectral density inside and
zero outside a frequency strip. This makes no difference to the following
discussion, as no reference will be made to any special type of elementary signal.
 Communication Theory and Physics 1165

(in which we consider one state of polarization only) can be represented

b y two systems of progressive waves in opposite directions, only one of
which represents the signal in which we are interested. W e count the
degrees of freedom in this wave s y s t e m - - w h i c h is w h a t we have done in
the last s e c t i o n - - a n d give it the e n e r g y ~T for each amplitude, disregarding
the phases. This is the application to the linear case of yon Laue's well-
k n o w n derivation o f Planck's law for the radiation density b y super-
position of plane waves (von Laue 1914).
The t h e r m a l e n e r g y ~T does not in itself represent " n o i s e " , i.e., an
uncertain disturbing factor. I t becomes disturbing only b y its
fluctuations. I n order to obtain the m e a n square fluctuations of the
energy, we a p p l y Einstein's formula
7._

~ e ~ : (eT-- ~T)~:kT2 ~ T, . . . . . . (2)

which gives $E~r:hv~-}-~ T . . . . . . . . . (3)

Einstein's i n t e r p r e t a t i o n of this e q u a t i o n is well k n o w n (cf. B o r n 1949).
The second t e r m has been identified b y H. A. L o r e n t z with the fluctuations
due to the interference of waves with r a n d o m phases.* B u t the first
t e r m suggests t h a t the e n e r g y is c o n c e n t r a t e d in light q u a n t a or p h o t o n s
of energy by, which fluctuate in a n y element as if t h e y were i n d e p e n d e n t
particles. As b o t h effects are present simultaneously, and are acting
as if t h e y were i n d e p e n d e n t of one another, it is not possible to m a k e
a simple physical picture of the process. B u t f o r t u n a t e l y this is n o t
necessary. Applying Einstein's a r g u m e n t to the case when a signal is
present, so t h a t the m e a n e n e r g y ~ in the cell exceeds the t h e r m a l m e a n
e n e r g y ~T, we obtain in A p p e n d i x I.
~e2= (E--~)~=hv~-~2~T-- ~T. . . . . . . (4}
Expressing the e n e r g y b y the n u m b e r o f photons N in the cell, so t h a t
e=-Nhp, ~T=NThv, this becomes
3N2= ( N - - N ) Z = N ( 1 ÷ 2 N ~ ) - - N ~ . . . . . (5)
with N~---- 1/[exp ( h v / k T ) - - l ] . . . . . . . . (6)
* In technical theories of thermal noise, it is usually forgotten that it is not
the noise power but its fluctuations which cause the disturbance. But if the
quantum effect is small and the se6ond term in equation (3) predominates, the
r.m.s, value of the noise power fluctuation is equal to the noise power itself,
hence this error is without consequences.
From equation (3), neglecting the quantum term, one can easily derive
Nyquists well-known rule (Nyquist 1928) that a resistance R can be considered
as containing a " noise generator " with a mean square electromotive force
E~=4kTRAv.
The proof can be given in the same form as Nyquist has done, by substituting
a cable with wave impedance R for the resistance. But one must add the,
condition that not only the noise power, but also its fluctuations follow the same
rule in the resistance as in the cable. This is no arbitrary rule ; the necessity o f
a universal law of mean square fluctuations follows directly from the second
principle, as Szil£rd (1925) has shown.
SER. 7, VOL. 4 I, NO. 322.--NOV. I950 4L
I 166 D. Gabor on the

NT, the number of " thermal " photons per cell, is a large number for
the frequencies used in electrical communications. One can use the
approximate formula
bT
- - kT
biT= ~ = ~ - ~=0.7~T

which gives biT=210 for a wavelength of 1 cm. and a temperature of

300°K. On the other hand for visible light, ~=5.10 -5 cm., the
approximate formula
N~=exp (--hv/l~T)=exp (--1/0.72~T)
gives N~=e-~a= 10-3% Hence for visible light, at ordinary temperatures,
there is practically no thermal noise, and the fluctuation becomes pure
" quantum noise ", which follows the law 3 N ~ : N .
These results enable us to construct a complete physical representation
of a signal. We see t h a t the state of an information cell is completely
determined by a stochastic number N, the number of photons. We can
now construct a scale or " l a d d e r " o f distinguishable states, on which
every step corresponds to a reasonably ascertainable difference. I t is
an evident suggestion to adopt the r.m.s, fluctuation of bi as the unit
step.* With this convention, the number of steps distinguishable below
a maximum level bi~ is, by equation (5),

2 . . . . . 2N~2
---- 1-~-2~r {INto(1+2NT)--N~]I/~--[NT(1 +NT)]I/2} ~ (1 +2N--T) 1/2" (7)

The last formula is valid for large signals.

In the useful terminology introduced by D. M. MacKay (1950), S is
the " proper scale " of the photons. Fig. 3 illustrates the representation
of a signal in three dimensions, with the photon scale at right angles to
the time-frequency plane. I f instead we plotted logz S, the ordinates
would give directly the equivalent number of binary selections or " bits "
This is the number of " y e s e s " or " n o e s " required to fix the position of
the signal on the ladder.
I t is of interest to enquire about the minimum energy required for the
transmission of the first " b i t " of information. By our convention, this
cannot be less than the r.m.s, value of the thermal energy fluctuations,
though it can be more. Combining equations (1) and (3), one obtains
- - hv
emin----(Se~)l'2 = 2 sinh (hv/2kT) . . . . . . (8)

* By the theorem of Bienaym6 and Tchebycheff, the probability of an error

k times exceeding the r.m.s, error is smaller than 1/k2, whatever the law of the
fluctuations may be.
 Communication Theory and Physics 1167

B u t , on t h e o t h e r h a n d , this energy c a n n o t be less t h a n one q u a n t u m , hr.

T h u s t h e e n e r g y required for the first " bit " is either emm as g i v e n b y
e q u a t i o n (8), or by, w h i c h e v e r is the larger of t h e two. As s h o w n in
fig. 4, t h e t w o lines cross at h v 0 : 0 . 9 6 k T , which is o n l y slightly less t h a n kT.

Fig. 3.

-_$

Representation of an arbitrary signal. A ladder or " proper scale " of

distinguishable values is erected in every cell, on which the occupation
is marked off.
Fig. 4.

f
0

The energy required for the transmission of the first " bit " of information.

T h u s u p to a critical f r e q u e n c y vcr, an e n e r g y kT is sufficient for the first

step, b u t no c o m m u n i c a t i o n is possible w i t h a n energy of less t h a n kT.
4L2
1168 D. Gabor on the

The interesting feature of this result is its generality; it applies to th~

unknown processes in the nervous system as well as to electrical
communications.
§ 3. S~A~STmS o~ S m N ~ s .
Up to this point, we have directed our attention only to one cell.
A short digression on signals and ensembles of signals m a y not be out o f
place before we return to the physical analysis of our results.
A signal is a system of, say, n cells, preferably, but not necessarily,
contiguous in the information plane. We consider a large number o f
such systems in a stationary transmission, in which they differ only by
their position in time, not in frequency, and we speak of them as of an
" ensemble ". Evidently the analogy with statistical mechanics is not
very perfect. I n statistical mechanics, we can either follow a system in
an ensemble over a long time, or we can look at all systems in the ensemble
simultaneously, while here we can take only the first view. I t is also
somewhat questionable whether the often used expression " ergodic "
is justified. In its original sense, due to WiUard Gibbs, it means t h a t
each system (each signal) spends equal times in all states compatible
with a given energy, which is not true for most stationary transmissions
usually considered as " ergodic " in communication theory. Thus, we
prefer to avoid this term.
The most important mean value in such an ensemble is t h a t of the
entropy. In order to clarify the connexions between physics and com-
munication theory, it m a y be useful to consider this problem in two stages.
In the first, we consider all configurations Of the system of n cells,
compatible with the energetic conditions of the transmission as equally
probable, and define the entropy as ]c log P, where P is the number o f
all these possible configurations. In the second stage, however, we give
them different probabilities or " w e i g h t s " . The first stage is in close
connection with physics; actually it is the problem of calculating the
entropy of an " ergodic " system, in the original meaning of the word.
According to quantum statistics, all simple, accessible states have equal
probability, and the levels of Planck oscillators are, of course, simple.
(Cf. Jordan 1933.)
As an example, let us estimate the mean entropy of groups of n cell~
in a transmission in which the mean energy level is S~, and the r.m.s.
deviation from the mean is ASn, for brevity. (The suffix n had to be added,
as the standard deviation is dependent on the size of the group.) For
simplicity, we assume (zJSn)~ S2, i.e., a small degree of "modulation".
This allows us to equate the number P of possible states to the number of
points with positive, integer coordinates inside a spherical shell in
n-dimensions, whose mean radius R is given by
R 2 ~ n S 2,
while its thickness is
2 AR----nAS~/R.
 Communication Theory and Physics 1169

Well-known formulae for the volume of a n-dimensional sphere give

p: (7r/4)~n nRn-12Ai~
F(½n+l)
- - F(½n-~-l)(Tr/4)tnn(nS-~)l/~ --~-ASn~~ (½e~rS'~)tn V(n)
ASns_
~ (9)

We have used Stirling's formula to approximate the F-function for

large arguments. For sufficiently large n, however complicated the
signals, they must behave as if they were independent, and we must have
asymptotically
n. (A S~)2-> const. ;
hence, apart from a constant, the entropy per cell goes to
k log (½e~rS2)1/2. . . . . . . . (10)
a result also obtained b y Shannon (1949).
Apart from the factor k, the entropy (I0) is also a measure of the
information capacity, per cell of the transmission, in accordance with its
definition as the logarithm of the number of possible, equally probable,
selections. Thus, the concepts of information capacity and of entropy
are closely parallel, even identical.
Consider now for a moment the central problem of statistical com-
munication theory, which has been investigated by Shannon, the
communication of symbols, as supplied by some source. If the source
,supplied them with equal probability, one could connect the possible
and equally possible states of the signal with the symbols b y an infinity
of possible codes, and transmit information at the maximum rate, given
by (10) multiplied by the number of cells per second. But if, as it is mostly
the case, the symbols have different probabilities Pi, it would be
uneconomical to give them equal shares in the available signal entropy.
Those which occur more often, and hence carry less information, should be
given less than the more unexpected ones. In this way, though it is
not, of course, possible to make the channel carry more information than
given b y (10), it is possible to make it carry more symbols, each symbol in
the average being worth less. Shannon (1948) has shgwn that in the
case of optimum coding each bit of channel capacity can carry 1/H
symbols, where H is defined as
H ~ - - Z p i log Pi bits/symbol, . . . . . (11 )
i.e., as the mean value of a symbol in terms of binary selections. This
" selective entropy " is defined by the source of symbols, and it is at an
appreciable remove from the physical entropy of the signal previously
discussed, as m a y be already seen from the fact that it appears as its
divisor. The interesting properties of the expression (11), demonstrated
by Shannon, follow rather from its mathematical form than from its
intrinsic relation with Boltzmann's "H"
1170 D. Gabor on the

There exists, however, a general limiting relation between selective

information and entropy in the thermodynamical entropy of the source
considered as a physical system, which was revealed by Szil£rd (1929).
This is t h a t a signal corresponding in information value to an s-fold
selection enables the receiving system to reduce the entropy of the
transmitter by a maximum of/c log s. It follows, as he proved in detail,
that any mechanism acquiring this information must increase the entropy
by a minimum of /~log s, thus safeguarding the second principle o f
thermodynamics.

§ 4. THE CONNECTION BETWEEN QUANTUM REPRESENTATION

AND CLASSICAL DESCRIPTION.
Comparing the representation of a signal, illustrated in fig. 3, with
the classical, mathematical description, it appears t h a t we have lost
something. Previously we had two data per cell ; we are left now with
one only, a function of the quantum number N, which corresponds to the
energy. What has happened to the phase ?
I t is not difficult to give an answer to this question on general lines.
N is not an exact datum but a stochastic number, which represents a
finite amount of information. But between one datum of finite accuracy
and two, there is no transcendental abyss of the kind which exists between
a simple and a twofold infinity. One finite datum can very well contain
two independent, finite data, provided that their aggregate information
is not more than the original. I t will be shown t h a t this is indeed the case,
and t h a t the maximum amount of information on amplitude and phase
are both contained in the single photon scale.
An electromagnetic signal can be physically analysed in various ways,
but it will be instructive to consider first extreme cases only. One
extreme is a counter, an instrument which records single photons. It
follows immediately from I-Ieisenberg's uncertainty principle t h a t the
" time resolution " of such an instrument cannot be better t h a n a whole
cycle, hence the phase remains entirely unobservable. The other extreme
is any classical field-measuring instrument, capable of recording the
electromagnetic field in the signal as a function of time. I t m a y be
called for brevity a proportional amplifier, as amplification of weak signals
is one of its essential functions. There is no need to consider intermediate
instruments, as it will be seen that in a certain range of very weak signals
every classical amplifier operates as a sort of " proportional c o u n t e r "
One limit for the operation of any such instrument is given by the well-
known uncertainty relation of the quantum theory of radiation (cf.
tIeitler 1944, p. 68)
ANA¢~>I . . . . . . . . . . (12)
where AN is the uncertainty in the photon number N, and A~ the
uncertainty in the phase ¢. But this gives merely an upper limit in our
case, as we want to determine amplitude and phase simultaneously,
 Communication Theory and Physics 1171

each as accurately as possible. I t will be seen that, in fact, the limiting

accuracy which is praetieally obtainable is much below what might be
expected from the inequality (12).
We will approach the problem by a detailed analysis of a particular
type of proportional detector-amplifier. Subsequently, we will t r y to
improve its performance to the extreme limit. I t will then be found t h a t
allspecial features of the device vanish from the formulm, thus the final
result can be considered as of general validity.
Assume t h a t the electromagnetic signal is introduced into a rectangular
wave guide, in the TEl0 mode, i.e., with an electric field in the x-direction,
independent of the x-coordinate, which is at right angles to the direction
of propagation z. An electron beam passes through two small holes
along the x-axis, in the plane of maximum intensity, with velocity v
and current intensity J. The alternating acceleration and retardation

Fig. 5.

X$1 etec rwn be~tm~

f
t
a
[1111rl(f,~(s~

l---z, 1
Analysis of an electromagnetic signal in a wave guide by an electron beam.

of the electrons produces velocity modulation in the beam, which b y

well-known methods can be translated into current modulation, i.e.,
producing an alternating component superimposed on the mean value J .
One method utilizes the " bunching " of the electrons which takes place
at. a certain distance from the guide, but one can also deflect the beam
by a constant field and make it play between two collecting electrodes
close together. In either case, one obtains an alternating current which
in first approximation is proportional to the d.c. current and to the
relative accelerations and retardations suffered by the electrons in the
wave guide.
Let us measure the energy exchange between the electrons and the
field in quantum units, hv, where v is the mean frequency of the signal.
(It will soon be seen that, in order to make the exchange intense, the
waveband of the signal must be made so narrow t h a t it is permissible
to take the arithmetic mean.) Let N be the mean number of photons in
1172 D. Gabor on the

an information cell, which passes through a cross-section of a guide in

a time At:l/v. During this time N : J A t / e electrons pass through it,
and exchange in the mean n quanta with the field, either by losing or
by gaining energy. The positive number n is the essential parameter
of the process. I f ~ is a large number, which is possible only if N is also
large, the interchange will be essentially classic; if ~ is small, quantum
phenomena will dominate.*
The detailed calculations may be found in the Appendices II. and III.,
only the results will be discussed here. The first contains the classical
theory of exchange, valid for large n and N, the second a wave-mechanical
calculation valid forsmall exchange parameters. I n the classical theory
the result is

- ~ - ~ \ hc ] O vb_l--\2bv/ j v ~" (13)

Thus the exchange parameter is proportional to the square root of the
photon number, as m a y be expected. Of the dimensions of the wave guide
the width b appears explicitly in the factor of N, while the depth a is
contained in the angle
O:m,a/v . . . . . . . . . (14)
which is one-half of the " transit angle ". c is the velocity of light.
In addition, there appears a factor which is the reciprocal o f
hc
2~re2 -- 137,

the fundamental number which connects photons with electrons.

Anticipating that the best measurement will require an intense
interchange, we now try to increase the coefficient of N in equation (13)
by all available means. First we make the factor sin ~"0/0 a maximum.
This is 0.723 and is obtained with 0 : 6 7 °, i.e., a transit angle of 134 °.
This disposes of the depth a of the wave guide. The optimum width b
is determined by the condition that the group velocity

u:[1 - A
must be as small as possible. But the smallest value is reached when U
vanishes at. the low-frequency limit, v--½Av, of the band. Substituting
those values into equation (14), one obtains in the optimum case

= 1-~\c] N. . . . . . . (]5)

* Quantal energy exchange between electrons and the field in a wave guide at
high quantum numbers has been previously discussed by Lloyd P. Smith (1946),
but we cannot agree with most of his results. Monokinetic electrons and
exchange of sharply defined quanta on the one hand, well-defined entrance phases
and short transit times on the other, are mutually exclusive phenomena by the
Uncertainty Principle; hence we believe that only certain averages over Smith's
detailed results have physical significance.
 Communication Theory and Physics 1173

All special features of the device have vanished in this formula, apart,
perhaps, from the unimportant factor, 1.5. But it is evident that the
factor of N must always be much smaller than unity, while its best value
as will be shown later, is just unity. There exists, however, a further
possibility for improving the performance. Assume that we can make
each electron perform repeated passages through the guide, each transit,
in the opposite direction exactly half a cycle after the last. (This is
possible in principle, as the optimum transit angle is about 134°.) I f
the frequency were known beforehand, the number of passages would
be limited only by the consideration that, by repeated gains or losses,
the electron would be bound to fall out of synchronism. But if the
signal had a single frequency, known in advance, there would of course be
no communication. However, even if the frequency is known beforehand
Fig. 6.

nnn

Ifalqql |

UU

Repeated passages.
only within v±½Av, one can make the number P of passages as great as
v/Av, without risking a phase error of more than i½~r, and it can be shown
that these passages are almost of equal value, so that ~ is increased very
nearly by a factor__ P. The number of passages required to make the
coefficient of N in equation (15) unity is
Popt= 9.5 (v)l/~ \~-~v] . . . . . . . (16)

i.e., at least of the order ten. But this number must not exceed v/Av ;
hence we obtain the condition that, in order to realize optimum conditions,
the frequency band must be so restricted that

>20-2 . . . . . . . . (17)

This is a somewhat surprising result. In the mathematical representation,

it did not matter whether we divided up the frequency band into broad
or narrow strips. But by the intervention of the number 137,
(20"2= (137/1"5)~/~), it turns out that only narrow frequency bands are
1174 D. Gabor on the

capable of accurate analysis b y means of electrons. (If ions with charge Z

were used, one would have to replace 137 b y 137/Z ~, and the condition
would be less stringent.)
Thus, at least in theory, the device could be perfected up to the
optimum performance. The practical difficulties are of course evident.
In practice one would rather replace the wave guide b y a " high-Q "
resonator, but this would somewhat complicate the theory.
We thus find, assuming P passages, in the " classical " case
(¢) (18).

while the corresponding wave-mechanical formula, valid for very weak

interchange, is (Appendix III.)
1.o f

Fig. 7.

1 f/~Oenera[ law
o 3 9 lo

Mean quantum exchange between electrons and photons. The initial tangent
is calculated from wave mechanics, the asymptotic parabola classically.
Apart from the factor 2/3, the coefficient of N is" the same in both cases,
but this time the quantum exchange ~ is proportional to the photon
number itself, not to its square root It can be said, therefore, that
f o r small photon concentrations, the device acts as a counter, at large con-
centrations, as a field-measuring instrument. The intermediate region is
difficult to calculate, but, as shown in fig 7, the two branches can be
connected by a plausible curve.
One can also interpret the results in this way: If there are few photons
present, there will be few collisions, and--as shown in Appendix I I I . - -
equal probabilities of gains and losses, at any instant. With increasing
photon concentration, repeated collisions will increase in number, and the
resulting loss or gain increases with the square root of the photon number
only, b u t this resultant has now a prevailing direction, which changes
 Communication Theory and Physics 1175.

its sign with the frequency of the signal. At this stage the " classical
field " has developed.
Having ascertained that, with certain reservations, we can make the
exchange as strong as w e l i k e , we ask the question: If we know the
average photoa number N, how must we adjust the electron beam
current J, and the exchange parameter ~ in order to measure the field
amplitude E with maximum accuracy ? And having made these adjust-
ments, how many steps shall we be able to distinguish in the scale of the
field amplitudes ? Evidently this question has a precise meaning only in
the " classical" range of large ~ and N, and the following considerations
relate only to this case. In order to simplify the problem, we neglect
the thermal noise, i.e., we put NT~-0 , so that the relative accuracy on
the photon scale would be 1/N 1/2, and the total number of steps in the
photon ladder 2N 1/2. The calculations are carried out in Appendix IV.,
here we give only the physical considerations
The quantity to be measured is the electric amplitude in the information
cell, which is proportional to the square root of the photon number. The
measured quantity, on the other hand, is the alternating electron current,
which, as mentioned above, is proportional to ~. J or to n. M for not too
strong signals, M being the mean number of exploring electrons per cell.
For the optimum, we impose the condition that the relative mean square
deviation of the quantity measured from the quantity to be measured
shall be as small as possible, i.e.,
(nM--CN1/~) 2
=rain, . . . . . . . (20)
(uM)~
where the proportionality factor C is determined from the condition
(nM--CNlI2)___0.
We have to choose n and M so as to satisfy the condition (20) for given N.
There must be an optimum, for these reasons: A too weak interchange
would leave the cell unexplored. A too strong interchange, on the other
hand, will interfere with the object of the measurement and spoil it. Though
in the mean electrons are as often accelerated as retarded, fluctuations
in the numbers M1 and M 2 of electrons which pass through accelerating
and retarding phases might produce extra photons, which could not be
distinguished from those belonging to the signal, or annihilate some.
The spurious photons are generated according to a law
~N-~(M1--M2)-~n(~MI--~M2) . . . . . (21)
as M1----M2-~½M. We have written 8.zN for this number, considering it
as a fluctuation which must be added to the natural fluctuation 3~N,
whose law is 8~N2=N. The two fluctuations must be considered as
independent.
It is already evident from the above that the fluctuations in the number
of beam electrons, i.e., the " shot effect ", play an important part in
these phenomena. A too weak current has a high relative fluctuation.
A too strong current, especially aided b y a large exchange, will again spoil
1176 D. Gabor on the
the object. It may be noted that we have here a type of uncertainty which
springs directly from the fact that photons and electrons are discrete, without
any reference to the physical values of h and of e.*
The relative error according to equation (20) is calculated on the basis of
equation (21), together with such evident assumptions as the independence
-of the fluctuations of n and M and of the "natural" part of 6N. We
assume also '' normal " shot effect, ~ M Z = M . The result is

This is a minimum for

- .- - - -
nM=%.M=2N n L N . .
from which%=2&< This gives the simple rule that, for optimum analysis
.of the signal, one must take one electron for every st'ep in the scale of
the photons, and the interchange 7 i must be itself equal to one
distinguishable step a t the level p. This again is a general rule, quite
independent of the special model from which we started.
Substituting these values, we obtain for the minimum of the mean
square relative error in the measurement of amplitudes

.As we are dealing with large photon numbers only, the first term a t the
right-hand side can be neglected with respect to the second. Thus we
.see, applying the same rule which we have used in constructing the
photon ladder, that the smallest distinguishable relative step in the
.amplitude scale is ~ ' times
2 larger than the square root of the corre-
sponding quantity in the photon scale. I n other words, the proper
scale of the amplitzLdes will contain a l w a y ~less than the sqmre root of the
number of steps in the photon scale; apart from the factor 4 2 for the
reason that the optimum setting is, of course, possible for one level only.
Having determined the proper scale of the amplitudes, a simple
.application of the Uncertainty Principle (Appendix IV.) shows that the
proper scale of the phase must also contain less steps than the square
root of the photon scale. Thus, summing up, we see that the classical
description of the signal, by being too detailed, gives in fact a somewhat
smaller total amount o f 6nformatiok than the qua&mn description.
The classical methdd of description, though theoretically inferior, may
of course be still the best practically in the range of frequencies used for
electrical communications, where efficient photon counters are not
* I t may be noted, however, that the shot effect, that is to say the random
arrival of t,he exploring electrons, is a practical rather than a fundamental
limitation. As Prof. R. E. Peierls has pointed out to me, it must be ideally
possible to measure amplitude arid phase simultaneously up to the limit given
by the relation 12. In fact if one could make the exploring electrons arrive at
regular intervals, all terms but the first on the right-hand side of equation 22
would vanish. But as it is practically impossible to suppress the shot effect to
any significant degree, our results retain their practical importance.
 Communication Theory and Physics 1177

available. Conditions are different in the optical region, where detectors

of the counter t y p e - - s u c h as the eye--are not far from perfection. In
this region, analysis in terms of electromagnetic waves is as yet technically
impossible, but it is interesting to note that, even if it were possible, it
would not be very practical. Progress in the field of microwaves is now
actually approaching a region where the two different methods of analysis
m a y become competitive. We see from our results that it takes about
a hundred million photons per information cell in order to define amplitude
and phase of the signal to 1 per cent each. Remembering that at 1 cm.
wavelength the number of thermal photons per cell is only about 200,
it m a y be seen that the time m a y not be far off when the imperfections
of the classical method of description will manifest themselves even in
electrical communications.
Will it ever be possible to generate and to amplify light b y an extension
of microwave techniques ? Generation m a y not be unthinkable, but
we can show at once that a light-amplifier for the reception of weak-light
signals is impossible. It follows from our results that an electromagnetic
signal in which amplitude and phase are defined with a relative accuracy •
must carry at least a power
h~•-4(A~/~) . . . . . . . . . (25>
Applied to visible light with ~ - 5 0 0 0 A., a line width of I A. and •~0.01,
we obtain 4.7 watts. Even if amplitude and phase were defined to
10 per cent only, the power would still be 0.0005 watts, a very high intensity
in terms of visible light. This example demonstrates that what we see
are always photons, not waves.
It m a y be hoped that these considerations have shown that the concepts
of information theory may well prove their usefulness when applied to
problems of physics.

]~EFERENCES.
BoR~, IV[., 1949, Natural Philosophy of Cause and Chanve (Oxford:
University Press), p. 81.
CARSON, J. R., and F~Y, T. C., 1937, Bell Syst. T. J., 16, 513.
GABOR, D., 1946, Journ. I.E.E., 93, III, 429 ; 1947, Ibid., 94, III, 369 ; 1947,
Nature, Lond., 159, 591.
I~EITLER, W., 1944, The Quantum Theory of Radiation (Oxford: University
Press, 2rid edn.).
JOrDAn, P., 1933, Statistische Mechanilc (Braunschweig : Fr. View•g).
vow LAu]~, M., 1914, Ann. d. Phys. [4] 44, 1197.
MACDON~LD, D. K. C., and Ko~P~E~, R., 1949, Proc. I.R.E., 37, 1424.
MACKAY, D. M., 1950, Phil. Mag. [7] 41, 189.
NYQws% H., 1928, Phys. Rev., 32, 753.
OSWALD, J., 1949, C.~., 229, 21.
OSWALD, J., 1950, Cables et Transmission, 2, 197.
VAn D ~ POL, B., 1930, Proc. I.R.E., 18, 1194 ; 1946, Journ. I.E.E., 93, III, 153.
SHA~O~, C. E., 1948, Bell Syst. T.J., 27, 379, 623 ; 1949, Proc. I.R.E., 37, 10.
S~rH, L. P., 1946, Phys. Rev., 69, 195.
SZmXRD, L., 1925, Zeitschr. f. Phys., 32, 753, 1925 ; 1929, Ibid., 53, 840.
TELLER, W. G., 1949, Proc. I,R.E., 37, 468.
VILLE, J. A., 1948, Cables et Transmissions, 1, 61.
1178 D. Gabor on the

APPENDIX I.
ENERGY FLUCTUATIONS IN AN INFORMATION CELL IN THE PRESENCE
OF A SIGNAL.
As in the case of t h e r m a l equilibrium, the m e a n square energy
fluctuation is the sum of a classical and of a q u a n t u m term.
The classical term. Consider, for simplicity, a (complex) Fourrier-
c o m p o n e n t E s of the signal, and a c o m p o n e n t E T of the t h e r m a l noise,
corresponding to two different (circular) frequencies, say, wl and ~o~.
The instantaneous e n e r g y density, resulting from the interference, is
proportional to
E s E ~ ' + E T E ~ + {EsE ~ exp [ j ( w l - - w 2 ) t ] + E ~ E T exp [ - - j ( % - - w2)t] },
where the asterisks stand for the conjugate complex values. The first
two terms are the energy of the signal and the energy of the noise, the
rest arises from interference. (Beats.) L e t us write
e=eS+eT+EST.
The m e a n value of the interference energy ~sT is nil, b u t its m e a n square is
2 *
%,r~--2EsEs . ETE~=2~se T.
Using this and the relations
ESEST ~ ~TEST = 0
which are evident, as there is no correlation between the signal and the
noise, one obtains
(~-~)'~ = [(Es + ~ + %T) - (% +~T)] 2 = ~-~ + 2 ~s~T"
.One knows the first two t e r m s from Lorentz's calculation, which gives,
in the absence of a signal
2 -2 -2
ET - - ET - ET

hence (e--~)2=2¢S~T+~,
or, as %=E--ET,
(~--~)~L=(2~--~T)~T . . . . . . . . (I.1)
This is the classical p a r t o f the fluctuation, arising from t h e interference
of waves. The quantum term is always given b y the " law of rare events "
(~--~)~u.=hv~ . . .. . . . . . (I.2)
as if the energy were present in the form of particles (photons), with
energy hr. Adding (I.1) and {I.2), one obtains e q u a t i o n (4).

APPENDIX II.
CLASSICAL ENERGY EXCHANGE BETWEEN A WAVE AND AN ELECTRON
BEAM.
I n a rectangular wave guide with cross-section a × b (fig. 5), the com-
p o n e n t s of the electric field are, in the TE01 mode,

= 0cos
 Communication Theory and Physics 1179

where V is the phase velocity

V = e E l - - (\267]c
~2-]1=_1

Thus the mean electric energy is, per unit length of wave guide,
1 i_ 2
8¢r 4 E° ab,
and the mean total energy, electric and magnetic, is twice as much.
This energy moves with the group velocity U=c~/V. The mean number
of photons in the information cell is obtained b y equating the energy
flux expressed by N and expressed b y E 0. As the cell occupies a time
A t = I / A v , the flux per second, i.e., the power, is
-- i 1 F [ c ~2]i/~.
hvNAv = ~ ½E~abU = ~ E2°abc h 1 -- \2b-v~] .J " (II. 1 )

Consider an electron which traverses the guide on the x-axis. Assume

that its velocity v is large enough to neglect its change during the transit.
Thus the number of quant~ exchanged--gained or lost--will be, in the
mean,

= E dx = I E dt
J --1/2a J to

where the mean is to be taken with respect to the entrance time t 0.

A simple calculation gives
= _2 sin 0 ea
7r 0 hvE° (0<~), . . . . . (II.2)

where 0 is half the " transit angle," i.e.,

O:~va/v.
The condition 0 < v for the validity of (11.2) is always satisfied in an
efficient arrangement, in fact, one must take 0 <½~.
Eliminating the field amplitude E 0 between (II.1) and (II.2), one
obtains a direct relation between the photon number N and the exchange
parameter n :

_i-kFfQ] 7-N
I 32 sine0 v [1 [ e "~27-1/~Av--
--137 r~ 0 vb_ -\2-b-v] J TN" (II.3)
As explained in the text, the coefficient of N must be made as large as
possible. The depth a appears only in the form sin 2 0/0, whose maximum
value 0.723 is reached at 0 = 6 7 °, i.e., at a transit angle of 134 °. The
width b, on the other hand, appears in the factor
1/b [1 / c \2ql/2
1180 D. Gabor on the

The smallest admissible value of U is obtained if the group velocity is

zero at the lower end of the frequency band, i.e., if frequencies below
v=½/lv are cut off. In this ease
b = ½ c / ( v - - l Av),
hence the best value of the factor in question is the reciprocal of
[rov
b i
o I 4.)J '

assuming Av/v < 1, i.e., the band to be narrow. Substituting these values,
one obtains
1-~ c \-~-] N . . . . . . . (II.4}
which is equation (15) of the text.
I f repeated passages are used, it is evidently permissible to consider
them of equal value, so long as their number P is small. I t m a y be asked
when this assumption will lead to an appreciable error. An estimate can
be made b y introducing the concept of the "instantaneous frequency"*
which will vary slowly between the limits v~½Av. A somewhat long
calculation, which m a y be omitted, gives the result that for large P's the
factor of ~ is not exactly P but approximately

t :,.i. ]=P .... ],

where 8v is the deviation of the instantaneous from the mean frequency.
Assuming that the instantaneous frequency is uniformly distributed in the
available band Av, one has 8ve=l/12(Av) ~. In the text, we have assumed
that the maximum number of admissible passages is Pmax=V/Av. Even
this gives an error of less than 14 per cent, which justifies our approximate
treatment.

APPENDIX Ill.
ENERGY EXCHANGE OF ELECTRONS AND W E A K FIELDS ACCORDING TO
QUANTUM MECHANICS.
We apply to the problem the standard perturbation method of wave
mechanics. As the beam width can be considered as small in comparison
to the spatial periods 0 f t h e field, it is sufficient to start from Schr0dinger's
one-dimensional equation with vector potential A z,
h OT h~ 02T ieh . @T
2~ri Ot = 8~r2m'Ox 2 ~-~-m-~Ax ~-~. (III.ly
We put
A x = A o cos ~ot, Ay----A~-~0, o~o=2~vo,

* This concept is due to I-Ielmholtz and has been first applied in communication
theory by Balth. v.d. Pol. (1930). Cf. also J. R. Carson and T. C. Fry (1937).
 Communication Theory and _Physics 1181

which corresponds to an electric field

E x = E o sin o~ot, Ev=E,=0 , Eo=Ao/oJoC,
~o is here the frequency of the field, which previously we have called v,
b u t in this cMculation we will reserve v for the frequency of the oscillators
.by which we will later replace the electron.
Assume t h a t the solutions for zero field are written in the form
W ~ = ~ exp (--i2~%t/h),
where % is the energy of the unperturbed electron in some state " n "
The general solution of equation (III.1) can be written in the form

With this substitution, equation (III.1) becomes

dc m e O~m (III.2)

Let us assume t h a t the functions ¢~(x) are orthonormal

f ¢*¢~ d x = ~ rim.

Multiplying equation (IIL2) by T*, integrating over the whole domain and
making use of the orthonomality conditions, we obtain ordinary differential
,equations for the coefficients %(0 in the form
dc r ~'eAo _ .
-- :i ~ Z c,~pmr{ex p [i(~o0+~o~)t]+exp [--i(wo--wm)t]} ,
dt men ,~
(111.3)
hr,O
where Pro,= ~ J~b, - ~ ~,~ dx

is the matrix element of the momentum p of the electron, and

~,~= ~ (~- ~m)

is the circular frequency which corresponds to the transition r - ~ m.
Assume now t h a t at the instant to the electron is in the state m = 0 .
Thus at the start the coefficient co is unity, all the others are zero. Let us
also assume that during the time of interaction, i.e., during the passage of
the electron, the other states c~ increase slowly enough to neglect tran-
sitions other than 0 -~ r. Thus we obtain, integrating equation (III.3) for
the coefficient cr at the time t,

" neA°-~-~Por,.~exp [i(oJo+¢%)t]--ex

c~(t)=" p(a~o+t%) [i(oJ0+c%)to]

_ exp [--i(~o--O~)t]--exp [--i(oJo--%)to! \

(~o-~,) J
SER. 7, VOL. 41, ~O. 322.--NOV. 1950 4M
1182 D . G~bor on the

The absolute square of c r is the probal~ility t h a t after the time t - - t o the-

electron will be in the state " r ", having absorbed a n e n e r g y corresponding
to the f r e q u e n c y ~or0=w r. Substituting E 0 instead of A 0, we find finally
for the p r o b a b i l i t y distribution of the electron a t time t the expression

{\mt, o] Ipo, is +

+ 2 cos 2Wot sin-2 ½°Jr(t--t°)\ (III.4)

C002-- Cot2 J "

There are several interesting features in which this differs from the classical
result. N o t e first, t h a t t h e expression is even in wr, hence an electron
has the same probability to gain or to lose a certain a m o u n t of energy.
N o t e also t h a t only the last t e r m depends on the phase of the electric field
at the instant t o, the others d e p e n d only on the interaction time t - - t o . B u t
even this last t e r m has a f r e q u e n c y twice t h a t of the electric field. I t is
r a t h e r d o u b t f u l w h e t h e r this corresponds to an e x p e r i m e n t a l l y observable
effect. Our calculation re]ares to weak fields, where ~ will be smaller t h a n
u n i t y ; b u t in order to observe even n---- 1, the electron b e a m m u s t be mono-
chromatized to such an e x t e n t t h a t its entrance phase becomes indefinite
within a whole cycle. This is of no i m p o r t a n c e for our subsequent
calculations, as we will consider only the m e a n value of (III.4), averaged
over the entrance times to, in which the last t e r m vanishes.
There remains now the problem to a p p l y e q u a t i o n (III.4), which is a
well-known result of wave mechanics, to our problem of an electron
traversing the oscillating field between two conducting plates. This means
calculating the m a t r i x of the (mechanical) m o m e n t u m p. We will evaluate
this b y a classical method, m a k i n g use o f the correspondence principle.
This is justified in our case, as the de Broglie wavelength of the electrons
is always v e r y small with respect to the distance a between the plates.
I n order to simplify the calculations, and to avoid singularities, we
replace the electron b y a plane density wave of t o t a l charge e (per unit
cross-section)

p(x, t)---- ~ e x p [--(t--to--x/vf/~-]. D(x). (III.5).

D(x) is an " annihilation factor " which is u n i t y inside the guide, i.e., for
--½a ~ x ~½a, and zero outside. A p a r t from this, i.e., ff the electron is at
more t h a n a b o u t w from the wails, the distribution is gaussian. ~ is a
small time, with which we will go u l t i m a t e l y to the limit zero. I t will
m a k e the calculations easier if for the start we ignore the factor D(x)
and m a k e use of it only towards the end.
W e m a k e use of the F o u r i e r formula

exp - - [(t--x/v)/~'] 2= V'(~)'r

;o exp [-- (~"rv)2 - 2~iv(t--x/v)] dr.
--00
 Communication Theory and Physics 1183

This is to say that we decompose the electron into travelling waves o f

density
p~ dv=e/v exp -- (Tr~v)2 . exp [--2~riv(t--x/v)] dr.
We have put the entrance time t 0 : 0 , as it plays no part in the mean values
which we want to calculate. In the following formulso we will drop also
the factor exp --(~rrv) 2, which cuts off the spectrum at very high fre-
quencies, but which is practically unity in the frequency region to be
investigated.
The electric field produced by the electron follows from Poisson's
equation
OE

Let us decompose this too into harmonic components by putting

E = I ® E~dv.
J --C~

O oob ain

for the progressive field waves which accompany the free electron. But in
a plane condenser with boundaries at x = ± ½ a , this field induces surface
charges
%(~=½a)=T~rE~(~a)=T~-~ex p --2~iu t T ~ ,

which cut off the field beyond x = ~=½a.

We eannow calculate the electric moment of the charges in the condenser,.
i.e., of the electron charge and the surface charges, whose sum total is zero.
The Fourier component of the moment in the frequency i n t e r v a l , , d, i~

dv {f½a xp, d x + la[%(½a)--%(--½a)] }

--½a
= , . ~ 5V sin (~rva/v) . exp (--27rivt) dr..

We now equate this to the " standard oscillator " of electromagnetic.

theory, i.e., to an electron of charge e which oscillates with a f r e q u e n c y ,
around an equal and opposite charge, with an instantaneous amplitude X,.
This gives
Y
X , d v = i 2-fi~v2sin (~rva/v) . exp (--2~ivt) dr.

The mechanical momentum of this equivalent oscillator is mX, dr. Thus

the momentum is, for the frequency interval , , dr,

P~ d r = mv sin (~rva/v) . exp (--2~zivt) dr. (III.6)

Try

We cannot introduce this directly into equation (III.4), because that

equation relates to discrete, steady states. In that case one has only to.
4M2
 1184 D. Gabor on the

take the absolute square of the matrix elements to form the quantities
]P0r [2 which figure in equation (III.4). But here we have to do with
continuous states, and we must also remember that the oscillators,
according to equation (III.6), replace the electron during the finite transit
time only. I n order to obtain the equivalents of t h e " oscillator strengths"
I P0r I~, we must now multiply the momentum amplitudes P~ dv with the
conjugate amplitudes P* dtz which belong to another interval/z, d/~ ; we
must integrate over the whole domain of the t~, and finally average over the
transit time a/v. I t is to say that we must calculate, instead of ! P0r [2,
V r½a/v (
]P0, [2 d r = 2 - dv | dt / P,P* d/z. (III.7)
a d - ½al. d -

We had to add the factor two, because in wave mechanics v is considered as

a positive quantity, while in the classical theory it runs from -- oo to -}- so.
Substituting from (III.6),
v
] P°~ ]2~2 \~r(mv~2vsinqrva/v/a ~-talv~½al°
dt :-~o~°~ sin
7r/~a/______v/~exp [27ri(/~--v)t] d/~.

After integration with respect to t, the double integral becomes

1 f~ sin~rt~a/v sin sin~r(l~--v)a/vdlz.

~Vith the substitutions x : (2/~/v)--1 and O=~rva/v, this is transformed into

~vv2f~_ ~ sin½0(lq-x).sin½0(1--X)l
-- x ~ dx = ~vv2~feosOx-cOSOo 1 -- x 2 dx-----lsinO'v

thus finally
(mv'~ 2v sin20
v dr.
] P0, I~ dv = 2 \--~--/ a --'-K- ..... (III.8)

This means simply that during the transit time a/v the coherence region of
any frequency v is v±v/2a, with a bandwidth v/a, whieh is the reeiproeal of
the transit time. A frequency v cooperates with any frequency/z inside
this band in producing the oscillator strength.
Substituting (III.8) instead of ]P0r 12 into equation (III.4),,where we
drop the last term whose mean value is zero, and replacing t--t o by the
transit time a/v, we obtain
1/2~re~ E~acv dv sin2 (~rvatv) ~Sing ~r<voq-v)a/v sin~r(vo--v)a/v~

(III.9)
This is the probability for an electron to be, after the passage, in a state
v, dr, that is to say, to have absorbed the energy hr. Thus the energy
spectrum of the originally monokinetic electrons after the passage through
the field is of the form
sin 2 Of sin 2 (0+0o) sin 2 (0--0o) ~
S ( O ) - - - - - ~ [ ~ff÷-~o)2 q- ~-~o)2 j , (III.10)
 Communication Theory and Physics 1185

where the frequencies (and energies) are expressed b y the half transit-
angles
O=rrva/v, Oo=rn,oa/V.
This function is shown in fig. 8 for a few values of 0o. Experimental
checking, t h o u g h difficult, m a y not be impossible (cf. Maedonald a n d
K o m p f n e r 1949).
The mean energy exchanged with the field, lost or gained, is by (III.10)
1 E~acv
~h~o = ~J o~ h~ I ~ I~ d~ - 137 4~3v~ F(0°)' (III.ll)

Fig. 8.
2

%,
1.5

-jsoo iso fzo 9o 8o 3o o w so 9o leoo ~ ,ao"

The function S(0), equation (III.10), which gives the energy distribution of
electrons after the transit. Gains and losses have equal probabilities.
The " resonance point " at which the electron has gained or lost one
quantum hv0 of the oscillating field is at 0 = 0o.

where F(0o) is the integral

F(0o)
; dO['sin 2 (0+0o)
~o ~ s in2 (0-00)7
Jo YL ~+-~o)~ ~ J
F o r 0 ~ 67 °, which was the most favourable case in the classical theory, t h e
value of the integral is found to be 1.22. Substituting this into equation
(III.11) and expressing, as before the field E o b y the n u m b e r of photons, N,
b y means of equation (II.1), and finally giving the wave guide the same
width b which was found the best in the classical case, one finds

137 c
where we have again written v for the field frequency, instead of vo. This
is the energy exchange in weak fields, discussed in the text.
1186 D. G a b o r on the

APPENDIX IV.
COI~TDITIO:NSFOR OPTIMUM EXPLORATION OF ELECTRIC FIELDS BY
ELECTRONS.
As explained in the t e x t , we a d o p t as the criterion of o p t i m u m measure-
m e n t the condition t h a t
(nM._CN~'~)~
(nM) 2
must be as small as possible. The p r o p o r t i o n a l i t y factor C is given b y the
condition
(riM -- CN ~/~)= 0,
or, as we are dealing here only with the " classical " case, in which the
fluctuations are relatively small, C = - n M . (N) -1~'. As the fluctuations of n,
the energy exchange p a r a m e t e r , and of M, the electron number, can be
s u p p o s e d as i n d e p e n d e n t of one another, we can write nM----~. M.
L e t us now first calculate

n2M2 = [ ( n + ~n)(M-- 3M)]2_

~ (~M -- ~SM ~-M3n)2 = ~2M'2-~-n23M2-}- M'~3n2
=n~2+~2~+~.
Here we have neglected the flunctuation t e r m of the highest order, and we
have assumed
3n~M-= 0, ~n ~-- n, ~M ~= M.
The first is the assumption of the relative independence of the fluctuations
o f n and M. The second is the assumption t h a t the a b s o r p t i o n of a
q u a n t u m is i n d e p e n d e n t of those previously absorbed or lost, which is
certainly admissible so long as the electron velocity is not changed
appreciably. The t h i r d assumption means shot effect w i t h o u t space charge
smoothing. This is a logical assumption in a t h e o r y in which each electron
is considered to interact singly with the field. W i t h these formulae, we
can now calculate

(nM--CN1/2)2=n~MU--2C(Yt+~n)(M +~M)(N1/2+ I~N(N) -1/2)

+ C~N1/2 +2l~ )
1~2 M )
1/22 ~ M 2n+ 1c 2 N ,
(IV.D
where the first t e r m is zero.
I n order to calculate the last term, we consider the fluctuation ~N o f
the photons as the sum of two i n d e p e n d e n t c o m p o n e n t s 8N1 and ~N2
~N=SNlq-~N2,
o f which the first represents t h e " n a t u r a l " f l u c t u a t i o n s , subject to the law
~BN~-=I~I, and the second is due to the exploring electrons, i.e., it is the excess
 Communication Theory and Physics 1187

•o f e m i t t e d over absorbed photons. L e t M s be the n u m b e r of b e a m elec-

trons which pass t h r o u g h the field in a r_etar_d~ng phase, Mz the n u m b e r of
a c c e l e r a t e d electrons. As in the m e a n MI-~Mz=½M, and as in the m e a n
e a c h electron produces or absorbs ~ photons, we can write
~N -----~(M1-- M 2) = ~(~M 1 - ~][ 2),
and ~N'=~ (~Ml-~-~M2)=/t (MI+M~)----n M,
where we have assumed 8M18M2~-0, in accordance with our assumption of a
normal, entirely r a n d o m shot effect.
W e now substitute these results into e q u a t i o n (IV.l), a n d replacing C~
b y its value ~M2/N, we obtain for the m e a n square relative error in the
m e a s u r e m e n t o f the amplitudes
(nM--CN1/~) ~ __ 1 1 _ [ 1 nM\
+ +

This is e q u a t i o n (24), discussed in the text.

I t is n o t so easy to give an e x a c t analysis of the errors in the measure-
m e n t o f phase, because devices for the running m e a s u r e m e n t of phase, with
uniform error, are b y no means simple. B u t the order of m a g n i t u d e can be
seen i m m e d i a t e l y ff we a p p l y the u n c e r t a i n t y principle to the exploring
electrons, in the form
8E~t~l.
8t is t h e u n c e r t a i n t y in the time at which the electron passes some fixed
plane, e.g., a fluorescent screen on which the w a v e f o r m is recorded. T h e
u n c e r t a i n t y in the energy e is, assuming t h a t the electrons were mono-
kinetic at the start,
BE=hymn,
n being the n u m b e r of q u a n t a which it m a y have absorbed from or lost t o
the field. B u t as $n is of the order ~1~2, we obtain
v3t,~(n )-l/2.
I n words, the oscillogram of the signal is t r a c e d with a spot whose m e a n
w i d t h in the direction of time is a fraction 1/V/~ of a cycle. Thus one
c a n determine the phase with the same order of relative error as the
amplitude.