Device Research Conference 2007

Transcription

1 Devie Researh Conerene GHz t, max InGaAs/InP DHBT in a novel dry-ethed emitter proess Erik Lind, Adam M. Crook, Zah Griith, and Mark J.W. Rodwell Department o Eletrial and Computer Engineering University o Caliornia, Santa Barbara, CA, , USA Xiao-Ming Fang, Dmitri Loubyhev, Ying Wu, Joel M. Fastenau, and Amy Liu IQE In. 119 Bethlehem, PA 18015, USA lind@ee.usb.edu, , ax

2 Latest UCSB DHBT w/ reratory emitter metal New reratory metal emitter tehnology 250 nm emitter width Gain ns (db) U 15 5 MAG/MSG H 21 I =22 ma, V =1.45V e e, =560 GHz t max J e =16 ma/μm 2, V =0.4V Frequeny (Hz) Salable below 128 nm width First RF results Simultaneously 560 GHz t & max BVeo = 3.3V

3 Standard igures o merit / Eets o Saling Small signal urrent gain ut-o requeny (rom H 21 ) nk BT = τ b + τ + C je + C + R qi 1 ( ) ( ex + R ) C 2π τ Thinning epitaxial layers (vertial saling) redues base and olletor transit times But inreases apaitanes Power gain ut-o requeny (rom U) max 8πR τ C e bb, e, Redue R bb and C, through lateral saling Charging time or digital logi More eiient heat transer τ C ΔV I ΔT J e We V πk InP e L ln W e e +L

4 What parameters are needed or THz HBTs? emitter nm width Ω μm 2 aess ρ base width, 20 5 Ω μm 2 ontat ρ olletor nm thik, 5 20 ma/μm 2 urrent density V, breakdown τ GHz max GHz power ampliiers GHz digital lok rate GHz (stati dividers) ρ <1 Ωμm 2 U. Singisetti DRC 2007 Ohmi ontats and epitaxial saling good or 64nm HBT node! Develop tehnology suitable or aggressive lateral saling

5 TiW emitter dry eth ormation Lit-o no good at <300 nm! Dry ething very high aspet ratio Optial Lithography h Mask Plate Cr O 2 plasma ICP Cl 2 O 2 eth ICP SF 6 /Ar eth SiO 2 TiW Reratory Ti 0.1 W 0.9 is thermally stable ρ < 1Ω/ m 2 possible no degradation o ontat resistane or anneals up to 400C

6 TiW dry ething results nm emitters routinely ormed Demonstrated salability down to 150 nm ~ 50-0 nm tapering during eth sets saling limit 150 nm Emitter metal height ~ nm

7 Wet eth does not sale Colletor Base TiW metal InGaAs emitter InP emitter base Minimum emitter width ~ 150 nm! InGaAs eth ~ 30-40n InP eth ~ 40 nm {1}A planes ast {111}A planes slow

8 Hybrid dry/wet eth Anisotropi dry eth InGaAs, part o InP emitter, Cl 2 N 2 Formidable problem InCl x ormation Extensive UCSB know how on dry ething Solution low power ICP 200C, low Cl 2 onentration Short InP wet eth TiW InP InGaAs InGaAs Base

9 Current UCSB TiW emitter proess a b d e SF Cr 6 /Ar ICP SiN x sidewall Cl 2 /N 2 ICP SiO2 Wet Eth HCl:H 3 PO 4 BHF TiW InGaAs n++ InP n InGaAs p++ Base Ti InGaAs n++ InGaAs n++ InP n InP n InP n InGaAs p++ Base InGaAs p++ Base InGaAs p++ Base InGaAs p++ Base 5 nm Ti layer or improved adhesion 25 nm SiN x sidewalls protets Ti/TiW during Cl 2 and BHF eth, improves adhesion Standard triple mesa BCB pasivation Emitter prior to InP wet eth

10 Layer struture nm olletor DHBT Energy (ev V) emitter valene band base J e = 0, 14, 21 ma/μm 2 ondution band olletor distane (nm) Objetive: J e = 0mA/μm 2, 15mA/μm 2, 30mA/μm 2 V be = 1.0 V, V = 0.0 V Thin olletor and base or dereased eletron transit time Colletor doping designed or high Kirk threshold, designed or 128 nm node Setbak and grade thinned or improved breakdown Thikness (nm) Material Doping m -3 Desription 30 In 0.53 Ga 0.47 As 5 19 : Si Emitter ap In 0.53 Ga 0.47 As 4 19 : Si Emitter 60 InP 3 19 : Si Emitter InP : Si Emitter 20 InP : Si Emitter 22 InGaAs : C Base 5.0 In 0.53 Ga 0.47 As 2 17 : Si Setbak 11 InGaAs / InAlAs 2 17 : Si B-C Grade 3 InP : Si Pulse doping 51 InP 2 17 : Si Colletor 5 InP 1 19 : Si Sub Colletor 5 In 0.53 Ga 0.47 As 2 19 : Si Sub Colletor 300 InP 2 19 : Si Sub Colletor Sbt Substratet SI : InP

11 RF & DC data 70 nm olletor, 22 nm base InP Type-I DHBT ) Gains (db MAG/MSG U H 21 I =22 ma, V =1.45V, =560 GHz e e t max J e =16 ma/μm 2, V =0.4V Frequeny (Hz) Emitter width ~ 250 nm First reported devie with t, max > 500 GHz BV CEO ~ 3.3 V, BV CBO = 3.9 V (J e, = 15kA/m 2 ) ) J e (ma/um 2 I, I (A) b Peak t 25 mw/μm 2 Peak max A je = 0.25 x 4.5 μm 2 A j = 0.6 x 5 μm 2 BV eo ~ 3.3 V V e (V) A je = 0.25 x 4.5μm 2, A j = 0.6 x 5μm 2 I b n b =2.01 n = 1.05 I Curren nt Gain Emitter ontat (rom RF extration), R ont < 5 Ω μm 2-11 Base: R sheet = 780 Ω/sq, R ont ~ 15 Ω μm 2 Colletor: R sheet = 11.1 Ω/sq, R ont ~.1 Ω μm V (V) e 0

12 RF data 70 nm olletor, 22 nm base InP Type-I DHBT τ (GHz) V V = 0.4V 0.0V -0.1V J (ma/μm 2 ) e C (F) C / J e A =0.25 x 5.5 μm 2 A =0.6 x 6 μm 2 je j V = -0.1 V 0.0 V 0.1 V 0.2 V 0.3 V V = 0.4 V J (ma/μm 2 ) e C /A (F/um 2 ) je No detetable inrease in C or high J e Indiates that Kirk threshold is not reahed Devie perormane limited by sel heating? Further saling needed or better thermal perormane! C = Q Q τ + Kirk eet inreases C = V...even in DHBTs b b+ = I I V V I

13 Breakdown perormane o InP HBTs Bre eakdown vo oltage, BV (V) e eo? InP/GaAsSb InP/InGaAs InP SHBT This work E Ev E Ev Type II DHBT Type I DHBT τ (GHz) Type I DHBT and Type II DHBT similar BV eo Type II DHBTs has low max due to base resistane SHBT have substantially lower breakdown

14 Current status o ast transistors max (GHz) m 200 GHz GHz 400 GHz 500 GHz 600 GHz = max τ Teledyne 700 UIUC DHBT nm NTT Fujitsu HEMT 600 Vitesse SFU/ETHz 500 UIUC SHBT UCSB nm NGST Pohang SHBT HRL 300 IBM SiGe Updated July nm t (GHz) popular ( τ (1 τ or + τ τ max max + 1 max metris : alone ) / 2 mw/ μm max muh better ) 1 power ampliiers: metris : PAE, assoiated gain, low noise ampliiers: F min digital : ( C ( R ( R ( τ lok b ex bb, assoiated gain, I, hene ΔV / I I + τ ), / ΔV ), / ΔV ), )

15 Emitter Proess will sale to 128 nm & below Emitter-base juntion has been saled down to 64 nm Pure W emitter- no tapering Tungsten 61 nm emitter-base juntion! Emitter metal peeled o during ross setion leave..

16 Conlusion 128 nm node 700 GHz t, xx GHz max easble hallenges: ontat resistivity, robust deep submiron proesses Dry eth based DHBT emitter proess sub 0 nm juntions easible First RF results: 250nm juntions: simultaneous t, max ~ 560 GHz 125 nm devies should ome soon! THz beore next DRC? This work was supported by DARPA SWIFT program and a Swedish Researh Counil grant.