When the Light Reading a Thermometer Heats the Thermometer

Light sent to read a thermometer can change the value it reports. At first this may sound like a minor detail. When the goal is distinguishing thousandths of a degree, however, the measurement's own contribution becomes central. In silicon ring resonators, light does more than carry information: circulating inside the ring, it can alter the material's optical properties and temperature. A stronger readout signal therefore does not necessarily produce a more trustworthy temperature result.

A research manuscript shared on arXiv in June 2023 by Siegfried Janz and colleagues at Canada's National Research Council investigates this problem through experiments and models. It separates optical-power effects from distortions of the spectral lines used for thermometry. The document considered here is the authors' preprint; no claim is made about peer review of a separate journal version. The knowledge cutoff is 13 October 2026. The point is not to announce a finished commercial product, but to understand where a sensitive sensor can go wrong. [1]

Turning temperature into wavelength

A ring resonator is a small optical structure in which light circulates along a closed waveguide. Another waveguide brings light into the arrangement and allows its output to be measured. Resonances occur at particular wavelengths; in the configuration used here, they appear as dips in the transmitted spectrum. Temperature changes silicon's effective refractive index, shifting the resonance wavelength. An established temperature response connects that shift to a temperature reading. [1]

The team investigated four rings fabricated in silicon-on-insulator structures and operating near a wavelength of 1,550 nanometers. Their waveguide cross sections were 220 nanometers thick and 500 nanometers wide. These dimensions do not describe the whole measurement system. A laser, photodetector, fiber connections, and temperature-comparison equipment are also involved. A miniature sensing region does not mean an entire laboratory has become one microscopic component.

Why circulating power matters

The advantage of resonance is that light can circulate effectively within the ring. The same feature allows circulating optical power to become much greater than the incoming power. Absorbed energy can contribute to local heating. The optical response used to measure temperature consequently becomes sensitive to the light interrogating it. The measurement question is how much of a shift comes from the surroundings and how much from the conditions used to read the sensor. [1]

For Ring A, the researchers observed a resonance displacement of approximately fifteen picometers at one milliwatt launched from the laser into the input fiber. Interpreted as temperature, that displacement produced an error of nearly 0.2 kelvin. The milliwatt is not the power absorbed by the ring or the power circulating within it. Coupling and transmission losses separate those quantities. Where a measurement condition is defined matters as much as its numerical value. [1]

Reducing power is not the entire solution

The first response might be to weaken the probe light. Less light can reduce absorption-related heating, but also weakens the signal. Detector noise and statistical fluctuations of light eventually restrict the measurement. The engineering task is not simply minimizing illumination. It is identifying which distortion dominates while preserving enough signal to locate the resonance reliably.

Moreover, low-power behavior was not merely a smaller version of heating. In the more sensitive bath measurements, some rings initially shifted slightly toward shorter wavelengths. The researchers described a competing optical contribution to the effective index, alongside the positive change dominating at higher power. They discussed possible roles for carriers and defect states, while explicitly placing detailed identification of the microscopic mechanisms outside the study's scope. A wavelength shift should therefore not automatically be interpreted as the surroundings becoming warmer or cooler. [1]

The consequences of a longer ring

Changing geometry offered another route. Ring A had a cavity length of one hundred micrometers, while Ring B had a length of 950 micrometers. These are optical path lengths, not diameters. With comparable narrow linewidths maintained, longer rings could reduce circulating power and power-dependent shifts. This does not establish that every larger ring is superior in every respect. Waveguide loss and coupling into the ring have to be designed together. [1]

A longer cavity also reduces the wavelength separation between successive resonances. If temperature changes substantially between observations, the identity of the tracked resonance order can become ambiguous. That does not mean the sensor can operate only across a narrow temperature range. It is an identification problem introduced by intermittent interrogation. Continuous tracking can help avoid the ambiguity. A change that reduces one error source thus creates another requirement for the readout procedure.

Connections have mechanical and optical costs

Some rings were tested on an exposed temperature-controlled stage, while others were enclosed in probes immersed in a stirred liquid bath. Bath measurements allowed comparison with a calibrated standard platinum resistance thermometer. The silicon chip was held with a spring arrangement intended to avoid unnecessarily restricting its expansion. This helped limit stress-induced changes in refractive index. The mechanical assembly around an optical thermometer is not incidental packaging outside the measurement. [1]

Rather than bonding the fiber ends directly to the chip, the probe held them at a distance. This sacrificed some optical coupling, but could reduce difficulties associated with adhesive stress, alignment, and reflections. Maximizing transferred light was not the team's only objective. They wanted a connection that remained dependable as temperature changed. A strong signal and a stable attachment are related but distinct design goals.

Finding the bottom of a spectral dip

Backscattering within a ring and reflections elsewhere in the optical path can split a resonance or distort its shape. Yet an imperfect-looking line does not necessarily generate a temperature-dependent error. Some distorted profiles in the study retained their shape as temperature changed. If reliably tracked, such a feature can still carry temperature information. The difficulty concerns how the procedure used to locate the resonance interacts with that shape. [1]

For particular reflection-induced distortions, the team showed that fitting an ideal curve across the line could introduce more displacement than tracking the minimum-transmission point. This does not make curve fitting universally wrong. When time-varying external noise dominates, fitting can be helpful; using a narrower spectral window and averaging repeated scans can provide a different compromise. Readout software is not an independent finishing step applied after the optics have done their work. Its choice of which feature counts as the temperature marker participates in the result.

The distinction also explains why a visually smooth fit is not sufficient evidence of a correct reading. A curve can average random variation while remaining biased by a stable distortion. Conversely, a direct minimum measurement can remain vulnerable to noise. Deciding between procedures requires understanding the relevant error source, rather than selecting whichever plot looks cleaner.

Keeping sensitivity in perspective

The study reports that appropriate design and readout choices can limit the effects of the examined impairments to approximately three millikelvin. This is not a statement that every photonic thermometer has that total uncertainty at every temperature. Resolving a smaller change is also different from knowing absolute temperature with the same accuracy. The manuscript explicitly identifies individual-device calibration and more stable wavelength measurements as requirements for further progress. Its substantive achievement is to treat probe light's influence as a measurable design variable instead of assuming that the light is an uninvolved observer. [1]

Sources

[1] Siegfried Janz, Sergey Dedyulin, D.-X. Xu, Martin Vachon, Shurui Wang, Ross Cheriton, and John Weber (2023), Measurement Accuracy in Silicon Photonic Ring Resonator Thermometers: Identifying and Mitigating Intrinsic Impairments, arXiv:2306.15643v1. Author record and license link; research manuscript. Original English explanation adapted with attribution from the CC BY 4.0-licensed manuscript; figures are not reproduced.

Source access and factual review: 15 October 2026. Knowledge cutoff: 13 October 2026. The main experimental, modeling, and concluding sections were directly read. Raw data and models were not independently reanalyzed or rerun. Cavity lengths, diameters, and optical-power definitions are distinguished; no numerical claim relies on the source's inconsistent descriptions of Ring C's length.