Semiconductor Thermal Metrology

Time-Domain Thermoreflectance (TDTR) Nanoscale Thermal Metrology Lab

Q: What is Time-Domain Thermoreflectance (TDTR) and how does it measure thermal conductivity?

Time-Domain Thermoreflectance (TDTR) is an ultrafast optical pump-probe technique that measures the thermal conductivity and interface thermal resistance of thin films, 2D materials, and bulk substrates. A femtosecond laser pulse train (~100 fs, ~80 MHz rep rate) is split into a modulated pump beam and a mechanically delayed probe beam. The pump beam periodically heats an aluminum or gold transducer film (~80 nm thick), while the probe beam measures minute changes in optical reflectance (dR/dT ~ 10^-4 K^-1) as heat diffuses into underlying layers. By fitting the ratio of the in-phase to out-of-phase lock-in signals (-Vin / Vout) over a delay range of 100 ps to 4 ns to an analytical multilayer heat diffusion model, thermal conductivity (k) and thermal boundary conductance (G) are uniquely extracted.

Q: What is thermal boundary conductance (Kapitza conductance, G) and why is it critical in GaN-on-Diamond?

Thermal boundary conductance (G), also known as Kapitza conductance, quantifies the heat flux across a material interface per unit temperature drop (q = G * Delta T). When phonons cross a dissimilar material interface (such as GaN and CVD diamond in RF power amplifiers, or Si and SiO2 in microprocessors), acoustic impedance mismatch and interfacial atomic disorder scatter phonons, creating a localized thermal resistance (R_th = 1 / G). In high-power GaN HEMTs, diamond provides extraordinary bulk thermal conductivity (>1,500 W/mK), but an inferior GaN/diamond interface (G < 30 MW/m^2K) creates a massive thermal barrier that causes excessive channel junction temperatures and accelerated device degradation.

Q: Why does TDTR measure the lock-in ratio (-Vin / Vout) instead of absolute probe intensity?

Measuring the lock-in ratio (-Vin / Vout)—where Vin is the in-phase signal and Vout is the out-of-phase (quadrature) signal—renders the measurement immune to probe beam intensity fluctuations, mechanical vibrations of the optical delay stage, laser pointing drift, and local surface roughness variations. While the individual voltage signals depend directly on laser power and detector gain, their mathematical ratio cancels out common-mode optical intensity noise, leaving a clean, highly reproducible thermal decay curve governed exclusively by intrinsic heat capacity, thermal conductivity, and interfacial conductance.

Q: How do picosecond acoustic strain echos calibrate the metal transducer layer thickness?

When the ultrafast pump pulse heats the aluminum transducer film, rapid thermal expansion launches a longitudinal acoustic strain pulse traveling into the metal at the speed of sound (v_sound ~ 6,420 m/s for Al). The acoustic wave reflects off the transducer-substrate interface and returns to the top surface at delay time t_echo = 2 * d / v_sound (approximately 25 ps for an 80 nm Al film). This acoustic echo causes a localized change in surface reflectance via the photoelastic effect. By measuring the exact time of the echo peak, the transducer thickness d is calibrated in situ with sub-nanometer accuracy, eliminating a major source of systematic error in subsequent thermal fitting.

Q: What are the CapEx and commercial payback economics of an industrial TDTR metrology workstation?

A turnkey industrial-grade TDTR metrology suite—including an ultrafast mode-locked Ti:Sapphire or Yb-doped fiber laser oscillator, 1.2-meter high-precision optical delay stage, high-bandwidth electro-optic modulator (EOM), balanced photodetector, high-frequency lock-in amplifier, motorized 300 mm wafer mapping stage, and automated multilayer thermal fitting software—requires a capital expenditure (CapEx) between $400,000 and $550,000. Annual operational expenses (OpEx) for laser diode pump replacement, chiller maintenance, and cleanroom technician time average $45,000 to $60,000. Conducting 1,000 to 1,500 commercial wafer and chip package testing runs per year at fees of $550 to $800 per sample generates $600,000 to $1,100,000 in annual revenue, achieving full payback in 1.1 to 1.7 years with an IRR exceeding 60%.

Model ultrafast optical pump-probe thermoreflectance, Kapitza boundary conductance (G), picosecond acoustic echos, and 20-year laboratory DCF economics.

Sub-Picosecond Acoustic Resolution

Thermal Material Preset

High-power GaN HEMT on chemical vapor deposited (CVD) diamond. Critical figure of merit is the GaN/diamond interface Kapitza conductance (G ~ 40 MW/m^2K) that governs channel peak junction temperature.

Penetration μ_th

6.25 μm

Thermal diffusion depth

Acoustic Echo t_echo

24.9 ps

Al film roundtrip

Lock-in Ratio

2.45

-V_in / V_out

Kapitza R_th

25.0 nK·m²/W

1 / G boundary barrier

Diffusivity D

10.29 cm²/s

Substrate k / C_v

Peak Temp Rise ΔT

2.8 K

Steady periodic heat

TDTR Parameters

Substrate (k) 1800 W/(m·K)

Interface (G) 40 MW/(m²·K)

Al Transducer (d) 80 nm

Modulation (f_m) 8.5 MHz

Pump 1/e² Radius (w₀) 12.0 µm

Larger spot sizes ensure strictly 1D cross-plane heat diffusion.

Optical Delay Time (t) 0.85 ns

Move delay stage from 10 ps to 3.8 ns to observe acoustic echo and thermal relaxation.

TDTR Ratio Signal (-V_in / V_out) vs Optical Delay

Lock-in Quadrature Decoupling

Multilayer Heat Diffusion & Penetration Depth

Cross-Plane Profile

Industrial TDTR Metrology Suite & Packaging DCF

20-Year discounted cash flow (DCF) model for commercial automated TDTR wafer mapping and thermal testing.

IRR: 66.4% | Payback: 1.4 Yrs

Equipment CapEx $480,000

Ultrafast laser, EOM, delay

Annual OpEx $52,000/yr

Chiller, diodes, engineer

Annual Runs 1,200

Wafer/package tests/yr

Testing Fee $650

Fee per thermal scan

Discount Rate 8%

WACC capital cost

20-Year NPV $5.68M $780k/yr Rev

Year	Revenue	OpEx	EBITDA	Taxes (21%)	Free Cash Flow	Discounted PV	Cumulative Cash

Engineering Guide: Time-Domain Thermoreflectance (TDTR) Metrology

1. Ultrafast Thermoreflectance & Pump-Probe Optics

Time-Domain Thermoreflectance measures thermal properties across sub-micron scales using a mode-locked laser emitting train pulses ($\approx 100\text{ fs}$ duration, repetition rate $f_{\text{rep}} \approx 80\text{ MHz}$). The output is split into two paths:

Pump Beam: Modulated electro-optically at radio frequency $f_m$ ($0.2\text{--}20\text{ MHz}$), creating a steady-periodic heat wave in a thin aluminum film ($\approx 80\text{ nm}$).
Probe Beam: Delayed along a motorized mechanical translation stage ($0\text{ to }4\text{ ns}$). Because optical reflectance depends linearly on temperature ($R(T) \approx R_0 + (dR/dT) \Delta T$), the reflected probe intensity tracks surface temperature decay in real time.

Al is the standard transducer because its thermoreflectance coefficient $dR/dT \approx 2.3 \times 10^{-4}\text{ K}^{-1}$ peaks strongly at $\lambda = 800\text{ nm}$, while exhibiting excellent thermal conductivity ($k \approx 200\text{ W/mK}$).

2. Multilayer Heat Diffusion & Hankel Transform Modeling

Cross-plane and in-plane thermal transport in the sample stack is governed by the Fourier heat conduction equation in cylindrical coordinates:

$C_v \frac{\partial T}{\partial t} = \frac{k_r}{r} \frac{\partial}{\partial r}\left(r \frac{\partial T}{\partial r}\right) + k_z \frac{\partial^2 T}{\partial z^2}$

Using Cahill's analytical model, the periodic temperature response is solved by taking the Hankel transform over radial spatial coordinates and summing over the Fourier series harmonics of the laser repetition rate. The thermal penetration depth is controlled by modulation frequency:

$\mu_{\text{th}} = \sqrt{\frac{k_z}{\pi C_v f_m}}$

3. Lock-in Ratio Detection (-V_in / V_out)

TDTR does not rely on absolute reflected intensity, which is vulnerable to laser flicker noise and optical stage misalignment. Instead, an RF lock-in amplifier detects the in-phase ($V_{\text{in}}$) and out-of-phase ($V_{\text{out}}$) photodiode signals. The ratio:

$\text{Ratio}(t) = -\frac{V_{\text{in}}(t)}{V_{\text{out}}(t)}$

cancels laser power fluctuations, stage wobble, and detector gain drift. Furthermore, $V_{\text{in}}$ is predominantly sensitive to thermal boundary conductance $G$ and near-surface layers at short delay times ($< 1\text{ ns}$), whereas $V_{\text{out}}$ captures steady-state accumulated heating into the bulk substrate, enabling simultaneous extraction of $k$ and $G$.

4. Picosecond Ultrasonics Acoustic Echo Calibration

The absorbed pump pulse triggers instantaneous thermal expansion that launches a coherent longitudinal acoustic strain wave into the metal transducer. Traveling at the speed of sound ($v_L = 6,420\text{ m/s}$ for Al), the acoustic pulse reflects off the interface and returns to the surface at:

$t_{\text{echo}} = \frac{2 d_{\text{Al}}}{v_L}$

For an $80\text{ nm}$ film, $t_{\text{echo}} \approx 25\text{ ps}$. Detecting this acoustic echo calibrates the transducer film thickness $d_{\text{Al}}$ with sub-nanometer precision in situ, eliminating thickness ambiguity in the subsequent thermal conductivity extraction.

Frequently Asked Questions About Time-Domain Thermoreflectance (TDTR)

Time-Domain Thermoreflectance (TDTR) is an ultrafast optical pump-probe technique that measures the thermal conductivity and interface thermal resistance of thin films, 2D materials, and bulk substrates. A femtosecond laser pulse train (~100 fs, ~80 MHz rep rate) is split into a modulated pump beam and a mechanically delayed probe beam. The pump beam periodically heats an aluminum or gold transducer film (~80 nm thick), while the probe beam measures minute changes in optical reflectance (dR/dT ~ 10^-4 K^-1) as heat diffuses into underlying layers. By fitting the ratio of the in-phase to out-of-phase lock-in signals (-Vin / Vout) over a delay range of 100 ps to 4 ns to an analytical multilayer heat diffusion model, thermal conductivity (k) and thermal boundary conductance (G) are uniquely extracted.

Thermal boundary conductance (G), also known as Kapitza conductance, quantifies the heat flux across a material interface per unit temperature drop (q = G * Delta T). When phonons cross a dissimilar material interface (such as GaN and CVD diamond in RF power amplifiers, or Si and SiO2 in microprocessors), acoustic impedance mismatch and interfacial atomic disorder scatter phonons, creating a localized thermal resistance (R_th = 1 / G). In high-power GaN HEMTs, diamond provides extraordinary bulk thermal conductivity (>1,500 W/mK), but an inferior GaN/diamond interface (G < 30 MW/m^2K) creates a massive thermal barrier that causes excessive channel junction temperatures and accelerated device degradation.

Measuring the lock-in ratio (-Vin / Vout)—where Vin is the in-phase signal and Vout is the out-of-phase (quadrature) signal—renders the measurement immune to probe beam intensity fluctuations, mechanical vibrations of the optical delay stage, laser pointing drift, and local surface roughness variations. While the individual voltage signals depend directly on laser power and detector gain, their mathematical ratio cancels out common-mode optical intensity noise, leaving a clean, highly reproducible thermal decay curve governed exclusively by intrinsic heat capacity, thermal conductivity, and interfacial conductance.

When the ultrafast pump pulse heats the aluminum transducer film, rapid thermal expansion launches a longitudinal acoustic strain pulse traveling into the metal at the speed of sound (v_sound ~ 6,420 m/s for Al). The acoustic wave reflects off the transducer-substrate interface and returns to the top surface at delay time t_echo = 2 * d / v_sound (approximately 25 ps for an 80 nm Al film). This acoustic echo causes a localized change in surface reflectance via the photoelastic effect. By measuring the exact time of the echo peak, the transducer thickness d is calibrated in situ with sub-nanometer accuracy, eliminating a major source of systematic error in subsequent thermal fitting.

A turnkey industrial-grade TDTR metrology suite—including an ultrafast mode-locked Ti:Sapphire or Yb-doped fiber laser oscillator, 1.2-meter high-precision optical delay stage, high-bandwidth electro-optic modulator (EOM), balanced photodetector, high-frequency lock-in amplifier, motorized 300 mm wafer mapping stage, and automated multilayer thermal fitting software—requires a capital expenditure (CapEx) between $400,000 and $550,000. Annual operational expenses (OpEx) for laser diode pump replacement, chiller maintenance, and cleanroom technician time average $45,000 to $60,000. Conducting 1,000 to 1,500 commercial wafer and chip package testing runs per year at fees of $550 to $800 per sample generates $600,000 to $1,100,000 in annual revenue, achieving full payback in 1.1 to 1.7 years with an IRR exceeding 60%.

Nanoscale Thermal Metrology Knowledge Check

1. What acoustic feature in the early TDTR delay curve allows in situ calibration of the aluminum transducer thickness?

A thermal runaway spike at 5 nanoseconds

A longitudinal acoustic strain echo reflecting from the film/substrate interface at ~25 picoseconds

A permanent optical blackout caused by laser ablation

2. Why does adjusting the pump modulation frequency (e.g., from 1 MHz to 10 MHz) alter the sensitivity of the TDTR measurement?

Higher frequencies change the color of the probe laser

Modulation frequency has no effect on thermal transport

Thermal penetration depth scales as $\mu_{\text{th}} = \sqrt{k / (\pi C_v f_m)}$, confining heat to the interface at high frequencies

3. Why is Kapitza thermal boundary conductance (G) a critical bottleneck in GaN-on-Diamond RF power devices?

Even with diamond's ultra-high bulk conductivity, phonon scattering across the interface creates a steep temperature jump that elevates junction temperatures

Diamond melts when in contact with gallium nitride

Kapitza conductance causes GaN to lose its electrical bandgap

Time-Domain Thermoreflectance (TDTR) Nanoscale Thermal Metrology Lab

6.25 μm

24.9 ps

2.45

25.0 nK·m²/W

10.29 cm²/s

2.8 K

TDTR Parameters

TDTR Ratio Signal (-V_in / V_out) vs Optical Delay

Multilayer Heat Diffusion & Penetration Depth

Industrial TDTR Metrology Suite & Packaging DCF

Engineering Guide: Time-Domain Thermoreflectance (TDTR) Metrology

1. Ultrafast Thermoreflectance & Pump-Probe Optics

2. Multilayer Heat Diffusion & Hankel Transform Modeling

3. Lock-in Ratio Detection (-V_in / V_out)

4. Picosecond Ultrasonics Acoustic Echo Calibration

Frequently Asked Questions About Time-Domain Thermoreflectance (TDTR)

Nanoscale Thermal Metrology Knowledge Check

Related Advanced Characterization & Metrology Labs

PDS Defect Spectroscopy Lab

MOKE Spintronics Lab

s-SNOM Nanophotonics Lab

SThM Thermal Lab

BLS Metrology Lab

Time-Domain Thermoreflectance (TDTR) Nanoscale Thermal Metrology Lab

6.25 μm

24.9 ps

2.45

25.0 nK·m²/W

10.29 cm²/s

2.8 K

TDTR Parameters

TDTR Ratio Signal (-Vin / Vout) vs Optical Delay

Multilayer Heat Diffusion & Penetration Depth

Industrial TDTR Metrology Suite & Packaging DCF

Engineering Guide: Time-Domain Thermoreflectance (TDTR) Metrology

1. Ultrafast Thermoreflectance & Pump-Probe Optics

2. Multilayer Heat Diffusion & Hankel Transform Modeling

3. Lock-in Ratio Detection (-Vin / Vout)

4. Picosecond Ultrasonics Acoustic Echo Calibration

Frequently Asked Questions About Time-Domain Thermoreflectance (TDTR)

What is Time-Domain Thermoreflectance (TDTR) and how does it measure thermal conductivity?

What is thermal boundary conductance (Kapitza conductance, G) and why is it critical in GaN-on-Diamond?

Why does TDTR measure the lock-in ratio (-Vin / Vout) instead of absolute probe intensity?

How do picosecond acoustic strain echos calibrate the metal transducer layer thickness?

What are the CapEx and commercial payback economics of an industrial TDTR metrology workstation?

Nanoscale Thermal Metrology Knowledge Check

Related Advanced Characterization & Metrology Labs

PDS Defect Spectroscopy Lab

MOKE Spintronics Lab

s-SNOM Nanophotonics Lab

SThM Thermal Lab

BLS Metrology Lab

TDTR Ratio Signal (-V_in / V_out) vs Optical Delay

3. Lock-in Ratio Detection (-V_in / V_out)