Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Achieving the Best Thermal Performance for
GaN-on-Diamond
J. Pomeroya*, M. Bernardonia, A. Saruaa, A. Manoia, D.C. Dumkab, D.M. Fanningb, M. Kuballa AH.H. WILLS PHYSICS LABORATORY, UNIVERSITY OF BRISTOL, BRISTOL, BS8 1TL, UK BTRIQUINT SEMICONDUCTOR, INC., 500 W. RENNER ROAD, RICHARDSON, TX, 75080, USA
1 H.4 CSICS 2013 Monterey
• Aim: Optimize thermal resistance in GaN-on-Diamond through measurement and modelling
• Review state-of-the-art GaN-on-Diamond transistor versus GaN-on-SiC
• Novel thermal resistance measurement:
• Substrate thermal conductivity
• Interfacial thermal resistance
• Validated transistor model for identifying
thermal bottle necks in GaN-on-Diamond
• Summary
Outline
2 H.4 CSICS 2013 Monterey
Thermal resistances near the HEMT channel:
GaN epilayer
+ GaN/substrate interface
+ Substrate
High RF output power density in GaN-based HEMTs requires improved thermal management
Motivation
3 H.4 CSICS 2013 Monterey
Multifinger GaN HEMT
thermal image
D G S
GaN
Substrate
heat
1µm
Thermal conductivity can be improved up to 5×, replacing SiC->diamond
- interface
Review: GaN-on-Diamond State-of-the-art
H.4 CSICS 2013 Monterey 4
40%
@15W
/mm
0 5 10 15
0
50
100
150
200
250
300
Tem
pera
ture
channel tr
em
pera
ture
ris
e [
oC
]
Power density [W/mm]
• Advantage over GaN-on-SiC already demonstrated: 10.8°Cmm/W (D.C. Dumka, F.4 CSICS 2013)
• How can we improve GaN-on-Diamond even further?
Peak channel temperature derived
from Raman measurement
100µm, 2 Finger HEMT
Historical Development
H.4 CSICS 2013 Monterey 5
GaN
Diamond
AlGaN
Using experimental feedback to aid design
GaN
κ≈160 W/mK
κ≈16 W/mK
GaN
Si
Diamond
AlGaN T.L.
0 5 10 15 20
0
50
100
150
200
GaN/Si
GaN/Tr. layer/Diamond
GaN/Diamond
Te
mp
era
ture
ris
e [
oC
]
Power density [W/mm]
GaN transistors originate from GaN-
on-Si growth
GaN-on-diamond Including
transition layer
Current design, T.L. removed
2x100µm HEMT comparison
Lets examine thermal resistance in more detail
Raman measured temperature
interlayer
Thermal Resistance Components
H.4 CSICS 2013 Monterey 6
1µm GaN
25nm dielectric
95 µm polycrystalline
diamond
Increasing thermal
conductivity along growth
direction
TBReff?: Effective thermal resistance of interface region, including dielectric + diamond nucleation layer (<100 nm)
κdiamond ?: Thermal conductivity depends on grain size. “Bulk” thermal conductivity measurements may be misleading for device modelling
160 W/mK
-> Aim: Separate these thermal resistance contributions
Raman Thermography Depth Mapping
7 H.4 CSICS 2013 Monterey
0.1 1 10 1000
50
100
150
200
250
Log. Depth [m]
Tem
per
atu
re r
ise
[oC
]
contact
GaN TBReff -
Substrate
Confocal depth mapping through transparent uniform materials
TBReff
Substrate thermal conductivity
Ungated transistor as a uniform heat source
Temperature gradient->thermal parameters
GaN substrate
Raman temperature mapping through polycrystalline diamond is challenging: Light absorption and stress variation
-200 -150 -100 -50 0 5040
60
80
100
120
140
160
mes
a ed
ge
cen
tre
Tem
per
aure
[oC
]Position [microns]
Surface Temperature Profile
H.4 CSICS 2013 Monterey 8
Polycrystalline Diamond
GaN
F.E. model of ungated HEMT ¼ cross section
Diamond thermal conductivity
+ GaN/diamond interface TBReff
Fit finite element model by adjusting two parameters:
contact
contact
mesa map
GaN
Diamond interface
• For highest accuracy, we measure the temperature in the uniform GaN layer, rather than the diamond.
-200 -150 -100 -50 0 5040
60
80
100
120
140
160
180
Measurement
Simulation
GaN
tem
per
atu
re [
Deg
. C]
Position [microns]
• Opaque diamond
• Effective diamond substrate thermal conductivity = 710±40 W/mK
• 70% increase over SiC
• GaN/diamond TBReff = 2.7±0.3 ×10-8 m2K/W
• Comparable to typical GaN-on-SiC TBR
Thermal Resistance Measurement
9 H.4 CSICS 2013 Monterey
3.36W
contact
contact
mesa map
Will result in lower transistor thermal resistance than GaN-on-SiC...
Validating Thermal Model
H.4 CSICS 2013 Monterey 10
0.1 1 10
60
80
100
120
140
160
180
Inte
rfac
e
DiamondTe
mp
erat
ure
[D
eg. C
]
Depth [m]
GaN
Simulation
GaN measurement
Diamond measurement
GaN
Diamond
Self consistency between measurement and model
contact
GaN
25nm dielectric
Probed region
1µm
0.5µm
Measurement
3.36W
Opaque diamond enables
measured diamond temperature
to be compared to model
Model input parameters are fixed
• Effective thermal conductivity < bulk thermal conductivity
• Effective thermal conductivity is relevant for transistor modelling
“Effective” Substrate Thermal Conductivity
H.4 CSICS 2013 Monterey 11
0.1 1 10
60
80
100
120
140
160
180
Inte
rfac
e
Diamond
Tem
per
atu
re [
Deg
. C]
Depth [m]
GaN
Simulation
GaN measurement
Diamond measurement
Region of highest sensitivity Lower
Higher
~30µm
contact
GaN
Thermal conductivity gradient
2-D-like heat diffusion
Wafer 2: Higher Grade Diamond
12 H.4 CSICS 2013 Monterey
0 20 40 60 80 100-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8Diamond
s
tre
ss [G
Pa
]
Depth [microns]
GaN
Image taken though diamond substrate
GaN surface temperature mapping approach can still be applied with high accuracy
Raman temperature mapping though translucent diamond is difficult, due to stress variations
• 1200 W/mK effective diamond thermal conductivity
• Thicker 50nm interlayer (w.r.t opaqe wafer), resulting in a 40% higher interface thermal resistance
Wafer 2: Thermal Measurement
H.4 CSICS 2013 Monterey 13
-140 -120 -100 -80 -60 -40 -20 0 20 4020
40
60
80
100
120
140
160 Measurement
Simulation
Ga
N t
em
pe
ratu
re [
De
g.
C]
Position [microns]
4.72W
contact
contact
mesa map
What is the relationship between interface thermal resistance, substrate thermal conductivity and transistor thermal resistance?
Transistor Thermal Model
14 H.4 CSICS 2013 Monterey
0.1 1 10 100
0
20
40
60
80
100
120
140
160
180
200
220
240
260
Tem
pera
ture
ris
e [
oC
]
Depth [m]
Measurement:
GaN Diamond
Simulation:
diamond
Rinterface
W/mK o
Cm2/W
710 2.7x10-8
GaN-on-SiC
GaN Diamond
Source Drain
Gate
GaN
Diamond
interface
1 µm
Opaque diamond wafer, 2×100µm HEMT, Pdiss = 15.3 W/mm
40%
Raman probe
Model validation: Agreement with measured temperatures
Parameters obtained
earlier
0.1 1 10 100
0
20
40
60
80
100
120
140
160
180
200
220
240
260
Tem
pera
ture
ris
e [
oC
]
Depth [m]
Measurement:
GaN Diamond
Simulation:
diamond
Rinterface
W/mK o
Cm2/W
710 2.7x10-8
1400 2.7x10-8
710 0
Reducing Transistor Thermal Resistance
15 H.4 CSICS 2013 Monterey
S
GaN
diamond
G D GaN-on-SiC GaN Diamond
Use validated transistor model to explore thermal resistance components
Current GaN-on-diamond
Increasing diamond thermal conductivity
OR decreasing interface thermal resistance
Eliminating GaN/Diamond interface resistance reduces transistor thermal resistance by a further 35%
• A 40% reduction in channel thermal resistance has been demonstrated for current GaN-on-diamond transistors versus GaN-on-SiC
• A further 35% reduction in transistor thermal resistance could be achieved by reducing the GaN/diamond interface thermal resistance
• A methodology has been developed for characterising the thermal resistance components of GaN-on-Diamond:
• Effective diamond thermal conductivity 750-1200 W/mK
• GaN/Diamond interfacial thermal resistance 2.7±0.3 ×10-8 m2K/W for 25 nm interlayer
Acknowledgement: This work is supported by the DARPA Near Junction Thermal Transport (NJTT) Program, monitored by Dr. Avi Bar Cohen, Dr. Joe Maurer and Dr. Jonathan Felbinger of DARPA. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of DARPA.
Summary
16 H.4 CSICS 2013 Monterey