Download Part 2

kg
= Seff. Vth
[4.7]
where Vth is the thennal velocity of the oxygen molecules and Seff is the effective
sticking coefficient of oxygenon the growing surface. Seffis boundedby zero and one (0
~ Seff ~1). For each temperaturein our experiment,we can approximatethe effective
sticking coefficient. To do so, we assumethat below 750 .C, desorptionof oxygen from
the wafer surfaceis negligible and that the thennal velocity of the oxygen is detennined
by the substratetemperatureand follows a Mawellian distribution of velocity given by
[4]:
Vth
= (21~~~0) J12
[4.8]
where M(O) is the massof oxygen and Tg is the substratetemperature.CombiningEqns.
4.6 through 4.8, we arrive at an estimateof Seff. The valuesfor Seff are listed in Table
4.1 for eachgrowth temperature.All valuesfor Seffare on the order of 10-4. This is two
orders of magnitude lower than the sticking coefficient of oxygen on a clean silicon
surfaceas detenninedby Hagstrom,Lander et al. and Ghidini et al. (Refs. 15,10-12of
Ch.2)
4.6 Discussionof the Reduction of Oxygen Incorporation in CVD
We associatethe reduction of the oxygen incorporation to (1) the hydrogen
passivationof the silicon surface(reducedSeff)and to (2) the fonnation of the boundary
layer leading to a reducedoxygenconcentrationat the wafer surface. The reductionin kg
is due to the reduction of the effective sticking coefficient, which is a property of the
surface reaction. We associatethis decreasein surface reactivity to oxygen with the
hydrogen passivation of the surface. An equally effective way to reduce the oxygen
contamination at the surfaceof the growing film is to create a large boundary layer to
reduce the mass-transfercoefficient and hencethe number of oxygen molecules which
can interact with the surface. We have detenninedthat the masstransfercoefficient (hg)
and the surfacereaction coefficient (kg) are related by the ratio, B, of Eqn. 4.6 and that
they are of the sameorder of magnitude. This implies that they are both important to our
62
I
experiments. Since their effects on the sticking are additive in nature and not
factor of the order of 100. For instance,if the surfacereaction were to occur infinitely
->
1) and the surfacereactionrateretainsits measuredvalue.
for oxygen incorporation under the assumptionthat we are well below the desorption
temperaturefor SiD. The line in Figure 2.1 (the Lander and Momson data) corresponds
the data of Lander and Morrison. The oxygen flux incorporating in the layers of our
experimentis approximately5 x 1013cm-2s-1.According to the Lander experiment,the
desorptiontemperatureto accommodatethis incoming flux is approximately850 .C. At
unexpected. Hydrogen passivationis expectedto tie up surfacebonds. Therefore, the
Assuming one open site is neededfor adsorption according to fIrst order Langmuir
63
-c~
\
1.2
Physisorption
1
+ Migration
"'"
=
.~ 0.8
1.=
g
u 0.6
"""
"'"
bO
filfSt or,der
:e
Langmuir
u
~
.. """
"'"
"'"
0.4
"'" ~
0.2
~
0
0
0.2
0.4
0.6
SurfaceCoverage
0.8
1
Figure 4.2. Schematicplot of the sticking coefficient of oxygenon the surfaceof silicon.
The ideal caseis that of fIrst order Langmuir adsorptionwhere the sticking probability
drops linearly with the number of open sites. Not many moleculesfollow this model.
Most moleculesare allowed to physisorbto the surfaceand migrate until a site is found to
chemisorb. In this situation, the sticking probability is relatively insensitive to surface
coverageuntil most of the sitesare covered. This plot is only a schematicrepresentation
for oxygenon silicon; the actualcritical points arenot known.
64
shown in Figure 4.2. The value of the oxygen sticking coefficient in the UHV clean
surfacecaseis
- 10-2and in the CVD caseit is -
10-4. a difference of a factor of -100.
This is closeto the first order Langmuir prediction.
4.7 High Purity CVD Growth
Since the incorporation of oxygen in the growing silicon film is reducedby 100
due to hydrogenpassivationand boundarylayer effects, we can tolerate 100 times more
oxygen in the gasflow than expectedfrom the experimentsperfonned on a clean silicon
surfacein UHV. This in turn meansthat in order to grow a silicon or silicon-gennanium
111mbelow 750 .C ( at a growth rate of 500 Nmin
at 10 Ton wiili an oxygen
concentrationin the film of less than 1018cm-3 (the peak solid solubility of oxygen in
silicon). we can tolerate 100 ppb oxygen in the total gas stteam (100 times more than
calculatedfrom the classicalLander and Morrison curve in Fig. 2.1). This is a level well
within the specificationsof semiconductorgaspurifiers availableon the markettoday. At
10 ppb (approximatelimit of purifiers), we can grow down to rates of 50 Nmin.
Although our experimentsare for 02. similar reduction in oxygen incorporation due to
residual water vapor would be expected. To confmn this. we have grown many lowtemperaturesilicon and silicon-germaniumepitaxial layersin our reactorwithout oxygen
inttoduction and with a load-lock to prevent contamination Figure 4.3 shows a SIMS
plot of a silicon-gennaniumlayer grown at 625 .C (100 A/min.) with a sttucture similar
to that of Fig. 3.6. However, in this case, the oxygen concentration in the silicongennaniumlayer hasbeenreducedto the backgroundresolutionof SIMS. Similar levels
of oxygen in silicon grown at 700 .C (30 Nmin. with DCS have beenobtained. This is
the first time that such low-oxygen concentrationshave been obtained in non-UHV
silicon CVD epitaxial layers. This is of tremendous technical imponance for
manufacturing.
65
1020
~
,
L-I1111111111111111111111111111-J
e
~1019
e
0
~
c
'-"
.-0
e
81018
c
0
U
0
0.5
1
1.5
Depth (J1m)
Figure 4.3. SIMS plot of silicon-germanium grown at 625 DC with little oxygen
contamination. The oxygen concentrationis below the detection limit of SIMS. This
sample was grown after the load-lock was installed on the system. The load-locks
preventsoxygen and water vapor from contaminatingthe chamberbetweendepositions.
66
propeniesfor deviceswill be discussedin the next chapter.
67
4.6 References
1.
A.S.
Grove, Ph~sicsandTechnolo~ of SemiconductorDevices,Wiley, New York
(1967).
2.
S.W. Wolf. R.N. Tauber.Silicon Processin&for the VLSI Era: Volume I - Process
Technolo~. Lattice. SunsetBeach.CA (1986): from H. Schlichting.Boundm
Laxer Th~. 4th Ed., McGraw Hill. NY, (1960).
3.
Perry and Chilton, Ed.,ChemicalEn&jneers'Handbook,5thEd., McGraw-Hill, New
York, (1973).
4
M.A. Morris, M. Bowker and D.A. King. "Kinetics of Adsorption.Desorptionand
Diffusion at Metal Surfaces."Com~Tehensive
ChemicalKinetics 19. Elsevier.
Amsterdam(1984).
68
Electronic
Properties
Introduction
All semiconductordevicesrely on recombination-generation
processeswhetherthe
lifetimes are long, as neededin p-n diodes, or short, as required in high speed
photoconductiveswitches. The electronicpropeniesof the silicon and silicon-gennanium
strainedlayersweredeterminedby studyingthe minority carrier lifetimes of the films. We
studiedboth the generationandrecombinationlifetimes of the films by fabricating several
device structures with oxygen concentrations of various levels. For quantitative
measurementswe studiedheterojunctionbipolar transistorsand MOS capacitors.and for
qualitative measurementsof recombinationlifetimes we studied a basic optoelectronic
propertyof d1esilicon-gennanium.
Recombination-Generation
Here we briefly discussrecombination and generationand review some of the
impottantcharacteristicsof semiconductors
andthe effectsthey haveon carrierlifetimes. A
semiconductorin a non-equilibrium state of carrier depletion will attempt to return to
equilibrium by the generationof carriers. Mechanismsfor carriergenerationaredepictedin
Figure 5.1(a). Caniers can be generatedby band-to-bandgenerationwherean electtonin
the valenceband absorbsenergy(an amount ~ Eg) andjumps into the conductionband
leaving behinda valencebandhole. This processcan also take placeby utilizing mid-gap
states.The electtonin the valencebandabsorbsa small amountof energyandjumps to an
energylevel in the forbidden gap; by absorptionof anothersmall amountof energy.it is
excitedinto the conductionband The final productof thesetwo mechanismsis the same:
an electronin the conductionbandanda hole in thevalenceband.
Recombination.on the other hand. is the reverseprocess. In a non-equilibrium
situationwherea surplusof carriersresidesin the semiconductor.recombinationoccursto
drive the material back into equilibrium. Recombinationmethodsare depictedin Figure
S.l b. Carriers can be annihilated by band-to-bandrecombination or through mid-gap
states.
Bam-to-bandoptical recombinationandgenerationprocessesareexn-emelyslow in
silicon and gennaniumbecauseof their indirect bandgaps. Phononsarerequiredin order
to conservemomentum. However.theseprocessescan be mediatedby localized statesin
the bandgap. As the numberof mid-gapstatesin the materialincreases,the rate of carrier
generation and recombination increases,allowing the recombination and generation
processesto occurquite rapidly. Thesestatescan be fanned in silicon by the incorporation
of ttansition metals such as Fe, Au, and Cu but also by defects. We will show that the
addition of large amountsof oxygen which can incorporatein silicon when grown at low
temperaturecanfonD thesestates.
5.3
Generation Lifetime Measurements
5.3.1 Pulsed MOS capacitor / Background
A pulsed metal-oxide-semiconductor(MOS) capacitor technique was used to
detemrinethe generationlifetimes of the films by employinga Zerbstanalysis[1] to extract
70
&:
.
.
t
Ev
-
Etrap
t
0
0
(a)
Ec
--t+ -
Ev
- -Etrap
0
(b)
Figure 5.1. Schematic diagrams of carrier (a) generation and (b) recombination. Both
processescan occur directly across the band gap or indirectly through mid-gap ttaps. In
indirect band gap materials such as silicon and silicon-germanium, direct recombination and
generation are slow and mid-gap statesare neededto speedthe recovery.
71
numerical values for generation times
We probed the minority-carrier generation
propertiesof Sil-xGexby driving an MOS capacitorinto deepdepletionandmonitoring the
recoveryof the devicefrom the deepdepletedstateto the invertedstate. This is seenasan
increasein capacitancedueto the shrinkingof the depletionregionin the semiconductorby
the creationof minority carriers. Thereis a direct relationshipbetweenthe rate of recovery
of a capacitor from deep depletion and the minority carrier generation time. Long
generationlifetimes are imponant for devices such as p-i-n photodiodes[2] and
heterojunctionbipolar b'ansistors[3] (HBT) which require low leakagejunctions under
reversebias. A morecompletedescriptionof the techniquefollows.
The layers under study in this experimentwere grown in the susceptor-free,rapid
thermal chemical vapor depositionsystemdescribedin Chapter3. The substrateswere
four inch, p-type, <100> Cz silicon with resistivitiesof 1-10 O-cm. The chemicalclean
andthe in situ hydrogencleanarealsothe sameasdescribedin Chapter3.
Each sttucturewas initiated with a 1.5 ~
buffer layer followed by the active area
of thedevice.Theactiveregionconsisted
of a 30 nm Sio.82Geo.18strainedlayer grown at
625 .C and a silicon cap layer of 30 nm grown at 850 .C for 30 secondsto simulateour
standardheterojunctionbipolar transistorgrowth process.All layersweredopedp-type to
a level of-3 x 1017cm-3with roron. The oxygenconcentrationsin the someof films were
kept below the detection limit of SIMS
(-2 x 1018cm-3) while in other samplesthe oxygen
concentrationswere > 5 x 1019cm-3. A crosssectionof the deviceis shownin Figure 5.2.
Control samplesof all-silicon layerswerealsogrown at 700 .C (similar doping and oxygen
levels) for parallel fabrication and testing. Although not specifically measuredin this
study. transmission electron microscopy (TEM) measurementson similar structures
fabricatedin our laboratoryhave shown that the Sit-xGex layers are fully strainedwith a
negligiblenumberof misfit dislocations(~ 10 ~
spacing).
The gatedielectricof the capacitorswerefonned by bothplasmadepositionof SiOl
and thermal oxidation of silicon Sttain relaxation, and hencethe fonnation of interface
72
w//////////////////////////,,@
Figure 5.2. Capacitance structure used for testing the generation lifetimes of the
Sio.82Geo.18
strainedlayer. The capacitoris pulsed into deep depletion and monitored
during recoveryto inversion.
73
defects,in the Sio.82Geo.18
film was avoidedby using a low-temperature(400 .C) plasma
deposition of silicon dioxide (10 nm) followed by a 600 .C anneal in nitrogen The
nitrogenannealservedto increasethe dielectric strengthof the material. A thennal oxide,
grown at 800 .C in dry 02. wasalsousedasthe gateinsulatoron somedevices.
In all devices,the oxide/semiconductor
interfacewasfonned with silicon insteadof
silicon-gennanium. (The SiO2/Geinterface has been observedto have poor electrical
properties[4].) Since the thermal oxide consumessilicon, the cap layer of silicon in the
fmal thennal oxide structuresis estimatedto be 25 nm thick as shownin Figure 5.2. The
aluminum gate and guard ring were formed by thermal evaporation, followed by
photolithographyand etching. The guardring surroundingthe gateprovides a meansof
limiting the activeareaof thedevice.
High frequ~ncy(1 MHz) capacitance-voltage
curvesof the capacitorsshowedwell
defmedregionsof accumulation,depletionand inversion for both the silicon and silicongennaniumburied layer capacitors.The dopingsand the oxide thicknessesextractedfrom
the curveswere consistentwith the valuesdiscussedearlier. Straightforwardcalculations
show that the depletionlayer in inversioncontainsthe buried layersof silicon-gennanium.
Generationlifetimes weremeasuredby applyinga voltagepulseto the gateof the capacitor
and monitoring the recovery of capacitancefrom deep depletion (Figure 5.3) as in the
Zerbst [1] technique.
5.3.1.1 Theory
The Zerbstanalysisconsistseffectivelyof plotting the generationratevs. generation
volume of the capacitorwith the slopeof the plot being relatedto the generationlifetime.
By analyzing the changein depletion layer width with time and relating it to the device
capacitance,the following (Zerbst)relationshipfor bulk semiconductors
is found: [1]
d ~~
'di
~c )
2niCo{£i
= 'tgNACi
~C - 1
[5.1]
whereCo, Ci, andC arethe oxide capacitance,inversioncapacitance,andthe instantaneous
74
190
- -- -- - -- ,--
C.mv
X
/~
180
/
p.
G:;
170
"
--Q)
~
"""""""
""
160
""
/
.~
a 150 I
p.
140
I
"
"
",,
" ""
"
~
0.2 V Pulse
/
/
0.5 V Pulse
0.7V Pulse
/
/
130
0
10
20
30
40
50
Time (s)
Figure 5.3. Capacitancerecoveryvs. time. Threedifferent voltagepulseswere appliedto
this capacitor,0- 0.2 V, 0 - 0.5 V, 0 - 0.7 V, and the capacitancewas recordedwith time.
The capacitors recover from deep-depletionto inversion (Cinv is shown on the plot).
Notice that therecoverytimesincreasewith increasedvoltagepulse.
;
i
.1
t1
!
75
(
capacitance,nj is the intrinsic carrierconcentration,NA is the doping concentrationand tg
is the generationlifetime. However,becauseof the multilayer heterostructure,the details
of generationmust be examined before the above equation can be used and lifetimes
extracted.
Generationprocessesin bulk semiconductorscan be modeled by the ShockleyHall-Readequation:
U
=
2 - np)
VthNT(ni
[5.2]
l/O'n (n + nl) + l/O'p (p + PI)
wheren andp arethe concentrationsof electronsand holes,respectively,Vthis the thermal
velocity of the carriers,NT is the trap density and O'nand O'pare the capturecrosssections
of the trapsfor electronsandholes,respectively,and
nl = nj e(ET-Ei)/kT
[5.3]
PI =nj e(Ei -ET)/kT
[5.4]
and
whereET is the energylevel of the trap andEj the intrinsic Fermi level.
The generationprocessdominatesonly if the np product is less than n~. If we
1
=
assumeO'n=O'p=0' andthatET Ej for simplicity,wefind:
n.12
G=
[5.5]
to (n + p + 2ni)
where to
= (O'NTVth)-I.
From this expression it is clear that the generation rate is
maximized when nand p are both less than nj. In this case,G is commonly written as
ni/tg, wheretg = 2to is calledthe generationlifetime. This expressioncannotbe appliedto
the entire depletionregion becausethe carrier concentrationsdecreasewith a finite slope
when moving from neutral (high carrier concentration)regionsto depletionregions. The
situation is further complicated in our multilayer structuressince the band gap of the
SiO.82Geo.18
layers is less than that of the silicon layers, resulting in a larger intrinsic
carrierconcentration.Under conditionswheren or p is greaterthannj, G can be muchless
76
than nJtg resulting in an over-estimateof the generationlifetime. This subtlety is often
overlookedin many experimentsresultingin reportsof an incorrectlyhigh t.
Assuminga bandgapreductionof 150meV in the Sio.82Ge().18
comparedto silicon
[5], ni in the SiO.82Geo.18
is approximately ni(Si).exp(15OmeV(lkT)
= 2.6 x 1011cm-3 at
room temperature.To illustrate this point. Fig. 5.4 showsthe logarithm of the calculated
carrier profIles as a function of depth into the capacitorat two different times during the
recoveryfrom deepdepletionto inversionfor a gate-substrate
voltagepulse from 0 to 0.7
V. This voltagerangeoperatesthe capacitorin the inversionregimefor the total rangeof
its recovery.
This diagram can be constructedin a straightforward manner from the
capacitanceassumingan elecuon quasi-Fermilevel calculatedfrom the inversion layer
density and a flat hole quasi-Fenni level in equilibrium with the substrate. The
discontinuityin the holeconcentrationreflectsthediscontinuityin thevalencebandbetween
the silicon and the silicon-germanium layers. As was stated previously. maximum
generationwill occur in regionswhere both the electronand hole concentrationsare less
than the local intrinsic carrier concentration,ni (ni is shownas the dashedline in Figure
5.4). During the recovery.this region is seento samplethe Sio.8~eo.18
layer. meaning
that the generationin the Sio.82Geo.18
layer is not artificially suppressedbecauseof high
carrier concentrations. This understandingof the generation-recombination
processesin
semiconductors
is neededwheninterpretingthe experimentaldata.
A standardZerbstplot is shownin Fig. 5.5 for a Sio.82GCO.18buried layer strucmre
with a thennal oxide. However.beforerelating the slopeof the Zerbstplot to the lifetime
asis generallydone,two points must be made. If we assumeas a worst casethat all of the
generationoccursin the silicon-gem1aniumfilms, the actualgenerationvolume (Sil-xGex
only) is lessthan the total generationvolume (includesSil-xGex and Si). Examinationof
Fig. 5.4 shows this correction, which reduces the extracted Sit-xGex lifetime, by
approximately50%. Also, the generationrateand hencethe recoverytime arerelatedby G
=ni/tg.
Sincethe intrinsic calTierconcentration(ni) dependsupon the bandgap,one must
77
18
I . d~1~":"
Ii'
l
Silicon
17
-'e
16
~
14
~
I
~~ In~' . . ~;1;:nft
. . I . . .
SiO.82GcO.18
Silicon
n
/
15
'
-
170pF
p\
I
(.)
--
145pF
c
--
bO
13
j
12
~
I
.
,
,.
11
n.(Si)
I
.~
n.(SiGe)
,
.
10
0
20
40
60
Depth(nm)
80
100
Figure 5.4. Logarithm of the catTierprofiles as a function of depthinto the semiconductor
at two points in die recoveryfrom deep-depletionto inversion. The dashedline represents
the carner concentrationswhenthe capacitanceis 145pF while the solid line representsdie
carrier concentrationswhen the capacitanceis 170 pF. The discontinuity in the hole
concentration corresponds to the valence band discontinuity at the silicon/silicongennaniuminterface. The effective generationregion is seento samplethe Sio.82Geo.18
layer.
78
0.35
-~
0.3
...
-
0.25
N~
0.2
~
~0
~
"0
0.15
0
0
0.05
0.1
0.15
0.2
0.25
0.3
(C/C - 1)
Figure 5.5. Zerbst plot of a recovering capacitor. The Zerbst analysis is effectively
plotting the generationvolume vs. the generationrate and extracting a lifetime from the
slope. The lifetime is directly proponional to nj and inverselyproponional to the slopeof
the Zerbstplot. The lifetime of this capacitoris - 2 ~.
79
use ni for Sio.82Geo.18when calculating 'tg from the recovery time (slope of the Zerbst
plot). Since generationis a them1ally-activatedprocessof emissionfrom traps to band
edges,we expectfastergenerationin a narrowergapmaterialfor identicaltrapdensitiesand
cross sections. (Lifetime, by defmition, dependsonly on trap density and capturecrosssection). A consequence
of this is that the generationrate (G) in the silicon-gennaniumis
muchgreaterthanthatin silicon for the samegenerationlifetimes.
5.3.1.2 Data and Interpretation
The generation lifetime (extracted with the Zerbst technique using the proper
corrections) of the low-oxygen content «1018 cm-3) Sio.82GeO.18buried layers is
approximately 1.45~ for the capacitorwith a thermal oxide. The lifetimes of the high
oxygencontentfilms (> 1019cm-3)weremuchlower andwerelimited by the measurement
set-up (- Ins). This lowerlifetime suggests
thata deeplevel ttap residesin theoxygen
dopedlayers. Severalcapacitorson eachwafer weretestedfor lifetimes and typical values
are seenin Table 5.1. Theselifetimes were found using an approximationto the Zerbst
analysisdescribedby Schroderand Guldberg [1]. The generationlifetime found above
using the Zerbst analysisand that found using the approximatemethod differ only by a
factor of approximately 1.5,
The recovery times (tr), shown in Table 5.1, are defined as the time required for the
capacitor to reach 90% of its final capacitance value after application of the voltage pulse.
This parameter illustrates the difference between the buried silicon-gennanium capacitors
and the all silicon devices due to the difference in the inninsic carrier concentrations. The
low-temperature all-silicon capacitors (grown at 700 .C) have total recovery times
approximately 20 times longer than their silicon-germanium counterpartswhich were
fabricated in parallel.
The larger recoverytimes in the silicon devicesdemonstratethat the generationin
the buried Sil-xGex devicesis most likely in the Sil-xGex layer and not at the Si/Si~
interface. (Both sampleshad Si at the SiOl interface.) If the generationwere surface-
80
Table 5.1. Recoverytimes and generationtimes of capacitorswith different oxides and
one with a large concentrationof oxygen. The silicon heavily-dopedwith oxygen has a
generationlifetime shorterthan the resolution of the measurement.The lifetimes of the
silicon-germanium layers were calculated using the increased intrinsic carrier
concentrationfor SiO.82Geo.18.
The silicon deviceswere pulsed from 0 V to 3 V while
the silicon-germaniumdeviceswere pulsedfrom 0 V to 0.7 V.
81
dominated, one would expect the total recovery time of the Sil-xGex and the silicon
capacitorsto be similar for largevoltagepulses. Furtherevidencethat the generationis not
surfacerelatedcomesfrom comparingthe lifetimes of the capacitorswith thermaloxides
andthosewith plasmadepositedoxides. We seethat the depositedoxideshavelifetimes of
the sameorder of magnitude. Although the silicon capacitorshad recovery times
- 20
times longer than that of the Sil-xGex capacitors,one finds similar lifetimes in the two
materialsbecauseof the differencein their intrinsiccarrierconcentrations.This implies that
the silicon and silicon-germaniumlayershavesimilar trap densitiesandproperties.
We alsoperformedtemperaturedependentmeasurements
of the total recoverytime
on someof the capacitors.Sincethe generationof carriersis a thermally activatedprocess
from trapsto the bandedges,this givesus a measureof the trap level within the bandgap.
An Arrheniusplot of the recoverytime is shownin Figure 5.6. From the slopewe find an
activationenergyof 0.6 eV which correspondsto the energydifferencebetweeneither the
trap level and the conductionbandor the trap level andthe valenceband. Our assumption
of the midgap trap level in Eqn.'s 5.3 and 5.4 provesto be reasonablesincethe band gap
of silicon is 1.12eV at room temperature.
Since we know the trap level and a maximum carrier lifetime, we can estimatea
capturecrosssectionof the oxygencontaminant. If we assume,as a worst case,we have
one trap per oxygen (10-20cm-3for < Ins lifetime), the minimum capturecrosssectionis
10-18cm-2. This is a lower limit since t was the upper limit due to the measurement
resolutionfor shon lifetimes.
By pulsing an MaS capacitor we have probed the minority carrier lifetimes in
oxygen-dopedand oxygen-freeepitaxial layers. The effect of oxygen on the lifetimes of
the films is dramatic,reducing the lifetimes by at leastthreeordersof magnitude,clearly
demonstratingthe detrimental effects of oxygen on the minority carrier lifetimes. The
lifetimes of the low-oxygen concentrationfilms are the longest lifetimes reponed for
silicon-germaniumstrainedlayersgrown by any technique.The long lifetimes demonstrate
82
103
--~
'--'
.-8
102
E
~
=0.6 eV
A
'c:'
~
>
0
U
~
~
~
101
~
~
10°
4.1
4
3.9
3.8
3.7
3.6
3.5
3.4
3.3
1000(r(K)
Figure
activation
5.6.
Temperature
energy
of
this
dependence
plot
is 0.6
of
recovery
time
for
an
all-silicon
device.
The
eV.
83
:1
I
I
that one can make low leakagejunctions with this material as neededfor heterojunction
bipolar transistorsandp-i-n photodiodes.
5.4 Recombination Lifetime Measurements
The recombination lifetimes were studied by analyzing the base current
characteristicsof HBT's. There is an intimate relationshipbetweenthe gain of a bipolar
transistorand the minority carrier lifetime in the base. The gain of the transistorincreases
with the minority canier lifetime in the baseregion. Long carrier lifetimes arerequiredto
minimize base currents, thus maximizing current gain.
We also used the
photoluminescenceproperties of the silicon-germaniumstrained layers to qualitatively
measurethe recombination lifetimes of the films. Long recombination lifetimes are
requiredin indirect bandgapmaterials(suchassilicon,gennaniumand silicon-gennanium)
in order to seeband-edgeluminescence.By studyingthe propertiesof the films with and
without oxygencontamination,we havedetenninedthe limiting effectson carrier lifetime
due to oxygenin theepitaxiallayers.
5.4.1 Heterojunction Bipolar Transistor Characteristics
We studiedthe basecurrentcharacteristicsof Si/Sil-xGex/Siheterojunctionbipolar
transistors(HBT's) to detenninethe lifetime limiting effectsof oxygenon minority carner
recombination lif8times. HBT's are important electronic devicesfor high speedcircuit
applications becausethey allow heavy basedoping while maintaining a high injection
efficiency becauseof the bandgapoffset betweenthe silicon emitter andthe Sil-xGexbase.
Most HBT work has beenreservedfor compoundsemiconductorsystemssuch as
GaAs/All-xGaxAs [6] and InP/Inl-xGaxP [7] but the potential compatibility of Sil-xGcx
with silicon VLSI manufacturingtechnologyis especiallyattractive. It is also possibleto
studythe effectsof oxygenon homojunctiondevices.but for historicalreasonswe choseto
study Sil-xGex-basedHBT basecurrents.
In the [«st Si/Sil-xGex/SiHBT's grown by low temperatureCVD, by C.A. King et
al. [3] at Stanford University, oxygen concenttationsreachedlevels of 1020cm-3 in the
84
baseregions. Thesewerethe first HBT's in silicon-germaniumto exhibit an ideal collector
currentincreasewith the addition of gennaniumto the base. At the time, it waswondered
if the oxygen was actually necessaryfor ideal Ic transistor characteristics. In these devices.
however. the base currents were not ideal and the high oxygen content in the base region
was blamed for the degradation. By removing the oxygen from the films, we were able to
study the effect of oxygen contamination on the basecurrent characteristics and also report
the first ideal base currents in Si/Sil-xGex HBT's grown in a non-UHV environment [8],
The following discussion assumesa knowledge of bipolar transistors - an excellent
referencefor the physicsofHBTs can be found in the doctoralthesisofE.I. Prinz [9] and
alsoa review article by lyer [10],
5.4.1.1 Theory
Figure 5.7 showsthe banddiagramsfor a silicon homojunctionbipolar transistor
(Bm and a Sil-xGexHBT (both npn structures).The layersof the HBT are similar to the
semiconductorlayersusedin the capacitorof Figure 5.2 with the appropriatedoping types.
The collector and emitter regionsof both devicesareformed in silicon but the baseregion
of the HBT is fonned by strained Sil-xGex and hence has a narrower band gap. The
collectorCmTentin the deviceis detenninedby the numberof electronswhich can sunnount
the potential barrier between the emitter and base and traverse the base region to be
collectedat the oppositesideof thedevice(collector). Becauseof the bandalignmentof the
Sit-xGeXtthe banier seenby the electronsat the emitter-basejunction is lower in the HBT
than in the BJT by an amountAEg. This lowering of the barrier in the HBT leads to an
exponential increase in the collector current over the BIT (Ie HBT - Ie BIT
exp(AEg/kT)[ll]).
Since the transistorgain (8) is proportional to !C. 8 increasesby the
sameexponentialfactor.
The base current, on the other hand, is ideally independent on the band gap
differenceof the Sil-xGex. In general.thereare four main sourcesof basecurrentin
heterojunction bipolar transistors. The primary source of base current in injection-
85
Silicon
0
6.EG I'
Ec
..
t
~ "'"
~,~
.
.
Slo.8GeO.2
""
,
'"
Silicon
Ic
'--.
'-
G IB
Ev
Base
Emitter
Collector
Figure 5.7. Band diagram for a heterojunction bipolar transistor and a silicon
homojunction device. The the baseregion of the HBT has a narrower band gap which
leads to a lowering of the conduction band in the HBT compared to the BJT. This
lowering of the conduction band increasesthe gain of the device. Also shown are the
collector current and the basecurrent.
86
efficiency limited devicesis due to the backinjection of holesfrom the baseto the emitter
(or holes in the baseregion which surmountthe potential barrier to be injected into the
emitter) (lb} in Fig. 5.8(a». As can be seenin Fig. 5.7. the barriers experiencedby the
holes in the HBT and the BJT are identical, and consequently the base currents of the two
devicesshouldbe equal,to first order. The basecurrentsare,however,greatly increasedif
the minority carrierlifetimesin the silicon-gennaniumarevery low
The three other sourcesof base current are neutral base recombination (IbV.
recombinationin the narrow bandgap materialof the emitter-basedepletionregion (Ib3).
andrecombinationon the emitter side (lightly dopedside)of the emitter-basemetallurgical
junction (lb4) (seeFigure 5.8). The total basecurrent density in a narrow band gap base
HBT with uniform base doping and gennanium profiles. where the minority carrier
diffusion lengthis much greaterthanthe basewidth. canbe expressedas: [12]
(DE nEi
2
1 W B n2
Bi
1 .7 n 2Bi XB )
JB - q+
+
exp
NE WE
2 'tB NB
+2
NB 'tBB
(~
1 q DEi WEB exp 2kT
'tBE
~)
kT
[S.6]
where DE is the minority carrier diffusion length in the emitter, WB and WE are the
Jretallurgicalwidthsof the baseandemitter,respectively,'tB is the minority carrierlifetiJre
in the neutral base.'tBB is the lifetime in the base-emitterspacechargeregion on the base
side of the metallurgical p-n junction (narrow gap) and 'tBE is me lifetime on me emitter side
of the base-emittermetallurgicaljunction (wide gap). NB is the basedoping,NE the emitter
doping, and XBthe depletionwidth extendinginto the baseregion from the metallurgical
junction. We distinguish the differencebetweenthe inbinsic carrier concenttationsin the
base (nBU and emitter (nE) regions. However, they are related, to IlI'St order, by nBi = nEi
exp(L1Bg/kT). The first tenDin Eqn. 5.6 is due to back injection of holes into the emitter
(Ibl). the secondtenDis due to neutralbaserecombination(Ib2). and the third tcnt1is due
to recombinationin the emitter-basespace-charge
regionextendingfrom the heterojunction
87
Base
Ec
Bv
(I)
Si-Ge
Ec
Ev
(b)
Figure 5.8. Base current componentsin bipolar transistors. (a) the four primary base
currents of a bipolar transistor are: Ibl due to back injection, Ib2 due to neutral base
recombination, Ib3 and 1b4due to recombination in the depletion region between the
emitter and the base. (b) is an expandedview of recombination in the space-charge
region. 1b4is the recombinationon the lightly doped emitter side of the junction while
Ib3 is due to recombinationin the depletionregion on the heavily-doped,n3n'Ow-gapside
of the junction. For Ib3 to dominate,the lifetime in the heavily doped material must be
much shonerthan the lifetime in the lightly dopedmaterial.
88
to the neuttal baseand the founh tenDis recombinationin the spacechargeregion on the
emitter sideof the p-n junction.
The first two terms and the founh tenD of Eqn. 5.6 are standardbase current
componentsassociatedwith bipolar transistorsand are describedin detail in Ref. [23] of
0l.1.
CurrentcomponentsIbl and Ib2havean exp(qVEBlkT)dependenceandare temled
'n=I' currents referring to the ideality factor associatedwith p-n junction diodes. The
slope of an 'n=I' cunent on a semi-log plot of current and voltage is
- ro mV/dec.The
founh cunent (Ib4)componenthasan exp(qVEBf2kT)dependenceand is tenDedan 'n=2'
current, which on a semi-log plot has a slope> 60 mV/dec. The third tenD is different,
however, from typical recombinationcurrent designationsand is also an cn=}' current
component. Appendix I containsthe derivation of the two spacechargerecombination
currents(Ib3, 1b4)and showsthe subtledifferencesin the assumptionsthat enter into the
their detennination. The characteristicsof '0=1' and 0=2' currentswill be important for
data interpretation.
The gain of the heterojunctionbipolar transistorscan be related to the minority
canier lifetime by dividing Eqn. 5.6 by the collector cunent of an HBT [11]:
JC
-
2
q nBi DN
NB WBexp
(~kT
[5.7]
assuminga flat germaniumanddopingprofile in the base. Since6 = IdIB = JdJB we fmd
thereciprocalof the HBT gain to be:
(
2
1 (~~
WB )
XB WB
8DB NEWE)+ 21 D;;;
+ (1.7
DB'tBEB
)
-.:QE.-
+ '21 (!!Ei
NB WEB WB )
~
tBEE
DBi2
DN
exp
(
::9:YEa:
2kT
[5.8]
The first term, again,dominateswhen hole injection into the emitter is the largestform of
basecurrent, and the secondtenDdominatesif the minority carrier lifetime in the neutral
baseregion is shott. The mird tenDis predominantif mereis a shott lifetime in me emitter-
89
basedepletionregion and the founh tenDis dominantif the lifetime is shon on the emitter
sideof the p-n junction. We will usetheserelationshipsin a later discussion
5.4.1.2 Data and Interpretation
Figure 5.9 is a Gumrnelplot (logarithm of IC and IB VS.baseemitter bias) for an
HBT fabricatedat StanfordUniversity by C.A. King [3]. Also shownon the sameplot is
a homojunctioncontrol devicemadeof all silicon layers(Bn'). This was the first repon of
a heterojunctionbipolar transistorin Si/Sil-xGexgrown in a non-UHV reactorvery similar
to the one used in our experimentsbut without a load-lock. At fIrSt glance. the device
performancelooks quite good - ideal baseand collector current slopes ('n=}' and 60
mV/dec) down to 10-12Amps and a current gain of 400. The HBT showsthe enhanced
collector current as expecteddue to the heterojunction(the baseregion was Connedwith
Sio.7oGCO.30).
but the basecurrent has increasedby a factor of almost 50 over the BJT.
This increasedbasecurrent was not expectedas discussedearlier in this section and is
characteristicof a shortcarner lifetime in the baseregionof the device. It wasreportedfor
this HBT that the oxygen concentration in the silicon-gem1aniumreached 1020cm-3.
From the argumentson n=1 currentgiven above,evenFrom the argumentson n=1
cunent given above,eventhoughthe basecurrentshaveideal slopesdown to 10-12A, it is
possiblethat the basecurrentof this HBT hasbeenincreasedby recombinationin the base
regionof the device. From Eqn. 5.8 therearetwo different lifetimes to consideras the gain
limiting mechanismand we can estimatea rangefor the minority carrier lifetime in this
device. If 'tB « 'tBB neutral baserecombinationwill dominatethe gain. In this casethe
estimated lifetime in the device is approximately 100 ps (DB = 6 cm2N/s and WB = 20 om
- given in King's report). If, however,'tBB« orB(which may be causedby increaseddefect
concentrationnearthe hetero-interface).the third term ofEqn. 5.8 dominatesthe gain and
(from the doping data given) the lifetime in the silicon-gennanium in the emitter-base
depletion region is
-
2 ps. The third condition, where 'tB = 'tBEB (which is reasonablesince
90
-<-c:
10.4
Q)
~
~
~
()
Q)
10
U)
co
CI)
~
0
(,)
Q)
91
both are material characteristicsof the base)is the samecondition as the first sinceWB »
XB. In any case,the gainof the gainof thedeviceis determinedby a carnerlifetime.
Figure 5.10 is a Gummel plot of an HBT and a BJT grown in our reactor and
processedby E. J. Prinz. The gennaniumconcentrationin the baseregion is 20%. The
oxygencontentin this baseregionis below the detectionlimit of SIMS «1018 cm-3). This
HBT exhibits the increasedcollector current as expec~ but also exhibits no increasein
basecurrent over the BJT. This implies the HBT is an injection-limited device. and the
base current is not limited by neutral base recombination or space-chargeregion
recombinationbut rather by back injection of carriersinto the emitter. Since we are not
limited by neutral baserecombination,we can extracta lower limit to the minority carrier
lifetime in the basefrom Eqn. 5.8, and we find it to be ~ 10,(XX}ps comparedto 100ps in
the silicon-germaniumwith an oxygenconcentrationlevel of 1020cm-3of King. This high
recombinationlifetime in the silicon-germaniumwithout oxygenis consistentwith the high
generationlifetime in silicon without oxygen.
The high concentrationof oxygenin the baseregion of the Stanfordheterojunction
bipolar transistordecreasedthe lifetime of the carriersby at leasttwo ordersof magnitude
from the ideal case. By eliminating the oxygen from the silicon-germanium,we have
increasedthe carrier lifetime in the baseanddemonstratedthe first near-idealbasecunents
in Si/Sil-xGcx HBT's (no increaseover a standardBm grown in a non-UHV
environment. The ability to grow high lifetime films at low temperaturein a non-UHV
reactor is very imponant for industry since it avoids unneededcomplexity of the
manufacttlringequipment
5.4.2
Photoluminescence
Photoluminescenceprovides a qualitative meansby which we can determinethe
relative magnitudecarrier lifetimes in the silicon-germaniumlayers. Photoluminescence
(PL) is a basicoptoelectronicpropeny of a material. Electronsand holesaregeneratedby
92
10-2
20% Ge
-- 10-<
'-"
I HBT
4
"'"
-
... "
c"...'"
...
"
"
"
...
=
= 1 0 -6
U
~
X
I BJT
c"
~
~
~
~
~
...
...
"
10-8
.-
.-
...
\ " ,,"
.-
",,'"
U
IBHBT
"
=~0
U
10-
10
IBBJT
10-12
0.1
0.2
0.3
0.4
0.5
Base-emitterBias (V)
0.6
0.7
Figure 5.10. Gummel plot of a PrincetonUniversity HBT. The oxygen concentrationin
the baseregion of thesedevices was below the detection limit of SIMS «1018 cm-3).
The HBT has the enhancedcollector current over the homojunction device as expected
and hasno increasein basecurrent asexpected.This HBT is an injection limited device,
neutralbaserecombinationis not the dominantsourceof basecurrent.
93
laser illumination and light emitted during the subsequentradiative recombinationof the
caniersis detected.A schematicofPL in an indirect bandgap is shownin Figure 5.11.
Conduction-bandto valence-bandrecombinationis an extremely slow processin
indirect band gap materialssuchas silicon and gcm1aniumsincephononsare requiredto
conserve momentum (k).
Since both radiative and non-radiative processes in
semiconductorsoccur simultaneously. the fastest recombination process will tend to
dominateme carrierrecombinationprocess.Relaxationprocessesare analogousto current
distributionin an electroniccircuit containingresistorsof different valuesin parallel;if one
resistor is significantly less than the other(s). most of the current passesthrough this
resistor. In order to see conduction-valenceband PL in silicon-gennanium. the nonradiative(low resistance)pathsmust not shunttheradiativepaths(i.e. we mustincreasethe
non-nKiiative recombination lifetime relative to the radiative recombination lifetime).
The first reportsof band-edgeluminescencein a strainedlayer of Sil-xGexwereby
Terashimaet oJ.[13] in thick epitaxiallayersof 4% germaniumcontentgrown by molecular
beamepitaxy (MBE) on silicon. If more germaniumwas addedto the layers, the layers
relaxed and were not strained. Band-edgephotoluminescenceof silicon-gennanium
strainedlayer quantumwells, however,hasprovenelusive. Noel et ale[14] at the National
ResearchCouncil of CanadahavereportedstrongPL (at 4K) in MBE grown Sil-xGexat an
energy level-ISO meV below the band-edge.Since this is significantly below the band
gap energy, this luminescenceis believed to be due to impurities or defects which
incorporateduring growth and give rise to levels within the bandgap. N~l also recently
explain the phenomenaby luminescencefrom gennaniumclustersConnedduring epitaxial
growth. Other work on strainedSiGelayer superlattices[15.16] hasyielded luminescence
which is almost certainly due to dislocationsrather than the superlatticesthemselves. In
our epitaxial layersgrown by CVD at low temperaturesandcontaininglow concentrations
of oxygen, we have observed the first band-edgephotoluminescenceof the silicongermaniumstrainedlayer quantumwells.
94
Figure 5.11. Schematicof photoluminescencein an indirect band gap semiconductor.
Carriers are excited fom the valenceband to the conduction band by laser illumination
creatingan electron-holepair. The carriersrelax to the lowestenergystateof the banduntil
they recombine. Phononsor alloy scatteringarerequiredto conservemomentumin these
materials. Long carrier lifetimes are neededor carriers will non-radiatively recombineat
traps.
95
The light detectedfrom our samplesis due to exciton recombinationand phonon
assistedtransitionsasopposedto defectrelatedrecombination. An exciton is an electronhole pair which behavesmuch like a hydrogenatom and in somecasesis free to move
throughoutthe material.[17]. The energyof the emitted light from the free exciton (FE)
recombinationpath is the band gap energyof the material reducedby the small binding
energy of the exciton (-15 meV). Excitons can also be bound (bound exciton, BE) to
impurity atomswhich lowers the binding energyeven funher from the free exciton case.
The phononassistedtransitionsoccur at energiesreducedfrom the exciton energyby the
characteristicenergyof the phonons.
Figure 5.12 is a PL spectrumtaken at 4 K, in an FrIR PL spectrometer,of a
Sio.soGeO.20
sttained layer quantum well of width 3 nm grown in our laboratory and
measuredby Thewalt and Lenchyshyn at Simon Fraser University. By reducing the
sampletemperatureto 4 K, many more featurescan be resolveddemonstratingthe long
lifetimes of our oxygen-freesilicon gennaniumlayers. Figure 5.12 is a feature-rich
spectnlm. The peaklabeledNP (at 1015meV) correspondsto the no-phonontransitionof
-
a round exciton in the silicon germaniumalloy andoccurs 20 meV below the band edge
of the Sil-xGex strained layer (after correction for quantum confinement and the
temperaturedependence
of the bandgap). This transitionis alloweddue to the randomness
of the alloy (localized alloy scattering)which satisfiesconditionson k conservation[18].
Also visible are the phonon-assisted
replicas(n-ansverse
optical- TO) correspondingto the
different chemical bondsin the silicon-germaniumalloy ( the Si-Si, Ge-Geand the Si-Ge
roods) and the transverseacoustic(TA) phononreplica. Thesewere the first suchfeatures
observedin strainedSil-xGexquantumwells [19]. At 77 K, the individual phononreplicas
are not resolvableand a broad peak encompassingthe different chemical bondsresults.
Figure 5.13 is a PL spectrumtakenat 77K showing the broadphonon-assisted
peak, note
this PL is due to a free excitonand not the bowxi exciton
96
I"
'.
..c
NP
-.e
10
~
';:
I
.r.f}
=
I
Si-Si
Q)
:s
~
Si-Ge
I
Ge-Ge
I
900
950
TA
1000
1050
Energy (meV)
Figure 5.12. 4K Photoluminescence
spectrumof a 30 A thick SiO.8oGeo.20
strainedlayer.
The no-phonon (NP) line is due to the randomnessof the silicon-germaniumalloy and
occursjust below the band-edgeof the silicon-germanium.The phononreplicasareclearly
visible and correspondto the different chemicalbondsin the material, Si-Si, Ge-Ge,and
Si-Ge. The luminescenceis due to boundexcitons.
97
The photoluminescencepropertiesof our rums are strongly dependentupon the
oxygen concentrationin the Sil-xGex layer. Figure 5.13 showsthe photoluminescence
spectraof two different samplescontainingSio.~eo.20 strainedlayerstakenat 77 K. The
two spectta illusttate the difference in the non-radiative recombination lifetimes of the two
films; notice the factor of 100 difference betweenthe two intensity scales. The upper
spectrumcon-esponds
to a Sio.8oGeo.20
layer of high quality. The lower spectrumof Fig.
5.13 is from the secondsamplegrown after venting the CVD chamberto atmosphere.The
oxygen concentrationof this samplereachedapproximately3 x 1018cm-3in the silicongennaniumlayer asdeterminedby SIMS. From the argumentpresentedaoove,this means
that the non-radiativerecombinationlifetimes differ by at leasttwo ordersof magnitude.
(We cannot assign an absolute number since we do not know the exact ratio of nonradiativeto radiativelifetime in eithersample.)
Increasingthe oxygenconcentrationin the silicon-gennaniumlayersto levels~ 5 x
1019cm-3removesall signsof band-edgePL at 77 K. The elimination of the PL peaksis
due to an introduction of bandgap statesin the alloy. As discussedin the introduction of
this chapter,the presenceof a set of localized traps within the energy gap of an indirect
band gap material provides an efficient path for non-radiative carrier recombination.
Whether the disappearanceof the PL is due to an oxygen complex or the formation of
defects in the material is still unclear. TEM was not performed on the structures to
determinea defectdensityanddefectetchingcould not beperrcm1edbecausethelayer were
too thin.
We have demonstratedthe ability to grow low oxygen content, high lifetime
epitaxial layers at low temperatureunder non-UHV conditions. By avoiding oxygen
contamination in our epitaxial films, we have succeededin demonstrating the first bandedge photoluminescence from sttained Sil-xGex quantum wells qualitatively demonstrating
long carrier lifetimes. MBE samples also have low oxygen concentration, but the best PL
is still from CVD grown material. MBE has a problem with metal contamination or some
98
77 K
.~
-1:".._/~"~~
c
'a
.£
~
...J
~
II
x 100
I~
900
950
1000
1050
Energy (meV)
1100
1150
Figure 5.13. PL spectra of silicon-gennanium layers grown in a well baked system and a
newly vented system. The top spectrum corresponds to a silicon-gennanium sample
grown in a baked-out system and the oottom spectrum COrI'Cs}X>nds
to a silicon-gennanium
layer grown as the second sample after the system was vented. Notice the factor of 100
difference retween the two scales. The luminescence is due to free excitons.
99
other lifetime killing defectincorporatingin the fIlm. We havealsodemonstratedthe first
Si/Sil-xGex/Si HBT's with ideal basecurrent characteristicsand reponed the longest
generationlifetimes measuredin silicon-gennaniumstrainedlayersto date.
5.5
Discussion of lifetime measurements
We havequantitativelyinvestigatedlifetimes in low-temperatureCVD silicon and
silicon gennaniumwith andwithout oxygen. Ghaniet aI. [20] comparedexperimentaldata
of silicon-germanium p-i-n diodes with SEDAN simulations of diodes to extract
By adjusting the minority carrier lifetimes in the diode
recombination lifetimes.
simulations,they estimatedthe minority carrier lifetimes of devicescontaining different
amountsof oxygen. The trend of lifetime with oxygen concentrationis consistentwith
qualitative measurementsof lifetimes by PL. At high concentrationsof oxygen, the
predictedlifetimes are comparableto the HBT data given above(2 ps for [0] - 2 x 1020
cm-3 and 100 ps for [0]
-
8 x 1019cm-3). From the Stanford ttansistorsand diodes and
our generationlifetime measurements,we can estimatea capture cross section for the
camers in the silicon and the silicon-germanium.Again, we assumeone trap per oxygen
atom. The resultsaredisplayedin Table 5.2. The crosssectionsarein the regi~ of 10-17
cm-2 for the different devicesfabricated. The cross sectionmeasuredby the generation
lifetime is limited by the measurement
set-upasmentionedbefore.
The microsU'Ucture
of the oxygenin the epitaxial rums or metal impurities gettered
at the oxygen sitesmay be the causethe short lifetimes associatedwith the heavily doped
samples.Given similarnumbersfor lifetime vs. oxygenconcentration
betweenus and
Ghani,metalimpuritiesareunlikely sincewe would expecta largervariationin lifetime and
cross section. Therefore,the low lifetimes are probably an intrinsic propeny of the high
oxygenconcenU'ation.Oxygencoordinatesasan elecU'onicallyinactive interstitial atom in
bulk Cz-silicon and thereforedoes not becomea recombinationcenter. In silicon films
grown by low temperatureCVD, however,the oxygencoordinatesdifferently asis evident
from the FI1R data of the high oxygen content films shown previously (Sec.3.5). This
100
rOl- cm-3! Material
1<1018
1020
IPrinceton
i Measurement l't
Rec.HBT I >
Sio.80Geo.20
Princeton
i Gen. Cap.
I Si
(sec)
10-8
I
I < 10-9
I
(J
(cm~)
~> 10-17
r~!!Da~YL
Bv + 0.6. or < 3 x 1.0-18
I Eo.- 0.6
Table 5.2. Capture cross section data detem1inedby recombination and generation
lifetime measurements
on different devices.
101
different configurationmay be due to the high concentrationsof oxygen in excessof solid
solubility. Thesepossibleconfigurationsinclude substitutionaloxygen and small SiOl.
plateletsbut the exactconfigurationis not known at this time. However,the configuration
of oxygenin low temperatureCVD fllms is clearly electronicallyactive.
102
5.6
1.
References
D.K. SchroderandJ. Guldberg,"Interpretationof SufaceandBulk Effects Using the
PulsedMIS Capacitor,"Solid State Electr. 14, 1285(1971);
M. Zerbst, " RelaxationEffects at SemiconductorInsulator interfaces," Z. Angew.
Phys. 22, 30 (1966) in german.
2.
H. Tempkin, T.P Pearsall,J.C. Bean,R.A. Logan and S. Luryi, "GexSil-x StrainedLayer SuperlatticeWaveguidePhotodetectorsOperatingnear 1.3 ~m," Appl. Phys.
Lett. 48(15), 963 (1986).
3.
C.A. King, J.L. Hoyt, C.M. Gronet, J.F. Gibbons, M.P. Scott and J. Turner,
"Si/Sil-xGex Heterojunction Bipolar Transistors Produced by Limited Reaction
Processing,"IEEE Elect. Dev. Lett. EDL-I0(2), 52 (1989).
4.
T.O. Sedgwick, "Dominant SurfaceElectronic Propertiesof SiO2-PassivatedGe
Surfacesas a Function of Various Annealing Temperatures,"J. Appl. Phys.39(11),
5066 (1968).
5.
R. People, "Indirect Band Gap of Coherently Strained GexSil-x Bulk Alloys on
<100> Silicon Substrates," Phys.Rev.B 32(2), 1405 (1985).
6.
P.M. Asbeck, M.-C. F. Chang, J.A. Higgins, N.H. Sheng,G.J. Sullivan and K.C. Wang, "GaAlAs/GaAs HeterojunctionBipolar Transistors:Issuesand Prospects
for Application," IEEE Trans. Elec. Dev. TED-36(10), 2032 (1989).
7.
Y.-K. Chen, R.N. Nottenburg, M.B. Panish, R.A. Hamm and D.A. Humphrey,
"SubpicosecondInP/InGaAsHeterostructureBipolar Transistors,"IEEE Elect. Dev.
Leu. EDL-I0(6), 267.
8.
E.J. Prinz and J.C. Sturm, "Base Transport in Near-ideal Graded-baseSi/SilxGex/SiHeterojunctionBipolar Transistorsform 150K to 370 K," IEDM Tech.Dig.
IEDM 90, 975 (1990).
9
E.J. Prinz, "," Doctoral Thesis,PrincetonUniversity (1992).
103
10
S.S. Iyer, G.L. Patton, J.M.C. Stork, B.S. Meyerson and D.L. Harame,
"Heterojunction Bipolar TransistorsUsing Si-Ge Alloys," IEEE Trans. Elec. Dev.
TED-36(10), 2043 (1989).
11. H. Kroemer, "HeterostructureBipolar Transistorsand Integrated Circuits,"
ProcJEEE 70(1), 13 (1982).
12. c.A. King, J.L Hoyt, J.F. Gibbons,"Bandgapand Transpon Propertiesof Sil-xGex
by Analysis of Nearly Ideal Si/Sil-xGex/Si Heterojunction Bipolar Transistors, II
IEEE Trans. Elect. Dev. 36(10), 2093 (1989).
13. K. Terashima,M. Tajima amd T. Tatsumi, "Near-band-gapPhotoluminescenceof
Sil-xGexAlloysGrown on Silicon by Molecular Beam Epitaxy," Appl. Phys. Leu.
57(18), 1925 (1990).
14
J.-P. Noel, N.L. Rowell, D.C. Houghton, and D.D. Perovic, "Intense
Photoluminescence
between1.3 and 1.8 J1mfrom StrainedSil-xGex Alloys," Appl.
Phys. Len. 57(10), 1037 (1990).
15. H. Okumura, K. Miki, S. Misawa, K. Sakamoto,T. Sakamoto and S. Yoshida,
"Observation of Direct Gap Propertiesin GenSimSttained-layerSuperlattices,"J.
Jour. Appl. Phys. 28(11), L1893 (1989).
16. R. Zachai, K. Eberl, G. Abstreiter,E Kasparand H. Kibbel, "Photoluminescencein
Shon-period Si/Ge Strained-layer Superlattices," Phys. Rev. Lerr.64(9), 1055
(1990).
17. 1.1.Pankove,QRticalProcessesin Semiconductors.Dover, New York (1975).
18. J. Weber and M.I. Alonso, "Near-band-gapPhotoluminescenceof Si-Ge Alloys,"
Phys. Rev. B 40(8), 5683 (1989).
19
X. Xiao, C.W. Liu, I.C. Sturm, L.C. Lenchyshyn,M.L. W. Thewalt, R.B. Gregory
and P. Fejes, "QuantumConfinementEffects in Sttained Silicon-germaniumAlloy
QuantumWells," Appl. Phys.Lett. 60(17), 2135 (1992).
104
20. T. Ghani, J.L. Hoyt, D.B. Noble, J.F. Gibbons, J.E. Turner and T.!. Kamins, It
The Exttaction of Minority CarrierLifetime from the Cmrent-voltageChararteristics
of Si/Sil-xGex Devices,"Proc. Mat. Res.Soc. 220, 385 (1991).
105
Chapter 6
Semi-insulating
6.1
Silicon
by O-Doping
Introduction
We have shownthat high concentrationsof oxygenin silicon degradethe minority
carrier lifetimes in the epitaxial layers. Short carrier lifetimes reduce gain in bipolar
transistors, increase recombination currents in diodes, and shunt the radiative
recombination processesrequired for photoluminescence. However, there are some
propertiesof the oxygen-dopedsilicon layerswhich can be exploitedto improve circuit and
device performance. By intentionally introducing oxygen into the silicon layers, we can
pin the Fermi level of the material nearthe centerof the band gap. Since thesefilms are
crystalline (aswas determinedin Sec3.4), we areable to grow crystalline silicon with low
oxygenconcentrationslayerson top of the semi-insulatingoxygen-dopedmaterial. In this
chapter we display the semi-insulating characteristicsof the oxygen-dopedfilms and
discusstheir potentialusesin the silicon materialsystem.We thendemonstratea MOSFET
fabricatedin crystallinesilicon grown on top of the oxygen-dopedsilicon (Si:O).
6.2
Semi-insulating Materials
Semi-insulatinglayersareirnponantto high speedelectronicsbecausethey provide
circuits and deviceswith intrinsically small capacitances.Device interconnectsfabricated
on semi-insulatingsubstrateshavelower intrinsic capacitancesand hencesmallerRC time
constantsthan thosefabricatedon semi-conductingsubsttates[1]. Low source-to-substtate
anddrain-to-substtatecapacitances
arealsogainedby fabricatingMOS transistorson semiinsulating substtates[1].
Semi-insulating layers require minimal electron and hole concentrations in a
material,which meansthe Fermi level is nearmid-gap This requiresthe fonnation of an
intrinsic material. This is difficult to achievein silicon becauseevensmall concentrations
of impurity atoms (1013cm-3 impurity atomsout of 5 x 1022cm-3 host atoms) yield a
resistivity on the order of 1000'o'-cm. Semi-insulatinglayers can be achieved,however,
by inttOOucingdeep-levelnapswithin the bandgapto pin the Fenni level.
A considerableadvantageof GaAs over silicon for high speedcircuits (other than
highercarrier mobilities) is the availability of semi-insulatingsubstrates.Beforeultra-pure
GaAs was available,dtesesubstrateswere Connedby introducing a deep-levelimpurity to
the substratelattice. Chromiumwas the metalof choicenot only becauseof its position in
the GaAs band gap (Ey + 0.74 eV) [2] but also because of its low solid-state diffusion
constant (2 xlo-13 cm-2s-1at fJ:X).C) compared to other deep-level impurities in GaAs [3].
SinceGaAsrequireslow-temperatureprocessingtechniques«700 .C), the chromiumdoes
not diffuse out of the subsb'ate
into the activedeviceareas.
Most deep-levelmetal impuritiesin silicon (Fe,Au, Cr...) areefficient trappingcenters
but diffuse quickly evenat moderatetemperatures.For example,Fe will diffuse 440 ~
in
20 min at 900 .C [3] (a standardprocessingcondition for an implant anneal) [3]. This
makesmetal impurities impracticalas mid-gapimpurities in substratesfor minority canier
devices. We propose using oxygen-dopedsilicon films, Si:O, to function as a scmiinsulatingmaterialon which devicequality silicon canbe grown.
107
6.3 Si:O Electronic Characteristics
We grew oxygen-dopedlayers by the processdescribedin Chapter3. We grew
severaldifferent layeredstructuresfor the variousdeviceswhich we would useto test the
electronic characteristicsof the Si:O material. The electronic characteristicswhich we
measuredwere:materialresistivity.Fenni level positionandcarriermobilities. The device
structuresusedto measurethe electtoniccharacteristicsareshownin Figure 6.1. We chose
to measurevertical transponpropeniesof the rums to avoid parallelconductionchannelsin
the conducting subsU"ates
in horizontal experiments.
6.3.1
Resistivity Measurements
We startedby measuringthe resistivity of the oxygen-dopedlayers using metal(Si:O)-(n-Si) and metal-(Si:O)-(p-Si) Schottky diodes. Figure 6.2 is a typical currentvoltage relationship for the n-type Schottky device. This device has an oxygen
concentration of 2 x 1020cm-3 and a film thickness of 10 J!D1.Notice that the voltage scale
is -1 V to 5 V and the full-scale forward-bias current is less than 10 11A. This
demonsttates the highly resistive nature of the oxygen-doped fllms. Similar results were
obtained for p-type devices. We then measured the resistivity of the films as a function of
temperature using the samedevice structure.
The electron concenn-ation(n) in silicon is related to the Fenni level (Eo position in
a material by:
[6.1]
where Nc is the effective densityof statesfor electronsin silicon and Ecis the conduction
bandenergy.k is the BoltzmannconstantandT the absolutetemperature.Sinceresistivity
(p) is inversely proportional to the carrier concentration (p-l
=q~nn). we obtain the
positionof theFermi level in the materialwith respectto theconductionbandedgefrom the
slope of a plot of resistivity Ys.T-t
We assumethat the dependence of ~n and Ef on
temperatureis smallover the rangeof the measurementFigure6.3 is a temperaturedependentplot of the resistivity of two Si:O layers(with the device structureof Fig. 6.1(a)
108
AI Contact
I
AI Contact
V/////~;~//////.J
1
v/////;:~//////J
p - Substrate
n - Substtate
V//////////////////////.A
V//////////////////////J
(b)
(a)
AI Contact
1
AlContact
V//////////////J
Y/~~~(~:~;:/J
P - Si No Oxygen
p - Substrate
n - Substrate
V//////////////////////.A
V//////////////////////J
(c)
(d)
Figure 6.1. Device strUcturesusedfor testing the electroniccharacteristicsof the oxygen
dopedsilicon layers. (a) Metal- Si:O - n and (b) Metal- Si:O - p were the devicesusedto
measurethe resistivity of the Si:O. (c) n - Si:O -n and (d) p - Si:O - p were used to
detenninethe canier mobilities.
1~
10 ~.
-1
-
"=
~
8
-
6
-
I
.
I
.
I
.
I
.
I
.
-
10 J1IDSi:O
-
4
='
u
2
-
0
-2
1
0
1
2
Bias (V)
3
4
5
Figure6.2. Room-temperature
Current-VoltagecharaCteristics
of the lreta!- (Si:O) - (n-Si)
device. Note that the full-scale currentis only 10 ~ for a 50 V bias. Thisdemonstrates
the high resistivity of the oxygen-dopedsilicon. The deviceareais 1.8x 10-3cm-3.
110
--
e
u 106
I
~
'--'
.-~>
.-
~
.~
Q)
~
105
2.4
2.6
2.8
3
3.2
3.4
lOOO(r(K)
Figure 6.3. Resistivity of Si:O as a function of temperature. The slope of the curves
(activation energy) gives us a position of the Fenni level in the material. The lighterdoped samplehas an activation energy suggestingthat the Fenni level is pinned at By +
0.4 eV or Ev + 0.6 eV. The datawas takenat +4 Volts. The trianglesaredata takenon a
sampledoped to 2 x 1020cm-3 with oxygen, and the circles are data taken on a sample
dopedto 5 x 1019cm-3with oxygen.
111
dopedto levels of 2 x 1020cm-3and S x 1019cm-3. We measuredthe resistivitiesat 4 V
forward bias to insure that any voltagedrop acrossthe Schottkycontactis negligible. The
slopesof Figure6.3
(-0.6 eV) show that the Fermi level is indeedpinned nearmid-gap
We have succeededin pinning the Fenni level near Ev + 0.6 eV or Ec- 0.6 eV. which
implies that the carrierconcentrationis within a factorof four of its intrinsic value.
6.3.2 Mobility
Measurements
We measuredthe carrier mobility properties using the n-Si:O-n and p-Si:O-p
structuresof Figure 6.1. Using the theoryofLampen and Mark [4] to describethe cUlTentvoltage characteristicsof insulatorsexhibiting space-charge-limited
current,we extracted
the carrier mobilities for oxygen-dopedfilms. We mention that we did not perfonn
horizontal b'ansponmeasurement(suchasHall mobility measurements).
therebyavoiding
parallel conductionin the silicon subsU"ate.
The theory of space-charge-limited
currentin
insulatorswith traps is quite involved, and we refer the interestedreaderto the work of
Lampert and Mark [4] for a full description. Here we give a physical picture of spacecharge-limitedcurrentin insulatorswhich containa backgroundof thennalcarriersandttap
levels within the band gap, and we discuss the current-voltage relationship in such
materials.
First we describethe cUITCnt-voltage
characteristicsof an insulatingmaterialwith a
backgroundof thenna! carriersin the absenceof traps within the band gap. As a bias is
applied to sucha material,carriersare injected into the insulator and contributeto spacecharge. Until the injected carrier concenb"ationsurpassesthe backgroundspace-charge
concentration,the chargeof the carriershaslittle effect on the el~c
field andOhm's Law
is obeyed. This con'espondsto a slopeof one on a logarithmic plot of current and voltage
(Figure6.4(a». The data Figs. 6.2 and 6.3 are taken from the linear regime. When the
injected carrier concentrationexceedsthe backgroundconcenttation,the space-chargeset
up by the injectedcarriersdominatesthe electric field and hencethe voltagedrop. Current
flow in a space-charge-limitedregime is proportional to the square of the voltage (1
112
--:-
~
'-'
-.
bO
j
Log V (arb.)
(a)
--:-
~
bO
j
Log V (arb.)
(b)
Figure 6.4. Current transport in an insulator with a backgroundof thenna! carriers (a)
without traps and (b) with a single trap level within the band gap. For both cases,Ohms
Law is followed at low bias (n=l), but for high bias the squarelaw prevails due to the
build up of space-charge.In (b), however, a suddenjump in current occurs due to the
fIlling of the trap level. A andJcrlowaredefmedin Eqn. 6.2.
113
oc
V2), whichcorresponds
to a slopeof two on a logarithmicplot of currentandvoltage
(seeFigure 6.4(a».
Insulatorswith trapshavea slightly different current-voltagerelationshipthan the
ideal insulator. We assumein the following that thereis a singletrap level within the band
gap and that the Fermi level is at the trap level in equilibrium. Figure 6.4(b) will help to
illustrate the conditionsof currenttransportin the semi-insulatingmaterial with traps. At
low bias and high bias the characteristicslook very similar to the ideal insulator of Fig.
6.4(a). As a bias is applied acrossthe deviceof Figure 6.1 (c), carriersare injected from
the n regioninto the semi-insulatingregion. Until the injectedcarrierconcentrationexceeds
the backgroundconcentration,the currentwill be due to the backgroundof thermalcarriers
in the material and OhmsLaw is obeyed,much like the trap-freecase. On the schematic
plot of logarithm of currentvs. the logarithm of voltageof Figure 6.4(b), this corresponds
to a slopeof one (n=l).
As we continueto increasethe bias acrossthe device,the carriersinjectedinto the
Si:O initially serveto fill the trapswithin the materialsincethe Fermi level is pinnedat the
trap level energy. When all of the trap sitesare filled, the current increasessharply with
voltage becausethe quasi-Fermi level is able to move above the trap level. This
correspondsto a suddenincreasein the currentwith a small increasein voltage. (A factor
of two changein carrierconcentrationcorrespondsto raisingthe Fermi level approximately
0.7 eV). In Figure 6.4(b), this correspondto the region labeled"jump".
Oncethe trapsare filled, all of the injectedcarrierscontributeto space-charge
and
the current characteristicsare similar to the trap-free condition. In Figure 6.4(b), this
correspondsto the regionwherethe slopeis two. The theoreticalexpressionfor the current
at the point wherethe currentjumps from the low currentlevel to the high currentlevel (Jcr
low)is [4]:
Jcr low
=A
q2 n02~ L
[6.2]
E
114
!
!
,
..
where A is the size of the currentjump at the critical point, q the electrOncharge,Dothe
thennal caITier background concenn-ation,~ the mobility, L the length of the insulator and £
thedielecbicconstant
Figm-e6.5 is a logarithmicplot of the currentandvoltagefor a p-Si:O-pdevicewith
a 3 JLmsemi-insulatingregion. By changingthe length of the semi-insulatingregion, we
can shift the critical voltage to different values(Vcr oc L2). Thediodeof Fig. 6.2 (L = 10
~)
doesnot showthejump in cun-entuntil 100V is applied. The featuresof Fig. 6.5 are
similar to thosedepictedin Fig. 6.4(b) demonstratingthe trapping propertiesof the film.
The mobility we extract from the p-Si:O-p device is approximately 100 cm2y-ls-l (for
holes)and for an n-Si:Q-n deviceit is about400 cm2y-ls.l (for electrons). We calculated
no from the position of the Fermi level determined from the temperaturedependent
measurements.Thesehigh mobilities may find applicationin high speedphotoconductive
switcheswhereshortlifetimes andhigh mobility materialsarerequired.
6.4
Silicon on Semi-insulating Layers
Semi-insulatingsilicon layerscan be usedfor many purposes,as discussedin the
openingsectionof this chapter. We now proposeusing Si:O asa semi-insulatingsubstrate
on which to fabricate MOS devices. Figure 6.6 is a device structure which we have
fabricatedto utilize the insulatingnatureof theoxygen-dopedsilicon.
Crystalline silicon on top of an insulating layer is advantageousfor MOS circuits.
The insulating layer providesdeviceisolation without the risk of bipolar latch-upfound in
CMOS circuits [1]. The insulating layer is also useful for reducing the parasitic
capacitances
of interconnectsand source-substrate
and drain-substtatecapacitances.
Devices with ideal sub-thresholdslopesof 60 mV/dec, which is important for switching
deviceson and off, can also be fabricatedby usingthin layers,which can be fully depleted,
on top of the insulator. Most insulators,however.are amorphous(e.g. SiD2) and make it
impossibleto grow crystalline silicon directly on top of the insulator. Silicon-on-insulator
15
10-3
p-Si:O-p
3 ~ Si:O layer
SquareRe~
--<
slope=2
'-'
-c: 10.7
~
Jump
a
OhmicRe~
slope=1
0.01
10
0.1
100
Bias (V)
Figure 6.5. Currentvs. Voltage for a p-i-p structurewith a silicon layer dopedto 2 x 1020
cm-3, with oxygen doped layer forming the intrinsic region. Notice the three regimes;
Ohm's Law, "jump" correspondingto trap filling, and squarelaw due to space-charge.
Thejump indicatesa singletrap level within the bandgap.
116
devicesare now fonned by SIMOX (separationby implantedoxygen) [5] or by growing
silicon on expensivesapphiresubsn-ates
6.4.1 Silicon-on-Si:O
Growth
To demonstratethe feasibility of this device structureusing Si:O as the insulating
layer, we fabricated the device shown in Figure 6.6 using standardMOS processing
techniques. We staned the growth processwith the cleaning sequenceand buffer layer
sequencediscussedin Sec 3.4. For n-channeldevices,the buffer layers were doped to
about 1017cm-3 p-type. and for p-channeldevices,the layers were doped 1016cm-3 ntype. The DCS and H2 flow rateswere 26 standardcubic centimetersper minute (sccm)
and 3 standardliters per minute (slpm),respectively.The temperatureof the substratewas
then lowered to 750 .C for one minute of growth of silicon (-100 A). The hydrogenflow
was then reducedand stabilized(-10 sec)beforethe DCS was switchedoff and the silane
switched on. The oxygen-dopedlayers were grown using 4% Sif14 in H2 as the Source
gassuchthat the total silaneflow was 10 sccmand the total hydrogenflow was2 slpm.
The oxygen was introducedto the reactorin an Argon carrier (280 ppm Ol in Ar)
gasthrough a calibratedmassflow conttoller. The flow of argonwas adjustedto give the
proper oxygen level. The oxygen was introducedinto the reactor 30 sec.after the SiH4
wasinjectedandwasleft flowing for thedurationof thelayer. Typically,a 1.5~ Si:O
layer was usedfor device isolation (30 min. growth). The oxygen-dopedlayer was not
intentionally dopedwith phosphorousor boron.
The cap (active) layer was grown at 1000 .C with silane as the source gas.
Following the oxygen-dopedlayer, a 30 sec. silicon layer at 750 .C (using silane) was
grown
(- 240 A)
before the temperaturewas raised to 1<XX>
.C. The active region was
grown to a thicknessof approximately0.5 ~
type dependinguponthe channelrequired
117
(with silane)and dopedin situ n-type or p-
Figure 6.6. Proposeddevice structureto demonstrateSi:O as a semi-insulatingsubstrate.
The oxygen-doped silicon remains crystalline during growth, allowing high quality
crystalline silicon (without oxygen)to be grown asthe active layer. We fabricatedseveral
p-channeland n-channeldevicesusing this structurewith gatelengthsranging from 2 ~
to 50 ~.
118
6.4.2 FET Processing
We processedthe devices using standard self-aligned SOI-MOS processing
techniques.We startedby isolatingthe activeareaswith a reactiveion etch sothat the mesa
walls reachedthe Si:O layer. We thengrew the 300 A gateoxide at l<XX>.Cin dry Ol and
delivered the wafers to David Sarnoff Laboratories for the poly-crystalline silicon
deposition,
Becausethe Si:O is not a perfect insulator, it conductswhen heavily doped. We
changedthe standardself-alignedgateprocessslightly to accommodatethis situation. We
implantedthe entirepoly-silicon layer for the gatedopingrefore the poly-crystallinesilicon
etch. The implant conditionswere4 x 1015cm-2at 100keV, using boronfor the p-channel
devices and As for the n-channeldevices. We then patternedthe gate by reactive ion
etching and openedthe sourceand drain contactholes. Another implant was performed
under the sameconditions as above to define the sourceand drain regions. To avoid
implanting the Si:O in the field regions,the photoresistusedto pattern the gate was also
used as the implant mask.
The devices were then cleaned thoroughly and the implant damage and gate drive-in
anneals were performed simultaneously. The n-channel devices had a 35 min., 02 anneal
at 950 .C
while the p-channeldevicesunderwentonly a 20 min. annealin 02. Oxide was
then depositedon the surfaceat 350 .C and annealedat 600 .C for 30 min. in Nz The
contact holes were then openedand Al evaporatedand patternedto finish the devices.
Fom1ing gas annealswere neededon some devices to improve the device performance.
Although the structures were not optimized for state-of-the-art perfonnance. the
devices fabricated do operate as FET's. Figures 6.7(a) and (b) are families of I-V curves
for p-channelFET's with 2 ~
is the bulk silicon wafer
gates.Figure 6.7(a)is the Si:O isolateddevice and 6.7(b)
(- 10 ohm-cm resistivity) processedin parallel with the Si:O
device. The semi-insulatinglayer of Fig. 6.7(a) was doped 5 x 1019cm-3 with oxygen.
The higher thresholdvoltage in the Si:O device is due to the doping level in the active
119
0
0
1
I
I
N
~
1
~
~I
1
~
1
~1
1
\0
~I
1
r-
(yw) :aU~llnJ U!"BJQ
N1
(yw) :au~llnJ u!"BJQ
I
00
\01
120
0
1
N
I
~>
1'-"
-cn
ro
;:Q'
'-'
.-=Q)-=
'-'-
~
.so
~Q)
cd-
Q)"O
0.",
->..-
I Q)
cOl)
~
cd
-cd "0
~O
.0>
SO
"'M
Q)Q)
"O~
'tSoS
gQ)
,-,oS
0.=
tIJQ)
~.~g
'"
.
~
Q)
go
= =
0'-
tIJ.-
["-'
~ ..'~.0
uC/)Ol)
C-.
'tSdoS
~
cd ~
Q»
'-'
Q) .~
'-'.c:u
O~cd
cd'> ~
~ Q) .9
~UOl)
e'~ ci
~OS
Ol)"O
Q)'"
~~
~Q)
C/)~
.SO
\I")
.-0 e
"r;5.o
c .S
9:
"0'"
~~
'-'Q)
.0
~ -=
cd~
~Q)
e-.a
Ol)~
~~
cd ~,
~
~
Q
=
~~
1I
1
I
--
~'a
\0
r-I
1
00
0
N
1
I
I~
cn
ro
~> '-"
~
=
.~
~EI 1'"\
~
Q)Q)
~C/)"O
O
-=oS
~ .5
.S
'9
~
\O-c
~"Cd~
~Ol)'-"
~:a
~~~
r-..:'~... u
Q)
.
00
1
W
""
,..
'0
.-4
.,
0
~
[8
0=1VIM
II II
~~
~
.,.
0
<v-)
~
'1'
0
(y-) 1
>
Q
IrI
0
II
>
-0
I
-b
>
>
Q
II
It')
9
..,U!"8JQ
I
>
II')
N
V)
-
V)-
d~
o~
0,
II
dO
,
-
N
-
""!
-
.
I
I
~
-
tQ
~
rI)
=
<=1>0
'1").£
O=
0'-
d~
-
""!
N
-I
NI
"1
I
-
V)
b
,...
.-4
-0
'0
~
..-
>O
121
~
";'
--
~
§.
in
~
"0
..c
~
~
.0
=
~
u
~
g
t.)
-.~
.~
~
-~
= .
.oij
c~
""'>
E--s
~~
OM
~.~
~
~
§
~.~
e.-8
]og
.;..;
~
g.:§
.$
.~~
~~
~~
Os
~§
"2->.
""'.g
~.~
'C ~
£5-
~
w
G~ ..
..c
~~
t.)
OC
~]
..cO
.
Ij
~~
~8
\0.>
a~
~-~
region. The sub-thresholdcharacteristicsare shownin Figures6.8(a)and (b). We did not
designthe Si:O structureswith optimumparameters,but the devicesarecomparable(in the
active region) to MOSFET's fabricatedin parallel on bulk silicon wafers. Both devices
have high leakage CWTentsin the off state which means the source is most likely a function
of devicethe processing,not the silicon quality of the SOl. The n-channeldevicesalsohad
largeleakagecurrentsfor both the controldevicesandthe testdevices.
6.5 Future of Si:O
The possibility of using Si:O as an insulatinglayer is a potentially exciting field of
study. We have alreadydemonstratedthe semi-insulatingand crystalline featuresof the
material. Now other characteristicsmust be studied. Someof the characteristicswhich
needto be investigateddeal mostly with the advantagesinherentto the silicon-on-insulator
(SOl) materialsystem.
SOl devices are inherently radiation-hard. Carriers generateddeep within the
semiconductorby alpha particles are blocked by the insulator and thereby kept from
diffusing into the active device areas. Sincethe Si:O layers contain mid-gap trap levels,
dtesettaps may preventcarriersfrom reachingthe active areasas well. SOl also reduces
short channeleffects in MOSFETs such as preventingpunch-throughof the sourceand
drain regions. Sincethe device sttucturewhich we fabricatedwas not optimized to study
theseeffectsin the Si:O devices.theseeffectscanremainan areaopento study.
There are somecharacteristicsof Si:O which may makeit more advantageous
than
SiO1, as the under-lyinginsulatorfor SOl.
SinceSiO1,is an insulator,makingcontactto the
activeareasubsttateis difficult anda self-biasingeffect [6] causesa changein the threshold
voltage of SOl MOSFETs. Since Si:O is not a perfect insulator. making contact to the
active area substratemay be easier Si02 also haspoor thermalconductivity(0.014
WIcmfC) as comparedto silicon (0.6 WIcmrC) [7]. Therefore. heat is not dissipated
easily from the active device areas.particularly in bipolar devicesin SOl. Since Si:O is
122
more than 99% silicon, the thermal conductivity should be much larger than amorphous
SiD2 andthe power dissipationbetterthanin SOl.
Another potential use for Si:O is as the substrate for high speed metal
semiconductor-metal(MSM) photodetectors.In silicon MSM devices,the initial photoresponseof the device can be as shon as 10 pSt but carriers generateddeep within the
substratelead to a long (>100 ps) tail in the recovery time of the devices [8]. By using
Si:O as a substratelayer, the carriersgenerateddeepwithin the substratemay be removed
quickly becauseof the densepopulationsof n-apstherebyeliminating the long carrier tail.
Preliminary studiesof carrier lifetimes by a pump andprobereflection technique[9] show
the lifetimes in the Si:O layersdopedto 5 x 1019cm-3with oxygento be on the orderof 10
ps.
6.7
Conclusions
Much work is neededto detennineif the Si:O layersnuly improve deviceperformance
to the sameextentasothersilicon-on-insulatorstructures.We havedemonstratedthe semi.
insulating nature of the oxygen-dopedsilicon films by studying the carrier transport
propertiesin the material. We haveshownthat the Fenni level is pinned nearmid-gapin
the silicon and that the films are highly resistive at room temperature. We have also
demonstrated
electronicdevicesfabricatedin crystallinesilicon grownon top of Si:O.
123
6.8 References
1.
K.E. Bean and W.R. Runyan. "Dielectric Isolation: Comprehensive. Current.
Future." J. Electrochem.Soc. 124(1). 5C (1977).
2
S.M.Sze,Ph~sicsof SemiconductorDevices.2ndEd., Wiley, New York (1981).
3
CRC HandbookofChemisErXand Ph~sics,72odEd, (19911992).
4
M.A. Lampert andP. Mark, CurrentInjection in Solids.AcademicPress,New Yark
(1970).
5.
PL.F. Hemment. "Silicon on Insulator Formed by 0+ or N+ Implantation." Proc
Mat. Res. Soc.53. 207 (1986).
6.
J. Tihanyi and H. Schlotterer,"Influence of the Floating SubstratePotential on the
CharacteristicsofEFSI MOS Transistors,"Solid StateElectr. 18,309 (1975).
7
A.S. Grove, Phxsicsand Technolo&xof SemiconductorDevices,Wiley, New York
(1967).
8.
Y. Uu, W. Khalil, P.B. Fischer, S.Y. Chou, T. Y. Hsiang, S. Alexandrou and R.
Sobolewski,"NanaoscaleUltrafastMetal-semiconductor-metal
Photooetectors,"
Dev.
Res. Conf., Boston (1992).
9
F.E. Doany, D. Grischkowskiand C.-C. Chi, "Carrier Lifetime Versus lonimplantationDosein Siliconon Sapphire,"
Appl.Phys.Lett. 50(8),4ro (1987).
124
Summary
In the previous six chapterswe have describedthe statusof oxygen in epitaxial
silicon and silicon-gennaniumlayersgrown by low-temperaturechemicalvapordeposition.
The motivation for this study stemsfrom the drive to increasedevice and circuit
performanceby improving the electronicmaterialin which the devicesarefabricated. Low
temperatureepitaxy provides a method for growing thin epitaxial layers with abrupt
interfaces. It can be usedwith silicon technologyto grow Si/Sil-xGex heterojunctionsto
improvedeviceperronnance.
The obstaclesassociatedwith low-temperatureCVD epitaxialgrowth techniquesare
rooted in the stability of oxygen on the silicon surface. We demonstratedthat oxygen
concentrationsfar abovethe peaksolid solubility canincorporateinto epitaxial silicon and
silicon-germanium layers grown at low-temperatures. These high concentrations of
oxygen were shown to reduce minority carrier lifetimes in heterojunction bipolar
transistors,p-n junction diodes,and MOS capacitors.The shortcarrier lifetimes associated
with oxygen contaminationof silicon-germaniumalloys also reducesthe non-radiative
recombinationlifetimes of the carriersin the Sil-xGex layers so that photoluminescence
processesare masked.
We also showedthat hydrogenpassivationof the silicon surfacecan reducethe
probability that oxygenwill stick to the growing interfaceand that the boundarylayer can
reducethe transportof oxygen to the growth interface,thereforereducing the amountof
oxygen incorporating into the growing film. This is a critical advantageof CVD over
MBE, and makesCVD possiblewith oxygen and water partial pressuresfar in excessof
what is predictedby classicalsurfacesciencestudies. We proceededto show that highlifetime, low oxygen concentration epitaxial films could, indeed, be grown at low
temperatUre
in a non-UHV environment.andthat improveddeviceperronnanceresulted.
In Chapter6, we exploited the electtonic propertiesof oxygendopedsilicon rums
for use as semi-insulatingepitaxial layers. We demonstratedthe crystalline and semiinsulating natureof thesefums and showedthat we could grow high-quality silicon f1lms
with low oxygen concentrationson top of the oxygen doped layers. This structure
provides many opportunitiesfor improved device performancesuch as MOS transistors
and high speedphotoconductors.
126
Appendix I
Recombination Currents in P-N Junctions
Here we discussthe different currentsassociatedwith recombinationin the spacechargeregion of p-n junctions. As seenin Chapter5, an extra term, due to recombination
in depletion region extending into the narrow band gap base, has been added to the
conventionalrecombinationcurrents. This extra current componenthas a slope of one
("n = 1") on a semi-logarithmicplot of current vs. voltage and is a dominant term if the
lifetime in the narrow bandgap region is short. The discussionof recombinationcurrent
is found in detail in Grove (Ref. 1 of Chapter4).
Recombinationwithin the space-chargeregion of a p-n junction gives rise to a
currentcomponentdescribedby:
W
Jrec = f U dx =
0
0
xp
f U dx +
f U dx
-Xn
[AI.I]
0
where W is the width of the space-chargeregion, Xn and xp are the depletion widths on
the n- and p-sides of the junction, U the Shockley-Hall-Read (SHR) (see Eqn. 5.2)
recombinationequationand Jrecthe recombinationcurrent-density.With the simplifying
assumptionsof mid-gaptraps (Et = Ei) and equalelectronand hole capturecrosssections
(an = ap = a) the recombinationrate,U, is given by:
pn - ni2
U = avthNt n + p + 2ni
[A1.2].
Under the assumptionof quasi-equilibrium (quasi-Fermi levels constant), the electron
hole productis constantthroughoutthe spacechargeregion and is given by:
pn = ni2 exp ~kT
[A1.3]
( )
whereVp is the forward biasvoltage.
For a given forward bias, U will have a maximum value at the location in the
space-charge
region wheren =p sincenp is a constant. This condition is satisfiedwhere
the intrinsic Fermi level (Ei) is halfway betweenthe quasi-Fermilevels for the holesand
the electrons,therefore,the carrier concentrationsare given by:
n
( )
= p =ni exp ~2kT
[A1.4]
and hence,U max is given by:
ni2( exp (W)
Umax = avthNt
)
-1
IV I
2ni(exp(~)
[A1.5]
+1)
or for Vp» kT:
Umax=2~exp (~
2kT)
1 ni
[A1.6]
where 'to = ovthNt and is the carrier lifetime. Assuming that the lifetime is the sameon
both sides of the junction, the approximation to the integral in Eqn. A 1.1 gives the
standardcurrent associatedwith recombinationwithin the spacechargeregiol! of a p-n
junction,
and
( )
Jrec =21~'to W exp ~
2kT
which is an "n=2" currentas discussedin Chapter5.
128
is
written:
[A1.7]
In the casewherethe lifetime is different on the two sidesof the p-n junction. the
must be evaluatedseparatelyon either side of the junction. In the
integralof Eqn.Al
caseof a bipolar transistorwe separatethe integral of Eqn. A 1.1 into two integrals,one
over the depletion region extending into the emitter. and the other integral over the
depletionregion extendinginto the base. SeeFigure AI.I
In the caseof the npn heterojunctionbipolar transistor with a narrow band gap
base, the baseregion is more heavily dopedthan the emitter (NB - 100NE) and most of
the depletion region extends into the n-type emitter (W
- Xn » xp).
Therefore. the
condition where n =p occurs in the emitter region and the integral over the n-depletion
region can be approximatedby Eqn. AI. 7. This term is denoted1M in Eqn. 5.6.
The integral over the depletion region extending into the p-type base yields a
separaterecombination terD1. In the narrow band gap material of the base, the hole
concentrationis much greaterthan the electronand intrinsic canier concentrations(p » n.
ni). Therefore the simplification to the SHR equation becomes:
.1 n
-
U-2;;
[Ai.8]
where 'tBE is the lifetime in the narrow band gap baseregion and is different than the
lifetime in the wider band gap emitter. This simplification leads to the "n=I" current
componentofEqn. 5.6. The caniers areaccountedfor by the potential which is assumed
to vary linearly acrossthe depletionregion extendinginto the base. The integrationover
the depletion region extending into the heavily doped base region then leads to the
following expressionfor the recombinationcurrent:
J~
-
1.7 Xu ni'-base
exp
(g!.yE!
kT )
[Al.9]
'tBE NB
We have assumedthat the basedoping is 70 times larger than the emitter doping which
leads to the constant 1.7 tenn in the numerator. (The baseto emitter doping-ratio was
taken from King in Ref. 3 in Chapter5 wherebygiving the sameexpressiondescribedby
King). This expressionis the "n=!" term designatedIb3 in Eqn. 5.6 and is a dominant
129
Figure AI. I. Expandedview of recombination in the spacecharge region. 1b4is the
recombination in the lightly doped emitter side of the junction while Ib3 is due to
recombination in the depletion region on the heavily-doped, narrow-gap side of the
junction. For Ib3 to dominate,the lifetime in the heavily doped material must be much
shorterthan the lifetime in the lightly dopedmaterial. This condition yield an "n=I" base
cunent.
130
term for recombination current if 'tBE is much shorter than the other lifetimes in the
device. This "n=1" current is characteristicof a short lifetime on the heavily dopedside
of the p-n junction and is not limited to heterojunctions or heterojunction bipolar
transistors.
131
I
i
!
Appendix II
Publications and Presentations Related to this Work
Journals
1. J.C. Sturm, X. Xiao, P.V. Schwartz,C.W. Liu, L.C. Lenchyshyn,and M.L.W.
Thewalt,to bepublishedin J. Vac.Sci.andTech.BI0 (4), July/Aug(1992).
2
L.C. Lenchyshyn, M.L.W. Thewalt, J.C. Sturm, P.V. Schwartz, E.J. Prinz, N.L.
Rowell, J.-P. Noel, and D.C. Houghton, "High Quantum Efficiency
Photoluminescencefrom Localized Excitons in Sil-xGex," Applied PhysicsLetters
60(25), 3174 (1992).
3.
V. Venkataraman,P.V.Schwartz and J.C. Sturm, "Symmetric Si/Sil-xGex Twodimensional Hole GasesGrown by Rapid Thermal Chemical Vapor Deposition,"
Applied PhysicsLetters 59(22),2871 (1991).
4. J.C.Sturm,P.V. SchwartzandE.J.Prinz, "Growthof Sil-xGex by RapidThermal
ChemicalVaporDepositionandApplicationsto Heterojunction
BipolarTransistors,"
J. Vac. Sci. and Tech. B9, 2011 (1991).
5.
,:;\f;,
Z. Matutinovic-Krstelj, E.J. Prinz, P.V. Schwartzand J.C. Sturm, "Reductionof the
p+
-n+
Junction
Tunneling
Current
for
Base
Current
Improvement
in
Si/SiGe/Si
Heterojunction Bipolar Transistors," IEEE Electron Device Letters EDL-12(4), 163
(1991).
"i:t:,;,
1;:~f
,""
c;;;:';'
~~i't;
';;'
~":;f,
~:
.'; ';
6. X. Xiao, I.C. Sturm,K.K. Goel and P.V. Schwartz,"Fabry-PerotOptical Light
Modulatorat 1.3 ~m in Silicon," IEEE PhotonicsTechnologyLetters 3(3), 230
(1991).
7.
J.C. Sturm, P.M. Garone and P.V. Schwartz, "Temperature Control of SilicongennaniumAlloy Epitaxial Growth on Silicon Substratesby Infrared Transmission,"
Journal of Applied Physics 69(1) (1991).
8
E.I. Prinz. P.M.Garone. P.v. Schwanz. X. Xiao and I.C. Sturm. "The Effects of
Base Dopant Outdiffusion and Undoped Sil-xGex Iunction Spacer Layers in
Si/Sil-xGex/Si HeterojunctionBipolar Transistors." IEEE Electron Device Letters
ED12.2. 42(1991).
9.
l.c. Sturm,P.V. Schwartzand P.M. Garone,"In-situ TemperatureMeasurementby
Infrared Absorption for Low-temperatureEpitaxial Growth of Homo- and Hetemepitaxial Layers on Silicon," Journal of Electronic Materials 19(10), 1051-1054
(1990).
10. P.V. Schwartzand I.C. StUrn1,"MicrosecondCanier Lifetimes in StrainedSilicongennaniumAlloys Grown by Rapid Thennal ChemicalVapor Deposition," Applied
PhysicsLetters 57(19),2004-2006 (1990).
11. P.M. Garone, ].C. Sturm, P.V. Schwanz, S.A. Schwarz and B. Wilkens, "Silicon
Vapor-phase Epitaxial Growth Catalysis by the Presence of Gennane," Applied
Physics Letters 56, 1275 (1990).
12. I.C. Stunn, P.V. Schwartzand P.M. Garone, "Silicon TemperatureMeasurementby
Infrared Transmissionfor Rapid ThermalProcessingApplications," Applied Physics
Letters 56,961 (1990).
3. J.C. Stunn, E.J. Prinz, P.M. Garoneand P.V. Schwartz,"BandgapShifts in SilicongennaniumHeterojunctionBipolar Transistors," Applied PhysicsLetters 54, 27072709 (1989).
Conferences
1.
P.V. Schwanz, C.W. Liu, I.C. Sturm, T. Gong and P.M. Fauchet, "Current
TransponPropertiesof Semi-InsulatingOxygen-DopedSilicon Ftlms for Use in High
SpeedPhotoconductiveSwitches," Electronic Materials Conference,Boston, MA
(1992).
2.
E.I. Prinz, X. Xiao, P.V. Schwartz and I.C. StUrD1,"A Novel Double Base
Heterojunction Bipolar Transistor for Low Temperature Bipolar Logic,"Device
ResearchConference,Boston,MA (1992)
3.
L.C. Lenchyshyn, M.L.W. Thewalt, J.C. Sturm, P.V. Schwartz, E.J. Prinz, N.L.
Rowell, J.-P. N~l, and D.C. Houghton, "Photoluminescence Spectroscopy of
LocalizedExcitonsin Sil-xGex,"ElectronicMaterial Conference,Boston,MA (1992).
4.
C.W. Liu. J.C. StUrIn. P.V. Schwanz and E.A. Fitzgerald. "Misfit Dislocation
Nucleation Mechanismsand Metastability Enhancementof Selective Sil-xGex/Si
Grown by Rapid Thennal Chemical Vapor Deposition." Proc. Symp.Mat. Res.Soc.
238 (1992).
133
5.
J.C. Stunn, Q. Mi, X. Xiao and P.V. Schwartz,"Well ResolvedBand-edgePhotoand Electro-luminescencein StrainedSi1-xGex QuantumWells and Superlattices,"
Proc. Int. Coni on Solid StateDevicesand Materials, in press(1991).
6.
J.C. Stunn, P.V. Schwartz, H. Manoharan,X. Xiao and Q. Mi, " High Lifetime
Strained Si1-xGex Films Grown by Rapid Thennal Chemical Vapor Deposition,"
Proc. Symp.EuropeanMaterial ResearchSociety (1991).
7.
J.C. Sturm, P. Prucnal, Y-M. Liu, H. Manoharan, Q. Mi, P.V. Schwartz and X.
Xiao, "Applications of Si1-xGex Strained Layer Alloys for Silicon-basedOptical
Interconnects," Proc. IEEE Sarnoff Symp.(1991).
8.
J.C. Stunn, P.V. Schwartz and P.M.Garone, "Non-invasive Silicon Temperature
Measurementby Infrared Transmissionfor Rapid Thennal ProcessingApplications,"
Proc. Symp.on VLSI Tech. 10, 107-119 (1991).
9.
J.C. Sturm, Q. Mi, P.V. Schwartz and H. Manoharan, "Band-Edge Exciton
Photoluminescencein Strained Silicon-Gennanium Alloy Films Grown by Rapid
Thennal Chemical Vapor Deposition" PresentedElectronic Materials Conference
Boulder, CO (1991).
10 Z. Matutinovic-Krstelj, E.J. Prinz, P.V. Schwartzand J.C. Stunn, "Reductionof the
p+ -n+ Junction Tunneling Current for Base Current Improvement in Si/SiGe/Si
HeterojunctionBipolar Transistors," Proc. Mat. Res.Soc.220, (1991).
11. J.C. Stunn, P.V. Schwartz, H. Manoharan, Q. Mi, L.C. Lenchyshyn, M.L.W.
Thewalt, N.L. Rowell, J.-P. Noel, and D.C. Houghton,"Well-Resolved Band-Edge
Photoluminescence
from StrainedSil-xGexLayersGrown by RapidThermalChemical
Vapor Deposition," Proc. Symp.Mat. Res.Soc.220, (1991).
12. X. Xiao, K.K. Goel, J.C. Sturm and P. V. Schwartz, "Data Transmissionat 1.3~m
using Silicon SpatialLight Modulator", Proc. SPIE Symp. 1476 in press(1991).
13. J.C. Sturm, P. V. Schwartz, E.J. Prinz and C.W. Magee, "Control of Oxygen
Incorporation and Lifetime Measurementsin Si1-xGex Epitaxial Films Grown by
Rapid Thennal Chemical Vapor Deposition," Proc. SPIE Coni on Rapid Thermal
Processing, 1393, 252-259 (1990).
14. X. Xiao, J.C. Sturm, P.V. Schwartz and K.K. Goel, "Vertical 1.3~m Optical
Modulator in Silicon-On-Insulator,"
Proc. 1990 IEEE SOSISOI Technology
Conference, 192, Key West, FL, October (1990).
15 J.C. Sturm, H. Manoharan, V. Venkataraman,P.V. Schwartz and P.M. Garone,
"Control of Individual Layer Growth Temperaturesby Rapid TemperatureSwitching
in Si1-xGex Multilayer Structures Grown by Rapid Thennal Chemical Vapor
Deposition," Proc. Inter. Confon Elec. Mat.,(1990).
16. J.C. Sturm, P.M. Garone, E.J. Prinz, P.V. Schwartz and V. Venkataraman,
"InterfaceAbruptnessin Epitaxial Silicon and Silicon-GermaniumStructuresGrown
by Rapid Thennal Chemical Vapor Deposition," Proc. Int. Conf on CVD XI 9012, 295 (1990).
134
,
I
i
17. I.C. Stunn, E.I. Prinz, P.V. Schwartz, P.M. Garone and Z. Matutinovic, "Growth
and TransistorApplicationsofSil-xGex Structuresby RapidThennal ChemicalVapor
Deposition," Proc. Topical Symposiumon Silicon BasedHeterostructures,Toronto,
Canada(1990).
18 P.V. Schwartz, J.C. Stunn and C.W. Magee, "Reduction of Oxygen Incorporation
and Lifetime Measurementsin Epitaxial Sil-xGex Films Grown by Low Temperature
Rapid Thermal CVD," Electr. Marer. Conf., SantaBarbara,CA (1990).
19. J.C. Stunn, P.V. Schwanz and P.M. Garone, "Temperature Measurement by
Infrared Transmission for Rapid Thennal ProcessingApplications," Proc. Symp.
Mal. Res. Soc. 157, (1990).
20. P.V. Schwanz,J.C. Sturm and P.M. Garone,"Extreme Supersaturationof Oxygen in
Low TemperatureEpitaxial Silicon and Silicon-gemlanium Alloys," Proc. Symp.
Mat. Res. Soc. 163, (1990).
21 E.I. Prinz, P.M. Garone, P.V. Schwanz, X. Xiao and I.C. Sturm, "The Effect of
Base-emitter Spacers and Strain-dependent densities of States in Si/Si1-xGex/Si
Heterojunction Bipolar Transistors," IEDM Technical Digest, 639 (1989).
22. J.C. Sturm, P.V. Schwartz and P.M. Garone,"Oxygen Incorporation in Low
Temperature(625 - 800 °C) Silicon and Siliron-GermaniumEpitaxial Layers Grown
By Vapor Phase Techniques in a Non-UHV System," Elect. Mater. Conf.,
Cambridge,MA (1989).
23. P.M. Garone, P.V. Schwartzand I.C. Sturm, "Growth Rate and Doping Kinetics in
Silicon-germanium Films Grown by Limited Reaction Processing," Proc. Symp.
Mat. Res. Soc. 146, 41 (1989).
24. J.C. Sturm, X. Xiao, P.M. Garone and P.V. Schwanz, "Electron-Beam-InducedCwTent(EBIC) Imaging of Defectsin Sil-xGex Multilayer Structures," Proc. Symp.
Mat. Res. Soc. 148, 341 (1989).
25 J.C. Stunn. E.J. Prinz. P.M. Garoneand P.V. Schw~ "ElectronTransponin
Silicon-GermaniumStrained-layerHeterojunctionBipolar Transistors."American
PhysicalSocietyMeeting.St.Louis.MO (March1989).
135