Thin-film thermocouples for differential thermal analysis

problems from changes in heat transfer coefficient between sample and sensor. This results in very flat base lines and highprecision measurement. It i...
0 downloads 0 Views 637KB Size
Thin Film Thermocouples for Differential Thermal Analysis W. H. King, Jr., and C . T. Camilli Esso Research ana' Engineering Co., Linden, N. J.

A. F. Findeis University of Alabama, University, Ala. Thin film thermocouples supported on quartz were developed for differential thermal analyses (DTA) over the range -125 OC to 500 OC. The low mass of the thermocouples, the small sample size, together with other design features practically eliminates problems from changes in heat transfer coefficient between sample and sensor. This results in very flat base lines and high precision measurement. It is this high degree of repeatability that distinguishes this DTA system from others. The system allows calorimetric values to be obtained from peak areas. Calibrations based only on indium yielded calorimetric data which agree well with the established literature.

PRECISE MEASUREMENT of thermal transients by the technique of Differential Thermal Analysis (DTA) has recently attracted a number of investigators. I n this paper, a new approach to DTA is described which results in the highly reliable nieasurement of thermal processes taking place in a wide variety of samples. This approach involves the use of vacuum evaporated thermocouples o n a quartz substrate. These thermocouples provide a sensing element possessing a number of advantages over previously described differential measuring devices. Thermocouple fabrication by the deposition of thin metal films has been described many times. Harris and Scholp ( I ) have made single and muhijunction thermocouples by sputtering bismuth-antimony and bismuth-tellurium on thin cellulose ester substrates (films). Rapid response thermocouples for use as radiation detectors in the infrared region of the spectrum have been described ( 2 ) . Hornig and O'Keefe (3) have thoroughly discussed the design of fast thermocouples for use with interrupted radiation in infrared spectrometry. The work reported herein is the first known application of such rapid response thermocouples to DTA. Differential thermocouples, thermistors, and resistance thermometers as applied to DTA have been described. Different problems are associated with each sensing method and the problems vary with the temperature range of the measurement. Ordinary differential thermocouples can readily be made, but precise matching of the junctions is difficult. Precise DTA measurements depend largely on matched thermocouples. Intimate coupling of the sample to the sensor is of equal importance. Thermistors may be matched precisely only over limited temperature ranges, and their upper temperature limits are severely restricted. Resistance thermometer techniques for DTA measurements imply that matched resistance thermometers be fabricated, and this is indeed difficult. Furthermore, the differential sensor should ideally contribute no heat to the system. Resistance thermometers fabricated for small sample holders and incorporated in the usual differential bridge type of circuit fail in this regard because of their necessarily small dimensions. The very nature of the resistance (1) L. Harris and A. C. Scholp, J. Opt. SOC.Am., 30, 519 (1946). (2) L. C. Roess and E. N. Dacus, Reu. Sci. Znstr., 16, 164 (1945). (3) D. F. Hornig and B. J. O'Keefe, ibid.,18, 474 (1947).

1330

ANALYT~CALCHEMISTRY

change with temperature results in a nonlinear response. Some of these disadvantages have been overcome in the nullbalance design of a differential scanning calorimeter by Watson et a / . ( 4 ) . The ideal sensing element should have high thermal impedance paths for all possible modes of heat flow except for that to the sensing elements, and the sensing elements for sample and reference should match perfectly. This ideal would require that the sensor, sample, and sample holder have n o mass; n o leads should be connected to the sensing element, and heat flow to the sample should be by a high thermal impedance path. This report presents a method of fabrication of a sensing element having nearly ideal characteristics for use in the technique of DTA. Light-weight, precisely matched thermocouples were produced by the evaporation of thin films of dissimilar metals which overlap to form the junctions. I n order to maintain high purity gold films, the vacuum evaporation equipment was used for gold only. Thus, possible contamination from other metals in the system was reduced. Nickel was subsequently electroplated onto the gold to form the thermocouple junction. I n this way precisely matched differential thermocouples may be fabricated. Gold and nickel have a favorable thermoelectric power which is intermediate between the standard Pt-Pt (10z Rh) and the chromel-alumel pairs. EXPERIMENTAL

Quartz slices, AT cut, 0.66 x 0.55 x 0.027 inch, similar to those used in quartz crystal oscillators for frequency control, were used as a substrate for the thermocouples. The quartz slices were first lapped with No. 500 carborundum abrasive. They were then cleaned in an ultrasonic detergent bath followed by a wash with water and finally with reagent grade methanol and then air dried. The quartz slices were then masked with aluminum foil and placed in a vacuum deposition apparatus to receive a gold film. The gold deposit on the quartz slice was then electroplated with nickel from a stirred bath containing 280 grams of NiS04.7Hz0, 30 grams of H3B03, and 20 grams of KF per liter of solution. A pure nickel anode was used in the plating bath. The thickness of the nickel plate was carefully controlled with a coulometer. The gold deposit was usually around 100 pg/cmz and the nickel plate was about 200 pg/cm2. Both nickel-goldnickel and gold-nickel-gold differential thermocouples were produced by this procedure. The sequence of steps in the fabrication of the thermocouples is shown in Figure 1. The thermocouples were electrically connected by the appropriate nickel or gold wires to avoid an additional thermal junction. A small drop of silver print paint was placed on each connection t o ensure good electrical contact. Copper wires were most often used in place of gold because the thermal E M F between C u and Au is small (0.15 pV/"C>. Both glass microscope slides and quartz plates were very satisfactory as substrates. Both are excellent insulators and

(4) L. S. Watson, M. J. O'Neill, J. Justin, and N. Brenner, ANAL. CHEM., 36,1233 (1964).

2

1

T"c

SWING SUPPORT ROD

HOLE FOR TEMPERATURE PROGRAM THERMOCOUPLE

M A S K AND E V A P O R A T E GOLD

ELECTROPLATE NICKEL

1 w-

3

\

-,

ALUMINUM OVEN -CYLINDER

_COOLING COIL AND ELECTRICAL HEATER

2-1'2'1 FIXED SUPPORT ROD

Figure 2. Temperature programmed oven Sensing elements and sample pans are attached to a fixed support. The oven assembly swings into position pendulum fashion

M A S K AND E V A P O R A T E GOLD Au

Ni

Au

Figure 1. Steps in making thermocouples good adherence of the metal films was obtained by roughening the surface. Quartz is free of transitions up to 575 "C. Mica was not satisfactory because adhesion was poor. I t was thought that the thickness of the quartz plates would affect results by providing better insulation under the thermocouple. This was tested by adding extra insulation (up t o 0.25 inch) under the quartz plate. The added insulation had n o observable effect on peak shape or amplitude in n-Cat transitions. Aluminum Sample Pans. The sample pans are considered a n important part of the system; the Perkin-Elmer DSC pans, covers, and pan crimper were used. These pans are 0.25-inch diameter aluminum and together with a cover weigh 21.10 mg with a standard deviation of 0.14 mg. Some samples in the form of films can be deposited on the thermocouples and the pan eliminated. Tests were made with and without pans. Precision is best when the samples are enclosed in the pans. Sample weights from 0.5 to 25 mg were tested and it was found that 5 t o 10 mg is a good compromise between resolution, signal strength, and repeatability. Making the pans from a better conductor than aluminum was considered ; however, materials with higher conductivities also have higher heat capacities, thus offsetting any advantage. Further, a subsequent test showed the aluminum pans to have excellent heat distribution properties. Oven Design. The design and shape of the oven is not critical. The first oven design tested was made from a tin can 2.5-inch diameter by 1.75-inch height. An electrical heater was wound on the outside of the can and then covered with 0.5 inch of asbestos insulation. A rectangular slot in the can permitted entry of the junctions. A more sophisticated design made from a 2.5-inch diameter X 2.5-inch long aluminum cylinder is shown in Figure 2. N o difference in performance could be detected when the center hole was 0.75-inch diameter or 1.5-inch diameter. A third design having a 0.5 X 0.74 X 1.5 inch rectangular shaped hole in an aluminum block was also used. The oven experience indicates that the thermal impedance between the oven walls and the samples is very high and does not change much with the hole size. The high thermal impedance coupling of the

oven to the sample is an important design consideration of this DTA system. The advantage of not using a heat sink t o homogenize the temperature increases sensitivity. This was pointed out by Mazieres (5,6). The oven designs allow programmed temperature operation from -120 'C t o over 500 "C. Cooling coils were fabricated from 0.25-inch diameter tubing. The oven was cooled by applying a partial vacuum to one end of the stainless steel coil and drawing cool nitrogen vapor and, subsequently, liquid nitrogen through the coil. By controlling the flow of nitrogen vapor either up or down, linear temperature programming was possible. OVERALL SYSTEM Figure 3 shows the electrical circuits for two of the systems used in this work. The first system used for calorimetric data employs a two pen strip chart recorder. A chart speed of 1.5 inches per minute was used. One pen records the differential temperature signal (AT) obtained from the Leeds & Northrup 9835-B amplifier. The other pen records the temperature signal ( T ) from a small chromel-alumel thermocouple placed near the differential junction. A Fincor Power Package programs the oven temperature from a cam cut t o provide linear temperature rates. The system, except for oven and thermocouples, was obtained as a package from the Leeds & Northrup Co., Philadelphia, Pa. The second system shown in Figure 3 was used t o obtain thermograms displaying differential temperature as a function of reference or sample temperature. This plot is convenient for research work because the same size graph is obtained regardless of heating rates. In this setup, a n X-Y recorder was employed, and the oven temperature was controlled by an F & M Model 240-M Linear Temperature Programmer. PERFORMANCE The thermocouples tested in the absence of sample holders or samples showed a drift of 0.5 pV from room temperature to 400 "C. With temperature raise rates of 200 "Cper minute, base lines were straight. Under normal run conditions of 10 OC/minute, a base line drift was typically 2 x 10-9 volt/ (5) C. Mazieres, Compt. Rend., 248, 2990 (1959). 36, 602 (1964). (6) C. Mazieres, ANAL. CHEM., VOL 40, NO. 8, JULY 1968

0

1331

'r------1

20

I

r-------I

I Oven I S

Au

IR

Au

Assly

I

1 Fincw Paver Package

NI

I

1

15

10 I

1 I

.Cu

I

NI Cu

- I

u

2

cu

D

9035 6. Ampllfler

0 0

50

100

150

200

TEMPERATURE, 'C

Figure 4. Thermocouple calibration

-

ATvs. T

Because thermoelectric power changes with temperature, calorimetric measurements must be appropriately corrected for this effect

Temperature Programmer

Figure 3, Electrical hook-up of DTA apparatus for strip chart and X-Yrecorder presentations

OC. In the few cases where drift and noise appeared, the trouble was usually due to poor electrical contacts. Drift is much less with this system than has been obtainable with the conventional type of differential thermocouples. The improvement is probably due to the exact match of composi-

0

2

4 6 8 10 SAMPLE WEIGHT MILLIGRAMS

12

14

Figure 5. Peak areas are linear functions of sample weight This is a good indication for a quantitative system

1332

0

ANALYTICAL CHEMISTRY

tion of the junctions resultjng from the method of fabrication. The noise level of the recorder trace could be maintained at 0.05 MVwith a time constant of about one second. The voltage response and the thermoelectric power of the thermocouple are shown in Figure 4. Temperature measurements were made with a chromel-alumel thermocouple placed adjacent to the gold-nickel thermocouple in the DTA oven. The oven was programmed at a rate of 2 OC/minute and the data were recorded with an X-Y recorder. The thermoelectric power was calculated from the slope of the goldnickel thermocouple L'S. temperature curve. The thermoelectric power of the gold-nickel junction increases with temperature and must be known for accurate calorimetric measurements with this system. The areas under DTA thermograms were shown to be a linear function of sample size for a wide range from 0.5 to 25 mg. A small positive deviation from linearity for exothermic reactions and a negative deviation for endothermic reactions would be predicted from the shape of the thermoelectric power curve, but area measurements with the planimeter are not precise enough to observe such deviations. A response curve for the melt transitions in indium and n-dotriacontane are shown in Figure 5 . The deviations from the curve were subsequently found to be the result of errors in the measurement of the mass of the sample and in area measurements with the planimeter. Repeat measurements on the fusion of a precisely weighed sample of indium resulted in area measurements reproducible to = t l . 2 %for one standard deviation. Numerous investigators have noted the utility of peak area measurements for calorimetric values from DTA thermograms (7). This approach has been strongly criticized by the classicists as thermometric rather than calorimetric. The authors choose the calorimetric term. With the sensing element described in this paper, precise calorimetric values may be obtained. The calorimetric values reported here were obtained using indium as the only standard. Area measurements were made with a K & E compensating polar planimeter. The thermoelectric power of the gold-nickel thermo(7) E. M. Barral, R. S. Porter, and J. F. Johnson, ANAL.CHEM., 36, 2172 (1964).

ANNEALED 4th RUN

f

ROTATION'

4*C/MIN.

F ANNEALED 3rd R U N

I DiRECT REPRODUCTION OF HEXATRIACONTANE ENDOTHERMS FROM RECORDING CHART: b.7 mg. SAMPLE 1 0 ' C h i n RATE x = 200 yV/in. (T), y 5 p'/in. UT)

-

Figure 6. AT us. T thermogram of (1) (2) (3) (4)

at two heating rates

Premonitory change Shoulder due to monoclinic to orthorhombic transition Orthorhombic to hexagonal transition Melting

couples is an increasing function of temperature, thus a correction must be applied based on the ratio of the thermoelectric powers at the transition temperatures of the sample and the standard. In the case of indium, the thermoelectric power at the fusion temperature is 22.6 pV/"C while that at the fusion temperature of n-dotriacontane is only 16.2 pV/"C. Thus, for an equivalent response, the peak area for indium would be 22.6/16.2 times as large as the n-dotriacontane area. Incorrect values for n-dotriacontane would result without such a correction. This correction is not unique to the goldnickel system. Chromel-alumel thermocouples, in the range of 100400 "C have a relatively constant, thermoelectric power. Thus, the correction is often ignored since it is small. At lower temperatures-Le., - 100 "C-the thermoelectric power of the chromel-alumel pair is lower by about 30% ; hence, such a correction is necessary. The problem of area measurement deserves further comment. Planimeter measurements are precise for well defined areas and 'reproducibilities of 0.5% are easily achieved. Precision deteriorates when decisions about the establishment of base lines or the resolution of over-lapping peaks are involved. The measurements can be made more reproducible from observer to observer only if one defines precisely both where to draw the base line, and the method of resolution of peak areas for overlapping transitions. A heat capacity change is usually observed in melting transitions; hence, the base line is not continuous. The base line both before and after the transition must be extrapolated. The first deviation from the extrapolation on both the lower and higher temperature ends of the transition defines two points between which the base line should be drawn. The points may also be established by a preliminary run at much greater sensitivity. In the case of overlapping transitions, the valley between the transitions was divided on the basis of the ratio of the peak amplitudes of the transitions. The region of overlap was divided assuming the trailing edge of the first transition was

Figure 7. AT os. T thermograms of a sample of n-Ca6 Slow cooling after the first run removes the pressure peak and the curves show a very high degree of repeatability. This figure is a direct copy from the recorder chart so the noise level and flatness of the base line are revealed.

identical in normalized shape with the trailing edge of the second transition. The value obtained for the heat of fusion of n-dotriacontane from the line average of Figure 5 was 38.9 cal/gram. More precise measurements on carefully weighed 7-9 mg samples of n-dotriacontane yielded a value of 38.4 + 0.46 cal/gram. This value is consistent with that found by others (4). The heat of fusion of silver nitrate was determined to be 16.29 cal/ gram, a value in excellent agreement with an accepted value of 16.216 (8). Thermograms of n-dotriacontane and n-hexatriacontane show the existence of a pre-rotational endotherm. This was described by Broadhurst as the monoclinic-orthorhombic transition (9). The presence of this transition is strongly dependent on the thermal and mechanical history of the sample as well as purity. This endotherm appears in freshly recrystallized dotriacontane and can be revealed again by thermal or mechanical treatment of the sample. A sample, after rapid quenching, upon reheating shows this prepeak endotherm. Squeezing the sample pan with the crimper will also cause it. This transition was not observed on a slow freezing cycle, nor is it observed on a heating cycle following the slow cooling. The prepeak transition can complicate calorimetric measurements of the rotational transition because the two are not always well defined. Thus, the ratio of melting to rotational transitions in samples prepared from crystalline n-dotriacontane was 2.46 0.17 when programming up in temperature, and 2.87 + 0.10 after the sample had

*

(8) K. K . Kelley, U. S . Bureau of Mines Bulletin 601, p 105, U. S. Government Printing Office, Washington, D. C., 1962. (9) M. G. Broadhurst, J. Res., Nail. Bur. Sfd.,66A, No. 3, MayJune 1962. VOL 40, NO, 8, JULY 1968

0

1333

o b

I 5

I

10

I 15

I

I

I

20

25

30

HEATING RATE W M I N .

Figure 9. Ratio of peak area to heating rate is almost a constant

Figure 8. n-Cs2thermogram of reference temperature US. A temperature showing the effect of heating rate on size, shape, and peak positions

been melted and slowly cooled. The latter value is in good agreement with the average ratio of 2.78 reported previously ( 4 ) . The calorimetric value for the rotational transition (orthorhombic-hexagonal) is thus 13.4 cal/giam, a value in excellent agreement with Differential Scanning Calorimeter ( 4 ) data. A value for the pre-rotational transition, based on the average ratios of the melt and rotational transitions, is approximately 2.2 cal/gram. Thermograms of n-dotriacontane a t two heating rates are shown in Figure 6. The presentation is from the X-Y recorder hookup of Figure 3 where the reference temperature drives the X axis and the temperature difference drives the Y axis. These are direct photographic reproductions of the original recordings to ensure no loss of detail. In the 10 "C,'min run, there is a slowly increasing departure from the base line before the first endotherm. This departure starts 10 "C before the first transition and probably corresponds to the premonitory density change described by Vand (IO). Vand showed that the density of n-C38 decreased 0.4% for 10 " C at low temperatures but from a point 10 " C below the first transition, the density fell 8% for this 10 "C interval. The 4 "C/min thermogram doesn't show the premonitory changes as distinctly, but a t the point marked 2, there is a shoulder which is probably the monoclinic to orthorhombic change noted by Broadhurst (9). These endotherms and subtle changes are not observed in the usual DTA equipment. They (10) V. Vand, Actu Cryst., 6,797 (1953).

1334

ANALYTICAL CHEMISTRY

are observed here because of the unusually flat base line. The endotherm marked 3 is the well known orthorhombic to hexagonal where the the molecules rotate in a crankshaft fashion, and the wax is mushy. Endotherm 4 is the melting of the hexagonal state. The existence of a pre-rotational peak in other normal paraffins has also been observed. For example, it is readily observed in n-hexatriacontane on the first run with a sample in a crimped container. The peak subsequently disappears in succeeding thermograms of the same sample after 10 "C/ minute cooling. This n-CaGpre-rotational peak may also be made to appear again by quenching o r recrimping the same sample. This is shown in the repetitive scans in Figure 7. These four endotherms were recorded with the X-Y technique and show several important points in addition to the odd behavior of the first run. Some idea of precision can be had by measuring the amplitude of the melt transition; these are 6.10, 6.09, 6.10, and 6.12 inches. The rotation peak is 5.27, 5.44, 5.43, and 5.43 inches. The rotation peak is not only smaller o n the first run, but it maximizes 0.5 "C lower than runs 2, 3, and 4. The missing energy is most likely in the pressure induced peak. The premonitory change mentioned before is readily observed on all four runs. In runs 2, 3, and 4, there is a small endotherm equal to 0.6% of the melt peak that occurs just prior to the beginning of the premonitory change. The cause of this tiny endotherm has not been discovered. HEATING RATE VS. RESPONSE

An increase in heating rate results in a n apparent sharpening of the transition if one records the differential temperature us. time. The increased rate actually makes the resolution of peaks poorer, which becomes apparent when one records the differential temperature us. reference temperature. I n Figure 8 the broadening of peaks with increased heating rate is observed together with movement of the peaks to remove apparent higher temperatures. At first glance one would not expect the peak position to change with heating rate, but the peak maximum is the point a t which the melting is completed. The point where melting starts is obtained by extrapolating the leading edge to the base line. The reason there appears to be a width to the melting point is due to the method of recording reference temperature L'S. temperature

difference. When the melting point is reached, the sample temperature stops a t the melting point but the reference continues to advance without interruption, thus both X and Y axes move during the transition. The area of the peaks must increase directly with the heating rate because the transition must absorb the same quantity of heat regardless of the speed. To absorb this energy in a shorter time, the temperature difference between sample and reference must be greater, thus the peaks are higher a t higher rates. In Figure 9 the peak areas were divided by the heating rates. Agreement to the expected horizontal line is reasonably good. The amplitude of the peaks is a linear function of the square root of the heating rate, and excellent agreement with a straight correlation line was obtained. All data points fell within 1 of the line. This correlation is better than that of the area measurements because the peak amplitude can be measured more precisely than peak area. Agreement of heating rate data with theory provides an extra measure of confidence in the calorimetric values determined with this new DTA system.

CONCLUSIONS

This new DTA probe has been successfully applied to waxes, polymers, inorganic salts, metals, and a wide variety of other materials. No correlation is necessary for response due to heating rate, and calorimetric data superior to any previously described D T A system may be obtained by direct area measurement followed by a correction for thermoelectric power. By using modern metal and insulator evaporation techniques, thin film thermopiles 5 to 10 deep can be made to improve sensitivity. In addition, the Differential Scanning Calorimeter (DSC) experiment is made possible by adding a thin film resistance heater under each thermocouple. It is believed that the new thin film system constitutes an approach to DTA superior to previously reported techniques.

RECEIVED for review February 23, 1968. Accepted April 24, 1968. Paper No. 172 presented April 5 , 1968, at 155th National ACS Meeting, San Francisco, Calif.

Quantitative Comparison of Powder Diffraction Patterns L. K. Frevel and C. E. Adams Chemical Physics Research Laboratory and Computation Research Laboratory, The Dow Chemical Co., Midland, Mich. A straightforward theoretically sound procedure i s described which permits analysts to compare diverse powder diffraction patterns in a quantitative nonsubjective manner. Criteria are established for valid matches of interplanar spacings and of their corresponding peak intensities. Fortuitous superpositions of two or more diffraction lines from two or more phases are resolved by an iterative correction. A computer program written i n Algol 60 for the Burroughs B 5500 yields a print-out of the stipulated matches with the original diffraction data and another print-out of the decomposed pattern corrected for any superposition of lines.

THEMOST DIRECT WAY to identify a single crystalline phase is to compare its powder diffraction photograph visually with those of certified standards exposed in the same camera under nearly identical conditions. Most analytical laboratories, however, do not possess a comprehensive file of standard films and thus have to rely on published data. Comparing a particular powder pattern with published patterns usually is not a straightforward procedure because of the wide variation in the relative merits of the diverse literature data. Table I summarizes the eleven modes of comparison practiced by diffractionists. With the recent development of automated photometry coupled with computerized processing of digitized data ( I ) , it is now possible to carry out objective quantitative comparisons of powder patterns with published patterns. THEORY

spacings and one for the corresponding intensities. The manner in which a merit rating is established for the interplanar spacings of a powder pattern becomes evident from Criterion 1 for a match of d, with d,

d,

- Ad,

- Ad, 6 d,,, 6 d,

Ad = dZ(4X-2 - d-2 11 1 2 A 6

film

numeric numeric /visual I numeric

film trace

(1) L.K. Frevel, ANAL.CHEM.,38,1914 (1966). (2) Ibid.,37,471 (1965).

...

numeric {:Eric

numeric numeric numeric

(3,))

then each pattern is assigned a merit rating for the interplanar

(2)

Table I. Methods of Comparing Powder Diffraction Patterns Form of diffraction data Type of comparison Unknown Standard d, A I , or I , film film visual visual film trace numeric visual film numeric numeric

trace

{

(1)

where d,,, is the cthd-spacing for standard s. The allowable error for any interplanar spacing d is given by Equation 2

trace (calc)

If {dv,Zv]represents the digitized data (2) of the powder pattern to be compared with a standard pattern d,,

+ Ad, + Ad,

{dv,IvI

9. i

(x>,l

(?i$ric numeric

numeric

...

numeric jvisual \numeric numeric

(machine(machine(print(printreadable) readable) out) out) d = interplanar spacing, A. I p = peak intensity of diffraction line in arbitrary units. I i = integrated intensity in arbitrary units.

VOL. 40, NO. 8, JULY 1968

1335