RAPID RESPONSE THERMOPILES

which Hecht's theory does not explain are thosepertaining to protanopia.

The demonstration by Granit of the existenceof modulators with widely separated sensitivitybands is obviously contrary to the assumptionthat Hecht's curves represent the sensitivitiesof individual cones. However, Hecht's theory isthe only one which gives an adequate descriptionof most color data (all except protanopia). If weconsider Hecht's curves in terms of sensationrather than in terms of cones, all contradictionsare removed and even protanopia can be ex-plained. Hecht's red, green, and blue curvesmay be considered as the curves for the sensationproduced by the combined action of the domi-nators plus any one of the three types of 'modu-lators. Hecht's curves then become the result ofthe fused action of both dominators and modu-

19 G. L. Walls, The Vertebrate Eye (Cranbrook Instituteof Science, 1942).

lators as measured at some higher center in thecentral nervous system. The mechanism of thisassumed fusion is not necessarily different fromthat of the fusion of the effect of any group ofmodulators to give the sensation of white orof some band of color. The explanation ofprotanopia offered under the Granit theory thenautomatically covers the phenomenon under theHecht theory.

The above combination of modern theoriesof color vision is apparently capable of explainingall of the basic types of data concerning colorvision and the various types of color blindness.Protanomaly and deuteranomaly are examples ofpartial protanopia and deuteranopia. The twotypes of total color blindness (achromatopsia),as pointed out by Granit,3 may be caused by acomplete absence of either cone or of modulatorfunction.

JOURNAL OF THE OPTICAL SOCIETY OF AMERICA VOLUME 36, NUMBER 10 OCTOBER, 1946

Rapid Response Thermopiles*

LOUIS HARRIS

Department of Chemistry, Massachusetts Institute of Technology, Cambridge, Massachusetts(Received June 17, 1946)

A technique has been developed for the evaporation of bismuth and of antimony upon thincellulose nitrate so that the metals have comparatively low electrical resistivities and com-paratively high thermal e.m.f.'s. A procedure has been developed for making bismuth-antimonyradiation thermopiles in which units of 50 thermal junctions have a receiving area of -0.O1 1 cm2

and a resistance of -70 ohms. These units are called "folded thermopiles." The response ofseveral folded thermopiles of different design operated at atmospheric pressure has been studiedover a range of frequencies. The "folded" thermopiles are faster than any thermopiles reportedheretofore, yet the response is greater for all frequencies above 5 cycles per second than forevaporated thermocouples operated in a vacuum. The thermopiles are rugged, relatively free ofmicrophonics, and show very little noise in addition to "Johnson noise."

INTRODUCTION

THE speed of response of a radiation thermo-T couple is greater (1) the smaller the thermalmass and (2) the greater the heat loss. Sputtered1

and evaporated2" thermocouples with speeds

*The work described here was carried out in whole or inpart under contracts OEMsr-126 and OEMsr 1147 betweenthe Office of Scientific Research and Development and theMassachusetts Institute of Technology.

I L. Harris and A. C. Scholp, J. Opt. Soc. Am. 30, 519(1940).

2 A. Stockfleth, M.I.T. Ph.D. Thesis (November, 1942).3 L. C. Roess and E. N. Dacus, Rev. Sci. Inst. 16, 164

(1945).

high compared to wire thermocouples have beenobtained by making the thermal mass as small aspossible. Further increase in speed of response,through decreased thermal mass, awaits the dis-covery of new materials and/or methods of ob-taining thinner deposits of metals having morefavorable electrical and thermal properties. Thegain in speed through increased heat loss wouldappear to have little advantage since the,4 volt/i watt/cm2 would decrease accordingly.However, it has been possible to achieve highheat loss and therefore higher speed of response

597

LOUIS HARRIS

TABLE I.

Thermal e.m.f. vs. copperResistivity (olm-cm) (microvolts per degree C)

'IhicknessesMetal (microns) average limits average limits

Bismuth 0.73 to 1.58 168X 10-6 131-212X10- 6 -57.5 -55.1 to -59.01.64 to 2.14 159 X 10- 6 -56.12.05 to 3.15 130X10 G -56.0

Antimony 0.48 to 1.66 95X10-0 60-119X10- 6 35.5 33.1 to 38.6

and also comparatively high Ai volt/,g watt/cm2 .This was accomplished by using thicker metaldeposits, so that most of the heat loss occurred bymetallic conduction, and by crowding manythermocouples into a small receiving area.

We shall describe in detail the construction andresponse tests of a fifty-junction thermopilehaving a receiving area 0.25 cmXO.44 cm. Eachelement is a single junction thermocouple formedby evaporating bismuth and antimony on oppo-site sides of a thin folded cellulose nitrate film.The overlapping junction of the two metals is atthe fold and is the "hot junction." Fifty elements(0.25 cm wide) are assembled in a miniature pressso that the thermojunctions, one above the other,form the receiving area. Electrical contact amongthe elements is made mechanically at approxi-mately 0.030 cm behind the fold. The contactlines form the cold junctions. Figure 3 shows thestructure of the thermopiles whose method ofmanufacture will be described in the next section.

CONSTRUCTION OF THERMOPILES

Method of "Evaporating" the Metals

Bismuth and antimony were deposited by"evaporation" on to cellulose nitrate by thefollowing procedure. A weighed quantity of thepure metal '(previously cast in vacuum) was evap-orated from a "boat"; the current through theboat was increased as the evaporation proceededso that the boat was at red (for antimony) oryellow (for bismuth) heat at the end of theevaporating process. Alongside the boat con-taining the metal to be evaporated was anotherboat (empty) maintained at red (for antimony)or yellow (for bismuth) heat, during the evapora-tion process. The evaporation of each metal wascarried out stepwise, so that about twenty-fivenliLutes elapsed for a complete evaporation. Avacuum of approximately 10- mm Hg was main-

tained during the evaporation process. Tantalumboats were used for the bismuth evaporation;molybdenum boats were used for the antimonyevaporation. The progress of the evaporationwas followed with a "meter." The meter wassituated in the bell jar directly above the boatfrom which the metal was evaporated.

The detailed procedure used for the evapo-ration of antimony has been described.4

We found, in agreement with Bussem, Gross,and Hermann,5 that it was necessary to form thebismuth deposit at an elevated temperature, inorder to obtain electrical resistivity not greatly inexcess of the macroscopic value. We obtained theelevated temperature by radiation, by operatingthe empty boat adjacent to the bismuth sampleboat at "yellow heat." Films of cellulose nitratewere occasionly decomposed by the high tempera-ture maintained.

Measurements of Electrical Properties ofEvaporated Bismuth and Antimony

Early experiments showed that the amount ofmetal deposited and the resistivity of the de-posited metal varied with the evaporation tech-nique as well as with the material upon which themetal was deposited. However, the results givenbelow are for the metals deposited upon thinfilms of cellulose nitrate. The cellulose nitratefilms were mounted on mica frames to facilitatehandling. The films were placed in the bell jar, inhorizontal position, about 15 cm above the boats.After completion of an evaporation the films wereremoved from the bell jar, leads were soldered totwo ends of the evaporated metal with Wood'smetal, and the resistance of each strip was de-termined. A strip about 2.2 cmX1.1 cm of the

L. Hlarris, J. App. Phys. 17, 757 (1946).W. Bussem, F. Gross, and K. Hermann, Zeits. f. Physik

64, 537 (1930).

598

RAPID RESPONSE THERMOPILES

metal with its backing was then cut out of themica frame and weighed. The metal with itsbacking was then mounted on a piece of mica andthe thermal e.m.f. vs. copper was measured. Astrip of the cellulose nitrate used had beenweighed previously so that the resistivity of themetals could be calculated. (The mass of thecellulose nitrate was about 0.35 X 10-3 gramwhile the masses of the metals varied from0.7 X 10-3 gram to 4 X 10- 3 gram. Massive densitywas assumed for the metals.)

The results obtained, using the techniquedescribed above, are found in Table I.

The bismuth deposits had a dull light-greyappearance; the antimony deposits were a bril-liant steel-grey and showed a crystalline patternto the unaided eye.

Burger and Van Cittert6 obtained a thermale.m.f. for evaporated bismuth-antimony of75 ,4v/0Cg. The values obtained here are about 25percent greater.- The lowest resistivity measurements for evapo-rated bismuth are those reported by Bussem,Gross, and Hermann5 ; their value of 160X-106ohm-cm is nearly the same as ours. Each value isabout 30 percent greater than that of the massivemetal.

The resistivity of evaporated antimony ob-tained here is about 110 of that reported byStockfleth.2 The resistivity for evaporated anti-mony is still about 21 times greater than that ofthe massive metal.

The Backing Support for the Metals

The backing support for the metals of thefolded thermopiles was a cellulose nitrate filmformed from a solution of "600-second" cellu-lose nitrate. The solution is poured on to aglass plate upon which it is spread evenly by aspreader bar or "knife." After the solvent has

evaporated, the films on the glass plate are cutinto strips 9 cmX 1.2 cm and then floated off onwater ready to be mounted.

Mounting or Folding the Cellulose Nitrate

In order to obtain a uniform fold of thecellulose nitrate film it is necessary to fold the9X1.2 cm strip over a "bar," the bar beingremoved after the fold has set. Various plasticmaterials were tried for the bar. (The plasticscould be removed with solvents to which thecellulose nitrate is inert.) However, better resultswere obtained by using a Nylon thread held tautby a spring, the thread being withdrawn me-chanically after the fold had set.

All subsequent operations are facilitated bymounting the lower ends of the cellulose nitratefolds on a thin strong material. For this purposewe use aluminum strips (0.6 cm wideX22 cmlongXO.006 cm thick). The aluminum strips arecut, straightened, and then anodized. Theanodized strips are mounted in a holder whichkeeps them taut. The distance (0.025 cm) fromthe top edge of the anodized strip to the fold ofthe cellulose nitrate is controlled by the spacingof the (0.0075 cm diameter) Nylon thread. Theadhesion of the cellulose nitrate to the anodizedstrip is effected by painting "glyptal" solution,three-fourths of the width, along the length ofand on both sides of the strip.

The holder (Fig. 1), with the anodized stripand the Nylon thread, is submerged in a vesselfilled with water, to the surface of which the

FIG. 2. Evaporation frame.

FIG. 1. Strip holder.

6 H. C. Burger and P. H. Van Cittert, Zeits. f Physik 66,210 (1930). FIG. 2a. Evaporation frame.

599

LOUIS HARRIS

cellulose nitrate film is transferred. The film isoriented so that its long axis coincides with thatof the anodized strip. The water is then drainedfrom the vessel so that, as the water surfacerecedes, the cellulose nitrate film follows and isleft in the folded position adhering to theanodized strip. After the film on a strip has dried,the Nylon thread is withdrawn, and the stripwith the attached film is ready to be transferredto an evaporation frame.

Frames for Holding Strips during Evaporation

The evaporation frames limit the deposition ofeach metal to one side of the folded cellulosenitrate strip and to a slight (0.015 cm) overlapat the fold. Each frame consists of four mainparts. One part (A) holds the strip taut, yetextendable (by means of a spring); a second part(B) serves as a supporting back for the strip; athird part (C) is a gold mask (0.005 cm thick)which helps to hold that part of the film whichextends beyond the edge of the anodized stripdown against the supporting back (B) and alsoconfines the metal deposits to bands 0.25 cmwide with a bare space 0.05 cm wide between thebands; the fourth part (D) holds the gold mask

DO NOr SCAL

FIG. 3.

in place. An evaporation frame is illustrated inFigs. 2 and 2a.

After one metal has been deposited, the framesare removed from the bell jar, parts B, C, and Dtransferred to the opposite side of part A, and thesecond metal is evaporated. Evaporation is ac-complished on four strips simultaneously. Fromeach strip twenty-two thermal elements 0.03 cmwide can be cut.

It has been necessary to make trial evapora-tions for each metal before the metal is depositedon the strips. A cylinder of metal (0.205 g ofantimony or 0.445 g of bismuth) is completely'evaporated stepwise on to cellulose nitrate filmsmounted on the mica frames described above. Atthe end of the evaporation process the films areremoved from the bell jar and the electrical re-sistance of the metal deposit is determined. Trialruns are continued until the deposits have aresistance of 3.8 ohms or less. The frames andstrips (illustrated in Figs. 2, 2a) are then placedin the bell jar and the evaporation of the metalcarried out.

Stacking and Mounting of Thermopiles

The folded strips are now cut along the barespaces into individual thermoelements using acutter similar to a paper cutter. The resistance ofeach element is measured with a special testingclamp whose jaws grip the element 0.03 cmback from the fold. Those elements havingresistances greater than 1.75 ohms are dis-carded.

The elements are then stacked in units of fiftyin a modified press. The folds of the elements areplaced against a glass plate for assembly so thatall the folds will be in the same plane. Pressure isapplied until the electrical resistance is at a mini-mum. The sides and back of the stack are paintedwith cellulose nitrate solution while the stack isstill clamped, thus cementing the elements into apile. Figure 3 illustrates one of the elements andalso shows how the elements are arranged in acompleted thermopile.

The cemented pile is then put into a "skeleton,"also a miniature press, and blackened with gold"black." After the skeleton has been insertedinto the outer casing of the housing, the piles areready for use.

600

RAPID RESPONSE THERMOPILES 601

TABLE-IT. Physical dimensions and properties of thermopiles.

a b c d

Length "hot" to "cold" junction 0.025 cm 0.025 cm 0.035 cm 0.015 cmWidth of each element 0.25 cm 0.25 cm 0.25 cm 0.25 cmThickness, antimony 0.6 X10-4 cm 0.6 X10-4 cm 0.6 X10-4 cm 0.6 X10-4 cmThickness, bismuth 1.07 X10-4.cm 1.07X10- 4 cm 1.07X10- 4 cm 1.07X10-4 cmThickness, cellulose nitrate 1.7 X 10-4 cm 1.7 X 10-4 cm 3.2 X O-4 cm 3.2 X 10-4 cmNumber of junctions in pile 50 50 50 25Active receiving area of pile 0.11 cm 20.11 cm 2

0.09 cm 20.05 cm 2

Thermopile resistance 63 ohms 84.5 ohms 56 ohms 32.5 ohms

All the receivers were covered with gold "black." Thermal e.m.f. (measured) Bi-Sb =92 iv/degree Centigrade. a and b adjacent to one another inthe same housing; c and d adjacent to one another in a second housing.

TABLE III. Response of thermopiles to steady radiation.

a+b c+d a+ba b c d (series) (series) (parallel)

,uv/g watt/cm2 0.044 0.0513 0.0468 0.0075 0.095 0.054 0.0445Area (cm 2) 0.11 0.11 0.09 0.05 0.22 0.14 0.22,uv/,u watt 0.40 0.467 0.52 0.15 0.431 0.386 0.202

ELECTLUCAL CHARACTERISTICS AND SENSITIVITYOF THE FINISHED THERMOPILES

Theory of Thermopile Response

When a thermopile is subjected to interruptedradiation, the a.c. component of the voltagegenerated by the thermopile may be evaluated asfollows:

1Vac.=knm(G/L)

Y (f, )(1)

where

V. = the r.m.s. voltage generated at the terminals of thethermopile,

k =a constant of proportionality,n = the number of "hot" junctions,m= the volts per degree for each junction,G=heat absorbed by thermopile receiver (calories/

cm2 /sec.),L = heat lost by thermopile (calories/degree/cm2/sec.),

andy(f, T) =a functional relation, where f is the frequency of

interruption and r is the relaxation time, or timeconstant.

The relaxation time may be expressed moreexplicitly as

heat capacity of receiver (calories/degree/cm2Tr =

heat lost by receiver in unit time (calories/degree/cm 2/sec.)

In the cases to be discussed here, L is identicalwith the denominator of (2).

For steady radiation, Eq. (1) reduces to

GVd.,. = nm-, (3)

L

where Vd.c. represents the voltage response tosteady radiation. Equation (3) permits calcula-tion of the d.c. response from the physicalcharacteristics of the thermopiles.

Physical Characteristics of Thermopiles Tested

We give in Table II the physical and electricalcharacteristics of four folded bismuth-antimonythermopiles.

Voltage Response to Steady Radiation

Each thermopile was connected to an L. andN., H.S. galvanometer and illuminated withradiation from a "standard" lamp. The responses,corrected for window losses, are given in Table I I I.

The response to steady radiation is nearly thesame for piles a, b, and c, while the response of dis much less. This is caused by the fact that d hashalf the junctions of a, b, and c, and the heat lossfor d is greater than a, b, and c, because of theshorter distance between the hot and cold junc-tions. The- results obtained with the piles in seriesserve as a check on the accuracy of the measure-ments. Since each junction acts as its own re-

(2)

LOUIS HARRIS

i5

/

0

I?

It

05

o I I - I - I I _ _ I I Da 20 40 60 80 /00

FREQUENCY-CYCLES PER SECOND

FIG. 4. Response of thermopiles to interrupted radiation.a, b, c, d, refer to thermopiles a, b, c, d, respectively.

ceiver separately, the voltage output for an ex-

tended image is proportional to the number of

junctions. The resistance of the pile increasesdirectly with number of junctions.

The heat loss per unit time (L) by the receiver

can be evaluated from *the measurements inTable III with the aid of Eq. (3). Calculationsshow that L is approximately 100 times greater

for the folded piles than for the single junctionevaporated couples operated in vacuum.

Voltage Response to Intermittent Radiation

Radiation from a ribbon filament incandescentlamp was focused on the thermopiles and inter-

rupted by a special rotary shutter. Previous tests

showed that this shutter passed radiation of

sinusoidally varying intensity. The response of

the thermopiles to this radiation at 20, 36, 52.8,84, and 100 c.p.s. was measured.

The signal from the thermopile passed to a

transformer, then to the grid of the first tube of

an untuned amplifier. The output from the

amplifier was fed to a GR wave analyzer. Thisvoltage having been recorded, the thermopile was

disconnected from the transformer and a resistorof equal magnitude substituted. A signal from an

oscillator, reduced in intensity by 106 times, was

applied through the resistor, transformer, andamplifier. The output (OP) of the oscillator was

adjusted until the same reading was obtained on

the wave analyzer as with the thermopile arrange-

ment. The oscillator was then connected directly

to the wave analyzer and the voltage output forthe setting OP determined. This reading dividedby 106 equals the output at the terminals of the

thermopile. The above procedure was repeatedfor each frequency. The voltage at zero frequencywas determined from the d.c. voltage generatedby the thermopile at the maximum transmissionof the sector mentioned above. This voltagedivided by 2.83 equals the voltage that would be

generated at zero frequency. The ratio of the

deflections of the galvanometer for the radiationat maximum transmission and that using the

standard lamp permitted calculation of the in-

tensity of the intermittent radiation used. The

results are plotted in Fig. 4.Although a, b, and c have nearly the same re-

sponse at zero frequency, the response to inteF-

rupted radiation for c and d is much less than that

for a and b. This is caused by the much heavier

cellulose nitrate backing for c and d. Thermopiled should be faster than c, because of the shorter

-distance between hot and cold junctions. Themeasurements indicate this.

Speed of response

It is of interest to compare the response ob-

served here with the response of receivers having

unique time constants. Since* v/v0 = (I +27rf 2 T2 )-

where v =voltage at frequency f, vo =voltage atzero frequency for same radiation intensity,

f= frequency of interruption of radiation, and

r=time constant, straight lines should be ob-tained when (v0/v) 2 is plotted against f 2. Figure 5shows the straight lines obtained for receivers

having time-constants of 0.01 sec. and 0.03 sec.

The folded thermopiles do not show such a linearrelation; therefore they cannot be characterized

by a unique time-constant.A "time-constant" calculated using zero fre-

quency voltages will give an apparently slow

response time while a "time-constant" calculatedfrom the high frequency data alone will give an

apparently fast response. The "time-constant" isdependent upon the duration of the signal for

such a thermopile and therefore does not have

the usual significance.

* This relation is valid only for case where heat con-duction loss is small compared to radiation loss.

602

RAPID RESPONSE THERMOPILES

Noise

The "noise" from the thermopiles was less thanthe "noise" of the amplifier available and couldnot be measured. The amplifier noise wasequivalent to 3 X 10-8 volt.

Discussion of Results

The "folded" thermopiles described here areconsiderably faster than any thermopiles re-ported heretofore, yet the response ( volt/Lwatt/cm2 ) is greater for all frequencies above cycles per second than for evaporated thermo-couples operated in a vacuum.

The calculations for the relaxation time indi-cate that the use of thinner cellulose nitratebacking material, or another backing material oflower thermal mass, should increase the "speed,"and thus the response to interrupted radiation.

The electrical resistance of the individualthermo-elements is 1.4 ohms whereas theresistance calculated from the resistivity of themetals when evaporated separately is -0.4ohms. The source of this additional resistance isnot known.

The optimum heat loss for maximum voltageoutput at different frequencies is not known.Until a satisfactory theory has been evolved, itwill be necessary to determine the optimumdimensions empirically. It seems safe to assume

0 5000 /0000

FRtQUENCY (CPs)2

FIG. 5.

that the thermopiles can be substantially im-proved by further study.

Although developed particularly for use assurface piles, the design can be modified easily toform linear thermopiles.

ACKNOWLEDGMENTS

This development is the outgrowth of someideas held jointly with Mr. Willard E. Buck, whocontributed in great measure to the success of theproject. The author also acknowledges his thanksto Mr. Alan C. Bemis, Dr. Benjamin M. Siegel,Dr. Walter Stockmayer, Miss Alette C. Curtis,and Mr. James L. Hildebrand.

603