On Entropy Reduction In A Thermodynamic System By Interference By Intelligent Subjects

By Leo Szilard, 1929

Annotated on Fermat's Library with 8 comments. Text and annotations as Markdown Original paper

Introduce Szilard
define perpetual machine of second type
describe Maxwell's demon
introduce Smoluchowski ideas
describe isothermal expansion of an ideal gas
how is this measurement having an impact on entropy?
typo!
rewrite these formulas
NASA TECHNICAL TRANSLATION
NASA TT F-16723
ON ENTROPY REDUCTION IN A THERMODYNAMIC SYSTEM BY
INTERFERENCE BY INTELLIGENT SUBJECTS
L. Szilard
(
N
ASA-TT-F-16723)
ON
7N
Tr10DY
F
EDUCTION IN A
N7F-17981
THEFMODYNA
I
II C SYST
E
M BY IN
T
--
D
F7
v
FPC7 BY
INTFLLIG° N'T SnBJFCTS (Scientific Translation
Service)
23 p HC $
-i
.
SO
CSCL 20M
Unclas
63/77 1u178
Translation of " ber die Entropieverminderung
in einem thermodynamischen System bei Ein¢riffen
intelligenter Wesen" Zeitschrift fur Physik,
Vol. 53, 1929, pp. 840- 856.
f^,-
REC;
I I q
i
4
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
WASHINGTON, D. C. 205
!ti
FEBRUARY 1976
r^
STANDARD TITLE PAGE
1. Report Ne.
FO
aeemma
r
r
l
A%aalen 04.
NASA TT F-16723
7.
RaNrlon
^
'o Careloli No.
S.
Report Dare
February
1976
4.
Title and Swbotle
ON ENTROPY REDUCTION IN
A THERMODYNAMIC SYSTEM
BY
INTERFERENCE BY INTELLIGENT SUBJECTS
6.
Performing orgentaarlon Code
).
Auther(s)
I.
Parlermrn
f
OraenieaNen Rar•rl
Na.
L.
Szilard
Io.
wort
unit
No.
11.
Contract er G'Ont Ne,
9.
Performinp'0rponleatian Name and Address
NASw— 2791
SCITRAN
Box 5456
11.
T rr•
of Rep orr end Period Ce.er.d
Santa Barbara
A 93108
Translation
12.
svonrorino A
g
ency Nome and Address
National Aeronautics and
Spaee Administration
Washington, D.C.
20546
14.
tpnserinili Agency Cede
IS.
Su
^^
I•men
^
ary Norse
;ranslation of "Uber die Entropieverminderung
in
einem thermodynamischeni
System bei Eingriffen intelligenter Wesen", Zeitschrift
Vol.
53,
1121,
pp.
840 - 856.
fur
Physik,
16.
A6atrecr
A perpetual motion machine involving several
types
of molecules is
I
discussed
in which intelligent being influence
the course of events.
The
recall
property of the individual leads to
a
continuous increase in
entropy.
i
r
ON ENTROPY REDUCTION IN A THERMODYNAMIC SYSTEM BY
INTERFERENCE BY INTELLIGENT SUBJECTS
L. Szilard,
in Berlin
/840*
We investigate what conditions cause one to construe an
apparent perpetual motion of the second type if one allows an
intelligent subject to influence a thermodynamic system. By hav-
ing subjects make such measurements, they produce a behavior of
the system which is clearly different from an isolated mechanical
system.
We show that a type of recall ability can offer
an occasion to a continuous reduction of entropy. This recall is
distinguished in a system in which measurements occur, and would
lead to a contradiction of the second law if the measurements
themselves did not necessarily proceed under entropy production.
At first this entropy production is calculated universally from
the requirement that it represents a complete compensation in the
sense of the second law (Equation 1). Using an apparatus which
is able to implement continuous measurements (under continuous
entropy production), the corresponding amount of entropy is
calculated. We found that it is ,just as large as is necessary
for complete compensation: the actual entropy production on
measurement, therefore, does not have to be larger than that deter-
mined by Equation (1).
* Numbers in margin indicate pagination in original foreign text.
ORIGINAL P GE
ffi
1
OF POOR
,
Q AIX1Y
There is already an historical objection to the general
validity of the second law of thermodynamics which actually
makes a quite disturbing impression. This is the objection
of the Maxwellian Demon which still comes up today in different
forms (and perhaps with some justification, inasmuch as quantita-
tive relationships seem to be hidden behind the precise questions
which have not yet been clarified). The objection is, in its
original formulation one which operates with a demon which
catches the fast molecules and allows the slow ones to pass. One
can refute this by the fact that we humans cannot, in principle,
discover the value of the thermal, varying parameters; but it
cannot be denied that we could measure the value of such oscillat-
ing parameters very well, and could then surely obtain data on the
costs of the warmth by arranging our concepts appropriately to
the result of the measurements. Of course, it remains to bp
decided whether we do not make an error if we do not consider
the interfering subject himself, in the system, along with his
life processes.
Except for this point, it is known today that in an isolated /841
system, in spite of the occurrence of oscillation, no perpetual
motion, second type, can be effective, i.e., no machine using
i
automatic heat of lowest temperature of continuous, finite
capability, can be effective. This would be a machine whose
effectiveness would have the result, even over long periods of time,
that a weight would be raised at the cost of heat in a heat
reservoir.
i
Ex
p
ressed in another way, one can also say that, if we want
to use the occurrences of variations in order -co get data on
the costs of the heat, we are in the same position as in a game
of chance at which we can win certain amounts from time to time,
but at which the expected value (mathematical hope) of winnings is
zero or negative. The same is true for a system
which is
not
2
isolated, but at which the external interferences are highly
periodic, as by machines moving periodically. We consider this
statement as proven [1] and will only consider the difficulties
which occur when subjects interfere in the system, and to try
to recognize the quantitative relationships occurring in this.
i
Smoluchowski writes [2]: "As far as our present knowledge
^p
reaches, there is no automatic, continuously effective perpetual
{
motion in spite of molecular vibrations, but such an apparatus could
function regularly if it were operated by intelligent subjects
in a suitable manner. ...
t
"Thus, a perpetual motion is possible if one considers the
experimenting person, according to the usual methods of physics,
as a kind of "Deus ex machina" which is continuously and exactly
informed of the instantaneous condition of nature. This Deus
ex machina" can set in motion or interrupt the macroscopic
natural processes at any moment. Thus, he does not at all need
to possess, like the Maxwellian demon does, the ability to
catch individual molecules,
but
would be completely different in
the above points from real persons. The production of any kind
of physical effect, by operation of the sensor or motor nervous
system, is always connected with depreciation of energy, except
that its entire existence is connected to a continuous dissipa-
tion of same.
"It is doubtful that real persons could produce continuous
/842
or at least uniform, work at the cost of heat at the lowest
temperature, considering these circumstances, even though our
lack of knowledge of life processes precludes a definitive
answer. :however, the questions ,just touched u?on lead beyond
actual physics ...."
3
k
r
I
t
It seems to us that the lack of knowledge of life processes
should not be disturbing'for recognizing that with which our work
is concerned. Because we may be certain that one can replace the
subjects (inasmuch as concerns their interference in a thermody-
namic system) by any device whose "life processes" can be followed.
Thus, one could determine whether a compensation actually occurs
for the entropy reduction which is produced by the interference
of the device in the system.
At first, we want to try to recognize what causes the
entropy reduction on interference of intelligent subjects in a
thermodynamic system. We shall see that it depends on a coupling
of a special type between various parameters of the system. We
shall observe an especially simple type of this threatening
coupling. We will shortly speak of a "measurement" whenever we
succeed in connecting the value of one parameter "y" (e.g., the
distance coordinate of the pointer of a measuring instrument) at
f
one instant, with the momentary value of a varying parameter "x"
of the system. Thus, one can draw conclusions on the value of
"
x
"
at the moment of measurement from the value of "y".
Then, "x" and "y" are separated again after the measurement so
that "x" can change, whereas "y" retains its value a little
while. Such measurements are no harmless interference; a system
in which such measurements occur indicates a type of remembrance
ability, in the sense that one can recognize what kind of value
a different parameter, Y', had at an earlier time by the condi-
tion parameter "y". And we'shall see that such ability of
recollection would damage the second law if the process of
measurement would take place without compensation. We will then
see that the second law is not so greatly threatened by this
entropy reduction as one migh
t
; originally think. We recognize
that the entropy reduction caused by the interference would be
completely compensated if the implementation of such measurements
were universally dependent on an entropy production of k
•
log 2,
4
for example. It would then be possible to prescribe a somewhat
general entropy law which is based universally on all measurements.
In conclusion, a very simple apparatus (not arbitrarily chosen) is
under consideration. This device
is in
the position to undertake
continuous "measurements", and we may easily follow its "life
processes". By direct calculation, one actually finds a con-
tinuous entropy production in the amount which is required by the
mentioned, general entropy law which is derived from the validity
of the second law.
The first example, which we consider typical, is the following,:
a standing hollow cylinder which is closed top and bottom, can be
divided into two unequal halves of volumns V
1
and V
2
by pushing in
a solid disc from the side. This disc forms a 'piston" which can
be pushed up and down inside the cylinder. An infinite heat
reservoir of some temperature T insures that a gas present in
the cylinder performs an isothermic expansion on movement of the
piston. Now, this gas should consist of one single molecule which
roams around in the whole cylinder according to its thermal
motion, as long as the piston is not shoved into the cylinder.
Let's now imagine a subject who shoved the piston into the
cylinder at some point in time, and simultaneously determines
(somenow) wnether the molecule is in the upper or lower half of the
cylinder, in volume V
1
or V
2
. If he finds that the former is the
case, then he pushes the piston slowly downward, until the base
of the cylinder is reached. Durin
g
; this slow pushing of the pis-
ton, the molecule naturally remains above the piston at all times,
but does not remain in the top half of the cylinder, but rather
strikes the top of the moving piston many times. In this manner,
the molecule does a certain amount of work on the piston. This
is the work which corresponds to an isothermic expansion of an
ideal gas consiating of one molecule, performed from volume V. on
volume V
1
+
V
2
. After a time, when the piston
has reached the
5
^
I
r
i
-.3
base of the container, the entire volume V
1
♦
V
2
is again avail-
/844
able to the molecule, and the piston is then pulled out. The pro-
cess can be repeated as often as desired, where the subject pushes
the piston up or down depending on whether the molecule is in the
upper or lower half. That is, the subject is so coupled to a
weight which is to be raised by a mechanical transfer of force,
that the weight is pushed up by the piston. In this manner, the
potential energy of the weight certainly increases continuously.
(The transfer of force to the weight is so represented, that the
force transferred from the weight to the piston, at any position
of the piston balances the average pressure of the gas). It is
clear that Lnntinuous work can be obtained at the cost of heat,
inasmuch as the life processes of the interfering subject are
not considered
1;. the calculation.
In order to recognize what the subject actually performs for
the system here, one should imagine that the motion of the piston
is done mechanically and that the whole activity of the subject
is exhausted in determining the height of the molecule and in
adjusting a lever left or right (which guides the motion of the
piston) according to which motion of the piston, upwards or
downwards, is required by the position of the molecule. That is,
the interference of the subject consists in the coupling of two
position, coordinates, namely an "x" coordinate which determined
the altitude of the molecule, with another coordinate "y", which
defines the position of the lever and which is thus decisive in
giving the piston an up or down motion.
One should imagine the
mass of the piston to be large and its speed to be sufficiently
high so that the.thermal agitation, which the piston has at the
operating temperature, can be neglected.
We shall differentiate two time periods in the processes of
the typical example under consideration here, namely:
6
ORIGINAL PAGE IS
OF POOR QUALITY
L
1.
#
.,
.The time period of measurement, at which the piston is
Just shoved into the middle of the cylinder and the molecule is
shut into either the upper or lower half. Thus, we limit the
x-coordinate to the interval
x
>0
or x <0
by choosing an ap
p
ro-
priate coordinate origin.
2.
The time period of utilization of the measurement, during
which the up or down motion of the piston is accomplished, the
"time period of entropy reduction". During this period, the x-
coordinate of the molecule is by no means limited to the
original x >0 or x <0.
If the molecule, at the time of
/845
measurement is in the upper half of the cylinder so that x> 0,
then it must strike the piston which is now moving down if it is
to contribute to this work. That
is,
the x-coordinate of the
molecule must enter the interval x <0. However, the lever will
retain its setting corresponding to the downward motion during
the entire time period. If this right position of the lever is
determined by y = 1 (and the corresponding left-position y = -1),
we see that the measurement of the position x > 0 is paired with
y = 1 during the time interval. We also note that later, y = 1
remains, even after x has passed into the other interval x < 0.
One sees that on utilization of the measurement, the coupling of
the two parameters, x and y, is lost.
We will speak quite generally about this, that a parameter
"
y" measures the value of a parameter "x", which is varying by a
probabili'y law, wherever the value of "y" changes to the value of
the parameter "x" at a definite point
in
time. Taking this
measurement is the underlying principle of entropy reduction by
interference by intelligent subjects.
It is now suggested to assume that taking the measurement is
principally connected with a
q
uite certain average
entropy pro-
duction. By this, agreement with the
second law is again
()IMLNAL PAGE I,9
Of' POOR QUALITY
7
i
^
I
restored. The amount of entropy occurring on measurement may
always be larger, but not smaller. More exactly, two q4antities
of entropy must be differentiated here, the first, S
l
, is pro-
duced if "y" has the value of 1 on measurement, and the other, S2,
when y - 1. We can not expect to learn anything, generally,
about S
1
or S
2
, individually. But we shall indeed see from the
assumption that the entropy produced by this "measurement" must be
compensated by the utilization of the resultant entropy reduction
of the system (in the sense of the second law), that the quite
general :elation
r
+e
s l
(l)
follows.
From this formula, one sees that we can make one of the
quantities, S
1
, as small as desired, but that the other quantity,
S
2
, becomes correspondingly large. It is also striking that
it is not at all dependent on the size of the interval observed.
/846
Of course, one can easily see that it could not possibly be other-
wise.
We can say the converse: as soon as the entropies S
1
and S2
(occurring by the measurement) satisfy inequality (1), we may rest
assured that the attainable entropy reduction is already completely
compensated by the measurement.
Before we go into the proof of inequality (1), we can see
from the observed mechanical example how this all fits together.
For the produced entropys S
1
and S
2
, we write the special formula:
S
1
S2 t k log 2.
8
I
This statement obeys inequality (1), and the average value of
measurement-produced entropy is:
S - k log 2
(in this special case, it is naturally independent of the ratio
of frequency w
l
, w
2
of both occurrences). In the observed example,
one achieves entropy reduction* by isothermic expansion:
—
1
1
—
—
l:
log
y,
k
log -
-°- .
V, +
V^
V1 •
T
t ,
regardless of whether the molecule is encountered in volume
V
1
or V
2
when the piston is pushed in. (Entropy reduction
equals the quotient of the amount of heat withdrawn from the heat
reservoir by isothermal expansion, and the temperature of the
respective heat reservoir.) Since in the present case, the fre-
quency w
l
, w
2
of both occurrences is in the ratio as the volume
V
1
, V
2
, then the average amount of produced entropy (negative
number) is:
_
W
,
+W ,.(+r,)
y!-
klog
y'-
-+
V, '
s
4
.
109-
5
—.
One sees that actually
V L
k log
V
'
,+
'
k log
11
+ k log 2 a 0
V
,+
V
,
V,+Is
V
.+
V
,
V,+V,
and thus S + s ' 0.
Thus, in the present special case, there would be a full
/847
compensation for the attained e-..ropy reduction on utilization of
the measurement.
Let us now observe other special examples and try to derive
formula (1) and define the above relationships with the aid of a
general observation. Let us now think of the entire system as
• The entropy produced is designated e
l
, a2.
9
r
i
Then:
w1 + w
2
= 1
/848
The distribution
of the y-
parameter may
be in
any ratio of the
two values 1 and -1, but is independent of the x-values in any
!
case. We now imagine an
interference
by a subject, by which
''y" receives the value 1
in
all molecules, "x" falls in the promi-
nent interval at the time concerned, and corresponds to the
j
value of -1
in
the opposite case. If, as a consequence of thermal
I
variation, the parameter "x" leaves the prominent interval for any
molecule, or, as we can express it here, if the molecule suffers
a monomolecular chemical reaction with respect to "x", then
the parameter "y" retains its value of 1 at first so that the
molecule 1:- able to "remember" during the entire later process
that 'x" .,rigirally fell in the prominent interval. The mono-
molecular chemical reaction causes the molecule, wr,icn was per-
meable to piston A, to become impermeable. 4e shall see shortly
1
what role this "remembrance" can play. We now push in the pistons
as described above so that we separate the two k17
,
ds of molecules
from eacn other without doing any work. Two containers are then
present, of these, one gets one f^action, the other gets the
other fraction. Now, each fraction alone takes up the sam vr)lur-e as
the mixture did before. in one of these containers, an eau'.11hrium
distribution with respect to the two "modifications in 'x'", is not
present. The quantity ratio of the two modifications ;fractions)
is
Still
w
1
:w
2'
If we let this equilibrium distribution occur
j
by itself at constant volume and temperature in each vessel, then
the entropy of the system definitely increases during the process.
Because the total heat production is 0, since the ratio of both
"modifications in x", w
1
:W
2
ioes not change. 1 we lead the
E
adjustment of the equilibrium distribution ir. both containers by
{
reversible paths, then the entropy of the surroundings would
decrease by the
sane amount. Thus, the
entropy increases by a
negative amount., and the value of entropy increase
per molecule
is:
.,!,
i',1G1; IS
11
-. CUOR QUALI'Ty
i = k (w
I
log 0
1
+ '0, log w.}
(2)
/849
(The entropy constants which we have assigned to the two "modifi-
cations in 'x " do not explicitly occur here, because the process
which belongs to one or the other modification leaves the total
number of molecules unchanged.)
Naturally, we could return the two ga3es to their original
volumes without doing any more work by simply pushing back the
piston, since the volume determined by the piston BB' contains such
molecules whose x-coordinates fall outside of the prominent
interval and for whom piston A is no longer permeable. So we see
that the entropy reduction found (2) does not mean a contradiction
to the second law. So long as we do not use the fact that the
molecules in vessel BB' of coordinate "y" can "remember" that the
x-coordinate of the molecules in this container originally lay in
the prominent interval, then complete compensation is actually
present for the calculated entropy reduction; the partial pressures
in both vessels are now smaller than in the original mixture.
But we can use the fact that all moleraies in vessel BB' have a
y-cco:dinate of
1
and those in the other vessel hav
,
_^ a y-coordi-
nate of -1, in order to return all molecules to the original vol-
ume. For this, we merely replace semipermeable wall A by
wall A*, which is semipermeable with respect to "y" and not to
"x", s- that it is permeable fcr molecules with y-coordinate 1,
and impermeable for the others. Conversely, we replace B' by
a piston B'* which is impermeable for molecules with y = -1 and
permeable for the others. Then both vessels can be pushed
together again without expending work. The distribution of
the y-coordinates to 1 and -1 has become statistically independent
of the x-values and, moreover, we can produce the original dis-
tribution to 1 and -1: thus we pass through a complete circular
r
r
12
I
^
-I
process. The only change that we have to register is the
attaine6 entropy reduction of the amount (2),
i =
k
(W
I
log w,
+
W ,
log w,).
Thus, if we do not want to admit that the second law is
violated, then we must conclude that the interference which
/850
causes the coupling between "x" and
tt
y", the measurement of
'IX" by "y", is insolubly connected to the entrop. production.
If a definite path for production of this coupling is given, and
if the necessary amounts of entropy produced are designated S, an
1
S
2
, then the average produced entropy is given by
t0, `, 4- M,
so
= S.
Here S
1
is the average entropy increase which occurs when "y" has
the value 1,
a
,
.-^
S
2
is that which occurs when "y" has the value
-1. In order for the second law to be consistent, this amount
of entropy must be larger than the entropy reduction s which is
attainable by the evaluation of the measurement according to (2).
Thus
,5+xzo,
(3)
W
1
S, +
w, S, + k
(tu
t
log
tr,
+ to, log to,) > 0.
must be valid. This inequality must be true for any values of
w
l
and w
2
*. Naturally the peripheral condition w
l
+ w
2
= 1 must
be retained. In particular, we are asking for those values of w
and w
2
for which the expression has a minimum for given S-values.
For this pair of values, w and w
2
, inequality (3) must still
hold. Under the mentioned peripheral conditions, the minimum
occurs when
`k + log m, = .k + log to,
(4)
The increase in entropy can only be dependent on the type of
measurement and its result, and not on how many systems
of one type or another
,
a
r
e
f
.
,
resenc .
13
I
I
if
i
'
II
is true. But then
e
+ e R
5
1.
This can be seen very easily if one introduces the equation
I+ log a, _ + log W
S
= A
i
Then
W, = e'.#
t
.
W. = t
;
.e
e.
(5)
If we substitute these values into inequality (3), there results
f
i
xe
i
(e
F
-*-e
k)z0.
Thus
0
holds.
But if we substitute the values
for w and w
2
from (5) in the equation w + w
2
= 1, we obtain:
_t,
-N
e
k
+ 1
6
=
e
—
J.
And because ^ ?O ,
/851
c
t
+
(6)
is also true. This formula must be generally true if the thermo-
dynamics are not to be violated.
As long as we let the interference be done by intelligent
subjects, a direct check must be excluded. However, we can try
to give simple, specific devices which cause such a coupling,
and observe whether entropy is actually produced and in what
quantity. Not until we have recognized that it depends on a
definite characteristic kind of coupling, a "measurement" will
will be needed to construct some complicated model which signifi-
cantly imitates the interference of the subject. Thus, for the
present, we can be satisfied with the creation of this special
coupling equipped with memory.
14
ORIGINAL PAGE IS
OF POOR QUALITY
In this example, the position coordinates of a moving pointer
i
are "measured" by the energy content of a body K. At first, the
pointer will purely mechanically effect that the body K be
,neasured by its energy content, according to the position of the
I
pointer with one of two distances, A or B. The pointer works by
IIII'
heat so that the body is connected to A as long as the coordinates
determined by the pointer fall in a certain prominent interval
i
"a" of arbitrary length. The opposite holds with B in interval
"b". Both distances are in thermal contact with a heat reservoir
of temperature^T
0
up to the moment of "measurement" . At this mo-
ment, distance A is cooled to temperature T
A
by a periodically
functioning mechanical device. This is done by successive contact
with heat reservoirs of intermediate temperatures in contact with
a heat reservoir of temperature T
A
. Simultaneously, B is
warmed to temperature T
B
in the same manner. Then, the distances
are again isolated from the heat reservoirs.
/852
We assume that the position of the pointer changes so slowly
that all operations sketched here proceed at one and the same
pointer position. If the position coordinate of the pointer
in the prominent interval fell. then the body would be connected
to interval A during the mentioned operation, and would be conse-
quently cooled to temperature T
A
. Conversely, the body is now
warmed to temperature T
B
. Its energy content is thus, according
tc the position of the pointer at the moment of "measurement",
small corresponding to temperature T
A
or large corresponding to
temperature 'I
B
. It retains its value even when the pointer
leaves the prominent interval in the course of time. After a
time when the pointer carries out its further oscillations, no
more conclusions can be made on the instantaneous position of
the pointer from the energy content of body K, rather, a sure
conclusi
r
in on the position of the pointer at the time of measure-
ment. Then the measurement is completed.
15
r
r
The mentioned periodically functioning mechanical device
which makes the interference for us should bring the now thermi-
cally isolated segments A and B into direct contact with heat
Z
reservoir ':
Q
after completion of the measurement. This has the
purpose of bringing body K (now in contact with one of the two
Intervals) back to its original condition as it was before the
i
"measurement". The direct connection of the intervals A and B,
E
and of the cooled or heated body K with the reservoir To,
i
necessarily causes an increase in entropy. This cannot be avoided,
because it would be useless to try to warm the interval A by
successive contact with reservoirs of intermediate temperatures to
temperature T
o
, or likewise to try to cool B to T
o
. After the
completed measurement we do not know which of the two intervals is
now in contact with body K, and likewise, we do not know whether
body K was last in contact with T
A
or T
B
. We also do not know
whether we should use intermediate temperatures between T
A
and
T
o
or between T
o
and TB.
The average amounts of entropy produced in this manner, S
1
and
S
z
, can be determined if the heat capacity as a function of
the temperature u (T) for body K is known, since the entropy is /853
to be calculated from the heat capacity. We naturally ab-stract
from the heat capacity of the intervals, A and B. If the position
coordinates of the pointer at the time of "measurement" were in
the prominent interval, and if the body was in contact-with
interval A, then the entropy on successive cooling of the heat
reservoirs is:
f^
Thus the entropy of thiq heat reservoir on direct contact with
heat reservoir T
o
is:
-r.-
16
OF
I
NAL' PAGE
P
OOR QUALM
Thus, the sum of the entropy is increased by:
To
S
_;t(T'A)-n(f^)
+ f
1 du dr
(7)
'^
T^
T dC
rA
Analogous to this, the entropy would increase by
To
S
(Te)
K (TO) +
fT
1 d
i
( 8 )
n
=
1'
dT d
T
when the body was in contact with interval B at the moment of
"measurement",.
Now let's evaluate these expressions for the special, simple
case when the body possesses only two energy states, a lower
and upper. If such a
,
.Vody with a heat reservoir of some tempera-
ture T is in thermal;^qntact, then the probability of its being
in the upper or lower
Mate is given by
P(
T
) = - . _t
l+.Qe-ir
Y
-
tr
Or
y(T)
^ —e
(9)
---^ .
i
t
ge
k
Here, "L" is the energy difference of the two states and ''g"
is the statistical weight. We can equate the energy of the lower
state to zero, without this being a limitation. There results*
`A =
f
]
VA) 4.
log
'
I
(1'A)1'(J'0) +
k
lo^
J'(l A)
9 (
7'
0) J'
VA)
J'
(7'^1
or
I
(10)
N
)r
= P
(TO
k
log
- - + k lob;
9 (
1
'e)1' Vo)
7
V.)
* See the appendix at the conclusion of the work.
ORIGWAL
PAGD
IS
17
•
OF POOR QUALIT14
/854
Here q and p are the functions of T given by formula (9) which
are to be taken here for the arguments T
o
, T
A
, and TB.
If we now want to determine that a certain conclusion
regarding the position coordinate of the pointer is possible from
the energy content of the body K, we must show that the body got
into the lower state whenever it comes into contact with T
A
, or
in the higher state when in contact with T
B
. This means that:
1)
(
LE
A) _ ^^ Y (
T
A) _ ^^^
P
(^ u)
= ll ^! (%'») _= 1
This cannot be proven, but cart be approximated if we let T A
go to
absolute zero, and the statistical weight "g" go to infinity. [In
going to the limit, T
o
is so changed that p(T
0
) and q(T
0
) remain
unchanged.] From the formula (10) results:
."A
log Y (
r
= — k log P
(
T
,)
;
S
D
= —
k
•)
( 11 )
S11 we find that:
,
A
till
C
t
+ C
r
= 1.
The limit can be realized by our previous conclusions; the use
of semipermeable walls according to Figure 1 allows a complete
utilization of the measurement: inequality (1) certainly cannot
be intensified.
As we have seen from this example, a simple device can
accomplish what interference by an intelligent subject would pro-
vide, with respect to that which is important. In this example,
we have pursued the "life processes" of a device and we can see
that the requis'_te amount of entropy is produced, gust exactly
as provided by thermodynamics.
Now if we form the expression a-S;
+,
_
18
f^
4
lY
C
i
APPENDIX
In the case under consideration, when the frequency of both
conditions depends on the temperature according to the formulas:
1(
T
) =
1
Y
q(T) — 4c tr
^+A •
tT
^.1 7a
IT
and thus the average energy of the body is given by
Y
tr
i#(T) = r9(T)
+!.4e
-'r
1+Dd- RT
then.the identity
_ r
1 du _ r!
fu(r
+klog
(1+0 er))
T
d T ;Fr I r
`
is valid.
With this, we can write the expression
F
'
(T
A)
1
d u
9^ _ -
^^--- -) J
T
dT dT
T,j
thus.
—
u(TA)'
-
uo)
+jn(7)+klo 11
}r!
T•
A
-
to
Y
6
t.0e
^T
A
and when we pass to the limits, we obtain
1
1t
ge
tT
$
S
A
= "(TA) \
(
r
To — T, l
t
k lob
-r.
t T^
1-}-ye
If we now substitute in the last expression according to (9),
Y
1
r
.
qt
-
7.
1
P(T)
for T
A
and T
o
, we obtain
(
1
r)
\ T. TAI
r (T.)
ORIGINAL' PAGE 1B
OF POOR QUALM
19
and if then, according to (12), we substitute
w
(T
A
)
-
"
9
(TA)
there results
A = 4('
A
)
(N
w
)+klog
P(TA)
$
r
—
T
I
T
A
j (T^
If we substitute according to (9)
n _. _A
.
Ing '/(T)
/856
for T
A
and T
o
, we obtain
N
.,
'/ 0"') k In
d
;
k
Ing
POO
We obtain the corresponding formula for S
A
by exchanging
Index A for B. Then we get:
/ )klo /r(7n)'l('lIr)
kIn,/,rl^r)
The former is the formula which is given in the text for SA.
We can also express formula for S somewhat differently, by
4
writing
`I
multiplying out and adding the first and last member. It then
becomes:
Sit
=Jr
(T„)
1 log
J'
(TO)
9(Tn)
r klo^ 7(TO)
)
This is the formula given in the text for SB.
20
REFERENCES
1.
Szilard, L.
ZS. f.
Phys., Vol. 32, 1925, p. 753•
2.
Vortrage fiber die kinetische Theorie der Materie u.
Elektrizitat (Lectures on Kinetic Theory of Matter and
Electricity), Leipzig,
1914, p. 89.
21

Discussion

Tap the paper where you want to comment
Introduce Szilard define perpetual machine of second type describe Maxwell's demon introduce Smoluchowski ideas describe isothermal expansion of an ideal gas how is this measurement having an impact on entropy? typo! rewrite these formulas