Solid-State Electronics Vol. 30, No. 11, pp. 1121-1125, 1987
`Printed in Great Britain
`
`$3.00 + 0.00
`0038-1101/87
`Pergamon Journals Ltd
`
`N-TYPE SIPOS AND POLY-SILICON EMITTERS
`
`Y. H. Kwark and R. M. Swanson
`
`Stanford University, Stanford, CA 94305, U.S.A.
`
`Abstract—N-type SIPOSandpoly-silicon emitters on silicon show potential for improved minority carrier
`blocking properties over conventional diffused emitters. This paper discusses experiments designed to
`elucidate the physical mechanisms responsible for this improvement and to optimize the process
`conditions. Emitters both with and without an intentionally grown chemical oxide under the SIPOS or
`poly-silicon film are investigated. Both poly-silicon and SIPOS emitters,
`in their optimized form, can
`achieve J,. of less than 2 x 10~'* A/cm?, an improvementofseveral decades over shallow diffused emitters.
`
`1. INTRODUCTION
`
`SIPOS, for Semi-Insulating Poly-crystalline Silicon,
`has becomethe generic acronym for deposited silicon
`films doped with oxygen. The physical properties can
`be varied over a wide range by varying oxygen
`content. Indeed, normal poly-silicon and SiO, can be
`thoughtof as the limiting cases of no oxygen andfull
`oxidation. The optical andelectrical properties can be
`varied somewhat continuously between Si and SiO,
`by adjusting the oxygen content. N and p-type con-
`duction can be obtained at
`low oxygen concen-
`trations by doping with phosphorus and boron.
`Thesefilms, pioneered at Sony, foundinitial applica-
`tion in their intrinsic form to the passivation of high
`voltage devices[1]. Later, doped films were success-
`fully applied to emitter contact structures, resulting in
`
`a 50-fold reduction of J,, for n-type emitters[2,3]. This
`paper discusses experiments designed to elucidate the
`physical mechanisms
`responsible
`for
`this
`im-
`provement and to optimize the process conditions.
`Further details are available in Ref. [4].
`As deposited, SIPOS is usually amorphous. An-
`nealing is necessary to activate the dopant and to
`produce films suitable for use as emitter contacts.
`During annealing it has been found that the film
`morphology changes. Si precipitates into small grains
`of crystalline Si and the oxygen segregates as SiO, on
`the grain boundaries[5]. One of the salient effects of
`oxygen incorporation is the suppression of grain
`growth, typical grain sizes being less than 100 A. At
`higher oxygen concentrations, carrier transport
`is
`dominated by tunnelling through these thin SiO,
`shells[6].
`
`10mm 0.0.;
`
`8mm 1.0.
`
`PREHEATER
`TUBES INLET
`
`FRONT VIEW
`(Approx. 4 scale)
`
`GROUND Giass.
`BALL JOINTS
`
`28/15
`
`MIXING
`
`CHAMBER
`
`PREHEATER TUBE - TOP VIEW
`OUTLET
`INLET
`
`
`
`"SLIDE" RETAINER
`
`
`ASSEMBLY
`
`
`
`TACK WELD
`
`TO MAIN TUBE
`
`aol
`
`
`TAPER TO FIT
`BALL JOINT
`
`MIXING
`CHAMBER
`
`
`SIDE VIEW
`
`OVERALL VIEW
`(Approx. $ scale)
`
`Fig. 1. SIPOS deposition tube—overall view.
`1121
`
`HANWHA1046
`
`HANWHA 1046
`
`

`

`SIPOS RESISTIVITY
`
`¥ ? N,0/SiH,
`PH,/SiH, =8*10"
`
`PH./SiH,
`POLY-LO 36E-5
`POLY-HI
`
`1.08
`
`800 900 1000 1100
`
`AsDep
`(650)
`ANNEAL TEMP.
`
`1122
`
`Y. H. Kwark and R. M. Swanson
`
`PREHEATER TUBE
`OUTLET
`/
`ORIGINAL TUBE
`
`PROFILE
`
`
`
`MIXING
`BAFFLE
`
`
`
`(ohm-cm)
`RESISTIVITY
`
`OBLIQUE VIEW
`OF ONE (OF TWO) PREHEATER
`OUTLET TUBE &
`MIXING CHAMBER
`(Not to exact scale )
`
`(15 min.)
`
`Fig. 2. SIPOS deposition tube—close up to mixing chamber.
`
`Fig. 3. N + SIPOS film resistivity vs T, and y.
`
`2. DEPOSITION
`
`Evaporation, atmospheric CVD, and low pressure
`CVD have all been used to fabricate SIPOSlayers.
`CVDtechniques have been found most successful for
`device applications. Due to the lack of a low-pressure
`deposition system, atmospheric CVD wasselected as
`the method for fabricating experimental films. Initial
`experiments in a cold-wall epitaxial reactor revealed
`that SIPOS is a difficult material
`to controllably
`deposit becauseofits non-stoichiometric nature. Use-
`able areas of deposition extended only several centi-
`meters over the susceptor. Accurate control of both
`temperature and flow dynamics was clearly needed.
`This suggested using a hot-wall system. The final
`configuration is shown in Fig.
`1. A carrier gas
`preheater is used so that the high flow of nitrogen
`does not cool
`the wafers. Figure 2 details the gas-
`mixing chamber. The design is such that the reactant
`species are mixed due to turbulance in the gas flows.
`The end of the tube is purged to minimize stagnant
`regions and gas-phase reactions. Useful deposition
`with this system extends over 6 inches on the
`susceptor.
`A plot of film resistivity as determined by four-
`point probe is shown in Fig, 3, y is the molratio of
`N,O to SiH,. The strong dependence ofresistivity on
`oxygen content
`is confirmed and anneals at over
`900°C are found necessary to fully activate the do-
`
`the »y =0 films
`pant. It will be shown later that
`actually contain significant oxygen (as much as
`several atomic percent), probably due to residual
`water or oxygen in the deposition ambient; however,
`resistivities comparable to that of poly-silicon are
`obtained nevertheless.
`
`3. MEASUREMENT OF J,.
`
`The relevant parameters which characterize the d.c.
`performance of an emitter are its saturation current,
`J... and specific contact resistance, R,. This report
`will concentrate mainly on the saturation current. J,,
`characterizes the emitter recombination current be-
`cause it is easily shown that, even in the case of
`SIPOS emitters, the minority-carrier hole current in
`low-level
`injection at the edge of the emitter-base
`junction space charge region follows the standard
`relation J, = J,.(exp qV,./KT — 1).
`In order to measure J,,, transistor test structures,
`as shown in Fig. 4, were fabricated. These used
`epitaxially deposited bases with uniform doping con-
`centration
`to
`allow extraction
`of
`the
`base-
`recombination component of
`the base
`current
`through modulating the base width via changes in
`collector-base voltage. Base recombination current
`was equivalent
`to a J,. of 8 x 10° A/cm? which
`places a lowerlimit on the emitter saturation current
`
`BASE
`
`EMITTER
`
`MITTER
`COLLECTOR
`EMITTE
`
`
`(TOPSIDE) BASE
`
`
`
`
`POLY
`
`
`
`\ SHALLOW
`
`N* DIFFUSED
`OUTDIFFUSED
`
`
`
`
`
`P-epi (base)
`EMITTER
`EMITTER
`N (COLLECTOR)
`
`Fig. 4. Cross-section of bipolar test structures.
`
`

`

`N-Type SIPOSand poly-silicon emitters
`
`1123
`
`200 p x 200p
`EMITTER
`
`RAW DATA
`
`Corrected for
`Series Resistance
`
`Z..RAW DATA
`
`200
`
`400
`
`600
`Vbe (mVolts)
`
`800
`
`10?
`
`10°
`
`76. x 76p
`EMITTER
`
`PERIMETER
`
`
`which can be reliably determined. The test devices
`had two diffused emitters of different junction depth,
`emitters A and B, that were either contacted with
`aluminum, as a control, or contacted with in-situ
`doped SIPOSor poly-silicon. In addition, emitters of
`type C with no prediffusion were fabricated. These
`relied solely on outdiffusion from the in-situ doped
`SIPOS or poly-silicon to provide sufficient surface
`doping concentration.
`The emitter saturation current was measured using
`standard Gummel plot methods. Devices of two
`perimeter-to-area ratios were incorporated to allow
`extraction of the perimeter component of the base
`current. Temperatures determined using a thermo-
`couple on the wafer chuck agreed with those deter-
`mined from the slope of the log collector current vs
`base-emitter voltage, after correcting for base-width
`modulation effects. Reported saturation currents are
`“corrected” to 300 K using the relation:
`n2(300 K)
`J,e(300 K) = Soe( T device)
`oe
`)
`oe device) 1 Taevice)
`
`4. EXPERIMENTAL RESULTS
`
`4.1. Conventional emitters
`
`The pre-diffused emitter deposition parameters
`are shown in Table 1. Phosphine at 1100 ppm was
`used as the dopant source in these conventional
`predep and drive-in diffusions. Gummel plots for
`both the large and small devices are shownin Fig. 5.
`The extracted perimeter- and area-related current
`components are also shown. The perimeter current
`shows a strong non-ideal component which was
`found,through the use of a field plate equipped
`device, to arise from recombination at the Si-SiO,
`interface in the emitter-base space-charge region.
`Illustrated in Fig. 5 is the method of removing the
`effect of series resistance by assuming that the col-
`lector current continues to follow an ideal ex-
`ponential form. Any additional voltage is assumed to
`come from resistive voltage drops and is subtracted
`from both the collector and base current plots.
`Table 2 shows the measured saturation current for
`an aluminum-contacted conventional emitter. The
`extracted value varies by less than 4% over 4 decades
`of base current. This is testimony to the accuracy with
`which the various parasitics (perimeter current and
`
`-4
`Corrected for
`Three types of emitter contacts were studied, con-
`10
`Series Resistance
`ventional aluminum contacts, SIPOS and_poly-
`silicon. In the latter two cases the emitters were
`-5
`fabricated both with and withoutan interfacial chem-
`3 10
`2
`ical oxide.
`=z 0
`we
`
`RAW DATA
`
`197
`
`108
`
`10°
`
`200
`
`I
`
`“Y)
`400
`
`PERIMETER
`
`600
`Vee (mVolts)
`
`800
`
`Fig. 5. Gummel plots for aluminum contacted large and
`small emitter devices.
`
`series resistance) have been accounted for, as well as
`the precision of the temperature determination. Sim-
`ilar procedures were used for the other emitter struc-
`tures.
`
`4.2. SIPOS emitters
`
`4.2.1. Role of oxygen. It has been postulated that
`the improvement that SIPOS emitters obtain is be-
`cause SIPOS hasa larger band gap than crystalline
`
`Table 1. Deposition parameters for the prediffused emitters
`Sheet
`Junction
`resistance
`depth
`
`Drive-in (T/£)
`
`Predep (7/1)
`
`EM-A
`EM-B
`
`800°C/30 min
`800°C/30 min
`
`1025°C/90 min
`1025°C/30 min
`
`60n/O
`72/0
`
`1.04
`0.66
`
`

`

`1124
`
`Y. H. Kwark and R. M. Swanson
`
`Table 2. Measured J, vs V,, for a conventional
`aluminum contacted transistor
`
`
`
`“12 a|10 [ At contact 1
`
`Joe
`Vin
`Joc
`Vee
`
`(10? A/em?)
` (V)_
`(10° Ajem’?)
`(V)
`1.18
`0.545
`1.18
`0.392
`1.17
`0.562
`1.21
`0.409
`
`NoChemox
`0.426
`1.21
`0.579
`117
`0.443
`1.20
`0.596
`1.16
`0.460
`1.19
`0.613
`1.16
`0.477
`1.19
`0.630
`115
`0.494
`1.19
`0.647
`1,13
`0.511
`1.19
`0.664
`1,13
`0.528
`1.18
`
`
`
`Prediffused
`Emitter
`
`(B)
`Outdiffused
`Emitter
`(C)
`
`
`
`
`4
`
`10!4
`
`1000
`900
`850
`800
`SIPOS ANNEAL TEMPERATURE (°C)
`
`
`
`
`
`
`
`silicon so that the injected minority carriers see a
`barrier at
`the SIPOS-Si
`interface[3]. To test
`this
`hypothesis, SIPOS contacts with varying oxygen were
`fabricated with the results shown in Fig. 6. For these
`experiments, the SIPOS was annealed at 900°C for
`13 min. No out-diffused emitter was incorporated.
`It is seen that there is no descernable influence of
`oxygen content on saturation current as would be
`expected from the heterojunction model. The film
`with y = 1.5 has an oxygen concentration similar to
`that of the Sony-transistors[3]. Even the y =0 (no
`intentional oxygen) film has a low saturation current.
`These films, however, were found to dissolve in HF,
`indicating the presence of some oxygen.
`As seen in Fig. 6, however, the film annealed at
`1000°C has J,, nearly an order of magnitude larger
`than those at 900°C. Clearly, anneal temperature is
`more important than oxygen concentration in deter-
`mining saturation current.
`4.2.2. J,, vs anneal temperature. A series of experi-
`ments was conducted to determine the influence of
`
`INFLUENCE OF OXYGEN CONTENT OF SIPOS
`ON EMITTER SATURATION CURRENTS
`
`LAYER 2
`
`ntNt
`
`TYPE 6-8
`B2L 7® Ta= 1000/13
`TYPE A
`
`B2UL TYPE B
`
`FA, J oO
`
`Fig. 7. Influence of anneal temperature on J,, for SIPOS
`contacted emitters without chemox.
`
`anneal temperature on saturation current. In addi-
`tion,
`it was decided to incorporate an intentional
`interface oxide on a portion of the devices on each
`wafer in order to help clarify the roles of interface vs
`bulk effects in the observed reduction in saturation
`current under SIPOS contacts. The interface oxide,
`called “‘chemox’’ here,
`is the result of omitting the
`final HF dip after an RCA clean. The SIPOS had
`y =0, or no intentional oxygen. Results are shown in
`Figs 7 and 8. The results for the aluminum contacted
`emitters are also shown.
`For devices with no chemox and a pre-diffused
`emitter, Fig. 7, J,. is relatively constant for anneal
`temperatures less than 850°C at a value of 1/20 that
`of aluminum contacted emitters. As the temperature
`increases above 900°C, J,, increases, eventually ap-
`proaching the value for the conventional emitter at
`
`
`
`io? +++ |
`[
`Al contact
`4
`
`ith
`
`Ch:
`
`x
`
`L
`
`[
`
`f
`
`
`
`Outdiffused
`Emitter
`(c)
`
`4
`
`1
`
`
`
`
`1
`
`2
`
`10%
`
`1
`a
`1
`1000
`900
`850
`800
`SIPOS ANNEAL TEMPERATURE (°C)
`
`1
`
`Y= N07 SiH,
`Fig. 6. Influence of oxygen content of SIPOS on J,..
`
`Fig. 8. Influence of anneal temperature on J,, for SIPOS
`contacted emitters with chemox.
`
`n_
`Prediffused
`E
`Emitter
`«
`Zi0'1
`(B)
`.
`i
`o
`r
`-
`ooPESSs05 T, = 900/13
`Fl
`TYPEA
`
`

`

`
`
`
`
`
`with Chemox
`
`1,
`wh
`——-.
`
`1000
`900
`850
`800
`POLY ANNEAL TEMPERATURE (°C)
`
`1125
`
`without a chemox. This suggests that either the
`chemoxis acting as a barrier to the out-diffusion of
`phosphorus, and thus preventslifetime reduction in
`the surface region, or alternatively, is more resistant
`to break-up during annealing than whatever oxide
`exists without the chemox.
`
`4.3. Poly-silicon emitters
`
`Whenit was discovered that the y = 0 SIPOSfilms
`actually had significant oxygen,
`it was decided to
`compare these results with the case of poly-silicon
`deposited with a hydrogen carrier in an epi reactor.
`Such films contain very little oxygen. The results are
`shown in Fig. 9. In this case the emitters without
`chemox had J, only a factor of two less than the
`aluminum contacted emitters. This reduction can be
`theoretically accounted for simply from the increase
`in thickness of the highly doped emitter due to the
`addition of the poly-silicon. The addition of a
`chemoxcauses J,, to be substantially the same as for
`SIPOSemitters. Once again, there is an additive effect
`between the pre-diffused and out-diffused emitters.
`
`3. DISCUSSION AND CONCLUSIONS
`
`N-Type SIPOSandpoly-silicon emitters
`
`C =i
`T
`T
`T™
`—<—- —
`
`e————Al contact
`
`he,
`
`No Chemox
`
`Outdiffused
`
`Emitter
`(c)
`
`Prediftused
`Emitter
`‘)
`
`44
`4
`1
`]
`
`4
`1
`4
`
`a
`E
`e
`qt.
`S109}
`®
`r
`<
`L
`
`r
`|
`}
`r
`
`“12
`10
`
`1o'*
`
`Fig. 9. Experimental J,, vs anneal temperature for poly-
`silicon contacted emitters.
`
`1000°C. If no pre-diffused emitter exists, J, starts
`rather high at 800°C,
`then exhibits a minimum at
`850°C of only 2 x 107'* A/cm’. This is 1/50 that of
`the conventional emitter. Finally, J,, again increases
`in their
`Both poly-silicon and SIPOS emitters,
`with higher anneal temperatures. Undoubtedly, in the
`800°C case the surface of the silicon has not been
`optimized form,
`can achieve J, of
`less
`than
`2x 10~-'* A/em’. For poly-silicon contacts this re-
`dopedsufficiently in the out-diffused case to limit the
`quires an interfacial chemox. The chemox is not
`minority-carrier concentration at the SIPOS-Si inter-
`necessary for SIPOS contacts. The similarity of re-
`face, resulting in poor performance.
`sults for the chemox poly-silicon contacts and SIPOS
`It is interesting to note that, excluding the 800°C
`case,
`J,
`for
`the
` pre-diffused
`emitter
`is
`contacts suggests that these both obtain their im-
`5x 10-'* A/em? plus that for the out-diffused case.
`proved performance through incorporating a thin,
`tunnelable interfacial oxide that impedes the flow of
`This suggests that the total /,. comes from a constant
`holes rather than asaresult of the increased band gap
`recombination in the diffused emitter plus that com-
`ing from the out-diffused portion. For such to be the
`of SIPOS. Preliminary results on p-type SIPOS emit-
`ters indicates that
`they have much poorer per-
`case the quasi-Fermi
`levels throughout the emitter
`formance. This is consistent with a barrier that has a
`need be constant. This is approximately the case for
`lowerelectron barrier height than hole barrier height,
`these devices which have passivated surfaces. The
`increase in J,, with anneal temperature is then ex-
`as does SiQ,.
`plained as the result of increased doping, and hence
`reduced lifetime,
`in the out-diffused portion of the
`emitter. Alternatively, the increase could be due to
`the break-up of an interfacial oxide. Such an oxide
`could form, for example, during the anneal as oxygen
`segregates from the SIPOS. Perhaps both methods
`are operative.
`incorporated, a slightly
`When the chemox is
`different picture emerges. At 800°C J,, could not be
`extracted due to excessive series resistance, This
`resistance decreases with higher anneal temperatures
`suggesting the break-up of the interfacial oxide and
`an increase in surface doping concentration. At
`850°C the results are substantially the same as with
`no chemox, Fig. 8. Annealing at 900°C, however,
`does not result in any increase in J,, as was the case
`
`1. T. Matsushita, T. Aoki, T. Ohtsu et al., [EEE Trans.
`Electron Dev. 23, 826 (1976).
`2. T. Matsushita, N. Oh-uchi, H. Hayashi and H.
`Yamoto, Appl. Phys. Lett. 35, 549 (1979).
`3. T. Matsushita, H. Hayashi and N. Oh-uchi, Jap. J. appl.
`Phys. (suppl) 20, 75 (1981).
`4. Y. H. Kwark, Stanford Electronics Laboratory Tech-
`nical Report, Stanford University (1985).
`5. M. Hamasaki, T. Adachi, S. Wakayama and M.
`Kikuchi, J. appl. Phys. 49, 3987 (1978).
`6. J. Ni and E. Arnold, Appl. Phys. Lett. 39, 554 (1981).
`
`Acknowledgement—This work was
`Department
`of Energy
`through
`Laboratories.
`
`supported by the
`Sandia National
`
`REFERENCES
`
`