Case study: electromagnetic radiation...
Transcript of Case study: electromagnetic radiation...
Dean P. Neikirk © 2001, last update March 30, 2001 1 Dept. of ECE, Univ. of Texas at Austin
Case study: electromagnetic radiation detection
• many modes of detection– rectification
• requires non-linear element (diode)• usually restricted to “low” frequencies
– photon detection• generation of electron/hole pairs• usually restricted to “short” wavelengths
– thermal detection• convert incoming E&M energy into heat• can be fairly independent of frequency/wavelength
Dean P. Neikirk © 2001, last update March 30, 2001 2 Dept. of ECE, Univ. of Texas at Austin
300 Å
radio microwavefar infrared
IRvisible
Frequency
Wavelength
sub-millimeter
1m 10 cm 1 cm 1 mm 100 µm 10 µm 3000 Å
100 MHz 1 GHz 10 GHz 100 GHz 1 THz 10 THz 100 THz 1015 Hz
millimeter
1016 Hz
Spectral range of common electromagnetic detectors
Liquid HeliumCooled
Photoconductors
Schottky Diode
Antenna-Coupled Microbolometer Photo Diode
LiquidNitrogenCooled
CCD
Conventional Bolometer
Dean P. Neikirk © 2001, last update March 30, 2001 3 Dept. of ECE, Univ. of Texas at Austin
Basic thermal radiation detector: bolometer
• convert incident electromagnetic radiation to heat, heat capacity C, attached to a heat sink at temp. Ts via thermal resistance R.
• thermal time constant τ ≈C•R• At “dc” Tbolo = Ts + Pbolo •R• temperature change ⇒ resistance change ⇒ voltage
difference via constant applied bias current• to maximize responsivity ⇒ maximize thermal resistance
Rth between bolometer and heat sink
Dean P. Neikirk © 2001, last update March 30, 2001 4 Dept. of ECE, Univ. of Texas at Austin
Output voltage and responsivity
thermalabsorbedbb ZPdTdR
ITdTdR
IV ⋅⋅⋅=∆⋅= ⋅
η⋅⋅⋅=⋅⋅==ℜdTdRZI
PP
dPdT
dTdRI
PV
thermalbincident
absorbedb
incident
η⋅⋅α⋅⋅=ℜ bolothermalb RZI
=η coupling efficiencydTdR
R1 bolo
bolo
=α
• depends on bias current, temperature coefficient of resistance α,bolometer resistance Rbolo, device thermal impedance Zthermal, and coupling efficiency η
Dean P. Neikirk © 2001, last update March 30, 2001 5 Dept. of ECE, Univ. of Texas at Austin
Figure of merit: Noise Equivalent Power
• figure of merit is really the minimum detectable input power– depends on both responsivity and noise
• for simple resistive bolometer the lowest noise level is Johnson noise– noise voltage depends of bandwidth
• units for NEP are normally volts/root hertz
– increased responsivity R tends to improve (decrease) NEP
BWRTk4V Bjohnson ⋅⋅⋅=
ℜ= noiseV
NEP
Dean P. Neikirk © 2001, last update March 30, 2001 6 Dept. of ECE, Univ. of Texas at Austin
Detector Size Relative to Wavelength
• critical in determining "coupling efficiency"• much larger than wavelength
– "classical" absorber– detector is its own "antenna"– typical figure-of-merit: specific detectivity (D*)
• typical units cm•(root Hz) / watt
• much smaller than wavelength: "micro-detectors"– very poor coupling– requires "antenna" structure– typical figure-of-merit: Noise Equivalent Power (NEP)
• D* ˜ (effective area)0.5 / NEP
Dean P. Neikirk © 2001, last update March 30, 2001 7 Dept. of ECE, Univ. of Texas at Austin
Single versus Multi- Mode Antennas
• single mode: use for point sources– one antenna, one detector– absorbed power: P = kT– effective area ˜ λ2
• multimode: use for distributed sources– n-element antenna "array," n detectors– P = nkT– effective area ˜ n λ2
– NEParray = v n NEPsingle
• regardless, D* ˜ λ2 / NEPsingle
Dean P. Neikirk © 2001, last update March 30, 2001 8 Dept. of ECE, Univ. of Texas at Austin
Optimum "Resistive" Absorbers: large detectors
•area >> λ2
•"multimode" system•impedance matched sheet
-absorption must be VERY strong -requires both resistive sheet and perfect mirror
•100 % absorption in resistive sheet at design wavelength
mirror
λ/4
resistive sheetRS = 377 ? /square
incidentradiation
reflectance = 0
Dean P. Neikirk © 2001, last update March 30, 2001 9 Dept. of ECE, Univ. of Texas at Austin
Optimum "Resistive" Loads: small detectors
• "detector" area << λ2
- behaves like classical "lumped" circuit element: resistor
• "absorption" requires an antenna
• efficiency depends on
- antenna gain and beam pattern
- detector/antenna impedance match
Dean P. Neikirk © 2001, last update March 30, 2001 10 Dept. of ECE, Univ. of Texas at Austin
Quasi-Optical Detection System
objective lens
substrate lens
antenna
substrate
˜ λ
detector
Dean P. Neikirk © 2001, last update March 30, 2001 11 Dept. of ECE, Univ. of Texas at Austin
Trade-offs: Small versus Large Thermal Detectors
• figure of merit improves as responsivity increases– NEP = Noise Voltage / Responsivity
• responsivity r (Volts/Watt)
depends on:– bias current– "thermometer" sensitivity: resistance change / temperature
rise– thermal impedance: temperature rise / power in
• how do these quantities scale with size?
r = Ibias ⋅d Rd T
⋅d Td P
Dean P. Neikirk © 2001, last update March 30, 2001 12 Dept. of ECE, Univ. of Texas at Austin
Optimizing responsivity: alpha (temp. coefficient of resistance: TCR)
• highest TCR from superconductors at their transition temperature– requires cooling system;– may present impedance matching (E&M absorption) problem due
to low sheet R • alpha of several bolometer materials
– Y-Ba-Cu-O (YBCO) @ Tc : ~ +0.5 - 1 K-1
– Vanadium Oxide (VOx): ~ -0.015 - 0.028 K-1 (room temp.)– Semiconducting YBCO: ~ -0.0299 - 0.0337 K-1 (room temp.) – Ag: ~ +0.0037 K-1 , Ni: ~ +0.005 K-1 , Au: ~ +0.0036 K-1
– Bismuth (Bi): ~ -0.003 K-1 (Room temp.)
Dean P. Neikirk © 2001, last update March 30, 2001 13 Dept. of ECE, Univ. of Texas at Austin
929088868482800
50
100
150
0
20
40
60
80
100ResistancedR/dT (AVE)
Temperature (K)
Det
ecto
r R
esis
tanc
e (ž
)
dR/d
T (
ž/K
)
Optimizing alpha: Superconducting Detector Strip
• temperature coefficient of resistance α:
– intrinsic semiconductors
– extrinsic semiconductors and metals: α ˜ 1/T
– transition edge materials
• YBCO• VO2
High TC YBCO Strip
Dean P. Neikirk © 2001, last update March 30, 2001 14 Dept. of ECE, Univ. of Texas at Austin
Optimizing responsivity: thermal impedance
• Limiting mechanisms: heat flow from small bodies– radiant: negligible ( W = σT4 W: total radiated power per unit area)– conduction: dominant for micron size objects– convection: can be significant for > tens of microns but negligible
at low pressure• Thermal conductivity of various material
– Silicon (bulk): ~ 1.3 W/cm-K– Si3N4: ~ 0.032 - 0.0385 W/cm-K– Bi (thin film): ~0.018 W/cm-K– SiO2: ~ 0.014 W/cm-K– Bi (bulk): ~ 0.0792 W/cm-K
Dean P. Neikirk © 2001, last update March 30, 2001 15 Dept. of ECE, Univ. of Texas at Austin
Thermal properties
• thermal diffusion equation
tC2
∂φ∂
κ⋅ρ=φ∇
– φis temperature change relative to ambient– ρ is material density– C is specific heat (units: energy · mass-1 · kelvin-1)– κis thermal conductivity (units: power · distance-1 · kelvin-1)
• if φwas a voltage this looks a lot like an electrical problem…
• or if φwas a concentration this looks a lot like a diffusion problem for constant thermal diffusivity D = κ/ρC
φ∇⋅ρκ=
∂φ∂ 2
Ct
Dean P. Neikirk © 2001, last update March 30, 2001 16 Dept. of ECE, Univ. of Texas at Austin
1-d “sinusoidal” heat flow
• let’s assume the power flowing from the “source” end of the bar is sinusoidally varying in time
x
z
ypower flow
tw
Tamb + φo Tamb
l
( ) ztjo ez ⋅γ−⋅ω⋅φ=φ
( ) ztjo
ztjo
2 ejCe ⋅γ−⋅ω⋅⋅γ−⋅ω⋅ φ⋅ω⋅κ⋅ρ=φ⋅γ⇒
κ⋅ρ⋅ω⋅=γ⇒ Cj
tC
z2
2
∂φ∂
κ⋅ρ=
∂φ∂
{ tC
2
2
2
z
d1 ∂φ∂
κ⋅ρ=φ∇
∂φ∂=
−
⋅
κ⋅ρ⋅ω⋅⋅
⋅ρ⋅κ⋅ω⋅⋅
⋅=⋅=φ l
Cjtanh
Cj1
wtP
ZP thermalend
Dean P. Neikirk © 2001, last update March 30, 2001 17 Dept. of ECE, Univ. of Texas at Austin
What does this really mean?
• consider “low frequency”: length of bar much less than thermal diffusion length
1Cj <<⋅
κ⋅ρ⋅ω⋅==⋅γ l
Ll
lthermal
ll ⋅κ
⋅ρ⋅ω⋅⋅⋅ρ⋅κ⋅ω⋅
⋅⋅
≈
⋅
κ⋅ρ⋅ω⋅⋅
⋅ρ⋅κ⋅ω⋅⋅
⋅=φ
≈
<<
CjCj
1wt
PCjtanhCj
1wt
P
umentarg
1
end
444 3444 214434421
{ κ⋅
⋅≈φ l
densitypower
"freqlow"
end wtP
• looks just like the resistance of a bar with “conductivity” κ
Dean P. Neikirk © 2001, last update March 30, 2001 18 Dept. of ECE, Univ. of Texas at Austin
Low frequency thermal response
• the actual input power looks like– Pin = Pave •[ 1 + cos(ωt) ]
• at “low frequency” the temperature “follows” (in phase) the input power
– temperature rises as power increases– temperature falls as power decreases
Pwt
Tκ⋅⋅
=∆ l
time
pow
er
Pave
tem
pera
ture
Tambient
Tmax
Dean P. Neikirk © 2001, last update March 30, 2001 19 Dept. of ECE, Univ. of Texas at Austin
High frequency thermal response
• consider “high frequency”: length of bar much longer than thermal diffusion length
1Cj >>⋅κ
⋅ρ⋅ω⋅==⋅γ lL
llthermal
{( )j1
C21
wtPL
wtP thermal
densitypower
"freqhigh"
end−⋅
ωκρ⋅
⋅=
κ⋅
⋅≈φ
44 344 21444 3444 214434421
κ≈
>>
⋅ρ⋅κ⋅ω⋅⋅
⋅≈
⋅
κ⋅ρ⋅ω⋅⋅
⋅ρ⋅κ⋅ω⋅⋅
⋅=φ
thermalL
1
1
end Cj1
wtPCjtanh
Cj1
wtP l
Dean P. Neikirk © 2001, last update March 30, 2001 20 Dept. of ECE, Univ. of Texas at Austin
High frequency thermal response
• the temperature “lags” the power by 45°• as the frequency increases the magnitude of the
temperature change decreases as √ω
time
pow
er
Pdc
tem
pera
ture
Tambient
Tave
{( )j1
C21
wtPL
wtP thermal
densitypower
"freqhigh"
end−⋅
ωκρ⋅
⋅=
κ⋅
⋅≈φ
Dean P. Neikirk © 2001, last update March 30, 2001 21 Dept. of ECE, Univ. of Texas at Austin
Superconducting transition edgebolometer
• Richards, early 60’s– few square millimeter “black” coated glass slide
• “bismuth black” used as broad-band absorber– suspended by silk threads in a cooled vacuum chamber– small superconducting “resistor” attached
• threads coated with thin metal to allow electrical access
• produced very high thermal resistance, moderate thermal capacitance
• how does this scale with size?– need to look at thermal model and various heat transport
mechanisms for small objects
Dean P. Neikirk © 2001, last update March 30, 2001 22 Dept. of ECE, Univ. of Texas at Austin
Heat Conduction into Substrate
• approximate the detector-substrate contact area as a hemisphere
• use thermal spreading resistance
– Rthermal scales as (area)-1/2
– cut-off frequency scales as 1/area
• for device on SiO2– 10µm x 10µm
• Rthermal = 2.5 x 104 K/W• cut-off f = 8.5 kHz
– 1µm x 1µm• Rthermal = 2.5 x 105 K/W• cut-off f = 850 kHz
• for “large” devices clearly need to remove the substrate to increase R!
a
l
w a ≈ l⋅w2 π
( )thermal
o
Lrr
tjo
o err
−−⋅ω⋅φ=φ
CjL thermal ⋅ρ⋅ω⋅
κ=
Pr2
1T
cetanresisspreadingthermal
o43421 ⋅κ⋅π=∆
area21
Rsubstrate
thermalsubstrate ⋅κ⋅π
≈
areaCf
substrate
substrateoffcutsubstrate ⋅⋅ρ
κ≈−
Dean P. Neikirk © 2001, last update March 30, 2001 23 Dept. of ECE, Univ. of Texas at Austin
Thermal losses: convection to air
• simple spherical model for convective cooling• hc: surface coefficient of heat transfer
– units: W / area•Kelvin
• κ: thermal conductivity• d: diameter• Nu: Nusselt number• for free convection from spheres Nu is approximately
• Gr: Grashof number• Pr: Prandtl number
dNuh c
κ⋅=
( ) 41PrGr45.02Nu ⋅⋅+=
Dean P. Neikirk © 2001, last update March 30, 2001 24 Dept. of ECE, Univ. of Texas at Austin
Thermal losses: Convection
• Gr: Grashof number
– g is acceleration due to gravity, ρ is density, µ is viscosity, βis the coef. of volume expansion, L is length of object
– for air near room temperature
• Pr: Prandtl number
• so as long as L < cm the Nusselt # is
( ) 2PrGr45.02Nu 41 ≈⋅⋅+=
2
Lµ
⋅∆⋅β⋅⋅ρ=32 Tg
Gr
( )temproomnearair72.0C
Pr ≈κµ⋅=
( ) ( )KelvininTmicronsinLL10x4.3Gr 317 ∆⋅⋅≈ −
Dean P. Neikirk © 2001, last update March 30, 2001 25 Dept. of ECE, Univ. of Texas at Austin
Thermal losses: Convection
• then hc, surface coefficient of heat transfer– units: W / area•Kelvin
• thermal conduction from sphere is just (area •hc)
• or
– looks just like thermal spreading resistance term!
d2h c
κ⋅≈
d8hd4G airc2
c ⋅κ⋅π=⋅⋅π≈
area81
Rair
airtoconvect ⋅κ⋅π≈
area21
Rsubstrate
thermalsubstrate ⋅κ⋅π
≈
Dean P. Neikirk © 2001, last update March 30, 2001 26 Dept. of ECE, Univ. of Texas at Austin
• use a simple bar model
– assume • absorbed power is uniformly dissipated throughout bolometer,
sinusoidally varying in time
• ends of detector are connected to perfect heat sinks
• thermal diffusion equation is then
What about heat sinking through the ends?
( )tj
bol
o
bol
bolbol2
22 e1
wtL2P
tC
xω+
κ⋅⋅⋅⋅−
∂φ∂
κ⋅ρ=
∂φ∂=φ∇
bolometer
electrical lead: heat sink!thicknesswidth
x
Length
0
( )tjoinput e1
2P
P ω+=
Dean P. Neikirk © 2001, last update March 30, 2001 27 Dept. of ECE, Univ. of Texas at Austin
• since is symmetric, solve with x = 0 at center– assume
• absorbed power is uniformly dissipated throughout bolometer, sinusoidally varying in time
• ends of detector are connected to perfect heat sinks
• if TCR is constant, then really want integral averaged temperature over the whole length of the bolometer
What about heat sinking through the ends?
bolometer
electrical lead: heat sink!thicknesswidth
x
Length
0
( ) [ ][ ]
−
⋅γ⋅γ⋅⋅
ω⋅⋅ρκ⋅+−⋅
κ⋅⋅⋅⋅=φ ω 1
2Lcoshxcoshe
Cj
2x
8L
wtL2P
x tj
bolbol
bol22
bol
o
κ⋅ρ⋅ω⋅≡γ Cj
Dean P. Neikirk © 2001, last update March 30, 2001 28 Dept. of ECE, Univ. of Texas at Austin
• ratio of time varying temp to time varying input power is the thermal impedance
• integral averaged temperature over the length
Heat sinking through the ends
bolometer
electrical lead: heat sink!thicknesswidth
x
Length
0
( ) ( )
⋅γ−⋅γ⋅⋅
⋅γ⋅+⋅
κ⋅⋅⋅=φ=φ ω
+
−∫ 2
Ltanh
2L
eL1
241wt24P
dxxL1 tj
3bol
o2L
2L
κ⋅ρ⋅ω⋅≡γ Cj
( ) ( )
⋅γ−⋅γ
⋅γ⋅
κ⋅⋅=ω
2Ltanh
2L
L1
wtL2Z 3
bolthermal
Dean P. Neikirk © 2001, last update March 30, 2001 29 Dept. of ECE, Univ. of Texas at Austin
• low frequency: |γ| •L << 1
• high frequency: |γ| •L >> 1
• same as parallel R-C circuit with
Heat sinking through the endsbolometer
electrical lead: heat sink!thicknesswidth
x
Length
0
κ⋅ρ⋅ω⋅≡γ Cj
{ }tj
bol
2o
1L
e1wtL24LP ω
<<⋅γ
+⋅κ⋅⋅⋅⋅
⋅≈φ 321
⋅
ω⋅⋅ρκ⋅−⋅
κ⋅⋅⋅⋅⋅≈φ ω
>>⋅γ
tj
bolbol
bol2
bol
o
1L
eC
j12L
wtL2P
321
bol1L
thermal wt12L
PZ
κ⋅⋅⋅≈φ=<<⋅γ321
ω⋅⋅
⋅⋅⋅⋅ρ≈>>⋅γ
j1
CLwt1Z
bol1L
thermal 321
bolthermal wt12
LRκ⋅⋅⋅
= bolthermal CLwtC ⋅⋅⋅⋅ρ= 2bol
bolthermthermoffcut LC
12CR1
⋅⋅ρκ⋅=≈ω −
Dean P. Neikirk © 2001, last update March 30, 2001 30 Dept. of ECE, Univ. of Texas at Austin
•assume heat sink at ends of bolometer
- thermal resistance directly proportional to electrical resistance, regardless of material!- for high thermal R want length to cross-sectional area ratio to be large- for high speed want L small!
•for silicon bar:
Small bolometers: heat conduction directly through device leads
10µm x 10µm x 0.5µmRthermal = 103 K/Wcut-off f = 1.6 MHz
1µm x 1µm x 0.5µmRthermal = 103 K/W
?? cut-off f = 160 MHz
bolthermal CLwtC ⋅⋅⋅⋅ρ=
bolomelectrical
bolthermal R
121
wt12LR ⋅
κσ⋅=
κ⋅⋅⋅=
bolometer
heat sinkthicknesswidth Length
2bol
boloffcut LC
12⋅⋅ρ
κ⋅≈ω −
Dean P. Neikirk © 2001, last update March 30, 2001 31 Dept. of ECE, Univ. of Texas at Austin
Actual simulation results for full three-D
geometry: Isothermal Profile into Substrate
•Bismuth Load•Quartz Substrate•Operating Conditions
-Input power = 4 µW-Length = 4 µm-Width = 2 µm
3210-5.0
-4.0
-3.0
-2.0
-1.0
0.0
T= .01
T= .02
T= .05
T= .1
T= .15
T= .3
T= .4
X - Direction from the centerof the detector (microns)
Y-D
irec
tion
from
the
subs
trat
e su
rfac
e (m
icro
ns)
heater element antenna lead
substrate
Dean P. Neikirk © 2001, last update March 30, 2001 32 Dept. of ECE, Univ. of Texas at Austin
Fringe Heat Conduction into Leads
Antenna LeadDetector/heater
Substrate
heat flow into the antenna from the substrate
Z
ZZ
sub
D-sub
DTDetector
1Z net
= 1ZD
+ 1Z sub
+ 1Z D − sub
• thermal resistance and speed increase as
bolometer shrinks
Dean P. Neikirk © 2001, last update March 30, 2001 33 Dept. of ECE, Univ. of Texas at Austin
Thermal impedance summary
• convection to air above bolometer
• conduction to “substrate”– substrate under bolometer
– air above bolometer
• conduction out “arms” of bolometer
• total from parallel combination (smallest R dominates)
• clearly Rthermal gets bigger as device area gets smaller• speed gets higher as volume of device shrinks
bol
thermalarms area12
LRκ⋅⋅
=
area21R
substrate
thermalsubstrate ⋅κ⋅π
≈
area81R
air
thermalconvect ⋅κ⋅π
≈
area21
Rair
thermalair ⋅κ⋅π
≈
1
R1R thermal
ithermal
−
= ∑
Dean P. Neikirk © 2001, last update March 30, 2001 34 Dept. of ECE, Univ. of Texas at Austin
“Micro” bolometers
• originally this term was coined by Hwang and Schwarz, and referred to bolometers that were smaller than a wavelength– T.-L. Hwang, S. E. Schwarz, and D. B. Rutledge, “Microbolometers
for infrared detection,” Applied Physics Letters, vol. 34, pp. 773, 1979.
– required antenna for coupling• to increase thermal resistance need to decrease cross
sectional area, and increase length, of “thermal link”• can use micromachining to form "free standing" films to
“remove” the substrate– bulk or sacrificial layer processes can be used
Dean P. Neikirk © 2001, last update March 30, 2001 35 Dept. of ECE, Univ. of Texas at Austin
Basic classes of antenna coupledmicrobolometers
• main classification set by dominant heat transport mechanism
– “substrate supported”: main heat loss to substrate under bolometer– “air-bridge”: main heat loss out ends of bolometer
bol
thermalarms area12
LRκ⋅⋅
=area2
1Rsubstrate
thermalsubstrate ⋅κ⋅π
≈
thermalarms
thermalsubstrate RR <
bol
substrate
areaL
1221
κκ⋅⋅π<
thermalarms
thermalsubstrate RR >
1areaL
122
bol
substrate <κ
κ⋅⋅π
Dean P. Neikirk © 2001, last update March 30, 2001 36 Dept. of ECE, Univ. of Texas at Austin
“Air-bridge” bolometers
• need
– easiest way to get this is to have κsub << κbol
– use micromachining to form "free standing" films to “remove” the substrate
• bulk or sacrificial layer processes can be used
• thermal resistance is then
– increase by decreasing cross sectional area, increasing length
bol
thermalarms area12
LRκ⋅⋅
=
1areaL
122
bol
substrate <κ
κ⋅⋅π
Dean P. Neikirk © 2001, last update March 30, 2001 37 Dept. of ECE, Univ. of Texas at Austin
Optimizing Responsivity
• clearly want max Zthermal, but improvement is not unbounded
• bias current limitationsfrom I2R heating: Imax
2•R = Pmax = (Tmax-Tamb)/Zthermal
– current density (electromigration/critical currents)
– detector Instability (problem only if α is positive)
r ≈ Tmax − Tambient ⋅ Rbol ⋅αbol ⋅ Rbolthermal
r ≈Jmax ⋅αbol ⋅l⋅Rbol
thermal
σbolr ≈ Jmax ⋅αbol
2 π⋅σbol ⋅κsub⋅ l
w
r ⋅Ibias < 1 ⇒ Ibias( )2 ⋅dRdT
< 1Rbol
thermal r < Rbol ⋅αbol ⋅Rbolthermal
substrate-dominated
alpha positive
η⋅⋅⋅=⋅⋅==ℜdTdRZI
PP
dPdT
dTdRI
PV
thermalbincident
absorbedb
incident
Dean P. Neikirk © 2001, last update March 30, 2001 38 Dept. of ECE, Univ. of Texas at Austin
Tri-layer sacrificial resist process
• use photoresist as sacrificial layer– tri-layer resist, middle layer soluble, top and bottom patterned
• lift-off patterning of materials
bolometer evaporation
bolometer evaporation
electrical lead evaporation
bottom sacrificial layer
electrical lead material
Photoresist Air-bridge
electrical lead evaporation
Dean P. Neikirk © 2001, last update March 30, 2001 39 Dept. of ECE, Univ. of Texas at Austin
Air-bridge Lithographic Technique
Dean P. Neikirk © 2001, last update March 30, 2001 40 Dept. of ECE, Univ. of Texas at Austin
Behavior in IR (~1 - ~10 µm)
• antenna-coupled microbolometer– requires sub-micron size for lumped model to apply
• conductor losses in antenna– use "non-resonant" design: simple bow-tie– achieve wavelength selectivity using “back short” (integral
micromachined mirror)• substrate absorption impacts efficiency
– micromachining removes substrate from antenna
Dean P. Neikirk © 2001, last update March 30, 2001 41 Dept. of ECE, Univ. of Texas at Austin
Quasi-Optical Detection System
objective lens
antenna
micromachined “tuning mirror”
≈λ
detector
Dean P. Neikirk © 2001, last update March 30, 2001 42 Dept. of ECE, Univ. of Texas at Austin
Impedance Matching: Composite Microbolometers
detector antenna lead
detector lead
load
substrate
•size and conductivity
constraint: Rbol ˜ 50-100?
-cannot use very low or
very high conductivity
materials
•separate function of load
and temperature sensor
-matched "heater"
-high alpha "detector"
Dean P. Neikirk © 2001, last update March 30, 2001 43 Dept. of ECE, Univ. of Texas at Austin
Composite Microbolometer
Signal Line
Antenna Detector
Heater
Dean P. Neikirk © 2001, last update March 30, 2001 44 Dept. of ECE, Univ. of Texas at Austin
Layout of Simulated Devices
Gold Antenna Leads
YBCO detector signal line
2.5 µm
X
Z
Y
NiCr Heater 2.5 µm x 4 µm
smallest region of symmetry
substrate
1500 Å Nichrome Heater
800 Å Sputtered SiO2
2500 Å YBCO
2300 Å Evaporated Gold
Y
Z
X
Dean P. Neikirk © 2001, last update March 30, 2001 45 Dept. of ECE, Univ. of Texas at Austin
Temperature profiles for various substrates
1.00.50.0-0.5-1.00.00
0.05
0.10
0.15
0.20
Ksub = 0.014 w/cm/K •
Ksub = 10 w/cm/K
Ksub = 0.05 w/cm/KKsub = 0.15 w/cm/K
Distance Normal to Substrate (microns)
Tem
pera
ture
abo
ve L
ocal
Am
bien
t (K
)
substrate YBCO NiCr
SiO2
freespace
SI or sapphire, 80 K
SiO2
•YBCO substrates:
MgO, YSZ, LaAlO3
•with buffer layer: Si,
sapphire
-most have either
high κor high εr
Pin = 4 µW
Dean P. Neikirk © 2001, last update March 30, 2001 46 Dept. of ECE, Univ. of Texas at Austin
YBCO Composite Microbolometer
•YBCO films
-grown on 400 Å MgO buffer
-sapphire substrate
-courtesy: Alex deLozanne,
UT-Austin
antenna
YBCO
microbolometer
Dean P. Neikirk © 2001, last update March 30, 2001 47 Dept. of ECE, Univ. of Texas at Austin
Responsivity Measurement
1.0e-58.0e-66.0e-64.0e-62.0e-60.0e+00.1022
0.1023
0.1024
0.1025
0.1026
0.1027
y = 0.10226 + 40.017x
•bolometers are easily
calibrated from dc I-V
-slope of resistance
versus power curve
-plot V/I versus I*V
•for composite bolometer plot
detector voltage versus power
in heater
-YBCO / heater : 40 V/Watt
Dean P. Neikirk © 2001, last update March 30, 2001 48 Dept. of ECE, Univ. of Texas at Austin
Noise Voltage
10000010000100010010110 -10
10 -9
10 -8
10 -7
10 -6
10 -5
T = 85.7
T = 86.25
T = 85.05
Modulation Frequency (Hz)
Vn/
(Hz)
^1/2 1/root(f)
Johnson Noise Limit
Dean P. Neikirk © 2001, last update March 30, 2001 49 Dept. of ECE, Univ. of Texas at Austin
Noise Equivalent Power (NEP)
10 510 410 310 210 110 010 -12
10 -11
10 -10
10 -9
10 -8
10 -7
Modulation Frequency (Hz)
NE
P (w
atts
/roo
t Hz)
Johnson Noise Limit
1/root(f)
•YBCO composite
structure
-˜ 3 x 10-11 W/vHz
•conventional Bi
microbolometer
-˜ 10-10 W/vHz at
room temperature
Dean P. Neikirk © 2001, last update March 30, 2001 50 Dept. of ECE, Univ. of Texas at Austin
“Micro” bolometers
• originally this term referred to bolometers that were smaller than a wavelength– required antenna for coupling
• but even for “large area” devices, to increase thermal resistance, need to decrease cross sectional area, and increase length, of “thermal link”– if too large will be slow
• use micromachining to form "free standing" films to “remove” the substrate– bulk or sacrificial layer processes can be used
Dean P. Neikirk © 2001, last update March 30, 2001 51 Dept. of ECE, Univ. of Texas at Austin
“classical” micromachined bolometers
• device area >> λ2
• if use CMOS compatible processing can combine with other electronics– J.-S. Shie and P. K. Weng, “Fabrication of micro-bolometer on
silicon substrate by anisotropic etching technique,” presented at Transducers '91 1991 International Conference on Solid-State Sensors and Actuators, San Francisco, CA, 1991.
– R. A. Wood, “High-Performance Infrared Thermal Imaging with Monolithic Silicon Focal Planes Operating at Room Temperature,” presented at IEEE International Electron Devices Meeting, Washington, DC, 1993.
• these devices are now widely referred to as “microbolometers”
Dean P. Neikirk © 2001, last update March 30, 2001 52 Dept. of ECE, Univ. of Texas at Austin
Simple model of “large” microbolometer
• assume absorbing region absorbs power uniformly over its entire area
bolometer, area Abol, thickness tbol
support arm(s), length Larm, width warm,
thickness tbol
heat sink(s)
armarmarm
armarmthermal wt
LR
κ⋅⋅≈
bolbolbolbolbolthermal CAtC ⋅ρ⋅⋅≈
Dean P. Neikirk © 2001, last update March 30, 2001 53 Dept. of ECE, Univ. of Texas at Austin
Equivalent thermal circuit
• bolometer is REALLY “distributed”
• but if assume:– power uniformly deposited
across bolometer– thermal resistance of arm is
much larger than that of bolometer
Ø bolometer area is approximately isothermal
Dean P. Neikirk © 2001, last update March 30, 2001 54 Dept. of ECE, Univ. of Texas at Austin
Simplified thermal equivalent circuit
• surface of bolometer is approximately isothermal– mainly contributes thermal capacitance
• arm thermal resistance is large
Pinheat sink
armarmarm
armarmthermal wt
LR
κ⋅⋅≈
bolbol
bolbol
bolthermal
CAt
C
⋅ρ⋅⋅⋅
≈
1Cj
R1Z bol
thermalarmthermal
thermal
−
⋅ω⋅+≈
( )( )2bol
thermalarmthermal
bolthermal
armthermal
armthermal
CR1
CRj1R
⋅⋅ω+⋅⋅ω⋅−⋅=
• conclusions:– make arm length-to-cross section ratio LARGE– make thickness of bolometer small (area already set by wavelength)
Dean P. Neikirk © 2001, last update March 30, 2001 55 Dept. of ECE, Univ. of Texas at Austin
Free Carrier Absorption in Silicon
1
10
100
1 10
Abs
orpt
ion
Leng
th (µ
m)
Wavelength (µm)
Bolometer absorber material choices
• metals– skin depth ˜ 10nm @ λo = 5
µm– impedance matching
requires very thin sheets• tens of Å ˜ few ohms/square
• semiconductors– below-gap absorption weak
• free carrier absorption
– absorption constant α ≈n •λ2
• would appear very thick layers necessary for efficient absorption
– heavy doping also required
n-type silicon, 1020 cm-3
Dean P. Neikirk © 2001, last update March 30, 2001 56 Dept. of ECE, Univ. of Texas at Austin
Free carrier absorption in thin silicon films
• transfer matrix method used for calculations
– easily handles multiple layers, complex index of refraction
• equivalent to microwave network ABCD matrices
t = 0.7 µm
t = 0.36 µm
air
silicon
airt
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
2 3 4 5 6 7 8 9 10
Abs
orpt
ion
Wavelength (µm)
Dean P. Neikirk © 2001, last update March 30, 2001 57 Dept. of ECE, Univ. of Texas at Austin
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
2 3 4 5 6 7 8 9 10
Nor
mal
ized
Pow
er
Wavelength (µm)
silicon
incident
t = 0.7 µm
transmittance
reflectancereflectancet = λ/ 2
t = 3λ/ 4
t = λ
Interference effects in moderately absorbing films
• t = λ/2 ⇒ half wave window
• t = 3λ/4 ⇒impedance inverter
• t = λ⇒ full wave window
absorbance
transmittance
Dean P. Neikirk © 2001, last update March 30, 2001 58 Dept. of ECE, Univ. of Texas at Austin
Impedance Matching using "Backshorts"
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
2 3 4 5 6 7 8 9 10
Abs
orpt
ion
Wavelength (µm)
•0.36 µm thick silicon sheet
-Rsdc ˜ 83 ? /square
•silver mirror place 2.5 µm behind sheet
•at λo = 5 µm:
λ/ 4λ/ 2
Zair ZSi Zair
short open
no mirror
with mirror
Dean P. Neikirk © 2001, last update March 30, 2001 59 Dept. of ECE, Univ. of Texas at Austin
0.36 µm Silicon Film with Mirror
incident
t = 0.36 µm
silicon
Ag mirror
2.5 µm
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
2 3 4 5 6 7 8 9 10
Nor
mal
ized
Pow
er
Wavelength (µm)
Si absorption
Ag absorption
reflectance
Dean P. Neikirk © 2001, last update March 30, 2001 60 Dept. of ECE, Univ. of Texas at Austin
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
2 3 4 5 6 7 8 9 10
Abs
orpt
ion
Wavelength (µm)
30 Å coating
60 Å coating
no coating
Enhanced Absorption using "Resistive" Coating
•0.36 µm Si film, Ag mirror 2.5 µm behind•front surface coated with thin Ag film
-no coating-30 Å ⇒ Rs
dc ˜ 6.7 ? /?-60 Å ⇒ Rs
dc ˜ 3.3 ? /?
λ/ 4λ/ 2
Zair ZSi Zair
thin Ag layer
Dean P. Neikirk © 2001, last update March 30, 2001 61 Dept. of ECE, Univ. of Texas at Austin
Is wavelength selectivity important?
• visible wavelength sensors: object “reflectivity” dominates– reflectivity/absorption are functions of wavelength– produces “color” in the image
• most energy emitted by an object at terrestrial temp. (300K) falls in the 3 - 14 µm range– “mid” to “far” IR sensing: emissivity of object dominates– emissivity is also a function of wavelength
• atmospheric propagation also influences band choice
– strong absorption bands• H2O (2.6, 2.7, 6.3 µm)• CO2 (2.7, 4.3, 15 µm)
– transmission windows: • 3-5 µm (MWIR)• 8-14 µm (LWIR) window
Dean P. Neikirk © 2001, last update March 30, 2001 62 Dept. of ECE, Univ. of Texas at Austin
“Enhanced” absorption inmicrobolometers
incident
thin conductor(absorber)
Mirror layer
• to “match” an absorber to free space requires– absorber: thin conductor with sheet resistance 377 ohms
Gap
( )4
λ⋅odd
– mirror placed [(odd integer)/4]*λbehind absorbing layer– essentially a Fabry-Perot cavity
Thin mirror layer
Thin conductorbolometer layer
ZairZair Zair
incident
Dean P. Neikirk © 2001, last update March 30, 2001 63 Dept. of ECE, Univ. of Texas at Austin
IR wavelength selective structure and transmission line equivalent model
n n-1 12k
Mirror Layer(Rmirror)Bolometer Layer (Rbolo)
Incident IR
Zn Zn-1
Z2 Z1 ZairZair ZairorZk
Dean P. Neikirk © 2001, last update March 30, 2001 64 Dept. of ECE, Univ. of Texas at Austin
Unit: Å �
Simulation of absorption of microbolometer
Si3N4 SiO2 Si3N4 Air Gap Si3N4 SiO2 Si3N4
1_COM2 750 4000 750 10000 750 4000 750
0.3_COM3 400 2800 400 3000 400 2800 400
0.3_COM4 300 2100 300 3000 300 2100 300
Dean P. Neikirk © 2001, last update March 30, 2001 65 Dept. of ECE, Univ. of Texas at Austin
Resolving multi-wavelength ambiguities
• for a given gap, response occurs at all wavelengths where λ= 4*gap/(odd integer)
• four “color” detector array– gap (air equivalent): 2.5 µm -> λ1 = 10 µm
• “ghost” response at 3.3 µm, 2 µm , etc.– gap: 1.5 µm -> λ1 = 6 µm
• “ghost” response at 2 µm & shorter– gap: 0.83 µm -> λ1 = 3.3 µm
• all “ghost” responses below 2 µm
– gap: 0.5 µm -> λ1 = 2 µm• all “ghost” responses below 2 µm
• assume external bandpass from just less than 2 µm to just over 10 µm
Dean P. Neikirk © 2001, last update March 30, 2001 66 Dept. of ECE, Univ. of Texas at Austin
dete
ctor
sig
nal o
utpu
t
gap:
Resolving multi-wavelength ambiguities
• I10µm = Sg=2.5µm - m3.3->10• Sg=.83µm - m2->10• Sg=.5µm• I6µm = Sg=1.5µm - m2->6• Sg=.5µm
2.5µm 1.5µm 0.83µm 0.5µm
I10µm
Sg=2.5µm
I6µm
I3.3µm I2µm
2µm (4g/3) “ghost”
3.3µm (4g/3)
2µm (4g/5)
Dean P. Neikirk © 2001, last update March 30, 2001 67 Dept. of ECE, Univ. of Texas at Austin
fabrication procedure
• STEP 1: formation of bottom membrane
– deposition of bottom membrane layers (Si3N4/LTO/SI3N4 )
• multi-lare stack used to compensate for stress in deposited films
– nitride in tension– oxide in compression
– backside masking & RIE– anisotropic KOH etching
Cross section view
Top view
Dean P. Neikirk © 2001, last update March 30, 2001 68 Dept. of ECE, Univ. of Texas at Austin
• STEP 2: formation of sacrificial layer and top membrane
– deposition of sacrificial
layer (LPCVD polysilicondeposition)
– deposition of top membrane layers (Si3N4/LTO/SI3N4 )
Top view
Dean P. Neikirk © 2001, last update March 30, 2001 69 Dept. of ECE, Univ. of Texas at Austin
• STEP 3: back side removal of poly, removal of sacrificial layer– remove backside
Si3N4/LTO/SI3N4 layers by RIE
– polysilicon by KOH– patterning & RIE to
remove top Si3N4/LTO/SI3N4
– remove sacrificial polysilicon by KOH
Top view
Dean P. Neikirk © 2001, last update March 30, 2001 70 Dept. of ECE, Univ. of Texas at Austin
• STEP 4:“self-aligned”microbolometer structure– deposition of back side
mirror coating layer by evaporator
– deposition temperature sensitive bolometer layer
Top view
Dean P. Neikirk © 2001, last update March 30, 2001 71 Dept. of ECE, Univ. of Texas at Austin
Wavelength selective micromachined IR sensors
Top view Top Membrane
Bottom Membrane
Dean P. Neikirk © 2001, last update March 30, 2001 72 Dept. of ECE, Univ. of Texas at Austin
Fabrication of proposed structure using IC comparable micromachining techniques
Top view Top Membrane
BottomMembrane
Bolometer layer
Silicon Nitride
Silicon oxide
Mirror layer
Silicon
bolometer active area
bottom membrane
top membrane
Dean P. Neikirk © 2001, last update March 30, 2001 73 Dept. of ECE, Univ. of Texas at Austin
Fabrication process for micromachined microbolometer
Si substrate Si3N4
•Deposition of silicon nitride as barrier layer
•Deposition of sacrificial poly and patterning
•Deposition of membrane layers (Si3N4/LTO/SI3N4 )
•Patterning of membrane layers using RIE
• Poly etching and Self-aligned structure using KOH
undercutting etching process
•Deposition of bolometer layer
SiO2Poly Cr
Active area 26 µm x 20 µm, 20 µm leg length, 3 µm leg width
3 um
20 um26 um
20 um
Dean P. Neikirk © 2001, last update March 30, 2001 74 Dept. of ECE, Univ. of Texas at Austin
Self aligned undercutting structure
Dean P. Neikirk © 2001, last update March 30, 2001 75 Dept. of ECE, Univ. of Texas at Austin
Fabrication issues
• legs are still quite compliant, fragile
• mechanical stability issues– tensile nature of Si3N4 + compressive nature of SiO2
– combination of these films ⇒ weak tensile
Dean P. Neikirk © 2001, last update March 30, 2001 76 Dept. of ECE, Univ. of Texas at Austin
Thermal impedance approximation
w
t1t2t3t4
To To
To To
l
Thot
w = 3 µm, l = 20 µm
)111(1
421 tttleg RRRR⋅⋅⋅++=
κ⋅⋅≈
wtl
RtWhere
Rleg ≈4.3 x 106 K/W
4leg
bolometer
RZ ≈
Zbolometer ≈ 1.1 x 106 K/W
• measured thermal impedance under ambient pressure is 6 x 104 K/W vs. calculated thermal impedance 1 x 106 K/W
– simple “still air” estimate gives thermal resistance of about 2x105 K/W– past work supports two-order reduction due to thermal losses to air at
ambient pressure
• Active area 26 µm x 20 µm, 20 µm leg length, 3 µm leg width
Dean P. Neikirk © 2001, last update March 30, 2001 77 Dept. of ECE, Univ. of Texas at Austin
Optical setup used for IR testing of microbolometers
IR HeNe
Visible HeNe
DFGChromex
A1 A2 A3
A4
CWdet A
det B
S
Flip top mount
6 in FL
Sample xyz stage
Microbolometer under Mirror
M
M
M
M
M
Splitter
• set-up in A. J. Welch’s lab, with thanks to Dan Hammer• two device types tested
– resonant cavity bolometer WITH mirror for enhanced wavelength dependent absorption, first resonance at 3 µm
– simple bolometer WITHOUT mirror
Dean P. Neikirk © 2001, last update March 30, 2001 78 Dept. of ECE, Univ. of Texas at Austin
IR spectrum of ultrafast laser system
∫ ∫ ∫ λ⋅λη⋅λ= )y,x,(S)(ddydxPabsorbedbolometer response ∝
fairly wide spectral featuresλo ± ∆λ
Dean P. Neikirk © 2001, last update March 30, 2001 79 Dept. of ECE, Univ. of Texas at Austin
Irradiance on and output voltage of resonant cavity microbolometer
0.00E+00
5.00E+02
1.00E+03
1.50E+03
2.00E+03
2.50E+03
3.00E+03
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Peak wa ve length (micron)
Ave
rage
d
Me a n
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 1 2 3 4 5
Peak Wave length(Micron)
Measured_with
Calcula ted_with
Dean P. Neikirk © 2001, last update March 30, 2001 80 Dept. of ECE, Univ. of Texas at Austin
Responsivity at 1.15 and 3.39 µm,narrow-band excitation
0
5
10
15
20
25
30
35
40
45
0 0.5 1 1.5 2 2.5 3 3.5 4
Wavelength(Micron)
measured_without
measured_with
calculated_without
calculated_with
with mirror
with mirror
without mirror
without mirror
Dean P. Neikirk © 2001, last update March 30, 2001 81 Dept. of ECE, Univ. of Texas at Austin
Measured responsivity of resonant cavitymicrobolometer
0.00
10.00
20.00
30.00
40.00
50.00
60.00
70.00
80.00
0 1 2 3 4 5
Peak Wavelength (micron)
Mean_without
Mean_with
with mirror
without mirrorwith mirror
Dean P. Neikirk © 2001, last update March 30, 2001 82 Dept. of ECE, Univ. of Texas at Austin
Power coupling efficiency
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 1 2 3 4 5 6 7
Wa ve length (micron)
simulation_with
simulation_without
measured_without
measured_with
Dean P. Neikirk © 2001, last update March 30, 2001 83 Dept. of ECE, Univ. of Texas at Austin
• 1/f noise dominant
• NEP (Noise Equivalent Power) at 1500 Hz
– ~ 5 x 10-8 W/(Hz) ½
• Detectivity (D*) at 1500 Hz – 6.9 x 104 cm (Hz) 1/2 /W
Noise measurement of the resonant dielectric cavity microbolometer
1.00E-07
1.00E-06
1.00E-05
1.00E-04
1.00E-0310 100 1000 10000
Modulation Frequency (Hz)
Vn/
Hz^
1/2
Dean P. Neikirk © 2001, last update March 30, 2001 84 Dept. of ECE, Univ. of Texas at Austin
Noise equivalent temperature difference (NETD)
• smallest temperature change in object space that can be resolved assuming a unit signal to noise ratio
• temperature resolution for IR “microbolometer” systems– NETD of uncooled camera: claim that 0.039 C° is enough for “TV
quality”– public results
• SBRC (1996) : 110 mK• Mitubishi Electronics (1996) : 500 mK• Lockheed Martin (1998): less than 100 mK
•
thermalthermal CR1TNETD
η∝ Where T: temperature, η: absorptivity, and
C: thermal capacitance
Dean P. Neikirk © 2001, last update March 30, 2001 85 Dept. of ECE, Univ. of Texas at Austin
Comparison of current commercial IR detectors
Summary of Important Specification of IR dectors
Near Infrared Mid IndiumAntimonide
UncooledMicrobolometer
Microbolometer Long QWIP
Type InGaAs InSb Microbolometer Microbolometer GaAs QWIPSpectral range 0.9 – 1.68 um 1 – 5.5 um 7 –14 um 7 – 14 um 8 – 9.2 umArray format 320 x 240 320 x 240 320 x 240 160 x 128 320 x 240Size 30 x 30 um 30 x 30 um 51 x 51 um 51 x 51 um 30 x 30 umOperatingTemperature
291 KUncooled
77 KCooled
313 K or 293 KUncooled
300 KUncooled
77 K or lessCooled
NedT < 20 mK < 100 mK < 100 mK < 30 mK