Hogar Buscar Colecciones Diarios Sobre Contáctenos Mi IOPscience

Un aparato para medir la conductividad térmica de líquidos

Este artículo ha sido descargado de IOPscience. Por favor, desplácese hacia abajo para ver el artículo de texto
completo.

1976 J. Phys. E: Sci. Instrum. 9 1073

(http://iopscience.iop.org/0022-3735/9/12/020)

Vea la tabla de contenidos de este número, o vaya a la página de inicio de la revista para obtener más

Download details:
Dirección IP: 130.102.42.98
El artículo fue descargado el 20/04/2013 a las 10:48

Tenga en cuenta que se aplican los términos y condiciones.

 sugiere fuertemente que para los líquidos muchas de las
Un aparato n para medir mediciones anteriores deben estar en error. Por

la conductividad térmica
de líquidos

C A Nieto de Castrot, J C G Caladot, W A Wakehami

and M Dix:
t Laboratorio de Quimica Estrutural, Complex0
Interdisciplinar, IST, Lisbon, Portugal
$ Departamento de Ingeniería Química y Tecnología
Química, Imperial College, Londres SW7 2BY, Reino
Unido

Recibido el 3 de mayo de 1976, en forma definitiva el 24 de

junio de 1976

El artículo describe un aparato para la medición de la

conductividad térmica de líquidos no conductores bajo su
presión de vapor de saturación. El instrumento, que se basa en
el principio de alambre caliente transitorio, ha sido diseñado
de manera que el elemento de medición se ajuste lo más cerca
posible a una fuente de línea infinita de calor en un fluido
infinito. En estas condiciones, la conductividad térmica del
líquido puede determinarse a partir de la pendiente de un
gráfico del aumento de temperatura del elemento de
calentamiento contra el logaritmo del tiempo. El sistema de
medición ha sido diseñado para proporcionar hasta 60 puntos
en esta parcela para cualquier estado termodinámico
particular del fluido bajo investigación. La reproducibilidad
del instrumento es del orden del 0,03 % y la precisión de
las mediciones se estima en un 0,1 %. Debido a la falta de
una teoría adecuada para los efectos de la transferencia de
calor radiativa, la precisión de los valores de conductividad
térmica no puede definirse inequívocamente, Los resultados
preliminares se presentan para el n-heptano a tres
temperaturas en el rango de 20 a 30°C.

1 Introducción
Históricamente, la conductividad térmica de los fluidos ha
demostrado ser una propiedad de transporte extremadamente
difícil de medir con precisión. En el caso de los gases Kestin
eta1 (1972) y de Grooteta1(1974) han demostrado de forma
concluyente que casi todas las primeras mediciones de la
conductividad térmica a bajas densidades han sido objeto de
grandes errores. Esta conclusión se ha basado en los resultados
de la teoría de Chapman-Enskog de mezclas de gases diluidos,
que en el caso de los gases monoatómicos arroja una relación
precisa y bien definida entre el gas viscosityof a y su
conductividad térmica (Kestin et a11972). Para los líquidos
no existe tal relación termodinámicamente consistente, por lo
que las declaraciones con respecto a la precisión de los datos
de conductividad térmica deben basarse en la inferencia. Por
lo tanto, la disparidad entre los resultados de los trabajadores
independientes para sistemas específicos, que en general es
mayor que la supuesta incertidumbre mutua de los datos,
 ejemplo, los datos disponibles para n-CsH10 muestran un
diferencial del k6 % aproximadamente un valor medio,
mientras que para n-ClsH3s el diferencial es del *2 %.
En este artículo describimos el diseño y funcionamiento de
un instrumento que elimina las principales fallas asociadas con
mediciones anteriores de conductividades térmicas líquidas, y
permite su medición precisa cerca de la temperatura ambiente
bajo su presión de vapor de saturación. La dificultad
subyacente asociada con la medición de la conductividad
térmica de los fluidos es el inicio de la convección libre en el
fluido cuando está sujeto a gradientes de temperatura en el
campo gravitatorio de la tierra. Este problema es más agudo en
los métodos de estado estacionario y solo se ha superado
ocasionalmente (Michels y Botzen 1952, Michels et a11962).
Incluso en algunas técnicas transitorias de la medida el inicio
eventual de la convección se ha observado (Pittman 1968) y
pudo haber viciado los resultados. Por lo tanto, aunque es
evidente que una técnica de medición transitoria proporciona
el mejor esquema posible para evitar el problema de
convección, debe prestarse la debida consideración al diseño
del aparato con respecto a este punto. Además, las
consecuencias experimentales y teóricas de esta elección de
técnica deben analizarse cuidadosamente, y el análisis
utilizado para optimizar el diseño del instrumento.
La teoría más reciente del esquema de medición de alambre
caliente transitorio fue originada por Haarman (1969, 1971,
1973) y posteriormente ha sido refinada y extendida por Healy
et a1 (1976). La aplicación de esta teoría al diseño de
instrumentos y a la medición experimental se ha limitado hasta
ahora a las mediciones de fase gaseosa (Haarman 1969, de Groot
et a11974), pero la mayor parte del análisis puede trasladarse
intacta al caso de líquidos no conductores. La única
complicación adicional introducida en la teoría del instrumento
surge debido a las propiedades de absorción del medio líquido
para la radiación (Pittman 1968, McLaughlin y Pittman 1971).
Sin embargo, este efecto solo conduce a una pequeña
corrección de las conductividades térmicas observadas y, por
lo tanto, puede ignorarse para el presente propósito del diseño
del instrumento y reconsiderarse cuando se presenten los
resultados definitivos.
La disponibilidad de una considerable literatura sobre el
tema general del método de alambre caliente transitorio nos
permite omitir aquí una exposición detallada de la teoría y
concentrarse en su lugar en su aplicación a las mediciones en
la fase líquida.
2 Principio de medición
En la técnica de alambre caliente transitorio un alambre de
platino delgado y recto sumergido en un fluido isotrópico e
inicialmente en equilibrio con él, se somete en el tiempo t =0
a un cambio de paso en el voltaje aplicado a él. Después de
este cambio, el cable aparece como una fuente de línea de
calor con una magnitud constante, q, por unidad de longitud, y
la evolución del tiempo de la temperatura del cable está
determinada por la conductividad térmica del fluido.
En la situación ideal de un alambre infinitamente delgado y
largo inmerso en, e inicialmente en equilibrio con, un fluido de
extensión infinita a una temperatura BO, el campo de temperatura
alrededor del cable después de la iniciación de una disipación de
calor constante en él en el tiempo cero es dado por (Carslaw y
Jaeger 1959)
hT(r, t)=B(r:t)-Be,=(-qi47t.h)Ei(-r2/4~t). (1)
Aquí, es la conductividad térmica del fluido, se supone que
es independiente de la temperatura, y es la coordenada
radial medida desde el centro del cable. Además la difusividad
térmica, K , del medio se da como
donde p, la densidad del fluido yCp,el calor específico de
presión constante, también se supone que es independiente de
la temperatura.
1073
C A Nieto de Castvo, J C G Calado, W A Wakehanz and M Dix
Para un alambre de radio se puede obtener la temperatura RI, R2and R1.iare década cajas de resistencia (Cropico tipo
del alambre estableciendo r= aand ecuación de RBB5) mientras que el resto de las resistencias del puente son
expansión(1)como resistencias de película metálica ultraestable (Welwyn Electric
Company, tipo Vishay) montado en un entorno de temperatura
hT(a, t )=(-q/4nh)In(4~t/a2C) controlada. Las resistencias R4, 113,Rs, RIOson todas iguales y
para valores suficientemente pequeños dea2/4~t . Aquí, la su valor será designado posteriormente por R.Los valores
constante de Cis Euler. Así la ecuación(3) proporciona la numéricos de todas las resistencias se recogen en la tabla 1para
relación más simple posible entre el aumento de temperatura del facilitar la referencia. La
cable y el tiempo, y permite la evaluación de la conductividad
térmica siempre que esta relación se pueda medir
Cuadro 1 Resistencias de los elementos en la figure 1
experimentalmente.
Aunque cualquier realización experimental de este principio
debe inevitablemente apartarse de la situación ideal descrita Resistencia Resistencia (a) Tolerancia (I;)
anteriormente, de Groot eta1 (1974) y Healy
eta1(1976)han demostrado que una elección adecuada de RL at 20°C 195.64 k 0.05
condiciones experimentales puede reducir algunas de estas RS at 20°C 58.81 k 0.05
desviaciones al nivel de pequeñas correcciones calculables del RI 0-1000 k 0.05
orden del 1 %, y hacer que otras sean insignificantes < 0,01%. Rn 0-1000 & 0.05
Así, al llegar al diseño del instrumento descrito en la siguiente R3, R4, Rg,Rio 4200 k 0.05
sección hemos realizado un estudio detallado de cada R5 2000 k 0.05
corrección de forma independiente, siguiendo el trabajo de Rs 4000 k 0.05
Healy etaZ(1976),modificando sus expresiones solo para R7 8000 0.05
hacerlas apropiadas para la fase líquida. Por lo tanto, es R8 16 000 & 0.05
innecesario parausto proporcionar una descripción detallada de Rii 106 k 0,05
la búsqueda de un diseño óptimo para el presente aparato. R12 5x 105 k 0.05
Preferimos presentar las formas y magnitudes finales de las Rl3 3.3 x105 -+0.05
correcciones que se aplicarán cuando se utilice la ecuación(3) k 0.05
R14 2.5 x105
para interpretar las mediciones con el aparato actual, R15 2.0 x105 k 0.001
entendiendo que sus magnitudes son un resultado k 0.01
R17 0-1000
deliberado del estudio de diseño realizado.

3 Diseño de instrumentos relés de conmutación 2-s7are relés de lámina de alta velocidad

La medición esencial que debe realizar el instrumento es la
determinación del aumento de temperatura de la fuente de
calor en función del tiempo después del inicio de una
corriente. En el presente trabajo esto se logra haciendo del
elemento calefactor un delgado cable de platino y conectándolo en
un brazo de un puente automático DCWheatstone, que es una variante de
a1 (1974). Los tiempos en los que la
la descrita por Andersonet
resistencia de este cable alcanza varios valores
predeterminados se encuentran mediante el uso de balanzas
secuenciales del puente para detener una serie de
temporizadores electrónicos iniciados en el inicio de la
disipación de calor.
La figura 1 muestra un diagrama esquemático del puente
automático de Wheatstone diseñado para realizar esta
medición. En este circuito RLrepresenta la resistencia de un
cable de platino largo y delgado y la resistencia Rsthe de un
cable que difiere solo de la primera en su longitud. Ambos
cables están encerrados en celdas llenas de líquido que se
describirán más adelante (63.2). Resistencias
(Kemreed M301C), SI es un relé de lámina humedecido por
mercurio (Hamlin MRC-DT), SI es un interruptor de selección
manual de baja resistencia al contacto y S i 1a interruptor manual
de un solo polo. Cambios de polaridad de tensión entre puntos
A andFare detectados por un comparador electrónico de alta
impedancia de entrada que se describirá más adelante
(53.1.1).Por último, el bridgeDCpower es proporcionado por
una unidad de voltaje constante (Hewlett-Packard type61 12A) y
está organizado en el punto Aof el puente está siempre cerca del
potencial de tierra.
La figura 1 muestra el puente en su posición de
restablecimiento antes de la medición. En esta configuración se
dispone que la resistencia del brazo superior derecho del puente
supera ligeramente la del brazo superior izquierdo. Además,
R17se ajusta de manera que R17= RL(O)+Rs(0) para garantizar
que al cambiar el SI de su posición de reposo a X no haya
ningún cambio significativo en la carga de la fuente de
alimentación. RL(O), Rs(0) son los

R,b I ~ T
resistencias de equilibrio de los dos cables de platino. La
operación de conmutación de ITO X inicia un ciclo de medición
TO
logrc circuit de la siguiente manera. El pulso generado a través de Sec y C por
la corriente que fluye en los dos cables de platino se utiliza para
abrir puertas entre seis contadores electrónicos y un reloj de
cristal controlado, por lo que la secuencia de tiempo se
I t inicia. El flujo de corriente en RL and Rsconduce a la
disipación de calor en estos cables y su temperatura y (en
consecuencia) aumento de la resistencia. Debido a la configuración
15 V inicial del puente, la polaridad de la señal entre los puntos A y F
cambiará en algún momento posterior indicando el paso a través de
un punto de equilibrio. Este evento es detectado por el comparador

I
que causa que se detenga el primer contador electrónico, registrando
así el tiempo de la condición de equilibrio. Posteriormente, el circuito
lógico asociado (83.1.2) hace que se abra el interruptor S7y se
cierre la repetición de R15BY R14en el circuito, aumentando así la
resistencia efectiva del brazo superior derecho del puente. El continuo
III aumento de temperatura de los cables RL and Rsthen causa una

pomporat- Circuito tológico inversión de polaridad adicional entre A y F que detiene la segunda
Figura 1 El puente de medición automático. C: 0.1 pF
contador, registrando el tiempo del segundo punto de
equilibrio. La siguiente resistencia en el brazo paralelo
reemplaza a RI4y el
1074
Un aparato para medir la conductividad térmica de líquidos 3.1 Componentes electrónicos de el instrumento
En esta sección se describen brevemente los distintos módulos
secuencia se repite hasta que los seis contadores se han
electrónicos del instrumento. Fueron diseñados para proporcionar
detenido, la última medición de tiempo se hace con todos los una resolución de tiempo de la inversión de polaridad del puente
interruptores S3-S7 open. El circuito lógico devuelve el ofiz 10,us,to store
sistema a su estado de restablecimiento inicial
automáticamente. Como resultado de esta única ejecución
experimental, por lo tanto, seis veces el equilibrio del puente se
almacenan en contadores y se pueden recordar de forma
independiente a una pantalla visual para fines de grabación.
Con el interruptor S11 en posición (a), la condición para la
inversión de polaridad entre A y F se puede escribir como

Aquí
(4
Ci={1 - [(B+l)/D]l/(l +S/D) bis)
con
(4
B= PR‘IR ter)
D=(2Rll/R+ 1)(1+ R’/R) (i=1 to 6) (44
y
R = R j + RsfR 7 f Ra. (44
Además,8es la fracción de R' incluida en el circuito de
comparación, y por lo tanto puede tomar cinco valores
discretos (figura 1).
Con el interruptor S11 en la posición (b) una condición
correspondiente para la inversión de polaridad entre Aand F se
puede escribir como

Aquí
(5
C2={l - [(B+l)/E] ,'(l + B/ E ) bis)
y
(5
E=(2R1t/R+ 1)(1+ PRiR). ter)
Se ha optado por la disposición experimental de que la
relación RLiRs se aparta insignificantemente de la relación
conocida de las longitudes de los dos alambres l d l l durante el
curso de la medición (54.1.2). Así las ecuaciones (4) y (5)
demuestran que el puente mostrado en la figura 1 permite , en un
ciclo, la medición de la diferencia de resistencia entre los cables
largos y cortos en términos de las resistencias conocidas del resto
de los elementos del puente en seis tiempos registrados desde el
comienzo del proceso de calentamiento. El ciclo de medición
descrito anteriormente puede, por supuesto, repetirse con el
interruptor S11 en posición (a) para todos los valores de P, y
luego de nuevo para todos / 3con el interruptor S11 en posición
(b). De esta manera, siempre que el estado termodinámico del
fluido en las células permanezca inalterado, se pueden obtener
60 mediciones del valor de RL- RS en varios momentos.
Debido a que los dos alambres de platino fueron elegidos
para ser idénticos a excepción de la longitud y están montados
en celdas idénticas, la medida anterior constituye la
determinación de la resistencia de un alambre de longitud
IL- ISbut sin extremos (de Groot etal1974). Por lo tanto,
siempre que se hayan determinado las características de
resistencia-temperatura de los cables, se puede calcular el
aumento de temperatura de este hipotético cable en función del
tiempo. Así, después de tener en cuenta pequeñas correcciones
(94) y calcular la disipación de calor en el cable por unidad de
longitud, q, la conductividad térmica se puede obtener de los
datos con la ayuda de la ecuación (3).
Para evaluar el rendimiento de la red de puentes con
respecto a la medición de una resistencia absoluta, se utilizó en
modo manual para determinar la resistencia estándar 100C2fa
(Tinsley typeS). El valor obtenido difería menos del 0,05 %
del valor calibrado.

the six times resulting from a single run, to allow their
subsequent recovery for recording purposes and to perform all
the switching operations described in the previous sections in
times commensurate with the desired operating characteristics
of the instrument.
3.1.1 The comparator A diagram of the comparator circuit is
given in figure 2. It consists of two input buffer amplifiers
c accidental triggering or resetting, so the ‘run’ button must be
+I5 v -'5 v
Figure 2 The comparator. Diodes, SSD 708; meter,
50-0-50 /-LA
(Analog Devices type 43K) which are connected in a balanced,
cross coupled mode to achieve a very high input impedance ( >
a)
loll and common mode rejection ratio (> 80 dB). The cross
coupling has the advantage that common mode signals are
passed at unity gain whereas differential signals are amplified.
The outputs of the two buffers are connected to a conventional
differential circuit (Ancom type 15A-2a), which is equipped
with a zero set control and a meter to allow offset adjustment.
The complete unit, which is mounted in a temperature-
controlled environment to minimize offset drift, has a band
width of 100 kHz, a gain of 4 x lo6 and a peak-to-peak noise
level of approximately 20 p V .
3.1.2 Control logic This unit controls the automatic sequencing
of the bridge opelation and the gating of the counters; a circuit
diagram is shown in figure 3. Sequencing is accomplished by
the six-stage shift register (SN74174) which is clocked by the
comparator through a 2 m s monostable. The monostable
ensures that only one pulse is passed to the register for each
bridge balance and effectively masks the comparator output
during the bridge resistor switching operation. Two sets of
gates are connected to the shift register outputs. The first set
selects for activation the appropriate buffer transistor from a
group which drives the high-speed reed relays of the bridge.
The second set feeds the corresponding counters from a master
clock (Meon type MClO5-TM). The system is provided with
an interlock to the counter unit in order to minimize the risk of
depressed before ‘reset’ and ‘start’ are operative.
To commence a run, the ‘run’ and ‘reset’ buttons are
depressed. The ‘reset’ button connects the 100 kHz clock to the
shift register clock input and thereby loads all stages to logic
state ‘1’. In this state all counters are gated off and all relays
de-energized. The ‘reset’ command also sets the bistable 1
(which provides a ready signal), enables the trigger monostable
2 and the ‘start’ button and resets all the counters. Depression
of the ‘start’ button enables gate (a) and fires the 30 ms
monostable 3 ~ i thea antibounce circuit 4.Monostable 3 then
energizes the start relay Si via gate (b). The current in the
bridge produces a pulse through the capacitor C (figure 1)
which is fed to the trigger monostable 2 . A 2 ms pulse is then
passed through gate (a) to the clear input of the shift register
1075
C A Nieto de Castvo, J C G Calado, W A Wakeham and 134 Dix
C,C< ioci
Bridge
Tr gger
setting the output of all elements to logic ‘O’, thereby
energizing relay S.i and starting all counters, but leaving relays
SG-& de-energized. The 2 ms pulse holds the shift register
clear while the relay and comparator settle.
Following the fist polarity reversal, detected by the
comparator, a logic ‘1’ is shifted into the first element of the
shift register. The first counter is thereby stopped, relay S S is
de-energized and relay s6 is energized. A similar process is
initiated by each successive polarity reversal of the bridge
until all the counters are stopped. The present logic circuitry
therefore differs from that of Anderson et aZ(l974) by virtue
of the sequencing of the switching of s3 - S~ .The logic
elements used throughout are ‘74’ series TTL.
The counting unit consists of six separate counters each of
six decades (SN7490) with separate clock inputs and a
common reset line. These counters are multiplexed to a six-
decade, seven-segment display via push buttons enabling
readout of any counter at will. This arrangement allows a great
saving in cost compared with an earlier version [Anderson et
a1 1974) with no loss of flexibility or precision.
3.2 Fluid cell design
Figure 4 shows the final design of the hot wire cells. Two cells
of this type have been constructed differing only in their
lengths. The outer wall of the cell is formed by a stainless steel
cylinder of internal diameter 10 mm fitted with detachable
sides allowing access to the interior. When in position, these
pieces fit flush with the inner wall of the tube and provide a
cylindrical outer boundary to the cell. The platinum wire of 10
pm diameter is supported between two threaded stainless steel
pins, being mechanically and electrically connected to them
with Wood’s metal. The pins are mounted in pyrophillite
bushes at either end of the cell.
Following assembly of the wire in the cell it was annealed at
red heat for a short time and then tightened by means of the
threaded pins until, with a small current flowing through it, there
was no observable deflection when a small bar magnet was
brought close to it. Subsequently the length of the wire between
the Wood’s metal connections was measured with a cathetometer
to kO.01 mm. Both cell assemblies were then washed in Decon 90
solution and acetone, the cylinder sides attached and the two cells
mounted in a cylindrical stainless steel vessel sealed with a rubber
O-ring, with their axes parallel
1076
c 1I2 v
Figure 3 The control
logic. Components:
all resistors 4.7 kQ;
Ind. is a 12 V indicator
bulb; NAND gates, +
of SN7400; Inverters,
&Of SN7404; NOR
gates, + of SN7402;
monostables, SN74121;
shift register, SN74174
to the vessel axis. The vessel is equipped with thermocouples
for temperature measurement, pipework for evacuation and
admission of the test fluid, and ceramic/metal seals for
electrical connections. The entire vessel was immersed in a
liquid thermostat bath providing stability of k 0.005 K at 288
K. The fluid temperature was measured uith a platinum
resistance thermometer immersed in the vessel. Details of this
part of the assembly are omitted here because of its standard
nature. During assembly extreme care was taken to ensure that
the axis of the enclosing vessel was vertical. The ~erticalityof
the vessel could be adjusted by means of levelling screws on
its supports, and could be checked with the aid of a spirit level
to within an angle of k 1’. The machining of the cell supports
within the vessel was such that the error in the \erticality of the
hot wires was insignificantly worse than this. An experimental
check of the correct alignment of the hot wires was also
possible ($4.1 1).
I
The resistance-temperature relation for the platinum wires
was determined by in situ calibration. For this purpose the
cells were filled with helium and their temperature measured
with the platinum resistance thermometer with an accuracy of
5 0.001 K. Resistance measurements were performed by a four-
terminal technique with the connections made externally to the
enclosing vessel. The resulting uncertainty in the resistance of the
wires (including possible lead resistances to the wires) was =
0.006 %. The calibration was carried out in the temperature range
8-41 “C and the resulting data were correlated with a maximum
deviation of 0.04 % by the relation
R( TyR(273.16)=1+ 3,9985x W3(T- 273516)
- 1 . 7 0 0 ~10-G(T-273*16)2 (6)
where Tis the absolute temperature.
It is noteworthy that a calibration repeated with the cells full
of n-heptane showed no significant deviation from this
correlation.
4 Measurement analysis
As indicated above, the design of the present instrument has
been performed so as to minimize deviations of its operation
from the simple mathematical description of it embodied in
equation (3). In this section we demonstrate how the present
experimental design has eliminated some of these deviations,
A n apparatus to measure the thermal conductivity of liquids
'A !
Figure 4 The hot wire cells
reduced others to negligible proportions and made the
remainder amenable to evaluation. The corrections are
arranged in every case so that their application brings the raw
experimental data on to the straight line characterized by
equation (3).
4.1 Effects eliminated entirely
4.1.1 Conwction The problem of convection in a vertical
transient hot wire cell of the type described here has been
discussed by a number of authors (e.g. Goldstein and Eckert
1960, Goldstein and Briggs 1964). The essential result of all
such analyses is that there exists an estimable time limit t c ,
measured from the initiation of the heating process, prior to
which the one-dimensional regime of heat transfer implicit in
equation (3) persists. For the present cells this time limit is
approximately 10 s so that, provided that the measurement
cycle is completed in a time less than this, convection is
nominally eliminated. In the present measurements the time for
one experiment cycle does not exceed 2 s and this criterion is
easily fulfilled.
On the other hand, de Groot et al(1974) have shown that, in
the gas phase, convective heat transfer can occur at times
shorter than the limit t c . However, the same authors have also
demonstrated that in the transient hot wire cell the effects of
convective heat transfer can be observed experimentally by
virtue of the departures they cause from the behaviour
predicted by equation (3). Therefore, while adopting 2 s as a
suitable experiment time for the present measurements, the
need to examine the experimental data for evidence of
convective currents is to be borne in mind.
In particular, it should be emphasized that all the discussion
of the convection problem has referred to a vertical wire. It is
clear that the onset of convection will occur at shorter times if
the wire is not truly vertical. Because the effect of convection
is to increase the heat transfer from the wire, such an effect is
clearly discernible as a curvature in the plot of A T against In t
at times following the beginning of convection. In this work no
such curvature was detected for times as long as 10 s, so it
may be concluded that the wires were sufficiently well aligned,
However, it is noteworthy that when the cells were displaced
by about 1’ from vertical, curvature of the h T against In t plot
occurred after 0.5 s.
4.1.2 Wire end effects The mechanical and electrical
connection of each platinum wire to a support in the fluid cell
necessarily implies the presence of longitudinal temperature
gradients in the wire and the fluid. Haarman (1971) and Healy
et a1 (1976) have provided differing analyses of this problem.
Both discussions show that the ‘end effects’ can be arranged to
cause only small (- 0.01 %) departures from the behaviour
predicted by equation (3), by making the platinum wires
sufficiently long. In formulating the present design we have
employed Haarman’s analysis (Haarman 1971) and have
chosen to use a long wire of length 0.15 m and a short wire of
length 0.04 m. Under these conditions, calculations (Castro et
a1 1974) show that the ratio of the resistances of the long and
short wires does not differ by more than 0.01 % from the ratio
of their lengths during an experimental run. Thus equations
(4) and (5) yield a measure of the difference of the resistance of
the long and short wires directly. The small remaining end
effects are therefore experimentally compensated and need not
be considered further.
4.2 Corrections rendered negligible
The corrections detailed by Healy et a1 (1976) which fall into
the present category include Knudsen effects and effects due to
viscous heating of the fluid (McLaughlin and Pittman 1971). In
the present system these effects never amount to more than
0.01 % of the measured temperature rise of the wire, and are
therefore negligible.
4.3 Remaining corrections
4.3.1 Heat dissipation per unit wire length A basic
presumption of the derivation of equation (1) was that the heat
dissipation per unit length of the platinum wires is constant.
Evidently, the use of a constant voltage source for the bridge
power and the changing resistances of the platinum wires
combine to vitiate this assumption. However, if the fixed
1077
C A Nieto de Castro, J C G Calado, W A Wakeham and M Dix
resistances of the bridge RI and Rz are chosen such that
RI + Rz = RL(O)+ Rs(0)= Ro
it can be shown (de Groot et a1 1974) that the heat dissipation
in the wires as a function of time is, to a sufficient degree of
accuracy,
Here,
AR(t j= A R L ( ~+)h R s ( t ) (8)
where ARL(t) and ARs(t) are the resistance increases of the
long and short wires respectively during the heating process.
Calculation for the bridge of figure 1 shows that q ( t ) departs
from the value for t = 0 by at most 0.06 % during the
measurement cycle. This departure may be expected to lead to
barely discernible deIiations of the plot of A T against In t
from linearity.
However, because the deviation from constant q is so small,
equation (7) also pro! ides a means of correcting the observed
temperature rises to those appropriate to q(0). This may be
performed most easily by an iterative technique, using q(0) to
generate estimates of h R ~ ( tand) LRs(t) and then equation
(7) to find q(t) . Finally, the proportionality between AT and q ( t )
may be used to obtain corrected \. alues of AT. In all cases the
correction is so small that the first iteration is sufficient, Indeed,
for the sample measurements reported here q did not deviate by
more than k 0.01 % from a constant value, so that this iteration
was unnecessary.
4.3.2 Finite physical properties of the wire Equation ( 1 ) was
derived on the assumption of an infinitely thin line source of
heat. In practice the platinum wire has a finite diameter, a
finite heat capacity (PC,),~and a finite thermal
conductivity .,A The consequences of all of these
considerations have been investigated by Healy et a1
(1976);and according to their analysis the average temperature
of the wire, which determines its resistance, is given by
where K ~ h,/(pCp),=.
The corrections to equation (3) incorporated in equation (9)
are of two types. The term h/2h, constitutes a small (-0.1 %)
time-independent contribution. It therefore plays no part in the
evaluation of the thermal conductivity of the fluid from the
slope of the plot of A Tagainst In t. The smallest of the
remaining time-dependent contributions, the term - ~ 1 2 1 4
~never~4 amounts to more than 1 x lO-3% of the temperature
rise and is therefore negligible. The other two, time-dependent
corrections are more significant and can amount to about 1 %
of the temperature rise at the shortest times employed in the
present measurements. However, the correction to be added to
the observed temperature rise,
may be estimated to within a few per cent. Thus, provided that
measurements for which the correction exceeds 1.0% of the
temperature rise are ignored, the resulting uncertainty in the
corrected temperature rise is only a few parts in ten thousand.
1078
For the purposes of this correction the wire radius was
measured with an electron microscope (a= 5-49k 0.08 pm).
4.3.3 Finite dimensions of the fluid cell The derivation of equation
(1) presumes that the test fluid is of infinite extent in the radial
direction. In practice the fluid is contained in a cylindrical vessel
of radius b ( 1 cm for this work). A correction to the observed
temperature rise, which is of greatest significance at long times, is
given by de Groot et a1 (1974) as
(11)
where g , are the consecutive roots of Jo(g,) = 0. JOis the zero-
order Bessel function of the first kind and Y Othe zero-order
Bessel function of the second kind.
The present experimental arrangement was chosen to ensure
that 6 T2 never exceeded 0.02 % of the wire temperature rise.
The remaining small correction is easily made if necessary by
adding the value of 6Tz to the observed wire temperature rise.
4.3.4 Variable fluid properties The fact that the density and
thermal conductivity of the liquid vary with temperature
introduces a further correction (Healy et al1975). This is most
conveniently viewed as a correction to the temperature with
which the thermal conductivity, obtained from the plot of A T
against In t, is identified (de Groot et a1 1974).
In the present work the slope of h T against In t is usually
obtained for the range h T ( t l ) = 5 " C to hT(tz)=lO"C. The
change- in the liquid thermal conductivity over this range is
only 1 % and so it can be adequately described by a linear
function of temperature. Thus it is possible to carry over the
analysis of Healy et a1 (1976) for the variation of the fluid
properties of a gas to the case of a liquid with only small
changes (Castro et al 1974). This analysis indicates that if the
temperature of the fluid before initiation of a run is BO, the
temperature to which the obserl ed thermal conductivity
refers, Tr, is given as
Tr= Bo++[LT(tl)+ AT(t2)l
with negligible error. In 46 the correctness of this temperature
assignment is confirmed experimentally.
In principle, the variation of the fluid properties with
temperature introduces a very small curvature to the plot of A
Tagainst In t (de Groot et a1 1974). However, in practice this
curvature cannot be discerned. Nevertheless, for the sake of
consistency, we have applied a correction for this effect as
detailed by de Groot et al(1974).
4.3.5 Radiation correction The final correction to be considered
involves the radiative heat transport from the heated wire to the
fluid simultaneously with the conductive process. Mani (1971)
has recently considered this problem for the transient hot wire
apparatus. However, because his detailed results are not generally
available, it is preferable to postpone a discussion of this last
refinement of the measurement until definitive thermal
conductivities are reported. At present it is sufficient to conclude
that, in view of the small temperature rises involved in the present
measurements, this final correction is not expected to amount to
more than 1 % in the thermal conductivity (Mani 1971).Thus even
if only a crude estimate of the effect can be made the overall
accuracy of the final thermal conductivity derived should be
maintained.
5 Precision and accuracy
The precision of the time measurement for the present
apparatus, with due regard to the noise level of the comparator
An apparatus to measure the thermal conductiuitj' of liquids
and the bridge network, is f 10 p. On the other hand, the
resistance change of the wire is determined with a precision of
f 0.01 %, estimated from the tolerances of the fixed resistors in
the bridge. Accounting for the uncertainties introduced by
applying the small corrections to the measured temperature rise
detailed earlier, we should expect that the maximum deviation of
any experimental point from the linearity predicted by equation
(3) for-a single run, consisting of six time measurements, should
be f 0.05 %. In order to demonstrate that this expectation is
fulfilled by the present apparatus we display in figure 5 the
deviations from linearity obtained for
from linearity are slightly degraded when a number of runs are
combined to cover the entire accessible time regime. Nevertheless,
the maximum deviation remains less than 0.1 % as can be seen
from figure 6. This graph contains the deviations of the results of
ten successive runs from the straight line providing the best fit to
the 60 points. The ten runs were performed under nominally
identical equilibrium conditions with n-heptane as the test fluid.
As a consequence of these considerations it is contended that the
precision of the present measurements is
k0,1 %.
The accuracy of the thermal conductivity value deduced
from the slope of hT against In t plots is determined by the
uncertainty in the resistance-temperature characteristics of the
wires, by the uncertainty involved in merely obtaining the
slope, and by any unaccounted corrections. Although the
former two points contribute only an estimated f 0.1 % to the
overall uncertainty, it is prudent to reserve a final estimate of
the accuracy of the thermal conductivity until a later
publication because no radiation correction can yet be applied.
However, the discussion in $4.3.5 indicates that allowing for
this correction the overall accuracy of the measurements will
be in the range f 0.2-0.3 %.

-010
0 01
~
005 015 CL
Tieiie s )
Figure 5 Deviations of experimental data from the
linearity predicted by equation (3) for individual runs,
Orunl;Arun2;*run3;@run4
four discrete runs with the fluid cells filled with n-heptane.
The maximum deviation is f0.06 %, in accord with the
foregoing description. It is also noteworthy that the average
deviation from linearity for each of the four runs is k 0.02%.
Because of unavoidable variations in the fluid temperature of
the order of f 0.01"C between successive runs, the deviations
I nominal temperatures by
6 Sample measurement
The apparatus described here has been used to perform
preliminary measurements of the thermal conductivity of n-
10
heptane at four temperatures near room temperature. These
measurements are in no way final results but are used here to
illustrate the correct operation of the instrument described and
the underlying theory for liquids. For these measurements the
n-heptane used was commercial BDH n-heptane treated with
4A molecular sieves to remove water, degassed under vacuum
and then distilled under its own vapour pressure.
For these measurements bath temperatures of 293.04 and 287.99
K were employed. At each temperature several different heat
inputs to the wire were employed. The different temperature rises
generated in consequence allow measurement of the thermal
conductivity at different reference temperatures (34.3.4).
Conditions were chosen such that reference temperatures for some
runs for both bath temperatures were almost the same. These tests
then not only provide data for the thermal conductivity of n-
heptane, but also enable the conclusions of $4.3.4 to be examined.
In addition, they provide support for our contentions with regard
to precision. Table 2 contains the thermal conductivities obtained
as a result of these measurements; they have been corrected to

Table 2 The thermal conductivity of n-heptane at

various reference temperatures

Bath 4 Reference h
temperature, (W-l) temperature, (W-L-K-1)
l- (K) Tr (K)

287.99 1.1874 293.60 0,1253

1.8425 295.65 0,1242
2.6303 298.15 0,1233
2.6307 300.15 0.1227
I 293.04 1.1762 298.15 0.1235
-0.10- 0.05 0.15 0.4 1. 0
Time f (5)
1.1493 298.15 0.1236
158254 300.15 0.1229
Figure 6 Deviations of experimental data from the 1,7842 300.15 0.1227
linearity predicted by equation (3) for a complete set of 2.6073 303.65 0.1217
measurements 2.5471 303.65 0.1219
2.5476 303.65 0.1220

1079
C A Net0 de Castro, J C G Calado, W A Wakeham and M Dix
application of small (0.1 %) temperature corrections with the
aid of dhidTgiven by Kandiyoti et a1 (3972). The table shows
that the repeated measurements do not differ by more than
0.15% even for different bath temperatures. Thus the use of the
reference temperature given by equation (12) is vindicated as are
our claims for precision.
Finally table 3 compares the present data with those
obtained in earlier studies (Kandiyoti et a1 1972, Vilim 1960,
Table 3 Thermal conductivity of n-heptane, A (W m-l K-l)
Michels A and Botzen A 1952 Physica 18 605-12
Michels A, Sengers J V and van der Gulik P S 1962
Physica 28 1201-15
Missenard F A 1965 Rev. Gen. Termique 4 409-28
Pittman J F T 1968 PhD Thesis University of London
Tsederberg N V 1965 Thermal Conductivity of Gases and
Liquids (Cambridge, Massachusetts : MlT Press)
Vilim 0 1960 Colln. Czech. Chem. Commun. 25 993-9

T=293.25 K T=303.15 K Source

0.1244 0.1214 Kandiyoti et al(l972)

0.135 - Vilim (1960)
- 0.126 Tsederberg (1965)
0.124 0.121 Jobst (1964)
0.127 0.124 Missenard (1965)
0.1254 0,1219 Present work

Tsederberg 1965, Jobst 1964, Missenard 1965). The agreement

with the most reliable data (Kandiyoti et aZl972) is gratifying,
being well within the claimed uncertainty for the earlier work
(1 %). However, because the present results are not corrected
for any radiation effects, any further comment would be
unwarranted at this stage.

Acknowledgments
We wish to thank Dr D Ferreira of the Laboratorio de
Biologica Celular of the Instituto Gulbenkian de Cihcia, for
electron microscope facilities, and Professor J Kestin of
Brown University for his interest and valuable advice.

References
Anderson G P, de Groot J J, Kestin J and Wakeham W A
1974 J. Phys. E: Sci. Instrum. 7 948-51
Carslaw H S and Jaeger J C 1959 Conduction of Heat in
Solids 2nd edn (London: Oxford University Press)
Castro C A N, Wakeham W A and Calado J C G 1974
Proc. 1st Nat. Congr. on Pure and Applied Mechanics,
Lisbon
Goldstein R J and Briggs D G 1964 J . Heat Transfer,
Trans. ASME C 86 490-500
Goldstein R J and Eckert E R G 1960 Int. J . Heat Mass
Transfer 1 208-18
de Groot J J, Kestin J and Sookiazian H 1974 Physica 75
454-82
Haarman J W 1969 PhD Thesis Technische Hogeschool,
Delft
Haarman J W 1971 Physica 52 605-19
Haarman J W 1973 Transport Phenomena: Am. Inst. Phys.
Conj Proc. No. 2
Healy J J, de Groot J J and Kestin J 1976 Physica 82 392-408
Jobst W 1964 Int. J . Heat iMass Transfer 7 725-32
Kandiyoti R, McLaughlin E and Pittman J F T 1972
1.Chem. Soc. Faraday T m m . I 6 8 860-6
Kestin J. Ro S T and Wakeham W A 1972 Physica 58
165-21 1
Mani K 1971 PhD Thesis University of Calgary, Canada
McLaughlin E and Pittman J F T 1971 Phil. Trans. R. Soc. Journal of Physics E: Scientific lnstruments 1976 Volume 9
A 270 557-602 Printed in-Great Britain 0 1976

1080