METHOD OF MEASURING THE DIELECTRIC CONSTANTS OF LIQUIDS. 315 XLIX.-A Method of Measusing the Dielectric Con-stants of Liquids. By LEONARD ALFRED SAYCE and HENRY VINCEXT AIRD BRISCOE. THE introduction of the triode valve has made possible many accurate methods for the measurement of physical constants and it has been employed by several workers in the determination of dielectric constants. Considerable accuracy has thus been attained in the measurement of dielectric constants approaching unity, v i x . those of the gases (Carman and Lorance PhyLysical Rev. 1922, 20 715; Gill Radio Rev. 1921 2 450; Wagstaff Phil. Hag., 1924 [vi] 47 66)) but the dielectric constants of liquids which are much greater than unity have not hitherto been measured with similar accuracy (Whiddington Proc.Camb. Phil. Xoc. 1921 20, 445; Hyslop and Carman Physical Rev. 1920 15 243). This is mainly due to the difficulty of securing absolute constancy in the operation of valve-maintained circuits (Whiddington Engineering, 1920 110 384; Wagstaff loc. cit.) and of obtaining a variable condenser combining a large range of variation with a su%cient accuracy of setting (Hyslop and Carman loc. cit.). Further in previous methods the dielectric has been stressed in a circuit directly coupled to a triode and in such circuits although the funda-mental frequency may be known yet the wave-form is not usually sinusoidal but may be resolved into many harmonics some of M* 316 SAYCE AND BRISCOE A METHOD OF considerable amplitude. Possibly this is not seriously disadvan-tageous in work upon gases but in the case of dielectrics in which the variation of dielectric constant with frequency is appreciable it is evident that precise measurements of this property can be made only if the applied stress be of sine wave-form.The following method has therefore been developed in order to secure the conditions here seen to be required for the accurate measurement of relatively large dielectric constants. General PrincipZes.-If oscillatory currents are maintained in the circuit L,C by means of a triode connected as shown in Pig. 1, the magnitude of the direct component of the current in the anode circuit uf the triode is a function of the magnitude of the oscillatory currents in L,C (Dowling Proc. Roy. Dublin SOC. 1921 16 185).If now a second oscillatory circuit L3C3 is loosely coupled to LlC2, little or no energy is withdrawn from the latter until the natural frequency of L3C3 approaches that of L,C,. Resonance between the two circuits is then shown by a sudden decrease in the direct component of anode current of the triode. If the coupling between L and L3 is sufficiently tight the reduction in the anode current will continue throughout a considerable change of L3C3 for the latter circuit is forced into resonance with L,C over a considerable range of adjustment and then suddenly " breaks away " from the control of L,C,. The coupling between L and L3 can however be so adjusted that when a gradual change is made in the adjustment of C the anode current steadily decreases to a minimum and thereafter steadily increases.These conditions afford an extremely sensitive null method of observing when L3C3 is adjusted to a given fre-quency i.e. the frequency of L,C,. If the frequencies of the two circuits are of the order of 106- per second the resonance point is shown so critically that with ordinary precautions C can be reset with an error not exceeding 0-005ppF. - If now another condenser, C4 of unknown capacity be connected in parallel with C, then in order to restore resonance between the two circuits a reduction i SIEASURING THE DIELECTRIC CONSTANTS OF LIQUIDS. 31 7 the value of C is necessary such reduction being a measure of the capacity of C,. Thus unknown capacities can be measured accw-ately in terms of C, the frequency a t which the measurement i.: made being that of L,C,.The coupling between the two ckcnits is so loose that the stress applied t o the dielectric of the unknonn condenser C, is very small and has a wave-form free from the distortion almost inevitable in a valre-maiiitaiiiecl circuit. BJ. this method the messurement of ail uilk110~11 condenser involr-cis two null galvanometer adjustments and in practice these are made within such a short time that errors due to slight variation3 ill the frequency of L,C, due to variations in batteries tic.. zic insignificant. The Bxperinaeiztal Nethod.-The practical application of the above principles to the measurement of the dielectric constants of liquids presents two main problems : (1). The construction of a condenser of fixed dimensions in ti liich (lither a vacuum air or the substance under investigation may be used as dielectric a t will.For convenience this is hereafter called the * ' dielectric container." (2). The construction of a variable condenser having the relatively wide range of 5OOpp.P and yet such that it can be set and reaid with an error not exceeding 0.005ppF. (1). The capacity of the dielectric container is measure6 'BI lien full of air or evacuated and again when full of the liquid under examination the ratio of the latter capacity to the former being the dielectric constant of the liquid a t the temperature and fre-quency employed. The type of condenser that has been found most suitable is shown in Fig. 2. It resembles an elongated Demar flask and consists of two glass tubes like large " boiling-tubes." one sealed inside the other.Access to the annular space betn-em these tubes is given by the tubes G and H. Platinuni electrodes me fused through the bottom of the inner tube and through t'lie side of the outer tube. Both internal surfaces of the annular apace are silvered to the height shown and these silver films constitute the "plates" of the condenser. It was thought advisable to thicken the silver films by several successive applications of the silvering solution. A small glass elbow-tube ,J is cemented to the outer tube where the platinum electrode pierces it aiid mercury, poured into the little cup so formed provides a means of making a connexion to the outer silver coating. A little mercury is also poured into the inner tube to cover the other platinum electrode and connexion with it is made by a long stiff wire.At the top of this wire is cemented a small glass cup containing mercury aiid having a little hook of glass rod fused on a t one side. A stiff wire 318 SAYCE AND BRISCOE A METHOD OF coming from the remainder of the apparatus normally rests on this glass hook but is transferred to the adjoining mercury cup in order to switch the condenser into circuit. The condenser is immersed in a thermostat so that the water-level is a few milli-metres below the top a,nd the thermostat and outer silver coating of the condenser are connected to the " earthed" side of the remaining apparatus. FIG. 2. FIG. 3. P Y ( 2 ) . The wide-range variable condenser is an assembly of six condensers of fixed capacity with a small one of variable capacity.The former are labelled from " a " to " f " and are made of the following approximate values C = lOppP C = ZOppF C = 40ppF Cd = 80ppF C = lGOppF Cf = 320ppP. The small variable condenser is variable over a range of 10ppcF. It is per-manently in circuit but any or all of the fixed-capacity condensers may be switched in parallel with it by means of mercury cups and bridges. Each fixed-capacity condenser is made with interleaved aluminium plates separated by air. One set of plates is connected to an earthed metal base-plate and to an earthed tin-plate screen which envelops the condenser the other set of plates being con-nected to a brass rod which acts as the '' live '' terminal MEASURING THE DIELECTRIC CONSTANTS or LIQUIDS.319 The construction of the small variable condenser is shown in Fig. 3. It consists essentially of a single semi-circular vane M, free to rotate between two semi-annular rings N. The frame-work of the condenser comprises two triangular brass end-plates P, held apart by three vertical rods Q. Each end-plate is provided with a* longitudinally-drilled steel screw R and these two screws are the bearings for the conical ends of a rotatable spindle S. The spindle carries a thick semi-circular brass plate M accurately turned on both faces whilst mounted on the spindle. The two fixed plates of the condenser are two semi-annular rings N made of heavy brass and accurately turned secured respectively to two triangular ebonite platforms T which may be fixed a t any suitable position upon the vertical supporting rods.At the top FIG. 4. of the spindle is clamped a brass boss V carrying a plane galvanometer mirror W and a fibre arm X about 15 om. long. The framework of the condenser and therefore the rotating vane is earthed by attachment to an earthed base-plate as also is a tin-plate screen which covers the effec-tive parts of the condenser. The two semi-annular rings are connected to a stiff wire which passing through a hole in the screen constitutes the '. live " terminal. The position of the spindle is read by means of a telescope and a semi-circular scale about 150 cm. in length as shown in Fig. 4. The mirror on the spindle is a t the centre of the scale and the telescope is aligned upon it thus a rotation of the spindle of 90" corresponds to about 1500 mm.on the scale. The reading of the scale can be made within 0.2 mm. and the adjustment of the rotation of the spindle with this degree of accuracy is accomplished quite simply in the following manner. To the tip of the fibre arm X a long thread is attached which passes to one end of a wooden lever L 30 cm. long pivoted a t its middle point. To the other end of the lever is attached a second thread which is wound upon a thin brass rod provided with a knob and rotating stiffly in a cork that is clamped rigidly. Turning the knob thus rotates the spindle of the condenser. The threads ar 320 SAYCE AND BRISCOE A METROD OF kept taut and the spindle is returned in the other direction by the tension of a stout rubber band.The wooden lever is introduced into the system in order that the adjusting knob may be near the hand of the operator whilst he is a t the telescope at a distance from the conde’hser. The design of variable condenser here described has the following important advantages : (1) Consistency of resetting is obtained by the use of conical bearings. (2) Great accuracy of setting and reading is attained in the manner already described. (3) Freedom from external capacity effects is ensured by the earthing of the moving system adequate screening and remote control. (4) The calibration curve is free from the irregularities that arise in the usual vane type of condenser through the effects of the spacing washers and supports.( 5 ) A simple adjustment of the range of capacity within wide limits is attained by alteration of the positions of the eborrite platforms T (Fig. 3). The complete assembly of the apparatus is shown diagram-matically in Fig. 1. The dielectric container C, and the wide-range condenser C3 are connected in parallel with an inductance, L3 which for frequencies of the order of 106- per second consists of 30 well-spaced turns of No. 16 S.W.G. copper wire wound upon a spirally-grooved ebonite tube 9 cm. in diameter. L and L, are “honeycomb ” inductances of the type commonly used in radio-telegraphy; C is a variable condenser of the ordinary vane type having a maximum capacity of approximately 300ppF ; and C is a mica-dielectric “ by-pass ” condenser of approximately 0.01 yF.B is a dry-cell battery of 50 volts and B a 4 volt. 100 amp.-hour accumulator ; G is a uni-pivot galvanometer having a sensi tivity of about 1pA. per scale division. R, a potentiometer of 5000 and R, a resistance of 9000 are adjusted so as to neutralise the greater part of the anode current through G (Dowling Zoc. cit. p. 175). Although the construction of Cv was adapted to secure a recti-linear relationship between scale reading and capacity it was intercalibrated throughout its range by substituting for Ca a very small fixed-capacity condenser the capacity of which was measured in terms of scale divisions of C at a large number of places along the scale of the latter. If the “ scale-division capacity ” ratio had been truly rectilinear the apparent value of the very small condenser would have been precisely the same at all points alon JIESSURISG THE DIELECTRIC CONSTAKTS O F LIQUIDS.321 the scale. The apparent values obtained gave the data necesaary for the preparation of a correction-curve to express all readings in terms of “ mean scale divisions.” Finally the capacities of the six fixed-capacity condensers C to Cf were measured in terms of scale divisions of Cv (for C = (‘v C 2 Ca + C, C 5 Ca + (3 t C’v, and so on). The capacity corresponding to a mean scale division of C is thus the unit in all measurements. Having thus intercalibrated C, the capacity of C, coiitainiiig dry air a t a known pressure and a t the temperature of the ther-mostat is measured in terms of C,. To do this C is switched out of circuit and with C a t a large setting L,C is adjusted to approximate resonance with L,C,.Exact resonance between t 110 two circuits is then obtained by varying C’ and is shown by a sharp minimum reading of the galvanometer G and the reading of C ‘ , is taken. C is then switched into circuit and C is reduced in value until resonance i q again obtained between the two circuits. This reduction in C is a measure of the capacity of Clcab). Finally, C4 is filled with the liquid under examination (at the temperature of the thermostat) and its capacity measured as beiore. The latter capacity C(q(liquid) divided by the former C4(sir) gives the dielectric constant of the liquid to a standard of 1 1, at the temperature and frequency of the measurement.Shortly lieforc or after the measurement the frequency is determined by means o€ a heterodyne waTY-e-meter (Sayce Expt. Wireless 1923, 1 70). The value of the dielectric constant so obtained can lie corrected to a vacuum standard by making use of recent deter-minations of the specific inductive capacity of eir (Carman anti Lorance Physicul Rev. 1922 20 715; GiIl Radio lieu. 1921 2, 450; Wagstaff Phil. Nag. 1924 [vi] 47 66). Experimentul Difficulties and Precautions.-When fist applied , the method gave inconsistent results due to slight temperature changes in the valve in the oscillator coils and particularly in the variable condenser C,. Sudden changes in the temperature of the valve were prevented by enveloping it in cotton-wool. The influence of sunlight upon the apparatus was so marked that it was found desirable to obscure the windows of the laboratory and illuminate the apparatus by artificial light.Ultimately the air of the room was kept a t a fairly constant temperature and as an additional precaution the condensers C to Cz were intercalibrated with C before and after each determination. Results.-The following results obtained with benzene are given, not as having any absolute significance but simply as an example of the degree of consistency so far observed in repeated meamre-ments on the same material. This we regard as a n importan 322 FAIRBROTHER AND MASTIN : criterion of the precision of the method. The benzene was rigorously purified by fractional distillation and fractional crystallisation and was finally distilled over phosphorus pentoxide but without any attempt to attain intensive drying.Frequency 65 x 103 cycles per second. E (air = 1) ............... 2.2380 2.2392 2.2396 2.2389 Dielectric Constant of Benzene a t 25.5". Mean 2.2389 Diff. from mean ......... -0.0009 +0.0003 +0.0007 -0.0001 Mean difference from mean &0.0005. It is interesting to make a comparison as to consistency between our results and those of Turner (2. physikaZ. Chem. 1900 35 385), who using the method of Nernst (ibid. 1894 14 622) applied to carefully purified benzene obtained the following data which are commonly cited as the most accurate available. The frequency was low but unspecified and the results were corrected to 18". F ............... 2.290 2-291 2.285 2.288 2.293 2.292 2.286 Diff.from Mean 2.289. mean ...... $0.001 +0*002 -0.004 -0.001 +0-004 +0.003 -0.003 Mean difference from mean &O-003. During our measurements no special precautions were taken to maintain room temperature constant and there is much evidence that even the small differences here recorded are due to small changes in the temperature of the various parts of the apparatus. Hence it seems probable that under more constant temperature conditions still greater precision may be attained by the method. Note.-Since the completion of the above measurements a further contribution to the measurement of dielectric constants has been made by Grutzmacher (2. Physik 1924 28 342). Results for benzene stated to the fourth decimal place and from internal evidence probably significant to the third decimal place are there given but as only one result is recorded for each temperature we are unable to compare the consistency of the measurements with that obtained by our method.ARMSTRONG COLLEGE, NE WCASTLE -UPON -TYNE. [Received October 1 Ith 1924. METHOD OF MEASURING THE DIELECTRIC CONSTANTS OF LIQUIDS. 315 XLIX.-A Method of Measusing the Dielectric Con-stants of Liquids. By LEONARD ALFRED SAYCE and HENRY VINCEXT AIRD BRISCOE. THE introduction of the triode valve has made possible many accurate methods for the measurement of physical constants and it has been employed by several workers in the determination of dielectric constants. Considerable accuracy has thus been attained in the measurement of dielectric constants approaching unity, v i x .those of the gases (Carman and Lorance PhyLysical Rev. 1922, 20 715; Gill Radio Rev. 1921 2 450; Wagstaff Phil. Hag., 1924 [vi] 47 66)) but the dielectric constants of liquids which are much greater than unity have not hitherto been measured with similar accuracy (Whiddington Proc. Camb. Phil. Xoc. 1921 20, 445; Hyslop and Carman Physical Rev. 1920 15 243). This is mainly due to the difficulty of securing absolute constancy in the operation of valve-maintained circuits (Whiddington Engineering, 1920 110 384; Wagstaff loc. cit.) and of obtaining a variable condenser combining a large range of variation with a su%cient accuracy of setting (Hyslop and Carman loc. cit.). Further in previous methods the dielectric has been stressed in a circuit directly coupled to a triode and in such circuits although the funda-mental frequency may be known yet the wave-form is not usually sinusoidal but may be resolved into many harmonics some of M* 316 SAYCE AND BRISCOE A METHOD OF considerable amplitude.Possibly this is not seriously disadvan-tageous in work upon gases but in the case of dielectrics in which the variation of dielectric constant with frequency is appreciable it is evident that precise measurements of this property can be made only if the applied stress be of sine wave-form. The following method has therefore been developed in order to secure the conditions here seen to be required for the accurate measurement of relatively large dielectric constants.General PrincipZes.-If oscillatory currents are maintained in the circuit L,C by means of a triode connected as shown in Pig. 1, the magnitude of the direct component of the current in the anode circuit uf the triode is a function of the magnitude of the oscillatory currents in L,C (Dowling Proc. Roy. Dublin SOC. 1921 16 185). If now a second oscillatory circuit L3C3 is loosely coupled to LlC2, little or no energy is withdrawn from the latter until the natural frequency of L3C3 approaches that of L,C,. Resonance between the two circuits is then shown by a sudden decrease in the direct component of anode current of the triode. If the coupling between L and L3 is sufficiently tight the reduction in the anode current will continue throughout a considerable change of L3C3 for the latter circuit is forced into resonance with L,C over a considerable range of adjustment and then suddenly " breaks away " from the control of L,C,.The coupling between L and L3 can however be so adjusted that when a gradual change is made in the adjustment of C the anode current steadily decreases to a minimum and thereafter steadily increases. These conditions afford an extremely sensitive null method of observing when L3C3 is adjusted to a given fre-quency i.e. the frequency of L,C,. If the frequencies of the two circuits are of the order of 106- per second the resonance point is shown so critically that with ordinary precautions C can be reset with an error not exceeding 0-005ppF. - If now another condenser, C4 of unknown capacity be connected in parallel with C, then in order to restore resonance between the two circuits a reduction i SIEASURING THE DIELECTRIC CONSTANTS OF LIQUIDS.31 7 the value of C is necessary such reduction being a measure of the capacity of C,. Thus unknown capacities can be measured accw-ately in terms of C, the frequency a t which the measurement i.: made being that of L,C,. The coupling between the two ckcnits is so loose that the stress applied t o the dielectric of the unknonn condenser C, is very small and has a wave-form free from the distortion almost inevitable in a valre-maiiitaiiiecl circuit. BJ. this method the messurement of ail uilk110~11 condenser involr-cis two null galvanometer adjustments and in practice these are made within such a short time that errors due to slight variation3 ill the frequency of L,C, due to variations in batteries tic..zic insignificant. The Bxperinaeiztal Nethod.-The practical application of the above principles to the measurement of the dielectric constants of liquids presents two main problems : (1). The construction of a condenser of fixed dimensions in ti liich (lither a vacuum air or the substance under investigation may be used as dielectric a t will. For convenience this is hereafter called the * ' dielectric container." (2). The construction of a variable condenser having the relatively wide range of 5OOpp.P and yet such that it can be set and reaid with an error not exceeding 0.005ppF. (1). The capacity of the dielectric container is measure6 'BI lien full of air or evacuated and again when full of the liquid under examination the ratio of the latter capacity to the former being the dielectric constant of the liquid a t the temperature and fre-quency employed.The type of condenser that has been found most suitable is shown in Fig. 2. It resembles an elongated Demar flask and consists of two glass tubes like large " boiling-tubes." one sealed inside the other. Access to the annular space betn-em these tubes is given by the tubes G and H. Platinuni electrodes me fused through the bottom of the inner tube and through t'lie side of the outer tube. Both internal surfaces of the annular apace are silvered to the height shown and these silver films constitute the "plates" of the condenser. It was thought advisable to thicken the silver films by several successive applications of the silvering solution.A small glass elbow-tube ,J is cemented to the outer tube where the platinum electrode pierces it aiid mercury, poured into the little cup so formed provides a means of making a connexion to the outer silver coating. A little mercury is also poured into the inner tube to cover the other platinum electrode and connexion with it is made by a long stiff wire. At the top of this wire is cemented a small glass cup containing mercury aiid having a little hook of glass rod fused on a t one side. A stiff wire 318 SAYCE AND BRISCOE A METHOD OF coming from the remainder of the apparatus normally rests on this glass hook but is transferred to the adjoining mercury cup in order to switch the condenser into circuit.The condenser is immersed in a thermostat so that the water-level is a few milli-metres below the top a,nd the thermostat and outer silver coating of the condenser are connected to the " earthed" side of the remaining apparatus. FIG. 2. FIG. 3. P Y ( 2 ) . The wide-range variable condenser is an assembly of six condensers of fixed capacity with a small one of variable capacity. The former are labelled from " a " to " f " and are made of the following approximate values C = lOppP C = ZOppF C = 40ppF Cd = 80ppF C = lGOppF Cf = 320ppP. The small variable condenser is variable over a range of 10ppcF. It is per-manently in circuit but any or all of the fixed-capacity condensers may be switched in parallel with it by means of mercury cups and bridges.Each fixed-capacity condenser is made with interleaved aluminium plates separated by air. One set of plates is connected to an earthed metal base-plate and to an earthed tin-plate screen which envelops the condenser the other set of plates being con-nected to a brass rod which acts as the '' live '' terminal MEASURING THE DIELECTRIC CONSTANTS or LIQUIDS. 319 The construction of the small variable condenser is shown in Fig. 3. It consists essentially of a single semi-circular vane M, free to rotate between two semi-annular rings N. The frame-work of the condenser comprises two triangular brass end-plates P, held apart by three vertical rods Q. Each end-plate is provided with a* longitudinally-drilled steel screw R and these two screws are the bearings for the conical ends of a rotatable spindle S.The spindle carries a thick semi-circular brass plate M accurately turned on both faces whilst mounted on the spindle. The two fixed plates of the condenser are two semi-annular rings N made of heavy brass and accurately turned secured respectively to two triangular ebonite platforms T which may be fixed a t any suitable position upon the vertical supporting rods. At the top FIG. 4. of the spindle is clamped a brass boss V carrying a plane galvanometer mirror W and a fibre arm X about 15 om. long. The framework of the condenser and therefore the rotating vane is earthed by attachment to an earthed base-plate as also is a tin-plate screen which covers the effec-tive parts of the condenser. The two semi-annular rings are connected to a stiff wire which passing through a hole in the screen constitutes the '.live " terminal. The position of the spindle is read by means of a telescope and a semi-circular scale about 150 cm. in length as shown in Fig. 4. The mirror on the spindle is a t the centre of the scale and the telescope is aligned upon it thus a rotation of the spindle of 90" corresponds to about 1500 mm. on the scale. The reading of the scale can be made within 0.2 mm. and the adjustment of the rotation of the spindle with this degree of accuracy is accomplished quite simply in the following manner. To the tip of the fibre arm X a long thread is attached which passes to one end of a wooden lever L 30 cm. long pivoted a t its middle point.To the other end of the lever is attached a second thread which is wound upon a thin brass rod provided with a knob and rotating stiffly in a cork that is clamped rigidly. Turning the knob thus rotates the spindle of the condenser. The threads ar 320 SAYCE AND BRISCOE A METROD OF kept taut and the spindle is returned in the other direction by the tension of a stout rubber band. The wooden lever is introduced into the system in order that the adjusting knob may be near the hand of the operator whilst he is a t the telescope at a distance from the conde’hser. The design of variable condenser here described has the following important advantages : (1) Consistency of resetting is obtained by the use of conical bearings. (2) Great accuracy of setting and reading is attained in the manner already described.(3) Freedom from external capacity effects is ensured by the earthing of the moving system adequate screening and remote control. (4) The calibration curve is free from the irregularities that arise in the usual vane type of condenser through the effects of the spacing washers and supports. ( 5 ) A simple adjustment of the range of capacity within wide limits is attained by alteration of the positions of the eborrite platforms T (Fig. 3). The complete assembly of the apparatus is shown diagram-matically in Fig. 1. The dielectric container C, and the wide-range condenser C3 are connected in parallel with an inductance, L3 which for frequencies of the order of 106- per second consists of 30 well-spaced turns of No.16 S.W.G. copper wire wound upon a spirally-grooved ebonite tube 9 cm. in diameter. L and L, are “honeycomb ” inductances of the type commonly used in radio-telegraphy; C is a variable condenser of the ordinary vane type having a maximum capacity of approximately 300ppF ; and C is a mica-dielectric “ by-pass ” condenser of approximately 0.01 yF. B is a dry-cell battery of 50 volts and B a 4 volt. 100 amp.-hour accumulator ; G is a uni-pivot galvanometer having a sensi tivity of about 1pA. per scale division. R, a potentiometer of 5000 and R, a resistance of 9000 are adjusted so as to neutralise the greater part of the anode current through G (Dowling Zoc. cit. p. 175). Although the construction of Cv was adapted to secure a recti-linear relationship between scale reading and capacity it was intercalibrated throughout its range by substituting for Ca a very small fixed-capacity condenser the capacity of which was measured in terms of scale divisions of C at a large number of places along the scale of the latter.If the “ scale-division capacity ” ratio had been truly rectilinear the apparent value of the very small condenser would have been precisely the same at all points alon JIESSURISG THE DIELECTRIC CONSTAKTS O F LIQUIDS. 321 the scale. The apparent values obtained gave the data necesaary for the preparation of a correction-curve to express all readings in terms of “ mean scale divisions.” Finally the capacities of the six fixed-capacity condensers C to Cf were measured in terms of scale divisions of Cv (for C = (‘v C 2 Ca + C, C 5 Ca + (3 t C’v, and so on).The capacity corresponding to a mean scale division of C is thus the unit in all measurements. Having thus intercalibrated C, the capacity of C, coiitainiiig dry air a t a known pressure and a t the temperature of the ther-mostat is measured in terms of C,. To do this C is switched out of circuit and with C a t a large setting L,C is adjusted to approximate resonance with L,C,. Exact resonance between t 110 two circuits is then obtained by varying C’ and is shown by a sharp minimum reading of the galvanometer G and the reading of C ‘ , is taken. C is then switched into circuit and C is reduced in value until resonance i q again obtained between the two circuits. This reduction in C is a measure of the capacity of Clcab).Finally, C4 is filled with the liquid under examination (at the temperature of the thermostat) and its capacity measured as beiore. The latter capacity C(q(liquid) divided by the former C4(sir) gives the dielectric constant of the liquid to a standard of 1 1, at the temperature and frequency of the measurement. Shortly lieforc or after the measurement the frequency is determined by means o€ a heterodyne waTY-e-meter (Sayce Expt. Wireless 1923, 1 70). The value of the dielectric constant so obtained can lie corrected to a vacuum standard by making use of recent deter-minations of the specific inductive capacity of eir (Carman anti Lorance Physicul Rev. 1922 20 715; GiIl Radio lieu. 1921 2, 450; Wagstaff Phil. Nag. 1924 [vi] 47 66).Experimentul Difficulties and Precautions.-When fist applied , the method gave inconsistent results due to slight temperature changes in the valve in the oscillator coils and particularly in the variable condenser C,. Sudden changes in the temperature of the valve were prevented by enveloping it in cotton-wool. The influence of sunlight upon the apparatus was so marked that it was found desirable to obscure the windows of the laboratory and illuminate the apparatus by artificial light. Ultimately the air of the room was kept a t a fairly constant temperature and as an additional precaution the condensers C to Cz were intercalibrated with C before and after each determination. Results.-The following results obtained with benzene are given, not as having any absolute significance but simply as an example of the degree of consistency so far observed in repeated meamre-ments on the same material.This we regard as a n importan 322 FAIRBROTHER AND MASTIN : criterion of the precision of the method. The benzene was rigorously purified by fractional distillation and fractional crystallisation and was finally distilled over phosphorus pentoxide but without any attempt to attain intensive drying. Frequency 65 x 103 cycles per second. E (air = 1) ............... 2.2380 2.2392 2.2396 2.2389 Dielectric Constant of Benzene a t 25.5". Mean 2.2389 Diff. from mean ......... -0.0009 +0.0003 +0.0007 -0.0001 Mean difference from mean &0.0005. It is interesting to make a comparison as to consistency between our results and those of Turner (2.physikaZ. Chem. 1900 35 385), who using the method of Nernst (ibid. 1894 14 622) applied to carefully purified benzene obtained the following data which are commonly cited as the most accurate available. The frequency was low but unspecified and the results were corrected to 18". F ............... 2.290 2-291 2.285 2.288 2.293 2.292 2.286 Diff. from Mean 2.289. mean ...... $0.001 +0*002 -0.004 -0.001 +0-004 +0.003 -0.003 Mean difference from mean &O-003. During our measurements no special precautions were taken to maintain room temperature constant and there is much evidence that even the small differences here recorded are due to small changes in the temperature of the various parts of the apparatus. Hence it seems probable that under more constant temperature conditions still greater precision may be attained by the method. Note.-Since the completion of the above measurements a further contribution to the measurement of dielectric constants has been made by Grutzmacher (2. Physik 1924 28 342). Results for benzene stated to the fourth decimal place and from internal evidence probably significant to the third decimal place are there given but as only one result is recorded for each temperature we are unable to compare the consistency of the measurements with that obtained by our method. ARMSTRONG COLLEGE, NE WCASTLE -UPON -TYNE. [Received October 1 Ith 1924.