Electrical and optical characteristics of vacuum-sealed polysilicon microlamps

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 39, NO. 6, JUNE 1992 I363

Electrical and Optical Characteristics of Vacuum- Sealed Poly silicon Microlamps

Carlos H. Mastrangelo, James Hsi-Jen Yeh, and Richard S . Muller, Fellow, IEEE

Abstract-A silicon-filament vacuum-sealed incandescent light source has been fabricated using IC technology and subsurface micromachining. The incandescent source consists of a heavily doped p+ polysilicon filament coated with silicon nitride and enclosed in a vacuum-sealed ( = S O mT) cavity in the silicon- chip surface. The filament is formed beneath the surface and later released using sacrificial etching to obtain a microstruc- ture that is protected from the external environment. The filament is electrically heated to reach incandescence at a temperature near 1400 K. The power required to achieve this temperature (for a filament 510 x 5 x 1 pm3) is 5 mW. The emitted optical power is 250 pW, and the peak in the spectrum distribution is near 2.5 pm. The radiation approximately follows Lambert's cosine law. The subsurface micromachining technique used to produce the evacuated cavity has applications in other micromechanical devices.

I. INTRODUCTION CANDESCENT-filament light sources emit broad- r band radiation that extends from visible wavelengths to

the far infrared. The broadband nature of the radiation makes these sources useful in spectrophotometry and infrared-signal generation.

Early miniaturized incandescent sources were used as displays [l], [2]. These devices consisted of thin-film tungsten filaments suspended from a glass substrate, and they were fabricated using hybrid-circuit technology. More recently, miniature light sources [3], [4] have been fabricated utilizing silicon IC technology. In these sources, the glowing filaments are electrically heated polycrystalline silicon beams [5] standing a few microm- eters above a silicon substrate. The filaments are exposed to the atmosphere; hence, these elements are particularly susceptible to rapid oxidation and contamination.

In this paper we describe the electrical and optical characteristics of a micromachined, vacuum-sealed incandescent light source or microlamp [6] in which an incandescent filament is suspended inside a cavity that is sealed by a window transparent to the emitted radiation. The filament is built beneath the silicon-nitride window and released using sacrifical etching techniques. The vacuum- sealed subsurface design protects the device and avoids filament oxidation and particle-contamination problems

Manuscript received March 4 , 1991; revised September 16, 1991. The review of this paper was arranged by Associate Editor S. D. Senturia.

The authors are with Berkeley Sensor and Actuator Center, Department of Electrical Engineering and Computer Science, and the Electronics Re- search Laboratory, University of California, Berkeley, CA 94720.

IEEE Log Number 9107566.

suffered by earlier devices [3], [4]. The vacuum-sealed microlamps require lower power since gas-conduction heat losses are eliminated.

11. MICROLAMP IMPLEMENTATION Fig. 1 shows a sketch of the device cross section. The

incandescent filament is positioned between a V-groove etched into the substrate and a low-stress silicon-nitride window [7], [8] transparent to the filament blackbody radiation. The window hermetically seals the cavity at a reduced pressure. The V-groove is approximately 25 pm deep. This deep groove is necessary to accommodate for the buckling of the filament at high temperature without contact to the cavity walls.

The filament consists of a p + polysilicon beam coated with low-stress silicon nitride. The filament was doped with boron to a concentration of lo2' cm-3 yielding a resistivity of 4 x lop3 s2 * cm. The conductive polysilicon and insulating silicon-nitride coating are 0.9 and 0.3-0.5 pm thick, respectively. In operation, the beam is electrically heated until it glows at a temperature near 1400 K.

The cavity seal is achieved by filling lateral etch channels [9], [lo] with additional silicon nitride after the filament has been released and the V-groove etched. The pressure inside the chamber is the same as that present in the LPCVD furnace during the nitride deposition (300 mT at 835 "C yielding 80 mT at 25 "C). A similar sealing technique was used by Sugiyama et al. [ l l ] , [lo] for the fabrication of an absolute pressure sensor. The silicon-nitride window is thick enough to undergo negligible deflection due to the pressure difference between the chamber and the outside environment. A window thickness 2.5-2.8 pm is adequate for this purpose.

Fig. 2 is a top-view photograph of three vacuum-sealed microlamps of various lengths. Fig. 3 shows a SEM photograph of a cleaved microlamp cross section. The filament inside the cavity was broken when the wafer was cleaved, but the picture clearly shows that the filament is not bonded to the bottom of the window nor to the V-groove walls. Fig. 4 shows device cross sections near a sealed etch channel. Notice that the channel is completely filled with silicon nitride. The surface of the nitride layer near the etch-channel seal is very smooth and shows no evidence of cracks.

The microlamp chips measured 4 x 4 mm. Each of these chips consists of an array of 48 lamps ranging from

0018-9383/92$03.00 0 1992 IEEE

I364 IEEE TRANSACTIONS ON ELECTRON DEVICES. VOL. 39, NO. 6. JUNE 1992

nitnde-coated polysilicon filament

/ silicon nitnde window

\ 150 to 600 pm in length. About 70% of the devices were operational.

111. MICROLAMP FABRICATION The microlamp fabrication process shown in Fig. 5, be-

gins with ( 100) silicon substrate wafers. A thin layer of low-stress silicon nitride 0.5 pm thick, is deposited on the samples and plasma-etched to define the edge of the V-groove. The etching is followed by the deposition of

layer [11] between the filament and the substrate. A 0.3 pm of low-stress silicon nitride [SI is then deposited over the PSG spacer to form the bottom of the filament encap- sulation. After the deposition, the residual oxide of the silicon nitride is carefully removed by a H F dip, and 0.9 pm of undoped polysilicon is grown. The samples are then ion-implanted with a 1OI6 cm-2 dose of boron, yielding a polysilicon resistivity of 4 X IOp3 fl cm after annealing. The filaments are then patterned using a plasma etch that stops at the nitride layer.

After the photoresist is removed and the samples are cleaned in piranha solution (5: 1 H202: H2S04). The residual oxide left on the silicon-nitride layer is again carefully nitride is deposited on the samples. Both top and bottom nitride layers are next patterned and etched to complete the Drotective coating that encamulates the filament. The

Fig. 1. Schematic cross section of the vacuum-sealed microlamp. 0.7 pm Of phosphosilicate glass (PSG) to provide a spacer

Fig. 2. Top view of three vacuum-sealed microlamps 390, 430, and 470 p n long, respectively. The silicon nitride windows are 55 pm wide.

and a 0.3-pm layer Of

Fig. 3. SEM photograph of a cleaved wafer showing a microlamp cross section. The broken filament inside the cavity is free and unbonded to the window. The depth of the V-groove is approximately 25 pm.

SEALED - A N C H O R 7 CHANNEL /

Fig. 4. SEM photographs of a microlamp cross section near a sealed etch channel. The channel is completely filled with silicon nitride, and it shows no evidence of cracks.

" silicon-nitride etching stops at the oxide spacer as shown in Fig. 5(b).

A 3-pm-thick layer of PSG is then deposited on the samples. Subsequently, the wafers are heated at 1050°C for 30 min. This heating cycle activates the filament dopants and reflows the thick PSG layer which is then etched in buffered HF to form a thick glass mesa on top of the filament. The high-temperature reflow step is necessary to achieve a satisfactory etch of the PSG mesa.

A subsequent deposition of 0.8-pm-thick PSG follows the mesa etch. This glass is patterned and etched with buffered HF to form an alternating array of etching channels [9] and anchors for the nitride windows as seen in the photograph of Fig. 4. The height of these 0.8-pm channels can be reduced somewhat. Sugiyama et al. [ l 11 satisfactorily used 0.15-pm-high channels.

After the PSG etch, a 1-pm-thick layer of low-stress silicon nitride is deposited on the samples. This layer represents approximately one half of the final thickness of the window. The nitride is then patterned and etched on the periphery of the microlamp down to the PSG of the etching channels as shown in Fig. 5(d). These openings on the nitride are the etching holes through which the sacrificial PSG and silicon substrate are etched.

After the etching holes have been opened, the samples are immersed in concentrated H F for 2.5 min to remove the PSG under the silicon-nitride window, thus releasing the filament. The wafers are then immersed in hot KOH for 90 min to anisotropically etch the V-groove in the substrate. The hydrogen evolution generated by the KOH aids

MASTRANGELO er a l . . CHARACTERISTICS OF VACUUM-SEALED POLYSILICON MICROLAMPS

~

1365

nitride window SG etch channel

p+ poly filament nitride etch channel

etched cavity I I I I

(C) ( 4 PSG mesa sealed channel

I I

(e) i f ) Fig. 5 . Simplified fabrication process. (a) Initial spacer deposition. (b) Ni- tride-coated filament deposition and definition. (c) PSG mesa and etching channel definition. (d) Nitride window deposition and etch hole definition. (e) PSG removal and silicon groove etch. ( f ) Etch holes sealed with additional nitride.

in the removal of the end products of the reaction. The samples are cleaned, and an additional layer of silicon nitride is deposited which fills the etching holes and hermetically seals the cavities. Since the gas ratio of NH3 to SiCI2H2 is not stoichiometric, a small amount of dichlo- rosilane is trapped in the sealed cavity which is responsible for the cavity pressure. The effects of the dichloro- silane trapped in the cavity have not been studied; however it does not seem to affect the device performance when compared to other unsealed devices operated inside vacuum systems. After this step, contact holes to the polysilicon filaments are opened, and the wafers are met- alized and sintered.

IV. MICROLAMP RADIATION The spectral density P(A) radiated at wavelength h by

a microlamp with filament of width w and length 1 is

P(A> = w j: €V(x), A)I,(W), h)r(A) Q!x (1)

where T(x) is the temperature along the filament, Zb (T , A) is the power-per-unit area radiated by an ideal blackbody filament, E is the filament emittance, and r(X) is the optical-transmission coefficient of the nitride window. The intensity Zh is

h W Zb (T , A) = c D( A)

4 (exp (hw/k ,T) - 1 ) '

2nc A

U = -

where D( h) is the spectral density of allowed electromag- netic modes per unit volume and c is the speed of light; k, and h = 27rh are the Boltzmann and Planck constants,

respectively. This density D( h) depends on the dimensions of the microlamp cavity [12], [13] and the optical properties of the various layers of which it is composed. If these dimensional effects are ignored, D( A) reaches its limiting value D, ( A ) [ 121, [ 131

(3) 87r

h D,(A) dx = 7 dx.

The optical transmission of the window r ( A ) is a function of the incidence angle and the optical properties of the window material [ 141. If the window is lossless, at normal incidence r ( A ) is

(4)

where n and r are the window refractive index and thickness, respectively. The transmission coefficient r ( A ) can be modified by both radiation tunneling [ 151, [16] and interference effects due to the proximity of the filament to the window (about 3 pm).

In order to evaluate the integral in ( I ) , we must first know T(x) as a function of distance along the filament. If the filament is in a vacuum, and the radiated energy is small compared to conduction losses along its length, the temperature profile is determined by the heat equation

1 r(x) = (1 + ((1 - n212/4n2) sin2 ( 2 7 r t n l ~ ) )

where J is the filament current density in the polysilicon conductive layer, po its resistivity, and K, the equivalent thermal conductivity

where K ~ ~ ~ ~ , K , ~ ~ , Apoly, and Anit are the thermal conductiv- ities and cross-sectional areas of polysilicon and silicon nitride. Approximate values of K~ and K~~~ of 0.3 and 0.03 W * cm-' K-' are reported in [17] and [8], respectively. Equation (5) cannot be solved analytically unless K,(T) and po(T) are specified functions of temperature; however, because of symmetry, its solutions are even with respect to x = 1/2. Assuming that ohmic power is generated roughly uniformly along the filament and that the variation in K , in (5) is small, we can solve (5) to find an approximate form for T(x)

where T, is the substrate temperature, and Tmax is the maximum filament temperature. If K, and po are independent of x , Tmax is foundusing the voltage-temperature (V-T) theorem [ 181, [ 191

TI7l.W

T, Vg = 8 j K,(T)Po(T> dT = 8Kepo(Tmax - Ts) (8)

where V, is the voltage applied to the filament. Hence,

~

1366

the total power PT radiated by the microlamp is a function only of T,,, which depends on the filament voltage v b as shown in (8). In order to keep the computation tractable, we assume that the filament behaves like a gray body with c ( A , T ) roughly constant and equal to cp This assumption is reasonable, since unlike intrinsic silicon [ 2 0 ] , heavily doped p-type silicon has a continuum absorption spectrum due to its high density of free carriers [21] - [23] . Hence

OD

pT(Tmax) = PT(V) = I,, p ( A ) A = cfpb(V) (9 )

where Pb ( v b ) is the total power emitted by an ideal blackbody filament with temperature T(x) of ( 7 ) as seen from outside the nitride window

An exact solution of (10) is only possible by numerical methods because of the complex wavelength dependence of r ( A). However, since I' ( A) ( 4 ) depends only weakly on A, we replace it with an average value r. Using ( 2 ) , (3), (7), and r( A) = r in (9 ) and (lo), we obtain PT as a function of the applied bias. The integration of (10) is easily obtained if it is carried on A first. the integration in x yields a fourth-order polynomial on (T,,, - T, )

PT(V) ~ F C ~ T ~ W Z ( T , , , - T,)4 + o((T,,, - Ts)3 )

(1 1)

where U, = 2 a 4 k ; / 1 5 c 2 h 3 is the Stefan-Boltzmann constant. Using the V-T theorem of (8) we find

Thus the total radiated power is proportional to V: where the exponent n is somewhat different from eight because r) depends slightly on the filament temperature.

V. ELECTRICAL MEASUREMENTS A. Current- Voltage Characteristics

Fig. 6 shows the I-Vcurve of a microlamp with a polysilicon filament 350 pm long and 5 pm wide. Initially, the device resistance increases with the applied voltage since p-type, heavily-doped polysilicon has a positive temperature coefficient of resistance (TCR) ,$ = 10-3"C-' [24 ] , and its resistivity obeys

Po = P,(l + f ( T - G)). (13) Fig. 7 shows the resistivity of the p+ filament as a function of temperature. These data were obtained with a microlamp which was bonded to a ceramic package and placed inside a high-temperature box oven with feed- throughs. The change in resistance ARb/Rb nearly obeys (13) at least up to 900 K.

If the filament is in a vacuum, at these lower temperatures, the microlamp voltage follows a nonlinear relation-

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 39, NO. 6, JUNE 1992

I I l l 1 1 1

0.00 1.00 2.00 3.00 4.00 5.00 6.00

Microlamp Voltage ( V )

Fig. 6. Microlamp filament I-V characteristics. Beyond the point P the I -V curve is time-dependent and irreversible.

0.7 - I

0.6 -

'

A R - R o

-0.1 1, , , 300.00 400.00 500.00 600.00 700.00 800.00

Temperature (K)

Fig. 7. Electrical resistivity of heavily doped p+ polysilicon as a function of temperature.

V, = 2 Jy tan ( 4 ~ J 2 P d 5). (14)

At higher bias, there is a kink a point P in the curve as shown in Fig. 6 beyond which the filament resistance decreases. Guckel et a l . [9] have suggested that at this point the polysilicon experiences thermal breakdown [25 ] , [26 ] . Other possible mechanisms for the resistance decrease include current-induced resistance decrease [27 ] , [24] and filamentation [ 2 8 ] , [ 2 9 ] . In the first mechanism, impurities are redistributed after localized melting at the polysilicon grain boundaries. In the second, the silicon melts in narrow filaments creating low-resistivity current paths. At temperatures above point P of Fig. 6 , the electrical characteristics change rapidly with time and are irreversible as a consequence of a permanent change in the microlamp resistance.

B. Vacuum Test The quality of the silicon-nitride seal was tested as fol-

lows. First, the low-bias I-V curves of a sealed device were measured inside a vacuum system at both atmospheric pressure and at 5 pT. Then the silicon-nitride window was punctured; this exposed the filament to the vac-

MASTRANGELO er a l . : CHARACTERISTICS OF VACUUM-SEALED POLYSILICON MICROLAMPS

uum-system pressure, and the measurements were repeated. Fig. 8 shows the I-V curves of the sealed and punctured devices. Note that the I-V characteristic of the punctured device at atmospheric pressure has a more extended linear range than the others because the filament is cooler in air than in vacuum for the same bias voltage. Note that the I-V curves for the sealed device are un- changed by the vacuum-system pressure which indicates that the window seal is hermetic. Furthermore, both I-V curves for the sealed device resemble closely the low- pressure I-V curve of the punctured device as would be expected when a partial vacuum is present in the microlamp cavity.

C. Turn-off Transient The maximum light-modulation frequency fT for the

microlamp is determined by the rise and flal times of the filament temperature T(x, t )

1 fT ~

7, + rf The transient behavior of T(x, t ) is determined by the time- dependent version of (5). The transient form of (3, assuming that both polysilicon and nitride thermal conduc- tivities are roughly independent of temperature, is

a2T + J 2 p o ( T ) 1 aT (16) - - ax K e a, at

where a, is its composite thermal diffusivity

) (17) P n i t e n i t A n i t + poly e p o l y ~ p o l y

Knit&[ + ~ p o ~ y Apo~y a c = (

where the corresponding p and e and A are their respec- tive densities, heat capacities, and cross-sectional areas. Approximatne values of ppoly = 2.3 and pnit = 3.0 g - cmP3 as well as Cpoly = 0.7 and enit = 0.8 J . g- ' are reported in [17] and [8].

During heating, the temperature time rate aT/at is only limited by the heat-generation term J2po. Since J can be arbitrarily large, it is possible to achieve [30] a very large rate and hence a short T, (microseconds or less) using very short current pulses with large J . Conceptually, the temperature rise time r, is then determined by the rate at which heat is generated in the filament.

If now J is abruptly set to zero, the filament gradually cools since heat is naturally removed by heat conduction to the substrate along the filament itself. This mechanism is entirely determined by the filament thermal properties and physical dimensions. The general solutions of (16) with J = 0 are

OD

T(x, t ) = T, + B,e-'/'" sin n = l

Leak Test - 7 ~

I on 0 punctured device ldtm

A rcalcd device I dlm 0 90

3 080 - 0oo0

~

I367

$ 0 7 0

5 0.60 .- 5 0.50

2 0.40

E 010 0.00

m 0 3 0

e .o 0 2 0

U

ealrd device. 5 p I o n

puncured device, 5 ton

0 00 I 00 2 00 3 00

Microlamp Voltage (V)

Microlamp I-V curves of sealed device at 1 atm ( A ) and 5 pT (0) Fig 8 and punctured device at 1 atm (0) and 5 pT (0).

At sufficiently long times the higher order terms of (18) become negligible, and T(x, t ) limits to its first eigen- function decaying with exponential time constant r1 = T~

(19)

It T(x, t ) is low enough such that po obeys the linear relationship of (19), we can find rf directly from observa- tions of the transient decay of the filament resistance Rh (0) at zero bias [31] since in this regime

- l 2 T I = - - a p 2 7y

Rb = i' p , ( T ) dX a A + BePf /? (20) Apoly 0

Rb has the same time dependence of T(x, t). Fig. 9 shows the resistance transient of a microlamp

510 X 3 X 1 pm3 from incandescence at t = 0 to room temperature at t = 00. This measurement was performed using the circuit described in [31] and shown in Fig. 10. This circuit is driven with a square voltage signal generator. In the negative cycle of the voltage waveform, the generator voltage T/b(t) is applied to the microlamp through a diode heating the microlamp. In the positive cycle, both diodes are off, and the voltage V,(t) is applied to the microlamp through a large series resistor &ig. In this cycle, the electrical power dissipated in the micro- bridge is negligible (since Vb is a few millivolts), and the microlamp filament gradually cools down to the substrate temperature. The small microlamp bias voltage provided by Rhi, and VR ( t ) is amplified by the noninverting op-amp gain stage, and the voltage Vo( t ) is recorded. The time- dependent voltage Vo(t) is directly related to the microlamp resistance since

where A,, is the gain of the op-amp amplification stage. Thus from measurements of Vo( t ) we can extract Rb(t) directly.

1368 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 39, NO. 6, JUNE 1992

thermal breedown region

I

8oo 1-- '---'- I n 3 5 0 X 3 X 1 " 3 I h

2 750 e,

700 I v1 .- 8

6.50 C e,

E 6.00

.-

:LM-~ ~ -L 0.00 5.00 10.00 15.00

Cooling Time ( ms )

Fig. 9. Microlamp turn-off cooling transient. The plot shows the transient resistance decay at zero bias of a 350 X 3 X 1 pm3 microlamp from incandescence at t = 0 to room temperature at t = 03. When the filament is incandescent, the resistance increases with decreasing temperature. This behavior is indicative of thermal breakdown in the filament polysilicon.

'8

D1 LM118

I

Fig. 10. Circuit used for the measurement of the zero-bias microlamp resistance during cooling.

ms. The rather long settling time is mainly a consequence of the thick nitride filament coating which has high heat capacity (or heat storage), but very low thermal conductivity. Since Tj >> T,, the frequency cutoff for the full variation of the signalf, 5: 7;' = 300 Hz.

D. Thermoelectric Effects It is observed that when a microlamp is biased to in-

candescence with a dc voltage, the position of its brightest spot is not the midpoint of the filament as predicted in (7). Instead, it is located at a distance 6 from it; furthermore, it is also observed that the sign of 6 changes when J is reversed as shown in Fig. 1 1 . This shift of the hottest spot is explained by the presence of thermoelectric effects which were ignored in ( 5 ) . The Thomson thermoelectric effect [32]-[35] by means of which heat is given off by current carriers coming from warmer to cooler regions in- troduces a heat-generation term linear in J and proportional to the temperature gradient which modifies ( 5 ) as shown in (23)

In (23) , the parameter u is the Thomson coefficient which is typically a function of temperature [33], [34]. The direction of u is positive if the net effect is that created by positively charged heat-absorbing current carriers which skew the temperature profile toward the cathode. How- ever, there is no correlation between the sign of u and the actual carrier charges since it is possible to get a positive u with negatively charged electron current carriers [ 181. The added linear term in (23) makes T(x, t ) no longer an even function of (x - 1 / 2 ) . If both K, and u are assumed to be constant, and pa is assumed to obey (1 3), we can solve (23) to show that the hottest (and brightest) spot is located at a distance 6 from the midpoint (x = 1/2).

w tanh ( p 1 / 2 ) - p tan (w1/2) w tan (w1/2) + tanh ( 0 1 / 2 )

with

In Fig. 9, note that when the filament is very hot, its zero-bias resistance increases with time; hence it decreases with temperature. This observation is consistent with the thermal breakdown behavior in the polysilicon resistance that occurs beyond point P in Fig. 7.

In the tail of the Rb(t) transient of Fig. 9, the filament temperature is sufficiently low such that (13) and (20) hold, and we extract a T ~ O ~ 3 ms. The corresponding ratio Anit /Apoly obtained from T~ and (19) is

A,,, (Kpoly - Ppoly C p o l y ~ c ) l 2 " 3 , a c = -

AFlY (Knit - Pnit C n i t a c ) rl a2 - - _ -

(22) in agreement with the process specifications.

The resistance settling time is approximately 2rf 5: 6

The distance 6 is known as the Thomson shift [36], [37], and the temperature profile is displaced toward the cathode if u is positive.

Fig. 12(a) is a photograph of a powered microlamp taken with the aid of an infrared-emission microscope im- aging system (KLA EMMI). Note the bright spot in the central region. Fig. 12(b) and (c) shows magnified photographs of the same device both 7 and -7. The small white circle in the photographs shows the approximate lo- cation of the brightest spot. Note that its position shifts a distance of 26 toward the anode when J is reversed, yielding a negative Thomson coefficient. This observation is contrary to our expectations since, at low temperatures,

I369 MASTRANGELO et al . : CHARACTERISTICS OF VACUUM-SEALED POLYSlLlCON MICROLAMPS

J beam

J beam A

x = o x = L / 2 x = L

Fig 1 I . Illustration of the thermoelectnc Thomson effect in the microlamp temperature profile. The linear heat-generation term shifts the position of the maximum temperature and the bnghtest spot in the direction opposite to the current if the Thomson coefficient U is positive

the majority of the carriers in the heavily doped polysilicon filament are positively charged holes. The negative U is indicative of negative current carriers. The sign dis- crepancy of (T tends to suggest that at the filament temperature at which the experiment was conducted, the polysilicon was already in a thermal breakdown (i.e., intrinsic) condition when a significant population of electron carriers is present that could be responsible €or the negative Thomson shift.

Table I shows estimated observed values of the Thom- son shift for different voltages in a 550 X 3 X 1 pm3 microlamp. A Thomson shift as large as 30 pm or 6% of the microlamp length is observed. The large shift suggests a practical application of this device as a spatial light modulator.

E. Constant- Voltage Resistance Drift The microlamp electrical characteristics drift at high

bias. In Fig. 13, we show the overall initial resistance drift of a microlamp device at constant bias voltage. The end of the curves represents the end of the measurements as all microlamps have continued to function. For times close to the origin, the device resistance decreases with time, but for longer times, it flattens out and eventually increases. This behavior can be caused by simultaneous changes of both devices resistivity and (unknown) temperature making the interpretation of the drift curve very difficult. In order to eliminate the temperature as a vari- able, we measured the zero-bias resistance Rb(0) (with a few millivolts of bias) after voltage-time stressing the devices. The zero bias resistance Rb(0) exhibits the permanent changes that occur in the polysilicon resistivity itself at room temperature.

Fig. 12. Microlamp images made with an infrared-sensitive microscope. (a) Upper device under dc bias, lower microlamp unbiased. (b) Biased microlamp with dc voltage positive on the left, circle indicates the point of highest emission intensity which is positioned toward the anode. (c) Biased microlamp with dc voltage negative on the left, again te highest emission point has shifted toward the anode; total shift of the brightest point is 30 wm .

TABLE I

(550 x 3 x 1 pm') THOMSON SHIFT VERSUS VOLTAGE

Voltage Current Thomson Shift (V) (mA) ( pm)

4.5 0.57 10 * 5 4.75 0.60 15 * 5 5.0 0.64 20 f 5 5.25 0.68 25 5 5 .5 0.74 30 j~ 5

Fig. 14 shows both the time dependence of the microlamp resistance Rh(5.7) biased at 5.7 V and at zero volts Rb(0). Both curves show the initial resistance decrease and the long-term increase. We propose the following

1370 IEEE TRANSACTIONS ON ELECTRON DEVICES. VOL. 39, NO. 6, JUNE 1992

400 l l L p l - - - I

10' lo2 I O ' 10' I O ' 10'

Voltage Stressing Time (hours)

Fig. 13 Constant-voltage resistance dnft of a microlamp as a function of stressing time.

0 1200

.9 I0.W 5 : R.00 - 5 E 4.00

6.00 t <. :

t I 7J ~ R J O V J j t L L L-~L~ ~

lo3 lo2 Id lo" IO '

2.00

Voltage Stressing Time (hours)

Fig. 14. Resistance drift of a microlamp at 5 .7 V (dashed line) and at 0 V (solid line) after being stressed at a constant voltage of 5 . 7 V . Note that both curves retain the same characteristics. The zero-bias curve reflects the changes occuring to the room-temperature polysilicon resistivity.

mechanism which explains this behavior. The resistance decrease is most likely associated with growth of the polysilicon grains [38]-[40] in the hot filament since larger grains yield a lower resistivity for the same dopant concentration [4 11-[44].

We believe that the resistance increase is caused by the thermal diffusion [45] of dopants. In this phenomenon, impurities diffuse from hot regions to colder regions of the filament where they "freeze"; therefore, the hot central region is gradually depleted of impurities, resulting in a central higher resistivity region, and an increase in its zero-bias resistance.

F. Constant-Intensity Voltage Drift

compares this voltage to that of a preset reference Vref. The output voltage of the controller V,,,, is automatically adjusted such it is increased if VIo& < Vref and decreased if VIock < Vref, thus adjusting V,,,, such that = Vref. The controller output voltage is then converted to a high- frequency square-wave voltage (1 kHz) and chopped at the lock-in amplifier frequency of 15 Hz. The resulting chopped square waveform V, is used to power the microlamp, closing the feedback loop.

Since the microlamp is driven with a high-frequency ac signal, all phenomena that depend on the direction of the electric field which may aggravate the microlamp resistance drift, such as the Thomson effect, are eliminated. The 15-Hz chopping is required by the lock-in amplifier to discriminate the microlamp light from background il- lumination. The net effect of the feedback circuit is to adjust the microlamp voltage Vb ( t ) continuously such that its radiation output remains at the present value given by

The controller consists of a comparator and a very slow integrator. A slow 30-s time constant of the integrator is necessary to prevent the oscillation of the loop; neverthe- less, it is fast enough to track the microlamp resistance drift. The output of the integrator is then inverted, switched, and amplified to convert it to a square wave. Finally, the square wave is chopped at the lock-in frequency and buffered before it is fed to the microlamp.

Fig. 16(a) shows plots of the microlamp voltage required to maintain a total (over the whole spectrum) microlamp radiation of 70 pW (corresponding to a power efficiency of about 2 .0%) as a function of time in log and linear scales. The curves of Fig. 16 represent roughly two months of continuous operation. During this time, the lamp has undergone 10' on-off cycles without failure. In Fig. 16(a), the microlamp voltage initially decreases (as Rb intially decreases) but it later increases with time. In practice, the device experiences a characteristic aging behavior and it has a finite lifetime. We believe that its lifetime is related to dopant thermal diffusion hence it may be longer if we replace the boron with much slower dif- fusing arsenic instead.

Vref.

G. Filament Buckling and Mechanical Resonance An examination of a microlamp without a window op-

erated inside a SEM reveals that the filament buckles at high temperatures. If bimetallic effects [47]-[49] are ab- sent, the buckling occurs when its average thermal strain exceeds the buckling strain of the beam [50], [51]

In this measurement we use the experimental setup of a2 t 2 (TR a, j' (T(x) - rT) du = a,T 2 - 7 + - (25)

where a, is the thermal expansion coefficient of the composite beam of thickness t , oR is the internal tensile stress of the beam, and E, its composite Young modulus. As- suming a thickness of 2 pm, a length of 500 pm, E, = 250 GPa, oR = l o 0 MPa, and a, = 3 X 10-6"c-', buckling occurs at T I 150°C. At higher temperatures, the

3 1 E, Fig. 15. A microlamp is placed inside a Faraday cage which contains a pyroelectric detector sensitive to the lamp radiation. The detector signal is converted to a voltage at the output of a lock-in amplifier v l o c k which is rep- resentative of the microlamp radiation. The operation of the lock-in and detector optical setup is described in more detail in Section VI. The output voltage of the lock- in amplifier is fed to a feedback controller. The controller

I o

MASTRANGELO er a l . : CHARACTERISTICS OF VACUUM-SEALED POLYSILICON MICROLAMPS 1371

12.00

; ll.OO ; 10.00 M

0 d 9.00

;> 8.00

- (a) $ 7.00 -

6.00 .- E 5.00

4.00

Vre, (intensity control)

I ................. * .................. * ......... 8

FARADAY CAGE

' 15Hz - 1 kHz

Fig. 15. Experiment setup used in the constant-intensity drift experiment

PT= 70 P W

grain thermal diffusio ~ growth ofdopants

lo3 162 14 10' I O ' 10' lo3 Time ( hours )

12.00 c---7 7 - 11.00 > - 10.00 ho

0 Y a 9.00

;> 8.00 -

4.00

I 5 1 0 ~ 5 x l ~ ~ m ~ I

0.0 500 1000 1500

Time ( hours )

Fig. 16. Voltage drift of a microlamp operating at an output radiation power of 70 pW in (a) linear and ( b ) log time scales.

buckling of the beam is very pronounced. If the filament operates at T,,, = 1400 K, the thermal strain is about 3 x causing a beam deflection of 18 pm. This deflection is much larger than the 3 pm of the beam-to-window spacing; thus a careful design of the multilayered filament

and stability analysis of its buckling modes are necessary to assure that its deflection is in the direction of the deep groove beneath.

The periodic expansion and contraction of the beam during heating and cooling can drive such structure to resonance. Sniegowski et al. [50] have analyzed the conditions necessary to drive beams to resonance by this mechanism. The efficiency of such drive is determined by the ratio of the resonance period to the thermal time constant of the beam.

where pm is the density of the beam. The drive becomes efficient when R 1 1. In our beams, this ratio is in the order of lop2 to lop3. Thus beam resonance is not a con- cern since there is no efficient drive mechanism to reach it.

VI. RADIATION MEASUREMENTS A. Experimental Apparatus

Fig. 17 shows the apparatus used to study the microlamp radiation. To carry out measurements, the microlamps are mounted on a ceramic 24-pin DIP (dual-in-line package) positioned inside a Faraday cage and biased by a square-wave pulse generator (Tektronix FG5010) in phase with a lock-in amplifier (EG&G 5203) operating at a frequency of 15 Hz. The radiation is sampled with a LiTaO, pyroelectric detector (Molectron P1-43) suspended above the microlamps. The detector is 3 mm in diameter, and it is housed in a metallic TO-5 package with a MgF, window. This detector has a flat responsivity of 3.4 X lo4 V W - ' for wavelengths between 0.1 and 8 Ccm.

1372 lEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 39, NO. 6, JUNE 1992

... .. . ... . . .. .. .. ... .. . . . .. . . . . ... ... . . . . . .. . .. .... .. .. .. . .. .

pyroelectnc ~

i detector

f = 1 5 H z

Fig. 17. Apparatus used for radiation measurements.

For the spectral measurements, a set of infrared and optical bandpass dielectric filters and absorbing glasses were mounted horizontally in a rotary carousel inserted between the microlamps and the pyroelectric detector. The filters had center wavelengths from 0.5 to 6 pm in 0.5-pm intervals and bandwidths of 0.1-0.3 pm. This filter-detector arrangement performed satisfactorily in con- trast to a conventional monochromator arrangement which experienced unacceptably high power losses.

B. Window Transmissions Stoichiometric silicon nitride is transparent [52]-[54]

to radiation with wavelengths between 0.28 and 8 pm. The low-stress nitride of the window is not stoichiometric [8]; rather it has a composition of Si, and a refractive index of 2.4. Fig. 18 shows the optical transmittance of a 1.3-pm-thick low-stress silicon nitride membrane as measured using a Fourier transform infrared spectropho- tometer. The oscillatory nature of the transmittance is caused by interference in the thin membrane as specified in (4). The nitride window is roughly transparent for wavelengths between 0.5 and 8 pm; hence it transmits most of the radiation emitted by the incandescent filament which has a center wavelength at about 2 .5 pm. The increased wavelength of the lower absorption edge in the low-stress nitride compared to the absorption edge of stoichiometric silicon nitride is consistent [55] with the ex- cess silicon in these films.

C. Radiant Intensity In this measurement, the detector is placed a few cen-

timeters above the microlamp chip. Since the lamp filament is much smaller than the detector, it can be regarded as a point source of radiation. Fig. 19 shows the radiant intensity +((e) (detected power per solid angle) of a microlamp at normal incidence ( e = 0) as a function of the applied bias. The radiation, emitted by the incandescent filament, is clearly visible to the naked eye for vb > 5.5 V. The power needed to reach this condition is approximately 5 mW for a 510 X 5 x 1 pm3 device. The slope n of the log @ versus log vb plot is between 7.5 and 7 for I/ > 4.5 V , as expected from (12).

Fig. 20 shows the radiant intensity as a function of the incidence angle 0. The data can be written as a Fourier

1.00 i

0.90 c - 0.80 c c-2

h

- 1 0 7 0 -

3 0.50’-

e, 0.60 1 0 ,

* .- E 0.40

-: i

thickness = 1.3 II m

0.3 1.0 3.0 10.0 300 Wavelength h ( F m )

Fig. 18. Optical transmission of a 1.3-j~m-thick low-stress nitride membrane.

I O

I O 2 0 3 0 4 0 5 0 6 0 100

Microlamp Voltage ( V )

Fig 19 Radiant intensity of a 510 X 5 x 1 pm3 microlamp at normal incidence as a function of the applied voltage

Incidence Angle 30

sn _ i ~

0 20 40 60 80 Incidence Angle 8 (

Fig. 20. Normalized angular radiant intensity as a function of incidence angle. The curve is independent of the azimuth angle $.

series m

+(e> = +(o) C a, cos (ne) n = l

= +(o) (cos (e) + a3 cos (3e) + h.0.t.) (27)

with u3 = 0.06. This relationship is independent of the azimuth angle 4. Note that the spatial distribution of (26) closely resembles Lambert’s cosine law. This behavior is expected because the filament thickness is of the same order as the penetration depth of the emitted peak radiation

MASTRANGELO et al . : CHARACTERISTICS OF VACUUM-SEALED POLYSlLlCON MICROLAMPS

~

1373

1.00 -

,$ 0.80 -

3 s 3 0.60 - B : 0.40 -

2 0.20 -

v)

f

I i - L

0.00 1, o!’ ’ , 0.00 2.00 4.00 6.00 8.00 I O 00

Wavelength h ( ~ m )

Fig. 21. Normalized spectral distribution of a battery of ten microlamps operated in parallel connection at 6 V. The distribution peaks near 2.5 pm. The dashed line represents a fit of (1) using T,,, = 1400 K .

TABLE I1 OPTICAL FILTERS AND ABSORBING GLASSES

Center Wavelength Half Width Area Filter Type ( r m ) (pm) ( r m )

~

2510t . BP1100* + KG-4S+ BP1500* + SP2000*

BP2600* BP3100* BP3600* BP4000* BP4600* BP5100* BP5300* BP5900*

BP2000* + KG-4S’

0.48 1.12 1.49 1.99 2.62 3.11 3.65 4.04 4.60 5.14 5.36 5.96

0.38 2.9 x IO-’

0.10 6.7 x IO-’

0.12 9.0 x lo-’

0.10 1.0 x lo-’

0.11 2.7 X lo-’

0.15 1.2 X IO-’ 0.18 1.5 X lo - ’ 0.18 1.3 X IO-’ 0.19 1.1 X lo-’

0.28 1.9 X IO-’ 0.16 8.8 X lo-’

0.14 1.1 x lo-’

Manufacturers: (t) Reynard; (*) Spectrogon; (+) Schott.

[21]. Hence the radiation is emitted in the bulk of the filament resembling that emitted by a partially transparent

Integrating the curve of Fig. 20 on an hemisphere above the microlamp, and using a(0) from Fig. 19 at 6 V, we obtain a total radiated power

body [56]-[58].

2 H

pT = j, +(e, 4) dol = jr jy2 * (e , 4) sin (e) de d+

I: 2.77+(0) (28) or 250 pW. Since the electccal power is 5 mW, the energy-conversion efficiency is 5 %.

D. Spectral Distribution The circles in Fig. 21 show normalized experimental

values of the radiation emitted by an array of ten microlamps distributed over an area of 4 mm2 as a function of wavelength for a bias voltage of 6 V. The spectral measurements are made by inserting a set of filters and absorbing glasses between the detector and the sample. The filter characteristics are summarized in Table 11. The peak of the distribution of the data in Fig. 21 occurs at nearly 2.5 pm. The dashed line represents a fit of (1) to the data

points using r ( A) of (4) and T(x) (7). The parameters T,,, I: 1400 K in (7) and Er = E (A , T) z 0.67 in (1) has been estimated in [59] from the peak wavelength of the spectral distribution and total radiative power emitted by the lamp. Note that the measurements support the interference effects near A = 4 pm, but they are not evident (although they may be present) at shorter wavelengths.

VII. SUMMARY We describe the optical characteristics of a vacuum-

sealed incandescent microlamp fabricated using IC technology. The incandescent source consists of a heavily doped p + polysilicon filament coated with silicon nitride and enclosed in a vacuum-sealed ( = 80 mT) cavity in the silicon chip surface. The filament is electrically heated to reach incandescence at a temperature near 1400 K. The lamp filament emits wideband visible and infrared light with a peak wavelength Ap 2.5 pm at a bias voltage of 6 V. An energy-conversion efficiency of 5 % is measured. The microlamp has applications as a wideband radiation source for infrared and optical spectroscopy while operating at low power and with reasonable radiation emission. The subsurface micromachining technique is also applicable to other microstructures that need protection from environmental conditions.

ACKNOWLEDGMENT The authors wish to thank S. Kumar of the Lawrence

Berkeley Laboratory and R. Anderson of the Berkeley Sensor and Actuator Center for some of the optical measurements. They also wish to thank Prof. C. Hu of the UCB Electrical Engineering Department, KLA Instru- ments Corporation for the measurements performed with the EMMI infrared system, and R. Hamilton and K. Vo- ros of the UCB Microfabrication Facility for their help with the high-temperature resistivity measurements.

REFERENCES

P. M. Alt, “Performance and design considerations of the thin-film tungsten matrix display,” IEEE Trans. Electron Devices, vol. ED-

F. Hochberg, H. K. Seitz, and A. V. Brown, “A thin-film integrated incandescent display,” IEEE Trans. Electron. Devices, vol. ED-20, pp. 1002-1005, Nov. 1973. H. Guckel and D. W. Bums, “Integrated transducers based on blackbody radiation from heated polysilicon films,’’ in Transducers ’85, June 11-14, 1985, pp. 364-366. G . Lamb, M. Jhabvala, and A. Burgess, “Integrated-circuit broadband infrared source,” NASA Tech. Briefs, p. 32, Mar. 1989. R. T. Howe and R. S. Muller, “Polycrystalline silicon micromechanical beams,” in Extended Abstract, Electrochem. Soc. Meet., May 9-14, 1982, pp. 186-189. C. H. Mastrangelo and R. S. Muller, “Vacuum-sealed silicon micromachined incandescent light source,” in IEDM Tech. D ig . , 1989, pp.

M. Sekimoto, H. Yoshihara, and T. Ohkubo, “Silicon nitride single- layer x-ray mask,” J . Vac. Sei. Technol., vol. 21, no. 4, pp. 1017- 1021, Nov./Dec. 1982. C. H. Mastrangelo, Y. C. Tai, and R. S. Muller, “Thermophysical properties of low-residual stress, silicon-rich, LPCVD silicon nitride films,” in Transducers ’89; also in Sensors and Actuators, vol. 23(A),

20, pp. 1006-1015, NOV. 1973.

503-506.

pp. 856-860, 1990.

1374 IEEE TRANSACTIONS ON ELECTRON DEVICES. VOL. 39. NO. 6, JUNE 1992

[9] H. Guckel and D. W. Bums, “Fabrication techniques for integrated sensor microstructures,” in IEDM Tech. Dig . , 1986, pp. 176-179.

[ lo] R. T. Howe, “Surface micromachining for microsensors and mi- croactuators,” J . Vac. Sci. Technol., vol. B6, no. 6 , pp. 1809-1813, Nov./Dec. 1988.

[ I l l S. Sugiyama, T. Suzuki, K. Kawahata, K. Shimaoka, M. Takigawa, et al . “Microdiaphragm pressure sensor,” in IEDM Tech. Dig . , 1986,

[12] H. P. Baltes and F. K. Kneubiihl, “Thermal radiation in finite cavities,” Helv. Phys. Acta, vol. 45, pp. 481-529, 1972.

[13] H. P. Baltes, R . Muri, and F. K. Kneubiihl, “Spectral densities of cavity resonances and black body radiation standards in the submillimeter wave region,” in Proc. Symp. on Submillimeter Waves, vol.

[14] 0. S . Heavens, Optical Properties of Thin Solid Films. New York: Academic Press, 1955.

[15] D. Polder and M. V . Hove, “Theory of radiative heat transfer between closely spaced bodies,” Phys. Rev. B, vol. 4 , no. 10, pp.3303- 3314, Nov. 15, 1971.

[ 161 C. M. Hargreaves, “Anomalous radiative transfer between closely- spaced bodies,” Phys. Lett., vol. 30A, no. 9 , pp. 491-492, Dec. 1969.

[17] Y. C. Tai, C. H. Mastrangelo, and R. S . Muller, “Thermal conductivity of heavily doped low-pressure chemical vapor deposited polycrystalline silicon films,” J . Appl. Phys. , vol. 63, no. 5 , pp. 1442- 1447, March 1, 1988. Also see vol. 66, no. 7 , pp. 3420, Oct. 1, 1989.

Oxford, UK: Clar- endon, 1957.

pp. 184-187.

20, 1970, pp. 667-691.

[18] F. L. Jones, The Physics of Electric Contacts.

[19] R. Holm, Electric Contacts. [20] T . Moss, Optical Properties of Semiconductors. London, UK: But-

terworths, 1959. [21] V . I . Fistul’, Heavily Doped Semiconductors. New York: Plenum,

1969. [22] C. H. Liebert, “Spectral emissivity of highly doped silicon,” Progr.

Astronaut. Aeronaut., vol. 20, pp. 17-40, 1967. [23] H. Y. Fan and M. Becker, “Infrared optical properties of silicon and

germanium,” Semiconducting Mater., pp. 132-147, 1951. [24] Y. Amemiya, T. Ono, and K. Kato, “Electrical trimming of heavily

doped polycrystalline silicon resistors,” IEEE Trans. Electron De- vices, vol. ED-26, pp. 1738-1742, Nov. 1979.

[25] H. A. Schafft, “Second breakdown-A comprehensive review,” Proc. IEEE, vol. 5 5 , pp. 1272-1288, Aug. 1967.

[26] K. Ramkumar and M. Satyam, “Negative-resistance characteristics of polycrystalline silicon resistors,” J . Appl. Phys., vol. 62, no. 1,

[27] K. Kato, T. Ono, and Y. Amemiya, “A physical mechanism of current-induced resistance decrease in heavily doped polysilicon resistors,” IEEE Trans. Electron Devices, vol. ED-29, pp. 1156-1 161, Aug. 1982.

1281 D. W. Greve, “Programming mechanism of polysilicon resistor fuses,” IEEE Trans. Electron Devices, vol. ED-29, pp. 719-807, Apr. 1982.

[29] C. N. Berglund, “Thermal filaments in vanadium dioxide,” IEEE Trans. Electron Devices, vol. ED-16, pp. 432-437, May 1969.

[30] W . G. Chace and E. H. K. Moore, Exploding Wires, vol. 2. New York: Plenum, 1962.

[31] C. H. Mastrangelo and R. S . Muller, “Thermal diffusivity of heavily doped low pressure chemically vapor deposited polycrystalline silicon films,” Sensors Mater., vol. 3, pp. 133-142, 1988.

[32] N. F . Mott and H. Jones, The Theory of the Properties of Metals and Alloys. New York: Dover, 1936.

[33] D. K. C. MacDonald, Thermoelectricity: An Introduction to the Principles. New York: Wiley, 1962.

[34] R. A. Smith, Semiconductors. Cambridge, UK: Cambridge Univ. Press, 1978.

[35] V. A. Johnson, Photo- and Thermoelectric Effects in Semiconduc- tors. New York: Pergamon, 1962.

[36] P. M. Davidson, “The theory of the Thomson effect in electrical contacts,” Proc. Inst. Elec. Eng., vol. 96, pt. I , pp. 293-295, 1949.

1371 A. Fairweather, “The closure and partial separation of a metallic contact,” J . Inst. Elec. Eng., vol. 92, pt. I, pp. 301-321, 1945.

[38] L. Mei, M. River, Y. Kwark, and R. W. Dutton, “Grain-growth mechanisms in polysilicon,” J . Electrochem. Soc., vol. 129, pp. 1791-1795, Aug. 1982.

1391 G. C. Jain, B. K. Das, and S . P. Bhattacherjee, “Grain growth in

New York: Springer-Verlag, 1967.

pp. 174-176, July 1987.

polycrystalline silicon,” Appl. Phys. Lett., vol. 35, no. 1, pp. 445- 446, Sept. 1978.

[40] C. D. Ouwens and H. Heijligers, “Recrystallization processes in polycrystalline silicon,” Appl. Phys. Lett., vol. 26, no. 10, pp. 569- 571, May 1975.

[41] J . Y. W. Seto, “The electrical properties of polycrystalline silicon,” J . Appl. Phys. , vol. 46, pp. 5247-5254, Dec. 1975.

[42] G. J . Korsh and R. S . Muller, “Conduction properties of lightly doped polycrystalline silicon,” Solid-Srate Electron., vol. 21, pp.

[43] N. C. Lu, L. Gerzberg, and J. D. Meindl, “A quantitative model of the effect of grain size on the resistivity of polycrystalline silicon films,” IEEE Electron Device Lett., vol. EDL-1, pp. 38-41, Mar. 1980.

[44] N. C. Lu, L. Gerzberg, C. Lu, and J . D. Meindl, “Modeling and optimization of monolithic polycrystalline silicon resistors,” IEEE Trans. Electron Devices, vol. ED-28, pp. 818-830, July 1981.

[45] P. G. Shewmon, Diffusion in Solids. New York: J . Williams Book Co., 1983.

[46] S . Wolf and R. N. Tauber, Silicon Processing for the VLSI Era, vol. 1.

[47] S. P. Timoshenko and J . M. Gere, TheoryofElastic Stability. New York: McGraw-Hill, 1961.

[48] S . P. Timoshenko and J . N. Goodier, Theory of Elasticity. New York: McGraw-Hill, 1969.

[49] W. Riethmuller and W. Benecke, “Thermally excited silicon mi- croactuators,” IEEE Trans. Electron Devices, vol. 35, pp. 758-763, 1988.

[50] J . J . Sniegowski, H. Guckel, and T. R. Christenson, “Performance characteristics of second generation polysilicon resonating beam force transducers,” in Proc. IEEE Solid-State Sensor and Actuator Workshop, June 1990, pp. 9-12.

1511 J. M. Gere and S. P. Timoshenko, Mechanics of Materials. New York: PWS-Kent, 1990.

1521 K. E. Bean, P. S . Gleim, R . L. Yeakley, and W. R. Runyan, “Some properties of vapor depositied silicon nitride films using the SiH,- NH,-H, system,” J . Electrochem. Soc., vol. 114, pp. 733-737, Aug. 1971.

[53] E. A. Taft, “Characterization of silicon nitride films,” J . Electro- chem. Soc., vol. 118, pp. 1341-1346, Aug. 1971.

[54] H. R. Philipp, “Optical properties of silicon nitride,” J . Electro- chem. Soc., vol. 120, pp. 295-300, Feb. 1973.

[55] e. a. V. I . Belyi, Silicon Nitride in Electronics. Amsterdam, The Netherlands: Elsevier, 1988.

[56] D. Y. Svet, Thermal Radiation; Metals, Semiconductors, Ceramics, Partly Transparent Bodies and Films. New York: Consultants Bu- reau, 1965.

[57] H. 0. McMahon, “Thermal radiation from partially transparent re- flecting bodies,” J . Opt. Soc. Amer., vol. 40, no. 6, pp. 376-380, June 1950.

[58 ] R. Gardon, “The emissivity of transparent materials,” J . Amer. Ce- ram. Soc., vol. 39, no. 8, pp. 278-287, 1956.

[59] C. H. Mastrangelo, R. S . Muller, and S. Kumar, “Microfabricated incandescent lamps,” Appl. Opt . , vol. 30, pp. 868-872, Mar. 1991.

1045-1051, 1978.

Sunset Beach, CA: Lattice Press, 1986.

Carlos H. Mastrangelo was born in Argentina in 1960. He studied at the University of California, Berkeley first as an undergraduate, receiving the B.S. degree with high honors in electrical engineering in 1985. He continued graduate studies at Berkeley for the M.S. (1988) and Ph.D. (1991) degrees carrying out research in the Berkeley Sen- sor and Actuator Center in the area of integrated silicon microsensor systems. His studies were partially supported by an ATT Fellowship.

He has recently joined the staff of the Ford Sci-

MASTRANGELO et al.: CHARACTERISTICS OF VACUUM-SEALED POLYSILICON MICROLAMPS 1375

entific Laboratory, Dearborn, MI. His research interests include silicon processing, micromachining, integrated microsensors, and CAD tools for microsensor design.

area of laser research.

James Hsi-Jen Yeh was born in Taiwan in 1969. He received the B.S. degree in electrical engineering at California Institute of Technology, Pasadena, in 1989, and began his graduate work in EECS at the University of California. Berke- ley, where he was named an IBM Fellow.

His research for the M.S. degree carried out in the Berkeley Sensor and Actuator Center has concentrated on optical sources, fabricated using silicon micromechanics. After completing the M.S. requirements he will begin Ph.D. studies in the

Richard S. Muller (S’57-M’58-M’62-SM’70- F’88) received the Mechanical Engineer degree from Stevens Institute of Technology, Hoboken, NJ, and the M.S.E.E. and Ph.D. degrees at the California Institute of Technology, Pasadena.

He joined the Department of EECS at the Uni- versity of California, Berkeley, in 1962 where he is now Professor as well as CO-Director (with Richard M. White) of the Berkeley Sensor and Actuator Center, an NSF-Industry-University research center. He has been awarded NATO and

Fulbright Research Fellowships at the Technical University, Munich, Ger- many.

Dr. Muller serves as Chairman of the Sensors Advisory Board and as a member of the Advisory Committee for the Electron Devices Society of IEEE. He is Chairman of the steering committee for the biennial TRANS- DUCER Conference. He is also a member of the National Academy of Engineering. Together with T. 1. Kamins, he is the author of Device Elec- tronics for Integrared Circuits, (New York: Wiley); 2nd ed. in 1986. He is a co-editor of Microsensors, a volume in the IEEE PRESS Selected Re- print Series, published in 1990.

Electrical and optical characteristics of vacuum-sealed polysilicon microlamps

Documents

Transcript of Electrical and optical characteristics of vacuum-sealed polysilicon microlamps