1. Introduction
The advancement of high-power photodiodes operating in the gigahertz frequency range has enabled the creation of fiber-based systems of long-haul transmission of analog microwave signals that provide the advantages of low signal loss, immunity to electromagnetic interference and light weight[1–5]. Various designs of high-power photodiodes have been considered so far. Achieving optimal photodiode parameters requires optimization of various design key points, including charge carrier transit time, structure capacitance and conductivity, and heat removal, which frequently implies a compromise in parameters. Most of the designs of the high-power photodiodes described so far were based upon the p–i–n semiconductor structures. A Schottky barrier photodiode design employing a back-illumination scheme provides several advantages over the p–i–n-based structures, which include a lower electrical resistance of the photodiode leads and a larger thermal conductivity of two massive bulk leads as well as the possibility to use the optical reflection of the non-alloyed Schottky metal contact to increase the total absorption efficiency in the absorbing layer of the structure[6]. We have recently demonstrated the Schottky barrier photodiodes in high-power operation[7]. In this paper we describe the design, manufacturing and characterization of high-power Schottky barrier back-illuminated mesa structure photodiodes based on InAlAs/InGaAs heterostructures, and discuss the advantages of this type of microwave photodiodes for the 1.2–1.6 μm wavelength range. The operation of 10 and 15 μm diameter photodiodes in the frequency ranges of ≥40 and 28 GHz, respectively, the output power of up to 58 mW at 20 GHz, the responsivity of up to 1.08 A/W, and a photodiode dark current as low as 0.04 and 0.1 nA for 10 and 15 μm diameter diodes, respectively, were achieved. A study of the amplitude-to-phase (AM-PM) noise conversion characteristics of the manufactured photodiodes showed that the Schottky barrier photodiodes provide an advantage in the amplitude-to-phase conversion factor. This makes the Schottky barrier photodiodes a promising candidate for use in high-speed analog transmission lines with stringent requirements for the phase noise such as optoelectronic oscillators, analog optic transmission links and distributed antenna arrays.
2. Photodiode fabrication
The heterostructures for photodiode manufacturing were produced by molecular beam epitaxy using a Riber Compact 21T system. The growth was performed on substrates of semi-insulating InP (001) which provided low free-carrier light absorption for the back-illuminated photodiode design. One of the most important stages in the growth sequence was the substrate cleaning procedure prior to the layer growth. In order to initiate the growth on an atomically clean surface, the substrate was annealed at temperatures up to 540 °C in an arsenic flux with an equivalent beam pressure of 2.5 × 10–5 Torr. The preparation of the substrate surface was monitored using a reflection high-energy electron diffraction (RHEED) system. The anneal was carried out until the appearance of surface reconstructions indicating that a clean surface has been reached. Fig. 1 displays RHEED images showing the creation of the (4 × 2) surface reconstruction which indicates the complete removal of residual oxides and the formation of a clean substrate surface suitable for the heterostructure growth. The details of surface transformations during the substrate preparation were discussed in Ref. [8].

The heterostructures used for the photodiode fabrication utilized a phosphorus-free InAlAs/InGaAs layer design which reduced the complexity of the growth procedure and system support in comparison with P-containing structures. The design of the heterostructure is given in Table 1 and the photodiode energy band diagram is shown schematically in Fig. 2. The structure consisted of a light absorbing layer which in turn comprised two parts, namely an undoped part and a thin n+-doped part, an n-side strongly doped contact layer below it, and a layer sequence used for the formation of the Schottky barrier grown above the absorbing layer. The thin n+-InGaAs absorbing layer and the graded bandgap InGaAlAs layer helped to eliminate the accumulation of charge carriers at the InAlAs/InGaAs heterointerfaces, which could lead to a significant increase of the carrier transfer time and, therefore, to a limitation of the photodiode bandwidth. The top protective InGaAs layer prevented the InAlAs surface from oxidation before the fabrication of the Schottky barrier. The InAlAs and InGaAs layers were grown at temperatures of 520 and 480 °C, respectively. The xInAs contents in the layers were calibrated using the photoluminescence and optical absorption techniques prior to the growth of each heterostructure. The quality of the grown single layers and heterostructures was evaluated using the X-ray diffraction[9], atomic force microscopy (AFM), and scanning electron microscopy (SEM). A deviation of the xInAs from the lattice-matched values of xInAs = 0.53 for InxGa1–xAs and xInAs = 0.52 for InxAl1–xAs alloys of not more than 0.2% was generally accepted throughout this study. The peak-to-peak roughness of the surface of the multilayer heterostructures, as observed by the AFM, did not exceed 0.8 nm which corresponded closely to the r.m.s. value of 0.113 nm that was obtained on single-layer samples previously[9]. A SEM picture of a cleaved as-grown heterostructure shows distinct layer boundaries (Fig. 3) which confirms the achievement of the coherent two-dimensional growth mode required for the synthesis of high-quality heterostructures.
Layer composition | Layer thickness | Layer description |
In0.53Ga0.47As | 30 nm | Protective layer |
In0.52Al0.48As | 30 nm | Barrier layer |
InGaAlAs | 50 nm | Graded bandgap layer |
In0.53Ga0.47As | 500–1200 nm | Undoped absorbing layer |
In0.53Ga0.47As:Si | 50 nm | Absorbing n+-layer (5 × 1018 cm−3) |
In0.52Al0.48As:Si | 300 nm | Contact n+-layer (5 × 1018 cm−3) |
InP(001) | > 350 μm | Substrate |

The photodiodes were fabricated by forming mesa structures by a two-step chemical etching of the heterostructures in an H3PO4 : H2O2 : H2O = 1 : 1 : 38 solution. The mesas with diameters ranging from 10 to 40 μm were manufactured. The ohmic contact to the n+-InAlAs layer was fabricated by depositing layers of Ge/Au/Ni/Au (20/40/20/200 nm) followed by an annealing at 385 °C in a hydrogen atmosphere. The Schottky barrier was formed by the deposition of Ti/Au (20/200 nm) on the InAlAs surface of the mesa structure. The use of a large bandgap InAlAs layer allowed one to increase the barrier height as compared to metal-InGaAs Schottky barrier structures. Prior to the barrier formation the InGaAs protective layer and native oxide of InGaAs protective layer were removed in succinic acid : peroxide = 5 : 1 and HCl : H2O = 1 : 10 solutions, respectively. The mesa surface was passivated with an 800 nm SiO2 layer synthesized by the low-pressure chemical vapor deposition technique at 250 °C. Contact leads of 4.5–5 μm thickness were formed by the electrochemical deposition of gold. The thick leads provided both a low electrical resistance and a high thermal conductivity to facilitate the heat removal from the diode. The mechanical strength of the structure was maintained by a polymer film. The substrate was thinned by mechanical grinding and chemical-mechanical polishing, after which an anti-reflection coating layer was deposited to minimize the light reflection. The wafer was then separated into individual back-illuminated diode chips. A photodiode chip is illustrated in Fig. 4.

3. Photodiode characterization
The DC parameters of the photodiodes were evaluated using an Agilent B1500A semiconductor device parameter analyzer, a Cascade M150 probe station, and a single mode fiber-coupled 1.55 μm laser used for responsivity measurements under continuous-wave illumination. A double diffraction grating spectrometer was employed to study the optical reflection spectra of Schottky barriers. Microwave characterization was performed using an experimental setup build around a distributed feedback laser diode with an erbium doped fiber amplifier, a 40-GHz bandwidth Mach-Zender electrooptic light intensity modulator, an Anritsu MS4645B vector network analyzer and an Anritsu MA24126A power meter. The photodiode chips were flip-chip mounted to carrier boards with 50 Ω grounded coplanar waveguide lines and end-launch K connectors.
The DC current–voltage dependencies of the diodes were typical of Au/Ti-InAlAs Schottky diodes and displayed a barrier height of 0.7 eV and an ideality factor of 1.2[10]. The average values of the dark current taken at a 3 V bias on bunches of photodiodes of 20 pcs each amounted to 0.04, 0.1 and 0.3 nA for diode diameters of 10, 15 and 40 μm, respectively. The responsivity of the photodiodes depends on the absorbing layer thickness and is also influenced by the amount of incident light that had passed through the absorbing layer and is reflected backwards by the Schottky contact that acts as a mirror. In order to assess the amount of the contact reflection, we evaluated the reflection coefficient RS of the Schottky contact. The RS was determined from the spectrum of the reflection coefficient R(λ) of a specially prepared large-square sample that was not separated into diode mesas. The substrate of this sample was not thinned and had no antireflection coating. This resulted in the development of interference fringes in the spectral range below the InGaAs absorbing layer bandgap Eg = 0.74 eV due to the light reflections between the semiconductor–air interface and the Schottky metal mirror. The interference period in the spectrum was approximately equal to 1 nm. The R(λ) spectrum taken at 45° angle of incidence is shown in Fig. 4. The spectral resolution of the spectrometer was set so as to average the interference fringes during this measurement. The spectrum comprises two parts. At photon energies hν above the absorbing layer bandgap the reflection coefficient is determined mostly by the reflection at the substrate–air interface. In this case, the reflection is dominated by the component of light polarized perpendicularly to the plane of incidence and R(λ) reaches 38% for hν above 0.85 eV. At photon energies hν < Eg the reflection by the metallic mirror of the Schottky barrier sets in, resulting in an increase of the total reflection coefficient to 70%. In this spectral range the structure was modeled as a transparent slab of InP with a mirror on the surface opposite to the air contact, while the InGaAs/InP boundary was disregarded in the calculation. Modifying the relationship for the reflection coefficient of a similar layered structure[11] and averaging the interference fringes in the spectrum, we infer for hν < Eg the reflection coefficient of the Schottky barrier of RS = 65% with an estimated uncertainty of about ±5%. In the following we assume that the spectral dependence of the found RS
The dependence of the photodiode responsivity at a 1.55 μm wavelength on the absorbing layer thickness is shown in Fig. 5. The photodiodes were illuminated from the substrate side using a cleaved fiber positioned opposite to the mesas. The measurement was performed on photodiodes with the largest mesa diameters in order to minimize the loss of light. The responsivity reached 1.08 A/W for an absorbing layer thickness of 1200 nm. The theoretical model values for the responsivity calculated according to Ref. [12] with the reflection of the Schottky barrier taken into account and assuming an absorption coefficient of α = 7050 cm–1 (see Ref. [13]) and a negligible reflection at the coated chip surface (the solid curve) approximate the experimental data well.
Although a larger thickness of the absorbing layer results in a larger responsivity of a photodiode, it also leads to an increase in the charge carrier transit time and therefore limits the photodiode bandwidth[14]. For this reason the measurements of the photodiode frequency response were performed on photodiodes with the smaller thickness of the absorbing layer. Fig. 6 depicts the frequency responses of photodiodes with mesa diameters of 10 and 15 μm that have low capacitances and display the largest bandwidths. Here, the drops on the low-frequency side of the response curves are caused by the frequency responses of the high-pass bias circuits on the chip carrier boards. The ‘wavy’ behavior of the curves is induced by two unfavorable side effects: the incomplete matching of the impedances of the photodiode and transmission line, and non-ideality of the light intensity modulator calibration. It is seen from the figure that the frequency range of the 15 μm diameter photodiodes is limited to 28–30 GHz while that of the 10 μm photodiodes extends over 40 GHz. The limitation of the frequency range of the 15 μm photodiodes is attributed to the larger RC time constant value of the circuit (here R = 50 Ω being the load impedance) due to the larger diode capacitance. The largest output microwave power was achieved from photodiodes with the 15 μm diameter. Fig. 7 shows the dependencies of the microwave power output of a 15 μm photodiode on the input modulated optical power taken at a modulation frequency of 20 GHz for various applied bias voltages. The maximum output power, which caused thermal failure of the devices, reached as much as 58 mW at a bias voltage of 3 V. The output power value attained compares well with the data on the partially depleted absorber (PDA) and uni-traveling-carrier (UTC) photodiodes of comparable sizes in this frequency range (see e.g. Refs. [2, 15]).
Another important characteristic of a high-power photodiode that can limit its application to microwave measurement and communication systems, phased antenna arrays and optoelectronic oscillators is the phase noise induced by the photodiode due to a variation of the output signal phase with a change in the input laser intensity. As a result of the dependence of the output signal phase on the input power level due to nonlinearities of the photodiode, the laser intensity noise is converted to an additional phase noise at its output[16–18]. The amplitude to phase noise conversion can be characterized by the conversion coefficient KPD = ΔφPD / ΔPopt [rad/W], where ΔPopt is the input light power change and ΔφPD is the output phase change. To evaluate the KPD
4. Conclusion
We described the design, manufacturing and characterization of high-power Schottky barrier InAlAs/InGaAs back-illuminated mesa structure photodiodes. These photodiodes of a 10 and 15 μm diameter operate in the frequency ranges of ≥40 and 28 GHz, respectively, and demonstrate the output RF power as high as 58 mW at a frequency of 20 GHz, the DC responsivity of up to 1.08 A/W, and a photodiode dark current as low as 0.04 nA. An AM-PM conversion coefficient of 1.5 rad/W has been measured, which makes the Schottky barrier photodiodes suitable for use in high-speed analog fiber-optic transmission links and optoelectronic oscillators with stringent requirements for phase noise.
Acknowledgements
The preparation and processing of heterostructures used in this work was supported by the Russian Science Foundation (grant number 19-72-30023).