🔬 optics

Stable thin-film lithium tantalate modulators operating at high temperature for uncooled operation

This paper demonstrates that thin-film lithium tantalate (TFLT) modulators maintain stable electro-optic performance and DC bias up to 120°C, with an observed reduction in Vπ at higher temperatures, establishing them as a robust candidate for uncooled, high-temperature co-packaged optics.

Original authors: Ayed Al Sayem, Shiekh Zia Uddin, Ting-Chen Hu, Alaric Tate, Mark Cappuzzo, Rose Kopf, Mark Earnshaw

Published 2026-05-01

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Ayed Al Sayem, Shiekh Zia Uddin, Ting-Chen Hu, Alaric Tate, Mark Cappuzzo, Rose Kopf, Mark Earnshaw

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to build a super-fast highway for data (information) to travel between a computer's brain (the processor) and its memory. In the past, this highway was made of copper wires, but they are getting too crowded and slow. Now, scientists are building a new kind of highway using light instead of electricity. This is called "co-packaged optics."

The problem? The computer's brain gets very hot, like a car engine running at full speed. Most light-based devices are like delicate ice sculptures; if you put them near that hot engine, they melt, wobble, or stop working correctly. They need a constant air conditioner (cooling system) to stay stable, which uses up a lot of extra energy.

This paper introduces a new type of "light switch" (called a modulator) made from a special crystal called Thin-Film Lithium Tantalate (TFLT). The researchers wanted to see if this switch could handle the heat of a computer engine without needing an air conditioner.

Here is what they found, using some simple comparisons:

1. The "Heat-Proof" Switch

Think of the old switches (made of Silicon or Lithium Niobate) like a rubber band. When it gets hot, the rubber band stretches and changes shape, making the light signal drift off course. You have to constantly pull it back to the right spot, which takes energy.

The new TFLT switch is like a piece of ceramic or stone. When the researchers heated it up to 120°C (which is hotter than a boiling pot of water and much hotter than a typical computer chip), the switch didn't wobble. It stayed exactly where it was supposed to be.

The Result: They could turn the light on and off (modulate data) perfectly, even in the heat, without needing to constantly adjust it.

2. The "Hotter is Better" Surprise

Usually, when electronics get hot, they get worse. But this switch did something funny: it actually got better when it was hot.

The Analogy: Imagine a door that is usually hard to push open. When you heat it up, the hinges loosen, and suddenly it swings open with less effort.
The Science: The researchers found that the "voltage" needed to flip the switch dropped by about 10% at high temperatures. This means the device became more efficient the hotter it got.

3. The Speed Test

They also checked if the heat slowed down the switch.

The Analogy: Imagine a race car. You might worry that driving it on a hot track would make the engine overheat and slow it down.
The Result: The TFLT switch raced at speeds over 50 GHz (that's billions of switches per second) whether it was at room temperature or boiling hot. The heat didn't slow it down at all.

4. The "Resonant" Ring

The team also tested a smaller, circular version of the switch (a microring resonator). These are like a singing bowl; you hit it at just the right frequency, and it rings loudly.

The Problem: Usually, heat changes the pitch of the bowl, so it stops ringing.
The Result: Even at high temperatures, they could use electricity to tune the ring's "pitch" perfectly. It stayed stable and didn't drift away.

Why This Matters (According to the Paper)

The paper concludes that because these switches are stable in the heat and don't need constant adjustments, we might be able to build computer systems that don't need active cooling for their optical parts.

Think of it like this: Instead of putting a fan on every single light-switch in a data center, we can just let them sit in the hot air and work perfectly. This saves a massive amount of energy and makes the whole system smaller and simpler.

In short: The researchers built a light-switch made of a special crystal that not only survives the heat of a computer but actually works better in it, opening the door for faster, cooler, and more energy-efficient computers.

1. Problem Statement

The rapid scaling of Artificial Intelligence (AI) and data-centric computing has driven the adoption of Co-Packaged Optics (CPO), which integrates optical transceivers directly with compute silicon to minimize electrical trace lengths and improve energy efficiency. However, this tight integration exposes photonic components to elevated local temperatures (often exceeding 100°C) and steep thermal gradients.

Current photonic platforms face significant challenges in this environment:

Silicon (Si) & SiGe: Silicon microring modulators suffer from high thermal sensitivity, leading to resonance shifts that require complex active thermo-optic (TO) bias control. SiGe electro-absorption modulators (EAMs) offer better thermal stability but suffer from high optical loss, limited wavelength ranges, and poor extinction ratios.
Thin-Film Lithium Niobate (TFLN): While TFLN offers high speed and low loss, it is plagued by the photo-refractive (PR) effect, which causes DC bias instability and limits optical power handling. To mitigate this, TFLN systems often rely on TO tuning, but the weak TO effect in LN requires high power (hundreds of mW per modulator) for heaters, which is impractical for dense CPO. Furthermore, the PR effect makes resonant TFLN designs unstable at high powers.

The Gap: There was a lack of experimental evidence demonstrating that Thin-Film Lithium Tantalate (TFLT) modulators could operate stably at the high temperatures required for uncooled CPO systems.

2. Methodology

The authors experimentally characterized TFLT devices under high-temperature conditions (up to 125°C) to evaluate their suitability for uncooled CPO.

Device Platform: Thin-film Lithium Tantalate (TFLT) waveguides and resonators. TFLT was chosen for its significantly lower photo-refractive effect compared to TFLN (109x more stable).
Experimental Setup:
- Devices were mounted on a temperature-stabilized hot plate monitored by thermocouples.
- Microring Resonators: Characterized for propagation loss and quality factors ( $Q_L$ and $Q_i$ ) at 25°C, 95°C, and 125°C using a tunable laser and fiber array.
- Mach-Zehnder Modulators (MZM): A 7 mm traveling-wave MZI was tested for DC bias stability, $V_\pi$ (half-wave voltage), and electro-optic (EO) bandwidth at 25°C, 91°C, and 120°C.
- Resonant Modulators: Coupling electro-optic modulators were tested for DC bias stability and resonance tuning capabilities at elevated temperatures.
Measurements: Included transmission spectra, $V_\pi$ extraction, long-term drift monitoring (1+ hours), and EO frequency response up to 50 GHz.

3. Key Contributions

First High-Temperature Demonstration: This is the first experimental demonstration of TFLT modulators operating stably at temperatures up to 125°C.
Elimination of Active Cooling: The study proves that TFLT devices maintain performance without active thermal control, addressing a primary barrier to CPO scaling.
DC Bias Stability: Demonstrated that TFLT modulators are intrinsically DC-bias stable at high temperatures, eliminating the need for power-hungry thermo-optic stabilization loops.
Resonant Modulator Viability: Showed that resonant TFLT modulators can be tuned via electro-optic bias (rather than thermal tuning) even at high temperatures, enabling compact, high-density designs.

4. Key Results

A. Propagation Loss and Quality Factors

Propagation loss in TFLT microring resonators was found to be largely unaffected by temperature.
The intrinsic quality factor ( $Q_i$ ) remained invariant across 25°C, 95°C, and 125°C, indicating that the material properties and waveguide losses are thermally robust.

B. MZM Performance (DC Bias and $V_\pi$ )

Bias Stability: The modulator bias remained stable at the quadrature point (Q-point) for over one hour at 120°C with no monotonic drift. Minor fluctuations were attributed to air convection from the hot plate affecting fiber coupling, not intrinsic device instability.
$V_\pi$ Improvement: Contrary to the expectation that performance degrades with heat, the half-wave voltage ( $V_\pi$ ) decreased by ~10% at 120°C (dropping from 3.8 V at 25°C to 3.4 V at 120°C). This is consistent with the temperature dependence of ferroelectric materials (like LN and LT) where the EO coefficient increases with temperature.
Bandwidth: The 7 mm MZM maintained an electro-optic bandwidth beyond 50 GHz at high temperatures, with no noticeable degradation in the EO response compared to room temperature.

C. Resonant Modulator Stability

Bias Tuning: Resonant modulators on TFLT demonstrated stable DC bias tuning at elevated temperatures.
Coupling Regime: While the resonator was under-coupled at room temperature, it transitioned to critically coupled at elevated temperatures due to thermo-optic changes in the coupling condition.
Stability: Resonance wavelength shifts remained stable over time (20 traces over 20 minutes) for applied DC voltages ranging from 0 V to 25 V, proving that electro-optic tuning is a viable, stable method for resonant devices in hot environments.

5. Significance and Impact

This work positions Thin-Film Lithium Tantalate (TFLT) as a premier candidate for next-generation uncooled Co-Packaged Optics (CPO).

Power Efficiency: By eliminating the need for active thermo-optic bias control (which typically consumes hundreds of mW per modulator), TFLT significantly reduces the power overhead and thermal load in CPO systems.
Scalability: The ability to operate at >120°C without performance degradation allows for denser integration of photonic components near high-heat silicon processors.
Compactness: The stability of resonant modulators at high temperatures enables the use of compact, high-density ring resonator designs, which are otherwise difficult to stabilize in TFLN due to the PR effect.
Performance Enhancement: The observed reduction in $V_\pi$ at high temperatures suggests that TFLT modulators may actually perform better electrically in the thermal environments typical of AI data centers.

In summary, the paper validates TFLT as a robust, high-performance, and power-efficient platform capable of meeting the stringent thermal and stability requirements of future AI-driven optical interconnects.