THERMOMETRIC LAG

By D. R. Harper 3d

CONTENTS
Page
Introduction 659
I. Lag op a common mercuriaIv thermometer 660
Fundamental considerations 660
Physical meaning of " X," the time lag of a thermometer 662
Falling meniscus 662
Principal equations 663
Methods of determining X —illustrative typical values of same 667
Variation of X with stirring 672
Tn water 672
In a kerosene oil 676
In air 678
II. Lag of a Beckmann thermometer 682
Additional assumptions 682
Justification of assumptions made 688
Special cases 692
III. Lag op ei<Ectricai. thermometers 694
Galvanometer lag 695
Resistance thermometers 698
Jaeger-Steinwehr method of computing lag of above 699
Thermoelectric thermometers 702
IV. Thermometric lag in calorimetry 702
V. Lag corrections in applied thermometry:
R6sume of practical instructions 711
VI. Summary 713

INTRODUCTION
When a thermometer is immersed in a warmer or in a cooler
medium for the pmpose of ascertaining the temperature of the
latter, it does not immediately indicate this temperature but
exhibits a time lag in reaching it. A certain time must elapse
after immersion before the reading is correct to within o?i, still

longer before correct within o?oi, etc., when the temperature of

659
66o Bulletin of the Bureau of Standards [Voi. 8

the medium remains constant. If the latter be varying, the ther-

mometer follows the variation in a definite way, maintaining a
difference which may be large or small, according to the rate of
the variation and the form of the thermometer.
It therefore becomes necessary under some conditions to apply
a correction to a thermometer reading ^ in order to obtain the
simultaneous value of the temperature of the medium in which
it is immersed. Thiesen ^ in his elaborate treatise on thermometry
devotes a section to the consideration of lag, and Guillaume^
also, under the title of " Defaut de Sensibilite." Neither of these
chapters conveys an adequate impression of the variation of the
lag of a particular thermometer under different conditions of
immersion, namely, nature of medium and rate at which it is
stirred. Also electrical thermometers, thermoelectric and resist-
ance, v/ere not in use as precision instruments when these papers
were published and are, of course, not considered. It has, there-
fore, seemed advisable to prepare a paper supplying some of these
omissions, and in so doing to incorporate as much of the general
theory as seems necessary for full comprehension of the subject.

I. THE LAG OF A MERCURIAL THERMOMETER

FUNDAMENTAL CONSIDERATIONS
The transfer of heat between a thermometer bulb and the
medium in which it is immersed may be expressed in the form
usually referred to as '*
Newton's law of cooling," namely, the rate
of heat transfer directly proportional to the difference of tem-
perature between the two. While direct experimental test of this
fundamental assumption is difficult, it is very easy to test many of
the equations deduced from it, some of which are so closely related
to the parent equa,tion that the test possesses all the force of a
direct one. For the ordinary or "chemical" form of mercurial
thermometer, the agreement between theory and experiment amply
justifies the assumption, which analytically expressed is

1 Throughout this paper the assumption is tacitly made, except where exphcitly stated to the contrary,

that all instrumental errors have been corrected for, i. e., cahbration of bore, zero error, etc.
2 Thiesen: Metronomische Beitrage, No. 3, p. 13; 1881.

3 Guiilaume: Thermometrie de Precision, p. 184; 18S9.

 1

Harper] Tkermometric Lag 66

where i*= temperature of bath at instant t.

6 = temperature of thermometer at same instant.

X = constant with respect to u, 6, or t.

The temperature of the bath, u, is assumed to be uniform through-

out the neighborhood of the thermometer. No other case would
be simple enough for convenient mathematical treatment. The
temperature of the thermometer, 6, is the integrated or average
temperature on which depends the average densit)^ of the ther.
mometric liquid and the volume of the envelope. It defines the
position of the meniscus in the capillary, i. e., the reading of the
thermometer. \ requires no ftnther definition at present than the
placing of the limitation that it shall be independent of 0, u, and t.
It will be found to be not independent of form of thermometer,
medium surrounding latter, conditions of stirring, etc., a variation
of a kind to be carefully distinguished from that denied to it in
integrating equation (i).
Integration of (i) leads to relations between quantities all of
which may be readily foimd by direct experiment, serving as a
criterion of the validity of the assumption. For the ordinary
form thermometer a numerical value for X, constant for a par-
of
ticular thermometer in any one medium stirred at a certain
rate, may be found and used in applying corrections to any
readings made under the same conditions, whenever a lag correc-
tion is necessary. For some forms of thermometer, e. g., those
having a layer of air separating part of the btdb from the bath,
the simple theory employing equation (i) fails.^ If we endeavor
to find numerical values for X for such a thermometer by the
equations deduced from (i), we are led to the conclusion that no
single value is obtainable; the various values that might be
computed from a few points on a curve of observations will differ
according to the temperatmre or time corresponding to the points
chosen. This is inconsistent with the limitation placed on X in

* Thiesen, loc. cit.

662 Bulletin of the Bureau of Standards Woi. s

deducing the equations, nullifying the whole theory for appli-

cation to such an instrument. Modification of equation (i) by
the introduction of additional assumptions leads to a treatment
that is satisfactory for some such forms of thermometer; for
instance, a calorimetric Beckmann, which has a considerable por-
tion of its mercury in a large inclosed capillary, between the bulb
and the fine capillary. (See Sec. II.)

PHYSICAL MEANING OF X
Physical interpretation of the quantity X is highly desirable.
This obtained from the equations (see pp. 664-665), but it may
is

be well to anticipate slightly and give two such interpretations

at this point. From the fundamental equation, which serves as
the definition of X,

it is seen by inspection that X has the dimensions of time.

Later
equations give the following interpretations as a definite nimiber
of seconds:
(i) If a thermometer has been immersed for a long time in a

bath whose temperature is rising at uniform rate, X is the number of

seconds between the time when the bath attains any given temperature
and the time when the thermometer indicates this temperature. In
other words, it is the number of seconds the thermometer "lags"
behind such a temperature.
(2) If a thermometer be plunged into a bath maintained at a
constant temperature (the thermometer being initially at a dif-
ferent temperatiure) X is the number of seconds in which the difference
,

between the thermometer reading and the bath temperature is reduced

to c^ times its initial value}

FALLING MENISCUS
Considerable attention has been paid to the lag of a thermom-
eter when its temperature is and the results pubUshed
decreasing,^
seem liable to misinterpretation. Without entering the very dif-

6 •~^=r^=o-4. approximatdy. fXhiesen, loc. cit.; Guillamae, loc. cit.

Harper] Thermometric Lag 663

of analysis of the surface tension and capillary forces

fictilt field

governing the motion of the meniscus, it is sufficient to point out

that with falling temperattires one may observe sticking of the
meniscus and even separation of the mercury column, indicating
that there is no certainty that the forces tending to return mer-

cury to the bulb accomplish the retiun at a rate proportional to

the rate at which the transfer of heat from the bulb takes place.
The usual equations of lag, being primarily equations of heat
transfer, must not, therefore, be expected to cover all cases of
falling meniscus. If a thermometer is plunged into a bath much
cooler than itself, the drop of the meniscus is apt to be so erratic
that a computation of the lag by the usual methods results in
values widely variable under exactly the same conditions and
oftentimes quite different from the lag measured with a rising
meniscus. The pseudo X so obtained may or may
not be related
to the temperattire lag of the bulb, and as the conditions are impos-
sible of specification the values thus determined fail of interpreta-
tion.
Because of the sticking of the meniscus and possible separation
of the mercury column, the readings of a mercury thermometer
with falHng temperatures are less reliable than with rising ones,
and in precision thermometry a falling temperatiire should be
avoided whenever possible. When its use is unavoidable the
difficulties are overcome to a considerable extent, in the case of a
small rate of fall, by subjecting the thermometer to a series of
rapid jars, as by an electric buzzer. There seems to be every
reason for believing that under such conditions the position of the
meniscus depends on the temperattire of the bulb in the same
relation as when the meniscus is rising, and the lag corrections to
be applied may be computed from the value of \ determined with
rising meniscus, for the medium and rate of stirring employed.

PRINCIPAL EQUATIONS

The fundamental equation

Yr&-^ (I)

suffices to determine the function u if the function 6 be known,

 ^

664 Bulletin of the Bureau of Standards [Voi. 8

since -^ may at once be determined. The converse problem of

determining when u is known requires solution of the differential

equation.

Solving \-r +0=u under the conditions that the thermometer

reads Oq at the time ^=0, in a medium at temperature Uq; the

reading 6 at any time t when the medium is at temperature u, is

An integration by parts puts this equation in a form showing

directly the temperature difference, — u.
1, 1 rt?i^i 1

dt (3)

From this equation the solution for special cases is obtained upon
c)u
substituting the proper expression for -^. The cases of most
importance are (a) constant temperature u^Uq] (b) linear rise,
u^Uq+H, usually at small rate, r; (c) exponential change of
temperature according to formula u^A+Be"^* the constant a
being usually small.

(a) Whenu = UQ, -^ =

and equation (3) reduces to

{0-u,) = {e,-u,)e-h (4)

Equation (4) states that an initial temperature difference (^0 — Uq) ,

between a thermometer and a constant-temperature bath in

which it is immersed, decreases logarithmically with time, becom-
ing in X seconds, e~^ times the original difference. From the
approximate values, €^^ = .001 and e-^ = .oooi, it will be seen
'The frequent occurrence of the conditions (a) and (b) is evident. The importance of (c) lies in the
whenever one body is exchanging heat with another according to Newton's law, its temperattire is
fact that
expressed by the exponential equation given. When a is very small, nearly always true in calorimetry for
instance, the curvature of this exponential function may be so slight that only where high grecision is sought
need any accoimt be taken of its departure from a linear function (case b)
Harper] Thermometric Lag 665

that times of 7X and 9\ seconds elapse before a difference of 10°

is reduced to o?oi and o?ooi respectively. ,

(b) u = UQ+rt. —^The value of -^ is r, which is to be substi-

tuted in equation (3) . Reduction and collection of terms, gives

the relation

0-u=-r\+{0,-u, + rX)e-^* (5)

The interpretation of equation very simple, except the

(5) is

exponential term. Consideration of the numerical values which

enter will show that this term is usually negligible. From two to
six seconds, according to the size of bulb, is the value of X for an
ordinary thermometer in water stirred rather vigorously. Taking,

as a mean, four seconds for substitution in the term e ^\ the

value, for t = one minute, is e-^^ about 3 X iq-^.This multiplies a

term (6Q — UQ+rX) not very large, so the product is insignificant.
Accordingly, with a value of X of foiu: seconds or that order of

magnitude, the lag of a thermometer, immersed in a bath the

temperature of which is increasing at a constant rate r, is repre-
sented less than a minute after immersion, by the equation.

0-u=-r\ (6)

A constant difference of temperature, numerically r\ thus

exists between the thermometer and the bath. Both rise at the
rate r, so that in X seconds either one increases its temperature by
rX. Accordingly, X seconds after the bath attains a given temper-
ature, the same is indicated by the thermometer.
The (6) to numerical examples will illus-
application of equation
trate the magnitude of the corrections necessary for lag in applied
thermometry. For a rate of rise of o?03 per minute, perhaps the
maximum allowable w^hen readings to single thousandths of a de-
gree are taken, the lag correction to any reading if X = 4 seconds
is +0?002.

Care must be taken in the application of equations (5) and (6)

to any case where X is large, e. g., in still air, where the value of X
may be 50 to 100 times the value for the same thermometer in
^666 Bulletin of the Bureau of Standards Woi. s

stirred water. Computation in this case requires consideration of

_i
the term e ^, which becomes very small only after a rather
long time elapses.
(c)ExpoNENTiAi. Change of Temperature u^A+Be'"*, —
it may be well to
Before proceeding to the solution for this case,
give an example of its occiurence and the interpretation of the
symbols.
Let a calorimeter, whose temperature may be designated u, be
exchanging heat with surroundings at constant temperature A
according to Newton's law,

du , ..

the solution of u — A^Be'"^, the equation written

which is

above. B is the initial value oi u — A, i. e., the value of the tem-

perature difference between the calorimeter and its surrounding
at any arbitrary time at which we may choose to start applying
the equation. Instead of B, therefore, may be introduced
another
constant, Uq, which, if defined as the value of u for the time zero,
will be in harmony with all the foregoing equations and permit
of using them.

liu^^A+Be-"^] Uo = A+B, u-A = {Uo-A)e"'\

Substituting the value of ^, namely, —a{uo — A)e~'^ in equa-

tion (3) and collecting

Every term in the coefficient of the exponential is a constant,

so that some value of t can be found, after which the term will be
negligible to any required order. For the values of a,\,Aj and u^
commonly met with in practice, this time is comparatively short.
The " steady state " is then said to be estabUshed, and the behavior
of the thermometer is given by

f-u = ~^{u-A). (8)

Harper] Thermoifietric Lag 667

The difference (d — u) is, therefore, dependent on u\ but when a

is u is small and u — A is
quite small, the change, with time, in
almost constant. The temperature of the thermometer then fol-
lows that of the bath in a manner almost similar to that expressed
by equation (6), namely, with a constant difference.^

METHODS OF DETERMINING X
The equations
which the derivations and applications have
of
just been discussed may be easily transformed into some which
can be conveniently employed in the determination of X.
Absolute Determination.^ Equation (4) may be written in the —
form

which, in logarithmic form, is

X=

This linear relation between time and logarithm of temperature

differences is very convenient in getting the best mean value from
a large number of readings of and t, because a graphical plot
(time against logarithms of temperattue difference) should be a
straight line of slope X.
The experimental details for carrying out the determination
are simple. The thermometer is cooled and plunged into a bath

* From the expansion e =1 l—at-{ — ^

) > one can conveniently study the function for values

of t less than 1/a. The expression for u is

u=A+B (.-.+<f-...)
If the precision of the work in hand permit of neglecting at with respect to unity we have u=A-\-B, or
the change in v. is insignificant during the time t considered.
If at be appreciable, but its square and higher powers are negligible, we have u= A-rB {\—at)= {A-\-B)—
aBt, which is the linear change u=uo+rt.
Calorimetric work, generally speaking, never permits of neglecting the first power. Ordinary work
does permit of neglecting the second and higher powers and the temperature changes, which are really
logarithmic, are treated as Unear. Precision work requires retention of at least one more term in the series
Bnd it may be as convenient to keep the exponential as such as to employ a quadratic expansion for it.
' Method given here is the classical procedvu'e, described by Thiesen, GuiUaume, and others.
 —

668 Bulletin of the Bureau of Standards [VcLS

which is stirred at the desired rate, and maintained at constant

temperature throughout the experiment. This condition may
be secured by the use of a bath of large heat capacity and at room
temperature. As the meniscus of the thermometer rises, the times
of passing chosen graduations are read off on a watch or recorded
on a chronograph. The later thermometer readings serve to give
TIM E IN SE'CONDS
the bath temperature t^o- Any
S5'
1
15 6 7 9
?5 10 tl reading may be taken as 0^ and
; .

-H time reckoned from that of this

f^

S3' ^
reading. Corresponding to
1/
f each reading 6, a value of log
01
/ 0—11
/
/
/_ ^ is computed and plotted
)
against time. The slope of the
/ best mean straight line is the
/ numerical value of \. If loga-
/
rithms to base lo instead of the
215°
O base € be used, the modulus must
IS
cc
/ be taken account of in com-
/ puting the slope.
" The chronograph should be
10'. L_ employed if the X to be deter-
1
mined is small. Fig. i illustrates
Fig. 1. Rise of thermometer cooled below 10'^ the rate at which the meniscus of
and plunged into bath maintained at 25^
a thermometer rises from io° to
for \= 2 seconds)
{
25° for X = 2 seconds (small bulb
chemical thermometer in well-stirred water). It makes evident
the desirability of securing time readings to a fraction of a second.
The agreement of results in a well-stirred water bath when a
chronograph is employed has been found to be well within 5 per
cent. Observations taken with a watch depart from a straight
line, on a logarithmic plot, to an extent making it difficult to

determine the slope within this limit of certainty.

Departtu-es of points either side of a straight line (logarithmic
plot) in an irregular fashion, indicate the degree of experimental
error in making readings, but departures in the nature of a con-
sistent curvature are a warning that the thermometer is one to
 '

Harper] Tkermometric Lag 669

which the equations just deduced should not be applied other than
as first approximations. SHght curvature may be considered to
define limits between which lies a mean value that may be used
as X in the equations above, for the approximate computation of
lag corrections to readings. Additional assumptions, modifying
equation (i), lead to a more perfect representation of the behavior
of many such thermometers. (See Sec. II.)

Relative Determinations. —The

in the values of \ difference
for two thermometers, under the same conditions, can be ob-
tained in a bath whose temperature is uniform and rising linearly.
If one be known under the conditions of the intercomparison, the
value of the other is thus obtained.

Applying equation (6) to two thermometers read at the same

instant
6^ — u= —r\\ 62 — u= —y\
Subtracting
e^-e,= -r{\,-x,) -
(10)

^2-^1=^(^1-^2) (11)

The rate of rise, r, is the same for both thermometers and for the
bath (p. 665) and therefore readily determined by observing the
,

progress of either thermometer. The difference {0^ — 0^ ^^ at any

^"^1, 02, refer here, as elsewhere, to readings correc±ed for bore, zero error, etc.; i. e., assutae a perfect
thermometer. Since 6\—9i is generally small and since only the comparative instrmnental correction is
required, much greater precision can be obtained by direct comparison of the thermometers at the time
of the experiment for lag, than by the use of tables of corrections pertaining to the instruments individually.
Convenient procedure is to carry out the lag experiment with the required rate of rise of temperature,
and then so diminish the supply of heat to the bath employed that the condition of constant temperature
is as nearly attained as may conveniently be, and intercompare the thermometers. The departure from
constancy of temperature must be in the direction of a slow rise, never a fall, because the sticking of a falling
meniscus vitiates thermometer comparisons.
The working equations will be:
Let 0=readiQg of thermometer if "perfect."
/?= observed reading.
e=correction for all instrumental errors.
5i-62=(i?l+£l)-(/?2-fe2)
so that equation (ii) is used in form
A2-Ai=-^[(2?i-2?2) + (n-£2)]

"where n— ^n is
determined directly from the mean of a number of pairs of readings taken in the inter
comparison at a very small rate r. Designating such readings with primes, equation (lo) is
(i?'2+£2)-(i?'i+ei)=o
or

because when r is small enough, no matter what ^i, A2, may be, the term r (Aj—Ai) may be neglected.
670 Bulletin of the Bureau of Standards [Vol. g

instant may be repeatedly obtained to give a reliable mean, and

if Xi be kno^^m the data suffices to give X^ with the simplest of
computations. X^ may be known by a previous application of
the method described on p. 667, or it may be computable from
other constants of the thermometer (e. g., see Section III on
platinum resistance thermometers).
Frequently X^ may be neglected, and the method becomes an
approximate absolute one for **slow" thermometers and more
convenient than the other. For example, if the result of an
experiment gives X2 — Xi = 3o seconds, and X^ is surely less than 3
seconds, X3 is determined to be 30 seconds ( ± the experimental
error) mthin 10 per cent. A little familiarity with thermometers
and their behavior in baths stirred at various rates enables one to
place by inspection an upper limit for X for many forms of ther-
mometers and thus apply the method just described to ''slow"
thermometers.
Illustrative Values of X. —
To illustrate the concordance of both
final results and the individual readings pertaining to a single
result, a short series of experiments performed in vigorously stirred
water is tabulated in Table I and plotted on Fig. 2. The method
employed was the logarithmic one described on page 667 et seq.
TABLE I

Lag in
Date Thermometer number Brief description of same well-stirred
water

1909 Sec.
Aug. 30 Chabaud, 77874 ©"-SO" thermometer divided in 09l (1° about 6.6 mm f
2.00
long). Convenient to read 0901. Bulb approxi- 2.13
mately 4,5 mm diameter and 25 mm long. 2.15
Do. Golaz, 4192 Open-scale calorimetric thermometer divided in 0902 [
4.82
(1° about 31 mm long). Convenient to read 09001.
J
4.95
Bulb approximately 9 mm diameter and 52 mm 1 4.96
long. [ 4.78
Do. 5951 "Einschlusz Faden." Bulb a long narrow thread of 52
mercury, 100 mm X 2.5 mm (diameter) surrounded
by an air space of about 3 mm, outer envelope of glass
approximately 8 mm in diameter inclosing the whole.
Do. 1787 Callendar type platinum resistance thermometer.
Fine platinum wire coil wound on a cross of sheet 15.3
mica and inclosed in a porcelain tube about 30 cm 15.7
long, 1 cm diameter, and 0.15 cm thick.
 —

Harper] Thermometric Lag 671

TIME IN SECONDS
30

Fig. 2. Typical curves obtained in determining X {vigorously stirred water) . To illustrate

degree of concordance obtained for individual readings forming one determination; and also
for slopes of the several determinations with a given thermometer

73764°— 13 5
672 Bulletin of the Bureau of Standards [Vol. 8

A brief description of the thermometers is included, giving an

idea of the value of X for different types. The lag of a platinum
resistance thermometer of the Callendar type is tabulated with

those of the mercurial instruments to show its comparative mag-

nitude.
VARIATION OF X WITH STIRRING

With a given thermometer of the usual type in a given medium

stirred at a certain rate, a definite numerical value for \ in the equa-
tion (i)

may be found and applied in this equation, or integrated forms

deduced from it. But this same value of \ must not be employed
if themedium be stirred at a different rate.
—
Lag IN Water. As an interesting example arising under con-
ditions of use, the following valueswere measured in a calorimeter
with a propeller stirrer rotated as slowly and as rapidly as the
convenience of the calorimeter stirring arrangement would permit.

TABLE II

Thermometer Golaz 4191

X MeanX

Sec. Sec.
4.22
1
Vigorous Rtirring 4.34
4.42 1
•
6.46
1
Slow stirring 6.40
6.57 1
"
15.5
1
No stirring 14.5 14.5
14

An increase of 50 per cent in the value of X is thus observed

upon decreasing from the normal vigorous stirring of calorimetric
use to a propeller speed of one-third the normal. Such an experi-
ment gives no definite relations, however, because the velocity of
 —

Harper] Thermometric Lag 673

flow past the thermometer bulb is not known. Accordingly, a

special apparatus was assembled in which the velocity of flow
could be measured. The schematic drawing, Fig. 3, renders
description unnecessary.
The velocity of the ciurent of water past the thermometer bulb
was computed from the quantity of water delivered in a definite

time and the area of the annular space -(1.19^ — 0.54^) = [Link]^ I

between the thermometer bulb and the surrounding glass tube.

TO CONSTANT TEMPERATURE,
CONSTANT-HEAD, WATER SUPPLY

THERMOMETER
(CENTERED BY GUIDES)

2000- C C
MEASURING FUS"K

Fig. 3. Apparatus for determining lag of thermometer in a stream

of liquid flowing at a definite rate

This does not take account of the " drag" near the walls, but the
nature of the problem in hand does not warrant such refinement.
With the stream flowing at the desired rate, and its temperature
that of the room (between 30° and 35°) the thermometer was
cooled in ice and dropped into place in the tube, centering by
guides. As the meniscus passed appropriate graduations between
10° and room temperature the observer made a record chrono-
graphically. This was repeated several times for each velocity
tested. The results are summarized in the accompanying Table
 »

674 Bulletin of the Bureau of Standards {Vol. 8

III and plotted on Fig. 4. The method of computing results was

the logarithmic plot method described on page 667, and the indi-
vidual points lie on a straight line very closely. The results of
the several determinations at each rate of stirring are in good
agreement, as shown by the table.

TABLE III

Value of \ for Various Velocities of Water Flow Past Bulb (in Apparatus
of Fig. 3)
Thermometer Chabaud 80659

Experi-
Mean Av. dev.
Velocity
MeanX Av. dev.
Date ments
flow cc
from mean
cm per sec. from
per sec. sec.

1911

July 13 0.62 [Link] 0.70 5.6 ±0.1«

Do 1.14 [Link] 1.28 4.6 [Link]
Do 1.67 0.01» 1.9o 4.2 O.I3

Do 1.87 0.02 2.1 4.1 O.O4

Do 3.3 (1 obs.) 3.8 3.54 O.O5

July6 4.9 (1 obs.) 5.5 3.24 O.O5

July 13 5.3 .0(2 obs.) 6.0 3.22 0.04

July6 9.6 0.1 10.9 2.89 0.03
Do 20.9 0.2 23.1 2.59 0.06
Do 33. 0.1 37 2.37 0.02
Do 40.0 0.1 45 2.41 0.03

As the velocity decreased, the logarithmic plots curved a little,

making accurate determination of X impossible. It is to be noted,
however, that considerable latitude in the value of X (vertical dis-
placement of a point) at the lower velocities would not greatly
affect the shape of the curve.
For the determination of lag imder conditions approximating
infinite velocity the thermometer was plunged into steam and read
as above described. The heat supplied instantaneously as steam
condenses on the bulb maintains its surface at the temperature of
the steam, less the drop through the layer of water formed there.
The thickness of this, if uniformly distributed over the bulb, would
be less than 0.05 mm at the conclusion of the experiment, a film
of the order of one-tenth the thickness of the glass in the bulb.
It may
be concluded that the behavior of the thermometer under
the conditions of this experi^ient is probably not very different
 —

Harper] Thermometric Lag 675

10.000
I MEAr9
{JO NEAREST SEC.) OF

LOO

8.00

7.00

6.00

5.00

4.00

aoo

ADDITIONAL POjKTS, BEYOND MARGIN

AT VEU 23,5 CM. PEB SEC. X=2._59G-»"
41 " M if X'=2:-39G—»•
100. o /* V %=2..30Q-*
co(?) X=2.L60-*- ASYK PTOTE OF CURVE (?)
8.00
3 4 5 6 7 8 9 10 11

VELOCITY OF WATER PAST BULB, IN CM PER SEC.

•Fig. 4. Variation of X with stirring
Thermometer Chabaud 80659 in water
 . . ;

676 Bulletin of the Bureau of Standards [Vol. s

from what its behavior might be if an infinite supply of heat could

be instantaneously brought to the surface of the bulb."
The values of X obtained in steam are given in Table IV. Here .

are also the values obtained in water at rest except for its own
nattiral convection.
TABLE IV
Limiting Values of /\ for Curve of Fig. 4 (Zero Velocity of Water Past Bulb
Infmite Velocity (?) of Any Medium Past Bulb)
Thermometer Chabaud 80659

Experi-
Av. dev.
Date Immersion Remarks ments from MeanX
mean

1911 Sec
July 10... Slightly superheated steam" . Equivalent of infinite stirring 6 ±0.04 2.16
(any medium)?
Do... Still water in calorimeter can. . Lag curves not straight lines. 3 11
Cf. pp. 668-9.

July 13... do do 3 1 9

12 Steam determinations in International Bureau form of steam-point apparatus (due to Chappuis),

Pressure of steam about 5 mm of vv^ater in excess of atmosphere.
The determinations
in imstirred water give wide latitude of
variation, probably due to differences in convection currents. A
mean value of X of lo seconds was obtained, using a large can of
water at constant temperature as the immersion bath.
The minimum value of X obtained in the experiments summarized
in Table III being 2.4 seconds, it seemed desirable to test higher
velocities and see whether the steam value of 2.2 seconds was more
closely approached. Modifications of the apparatus for the pur-
pose of securing a larger flow somewhat impaired the accuracy of
measurement. One hundred cm per second was attained and the
mean X found was 2.3 seconds (the statement of a third figure not
being warranted by the results)
Lag in Kerosene Oii.. On accoimt of the great variation in —
the rate of change of the lag with the velocity as shown on Fig. 4,
a short investigation with another liquid was imdertaken to see
11 If this determined in this way should be independent of the medium; i. e., a characteristic
be true, the A
constant of the thermometer. Any other vapor of high latent heat should give the same result. An experi-
ment in alcohol gave a value of A (2.5— seconds) somewhat larger than in steam (2.2 seconds), being about
equal to the A foimd in water at the highest velocity measured with the apparatus of Fig. 3 (A=2.4 seconds at
45 cm per second. Table III).
 — ,

Harper] Therviometric Lag 677

whether the same general form of curve (when the same units of
measurement were employed) would be found. A kerosene oil was
employed with the same thermometer and gave a very similar plot.
As might be expected, the more viscous liquid gives a greater
value of \ for a given velocity (Fig. 5) It must be noted that the .

values given were all obtained in one tube (internal bore 1 2 mm)
and the question of change in the relations at that velocity where
the flow through a tube changes from a steady drift to a turbulent
motion, involving experiments in tubes of several sizes, would
enter into a more complete study of the subject.

13

12

n
JO \
\

CO
9
\
\ \
Q
\
1

2
V
Q i

\
^^
^
1

-^6

5
t

\
\
m W # ^ ,(KE ROSE NE)

A 44->(

\ -%
% rl:

4
V. «<^6
Q -Ti/Efi MOMf ;xER_ N-^ FERj LkEc rawn!^TO SCALE ,OFj XIRVE
-ABC)VE)
1
8 10 12 U 16 18 20 22 2i g6 28 30
VELOCITY PAST BULB IN CM. PER SEC.

Fig. 5. Variation of\ with stirring

Thermometer Chabaud 80659 in a kerosene oil

The data obtained for kerosene oil, including the density and
viscosity of the oil at 20° C, are summarized in Table V, and the
results plotted as Fig. 5. The discrepancies between individual
observations are more likely due to lack of care in obtaining the
data than inherent in the method or the heat-convecting prop-
erties of the oil. No great pains were taken concerning tempera-
ture regulation of oil, and the observations were taken more hastily
678 Bulletin of the Bureau of Standards {Vol. 8

than for the work in water, as an accuracy of 5 per cent seemed

quite sufficient to illustrate what was desired. No observations
are discarded in making Table V.
TABLE V
Lag of a Thermometer in Oil. (Apparatus of Fig. 3)
Thermometer Chabaud 80659

fSfixture ol kerosene oils: Density, at 20° C, 0.870 g per cc; viscosity, at 20" C, 5° Engler," equivalent to 0.31
dynes per cm*]

Experi-
Mean flow Av. dev.
Velocity 1
Av. dev.
Date ments ccper
from mean
cm per |
Mean X from mean
sec. sec.
1

1911 Sec.
Nov. 13 6 40-50 H(20)
Nov. 11 3 0.96 ±0.04 1.08 13.0 0.1
Nov. 13 4 1.80 0.15 2.03 10.6 0.2
Nov. 11 4 3.2 0.15 3.6 8.2 0.0s
Nov. 10 6 7.5 0.33 8.5 6.7 [Link]
Nov. 11 4 11.9 0.2 13.5 5.9 O.O7
Do. . ... 4 38 1.5 44 4.8 0. Ij

13 A
viscosity of s° Engler means that a fluid of that viscosity runs through the efflux tube of an Engler
viscosimeter at a rate one-fifth that for water at same temperature.
" Determinations of the lag in this oil imstirred showed large variation, owing probably to the relatively
high viscosity of the oil interfering with convective interchange of heat between the liquid near the bulb
and the mass of the liquid.

Lag of a Thermometer in Air.^^ —^The lag of a thermometer

employed to measure the temperature of gas has considerable
practical interest because it is great enough to affect the results
quite appreciably. It is not, however, easy to make proper cor-
rection in the usual case, because the thermometer is subject to
»5 subject a large number of papers have been contributed, chiefly in the meteorological journals.
To this
Most must be characterized as little more than qualitative investigations, because the experi-
of these
menters do not more closely specify the velocities employed than as "strong wind. " hghtly moving vdnd,"
'
'

''
qtdet room," etc. The most complete quantitative papers are by Hergesell, Meteorologische Zeitschrift,
14 pp. 121 and 433, 1897. Simultaneously with this appeared the paper of J. Hartmann, Zeitschrift fiir In-
strumentenkunde, 17, p. 14 (1897), which is qualitative only.
Wnhelm Schmidt, Meteorologische Zeitschrift, 27, p. 400 (1910), contributes data for over a dozen ther-
mometers, mercurial, alcohol, toluene, and metallic expansion, under a variety of conditions, but vtitb-
out quantitative specification thereof.
de Quervaia, in the same journal, 28, p. 88 (1911), reviews previous work and deduces therefrom some
formtdse for which it appears that he claims great generaUty. Certainly they are not imiversal in their
appUcation, for they fail to agree with the observed behavior of the thermometer used as an example in
the present paper, a thermometer of the ordinary' "chemical" type in a current of air. Brief reviews by
de Quervain summarize the saUent features from the following papers:
Dufour in 1864 (see Meteorologische Zeitschrift, 14 p. 276, 1897); Hartmann, loc. dt.; Hergesell, loc. cit.;
Valentin, Meteorologische Zeitschrift, 18 p. 257, 1901; Maurer, ibid., 15 p. 182, 1898, 21 p. 489, 1904.
Rudel, in the same issue of the ileteorologische Zeitschrift, 28, p. 90, 191 1, gives the data for some ther-
mometers and reviews the work of Krell, Zeitschrift fiir Heizung, Liiftimg, und Beleuchtung, 11, 1906-7.
Msrvin, Monthly Weather Review, 27, p. 458, 1899, gives some data pertaining to the lag of kite ther-
mographs. ,
 —

Harper] Thermometric Lag 679

drafts of widely different velocities X holds

and no single value for
for more than a few moments. To
the magnitudes
illustrate
involved, it may be well to anticipate the data tabulated below
and discuss the lag in air of the thermometer whose constants in
water and oil have just been given. In still air the mean X found
was 190 seconds. The time of 7X seconds for an initial differ-
ence of 10° between the thermometer and still air in which it
might be immersed, to be reduced to o?oi (see p. 665) is accord-
ingly over 20 minutes. The reading of a gas temperature to
htmdredths of a degree with a mercurial thermometer must there-
fore be imdertaken with due lapse of time permitted after immer-
sion. Then, too, in a space warming a degree in 15 minutes (if

TO BLOWER

J
«fr=^ /?

V^
"ANT"^^
CONSTANT
TUBE
TEMPERATUeE /
eOPPEB WORM GAS METEa PRESSURE STEADYING
WATEFl
1 TURNaKocy. Er.5s2830 e.c. capacity

Fig. 6. Apparatus for determining lag of thermometer in a stream of gas flowing at a definite
rate

there be no drafts) the error in any reading of this thermometer

after reaching the equilibrium state, would be over o?2o. In
the presence of drafts the numbers cited are considerably reduced,
but they are worth consideration as examples of the error under
the worst conditions.
To study the variation of the lag with drafts in air, thermom-
eter Chabaud No. 80659 was immersed in currents of varioiis
velocities in the same U tube as that employed for water and oil
(Fig. 3). The current of air was measured by a gas meter in
series with the tube. The area of the annular orifice past the
thermometer being the same as for the experiments with water,
the computation on page 673 of 1.13 cm/sec velocity for each cc
per second current holds good here. The accuracy claimed for
the tabulated velocities is but 5 per cent.
68o Bulletin of the Bureau of Standards [Vol 8

Since the wet meter raised the humidity of the air above the
point where dew condensed on the thermometer bulb, when
cooled sufficiently for the experiment, a drying tower of CaClg
was necessary. A copper worm in a thermostatic water bath was
introduced to steady temperature fluctuations. The apparatus
is shown diagrammatically in Fig. 6.
The resistance of the piping employed placed the upper limit,
for the apparatus shown, at a current of about 90 cm per second.
To obtain some points on the curve at higher velocity the appa-
ratus was modified by substituting a dry meter, of larger capacity,
dispensing with the drying tower and also the copper worm.
With this arrangement, velocities up to 1000 cm per second were
obtained. The value of X of the thermometer is, at this velocity,
23 seconds, whence it appears that an almost inconceivably large
velocity of gas past the bulb of a thermometer would be necessary
to supply heat as fast as the surface can transmit it to the in-
terior, corresponding to a value of 2.2 seconds for X, the value in
steam (p. 674).
The logarithmic method described on page 667 was modified
slightly to a more convenient form to avoid plotting lines and com-
puting logarithms. This is possible when the motion of the ther-
mometer meniscus is comparatively slow throughout the scale;
i. e., when X is large. Every X seconds a given difference of tem-
perature between medium and thermometer is reduced to e-^
times its initial value. From tables of e-^ we obtain the following
data for an initial difference of 10°.

TABLE VI
Temperature difference For bath at 30°,
(£-^) At time t.
equals— thermometer reads—

eO=l 10900 20900

X
e-i= .6065 6.06 23.94
2

f-i= .3679 A 3.68 ;>6.32

3;,
e-3= .2231 2.23 27.77
2

£-2= .1353 2A 1.35 28.65

5/
0.82 29.18
e-§= .0821 T
£-3= ,0493 3;. 0.50 29.50
7;.
£-?= .0302 0.30 29.70
2
-*= .0183 4/ 0.18 29.82
 ' 1

Harper] Thermometric Lag 68

Cooling a thermometer and plunging it into a bath at 30° one

would read off the times corresponding to the thermometer read-
ings tabulated in the last column. These should occur at equal
intervals of half X. thermometer readings will be odd
Most of the
valued and not easily carried in mind, no matter what temperature
range is chosen, whence the method is hardly to be recommended
imless the observation interval be great enough to permit of ref-
erence to notes between readings. Also with a rapidly moving
meniscus odd values can not be timed as accurately as can coinci-
dences with even graduation lines.

JJUU
" ... ~

150

Q (

Z \

Ul
\\
«100
z V\
y< x
.

>;x.

.^ -0

50
r^ '--
— D-, , .

,A T 60 CN .PE. SE
. *
AL ON
Ull 670
BEY DND MARGKvi ^Gj ->A T 10 j:. rt

_ _ L __ _L
50 100 150 200 250 300
VELOCITY OF AIR PAST BULB, IN CM .PER SEC.

Fig. 7. — Variation of\ with current, in air. Thermometer Chabaud 80659

It is very easy to see that a slight change in the bath tempera-

ture affects the time-differences for the later readings by a large
percentage consequently the first three or four differences are the
;

more dependable, and it is desirable to employ a table of readings

made out for every \/2 seconds rather than for any greater interval,
to secure a number of intervals before the temperature difference
grov/s too small.
The results of the experiments in air are summarized in Table
VII and shown graphically on Fig. 7.
 .

682 Bulletin of the Bureau of Standards [Voi.s

TABLE VII

Values of A. in Air Passing Thermometer Bulb at Different Rates

Thermometer Chabaud 80659

Velocityis Av. dev.

Experi-
Date Apparatus, etc.
ments
cm per MeanX from
sec.

1911 Sec.
July 25.., Inclosed space—air at rest except convection 3 190 ±11
July 27... U tube, etc. (no current) 1 190
Sept. 14.. U tube, etc.. Fig. 6 (wet meter) 3 6.3 144 5

Do.. ....do 3 11 126 2

Do.. do 3 26 94 4

Aug. 19. ....do 1 28.6 86

Do.. ....do 2 47 73

July 27... Dry meter (tower and worm removed) 1 52 70

Aug. 17.- Fig. 6 (wet meter) 4 86 64 2

July27... Dry meter, etc 2 95 60 0.5

July 26.. do 2 182 46 0.0

Do.. -...do 2 325 37 0.5
Do.. do 2 60, 30 0.5
Do.. do 3 67, 28 0.0
July 28.. do 2 106, 24 0.5

July 25.. Suspended vertically in greatest draft of high-speed 3 ? 24 0.5

horizontal fan.

18The average deviation from the mean velocity is without significance; as to the precision read, different
observations were in exact agreementamong themselves. Readings were taken to about i per cent, bu^the
gas-meter caUbration was sHghtly uncertain, making a systematic error greater than this possible.

n. LAG OF A BECKMANN THERMOMETER

ADDITIONAL ASSUMPTIONS

It is found that the behavior of the ordinary type of Beckmann

thermometer is not completely represented by the equations that

have been developed, whence the assumption =^7 = :r(^ — ^) , equation

Ot A
(i) (see p. 66 1 ) is not justifiable. A consideration of the form of
the instrument suggests a reason for this. The main bulb is Uke
the bulb of the type of instrament considered above. But in addi-
tion the large capillary, which is common in such instruments,
between the bulb and the zero of the scale, acts as a second smaller
 —

Harper] Thermometric Lag 683

bulb.i^ Inclosed in a tube with an air layer between it and the

bath, it is quite slow to assume the bath temperature, yet the
amoimt of mercury in this secondary bulb is sufficient to appre-
ciably affect the position of the meniscus in the small capillary.
There are thus two biilbs, the temperature of each of which may
dA I
be expressed by the law --rr = -{u — A) where uis the temperature
dt A.

of the surroimding medium and A is the average temperature of

the bulb, as previously defined (p. 661).

—- = -- (u — B) for main bulb (12)

dt A.B

_- = - {u — C) for large capillary (13)

dt A.-

B and C together define the position of the meniscus in the bore

of the thermometer, this being the read-
ing 6. If the relation of ^ to 5 and C be
stated (this being equation 14 below) it

will be possible to eliminate from the three

equations (12), (13), (14) the two quan-
tities B and C, which are not directly de-
{-GLASS
terminable, and leave a relation connect-
ing 6 with u, quantities in which interest
BOUNDARY FOR centers.
OFTEN NOT
•fc"

WELL DEFINED
In stating the dependence of 6 upon B
and C one has to bear in mind
Firstly, that the volume of the mercury
Fig. Z.—Section through Beck-
originally in the bulb will change propor-
mann thermometer
tionately to the change in the temperatiure
B, and the volume of that originally in the large capillary will be
likewise related to C.
Secondly, the position of the meniscus in the bore will vary as
the sum of these two voliunes changes.
1^ Total immersion is asstimed in this paper, along with the assmnption previously mentioned that the

thermometer is instnmientally perfect. The difference in the behavior of any thermometer between total
and partial immersion must be considered as part of the theory of emergent stem corrections and can not
be included here. In the experiments tabulated in this section the thermometer was immersed to the top of
the large capillary, the results given being on the basis of total immersion.
684 Bulletin of the Bureau of Standards [Voi.s

Whence dO will be the sum two quantities proportional to

of
dB and dC, respectively, and, no constant of integration
since
except zero could satisfy the obvious relation ^ = when both
B and C are 0, must be related to B and C exactly as dO to
dB and dC, although this does not imply that the temperature of
the thermometer as a whole is to be thought of as an addition of
the temperatures of the several parts.
If unit quantity of mercury be distributed between two bulbs,
k in the first and i-k in the second, the effect of each in defining the
position of the meniscus will be in the proportion ^ to i-k, besides
the effect of the respective temperatures. We may take for the
imit of quantity the total amount mercury in any thermometer
of
and so omit factors of proportionality and state the relation in the
simple form
e^kc+{i-k)B (14)

where k is the fraction of the total volume contained in the large

capillary.
Omitting the steps of the elimination of B and C from (12), (13),
and (14) and collecting, the resulting equation is

^.K^+{^. + K)^ + e = [k\s+{i-k)K]^^+u (15)

If equations deduced from it are verified by experiment this equa-

tion, which can not be directly tested, will be justified. We shall

find that a sufficient agreement obtains to do this, the equations
developed from it proving to be fairly close approximations to
exact statements of the behavior of this type of thermometer. It
must be noted that k can not be accurately determined, since the
end of the bulb and beginning of the large capillary is not thermally
a definite location, even if mechanically it were so. (^ig. 8.)
To obtain the primitive of (15) requires considerable mathe-
matical manipulation, but presents no difficulties as the steps
follow common textbook suggestions in ^^ order, being type-form

18 See, for instance, A. R. Forsyth: A Treatise on Differential Equations — " General linear equations with
constant coefhcients," p. 64 (3d ed., 1903); Notes on particular integral and complementary function in
the section on General Linear Equations of the Second Order, p. 98; " Method of variation of parameters,"
p. no.
Harper] Thermometric Lag 685

processes. The result is

e = u-\-A,e-~^J + A,e~-^ci-ke~^n "^e^c'dt- {i -k)€-^B' I ^e'^^.'dt (16)

where A^ and A 2 are the arbitrary constants of integration and

must be fixed by assigning two definite conditions. One of these
may well be to assign simultaneous values to all the variables,
defining for the time that the corresponding bath temperature
be Uq and thermometer reading 0^. The most convenient second
condition to impose is that all parts of the thermometer shall be
at the same temperature at this time zero, i. e., Bq = Cq. Since
= kC+{i —k)B, whenever B = C, either of these equals 6, so
that this second condition is expressed by the relation

^o = ^o = Co (17)

Whenever the thermometer remains medium of constant

in a
temperature for a considerable time all parts come to this tempera-
ture and the condition (17) is then fulfilled at every instant; so
that by taking as = the instant of the transfer to any medium
if

at a different temperature (for immersion in which latter medium

the equation is to be applied) both conditions outlined can be
readily satisfied in practice.
Ai and A^ are given by the two equations

gXc'gj^_(l^^) ^J^^dt (18)

J

from the substitution of initial values in (i 6) , and

from substitution in the first derivative of (16) of the first deriva-

tive of (14) and subsequent reduction of the result by the use of
(12) and (13) to a form where the values at the time t = (relation
1 7) can be substituted to give the form as written.

Equations (18) and (19) are not worth solving explicitly for A^,
A 2 for substitution in (16), because the special cases arising in
practice are more easily referred to these implicit forms. The
cases arising more frequently in laboratory practice will be treated
686 Bulletin of the Bureau of Standards [Voi.s

but one of these must be developed at this point.

in a later section,
By Beckmann thermometer plunged
considering the behavior of a
into a bath maintained at constant temperature, and the behavior
predicted b}^ the theory here developed, we shall find that an
excellent test of the validity of the theory is afforded.
^u
(a) Constant Temperatmre. — If u=Uo, ^= and equation (i6)

becomes
e==u, +A,€-k' + A,€--y (20)

Equations (18) and (19) have the forms

(h. l^\ (^0-^0)=^^+^^

From, which A^ and A 2 may be readily obtained.

A,^{i-k){e,-u,) (21)

A,=k{e,-u,) (22)

^®
Equation (20) in its complete form is

|i:^=(i-^)e-l' + ^e-b (23)

under the conditions, at t = o, imposed above (p. 685).

When k= or k=i or X^ = Xb, any one of which conditions
corresponds to a single bulb instead of a compound one, this
equation reduces to (4) of p. 664 as is necessary.
The properties of functions of e'^ depend so entirely on the
nimierical value of the exponent that it is rather difficult to
generalize, but a few remarks may assist to a clearer picture of

19 An equation of this general form was proposed by Thiesen (loc. cit.), "with a bare statement that the
idea had occurred to him of separate,independent lags for two parts of those thermometers which failed
of expression by the more common single lag equations. No derivation of the equation is given, and
1 1
—3- t —T.t
it is left with vindetermined coefficients in the form 9—L'o=A€ ^^ +Be ^^ , with the suggestion that
_1
this form be tried in cases where 6— U«F=Ae ^ obviously fails to express the behavior of the instrument.
 —

Harper] Thermometric Lag 687

the curve theoretically representing the behavior of a thermometer

of the usual Beckmann type plimged into a constant temperature
bath. To illustrate the discussion, Fig. 9 is inserted, although it
has been necessary to exaggerate greatly the value of k from the
usual size, in order to separate sufficiently the curves, on the scale
which must be here employed, to make clear which ftmction is
which.
•

1.0 \
'\\
\
\
\^
.8

\\ f t
40
\
.6 \ M- i^

ii'
\
\^
A <tS «J V/>

^^

h ^ ^^
f'V^,
^^>
'
i
,^ t

i-
40
'
J

rs} "^ -^
—» __
— — -- rzr
'"*--
f^ft; ' - -
(TE ^M DUE "to CAP icci^rv) ~~~
0.0 m:
2 4 6 8 10 12 14 16 18 2i 26 28 30 34 36 38 40
t

Rg. 9. Curves to explain theory of Beckmann lag {exaggerated value ofk)

--1+ r -A,
\b=8 sec. \c=40 sec. k=0.1 y=e ^^ +£e a/

Let US confine our attention to the two terms separately, that

1

due to the capillary, ^, and then that due to the bulb,

I — K'

e~~^B. K is small with respect to unity, and Xg is generally

several times Xj,. Fixing in mind X^ as the interval of time for use

as a convenient comparison unit, t must be large before e"^^'

reduces very greatly from imity; whence for a considerable time

k _i
the term —^^
^^ is very little different from k. On Fig. 9, with the
numbers employed, it starts at the value 1/9 and very gradually
approaches the axis.
73764°— 13 6
688 Bulletin of the Bureau of Standards [Voi. 8

The term e V decreases at a rate very much more rapid than

At
that of e~^c , \b being but a fraction of X^, so that although its
initial value unity is very large in comparison with k, it reduces

to this value, k, before very long, and continuing to decrease dis-

appears numerically while the other term is appreciable.
The svmi of the terms in equation (23) is accordingly almost
exactly the first for small values of t and the second after con-
siderable time. This point is illustrated by Fig. 9, and the actual
magnitudes which are involved in a practical case can be pictured
by supposing k diminished perhaps as much as thirty times. At
the one end the difference between the resultant and its first
term would be reduced from 10 per cent of either, as there shown,
to a very small fraction of it, or for most purposes the curves might
be considered identical. At the other end the total magnitude
of the resultant would of course be much smaller, but it would,
at some time, bear exactly the same relation to the second term
that it does on the plot, namely, become practically equal to it,
because the first term is diminishing at all points at a rate much
greater than that of the second. It may also be well to call
attention to the fact that as k is smaller, the point where the first
and second terms are the same size, about 22 seconds on Fig. 9,
displaces more and more toward the right.
The conclusions that have been reached predict that if a ther-
mometer of this type be plunged into a bath 10° warmer than it is,
it will cover the first 9°, say, of its rise in almost the same way as

would a thermometer with a single lag constant X^ and will cover ;

(about) the last o? i in nearly the same way as would a thermometer

of single lag constant X^; the interval between corresponding to
neither.
JUSTIFICATION OF ASSUMPTIONS MADE
The method of testing out the theory proposed is at once sug-
gested. The value of k may be approximately determined for a
thermometer with a secondary " bulb, such as the large capillary
*'

below the scale in the usual type of Beckmann. Then Xb and X^

may be approximately found by using the two ends of the observed
"lag ciu-ve" of that thermometer, obtained by the logarithmic
Harper] Tkermometric Lag 689

method described on page 667. If, then, the middle of the curve
computed from these values of X^, X^, and k in the equation (23) be
the same as the middle of the curve obtained by direct observation,
it is fair evidence that the function written is a proper one to repre-
sent the behavior of the thermometer. The test was made and
the results with one thermometer are given in full to make clear
the procedinre.
Beckmann No. 5952 has, in its large capillary (Fig. 8) about 20°
of merctu-y inclosed by the outer glass tube so as to have only
poor thermal contact with the medium of immersion. 6300° being
approximately the volume^^ of the mercury in a thermometer bulb,
the value of k is 20/6300 = 0.0033; and i—k is imity within the
limits of accttracy of this computation. Substituting these num-
bers in equation (23),

^—^=.^.+.003,.

should be the equation to give the reading (6) of this thermometer,

at any time (t) after immersion in a bath maintained at constant
temperatinre (Uq) provided ,
all parts of the thermometer were at the
same temperature (^0) when the instrument was introduced into

the bath. If Oq be below Uq the first 0.9 of the rise must fol-

low very closely the equation ^ —^= [Link]^^b , and applying the
methods of page 667 to simultaneous readings of thermometer
and time an approximate value of X^ may be determined. A num-
ber of such experiments made with thetmometer No. 5952 gave a
mean value of 8.7 seconds for X^.
By the time the quantity (Uq — 0) is reduced to i per cent of

(Uq — Oq) the term€~^B^ is smaller than 0.0032 e~^c' if \

be about
five times \b, and a little later may, for first approximations be
neglected. The equation
0-Uo = 0.00326-h
^
e.-t/o
*• The unit of volume being the volume that forms one degree in the stem of the particular thermometer of

which the bulb is a part. The number 6300 depends somewhat on the glass, but the relative expansion
ooeffidettt is seldom far from 0.00016=1/6300 (approximate).
690 Bulletin of the Bureau of Standards [Vol. 8

applied for a number of readings after 6 has almost reached Uq

gives Xc • Since the temperature differences to be read are extremely
small, little more than the order of magnitude of \c may be
deduced. This however is all that is necessary, as an error of 20
to 30 per cent would not greatly influence ntunerical values in the
complete equation. The mean of a number of experiments
indicated 50 seconds to be the best value to use for Xc-
In Table VIII is summarized the computation of
1
^
_i
F^(i — .0032) e '^B +.0032 € ^c*

for given values of t; \b taken as 8.70 seconds, \c, 50 seconds.

TABLE VIII

t (l-.0032).-g-fo .0032 e-^ F logio F logio e-g^o

10 0. 3160 0.0026 0.3186 9.503 9.501

20 .1002 .0021 .1023 9.010 9.002
30 .0317 .0018 .0335 8.524 8.502
40 .0101 .0014 .0116 8.064 8.003
50 .0032 .0012 .0044 7.643 7.505
60 .0010 .0010 .0020 7.301 7.005
80 .00010 .00065 .00076 6.881 6.007
100 .00001 .00043 .00044 6.643 5.008
120 . 00000 . 00029 . 00029 6.462 4.010 .

The second and fourth columns show at a glance the extent of

the agreement of the function F with
term, the fifth and its first

sixth columns showing the same for the logarithms, which are
more apt to be employed in computing a lag experiment. The
sixth column is of course a straight line, when plotted against
time; the fifth is so up to about 40 seconds, after which it rapidly
assumes marked curvature. This length of time after immersion
was accordingly the interval available for the determination of Xp.
This curvature of the logarithmic plot, due to the term involving
the lag of the capillary, distinguishes it sharply from the plot due
to a single lag, which is linear. The functions themselves are of
the same general shape as illustrated on Fig. 9, so that in comparing
the function F to the corresponding curve obtained by experi-
ment, it is better to compare logarithms than direct values. This
is done in Fig. 10.
Harper] Thermometric Lag 691

-
\f.i)U

\ f

\ \

^
9.00 \
\
\\

V
8.50
\
V
\\

k Q SERIES OF POINTS COMPUTED

\ X,

FROM LOG 10 5 — :, OBTAINED

\ V BY OBSERVING
EXPERIMENTS.
t/

IN EACH
IN TWO
, ^q"*^
^. Uo=29?!
BECKMANN THERMOMEl bK NO, 5952

\\ y

C\
Y y

7.50 \\
\^
\
\
\
\\
4 ^-K
7.00
^\ \)
6 \
k -A \
^Nd
\ \
V T
^h".
C<o
;

\ ^,
"^
\
t.0 20 30 40 50 60 70 90 100
t IN SECONDS
_ t t_
~^0
[Link].—Lo^iQ F F=il-0.0032)€ ^^0_^o.0032€
692 Bulletin of the Bureau of StaTidards [v^t. s

If equation (23) represents the facts, log F, of Table YIII, and

B— U
log ^ jj-, as obtained experimentally for this thermometer, must
coincide. A number of experiments were made by cooling it to 0°
throughout and plunging it into a bath maintained at 29°. 10,
reading as it warmed up. The data from two representative
experiments gave the circles and crosses plotted on Fig. 10. The
agreement with logio ^ {^^^ curve as shown) seems to justify the
theory proposed for this type of thermometer.

SPECUL CASES
The more important special cases for which the theory^ will be
developed somewhat in detail are three, (a) constant temperature,
(b) linear change, (c) exponential change, according to the law
u = A+B€-''K
The application of the equations to practical problems proceeds
along the lines already given in the section devoted to common
or "chemical" thermometers.
(a) Constant Temperature. —The equations for this condition
have been derived above (p. 686) and need not be repeated. The
only question of lag v^hich arises is: How long after immersion of
the thermometer in a constant temperature bath must an observer
wait to secure a given accuracy in reading ? is computed The time
from equation (23), following the principle outlined for an ordinary
thermometer in the discussion following equation (4). However,
as the time for a Beckmann to attain an ''equilibrium condition"
when plunged into a liquid bath is greater than that for a common-
type chemical thermom-eter, and it may be advisable to summarize
the computations of an example to show the order of magnitude.
The quantities involved are
_i i_

{i-k)€ ^J and ke'^c*

For the thermometer No. 5952 (in stirred water).
fe-= 0.0032 Xjj = 8.70 sec. Xc = 50sec=
_i
The term {i—k)e ^g' becomes io~^ after about 9X3 seconds,
_i
or 78 seconds, but at that time ke ^c^ is 0.0032 X 0.21 1, or 0.00067.
Harper] Tkermometric Lag 693

This term does not diminish to iq-^ tmtil t reaches the value of
174 seconds. At this time the first term is so small that the smn
is identical with the latter value. Accordingly, about three min-
utes is the time for a temperature difference of 10° to be reduced
to o?ooi when this Beckmann thermometer is plimged into a
bath of water vigorously stirred.

(b) Linear Rise of Temperature. —m =UQ + rt, ^= r. Sub-

stituting this value of -^ inequation (16), (p. 685), and performing

the integrations, the result is

e = ui-A^e~~^B' + A,e~~^c'~rk\^-r(i-k)\B (^4)

Aj and A may
2 be determined, under the conditions imposed

(p. 685), by equations (18) and (19). Replacing -^ of these

equations by r and carrying out the solution ior A^, A^,

A,= (i-k){e,-Uo + r\,) (25)

A2-=k{e,-U, + r\,) (26)

These values may be put into equation (24) to give its complete
form. A considerable time after immersion the terms containing
_i
them reduce to negligible size because of the factor e ^\ and the
equation has the simple form.

0=-u — r{i—k)\s — rk\c (27)

whence the thermometer follows the bath in which it is immersed

with a constant difference of temperature existing between them of

r[{i-k)Xs + k\,] (28)

Although \c is generally several times X^, k is usually so small

a fraction of i —^
that the last term is quite small with respect to
the first, being negligible more often than not.

For the thermometer previously discussed, Xb = 8.7o seconds

and {i — k)\B + k\c = ^'^3 seconds, a difference less than 2 per
694 Bulletin of the Bureau of Standards [va. 8

cent and of the order of the uncertainty of \b as determined.

The lag of this thermometer in the equihbrium condition when
immersed in a medium whose temperature rises Hnearly is prac-
tically that of the bulb alone, the value of k and the lag of the
capillary not entering the result to an appreciable extent. While
the equilibrium condition being established, however, their sig-
is

nificance is not to be overlooked. From (24), (25), (26), neglect-

6 —u
ing small terms, it is seen that the ratio ^ ^r is diminished with

_1 _1
time at a rate dependent on ( i — ^) e -^b* + ke ^cK This exact func-
tion was carefully examined bath maintained
for the case of a
at constant temperature, (p. 692) and the computations need not
,

be repeated. About three minutes after immersion this ther-

mometer would differ from the steady state by lO"^ times the
initial temperature difference.
(c) Logarithmic Temperature Change, u = A-\-Be~''K —^The
occurrence of this case has been explained in the footnote to p. 664,
and the detailed development of pages 666-7 permits us to dismiss
the form with a bare statement of the solution. Placing the value
^u
of -y, namely —aBe'""^ in equation (16), and dropping the terms

containing an exponential, in accordance with the discussion fol-

lowing equation (7) , we get for the solution in the steady state,

ie-u)^J^k^^^-,i^-k)^^iu-A) (29)

III. LAG OF ELECTRICAL THERMOMETERS

Electrical thermometers in common use fall into one of two
classes, thermoelectric or resistance. Some form of galvanometer
is necessary as an indicator for either, and the lag of this galva-
nometer is to be added to the lag exhibited by the thermocouple
or resistance coil in acquiring the temperature under measurement.
It will often be found that the galvanometer lag is the greater
portion of the whole; in fact, that frequently it is the only portion
which need be considered at all.
Harper] Thermometrtc Lag 695
GALVANOMETER LAG
An expression for the lag of a galvanometer is easily derived
from the familiar equations governing the behavior of the instru-
ment, whence it is quite unnecessary to treat the subject in
detail, but the general method of deducing the required expression
may well be summarized for the most common case, that of a
D'Arsonval instrument, under the condition of critical damping.
The fimdamental equation for the motion of the coil

^J^ + % + ^^ = ^ (30)

iV = applied moment (in general, a function of t)

6 = displacement
t =time
/ = moment of inertia of moving system
jK" =^ damping coefficient
T = elastic coefficient
has three different solutions according as K^ is greater than,
equal to, or less than 4/7, leading respectively to the equation
of motion if overdamped, critically damped, or underdamped.
Critical damping occiu-s when K^ = 4/T, and the solution ^^ is

e = A€ ^I'
+ BU 2/'
+€ 2/i t je^i'dt- jte^i'dt (31)

For nearly all thermometric work, the function is linear. A N

Wheatstone bridge or potentiometer is approximately balanced,
and the unbalanced emf causes a deflection of the galvanometer
according to the equation just written. This unbalanced emf
changes in direct ratio to the temperature change, and may be
taken as linear over the range of any one reading.
Let the moment Nq be impressed on the galvanometer coil by
closing the circuit at time zero, and let this moment decrease at

21 See, for instance, A. R. Forsyth: A Treatise on Differential Equations—" Linear equation with con-
stant coefficients—Case of 'equal roots,' " p. 64 (3d. ed., 1903). Notes on particular integral in the sec-
tion on the General Linear Equation, p. 98, or "Method of variation of parameters," p. 110.
 .

696 Bulletin of the Bureau of Standards [Voi. 8

constant rate r (passing through nil and increasing in opposite

direction). The function A^ will be

N = N,-rt (32)

TV
The moment is zero at the instant t = ~^y which is accordingly the
r
time of the temperature corresponding to exact balance of the
bridge or [Link] the deflection, 6, of the galvanometer

be not zero until X seconds later, or at time —

jV
^ + X, it is evident
r
that an error, due to lag,is made in the usual manner of reading

such instruments when measuring changing temperatures.

Omitting all the steps of substitution of (Nq — rt) for A^ in the
equation (31) and of determination of A and B for the initial
conditions, stated below,

O-'^k-rt + Hr (§^o + ^) + ^o + ^r]) (33)

if, at time zero, moment N^ be suddenly impressed on the coil by

closing the circuit when the coil is at rest f -77 =o j in its equilib-

rium position (^]o

= 0)
For most galvanometers the value of — when
=: critically damped
K
will be found to be such that in a few seconds the term e 2/* is
very small with respect to the other terms, and the factor multi-
plying it is not large, so the equation simplifies to

When the deflection, 0, is zero,

'4-.^ to)

The lag, X, of the galvanometer behind an emf changing linearly

is accordingly, after a short time, ^ seconds (see eq. (32) ). As

Harper] Thermometric Lag 697

an illustration of the magnitude of this quantity, a galvanometer

much used by the author has the value ==0.4 second, whence

its lag is 1.6 seconds. The must be closed at least 5 sec-

circuit
onds before taking a reading, an accuracy of o.i per cent of the
if

deflection is desired, as shown by the following computation,

which gives the time that must elapse before the term in €-27
becomes [Link].
T K
If ^ = 0.4 second, e~27' = e"^'^^^ = (about) 0.00 1 for ^ =5 seconds.

For the determination of — for any particular galvanometer,

many methods might be divised from the common equations dis-

cussed in the numerous papers on galvanometers. One may be
outlined here. The free period, T, of a system of moment of
inertia / and elastic coefficient t is 27r^ / -, from which relation it

follows that
47r2/

The condition imposed by critical damping is that

K2 = 4/t
from which
i67r^7^
T\
K _4.7r

The free period of a galvanometer system is very little different

from that in which it vibrates under any conditions not closely
those of critical damping, so that the period observed when swing-
ing as little damped as possible, will usually suffice to give T with

high accuracy, -j is thus given very directly; the lag of the

galvanometer, being ^, is of the extremely simple form — seconds.

698 Bulletin of the Bureau of Standards {Voi. 8-

RESISTANCE THERMOMETERS
The resistance thermometers studied were found to be either
very fast or very slow in comparison with mercurial thermometers.
The well-known type made after the design of Callendar, consist-
ing of a platinum coil wound on a mica frame and inclosed in a
glass, quartz, or porcelain tube, is quite slow. Immersed in well
stirred water, the values of X measured usually lay between 15
and 30 seconds, though even this value was exceeded. A de-
parture from the straight-line plot by the logarithmic method
described on page 667 was evident when the temperature difference
r}0 T
was small, indicating that the equation -j-.^^ {u — d)is only a first
at A,

approximation to the statement of the behavior of such thermom-

eters. The deviation was quite marked in some instances and
always in the direction and of the general curvature exhibited by
the Beckmann thermometers discussed in an earlier section. It
appears that the two-term formulae there developed are better
equations to apply empirically to a Callendar type resistance
thermometer than the simpler equations. This may perhaps be
explained by the fact that the temperature of the platinum coil
is partly determined by that of the inner surface of the containing
tube, for which the lag is relatively small, and partly by that of
the support, for which the lag is relatively large. The conditions,
therefore, resemble those considered in the section on Beckmann
thermometers.
The very fast resistance thermometers were of the type in
general use in this Bureau, in the range o-ioo°C, an improved
form of the instrument described in this Bulletin in 1907 by
Dickinson and Mueller,^^ and will be more fully described in a
future paper. The essential features respecting the lag of these
instruments are the small heat capacity of the enveloping sheath
and the intimate thermal contact between this and the resistance
coil. Attempts to measure the lag (in liquids) gave no results,
merely indicating it to be smaller than the method would admit
of determining, namely, considerably smaller than the galva-
nometer lag which was about one and one-half seconds.
*2 Calorimetric Resistance Thermometers and Transition Temperature of Sodium Sulphate: This Bul-
letin, 3, p. 641, Reprint No. 68.
fjarper] Thermometrtc Lag 699
JAEGER-STEINWEHR METHOD OF COMPUTING THE LAG OF A
RESISTANCE THERMOMETER

A method of computing the lag of such thermometers was

proposed by Jaeger and Van Steinwehr, but in the form in which
they pubHshed ^^ it, only a lower limit is placed upon the value
of the lag. The method depends upon measuring the heating of
the coil by different intensities of current.
Writing the equation governing the transfer of heat between the
coil and the medium in which the thermometer is immersed as

P{iu-0) (X)

^ = temperature of coil
u = temperature of medium
X=lag (in seconds)
the rate at which heat is transferred will be

M = heat capacity of the system cooling (or warming).

If 6 be greater than u, the thermometer coil will lose heat and
in time dt will lose a quantity dH

dH = ^-{e-u)dt (36)
A.

If there be any electric current, i, in the coil, heat will be gen-

erated at the rate Rt^, where R is the resistance. This will tend
to raise the temperature of the coil above that of the medium in
which the thermometer is immersed. It will rise until the dissi-
pation, which is proportional to {0 — u), equation (36) equals the ,

generation. This heat generated in time dt, (J being number of

joules in a calorie) is

23 Jaeger and Von Steinwehr:

2^
Zeitschrift Instntmentenkunde, 26, p.
fiir 241; 1906.
700 Bulletin of the Bureau of Standards [Voi. 8

and so in the equilibrium state of the thermometer

(37)
l^'-^^-Sj^
Direct measurement of {6 — u) is rather difficult, if at all possible,
but indirect determination of the value by employing two or
more values of i is quite easy. Applying (37) to such a series,
u being kept constant

Ue,-u)

etc.

The difference in does not change R appreciably for the second

members of the two equations, whence subtracting

The value of M
proposed by Jaeger and Von Steinwehr is the
water equivalent of the platinum resistance coil, computed from
dimensions, density, and specific heat. Obviously, this gives
merely a minimum value. The water equivalent of a portion of
the silk and shellac wrapping about the wire of their thermometer
^^
should have been included. This being doubtless several times
24 When a cylindrical heat source is surrounded by an annular covering (inner radius a and outer radius b),
whose outer surface is maintained at a definite constant temperature (0), the equilibrium distribution of
temperatures is expressed by

log r
V=C ?

Integrating through the cylindrical shell, to determine the position of the boimdary for considering the
heat transfer as between two bodies, one the core and a portion of the annular covering, the other the
 .

Harper] Therviometric Lag 701

that of the wire alone, the value of X as 1/33 of a second, obtained

in the computation, is many times too small. Nevertheless, the
lag of such a form of thermometer is small compared with that
of any ordinary galvanometer.
The thermometers of the Dickinson-Mueller type (improved
form) possess the following constants Heat capacity of platinum :

coil, [Link] cal. per degree C. Heat capacity of mica in region of

coil, 0.16 cal. per degree C.
The mica between the head of the thermometer and the resist-
ance coil, supporting and insulating the leads (with a heat capac-
ity of 0.30) and the platinum sheath (with a heat capacity of 0.60)
can not be supposed to be heated by the coil appreciably above
the temperatirre of any liquid in which the thermometer is
immersed, and play no part in the computation tmder such a
condition. This is equivalent to the statement that the whole
temperature drop between the coil and the bath is to be foimd in
the mica and air spaces separating these two. The value 0.17 for
M places a safe upper limit, and the value 0.0 1 is the certain mini-
mum for such a form of thermometer.
The factor

i J^f (^ee eq. 38)

has been foimd, for the thermometer investigated when immersed

in well-stirred water, to be 8.0 (

——— - j r-^
j,
giving to X, for the

values of M just stated, the limits and 0.08 seconds.

1.4 seconds
For a close winding, such that the middle mica plate is quite
inclosed, the water equivalent should include all of this plate and
one-half of each of the outer plates, separating coil from sheath.
This is, in all, two-thirds of the mica; the corresponding value of

outside medium and [Link] of covering, this boimdary is located so that a fraction of the covering

belongs to the core.

WiTO'-i
If 6= loa, as in the case of a [Link] wire covered to a total diameter of i mm, the fraction evaluates to
0.20; and for 6= sa, 0.27 of the total heat capacity of such a covering adds to that of the core. (In either case
it should be borne in mind that the heat capacity of the whole covering is many times (99 or 24) that of the

core if volume specific heats be the same, i. e., the total heat capacity is 20 times, or 7 times the heat capacity
of the core.)
 —

702 Bulletin of the Bureau of Standards [Vol. 8

M is O.I I and X would be 0.9 second. However, the thermometers

investigated were of very ''open" winding (see Fig. 11), pitch 0.7
mm. and wire o.i mm. diameter. The mica to be included with
the wire, roughly estimated, is that inclosed in the dotted circles,
about one-third the total. The most probable value of X in a
well-stirred liquid is thus less than one-
half second.

"^
THERMOELECTRIC THERMOMETERS

The lag of a thermocouple in acquiring

the temperature of a medium in v^^hich
it is immersed is, like the lag of a re-
sistance thermometer, principally a ques-
tion of the form of
mounting. Prob-
ably there are almost as many forms in
use as there are makers of thermocouples,
for there seems to be plenty of latitude for
variation in this respect without impair-
ing the usefulness of the finished instru-
ment. Consequently it was deemed im-
Fig. 11. Section through re- important to test any particular forms
sistance thermometer of of thermoelectric thermometer for lag.
the Dickinson-Mueller type
Dependent on the moimting of the junc-
{greatly magnified)
tion one would no doubt find lags
ranging from a small fraction of a second to perhaps 30 sec-
onds for immersion in a well-stirred water bath. The impor-
tant point is that for all v/ork in that part of the temperature
scale where high precision is attainable, so that lag corrections
might be appreciable, it is possible to design a thermocouple with
a lag as small or smaller than that of an ordinary galvanometer
and generally, if not always, quite negligible.

IV. THERMOMETRIC LAG IN CALORIMETRY

In view of the different conclusions that have been reached by
authors ^^ v/ho
have considered the effect of thermometric lag on
2^ Jaeger and von Steinwehr:
Verh. Deut. Phys. Gesells., 5, p. 353; 1903. Richards, Henderson, and
Forbes: Proc. Amer. Acad., 41, p. i; 1905; or Zs. fiir Pliys. Chem., 52, p. 551; 1905. Jaegeraad von Stein-
wehr: Zs. fiir Phys Chem., 54, p. 428; 1906. WTiite: Physical Review, 27, p. 526; looS,
 :

Harper] Tkermometric Lag 703

calorimetric measurements, it was deemed of sufficient importance

to consider the question in some detail. The conclusions reached

have been arrived at, by a different analysis, by W. P. White.^*'
Classic procedure is to divide the ordinary calorimetric experi-
ment into three parts, designated, respectively, as preperiod,
middle period, and after period. A precise measurement of the
calorimeter temperature, before adding the supply of heat whose
determination constitutes the object of the experiment, defines the
instant which separates the first two periods, and similarly a pre-
cise temperature measurement after the heat is added marks the
dividing line between the second and third periods. Considera-
tions of lag might therefore be said to pertain wholly to the middle
period, but it is more convenient to treat the first temperature
mentioned as the close of the preperiod rather than the beginning
of the middle one, and to assign the second temperatmre similarly
to the afterperiod. In this way the investigation is split up into
three parts. Two of these concern the error, due to lag, in deter-
mining the temperatmres mentioned, and are treated by investi-
gating the conditions pertaining to the exchange of heat between
the calorimeter, the thermometer, and the jacket, in the steady
state, during preperiod and afterperiod. The third part involves
the lag errors in the temperatiu-e readings used in the computation
of the cooling correction, or is a middle-period function.
Numerous methods, differing radically in many ways, have been
devised for performing calorimetric computations, but the funda-
mental relations underlying the " cooling correction " are the same,
by whatever method the details be accomplished. It is generally
assumed that the calorimetric aggregate exchanges heat with its
envelope according to the law commonly called Newton's law of
cooling

where u is the calorimeter temperatmre at the instant t and U the

jacket temperature. In the simplest case the latter is constant
throughout the three periods (f/ = A). a, the cooling constant,
26 Wiiite: Physical Review, 31, p. 562; 1910.

73764°— 13 7
704 Bulletin of the Bureau of Standards [Voi. 8

and A are assumed possible of determination in the course of the run

and will be discussed later.
After assembling the calorimeter any necessary time may be
allowed to elapse for the effect of the initial conditions to be
obliterated, so that during the preperiod the calorimeter tempera-
ture approaches that of the jacket according to a curve really
logarithmic but of ctirvature so small it may frequently be con-
sidered linear. At a given instant, t, is commenced a supply of
heat to the calorimeter and the middle period begins. If this
supply could be distributed instantaneously, its measurement
could be accomplished with no further process than the exact
determination of the temperatures just before and just after the
addition of the heat. No matter how long the middle period, the
same result would ensue if the calorimeter could be perfectly
insulated from external sources or sinks. In taking accoiint of
the loss or gain of heat in the middle period it is quite customary

to compute, not a quantity of heat, but rather a temperature cor-

rection, to apply to the observed difference. The result is thereby
stated in terms of the perfectly insulated calorimeter of the same
heat capacity.
After the supply of heat to the calorimeter ceases, the tempera-
tvne of the latter tends to a steady state of approaching that of
the envelope at an almost constant rate. At any time after this
state is attained, the precise measurement of the temperature is
made, of which the time, tz, marks the close of the middle period
and beginning of the afterperiod.
After proper instrument corrections are applied, the thermome-
ter readings give 6^ and 6^, which in turn, by appropriate lag
corrections would give the observed rise of temperature, U2 — u^.
The additional increment of temperature resulting from the " cool-
ing correction" is given by applying

du= — a{u^A)dt
to each instant of the middle period, t^ to t^', whence the total " cor-
rection" is

K^^-aV\u- A)dt
 .

Harper] Tkermometric Lag 705

The equivalent rise of temperattire, as if there were no heat losses

during the middle period, is accordingly

u^ — u^ + oc (u — A)dt
I

lOr which it is customary to substitute a similar formula contain-

ing thermometer readings (corrected for instrument errors) on the
hypothesis that they represent actual temperatinres without lag.
If any error due to lag occur, let it be designated e

e==[0,-e, + a' r\0 - A')dt] -[u,-u, + a [''{u - A)dt] (39)

and on the assumption that a' = a, A' = A. (See p. 706)

e=(d,- u,) - {6, -u,)+cx r\e -u)dt (40)

If the thermometer used in the middle period obey the equation

(i), ^t/^tC^""^) (P- 661), equation (40) is very easily reduced. If

it do not, as, for instance, when a Beckmann thermometer is em-

ployed, the mathematical manipulation is more complex. This
case will be discussed after taking up the one first mentioned.
From (i) we obtain directly.

{e-u)dt= -\~dt=-\(e,-e,)

so that equation (40) has the form

e^{0,- u,) - {0, - u,) - a\{6, - 6,) (41)

The values 0^ and 6^ must be obtained in the preperiod and

afterperiod in the steady state of heat exchange between the
calorimeter, its surrounding jacket and the thermometer, for which
condition the relations are given ^^ by equation (8) as explained
in the derivation of that equation, case (c) page 666.
• 2' This relation is also derived by W. P. White, loc. cit.
 :

706 Bulletin of the Bureau of Standards [Voi. 8

Applying equation (8) to the conditions occurring at /j and t^.

I — n'gA.g

between which equations the jacket temperature A is to be elimi-

nated. It is almost universal to employ the same thermometer
in preperiod and afterperiod so that ^2="'^i- I^et us assume that

the heat capacity of the calorimeter remains unaltered, as, for

instance, by adding some substance in the course of a method of
mixtures experiment; then 0^2 = ^1 = ^- ^^^ elimination of A gives

(i-a\){0,-0,) = {u,-U,) (42)

But (41 ) is (i — a\) (^2 — ^1) = (^2 -u^) +e

Therefore a{\-X){0,-e,) :=e (43)

The due to thermometric lag is thus proportional to the

error
difference of lag of the two thermometers used in the end periods
and the middle period, vanishing if the same thermometer be
employed in both.
The discussion is not quite complete without a consideration
of the assumption a^ = a and A' = A made at the time of defining
£ and passed over almost without comment (p. 707). Examina-
tion of the magnitudes involved makes it quite evident that the
assumption is always true to the extent that deviations from it are
numerically insignificant for the highest precision work yet accom-
plished. For the Regnault-Pfaundler ^^ and allied methods of
computation it may be shown to hold to the extreme limit of
being an exact statement of the relations involved. The proof
follows
From the primary equation

'8 Sytiopsis of this method may be found in Berthelot "Traite Pratique de Calorimetrie Chimique,"
aded., p. 117.
Harper] Thermometric Lag 707

applied to the instants t^ and ^2, may be obtained two equations

containing a and A , which may be solved for either. Solving,

_ dul dir\

u,-n.
whereas we use a value computed from a similar expression con-
necting thermometer readings,

«f' =
i dt\.
0,-0,
Now by equation (8)

dO_/ a\ \du I du
dt \ I —oK/dt I — a\ dt
(44)

^1 _^1 I (du\ _du'\ \

dt Ji dtj2~ I — <arX\ dt Ji dt _ 2/

and by equation (42)

0,-0,
I — ti'X {u,-u,)

from which it is evident that «f' is identically a. Solving for A ,

for instance in the form,

I du
A=u,+ a dt

an exactly similar process to the above shows that A ' is identically

a diM

byequation(i)
="'-'^S]i+ai]i

*'Mi+^ -^ by equation (44)

Jj

=A Q. E. D.
7o8 Bulletin of the Bureau of Standards [Voi.s

When a thermometer obeying the law -z- = ~-{u — 6) is employed in

at \
a calorimetric experiment, no account is to he taken of lag, if the
same thermometer he used for preperiod, middle period, o/rid after-
period. The important restriction on the generahty of this
proposition that the upper temperature measurement be post-
poned until the steady state of heat exchanges is well estabUshed,
need hardly be included in the italicized conclusion because no
measurement taken before the restriction was complied with
would satisfy the more fundamental requirements of the calori-
metric determination.
A thermometer which does not obey the law above stated must
be examined by a process similiar to that outlined in the preceding
pages. The only type we will consider here is the Beckmann,
general equations for which have been deduced in Section II.
The conclusions of the investigation show also that not merely
is any error due to lag quite inappreciable for ordinary conditions,

but that it is mathematically zero as for the case above. The

expression for the error is a little tedious of derivation, although
not so very difficult.
Tet us start from equation (40) which is an entirely general
expression for e, the total error due to thermometric lag, with
no limitation as to the form of thermometer, except in so far
a^ might be said to lie in the assum^ption that a' and A ' computed
from, lagging thermometer readings may be used interchangeably
with a (the real cooling constant of calorimeter) and A (real jacket
or "convergence" temperature), the values vv^hich would be
obtained with a lagless thermometer. This assumption, briefly
discussed before, is quite evidently true for a Beckmann ther-
mometer.
Equation (40) is £ = {Q^ — '^2) — (^1 — u^-\-a f '{Q - u)dt, and the

difficulty in reducing it, when a Beckmami thermometer is em-

ployed, is the determination of the value of the last term for a

thermometer obeying the lavv^ expressed by the relation (15).

(The explanation of which quite uninterpretable equation, and
the notation employed for this type of instrument, are to be found
in the paragraph on p. 683, containing equations (12) and (13).)
Harper] Tkermometric Lag yog

The expression for (u — 0) developed in the general treatment

contained in Section II (equation (i6)), imfortunately contains
u under an knowledge of the function
integral sign, presupposing a
before further reduction of the relation. It is very desirable to
make this investigation entirely independent of the form of the
curve which the calorimeter temperature rise follows, so that this
expression can not be employed as there written. Tracing it back
to its genesis we are obliged to use equations (12), (13), and (14)
as follows*
= kC + (i-k) B (14)
=
u ku + {i—k) u (Identity)

B-u=-K-^ (12)

from which

and
^— K-^^f)+(-<-^^f)
= -a[kXe(C,-C,) +{l-k)\B{B,-B,)]
and equation (40) , with the integration performed, becomes

^=^{02-U2)-{0,-u,) -a[k\c{C,-C,) +(i -k)\B(B,-B,)] (45)

The Bs and Cs can not be united to 6s on account of the factors

Kc and \b multiplying them, so the 6s must be split into parts to
permit of collecting terms. This is accomplished by means of
(14) and the identity just beneath it, a few lines above. Applying
to the instants t^ and /j,

6^— u^ = k{C^ — u^ + (i —k) {B^ — Ui)

6^-U2 = k{C2-u^ +(i -k){B2-u^

from which a value for {62 — u^ — {6^ — u^ may be substttuted in

equation (45) giving

s = ^[(^2 - Ci) (i - flrXc) - (^2 - "^^i)]

+ (i -k)[{B,-B,){i-'a\B)-{u,-u,)] (46)
7IO Bulletin of the Bureau of Standards [Voi.s

We turn now to the equilibrium state in the preperiod and after-

period and will pass by the case where thermometers of different
lag are employed in these and the middle period. The relations
of C and B to u are given by equations (12) and (13).

and u is determined by the relation

du , ..

SO that from the discussion of case (c) in Section I (p. 666) it may
be seen that equation (8) states the relation of 5 or C to ^^ when the
steady state obtains.

B-u= '^\ C-u = -^^{u-A)

I — a\B (u-A) — I a\c
(8)

from which, applied at instants t^ and /j come the relations

(i — aXs) {B2 — B^ =U2 — U^

(i - a\;) (C2 - Ci) = U2 - Wi

and equation (46) is at once

No error is made by employing thermometer readings,

therefore
uncorrected for lag, throughout the computation when a Beck-
mann instrument of the usual form is used. The restriction men-
tioned before that this conclusion is true only when both 6^ and
62 are measured in the steady state of heat exchanges requires
more emphasis than in the case of an ordinary thermometer,
because the Beckmann type is so much slower to reach this state,
as shown on pages 692-694.
When thermometers with different lag, whether Beckmann
or "chemical" type 'or electrical, are employed for the middle
period and the end periods, when the heat capacity of the calorime-
ter is greatly changed in the course of a run, or when the adia-
batic method of calorimetrv is used, with thermometers of differ-
 .

Harper] Tkermometrtc Lag 711

ent lag in the calorimeter and its jacket, the corrections because of
lag may be appreciable and may be computed from the most con-
venient of the foregoing equations.

V. LAG CORRECTIONS IN APPLIED THERMOMETRY

Statements and conclusions relating to thermometric lag of
importance in applied thermometry, are here summarized for
convenience. As has been previously suggested, a determination
of X to an accuracy of only 50 or 100 per cent is quite often sufficient,
as lag corrections are usually exceedingly small.

RESUME OF PRACTICAL INSTRUCTIONS

1 . Constant Temperature. —In
a bath at constant temperature
no correction for lag is made
if the thermometer reading be
to be
taken a sufficient interval of time after introducing the instru-
ment into the bath. A convenient interval to remember is loX
seconds, in which time the initial difference of temperature is
reduced to e-^^( =0.00004) times itself. For all ordinary chemical
thermometers in liquid baths stirred even slowly, loX will be rather
less than a minute, and decrease from this value if the stirring be
increased. In a gas where X may have any value between 15
seconds and 10 minutes or more, depending upon' conditions of
stirring and form of thermometer, it may not be convenient to
wait loX seconds for a reading. Computation on the basis of
reduction every X seconds of any temperature difference between
bath and thermometer to a value e-^ times that at the begin-
ning of the period of X seconds will show the interval which must
elapse before the thermometer reading is correct within the
allowable error.
2 Linear Change of Temperature. —
In a bath whose temperature
is a uniform rate of r units per second, the correction to a
rising at
thermometer reading at any time, to get the bath temperature at
the same instant is -1- rX imits. As a concrete example of the
largest error likely to occur were lag neglected, may be reviewed the
numbers in an intercomparison of a 3-second and a 15-second ther-
mometer in a comparison bath rising at about the most rapid rate
with which it would be practicable to obtain readings reliable to
single thousandths, about o?o3 per minute. One thermometer
 .

712 Bulletin of the Bureau of Standards {Voi.s

would lag behind thebath by (0.0005 X 3) degrees, the other by

(o?oco5Xi5), making a difference of o?oo6 in their indications
due to lag. So rapid a rate of rise when reading single thousandths
and so great a difference in the values of X for two thermometers
in the same bath is of quite infrequent occurrence, whence it is
not often that a lag correction will be foimd necessary.
3. General Case of Changing Temperature, —
In a bath whose
temperature is changing according to any more complicated law,
but so as to be a continuous single valued ronction of time, the lag
correction to any reading of a thermometer immersed in it is foimd
by reading temperatures at sufhciently short inter^^als to plot the

function 6 with time so as to have the value -^ at any point.

The fundam^ental equation (i) (p. 661) then gives u = 0-^\^ the

lag correction bemg H-'^sri-

4. Calorimetry. —No correction for thermometric lag is to be

made if all of the following conditions hold:
(i) The thermometer be one whose behavior accords with

(Symbols explained on p. 661), or is a Beckmann instnmient of the

usual form Vvdth a large inclosed capillary just above the bulb.
(2) The same thermometer (or different ones with the same lag,
X) be employed for preperiod, [Link]-period, and middle-period
temperature readings.
(3) The lower and upper temperatures be read after the calo-
rimeter, its jacket, and the thermometer have certainly "attained
the steady state" of heat exchange (usually a close approxima-
tion to linear rise or fall of temperature of c alorimetric aggregate)
"Attaining steady state" must, of course, be interpreted in the
sense used physically for any mathematically asymptotic relation.
There is seldom any doubt about this respecting the lower tem-
perature reading; it is only the upper one concerning which the
observer must be cautioned.
(4) The heat capacity of the calorimeter be the same during the
three periods of the experiment.
Harper] Thermometric Lag 713

VI. SUMMARY
Thermometric lag is conveniently expressed by employing a
1.

quantity X, whose significance may be stated as follows:

(i) If a thermometer has been immersed for a long time in a

bath whose temperature is rising at a uniform rate, X is the number

of seconds between the time when the bath attains any given tem-
perature and the time when the thermometer indicates this tem-
perature. In other words, it is the number of seconds the ther-
mometer *'lags" behind the temperature; or ^
(2) If a thermometer be plunged into a bath maintained at a
constant temperature (the thermometer being initially at a differ-
ent temperature) X is the number of seconds in which a difference
,

between the thermometer reading and bath temperature is

reduced to i/e of its initial value.
The fundamental equation of heat transfer, commonly referred
to as Newton's law of cooling, is stated in terms of X, for appHca-
tion to problems in thermometric lag, and the principal working
equations derived therefrom are reviewed.
2. To express analytically the lag of the common form of Beck-

mann thermometer, the simpler theory was modified to take into

account the fact that the lag of the bulb and that of the large
capillary, between the bulb and the fine capillary, are different.
3. Methods of determining lag are discussed and experiments
are cited to test the theories as applied to ordinary ''chemical"
thermometers and to Beckmann thermometers.
4. The large variation in the lag of a given thermometer with
the nature of the medium in which it is immersed, and with the
rate of stirring of this medium, is brought out by experiments in
water, in a viscous kerosene, and which these media
in air, in
were forced past the bulb of the thermometer at different measured
rates.
Values from the curves obtained are:
30 These two interpretations of A are mutually consistent. The definition of the quantity is most
logically made by designating it as the "constant" of the fimdamental equation, and then deducing the
interpretations here given.
714 - Bulletin of the Bureau of Standards [Vol. 8

^ in seconds. Small-Bulb "Chemical" Thermometer

Vel. past bulb in cm/sec. 1 5 10 50 100 500 1000 00

Water 10.0 5.1 3.3 2.9 2.4 2.3

2.2 sec.
Oil 40 to 50 13.4 7.5 6.4 4.8 (any
Air 190 170 148 128 71 58 33 25
medium)

5. Part, and frequently the largest part, of the lag of a thermo-

electric or electrical resistance thermometer is the lag of the gal-
vanometer. A d'Arsonval galvanometer, critically damped, is
-
shown to
^ .
lag 4 X
Moment of
:pr
Dampmg
Inertia of
;
^-
coemcient.
——^
Moving ——
—System
- seconds
,

behind an emf changing linearly with time, after the steady state
of motion is attained. A close approximation to this value is lag =
T/tt seconds, v/here T is the complete period of the moving system,
oscillating much imderdamped.
6. Types of resistance thermometers were tested for lag. The
Callendar type, in a liquid bath, was fotind to lag greatly in com-
parison with an ordinary "chemical" mercurial thermometer, and
the empirical expression of the lag is of the same form as that
developed for a Beckmami thermometer. The Dickinson-Mueller
type of resistance thermometer bulb in a liquid bath was found to
lag much less than the fastest of mercurial thermometers.
7. The Jaeger-Stein wehr method of computing the lag of a

resistance thermometer, from the heating effect of the measuring

current and the heat capacity of the thermometer, is critically
discussed. The lag of a Dickinson-Mueller thermometer in a
well-stirred hquid bath is shown to be about one-half second.
8. The necessary corrections that must be applied to the

observed readings of a thermometer to correct for the effect of

its lag under the usual conditions of use are discussed in some

detail.

9. An given of the fact that in an ordinary

analytic proof is

calorimetric experiment, in which the same thermometer is used

to determine temperatures in the "preperiod," the "middle
period," and the " afterperiod," the correction for lag in the middle
period neutralizes the corrections for lag in the preperiod and the
afterperiod.
Washington, March 5, 191 2.