A research team from Germany's Ferdinand-Braun-Institut (FBH), the University of Strathclyde in the UK, and the University of Cambridge has demonstrated optical wireless communication at up to 1.5 Gbps using LEDs that emit far-UVC light. On September 30, 2026, FBH announced that this is the highest speed achieved to date for optical wireless communication using wavelengths below 235 nm.
Far-UVC light is little affected by sunlight reaching the Earth's surface, which makes it easier to keep background noise low even outdoors. Showing that gigabit-class communication is possible with LEDs is significant. However, the experiment was a face-to-face link between a transmitter and receiver over just 30 cm, indoors, and the LED with the widest response bandwidth did not post the highest data rate. The results suggest that, beyond making LEDs smaller and improving signal processing, delivering enough light to the receiver is just as important.
Up to 1.5 Gbps over 30 cm
The study by Hichem Zimi and colleagues was published as a peer-reviewed conference paper in the proceedings of CSNDSP 2026. The conference was held July 15–17, and the university's records list the paper's publication date as August 24. FBH's announcement at the end of September presented the findings to a wider audience. University publication record
The transmitter was an aluminum gallium nitride (AlGaN) LED fabricated by FBH. The paper treats it as a "235 nm" LED, but the measured emission peak was roughly 233–234 nm, with a full width at half maximum (FWHM) of about 10 nm.
FWHM indicates the width of the wavelength range in which the light is at least half as intense as at the peak. In other words, this LED does not emit a single wavelength; it emits across a certain range centered on the peak.
The team focused the LED's light with two lenses and received it with an avalanche photodiode (APD) placed about 30 cm away. An APD is a light-receiving device that converts weak light into an electrical signal while amplifying it internally.
According to FBH, the measurements were made under ordinary indoor lighting. The use of a wavelength that resists solar noise is an advantage, but the 1.5 Gbps figure was not confirmed in an outdoor environment.
The speed gain owes as much to how the signal is sent as to improvements in the LED itself. The team had previously demonstrated 650 Mbps using the same type of microstructure LED, combining on-off light modulation with equalization.
This time they used DC-biased optical OFDM. This method divides data across multiple frequency components and varies the number of bits and the power assigned to each according to its signal-to-noise ratio (SNR). Components in good condition carry more information, while those prone to noise carry less, making efficient use of the limited bandwidth.
The maximum 1.5 Gbps is a transmission rate evaluated against a bit error rate of 3.8×10⁻³. This does not mean every received bit was error-free; it is a threshold that assumes forward error correction (FEC) is used.
In addition, waveform generation for transmission and demodulation after reception were done offline. The figure therefore reflects the performance of the communication scheme itself, and should be distinguished from a finished, continuously operating device or the data rates users would actually experience. Conference paper by Zimi et al.
The widest-bandwidth LED did not give the highest speed
The experiment compared three LED types: the standard "Std," the intermediate-size "S2," and "S3," which has the smallest emission area. All three chips are 1 mm square, but the shape and area of the part that actually emits light differ.
In S2 and S3, multiple small square emitters are arranged side by side and connected electrically in parallel—22 in S2 and 8 in S3.
Shrinking the emitters reduces junction capacitance, the property by which a semiconductor junction stores charge. Pushing the same current through a smaller area also raises current density, which can shorten the recombination time of the carriers involved in light emission. The aim of miniaturization is to let the LED follow changes in the electrical signal more quickly.
However, the ranking of data rates did not match the ranking of bandwidth.
S3, with the smallest emission area, had the widest −3 dB bandwidth of the three at 140.8 MHz. Yet its top data rate was only 1.3 Gbps, below the 1.5 Gbps recorded by the standard LED at 65.8 MHz.
| LED design | Current at bandwidth measurement | Optical output at same condition | −3 dB bandwidth | Max data rate | Current at max rate |
|---|---|---|---|---|---|
| Standard (Std) | 70 mA | 1.73 mW | 65.8 MHz | 1.5 Gbps | 100 mA |
| Intermediate (S2) | 20 mA | 0.64 mW | 91.3 MHz | 1.5 Gbps | 70 mA |
| Smallest (S3) | 20 mA | 0.63 mW | 140.8 MHz | 1.3 Gbps | 50 mA |
Source: Zimi et al., CSNDSP 2026. Combines the bandwidth and optical output listed in Table II of the author manuscript with the maximum speeds and drive currents from Section III and Fig. 6. Bandwidth and optical output were measured at the same operating point, but the maximum data rates were obtained at different currents. The communication distance was about 30 cm, evaluated against the bit error rate described above. Published author manuscript
The −3 dB bandwidth is the frequency at which the LED's response to the applied electrical signal falls 3 dB below its low-frequency level. A larger value means the LED can convey faster changes, but the number does not translate directly into bits per second.
With OFDM, multiple bits can be carried on a single frequency component, so bandwidth in MHz cannot simply be converted into a data rate in Mbps.
The researchers cite received light level and noise as the reason S3's wide bandwidth did not translate into the highest data rate. An LED with a smaller emission area produces less optical output, which lowered the SNR obtained at the receiver.
Even if an LED can flicker quickly, the receiver cannot read those changes accurately unless enough light reaches it. As a result, it becomes harder to load many bits onto each frequency component.
In short, making the LED smaller and widening its bandwidth does not by itself raise the data rate. The paper also says the effects of heat and current flow on performance could not be fully separated.
To go faster, it will be necessary not only to speed up the LED's response but also to collect light more efficiently onto the receiver and raise the receiver's sensitivity.
Low solar noise doesn't guarantee long range
Solar ultraviolet below 280 nm is almost entirely absorbed in the upper atmosphere and barely reaches the ground. Compared with optical communication using visible or infrared light, sunlight is less likely to become background noise for a receiver in the UVC band. This is one reason FBH hopes for outdoor applications. FBH announcement
However, low solar noise is a separate matter from whether the transmitted signal arrives strongly enough over long distances.
Light from an LED spreads with distance, and the fraction entering the receiver shrinks. Extending the distance tends to lower the SNR, which governed the data rate in this experiment as well. Even if external noise from sunlight is suppressed, the problem of the received light itself being weak remains.
A clue to achievable range comes from a separate study by the same group, published in Optics Express in 2025.
In line-of-sight communication using commercial LEDs, a 235 nm LED achieved 1 Gbps at 2 m and 0.5 Gbps at 10 m. A 255 nm LED, meanwhile, recorded 2 Gbps at 2 m and 0.9 Gbps at 30 m.
Summarizing the earlier work as "up to 2 Gbps at 30 m" or "gigabit-class communication up to 30 m in the 235 nm band" would therefore combine results obtained under different conditions. 2025 prior paper
The earlier and current experiments differ in both the LEDs used and the communication distance. Directly comparing the current 1.5 Gbps at 30 cm with the earlier 1 Gbps at 2 m, and calculating a percentage improvement from miniaturization, is not possible.
Still, the results show that range and data rate vary with wavelength and transmitted light level, and that research on gigabit-class UVC optical wireless communication is continuing to advance.
Far-UVC also has a property whereby Rayleigh scattering by molecules in the air grows stronger at shorter wavelengths. If part of the transmitted light scatters in other directions in the air, a signal might be received even when the transmitter and receiver are not directly facing each other.
This does not mean the light passes through walls. It is the idea of non-line-of-sight communication, in which light scattered in the air is received via a path different from the direct one.
FBH says it will next investigate how far such non-line-of-sight communication can be used and how environmental conditions affect it.
The 1.5 Gbps figure is an experimental result with a direct optical path between transmitter and receiver. The same speed cannot be expected where obstacles are present, and how efficiently weak scattered light can be received is the next challenge.
Wavelength alone can't settle safety near people
Light at around 233 nm is absorbed more strongly at the skin's surface and penetrates less deeply than the 254 nm light traditionally used for germicidal UV. This property has led researchers to explore whether far-UVC can be used in spaces where people are present.
However, it cannot be judged safe for humans on wavelength alone. The communication paper itself describes safety as an area that still requires research.
In a separate study reported in Photochemistry and Photobiology, David Welch and colleagues used a 233 nm LED fitted with a filter that suppresses unwanted longer-wavelength light, and examined effects on human skin tissue and other samples.
Skin tissue from two donors was irradiated with 100 mJ/cm² and examined 30 minutes later. Cyclobutane pyrimidine dimers (CPDs), a type of DNA damage, were found in about 8% of epidermal cells. At the same dose, 254 nm light produced them in about 45%.
The result shows that 233 nm light is less likely than 254 nm to affect deeper tissue, but the effect on DNA was not zero. Skin safety study by Welch et al.
The skin tissue used in that study came only from two donors of the same skin type. The authors also say further verification is needed, including people of different ages and skin colors.
The study also used an irradiation device with an optical filter. Even though both involve LEDs near 233 nm, its results cannot be applied directly to the communication LEDs in this work.
To consider effects on the human body, it is necessary to check not only the peak emission wavelength but also how much of the light falls in surrounding wavelengths, and the cumulative dose determined by the intensity and duration of exposure. Results on skin effects alone also cannot guarantee long-term safety, including for the eyes.
For practical use, devices will need to be designed so that enough light reaches the receiver for communication at the required distance while keeping the far-UVC dose people receive within acceptable limits.
If receiver sensitivity can be raised or light collected more efficiently, it may be possible not only to extend speed and range but also to reduce the amount of light that must be transmitted.
Whether the desired distance and speed can be achieved together within exposure conditions people can tolerate will be key to moving high-speed far-UV LED communication from the laboratory to environments where people are present.