TFAST

1

Low leakage current metamorphic
InGaAs/InP DHBTs with f? and fmax gt 268 GHz grown
on a GaAs substrate
Z. Griffith, Y.M. Kim, M. Dahlström, A.C.
Gossard and M.J.W. Rodwell Department of
Electrical and Computer Engineering University of
California, Santa Barbara, CA, USA griffith_at_ece.uc
sb.edu, 805-893-8044, 805-893-5705 fax now with
Sandia National Laboratories ymkim_at_sandia.gov
505-284-1625, 309-401-9210 fax
2
Motivation for metamorphic InP HBTs
Parameter InP/InGaAs Si/SiGe benefit
(simplified) collector electron velocity 3E7
cm/s 1E7 cm/s lower tc , higher Jbase electron
diffusivity 40 cm2/s 2-4 cm2/s lower tbbase
sheet resistivity 500 Ohm 5000 Ohm lower
Rbbcomparable breakdown fields Consequences, if
comparable scaling parasitic reduction 31
higher bandwidth at a given scaling
generation31 higher breakdown at a given
bandwidth Problem for InP HBTs SiGe has much
better scaling parasitic reduction,
cost Present efforts in InP HBT research
community Development of low-parasitic,
highly-scaled, high-yield fabrication
processes Why metamorphic InP/InGaAs
DHBTs?Access the properties of InP in a GaAs
manufacturing environment with lowered costs
3
Metamorphic HBTs

• Lattice mismatch between substrate and
  epitaxial device layers
• Thick intervening buffer layer needed to
  suppress mismatch defects

4
What are the potential problems ?

• Leakage current through
• threading dislocation

• Reliability issues

• DC performance and leakage
• Rf performance f? , fmax

• Thick buffer layer potentially poor thermal
  conductivity

5
Vertical TEM image of InP metamorphic buffer HBT
DHBT structure
InAlAs current blocking layer
InP metamorphic buffer
1?m

• Defect density drops after 1mm buffer layer
  thickness
• Threading dislocation density of metamorphic
  buffer 107 to 108 cm-2

6
Structure of metamorphic DHBT--mHBT

• 1.5 µm InP metamorphic layer grown at 470?C
• 200 nm collector
• 20 nm setback layer between base and collector
  grade
• 30 nm base with carbon doping at 4?1019 cm-3
• The base has no gradethis will reduce ?eff,
  the effective collector velocity
• Higher collector voltage Vcb will be needed to
  maximize device performance

7
Characterizing InP emitter wet etch0.5 ?m wide
1.0 ?m
Emitter width
s/c undercut
112 nm
Emitter width
1.24 ?m
Contact width
s/c undercut
Contact width

• IQE lattice matched material
• 30 nm InGaAs, 160 nm InP
• Base surface looks cleaned of n InP

• UCSB metamorphic material on GaAs
• 10 nm InAs, 30 nm InGaAs, 120 nm InP
• Base surface looks rough with n- InP

8
Characterizing emitter wet etch1.0 ?m wide

• Metamorphic UCSB material 100Å InAs, 300Å
  InGaAs, 1200Å InP

n- emitter material not clearing underneath
emitter contact
Strand of InP sticking outleakage to base contact
9
SiO2 dielectric sidewall process
A. PECVD deposition
C. Emitter wet etch
B. RIE etch back
D. Base contact deposition
10
mHBT devices with sidewall

• Emitter width ? 0.65 ?m
• Sidewall thickness ? 75 nm
• Base contact width ? 300 nm x 2
• Isolation etch ? 600 nm into InP buffer layer

• Sidewall looks rough, no evidence of shorting
  from pinched TLMs
• total width of base gap resistance ? 75 to
  125 nm

11
Cross-sectional images of mHBT devicerun11
BCB
SiO2 sidewall
BCB
subcollector
metamorphic buffer
Semiconductor undercut during foot removal
Collector and emitter undercut
GaAs substrate

• Note the high planarity of the BCB passivation

12
DC and rf measurement of mHBT w/ lattice matched
results
Common-emitter characteristics
Measured rf gains, DC-30 GHz and 75-110 GHz

• VBR,CEO ? 5.7 V, ? ? 32-37
• Device dimension 0.5 ? 7 ?m2 emitter junction
• 0.3 ?m base ohmic contact width
• 1.3 ?m measured base mesa width

• Measured f? 268 GHz, fmax 339 GHz
• Ic 10 mA and Vce 1.75 V
• Je 2.9 mA/?m2, Vcb 0.8 Volts
• Emitter junction to ambient temp increase ?T ?
  51 K

13
f? , fmax , and Ccb versus bias for mHBT

• Ccb vs bias plot
• 200 nm collector, 30 nm base
• Ccb, minimum ? 6.0 fF
• circuit design figure of merit Ccb/Ic 0.53
  ps/V

• f? and fmax variation under bias
• JKirk 2.0 mA/?m2 for Vcb 0.5 V
• peak f? , fmax 256 GHz, 321 GHz
• JKirk 2.9 mA/?m2 for Vcb 0.8 V
• peak f? , fmax 268 GHz, 339 GHz

14
Comparison of measured f? , fmax for LM-DHBT and
mHBT
Dahlström et al, IPRM 2002

• metamorphic DHBT w/ 200 nm collector thickness
• Base thickness 30 nm, constant doping 4?1019
  cm-3

• Lattice matched DHBT w/ 217 nm collector
  thickness
• Base thickness 30 nm, graded doping 8?5?1019
  cm-3

• Design similarities
• Same emitter design, base-collector grade
  (setback, grade, and pulse doping), collector
  doping of Nd 3?1016 cm-3
• LM-DHBT has higher effective collector velocity
  (base grade) and lower sheet resistance ? higher
  f? and fmax
• mHBT makes up in speed because of increased
  device scaling compared to LM-DHBT

15
Comparison of Thermal and Gummel performance

• Thermal resistance of LM-DHBT and m-DHBT
• Identical base and collector layer structures
• 200 nm collector, 30 nm base
• Average ?JA, LM-DHBT is ? 2.5 ?C/mW
• Average ?JA, metamorphic DHBT is ? 3.1 ?C/mW
• ? 25 increase of ?JA for m-DHBT vs LM-DHBT
• Test conditions Ic 5 mA, Vce 1.5 V, P
  7.5 mW

• Gummel plotmetamorphic DHBT w/ BCB passivation
• 200 nm collector, 30 nm base
• total leakage current ? 90 pA
• lt 10 the leakage seem w/ polyimide
  passivation
• leakage current / base-collector area ? 6.0 pA
  / ?m2

16
Concluding thoughts

• Summary
• Through the use of dielectric sidewall and
  advanced scaling, metamorphic DHBTs from UCSB
  have duplicated performance of their
  lattice-matched counterparts
• Transitioning to planar BCB passivation, leakage
  currents are suppressed
• InP metamorphic buffer layer provides minimal
  possible thermal resistance for mHBTs
• Future work
• Do these devices provide lattice-matched
  reliability? tests need to be performed
• Demonstration of small scale analog and digital
  circuits

This work was supported under by the Office of
Naval Research, programN00014-01-1-0065
17
Conclusion
Thank you
18

Extra slidesremoved due to time
19
Technical Objective
Develop InP-based HBTs for Naval Radar/Comms ICs
better device bandwidth breakdown than GaAs or
SiGe bipolar transistors microwave ADCs,
DACs, digital frequency synthesis
Develop InP HBTs on GaAs substrates for better
cost yield
For low cost, high volume processing ? wafer size
is critical ? compatibility with process tools
in existing GaAs HBT process lines ? GaAs
substrates, processes 6" diameter now Large
InP substrates ? high cost, high breakage,
only 4" available today ? breakage much worse
with 6" wafers grow InP-based HBTs on GaAs
substrates for cost and manufacturability
Requirements for metamorphic HBTs Low substrate
leakage, low junction leakage, low material
roughness high reliability, high performance, IC
demonstration
20
mHBT TLM data, run 11non pinched
SEM of gap for non-pinched base TLM

• Base contact resistance ? 13.6 ???m2, emitter
  contact resistance ? 20-24 ???m2
• Base sheet resistance ? 1000 ?/sq
• -- this is typical for 4?1019 base doping (?p ?
  40), 300Å base on a non-pinched TLM
• With a transfer length 0.12 ?m, thin base
  contacts will be goodneglecting metal resistance
• SEM measurement of TLM spacing used to validate
  Rcont extrapolation

21
Experimental Measurement of Temperature Rise
Temperature rise can be calculated by measuring
IC, VCE and ?VBE
22
Updated interconnect topology for CPW interconnect
Cross-section of final DHBT w/ interconnect

• Interconnects suspended on ? 1.7 ?m thick BCB
  (benzocyclobutene)
• Undoped InP metamorphic buffer layer leakage
  suppressed
• CPW slot-line modes and substrate modes reduced
  some with BCB intervening layer

23
Finished devicehow planar is BCB coating?

• Device / processing details
• Thickness of BCB as spun onto sample 1875 nm
• Height from top of emitter to semiconductor
  floor 1.8 ?m
• Thickness of BCB after planar etchback by ICP
  1667 nm
• SEM shows bulge in BCB over the device ? 215 nm