High-Responsivity Waveguide UTC Photodetector with 90 GHz Bandwidth for High-Speed Optical Communication

Abstract

A directly coupled waveguide uni-traveling carrier photodetector (UTC-PD) with high responsivity and broad bandwidth is demonstrated. The device’s epitaxial structure was carefully optimized via optical simulations to enhance quantum efficiency. Furthermore, the fabrication process was refined to introduce a vertically defined mushroom-shaped mesa structure, which effectively maintains high responsivity while facilitating further improvement in bandwidth performance. As a result, the fabricated device, without the use of an anti-reflection coating, simultaneously achieves a responsivity of 0.49 A/W and a 3 dB bandwidth of 90 GHz.

1. Introduction

With the rapid advancement of information technologies in modern society and the development of emerging internet technologies—such as artificial intelligence (AI), virtual reality (VR), three-dimensional (3D) media, and the Internet of Everything (IoE)—the volume of network communication has experienced an exponential increase [1]. To enhance network communication efficiency, it is essential to reduce the complexity and cost of transmitters and receivers within the existing fiber optic transmission framework, while also improving the transmission rate of individual carriers in the channel [2]. Concurrently, emerging technologies, such as free-space communication, have been proposed, with modulated optical carriers serving as a key feature for transmission in atmospheric channels. Additionally, there is an increasing demand for photodetectors with enhanced frequency response, larger dynamic range, and improved linearity at the receiver. Therefore, research on high-power, high-bandwidth, and high-responsivity photodetectors is of critical importance [3,4,5].

Compared to conventional PIN-type photodetectors, InGaAs photodetectors with a uni-traveling carrier (UTC) structure employ a p-type absorption layer, allowing the response time of photogenerated holes to fall within the dielectric relaxation time range (on the order of femtoseconds) [6]. As a result, only the electron transit time needs to be considered in the carrier transport process, enabling the realization of high-bandwidth performance [7]. Subsequently, researchers proposed an improved UTC photodetector structure, whose most distinctive feature is the introduction of a highly doped “cliff” layer between the absorption and collection layers. This modification enhances the device’s saturation current, thereby enabling higher output power [8,9,10,11]. However, achieving both high-speed operation and high responsivity in UTC photodetectors remains a critical design challenge, particularly in terms of improving the overall photoelectric conversion efficiency. Among the contributing factors, the optical coupling method plays a crucial role in device performance. Early research focused primarily on surface-illuminated structures. For example, Li W., Beling A., Campbell J. et al. [12] investigated a surface-illuminated UTC photodetector incorporating a cliff layer, achieving a bandwidth of 12.8 GHz and a responsivity of 0.36 A/W. Nevertheless, such structures suffer from a fundamental trade-off between bandwidth and responsivity, often referred to as the “bandwidth-efficiency limitation”, making it difficult to simultaneously attain high values in both metrics. As a result, waveguide photodetectors have been introduced, offering the advantage of decoupling the directions of carrier transport and optical propagation, thereby overcoming the aforementioned theoretical limitations. In practical applications, waveguide-based photodetectors—despite their superior bandwidth performance—often suffer from low responsivity due to optical facet coupling losses. For instance, Qinglong Li et al. from Joe Campbell’s group [13] successfully demonstrated a high-power MUTC photodetector with a bandwidth exceeding 105 GHz using an evanescent-wave coupling scheme. The measured responsivity was reported in the range of 0.1 to 0.16 A/W. Similarly, Bassem Tossoun et al. [14] developed a modified UTC photodetector with improved capacitive characteristics, achieving a 3 dB bandwidth of 85 GHz and a responsivity of 0.26 A/W. Marcel Grzeslo et al. [15] designed high saturation waveguide-type MUTC-photodiodes, with a responsivity of 0.25 A/W and a 3 dB bandwidth lower than 100 GHz.

In response to these limitations, we propose a directly coupled UTC photodetector. This design enables efficient coupling with optical fibers, significantly reducing insertion loss. It presents a promising solution for achieving both high-speed operation and high responsivity. By employing direct coupling techniques, the UTC photodetector’s optical responsivity is substantially enhanced. Additionally, the active layer structure and doping profile are carefully optimized based on the UTC photodetector’s bandwidth characteristics. The fabricated waveguide photodetectors demonstrate strong potential for use in high-speed optical communication systems.

2. Device Structure Design

The epitaxial structure of the directly coupled waveguide photodetector was optimized in this work. As shown in

Table 1. Epitaxial structure of the waveguide UTC-PD.

Composition	Thickness [nm]	Doping [cm−3]	Function
In0.53Ga0.47As	80	P-2.0 × 1018	P-contact
InP	200	P-1.0 × 1018	Cladding
InGaAsP (Q1.3)	900	P-3.0 × 1018	Cladding
In0.53Ga0.47As	25	P-2.5 × 1018	Absorber
In0.53Ga0.47As	25	P-1.0 × 1018	Absorber
In0.53Ga0.47As	25	P-5.0 × 1017	Absorber
In0.53Ga0.47As	75	u.i.d	Absorber
InGaAsP (Q1.3)	20	N-5.0 × 1017	Cliff
InGaAsP (Q1.3)	200	N-1.0 × 1016	Collector
InP	100	N-2.0 × 1018	Collector
InGaAsP (Q1.3)	580	N-2.0 × 1018	N-contact
InP	500	u.i.d	Buffer
InP	300,000		Substrate

, the epitaxial layers were grown on a semi-insulating InP substrate using metal–organic chemical vapor deposition (MOCVD). The structure includes a 75 nm graded p-type doped layer designed to facilitate electric field-assisted electron acceleration in the absorption region, along with a 75 nm undoped layer that ensures carrier depletion in the same region. A moderately thick p-doped InGaAsP layer was incorporated to improve coupling efficiency with a 3 μm tapered optical fiber while maintaining acceptable electrical performance. Although increasing the thickness of this layer enhances optical coupling, it also leads to higher series resistance. Conversely, reducing the thickness may decrease resistance but at the cost of lower coupling efficiency. Therefore, the thickness of this layer must be carefully optimized to balance optical and electrical performance in the overall device design [16]. In addition, a 20 nm moderately n-doped InGaAsP layer (doping concentration: 1.0 × 10¹⁷ cm⁻³) was inserted between the carrier collection layer and the In_0.53Ga_0.47As absorption layer to enhance the electric field strength within the depletion region and promote electron thermionic emission.

The overall bandwidth is jointly limited by carrier transit-time bandwidth ($f_T$) and the RC-limited bandwidth ($f_{\mathrm{R}\mathrm{C}}$), as described by

$${\displaystyle \frac{1}{{f^2}_{3dB}}}={\displaystyle \frac{1}{{f^2}_T}}+{\displaystyle \frac{1}{{f^2}_{RC}}}$$

(1)

In the UTC-PD, the carrier transit-time bandwidth ($f_T$) is strongly influenced by the electron transport velocity within the absorption layer. As the initial electron velocity is positively correlated with the strength of the electric field in the depletion region, an enhanced field—enabled by the proposed epitaxial structure—can effectively improve the electron transport velocity. Consequently, the optimized epitaxial design contributes to an improved overall bandwidth performance of the device. The RC-limited bandwidth is governed by the device’s total resistance and capacitance, both of which must be minimized to enhance overall bandwidth performance.

On the other hand, the RC-limited bandwidth is determined by the total resistance and capacitance of the device, both of which must be minimized to enhance high-frequency performance. Beyond reducing parasitic parameters, our design focuses on the impact of the active region size, as a smaller active area reduces device capacitance and thus increases bandwidth. However, in conventional waveguide photodetectors, overly shrinking the active region diminishes the top metal contact area, leading to higher series resistance and degrading the RC-limited bandwidth. To address this trade-off, a mushroom-shaped structure was introduced. This architecture maintains a large top metal contact area while employing selective etching to reduce the underlying active region size, thereby lowering capacitance and enhancing RC bandwidth. As a result, the overall speed performance of the device is significantly improved.

To implement a mushroom-shaped photodetector structure by precisely controlling the active area, two additional InP layers were incorporated into the aforementioned epitaxial design. Finite-difference time-domain (FDTD) simulation was then conducted to evaluate the optical impact of this structural configuration, particularly its influence on quantum efficiency (QE). The relationship between quantum efficiency ($\eta$) and optical responsivity (R) is described by the following [17]:

$$\eta =\left(1-R\right)\times \left(1-e^{-\alpha \Gamma L}\right)\times {\eta }_C$$

(2)

Here, R represents the power reflection loss, α is the absorption coefficient of InGaAs, Γ denotes the optical confinement factor of the InGaAs layer, L is the length of the photodetector, and ${ \eta }_C$ is the coupling efficiency with the optical fiber. Among these parameters, the device length L and coupling efficiency are key factors considered in the simulation-based structural design.

Based on the aforementioned theoretical framework, finite-difference time-domain (FDTD) simulations were performed to optimize the device structure, as illustrated in Figure 1d. A Gaussian beam with a beam waist diameter of 3 μm was employed as the input source, and the operating wavelength was set to 1500 nm. The corresponding simulation results are shown in Figure 1a, where the influence of the p-InGaAsP layer thickness in the upper cladding on quantum efficiency was analyzed. The results indicate that when the p-InGaAsP thickness approaches 900 nm, the improvement in quantum efficiency begins to plateau. Further increases in thickness yield only marginal gains in optical coupling efficiency while potentially introducing drawbacks such as increased series resistance. Similarly, Figure 1b illustrates the impact of the two additional InP layers on quantum efficiency. With the p-InGaAsP layer thickness fixed at 900 nm, increasing the p-InP layer thickness beyond 200 nm exhibits a similar trend, whereas a thicker n-InP layer in the lower cladding leads to a noticeable degradation in quantum efficiency. Taking practical fabrication feasibility into account, an optimized configuration was selected, consisting of a 900 nm p-InGaAsP, a 200 nm p-InP upper cladding and a 100 nm n-InP lower cladding. This design achieved high responsivity in the simulation results.

Additionally, simulations exploring the influence of photodetector width and size on quantum efficiency are shown in Figure 1c. When the PD length is less than 10 μm, quantum efficiency significantly decreases, which adversely affects the optical responsivity of the device. Similarly, if the PD width is smaller than the diameter of the testing optical fiber, facet coupling mismatches may occur, leading to reduced optical coupling efficiency. Therefore, five different PD designs with sizes of 5 × 20 μm², 5 × 15 μm², 4 × 20 μm², and 4 × 15 μm² were selected for chip-level experiments.

3. Device Fabrication and Characterization

The waveguide UTC-PD based on the proposed structure has been successfully fabricated. The device adopts a double-rectangle mesa structure, which was defined using inductively coupled plasma (ICP) dry etching. After that, wet etching was employed to further define the active region and simplify the photolithography process. Under dry etching conditions, the preliminary etching of the P-mesa was achieved by employing a CH₄/Ar/Cl₂ gas mixture, coupled with precise control of the etching time. The wet etching technique was initially introduced to mitigate potential surface leakage current issues caused by the dry etching process. Due to the inherent selectivity of wet etching, this approach also enabled further optimization of the active region, facilitating the formation of a mushroom-mesa structure and thereby enhancing the device bandwidth. Benefiting from the previously introduced thin InP layers, the modified epitaxial structure allows for both favorable optical and electrical performance while enabling vertical etch depth control via selective chemical etching. In the lateral direction, a mixed acid solution of H₂SO₄/H₂O₂/H₂O with a specific ratio was employed to promote inward etching of the mesa sidewalls, while an ice-water bath was used to maintain low temperatures for stabilizing the etching rate. A high-speed photodetector featuring a mushroom-shaped mesa structure fabricated through this process is illustrated in Figure 2a. Damage to the upper InGaAs layer during etching can result in increased series resistance or metal delamination. To mitigate this, etching parameters—including time, acid concentration, and temperature—were systematically optimized through controlled experiments. Further parametric analysis showed that mesa width had a negligible effect on lateral etching. A mixed acid solution with a concentration ratio of 3:1:5 was found to provide optimal lateral etch control. Consequently, an etching time of 40 s was selected for the final process. This mushroom-shaped structure enables adequate lateral definition of the active region while preserving the integrity of the thin InGaAs P-contact layer, thereby preventing excessive resistance and avoiding issues such as metal delamination that could compromise the device’s RC-limited bandwidth.

Following mesa definition, P-type and N-type metal contacts were fabricated using electron-beam evaporation and magnetron sputtering, respectively. To ensure high-quality ohmic contact, a combination of surface pretreatment and post-deposition rapid thermal annealing was applied. The entire chip was subsequently passivated with a sufficiently thick silicon dioxide layer. Polyimide (PI) was used to planarize the mesa edges, and a Ti/Au metallization stack was deposited to form the electrode interconnections. The final fabricated waveguide UTC-PD is shown under an optical microscope in Figure 2b and a SEM image in Figure 2c.

Figure 3 presents the dark current characteristics of the UTC-PDs. Current–voltage (I–V) curves were measured for waveguide UTC-PDs of various device sizes. Under standard laboratory conditions with normal illumination and room temperature, the dark current measured at a bias of −2 V ranges in 10⁻⁹ A, demonstrating excellent low leakage current performance. The responsivity measurements were carried out using a tapered fiber with a diameter of 3 μm, which was directly coupled to the photodetector following the dark current test. The responsivity was defined as the ratio of the measured photocurrent to the incident optical power. A maximum responsivity of 0.49 A/W was achieved for the waveguide UTC-PDs, corresponding to an effective area of 80 μm². While a reduction in device size slightly affects the responsivity, the impact remains relatively minor.

To optimize device performance and extend the RC-limited bandwidth, certain photodetectors were specifically designed with integrated RC circuit compatibility. Although this approach may introduce some RF current loss, it effectively broadens the bandwidth and enhances the high-frequency response of the device [18]. As shown in Figure 4, the frequency responses of multi-sized photodetectors (PDs) were characterized using a heterodyne optical measurement system. This system employs a dual-channel tunable laser to generate two optical signals with a controlled frequency offset. The vector superposition of these two signals produces an optical beat frequency that is tunable and serves as the test signal. Three different high-frequency probes were used to detect RF signals, covering spectral ranges of DC–64 GHz (V-band: 50–75 GHz), W-band (75–110 GHz), and D-band (110–170 GHz). Optical coupling was achieved via a tapered optical fiber with a 3 μm mode field diameter (MFD), ensuring accurate light delivery. All RF link losses—including those from high-frequency probes, bias-Ts, and waveguide transitions—were carefully calibrated.

Figure 5 presents the normalized frequency response of mushroom-shaped waveguide photodetectors with different geometric structures. To maintain uniform device current, all measurements were conducted under small-signal conditions below the saturation photocurrent, with a maximum measurement frequency of 220 GHz. Devices were biased at −3 V during testing. Under these conditions, a 3 dB bandwidth of 90 GHz was obtained. The mushroom-shaped waveguide photodetectors fabricated using the proposed process exhibit excellent bandwidth and responsivity, validating the effectiveness of the preceding optical analysis and bandwidth optimization strategies.

Table 2. Performance comparison of reported photodetectors with different structures.

References	Device Type	Responsivity	Bandwidth
[12]	UTC-PD	0.36 A/W	12.8 GHz
[13]	UTC-PD	0.16 A/W	105 GHz
[14]	UTC-PD	0.26 A/W	85 GHz
[15]	UTC-PD	0.25 A/W	100 GHz
[19]	PIN	0.65 A/W	20 GHz
[20]	PIN	0.9 A/W	25 GHz
[21]	WG-UTC	0.75 A/W	14 GHz
[22]	UTC-PD	0.33 A/W	52.2 GHz
[23]	UTC-PD	0.27 A/W	170 GHz
Our work	UTC-PD	0.49 A/W	90 GHz

presents a performance comparison of various InGaAs photodetectors. Compared with conventional PIN photodetector structures, our device exhibits a slightly reduced responsivity but a significantly enhanced bandwidth—an expected and typical characteristic of UTC photodetectors. In comparison with recently reported UTC-based photodetectors, our device also demonstrates competitive responsivity and bandwidth performance. These results indicate that both the device design and the fabrication process were successfully optimized to achieve a desirable balance between high responsivity and high bandwidth. This makes the proposed photodetector a promising candidate for applications in high-speed optical communication and integrated photonic systems.

4. Conclusions

In this study, we propose a directly coupled waveguide uni-traveling carrier photodetector (UTC-PD) that simultaneously achieves high responsivity and high bandwidth. Various optimization strategies are explored, including enhancing optical coupling efficiency through epitaxial structure design to improve quantum efficiency, and implementing a mushroom-shaped device architecture during fabrication to boost bandwidth performance. The fabricated waveguide UTC-PD, without the use of anti-reflection coating, demonstrates a responsivity of 0.49 A/W at a wavelength of 1550 nm and achieves a maximum bandwidth of 90 GHz. These results indicate strong potential for this device in high-speed optical communication applications.