Current Density Limits in InP DHBTs: Collector Current Spreading and Effective Electron Velocity

Mattias Dahlström1 and Mark J.W. Rodwell

Department of ECE

University of California, Santa Barbara

USA

mattias@us.ibm.com 802-769-4228

Special thanks to:Zach and Paidi for processing and development work

(1) Now with IBM Microelectronics, Essex Junction, VT

This work was supported by the Office of Naval Research under contracts N00014-01-1-0024 and N0001-40-4-10071, and by DARPA under the TFAST program N66001-02-C-8080.

Introduction

What limits the current density in a HBT?

• Heating– High thermal conductivity InP ☺– Low thermal conductivity InGaAs

– Low Vce ☺

• Kirk effect – Injected electron charge in collector deforms

the conduction band current blocking– thin the collector, increase collector doping

Collector in HBT under current (simulation)and measured effects on ft and Ccb

-2

-1.5

-1

-0.5

0

0.5

0 100 200 300 400

J=0mAJ=1mAJ=2mAJ=3mAJ=4mAJ=5mAJ=6mAJ=7mAJ=8mA

E (

eV

)

Position (A)

-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

50 100 150 200 250 300 350

E (

eV

)

Position (A)

Ec

EvEmitter

Collector

Base

InGaAs

InGaAs

InGaAs

InP

InP

InGaAlAs

At some current density Jkirk device performance will degrade due to the Kirk effect

200

220

240

260

280

300

2 2.5 3 3.5 4 4.5 5 5.5 6

f t (G

Hz)

Je (mA/um2)

We=0.5 m

We=0.7 m

We=0.6 m

16

16.5

17

17.5

18

18.5

0 1 2 3 4 5 6 7 8

Ccb

(fF

)

Je (mA/um2)

Vce

=1.5 V

Vce

=1.3 V

eff

ccc qv

JNN

Current blocking and base push-out effects ft and Ccb – the Kirk effect

High current

0

2

4

6

8

10

0 0.2 0.4 0.6 0.8 1 1.2 1.4

J Kirk

mA

/m

2

Web

(m)

Vcb

=0.75 V

Tc=217 nm

Vcb

=0.3 V

Tc=150 nm

Observation: The Kirk current density Jkirk depends on the emitter width

Jkirk extracted from ft and Ccb vs Je, extracted from S-parameter measurements at 5-40 GHz

Collector current spreads laterally in the collector

=0.14 m for Tc=150 nm

=0.19 m for Tc=217 nm

Sources of error:

Coarse Ic

Ohmic losses reduces Jkirk by max 4 %

Device heating not important - low Vcb

Extraction of the current spreading distance Poisson’s equation for the collector

0

1

2

3

4

5

6

-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1

I Kir

k / L

eb

Web

(m)

Tc=217 nm

Tc=150 nm

2

2'

eb

ebkirkkirk W

WJJ

Averaged data points

22/

222)(2

2

222 ebec

gradesetcc

gradec

gradesetccC

c

bbicbeff

ebeKirkKirk

WLT

TTTTNNqT

TT

TTT

q

EqN

qT

VVv

WLJI

Plot Ikirk/L vs. emitter junction width Web

Current spreading important as emitter width We scales to Jkirk will be much higher !

Poissons equation for the composite collector:

Collector velocity extraction from Vcb

There is no evidence of velocity modulation

0

1

2

3

4

5

-0.4 -0.2 0 0.2 0.4 0.6 0.8

J kirk

(m

A/

m2)

Vcb

(V)

DHBT 17 T

c=217 nm

=190 nm

DHBT 19 T

c=150 nm

=140 nm

∂Jkirk/∂Vcb provides effective electron velocity!

2

2

22/)(2

0,0

222 ebec

cb

cb

effebec

gradesetc

c

bbicbeff

cbcb

Kirk WLqT

V

V

vWLT

TTTT

qT

VVv

VV

I

cbVcbVeffv

Method requires and veff to be constants with regards to Vcb

over measured intervalLinearity of fit indicates this is correct But how can veff be constant with regards to Vcb? -L scattering should lead to velocity modulation!

Tc=150 nm: vsat= 3.2 105 - 3.9 105 m/s

Tc=217 nm: vsat=2.3 105 - 3.2 105 m/s

Why is there no Vcb dependence on veff?

veff is extracted at the Kirk current condition near flat-band at bc interface -L scattering removed from bc interface minimum Vcb influence on veff

-L scattering occurs when electrons in the band scatters to the slower L band veff reducedLarger Vcb -L scattering closer to the bc interface veff reduced

-1

-0.5

0

0.5

1

1.5

0 40 80 120 160

En

erg

y (e

V)

Distance (nm)

EcL

Ec

EcL

Ec

Vcb

= -0.05 V , Je=4 mA/m2

Vcb

=0.2 V , Je=6 mA/m2

Simulated @Je= Jkirk Vcb changes Je= Jkirk(Vcb)

-1

-0.5

0

0.5

1

1.5

0 40 80 120 160

Ene

rgy

(eV

)

Distance (nm)

EcL

Ec

EcL

Ec

Vcb

= -0.05 V , Je=4 mA/m2

Vcb

=0.2 V , Je=4 mA/m2

Simulated @Je<Jkirk Vcb changes Je fixed

Mesa DHBT with 0.6 mm emitter width, 0.5 mm base contact width

Thickness (nm)

MaterialDoping (cm-3)

Description

40 In0.53Ga0.47As 3∙1019 : Si Emitter Cap

80 InP 3∙1019 : Si Emitter

10 InP 8∙1017 : Si Emitter

30 InP 3∙1017 : Si Emitter

30 In0.53Ga0.47As 8-5∙1019 : C Base

20 In0.53Ga0.47As 3∙1016 : Si Setback

24InGaAs/ InAlAs SL

3∙1016 : Si Grade

3 InP 3∙1018 : Si Delta doping

100 InP 3∙1016 : Si Collector

10 InP 1∙1019 : Si Sub Collector

12.5 In0.53Ga0.47As 2∙1019 : Si Sub Collector

300 InP 2∙1019 : Si Sub Collector

Substrate SI : InP

Typical layer composition

DHBT-19 with 150 nm collector

Z. Griffith, M Dahlström

Device results at high current density higher than original Kirk current threshold

0

2

4

6

8

10

12

0 1 2 3 4 5 6

device failure

18 mW/um2

design limit 10 mW/um 2

J max

(m

A/u

m2 )

Vce

(V)

8 m emitter metal length, ~0.6 m junction width

biased without failure (DC-IV)

No RF driftafter 3-hr burn-in ECL

bias points

Low-current breakdown is > 6 Volts

this has little bearing on circuit design

Safe operating area is > 10 mW/um2

these HBTs can be biased ....at ECL voltages

...while carrying the high current densities needed for high speed

0

2

4

6

8

10

12

14

0 0.5 1 1.5 2

J e (m

A/

m2 )

Vce

(V)

Ajbe

= 0.5 x 7 m2 Ib step

= 0.4 mA

Vcb

= 0 V

peak (f, f

max) bias

Tc=150 nm

0

5

10

15

20

25

30

35

1010 1011 1012

Gai

ns

(dB

)

Frequency (Hz)

ft = 369 GHz

fmax

= 460 GHzU

H21

MAG/MSG

Ajbe

= 0.6 x 7 um2

Ic = 35 mA

Jc = 8.3 mA/um2, V

cb= 0.35 V

Conclusions

• Current spreading

0.14 m for Tc=150 nm

0.19 m for Tc=217 nm

(first experimental determination for InP)

• veff=3.2∙105 m/s for both 150 and 217 nm Tc

• Large effect on max collector current for sub- InP HBTs. Jkirk increases drastically

• Must be accounted for in collector isolation by implant or regrowth (provide room for current spreading)