1. Introduction

Semiconductor avalanche photodetectors (APDs) have been widely used as compact alternatives to photomultiplier tubes in modern laser detection and ranging (LiDAR) systems [1]. The 1064 nm has been chosen as an optimal operating wavelength for laser ranging applications mainly due to the consideration that the high power 1064 nm neodymium-doped yttrium aluminum garnet (Nd:YAG) diode-pumped solid-state lasers with high pulse repetition rates and subnanosecond pulse widths have been well developed which are approximately twice as efficient as the frequency doubled 532 nm Nd:YAG lasers and offer a mature choice for the LiDAR emitter [2]. Furthermore, the solar irradiance disturbing for laser detection at 1064 nm is reduced by three times in comparison to that at 532 nm [3]. In regards to APDs for 1064 nm photon detection, potential candidates are the Si, the In_0.53Ga_0.47As (referred to as InGaAs) and the In_xGa_1−xAs_yP_1−y (referred to as InGaAsP). Near-infrared enhanced Si (NIR-Si) APDs can offer responsivities up to 15 A/W at 1064 nm by increasing the device thicknesses and operating at higher temperatures, albeit at the expense of reduced response speed and higher thermal noise equivalent power [4]. InGaAs/InP separated absorption, grading, charge and multiplication (SAGCM) APDs can provide a high quantum efficiency (QE) up to 80% at 1064 nm [5]. However, thermal generation current from the narrow band gap InGaAs absorber (0.73 eV) has contributed a high dark current, which essentially limits their signal noise ratio at high gains [6]. In contrast, InGaAsP/InP SAGCM APDs with an accurately tailored 1.0 eV InGaAsP absorber have been shown to be the most promising 1064 nm detectors that can achieve high QE over 90% and greatly reduced dark current simultaneously [7]. Benefiting from the larger band gap in the InGaAsP, the dominant source for dark current under high reverse biases shifts to the trap-assisted tunneling (TAT) in the InP multiplier compared with the InGaAs/InP APDs, which enables enhanced linear- and Geiger-modes operation performances at 1064 nm [8].

InP has been widely used as a hole-initiated multiplier in InGaAsP APDs [7, 9, 10] with structures similar to the InGaAs/InP SAGCM APD. In_0.52Al_0.48As (referred to as InAlAs) that can offer a higher ratio of electron to hole impact ionization coefficient k_eff (k_eff =α/β) than the ratio of β/α for InP, has been proved to be an excellent electron-initiated multiplication material for InGaAs APDs [11]. Better performances for InGaAs/InAlAs APDs have been demonstrated progressively compared with InGaAs/InP ones, e.g. larger maximum gain factors (M) in excess of 100 [12], lower excess noises factors (F(M)) with typical k_eff of 0.15∼0.25 [13], and higher bandwidths up to 30 GHz at an M of 3.6 [14]. Therefore, InGaAsP APDs with InAlAs multipliers are expected to exhibit performances surpass that with InP. Nevertheless, such APDs have not been demonstrated to date. Challenges mainly lie in the epitaxy of quaternary InGaAsP on ternary InAlAs surfaces on the one hand, where rugged interfaces can easily occur due to dissolution of group V atoms at the beginning of growth [15, 16]; On the other hand, a proper band grading layer needs be designed to ensure efficient transit of carriers from InGaAsP to InAlAs.

Here we demonstrate an electron-initiated 1064 nm InGaAsP SAGCM APD with a p type InAlAs multiplier, a tailored InGaAs/InAlAs digital alloy superlattice (DGSL) grading structure and an InAlAs field-control layer. High unity gain QE, low dark current, a room temperature (RT) maximum M of 5, and a low multiplication noise with an effective k_eff of ∼0.2 are achieved. Factors limiting the maximum M are also discussed.

2. Experimental details

The APD was grown by a VG Semicon-V80 gas-source molecular beam epitaxy system. Figure 1(a) schematically illustrates the APD structure, which in turn consists of an 800 nm InAlAs buffer layer (n contact) heavily doped to n-4×10¹⁸ cm⁻³, a 500 nm InAlAs multiplication layer doped to p-6×10¹⁶ cm⁻³, a 70 nm InAlAs charge sheet layer doped to p-6×10¹⁷ cm⁻³, a 70 nm InGaAs/InAlAs DGSL band grading layer doped to p-1×10¹⁷ cm⁻³, a 1500 nm InGaAsP light absorption layer doped to p-1×10¹⁷ cm⁻³, a 450 nm InAlAs windows layer heavily doped to p-3×10¹⁸ cm⁻³, and a InGaAs p contact layer heavily doped to p-3×10¹⁸ cm⁻³, with silicon and beryllium as the n- and p-type dopants, respectively. P type doping in the multiplication region was chosen to ensure full depletion of the p type absorber under high reverse biases. With a moderate electric-field (E-field) formed inside (<150 kV/cm), the photo-generated electrons can be efficiently injected into the multiplier, ensuring a high unity gain QE. The DGSL grading structure consists of alternated ultrathin multilayers of InAlAs and InGaAs [17], thicknesses of which in each period were gradually varied in opposite directions from 95 and 5 Å to 63 and 37 Å, respectively, for 7 periods of SLs. The finally effective composition of the DGSL alloy was tailored to be In_0.52Al_0.19Ga_0.29As to attain a zero conduction band offset with respect to the InGaAsP absorber. After growth, circular mesas with variable diameters ranging from 20 to 200 μm were delineated by lithography and dry etching. SiN_x passivation and Ti/Pt/Au liftoff metallization techniques were applied to finish the device fabrication. No anti-reflection coating was applied.

 

Fig. 1 (a) Schematic device structure for the InGaAsP/InAlAs SAGCM APD. (b) The simulated 300 K E_C and E_V band edge line-up and the corresponding E-field profile along the device at −16.5 V. Upper panel: enlarged view of the smooth DGSL band grading.

Download Full Size | PDF

Figure 1(b) shows the 300 K conduction (E_C) and valance (E_V) band edge line-up and the corresponding E-field profile along this APD at −16.5 V simulated using Silvaco Atlas with nominal structural parameters as well as default materials parameters [18]. The 0 nm reference point corresponds to a position inside the n⁺ InAlAs contact layer with a distance of 200 nm to the n⁺-p⁻ interface. The E-field is well-confined in the avalanche region while its intensity is nonuniform and a high peak intensity up to 550 kV/cm is generated close to the pn junction interface, originates from the relatively high doping density. The enlarged view clearly shows the well-graded conduction band offset between the InGaAsP absorber and the InAlAs charge layer that ensures smooth injection of photo-generated electrons into the multiplier.

3. Results and discussions

3.1. Material quality validation

The material quality of this APD wafer was validated by measuring its photoluminescence (PL) and x-ray diffraction (XRD) in a selectively step-by-step chemical etching method. Samples were excited by a 633 nm laser at a power density of 0.5 W/cm² in a PL setup connected to a OMNIC IS50 FTIR spectrometer. Figure 2(a) shows the RT PL spectra of the as-grown wafer and the wafer after etching away the InGaAs contact and InAlAs window layers by dipping into a mixed solution of phosphoric acid and hydrogen peroxide, correspond to the PL emission from the p⁺ InGaAs contact and the p⁻ InGaAsP absorber, respectively. The InGaAsP absorber exhibits a stronger PL emission at 1280 nm tenfold that from the InGaAs contact layer, indicating a high crystal quality of the absorber. Figure 2(b) shows the XRD (004) rocking curves for this APD wafer before and after etching, measured using a Philips X’pert MRD diffractometer. The peak at the left shoulder of the InP substrate peak with a small lattice-mismatch with respect to InP substrate down to 3.5×10⁻⁴ remains almost unchanged after etching, and thus is attributed to the InGaAsP absorber. The broader peak at the right shoulder of the InP peak reveals an increased mismatch from −1.35×10⁻³ to −1.56×10⁻³ after etching, which is identified to be from the ternary InGaAs and InAlAs layers. Furthermore, this peak becomes sharper with narrower width after removing the InGaAs contact and InAlAs window layers, suggesting a degraded crystal quality for the ternary layers grown on InGaAsP, which is probably a result of the dissolution of group V atoms at the InAlAs/InGaAsP interface [16].

 

Fig. 2 (a) 300K PL spectra for the p⁺ InGaAs contact and the p⁻ InGaAsP absorber (after etching away the p⁺ InGaAs and InAlAs layers). (b) XRD (004) rocking curves for the APD wafer before and after etching, respectively.

Download Full Size | PDF

3.2. Current-voltage and gain curves

Figure 3 shows the RT photo and dark reverse current-voltage (I–V) curves and the corresponding M for a 20 μm diameter APD. Results for an APD without a grading layer were also shown in Fig. 3 for comparison. The RT dark I–V measurements were carried out in a shielded probe station connected to a HP4156B semiconductor parameter analyzer. For the photocurrent measurements, a 1064 nm semiconductor laser was coupled on the device mesa using a 9 μm diameter single-mode optical fiber with the same tip power of 10 μW. An InGaAsP/InP PIN photodiode with the same absorber thickness of 1.5 μm was also fabricated to serve as a unity gain reference, for which the 1064 nm responsivity at zero bias was calibrated to be 0.41 A/W at 300 K (a QE of 48%) [19]. A noticeable change of slope in the photo I–V curve can be seen at around −10 V for the APD with the grading layer, followed by building up of enough high E-field at −16 V where the onset of unity gain occurs. Further increasing the reverse bias results in the onset of multiplication of photocurrent. The M is calculated by M=(I_p−I_d)/(I_up−I_ud), where I_p and I_d are the photo and dark currents, respectively. I_up and I_ud are the photo and dark currents at unity gain, respectively. The calculated maximum M is around 5 for this APD, as also shown in Fig. 3. The same maximum M were observed for fiber tip powers up to 300 μW, indicating a very good photoresponse linearity on input optical intensity for this APD. The I_d remains below 1 nA before a steep breakdown at −30 V. The I_d at 90% of the breakdown voltage (V_B) is 470 pA, on the same order of magnitude comparing to that of a planar InGaAsP/InP APD with the same size of active area [7].

 

Fig. 3 RT photo and dark reverse I–V curves and the corresponding gain factors for 20 μm diameter APDs with (w) and without (w/o) a grading layer.

Download Full Size | PDF

For SAGCM APDs, the maximum M is a result of multiple factors including the α and β, the E-field intensity and distribution, and the thickness of the multiplier region. The α for this APD is primarily determined by the local E-field intensity E according to the local field theory [20] α = a_e exp(−b_e/E^n_e), where a_e and b_e are constants related to materials and n_e is close to unity. Therefore, increasing the mean E-field intensity in the 500 nm InAlAs multiplier region will directly increases the M. Given all other structural parameters are fixed (e.g. the thicknesses and doping densities for both the absorber and the multiplier), the electric field profile (i.e. the intensity and the distribution) can be modified by simply changing the charge layer thickness and doping density [21]. However, apart from the local E, α is also considerably affected by the electron scattering and recombination processes in the multiplier material. High density doping in the multiplier causes stronger impurity scattering and lowers the saturation mobility, leads to a reduced α. In addition, the electron recombination rate will also be enhanced due to higher density of impurity defect states, further lowers the effective α. The p-type InAlAs doped to 6×10¹⁷ cm⁻³ in this APD was compensated doped using beryllium, which means an essentially higher impurity concentration and thus deteriorated α in comparison to that for undoped InAlAs. Thereby, growth of p-type InAlAs with very low doping density will also help increase the maximum M. Moreover, more intrinsic multiplier also helps lower the peak E-field intensity in the multiplication region and increase the V_B [22]. In addition, another effective route for enhancing the maximum M is reducing the multiplier thickness by utilizing the "dead space" effect for thin multiplier APDs [23], albeit as a trade-off the I_d would also see a considerable rise due to the much increased tunneling leakage under higher E-field intensity.

From Fig. 3, the APD without grading shows a much lower unity gain responsivity down to 0.05A/W as well as a lower maximum M of ∼2. The reason behind such performance degradation is the much reduced probability for the photo-generated electrons to transit across the large band offset at the InGaAsP/InAlAs interface (∼460 meV). The minority carriers would be pinned at the interface and recombine there, leading to suppressed I_p and gain. The RT I_d at 90% V_B is 220 pA, lower that for the APD with grading, which is due to similar blocking effect for dark carriers from the absorber region. Furthermore, the V_B decreased by 1.6 V for the APD without grading, indicating that inserting of the grading layer also helps delay the onset of breakdown and increase the gain.

3.3. Spectral responsivities and gain

Bias- and temperature-dependent spectral responsivities were acquired by using a FTIR setup [24]. The APD was TO packed and mounted on the cold finger of a liquid nitrogen cryostat which controlled the temperature from 77–300 K. Reverse DC biases were applied through a bias tee network while the AC photo response signals were fed into a preamplifier connected to a OMNIC IS50 FITR spectrometer. Figure 4(a) shows the measured spectral responsivities biased from −18 V to −28.9 V at 300 K corrected using the FTIR instrument transmission function. The spectrum for the zero-biased PIN detector at 300 K is also shown for comparison. The absolute spectral responsivities for both the APD and the PIN are scaled according to the 300 K responsivity at 1064 nm for this PIN (0.41 A/W). The spectral responsivity monotonically increases with increasing the reverse bias, with a maximum 1064 nm responsivity of 1.8 A/W, corresponds to a maximum M of 4.4. Further increase the reverse bias results in a saturation behavior due to the significantly increased I_d near breakdown. Temperature-dependent spectral responsivities at a fixed reverse bias of −25 V are shown in Fig. 4(b). Enhanced 1064 nm responsivities from 0.7 to 1.6 A/W as the temperature decreases from 300 to 77 K are observed, which essentially originates from the suppressed electron-phonon scattering at lower temperatures and thus less kinetic energy loss of electrons in the multiplier [22].

 

Fig. 4 (a) Bias-dependent spectral responsivities at 300 K and (b) temperature-dependent spectral responsivities at −25 V for a 200 μm diameter APD. The spectrum for the PIN detector at zero bias is also shown as the unity gain reference.

Download Full Size | PDF

3.4. Temperature-dependent I–V characteristics

Temperature-dependent reverse bias I–V characteristics were measured in a Lakeshore TTPX cryostat probe station, which controlled the temperature from 77 to 300 K and was connected to a HP4156B semiconductor parameter analyzer. Figure 5(a) shows the temperature-dependent reverse bias I_d-V and the 77 K I_p-V curves for a 20 μm diameter APD. The I_d remain below 2 pA and almost unchanged for T ≤200 K, while for T>200 K a noticeable climb of I_d up to 200 pA at V>−20 V was observed. Such an increase of I_d originates from the thermal excitation of trapped carriers from defect energy levels in the InAlAs avalanche region, which is further verified by the activation energy (E_a) analysis. The Arrhenius plot of I_d at −20 V as a function of inverse temperature is shown in Fig. 5(b). I_d exhibits a linear dependence on 1/T at T>200 K and was fitted using the E_a model formulated by I_d∝exp(−E_a/kT), where k is the Boltzmann constant, T is the temperature and E_a is the activation energy of I_d. The extracted E_a is 0.27 eV, consistent with that measured in InGaAs/InAlAs APDs [22,25], indicating the dominant generation source of I_d is the TAT current in the InAlAs multiplier. The avalanche multiplication of dark carriers would also contribute to such climb of I_d but only plays a minor role. The V_B monotonically increases with increasing temperature with a positive temperature coefficient of 4.2 mV/K, as shown in Fig. 5(c), which is a signature of the occurrence of avalanche breakdown [26].

 

Fig. 5 Temperature-dependent reverse bias I_d-V and the 77 K I_p-V curves for a 20 μm mesa. (b) Arrhenius plot of I_d at a reverse bias of −20 V. The fitted E_a at T>200 K is indicated. (c) The V_B as a function of temperature. The line is a linear fitting.

Download Full Size | PDF

3.5. Excess noise factor

The F(M) is characterized as a function of gain by using a SRS-SR785 network signal analyzer and a Nd:YAG 1060 nm continuous laser, as shown in Fig. 6. The F(M) of an electron-initiated InGaAs/InAlAs [27] and a hole-initiated InGaAs/InP [28] SAGCM APDs with approximately the same multiplication region width of 500 nm are shown for comparison. Theoretical curves calculated using local-field noise theory [29] with k_eff from 0 to 0.5 are also plotted for reference. The multiplication noise of commercial InGaAs/InP APDs is characterized by a value of k_eff ∼0.5, while InGaAs/InAlAs APDs typically exhibit k_eff between 0.1–0.3 [14,27,30]. Thinner multiplier thickness tends to decrease the k_eff benefiting from the dead-space effect [27]. From Fig. 6, it appears that this InGaAsP/InAlAs SAGCM APD exhibits a low F(M) with a k_eff of ∼0.2, well consistent with the typical k_eff reported in InGaAs/InAlAs APDs. Such consistency indicates the advantage of low F(M) for electron-initiated multiplication in InAlAs is well conserved for 1064 nm InGaAsP/InAlAs APDs, and the InGaAsP/InAlAs is more competitive from the prospect of low multiplication noise when compare with InGaAsP/InP.

 

Fig. 6 Measured F(M) versus M curves for this APD. F(M) of InGaAs/InP [28] and InGaAs/InAlAs [27] APDs with the same multiplier width of 500 nm are plotted for comparison. The lines are theoretical curves in local-field noise theory [29] with k_eff from 0 to 0.5.

Download Full Size | PDF

3.6. Comparison with mainstream 1064 nm APDs

In Table 1, we have compared the linear APD performance parameters between this work and several mainstream commercial APDs that can be operated at 1064 nm, including the NIR-Si [31], the InGaAs/InP [32], the InGaAs/InAlAs [33] and the InGaAsP/InP [7] APDs. The high unity gain QE of 40% for the NIR-Si APD at 1064 nm is realized by utilizing a thicker absorber and a higher operating temperature in a punch-through structure, which unavoidably gives rise to a slow response speed (<200 MHz) and a much higher V_B up to −400 V. Other advantages of Si APDs such as the high M, the low F(M) and the low I_d are retained. For InGaAs SAGCM APDs with InP [32] and InAlAs [33] multipliers, identical spectral response ranges of 1000–1700 nm, the same high unity gain QE of 70% at 1064 nm, very close V_B of around −40 V and similar J_D of around 3∼4×10⁻⁴ A/cm² at 0.9V_B are achieved, whereas the electron-initiated InGaAs/InAlAs APDs can offer a higher M and a lower F(M) compared with the hole-initiated InGaAs/InP ones. The InGaAsP/InP APDs that are specifically tailored for 1064 nm detection [7] offers an enhanced unity gain QE up to 90% at 1064 nm while the photo response only covers a much shrank spectral range of 900–1150 nm. An order of magnitude lower J_D of 4.6×10⁻⁵ A/cm² at 0.9V_B and a much higher V_B up to −89 V are also attained in comparison to that for InGaAs/InP APDs, which benefit from the lower tunneling leakages in the 1.15 eV InGaAsP absorber. The linear gain as well as the excess noise characteristics remain the same as that for InGaAs/InP APDs with a typical k_eff of ∼0.5. In contrast, the unity gain QE, the V_B and the M_max for the presented InGaAsP/InAlAs APD revealed lower when compared with these state-of-the-art APDs, which is attributed to the by no means optimal materials and processing. Nonetheless, the superior properties of lower I_d for InGaAsP absorbers and lower F(M) for InAlAs multipliers were still successfully demonstrated. Enhanced performances are achievable via further optimizing the E-field profiles, lowering the doping densities in both the multiplier and the absorber, applying of anti-reflection and back-reflection coatings, and etc.

Table 1. A comparison of the linear performance parameters between several mainstream commercial APDs that can be operated at 1064 nm and this work. M_max-maximum M, J_D-dark current density.

View Table

4. Conclusion

In conclusion, we have fabricated an electron-initiated 1064 nm InGaAsP/InAlAs SAGCM APD. A tailored InGaAs/InAlAs DGSL grading layer was introduced to ensure smooth injection of photo-generated electrons into the InAlAs multiplier. A unity gain QE of 48%, a low RT I_d of 470 pA at 90% V_B, and a maximum M of 5 are demonstrated. The dominant source of I_d is identified to be the TAT current in InAlAs multiplier from temperature-dependent I_d-V measurements. Furthermore, low multiplication noises with a k_eff of ∼0.2 are also observed. The maximum M is currently limited by the non-optimized electric field profiles, and can be readily enhanced by modifying the doping and thickness parameters for the multiplier and the charge layer. These results further indicate the combination of InGaAsP/InAlAs could be more promising in comparison to InGaAsP/InP towards the development of high performance 1064 nm APDs.