We demonstrate a series of multifunctional micro-photodetectors (μPDs) designed for high-speed ultraviolet-A-(UVA)-light detection and blue-light illumination based on InGaN/GaN triple-quantum-well devices grown on patterned sapphire substrates. At forward voltage bias, the device operated as a light-emitting diode with a peak emission wavelength of ∼450 nm. When switching to reverse voltage bias, the device exhibited a dual-band responsivity of 0.17 A/W at 370 nm and 0.14 A/W at 400 nm, indicating effective UVA light detection. Furthermore, size-dependent emission and detection behaviors were investigated with the device's active area having radii ranging from 15 to 50 μm. For μPDs, the −3-dB bandwidth increased with the reduced device radius and reached a maximum of 689 MHz for the 15-μm device under −10-V bias. High responsivity and fast modulation speed contributed to 2-Gb/s UVA optical wireless communication based on direct-current-biased optical orthogonal frequency-division multiplexing. The research offers a promising approach to simultaneous high-speed communication and illumination in the UVA-blue-light optical spectral regime, and the dual-band responsivity feature is potentially useful for increasing channel capacity.

Optical wireless communication (OWC) has emerged as a compelling alternative to traditional radio frequency (RF) technologies, particularly when RF faces bandwidth constraints and electromagnetic interference.1–3 OWC offers the advantages of high data rates, enhanced security, and energy efficiency, garnering significant research interest in this domain.4 Among the OWC spectrum, the ultraviolet (UV) wavelength brings many distinctive advantages. First, ultraviolet-A (UVA) wavelength extends the usable spectral band of visible-light communication to the UV regime, which paves the way for higher channel capacity and throughput by facilitating broader data transmission. Another perspective of light in the UVA wavelength range is that it exhibits more scattering features than visible light in water,5,6 making it advantageous for non-line-of-sight (NLoS) and diffused LoS communication in underwater wireless optical communication (UWOC) applications.7,8 Moreover, UVA wavelengths within the UV spectrum are relatively less harmful to humans when compared to shorter UV wavelengths, providing safe usage with sustained human activity.

Conventional Si-based photodetectors (PDs) face challenges in UV detection due to their inherent sensitivity to visible light, leading to potential interference and noise from ambient sources.9 Their structure and properties also result in inefficient light absorption in the UV spectrum.10 Prolonged exposure to UV can further reduce the lifespan of Si-based devices, and the heat generated during this process elevates the noise level.11 In contrast, GaN, with a direct wide-bandgap of 3.4 eV, allowing for efficient light emission and absorption in the UV range.12 GaN also holds an advantage over silicon in operating effectively in harsh environments (e.g., space) with its heightened resistance to cosmic rays.13,14 Moreover, its unique properties, such as high electron mobility and thermal stability, combined with bandgap tunability,15 make GaN an optimal choice for UV-based optoelectronic devices.

Up to now, many efforts have been devoted to achieving high-performing GaN-based PD. Li et al. designed an InGaN-based resonant cavity for enhanced light absorption and responsivity by making high-reflectivity lateral porous GaN distributed Bragg reflectors.16 Pereiro et al. investigated the impact of device design for p-i-n InGaN/GaN-based multiple quantum well (MQW) PDs. With meticulous optimizations of epitaxial structure design and fabrication process, an optimal responsivity of ∼100 mA/W was achieved.17 In our prior work, by utilizing a semipolar InGaN/GaN MQW μPD with an enhanced overlap ratio of electron–hole wave functions, a 347-MHz bandwidth, and 1.55-Gb/s data rate were demonstrated in a visible-light communication system.18 Carrano et al. investigated using a 2-μm thick intrinsic layer to have low capacitance for metal–semiconductor–metal PD.19,20 Reducing the PD's active area can lead to diminished capacitance that influences the device response speed.21–24 it can potentially improve communication performance, producing a more responsive PD. Additionally, during the PD fabrication, the dry etching process can also induce sidewall damage, which may generate leakage current and escalate noise. Thus, appropriate sidewall passivation is essential to curtail the dark current level, paving the way for a PD with a higher signal-to-noise ratio (SNR).25–27 

Herein, we demonstrated a comprehensive investigation of μPDs with an optimized fabrication process and analysis of the performance metrics regarding size-dependent light emission, detection, and high-speed UVA-based OWC. Varying active areas having radii from 15 to 50 μm facilitated a detailed study of the modulation bandwidth with resistance–capacitance (RC) time constant, revealing the underlying relationship between the device active area and modulation speed. This study provides profound insights into the optimization of large-scale fabrication of InGaN/GaN-based optoelectronic devices for high-performance data and energy transmission systems.

The structure for the μPD is depicted in Fig. 1(a), which was grown on patterned sapphire substrates (PSS) using metal-organic chemical vapor deposition (MOCVD). The growth process began with thick Si-doped n-GaN on the PSS, followed by three pairs of InGaN/GaN MQWs and Mg-doped p-GaN. The μPD was fabricated using the process shown in Fig. 1(b). Initially, a 300-nm SiO2 hard mask was deposited using plasma-enhanced chemical vapor deposition (PECVD) at a temperature of 300 °C. The μPD active area mesa was patterned into circular shapes with radii ranging from 15 to 50 μm using the direct laser writer. Subsequently, a two-step inductively coupled plasma reactive ion etching (ICP-RIE) process was conducted. After etching, the SiO2 hard mask was removed by immersing the chip in buffered oxide etch (BOE) solution, which could also effectively mitigate the sidewall damage resulting from the dry etching process.28 Directly after depositing a 5-nm-thick adhesion layer of Ni using an e-beam evaporator, a 120-nm layer of indium-tin-oxide (ITO) was deposited using the RF magnetron sputter. After annealing the ITO layer at 250 °C in an O2 atmosphere for 10 min, the n-metal contact of Ti/Au (50 nm/190 nm) was deposited using RF magnetron sputtering.

FIG. 1.

(a) 3D Schematic diagram of InGaN/GaN-based MQW μPD. Insets show the SEM (scanning electron microscopy) cross section image of the epitaxial structure (top right) and the optical microscope image of fabricated 15 to 50-μm μPDs (bottom left) with 50-μm emission spectra (bottom right). (b) Diagram of μPD fabrication process.

FIG. 1.

(a) 3D Schematic diagram of InGaN/GaN-based MQW μPD. Insets show the SEM (scanning electron microscopy) cross section image of the epitaxial structure (top right) and the optical microscope image of fabricated 15 to 50-μm μPDs (bottom left) with 50-μm emission spectra (bottom right). (b) Diagram of μPD fabrication process.

Close modal

For sidewall passivation and n-metal isolation, the deposition of SiO2 was executed in a two-step procedure. Initially, the sample was cleaned using N2 and NH3 plasma for 30 s each. This method aids in suppressing the formation of pinholes29 and diminishing carbon and oxygen contamination30,31 in the SiO2. Subsequently, 350 nm of SiO2 was deposited using PECVD, which serves three essential functions: (i) Sidewall passivation, effectively reducing surface states and subsequently minimizing non-radiative recombination; (ii) As anti-reflection layer for enhanced light absorption. (iii) Preventing short-circuit between n-/p-contact, ensuring a low dark current for photodetection. Following the sidewall passivation, dry etching was utilized to remove the SiO2 and expose ITO for p-contact. The p-metal comprised of Ni/Au (50 nm/190 nm) was sputtered onto the ITO. Annealing was performed at 400 °C rapid thermal processing (RTP) for 1 min in a nitrogen atmosphere to enhance the Ohmic contact.

With the p-i-n MQW structure exhibits distinct behaviors under different biases. Under a forward bias, the p–n junction promotes the recombination of electrons and holes for light emission. When a reverse bias is applied, it separates photon-generated electrons and holes and allows for optical–electrical conversion, enabling the device to function as a PD. The bright blue-light emission obtained under 1-mA injected current is shown in Fig. S1. The current–voltage (I–V) measurement was carried out using a sourcemeter, and light power (L) was measured using a powermeter with a Si-based PD. Figure 2(a) shows the L–I–V and calculated external quantum efficiency (EQE) curves of 50-μm μLED. By linear fitting the I–V curve, the turn-on voltage was obtained ∼4 V. As the current increases, the normalized EQE exhibits about 40% decrease. This droop could be attributed to various factors such as Auger recombination, thermal effects, and the phonon bottleneck phenomenon.32–34 The emission spectra of each device were measured as shown in Fig. S2. Figure 2(b) summarizes the extracted full-width at half-maximum (FWHM) and peak wavelength from the emission spectra vs current density with device radii ranging from 15 to 50 μm. The results show that the smaller device tends to have a shorter emission wavelength and a narrower FWHM with higher carrier and optical confinements. Notably, an increase in current density resulted in a blue shift of approximately 1–2 nm due to band-filling effects, and the shift was more significant in devices with smaller LEDs. At 50 μm, the area-to-volume ratio is lower, resulting in comparatively less heat dissipation. Therefore, beyond 800 A/cm2, there is a tendency for a red shift to occur. According to our previous work,35 when serving as the transmitter, the 10-μm μLED could reach 57 MHz of −3-dB modulation bandwidth and 600-Mb/s net data rate using on–off keying (OOK) modulation.

FIG. 2.

μLED characterizations of fabricated devices. (a) L–I–V curve of 50-μm device with EQE corresponding to different current density. (b) Extracted FWHM and peak wavelength from emission spectra vs current density with varying device radii.

FIG. 2.

μLED characterizations of fabricated devices. (a) L–I–V curve of 50-μm device with EQE corresponding to different current density. (b) Extracted FWHM and peak wavelength from emission spectra vs current density with varying device radii.

Close modal
To characterize the photodetection properties, the devices were initially assessed with a semiconductor parameter analyzer for broadband responsivity measurement, which has a current measurement limit of ∼1 × 10−13 A. A 500-W mercury-xenon [Hg(Xe)] arc lamp illuminated the μPD. The light was channeled through a monochromator equipped with a grating mirror. Each wavelength of light was calibrated to 0.8 mW/cm2 using neutral density (ND) filters. Figure 3(a) displays the responsivity results from 0 to −5 V in the 230 to 440 nm wavelength range. The formula used to calculate the responsivity, R λ, is as follows:
R λ = I Photo I Dark P inc ,
(1)
where I Photo and I Dark denote photocurrent and dark current, respectively. P inc represents the incident light power received by μPD. Figure 3(a) shows that the responsivity gradually increased from 230 to 350 nm across all voltages. This trend matches well with the absorption of the ITO layer, which is opposite to the transmission spectrum as shown in Fig. 3 orange line, which implies that light within this specific wavelength range is dominantly absorbed by the ITO layer before going into the active layer and above 350 nm at −5 V, two distinct peaks occurred around 370 nm (0.17 A/W) and 400 nm (0.14 A/W), which could be attributed to different band edge absorption by GaN and InGaN layers. This dual-band photoresponse allows for the detection of a broader range of light signals demanded by multichannel OWC applications. Furthermore, an exponential cutoff in responsivity occurred after 420 nm due to light transparency below the bandgap, following Urbach's rule.36 It is worth noting that at zero bias, the responsivity is 0.149 A/W at 370 nm, showing only a 13% decrease when compared to the responsivity at a −5-V applied voltage, indicating the potential of μPD to be used as a self-powered device.
FIG. 3.

μPD characterizations. (a) Responsivity of μPD with varying wavelengths and different reverse biases under a light irradiance of 0.8-mW/cm. The orange line indicates ITO transmittance. (b) Dark current and photocurrent of μPD when illuminated by a 375-nm LD with laser power density ranging from 2.3 × 10−2 to 1.3 × 105 mW/cm2. (c) Photocurrent density received by μPD at −10 V vs laser power. (d) Specific detectivity and NEP vs reverse bias under 1.3 × 105 mW/cm2 irradiance.

FIG. 3.

μPD characterizations. (a) Responsivity of μPD with varying wavelengths and different reverse biases under a light irradiance of 0.8-mW/cm. The orange line indicates ITO transmittance. (b) Dark current and photocurrent of μPD when illuminated by a 375-nm LD with laser power density ranging from 2.3 × 10−2 to 1.3 × 105 mW/cm2. (c) Photocurrent density received by μPD at −10 V vs laser power. (d) Specific detectivity and NEP vs reverse bias under 1.3 × 105 mW/cm2 irradiance.

Close modal

According to the result of broadband responsivity, a 375-nm laser diode (LD) was selected to achieve an optimized photodetection performance for detailed characterizations. Figure 3(b) illustrates the power-dependent IV curves from −10 V to +5 V. The dark current gradually increased from 6 × 10−15 A at 0 V to 1.7 × 10−11 A at −10 V. At a voltage lower than −2 V, it reached the measurement limit of the equipment, resulting in noisy data being observed. Such a low-level dark current signifies the effective sidewall passivation achieved through BOE wet etching and plasma treatment before SiO2 deposition. The LD power was changed from 2.35 × 10−2 to 1.37 × 105 mW/cm2 with ND filters. The overall photocurrent followed the linear change with the laser power density. Figure 3(c) shows the linear power dependence between log-scale photocurrent density and the light power density at −10 V, with the blue line representing the linear fitting. The linear dynamic range (LDR) of μPD can be calculated with LDR = 20 × log(Psat/Pmin). Psat represents saturated power, i.e., the incident light power, which makes the photocurrent stop changing with increasing optical power, and Pmin is lowest detectable power. In our experiment, even at the highest light power density of 1.37 × 105 mW/cm2, no points deviating from the linear trend, indicating no μ PD saturation were observed. Therefore, the LDR is estimated to be >129.3 dB.

Furthermore, the specific detectivity (D*) and noise equivalent power (NEP) were investigated, which provide insights into the device's ability to discern weak signals in the presence of noise. The formulas for D* and NEP are as follows:
D * = ( R λ ) A 2 e I Dark = A NEP .
(2)
A represents the active area of the μPD, and e stands for the elementary charge. Under LD illumination, the specific detectivity exhibited an exponential increase with decreasing voltage, attaining values of 3.2 × 1015 and 5.7 × 1016 cm Hz1/2 W−1 when operated at −10 and 0 V, respectively [see Fig. 3(d)]. The NEP value decreased exponentially with the reduction in voltage, showing values of 1.0 × 10−14 and 8.1 × 10−16 W/Hz1/2 at −10 and 0 V, respectively. However, the dark current level exceeded the measurement limits, resulting in an inaccurate calculation with noise. As the transparent data points indicated, data beyond −2 V was not precise, and values were smoothed for prediction. Large LDR, high detectivity, and low NEP ensured a high-performance operation for the following OWC applications.
To comprehensively explore the influence of the device active area on modulation speed, we conducted capacitance–voltage (C–V) measurements using the semiconductor parameter analyzer and calculated the series resistance from I–V curve to estimate the RC limit. Figure 4(a) shows the capacitance and series resistance values across the active area. As the active area of the μPD increases, the capacitance increases linearly, while the resistance exhibits an inverse trend. After fitting the measured resistance and capacitance values and multiplying them, the RC constant was expected to be lowest for a circular device with a radius of 27 μm as shown in Fig. 4(b). However, this theoretical prediction can be deviated from practical results due to the variation in the growth/fabrication process and measurement setups. The modulation frequency of the photodiode is determined by a series of time constants, which follows the equation:37 
f 3 d B = 3 2 π ( 1 τ R + 1 τ N R + 1 τ R C ) ,
(3)
where τ R, τ N R represent radiative, non-radiative lifetime and τ R C is RC time constant, and these time constants are garnered by device active area. For example, as the active area decreases to a specific value (2290 μm2 in this study), τ R C reaches the minimum as shown in Fig. 4(b). Additionally, smaller samples allow for higher current density injection with lower self-heating effect, which can further decrease τ R, which is inversely proportional to the current density. Consequently, PDs with a smaller active area usually exhibit faster modulation frequencies.
FIG. 4.

(a) Measured series resistance (black dots) and capacitance (blue dots) with fitting curves vs μPD active area. (b) Estimated RC time constant. (c) Normalized frequency response for a 50-μm μPD. (d) −3-dB bandwidth for varying-active area μPDs vs reverse bias ranging from −10 to 0 V.

FIG. 4.

(a) Measured series resistance (black dots) and capacitance (blue dots) with fitting curves vs μPD active area. (b) Estimated RC time constant. (c) Normalized frequency response for a 50-μm μPD. (d) −3-dB bandwidth for varying-active area μPDs vs reverse bias ranging from −10 to 0 V.

Close modal

UVA-based OWC link was established to measure the small-signal frequency response. The measurement setup is illustrated in Fig. S4. The 375-nm UVA LD was employed as the transmitter, and the emitted light was focused on the μ PD chip as receiving module. Bias-tees were used on both sides to combine the driving DC and modulation signal for the transmitter while applying reverse bias and extracting photocurrent. A vector network analyzer (VNA) was used to generate sinusoidal signals from 300 kHz to 1 GHz and analyze the receiving signal from μ PD to obtain frequency response through S21 measurement. Figure 4(c) shows the normalized −3-dB modulation bandwidth for a μ PD with a radius of 50 μm. As the reverse bias increased from 0 to −10 V, a noticeable increase occurred in the −3-dB modulation bandwidth from 13.8–296 MHz. As shown in Fig. 4(d), the modulation bandwidth increased with reduced device active area and increased reverse bias. At −10 V, the −3-dB bandwidths for device radii of 50, 30, 20, and 15 μm are 257, 296, 641, and 689 MHz, respectively. When the bandwidth of μPD exceeds 600 MHz, the system's speed are no longer limited by the μPD but determined by the LD, which has a maximum bandwidth of 600 MHz when mounted.

The broad modulation bandwidth facilitated high-speed data transmission using the DC-biased optical orthogonal frequency-division multiplexing (DCO-OFDM) method to transfer quadrature amplitude modulation (QAM) signal. The modulation signal was generated using an arbitrary waveform generator with a sampling rate of 1.2 GSamples/s, The signal received by μPD was recorded through an oscilloscope with a sampling rate of 2.5 GSamples/s. Figure 5(a) shows the spectral efficiency of 15 and 50-μm μPDs during the implementation of DCO-OFDM with the constellation diagrams shown in Fig. S5. After estimating the SNR of each subcarrier by sending a 4-QAM test signal (see Fig. S6), the spectral efficiency can be obtained by using the Shannon limit of log2(1 + SNR), which determines the data rate that can be sent over a unit frequency. Integrating the spectral efficiency over the used bandwidth (until 600 MHz) provides gross data rates of 1.57, 1.86, 2.20, and 2.23 Gb/s with 15-, 20-, 30-, and 50-μm devices, respectively. After removing 7% overhead for forward error correction (FEC), the corresponding net data rates were achieved of 1.40, 1.63, 1.95, and 1.99 Gb/s as compared in Fig. 5(b), and the average bit error ratios (BER) for these devices are all within the FEC limit of 3.8 × 10−3. It is evident that although the device with a smaller active area exhibited wider bandwidth, the data rate shows the opposite trend. This is attributed to the fact that, within a fixed usable bandwidth of the transmitter, the reduced active area significantly diminishes the SNR due to the limited photon absorption and photocurrent generation. The data rate of smaller μPD can be further improved if the usable bandwidth can be extended with a faster laser.

FIG. 5.

Implementation of UVA-based OWC using DCO-OFDM. (a) Net data rates with varying-active area μPDs. (b) Spectral efficiency of each subcarrier for μPDs with radii of 15 and 50 μm.

FIG. 5.

Implementation of UVA-based OWC using DCO-OFDM. (a) Net data rates with varying-active area μPDs. (b) Spectral efficiency of each subcarrier for μPDs with radii of 15 and 50 μm.

Close modal

In conclusion, we demonstrated a comprehensive study of multifunctional μLEDs/μPDs fabricated on an InGaN/GaN MQW-based chip. These devices, capable of operating as high-efficiency LEDs and high-speed photodetectors, offer significant advantages for high-speed OWC, particularly in the domain of UVA detection. With an optimized sidewall passivation process, an extremely low dark current in the fA scale was achieved, allowing for a broad LDR exceeding 129.3 dB. The device also exhibits dual-band responsivity, with peaks at both 370 nm (0.17 A/W) and 400 nm (0.14 A/W). The size-dependent investigation of modulation bandwidth revealed an increase with decreasing device active area, peaking at 689 MHz with 15-μm-radius μPD under −10-V bias. Significantly, employing OFDM yielded data rates up to 2 Gb/s with a single 50-μm μPD, holding the promise of higher speed (up to Tb/s) data reception if the array of μPDs can be fabricated and operated in UVA-based OWC systems.

See the supplementary material for μLED spectra, μPD communication measurement setup, and more detail result of DCO-OFDM.

This work was supported by King Abdullah University of Science and Technology (KAUST) (Grant Nos. BAS/1/1614-01-01 and ORA-2022-5313).

The authors have no conflicts to disclose.

Tae-Yong Park: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Yue Wang: Conceptualization (supporting); Data curation (supporting); Formal analysis (supporting); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Omar Alkhazragi: Data curation (supporting); Formal analysis (supporting); Investigation (supporting); Methodology (supporting); Writing – review & editing (supporting). Jung-Hong Min: Methodology (supporting); Resources (supporting); Writing – review & editing (supporting). Tien Khee Ng: Funding acquisition (supporting); Project administration (supporting); Supervision (supporting); Writing – review & editing (supporting). Boon S. Ooi: Conceptualization (equal); Funding acquisition (lead); Project administration (lead); Supervision (lead); Writing – review & editing (supporting).

The data that support the finding of this study are available within the article and its supplementary material.

1.
H.
Haas
,
C.
Chen
, and
D.
O'Brien
, “
A guide to wireless networking by light
,”
Prog. Quantum Electron.
55
,
88
111
(
2017
).
2.
H.
Burchardt
,
N.
Serafimovski
,
D.
Tsonev
,
S.
Videv
, and
H.
Haas
, “
VLC: Beyond point-to-point communication
,”
IEEE Commun. Mag.
52
(
7
),
98
105
(
2014
).
3.
D.
Tsonev
,
S.
Videv
, and
H.
Haas
, “
Light fidelity (Li-Fi): Towards all-optical networking
,”
Proc. SPIE
9007
,
900702
(
2014
).
4.
H.
Kressel
,
Semiconductor Devices for Optical Communication
(
Springer
,
Berlin/Heidelberg
,
1982
).
5.
X.
Sun
,
W.
Cai
,
O.
Alkhazragi
,
E.-N.
Ooi
,
H.
He
,
A.
Chaaban
,
C.
Shen
,
H. M.
Oubei
,
M. Z. M.
Khan
,
T. K.
Ng
,
M.-S.
Alouini
, and
B. S.
Ooi
, “
375-nm ultraviolet-laser based non-line-of-sight underwater optical communication
,”
Opt. Express
26
(
10
),
12870
(
2018
).
6.
A.
Morel
,
B.
Gentili
,
H.
Claustre
,
M.
Babin
,
A.
Bricaud
,
J.
Ras
, and
F.
Tièche
, “
Optical properties of the ‘clearest’ natural waters
,”
Limnol. Oceanogr.
52
(
1
),
217
229
(
2007
).
7.
Y.
Wang
,
J.-X.
Wang
,
O.
Alkhazragi
,
L.
Gutiérrez-Arzaluz
,
L.
Gutiérrez-Arzaluz
,
H.
Zhang
,
C. H.
Kang
,
T. K.
Ng
,
O. M.
Bakr
,
O. M.
Bakr
,
O. F.
Mohammed
, and
B. S.
Ooi
, “
Multifunctional difluoroboron β-diketonate-based luminescent receiver for a high-speed underwater wireless optical communication system
,”
Opt. Express
31
(
20
),
32516
32528
(
2023
).
8.
C. H.
Kang
,
A.
Trichili
,
O.
Alkhazragi
,
H.
Zhang
,
R. C.
Subedi
,
Y.
Guo
,
S.
Mitra
,
C.
Shen
,
I. S.
Roqan
,
T. K.
Ng
,
M.-S.
Alouini
, and
B. S.
Ooi
, “
Ultraviolet-to-blue color-converting scintillating-fibers photoreceiver for 375-nm laser-based underwater wireless optical communication
,”
Opt. Express
27
(
21
),
30450
30461
(
2019
).
9.
J.
Romijn
,
S.
Vollebregt
,
L. M.
Middelburg
,
B. E.
Mansouri
,
H. W.
van Zeijl
,
A.
May
,
T.
Erlbacher
,
J.
Leijtens
,
G.
Zhang
, and
P. M.
Sarro
, “
Integrated 64 pixel UV image sensor and readout in a silicon carbide CMOS technology
,”
Microsyst. Nanoeng.
8
(
1
),
114
(
2022
).
10.
C. H.
Kang
,
I.
Dursun
,
G.
Liu
,
L.
Sinatra
,
X.
Sun
,
M.
Kong
,
J.
Pan
,
P.
Maity
,
E. N.
Ooi
,
T. K.
Ng
,
O. F.
Mohammed
,
O. M.
Bakr
, and
B. S.
Ooi
, “
High-speed colour-converting photodetector with all-inorganic CsPbBr3 perovskite nanocrystals for ultraviolet light communication
,”
Light: Sci. Appl.
8
(
1
),
94
(
2019
).
11.
N. E.
Helwig
,
S.
Hong
, and
E. T.
Hsiao-wecksler
,
CCD Image Sensors in Deep-Ultraviolet
(
Springer-Verlag
,
Berlin/Heidelberg
,
2005
).
12.
S. K.
Zhang
,
W. B.
Wang
,
I.
Shtau
,
F.
Yun
,
L.
He
,
H.
Morkoç
,
X.
Zhou
,
M.
Tamargo
, and
R. R.
Alfano
, “
Backilluminated GaN/AlGaN heterojunction ultraviolet photodetector with high internal gain
,”
Appl. Phys. Lett.
81
(
25
),
4862
4864
(
2002
).
13.
J. M.
Osheroff
,
J. M.
Lauenstein
, and
R. L.
Ladbury
, “
LET and range characteristics of proton recoil ions in gallium nitride (GaN)
,”
IEEE Trans. Nucl. Sci.
68
(
5
),
597
602
(
2021
).
14.
G.
Patterson
and
J.
Roberts
, “
Gallium nitride—Delivering its promise in automotive applications
,” in
6th Hybrid and Electric Vehicles Conference (HEVC 2016)
,
2016
.
15.
F.
Rasouli
and
Z.
Hemmat
, “
Simulation of InGaN-based near-ultraviolet/visible dual-band photodetector
,”
Opt. Quantum Electron.
48
(
1
),
23
(
2016
).
16.
J.
Li
,
C.
Yang
,
L.
Liu
,
H.
Cao
,
S.
Lin
,
X.
Xi
,
X.
Li
,
Z.
Ma
,
K.
Wang
,
A.
Patanè
, and
L.
Zhao
, “
High responsivity and wavelength selectivity of GaN‐based resonant cavity photodiodes
,”
Adv. Opt. Mater.
8
(
7
),
1901276
(
2020
).
17.
J.
Pereiro
,
C.
Rivera
,
A.
Navarro
,
E.
Munoz
,
R.
Czernecki
,
S.
Grzanka
, and
M.
Leszczynski
, “
Optimization of InGaN–GaN MQW photodetector structures for high-responsivity performance
,”
IEEE J. Quantum Electron.
45
(
6
),
617
622
(
2009
).
18.
C. H.
Kang
,
G.
Liu
,
C.
Lee
,
O.
Alkhazragi
,
J. M.
Wagstaff
,
K. H.
Li
,
F.
Alhawaj
,
T. K.
Ng
,
J. S.
Speck
,
S.
Nakamura
,
S. P.
Denbaars
, and
B. S.
Ooi
, “
Semipolar ( 20 2 ¯ 1 ¯) InGaN/GaN micro-photodetector for gigabit-per-second visible light communication
,”
Appl. Phys. Express
13
(
1
),
014001
(
2020
).
19.
J. C.
Carrano
,
T.
Li
,
P. A.
Grudowski
,
C. J.
Eiting
,
R. D.
Dupuis
, and
J. C.
Campbell
, “
Comprehensive characterization of metal-semiconductor-metal ultraviolet photodetectors fabricated on single-crystal GaN
,”
J. Appl. Phys.
83
(
11
),
6148
6160
(
1998
).
20.
J. C.
Carrano
,
T.
Li
,
C. J.
Eiting
,
R. D.
Dupuis
, and
J. C.
Campbell
, “
Very high-speed ultraviolet photodetectors fabricated on GaN
,”
J. Electron. Mater.
28
(
3
),
325
333
(
1999
).
21.
K. J.
Singh
,
Y. M.
Huang
,
T.
Ahmed
,
A. C.
Liu
,
S. W. H.
Chen
,
F. J.
Liou
,
T.
Wu
,
C. C.
Lin
,
C. W.
Chow
,
G. R.
Lin
, and
H. C.
Kuo
, “
Micro-LED as a promising candidate for high-speed visible light communication
,”
Appl. Sci.
10
(
20
),
7384
(
2020
).
22.
R. X. G.
Ferreira
,
E.
Xie
,
J. J. D.
McKendry
,
S.
Rajbhandari
,
H.
Chun
,
G.
Faulkner
,
S.
Watson
,
A. E.
Kelly
,
E.
Gu
,
R. V.
Penty
,
I. H.
White
,
D. C.
O'Brien
, and
M. D.
Dawson
, “
High bandwidth GaN-based micro-LEDs for multi-Gb/s visible light communications
,”
IEEE Photonics Technol. Lett.
28
(
19
),
2023
2026
(
2016
).
23.
K.-T.
Ho
,
R.
Chen
,
G.
Liu
,
C.
Shen
,
J.
Holguin-Lerma
,
A. A.
Al-Saggaf
,
T. K.
Ng
,
M.-S.
Alouini
,
J.-H.
He
, and
B. S.
Ooi
, “
32 Gigabit-per-second visible light communication link with InGaN/GaN MQW micro-photodetector
,”
Opt. Express
26
(
3
),
3037
(
2018
).
24.
P.
Tian
,
J. J. D.
McKendry
,
J.
Herrnsdorf
,
S.
Zhu
,
E.
Gu
,
N.
Laurand
, and
M. D.
Dawson
,
Micro-LED Based Optical Wireless Communications Systems
, 1st ed. (
Elsevier, Inc
.,
2021
).
25.
Q.
Chen
,
J. W.
Yang
,
A.
Osinsky
,
S.
Gangopadhyay
,
B.
Lim
,
M. Z.
Anwar
,
M.
Asif Khan
,
D.
Kuksenkov
, and
H.
Temkin
, “
Schottky barrier detectors on GaN for visible–blind ultraviolet detection
,”
Appl. Phys. Lett.
70
(
17
),
2277
2279
(
1997
).
26.
V.
Adivarahan
,
G.
Simin
,
J. W.
Yang
,
A.
Lunev
,
M. A.
Khan
,
N.
Pala
,
M.
Shur
, and
R.
Gaska
, “
SiO2-passivated lateral-geometry GaN transparent Schottky-barrier detectors
,”
Appl. Phys. Lett.
77
(
6
),
863
865
(
2000
).
27.
H. Y.
Liu
,
W. C.
Hsu
,
B. Y.
Chou
, and
Y. H.
Wang
, “
A simple passivation technique for AlGaN/GaN ultraviolet Schottky barrier photodetector
,”
IEEE Photonics Technol. Lett.
26
(
2
),
138
141
(
2014
).
28.
M. S.
Wong
,
D.
Hwang
,
A. I.
Alhassan
,
C.
Lee
,
R.
Ley
,
S.
Nakamura
, and
S. P.
DenBaars
, “
High efficiency of III-nitride micro-light-emitting diodes by sidewall passivation using atomic layer deposition
,”
Opt. Express
26
(
16
),
21324
(
2018
).
29.
Y.
Choi
,
T.
Kim
,
H.
Lee
,
J.
Park
,
J.
Park
,
D.
Ryu
, and
W.
Jeon
, “
Bottom-up plasma-enhanced atomic layer deposition of SiO2 by utilizing growth inhibition using NH3 plasma pre-treatment for seamless gap-fill process
,”
Sci. Rep.
12
(
1
),
15756
(
2022
).
30.
V. S.
Voitsenya
,
S.
Masuzaki
,
O.
Motojima
, and
J. W.
Davis
, “
Plasma cleaning of oxides from surfaces: The state of the art
,”
AIP Conf. Proc.
1282
,
96
102
(
2010
).
31.
S. C.
Liu
,
G. M.
Dai
, and
E. Y.
Chang
, “
Improved reliability of GaN HEMTs using N2 plasma surface treatment
,” in
Proceedings of the International Symposium on the Physical and Failure Analysis of Integrated Circuits
(
2015
).
32.
Y. C.
Shen
,
G. O.
Mueller
,
S.
Watanabe
,
N. F.
Gardner
,
A.
Munkholm
, and
M. R.
Krames
, “
Auger recombination in InGaN measured by photoluminescence
,”
Appl. Phys. Lett.
91
(
14
),
141101
(
2007
).
33.
M.-H.
Kim
,
M. F.
Schubert
,
Q.
Dai
,
J. K.
Kim
,
E. F.
Schubert
,
J.
Piprek
, and
Y.
Park
, “
Origin of efficiency droop in GaN-based light-emitting diodes
,”
Appl. Phys. Lett.
91
(
18
),
183507
(
2007
).
34.
A. R.
Vasconcellos
,
R.
Luzzi
,
C. G.
Rodrigues
, and
V. N.
Freire
, “
Hot-phonon bottleneck in the photoinjected plasma in GaN
,”
Appl. Phys. Lett.
82
(
15
),
2455
2457
(
2003
).
35.
T.
Park
,
Y.
Wang
,
O.
Alkhazragi
,
J.
Min
,
T. K.
Ng
,
O.
Bakr
,
O.
Mohammed
, and
B. S.
Ooi
, “
Dual-wavelength visible light communication through perovskite-integrated InGaN micro-LEDs
,” in
IEEE Photonics Conference (IPC)
(
IEEE
,
2023
).
36.
J. I.
Pankove
and
D. A.
Kiewit
, “
Optical processes in semiconductors
,”
J. Electrochem. Soc.
119
(
5
),
156C
(
1972
).
37.
M.
Haemmer
,
B.
Roycroft
,
M.
Akhter
,
D. V.
Dinh
,
Z.
Quan
,
J.
Zhao
,
P. J.
Parbrook
, and
B.
Corbett
, “
Size-dependent bandwidth of semipolar ( 11 2 ¯ 2) light-emitting-diodes
,”
IEEE Photonics Technol. Lett.
30
(
5
),
439
442
(
2018
).
Published open access through an agreement withKing Abdullah University of Science and Technology

Supplementary Material