# High performance planar germanium-on-silicon single-photon avalanche diode detectors

## Abstract

Single-photon detection has emerged as a method of choice for ultra-sensitive measurements of picosecond optical transients. In the short-wave infrared, semiconductor-based single-photon detectors typically exhibit relatively poor performance compared with all-silicon devices operating at shorter wavelengths. Here we show a new generation of planar germanium-on-silicon (Ge-on-Si) single-photon avalanche diode (SPAD) detectors for short-wave infrared operation. This planar geometry has enabled a significant step-change in performance, demonstrating single-photon detection efficiency of 38% at 125 K at a wavelength of 1310 nm, and a fifty-fold improvement in noise equivalent power compared with optimised mesa geometry SPADs. In comparison with InGaAs/InP devices, Ge-on-Si SPADs exhibit considerably reduced afterpulsing effects. These results, utilising the inexpensive Ge-on-Si platform, provide a route towards large arrays of efficient, high data rate Ge-on-Si SPADs for use in eye-safe automotive LIDAR and future quantum technology applications.

## Introduction

Near room temperature semiconductor-based single-photon avalanche diode (SPAD) detectors have become the accepted optical detection approach in a variety of emerging application areas in the visible and short-wave infrared spectral regions1,2,3. SPAD detectors are avalanche photodiodes biased at fields above avalanche breakdown, in Geiger mode, where a self-sustaining avalanche current can be triggered by an incident single-photon. After the photo-induced avalanche, the detector must be reset, or quenched, before the next detection event. Such detectors typically have a temporal jitter of hundreds of picoseconds, allowing ultra-sensitive measurement of fast optical transients. At wavelengths below 1 µm, silicon-based SPADs have been used in a range of quantum photonic applications, including experiments in quantum foundations4 and fibre and free-space quantum communications demonstrations5,6. In applications such as Light Detection And Ranging (LIDAR), Si-based SPAD detectors have also emerged as a candidate technology due to the high sensitivity and the picosecond temporal response which has resulted in enhanced range and improved surface-to-surface resolution7. Si SPAD detectors have been integrated with standard Si CMOS processes to produce ultra-sensitive, large format detector arrays with integrated electronics8. This low-cost technology has allowed time of flight systems to be adapted and developed for use in the automotive9,10,11 and smartphone industries12.

There are a number of clear advantages in extending the spectral range of SPAD detectors into the short-wave infrared (SWIR) region, beyond the detection spectrum of Si-based SPADs. Firstly, compatibility with the optical fibre low-loss telecommunications windows is a fundamental advantage in many fibre-based applications. Secondly, in free-space applications such as LIDAR and range-finding, the optical power of the laser source is limited by laser eye-safety thresholds. This eye-safety threshold increases by approximately a factor of 20 or more when the laser wavelength is increased from 850 to 1550 nm (IEC-60825-1 standard), permitting increased optical power whilst maintaining eye-safety in active imaging applications. Consequently, this results in an increased maximum attainable range and/or potential improvements to depth resolution. Thirdly, solar radiation, which typically acts as the background level in most single-photon LIDAR systems, decreases considerably in the SWIR13. Finally, operation in the SWIR will mean enhanced atmospheric transmission, .especially through obscurants such as smoke, smog, fog and haze14,15.

In the SWIR region, the most widely used single-photon detectors are InGaAs/InP SPADs and superconducting nanowire single-photon detectors (SNSPD)1,2. Generally, SNSPD devices have had superior single-photon detection performance; however, the cryogenic operating temperatures, typically below 3 K, limit their use in certain key application areas. InGaAs/InP SPADs are the dominant single-photon detector in the SWIR region and have been used in a range of quantum communications experiments, notably in long distance quantum key distribution demonstrations16. They are typically operated at temperatures between 220 and 255 K which are achievable using Peltier cooling, which have allowed compact detector modules to be used in LIDAR field trial scenarios17. For example, InGaAs/InP SPADs have been used in LIDAR and depth profiling experiments to good effect in the kilometre range17,18. Arrays of InGaAs/InP SPADs can give high-performance detection at telecommunications wavelengths19; however, two-dimensional arrays may prove expensive for the low-cost, high volume automotive and autonomous vehicle LIDAR markets. One issue that has made use of InGaAs/InP SPADs challenging has been the effect of afterpulsing—described below—which has severely limited the count rates possible of these detectors.

In this paper we report on the first normal-incidence, planar Ge-on-Si SPAD. We demonstrate high-performance single-photon operation, illustrating high-efficiency detection and low afterpulsing effects. These results illustrate clear potential for integration with Si CMOS for low-cost SPAD array imaging in the SWIR bands.

## Results

### Design, fabrication and preliminary characterisation

The devices described in this paper use the SACM structure as shown in Fig. 1. The incident SWIR radiation is absorbed in the Ge absorption region and the signal amplification takes place in the Si multiplication region. In between these regions, a selectively implanted charge sheet is used to control the electric field so that the field is high enough in the multiplication region to ensure that avalanche breakdown is reached and low enough in the absorption region to prevent band-to-band and trap-assisted tunnelling. A modest electric field, however, is maintained in the Ge layer to allow efficient drift of photogenerated electrons into the multiplication region. The electric field profile at 5% excess bias above the breakdown voltage is shown in Fig. 2a. The SPAD is biased above the avalanche breakdown voltage and when an incident photon is absorbed in the Ge layer, an electron−hole pair is created and the electron drifts into the Si multiplication region. Here it is accelerated, gaining sufficient kinetic energy to undergo impact ionisation, creating an electron−hole pair. The secondary electrons and holes are in turn accelerated and impact ionise, creating further electron−hole pairs. Further impact ionisation of both holes and electrons rapidly creates a large avalanche current which can be self-sustaining if the device is biased above avalanche breakdown. Under these conditions, this results in a detectable electronic signal that can be timed relative to the initial laser pulse. After detection, it is necessary to bias the SPAD momentarily below avalanche breakdown to quench the avalanche, after which the SPAD can return to its quiescent state ready to detect further incident photons.

Finite element analysis modelling using Silvaco ATLAS software was used to design the SPADs, as shown in Fig. 2a. The charge sheet doping levels and the thicknesses of the multiplication and absorber regions were determined, as well as the optimum overall design dimensions of the SPAD. This was necessary to ensure that the electric field profile throughout the SPAD was appropriate to give high performance, as discussed above. In the simulations shown in Fig. 2a, the electric field profiles are shown for a potential held at 5% above the avalanche breakdown voltage, or 5% excess bias. It is clear from Fig. 2a that there is a low electric field in the Ge absorber at breakdown, and, crucially the high electric field is confined to the centre of the SPAD preventing carriers originating at the sidewalls from causing breakdown events.

The planar SPAD growth and fabrication process is described in the Methods section. A mesa design with exposed sidewalls, similar to Fig. 2b, was also fabricated and used as a control during characterisation. To ensure high yield fabrication, device diameters ranged from 100 to 200 µm, resulting in very large cross-sectional areas compared to previous Ge-on-Si SPADs27,28. Future SPADs will be significantly smaller than this—with an aim to reduce the diameter to around 10 µm in order to reduce DCRs further.

Figure 2c shows the dark current at an operating temperature, T = 100 K for both the planar and mesa etched SPAD structures fabricated from the same wafer. The mesa etched structure has a similar microstructure to the planar geometry in Fig. 1, except that an etch process was used to create a mesa of diameter less than the diameter of the ion-implanted charge sheet, etched to a depth just below the charge sheet and into the multiplication layer (Fig. 2b). The planar SPAD has a sharp breakdown indicating a low multiplied dark current, previously found to be a strong indicator of the desired low DCR performance30. The mesa etched SPAD has a much softer breakdown with a dark current 50 times higher than the planar structure immediately before breakdown. This indicates that, as expected, significant surface generation is present and suggests that it will have high DCRs compared to the planar SPAD. Indeed, it was not possible to characterise the mesa etched SPAD above breakdown due to its prohibitively high DCR. Figure 2d demonstrates the dark current and photocurrent of the planar SPAD as a function of reverse bias at T = 78 K. The dark current before breakdown is less than 1 nA and the SPADs exhibited good uniformity, with little variation in dark current between devices. Photocurrent measurements at a wavelength, λ, of 1310 nm demonstrate clear punchthrough at 20 V where the electric field reaches the absorption region and photoexcited electrons can drift into the multiplication region. The device yield was over 90%, which at this early stage, is very encouraging for the eventual realisation of Ge-on-Si SPAD focal plane arrays.

### Time-correlated single-photon counting characterisation

After preliminary characterisation, SPDE, DCR and jitter measurements were taken using the time-correlated single-photon counting (TCSPC) technique, as described in more detail in the Methods section. In these measurements, an electrical gating approach was used to switch the detector to above avalanche breakdown, into the Geiger mode, for a duration of 50 ns in synchronisation with the arrival of the attenuated laser pulse. This gated detector approach was used at a low frequency of 1 kHz in order to fully quench the avalanche and avoid the effects of afterpulsing (described below). The SPAD detectors were initially cooled to T = 78 K for SPDE, DCR and jitter measurements using λ = 1310 nm laser radiation. The SPDE and DCR as a function of excess bias at T = 78 K, 100 K and 125 K are shown in Fig. 3. It should be noted that the detectors used had a large area (100 µm diameter), and it is fully expected that the DCR will be considerably lower with reduced area devices, as previously reported in all-Si SPADs31.

The measured DCR demonstrates a vast improvement when compared to previous Ge-on-Si work. Warburton et al. reported on mesa geometry Ge-on-Si SPADs with a DCR of 5.5 MHz for a 25 µm diameter SPAD at T = 100 K28. This corresponds to 11,200 counts/s/µm2 which is approximately three orders of magnitude higher than the 18.3 counts/s/µm2 reported in this work. It should also be noted that, under these conditions, the SPDE reported in that paper is 4%, compared to 26% for the SPAD reported in this paper. There is a similar relationship when our results are compared to results from Martinez et al., who reported a DCR of 500 kHz for a 1 µm wide by 15.9 µm long rectangular waveguide SPAD at T = 80 K29. This corresponds to 31,400 counts/s/µm2 which is over three orders of magnitude higher than the 6.37 counts/s/µm2 reported in this work. They report an SPDE of 5%, compared to 22% for the SPAD reported in this paper. This considerable reduction in DCR has resulted from the carefully designed electric field profile of these planar geometry detectors which means that the high electric field is confined within the SPAD, preventing surface states contributing significantly to the DCR. Most dark counts in these SPADs are now likely to originate from dislocations arising from the Si/Ge interface and from thermal excitation throughout the volume of device. In order to fully ascertain the relative contributions to DCR, we are initiating a series of measurements on samples with different diameters and Ge thicknesses.

Figure 3 demonstrates that the SPDE increases with excess bias to a maximum of 38% at T = 125 K, significantly higher than previous SPDEs reported for Ge and Ge-on-Si SPADs22,27,28,29 and comparable to the highest values recorded for InGaAs/InP SPADs at T = 225 K32,33,34. This is due in part to the high excess bias applied across the SPAD, attainable due to the low DCR, increasing the breakdown probability in the multiplication region. The relatively thick 1.5 µm Si multiplication region increases the breakdown probability, the likelihood of a self-sustaining avalanche occurring on arrival of the primary electron. The uniform electric field in the multiplication region, caused by minimal residual doping in the lower part of the Si multiplication layer, results in a uniform impact ionisation rate throughout, increasing the breakdown probability still further. The optimised electric field also ensures the efficient transit of photoexcited electrons into the multiplication region. Significantly, there is no conduction band energy barrier between the Ge absorption region and the Si charge sheet ensuring the photoexcited electrons can easily pass between the two regions. Indeed the Si Δ-valley of the conduction band edge is 235 meV below the Ge L-valley conduction band edge in the absorber if calculated using the deformation potentials in ref. 35. This is an advantage over InGaAs/InP SPADs, which possess an energy barrier step that photoexcited carriers must overcome to reach the InP multiplication region. These SPADs require an InGaAsP grading layer between the InGaAs and InP regions to reduce carrier accumulation at the absorber−charge sheet interface. Finally, an antireflection coating is used to reduce reflection from the top surface of the SPAD to less than 1%. For Ge-on-Si SPADs, the absence of a conduction band barrier at the Ge/Si heterointerface for photogenerated electrons to overcome should ensure that the SPDE remains high at elevated temperatures as the DCR is improved in future design iterations. With these samples, measurements at higher temperatures were limited by the increasing DCR rate due to increasing thermal generation rates; however, a reduction of the detector area in future work is likely to reduce the DCR further and allow a significantly higher operating temperature.

The high SPDE has been achieved despite the use of a relatively thin Ge absorption region. Using absorption coefficients for single crystal Ge at T = 77 K36 we have calculated that less than 50% of the λ = 1310 nm radiation is absorbed in the 1-µm-thick Ge absorber throughout the operational temperature range. Beer-Lambert’s law indicates that a 2-µm-thick Ge absorber will increase the absorption to over 70%, which should lead to an SPDE of greater than 55% at T = 125 K. This figure is significantly higher than reported SPDEs for InGaAs/InP SPADs32,33,34. Even thicker Ge layers should provide higher absorption still and we will examine thicker Ge absorbers in future work.

Noise equivalent power (NEP) is a figure of merit for SPADs calculated from the SPDE and DCR of the detector using

$${\mathrm {NEP}} = \frac{{hv}}{{{\mathrm {SPDE}}}}\sqrt {2{\mathrm {DCR}}},$$
(1)

where h is Planck’s constant and v is the frequency of the incident radiation. This can be used to compare detectors, with lower values indicating improved performance. At T = 78 K, we have calculated a record low NEP for a Ge-on-Si SPAD detector of 1.9 × 10–16 W/Hz1/2, 50-fold lower than the NEP of the Ge-on-Si SPAD reported in ref. 28. NEP values of 3 × 10–16 and 7 × 10–16 W/Hz1/2 were calculated for T = 100 K and 125 K respectively.

Figure 4 shows a timing histogram taken at an excess bias of 5.5% and T = 78 K when measured at λ = 1310 nm. The jitter full-width-at-half maximum (FWHM) is 310 ps, which is a reasonable value considering the increased width of the multiplication region37,38. Wider multiplication regions generally improve the SPDE but the increased variance in the avalanche build-up time increases the jitter. It is expected that the jitter will reduce as the device diameter is decreased, as found previously in Si SPADs39, and by improving the electronic packaging of the cooled device. Indeed preliminary measurements on Ge-on-Si SPADs with a diameter of 26 µm show a jitter of ~175 ps, closer to the performance of commercial InGaAs/InP SPADs.

The wavelength dependence of the Ge-on-Si SPAD detector efficiency will vary as the Ge bandgap changes with temperature. Figure 5a demonstrates the normalised wavelength dependence of the SPDE as a function of temperature. The high-efficiency SPDE region is related to direct bandgap absorption between the conduction band and the valence bands at the Γ-point. Absorption at longer wavelengths is related to significantly weaker indirect absorption into the L-valleys. At room temperature the direct bandgap of Ge is 0.80 eV35 but this increases to 0.88 eV at T = 78 K, reducing the detection cut-off wavelength35. Using a tunable laser we were able to vary the wavelength of the radiation incident on the SPAD from 1310 to 1550 nm to obtain accurate cut-off wavelengths at various temperatures. By defining the cut-off wavelength, λc, as the wavelength at which the detector’s SPDE is 50% of the λ = 1450 nm value. It can be observed that λc increases from 1468 nm at T = 125 K to 1495 nm at T = 175 K, increasing at a rate of approximately 0.54 nm/K, in agreement with values calculated using the Varshni temperature-dependent bandgap parameters in ref. 40, and shown in Fig. 5b. Using this device geometry, we expect λc to reach 1550 nm at T = 245 K. However, if the Ge absorber was increased in thickness to 2 µm, λc will be increased to longer wavelengths and will reach 1550 nm at the lower temperature of 220 K, as shown in Fig. 5c. This should be readily achievable using smaller diameter devices which exhibit lower DCR rates and allow higher temperature operation (as discussed above). This relatively high operating temperature is achievable using Peltier cooling which will permit Ge-on-Si SPADs to be used effectively in compact, low power LIDAR systems.

### Afterpulsing characterisation

One critical difference between Ge-on-Si SPADs and the InGaAs/InP SPAD alternative is a realistic potential of a significant reduction in the deleterious effects of detector afterpulsing. This phenomenon occurs when carriers are trapped after an avalanche event and then released later, resulting in an increased background level. Afterpulsing can be mitigated by using a long hold-off time (typically >10 µs) after each event in order that trapped carriers can be released prior to the detector being re-activated. This approach, however, increases the dead-time and restricts the maximum count rate possible. Afterpulsing is recognised as one of the main limitations of InGaAs/InP SPADs, severely affecting their performance, even at modest count rates. Afterpulsing in InGaAs/InP detectors originates mainly in the InP multiplication layer from deep level trap states41,42,43, and the expectation with Ge-on-Si SPAD detectors is that the high-quality Si multiplication layer will have a lower density of such states. For the first time, we show a comparison of an InGaAs/InP SPAD with a Ge-on-Si SPAD under nominally identical operating conditions.

To investigate the afterpulsing mechanism in Ge-on-Si SPADs, we examined the afterpulse lifetime as a function of temperature in the range 78−125 K, and fitted exponential decays. By fitting Arrhenius plots, we deduced activation energies in the region of 80–90 meV across a range of overbias levels to attempt to ascertain the origin of the traps. Native Si surfaces, native Ge surfaces and GeOx at Ge surfaces have been shown to have trap states close to 420 meV45, 130 meV45 and from 200 to 300 meV46, respectively. This provides further evidence that the planar geometry is reducing the effects of traps and other impurities at the exposed surfaces. Hence the afterpulsing is unlikely to be related to surface states on the passivated Ge or any exposed Si surfaces.

Figure 6c shows the band structure of the Ge-Si heterointerface calculated at T= 125 K using the deformation potentials from ref. 47 without any applied electric field for clarity and includes the three main trapped states for dislocations in Ge plotted as dashed lines inside the bandgap48. Whilst the 80−90 meV activation energy extracted from Fig. 6b is close to the 70 meV acceptor and the 90 meV donor trapped states at the valence band edge from the dislocations in the Ge, it is not clear what mechanism could be responsible for afterpulsing with these states. None of the metal impurities in Ge and Si with energies close to these values are expected to be at any trace levels in the present devices Co (90 meV acceptor), Zn (90 meV acceptor), Hg (87 meV acceptor) and Cr (70 meV acceptor)49. Metal impurities in Si with similar energies are Bi (69 meV donor), Ga (72 meV acceptor) and Al (67 meV acceptor)49. Whilst Al has been used for the contacts and bond-pads, these are at the top of the Ge and on the back of the Si wafer and the avalanche region of the device is buried so there should be no Al at trace levels close to the avalanche region of the device.

Dislocation traps in Si have been measured to be centred energetically at 807, 870, 940 and 1001 meV above the valence band at T = 12 K for the D1−D4 trap states, respectively50. The last of these dislocation trap states corresponds to ~130 meV below the Si conduction band at T = 125 K50 (see Fig. 6c) and the linewidth has at least 10 meV of thermal broadening at this temperature. It is well known that during relaxation of a Ge or Si1−xGex heterolayer grown above the critical thickness on a Si substrate, some threading dislocations can be injected into the Si substrate51 and some level of strain will be transferred into the Si close to the heterointerface which will reduce the energy between the trap state and the conduction band edge. The Si D4 trap state has been identified as originating from relaxed dislocations52 and with the uncertainty of thermal broadening and strain combined with the uncertainty of the afterpulsing excitation energy being extracted from three temperatures, this D4 dislocation trap in Si is a good candidate for the origin of the afterpulsing. Limited area growth of Ge on Si has already demonstrated significant reductions in threading dislocations densities both into the Ge heterolayer and into the Si substrate53. This would be one test to determine if the afterpulsing could be further reduced, thereby confirming if the D4 trap state is responsible for the afterpulsing mechanism. Further work is therefore required to confirm this afterpulsing process and to reduce the afterpulsing probability further.

## Methods

### Device fabrication

Five structures were grown on 150 mm diameter n++-doped Si (001) substrates. Firstly, a 1.5-µm-thick Si multiplication region was grown by a commercial reduced pressure chemical vapour deposition (RP-CVD) system. Finite element analysis modelling using Silvaco ATLAS software was used to determine the optimum charge sheet density to provide low electric fields in the Ge absorber and high electric fields in the Si avalanche region. Photolithography was used to define the charge sheet regions which were then implanted with boron acceptors at an energy of 10 keV. Different charge sheet doses were implanted in each of the five wafers to account for fabrication tolerances and ensure that the optimised electric field profile was achieved. After implantation the boron dopants were activated at 950 °C for 30 s using a rapid thermal annealer. After Radio Corporation of America (RCA) cleaning, a 1-µm-thick, nominally undoped Ge absorption region and a 50 nm p++ Ge top contact layer were grown on top of the selectively implanted Si layer using RP-CVD. A trench etch through the Ge was performed at a lateral distance of 10 µm from the charge sheet, in order to electrically isolate the SPADs, as shown in Fig. 1. This electrical isolation was required due to the conductive path formed by the background doping level found in the Ge layer. Metal contacts, GeO2 passivation, anti-reflection (AR) coatings and bond-pads were subsequently deposited.

## Data availability

All relevant data are available from the Heriot-Watt University data archive54.

## References

1. 1.

Buller, G. S. & Collins, R. J. Single-photon generation and detection. Meas. Sci. Technol. 21, 012002 (2010).

2. 2.

Hadfield, R. H. Single-photon detectors for optical quantum information applications. Nat. Photon. 3, 696–705 (2009).

3. 3.

Cova, S., Ghioni, M., Lotito, A., Rech, I. & Zappa, F. Evolution and prospects for single-photon avalanche diodes and quenching circuits. J. Mod. Opt. 51, 1267–1288 (2004).

4. 4.

Shadbolt, P., Matthews, J. C. F., Laing, A. & O’Brien, J. L. Testing foundations of quantum mechanics with photons. Nat. Phys. 10, 278–286 (2014).

5. 5.

Gordon, K. J., Fernandez, V., Townsend, P. D. & Buller, G. S. A short wavelength gigahertz clocked fiber-optic quantum key distribution system. IEEE J. Quantum Electron. 40, 900–908 (2004).

6. 6.

Pugh, C. J. et al. Airborne demonstration of a quantum key distribution receiver payload. Quantum Sci. Technol. 2, 024009 (2017).

7. 7.

Buller, G. S. & Wallace, A. M. Recent advances in ranging and three-dimensional imaging using time-correlated single-photon counting. IEEE J. Sel. Top. Quantum Electron. 13, 1006–1015 (2007).

8. 8.

Charbon, E. & Donati, S. SPAD sensors come of age. Opt. Photonics News 36, 34–41 (2010).

9. 9.

Niclass, C., Inoue, D., Matsubara, H., Ichikawa, T. & Soga, M. Development of automotive LIDAR. Electron. Commun. Jpn. 98, 1476–1480 (2015).

10. 10.

Gyongy, I., Dutton, N. A. W. & Henderson, R. K. Single-photon tracking for high-speed vision. Sensors 18, 323 (2018).

11. 11.

Takai, I. et al. Single-photon avalanche diode with enhanced NIR-sensitivity for automotive LIDAR systems. Sensors 16, 459 (2016).

12. 12.

Pellegrini, S. et al. Industrialised SPAD in 40 nm technology. In 2017 IEEE Int. Electron Dev. Meeting, San Francisco, CA, https://doi.org/10.1109/IEDM.2017.8268404 (2017).

13. 13.

Bird, R. E., Hulstrom, R. L. & Lewis, L. J. Terrestrial solar spectral data sets. Sol. Energy 30, 563–573 (1983).

14. 14.

Tobin, R. et al. Three-dimensional single-photon imaging through obscurants. Opt. Express 27, 4590–4611 (2019).

15. 15.

Arnulf, A., Bricard, J., Curé, E. & Véret, C. Transmission by haze and fog in the spectral region 0.35 to 10 microns. J. Opt. Soc. Am. 47, 491–498 (1957).

16. 16.

Korzh, B. et al. Provably secure and practical quantum key distribution over 307 km of optical fibre. Nat. Photon. 9, 163–168 (2015).

17. 17.

McCarthy, A. et al. Kilometer-range depth imaging at 1550 nm wavelength using an InGaAs/InP single-photon avalanche diode detector. Opt. Express 21, 22098–22113 (2013).

18. 18.

Pawlikowska, A. M., Halimi, A., Lamb, R. A. & Buller, G. S. Single-photon three–dimensional imaging at up to 10 kilometers range. Opt. Express 25, 11919–11931 (2017).

19. 19.

Jiang, X. et al. InP-based single photon detectors and geiger-mode APD arrays for quantum communication applications. IEEE J. Sel. Top. Quantum Electron. 21, 3800112 (2015).

20. 20.

Michel, J., Liu, J. & Kimerling, L. C. High-performance Ge-on-Si photodetectors. Nat. Photon. 4, 527–534 (2010).

21. 21.

Sammak, A., Aminian, M., Nanver, L. K. & Charbon, E. CMOS-compatible PureGaB Ge-on-Si APD pixel arrays. IEEE Trans. Electron Dev. 63, 92–99 (2016).

22. 22.

Buller, G. S. et al. Time-resolved photoluminescence measurements of InGaAs/InP multiple-quantum-well structures at 1.3 μm wavelengths by use of germanium single-photon avalanche photodiodes. Appl. Opt. 35, 916–921 (1996).

23. 23.

Loudon, A. Y. et al. Enhancement of the infrared detection efficiency of silicon photon-counting avalanche photodiodes by use of silicon germanium absorbing layers. Opt. Lett. 27, 219–221 (2002).

24. 24.

Kang, Y. et al. Monolithic germanium/silicon avalanche photodiodes with 340 GHz gain–bandwidth product. Nat. Photon. 3, 59–63 (2009).

25. 25.

Derose, C. T. et al. Ultra compact 45 GHz CMOS compatible germanium waveguide photodiode with low dark current. Opt. Express 19, 527–534 (2011).

26. 26.

Martinez, N. J. D. et al. High performance waveguide-coupled Ge-on-Si linear mode avalanche photodiodes. Opt. Express 24, 19072–19081 (2016).

27. 27.

Lu, Z. et al. Geiger-mode operation of Ge-on-Si avalanche photodiodes. IEEE J. Quantum Electron. 47, 731–735 (2011).

28. 28.

Warburton, R. E. et al. Ge-on-Si single-photon avalanche diode detectors: design, modeling, fabrication, and characterization at wavelengths 1310 and 1550 nm. IEEE Trans. Electron Dev. 60, 3807–3813 (2013).

29. 29.

Martinez, N. J. D. et al. Single photon detection in a waveguide-coupled Ge-on-Si lateral avalanche photodiode. Opt. Express 25, 16130–16139 (2017).

30. 30.

Webster, E. A. G. & Henderson, R. K. A TCAD and spectroscopy study of dark count mechanisms in single-photon avalanche diodes. IEEE Trans. Electron Dev. 60, 4014–4019 (2013).

31. 31.

Sciacca, E. et al. Silicon planar technology for single-photon optical detectors. IEEE Trans. Elec. Dev. 50, 918–925 (2003).

32. 32.

Zhang, J., Itzler, M. A., Zbinden, H. & Pan, J.-W. Advances in InGaAs/InP single-photon detector systems for quantum communication. Light Sci. Appl. 4, e286 (2015).

33. 33.

Tosi, A., Calandri, N., Sanzaro, M. & Acerbi, F. Low-noise, low-jitter, high detection efficiency InGaAs/InP single-photon avalanche diode. IEEE J. Sel. Top. Quantum Electron. 20, 3803406 (2014).

34. 34.

Amri, E., Boso, G., Korzh, B. & Zbinden, H. Temporal jitter in free-running InGaAs/InP single-photon avalanche detectors. Opt. Lett. 41, 5728–5731 (2016).

35. 35.

Paul, D. J. 8-band k.p modelling of the quantum confined Stark effect in Ge quantum wells on Si substrates. Phys. Rev. B 77, 155323 (2008).

36. 36.

Dash, W. C. & Newman, R. Intrinsic optical absorption in single-crystal germanium and silicon at 77K and 300K. Phys. Rev. 99, 1151–1155 (1955).

37. 37.

Itzler, M. A. et al. Single photon avalanche diodes (SPADs) for 1.5 µm photon counting applications. J. Mod. Opt. 54, 283–304 (2007).

38. 38.

Wang, S. et al. Analysis of breakdown probabilities in avalanche photodiodes using a history-dependent analytical model. Appl. Phys. Lett. 82, 1971–1973 (2003).

39. 39.

Lacaita, A. & Mastrapasqua, M. Strong dependence of time resolution on detector diameter in single photon avalanche diodes. Electron. Lett. 26, 2053–2054 (1990).

40. 40.

Varshni, Y. P. Temperature dependence of the energy gap in semiconductors. Physica 34, 149–154 (1967).

41. 41.

Buller, G. S. et al. Semiconductor avalanche diode detectors for quantum cryptography. IEEE LEOS Newsl. 20, 20–24 (2006).

42. 42.

Jiang, X., Itzler, M. A., Ben-Michael, R. & Slomkowski, K. InGaAsP–InP avalanche photodiodes for single photon detection. IEEE J. Sel. Top. Quantum Electron. 13, 895–905 (2007).

43. 43.

Korzh, B., Lunghi, T., Kuzmenko, K., Boso, G. & Zbinden, H. Afterpulsing studies of low-noise InGaAs/InP single-photon negative-feedback avalanche diodes. J. Mod. Opt. 62, 1151–1157 (2015).

44. 44.

Tosi, A., Mora, A. D., Zappa, F. & Cova, S. Single-photon avalanche diodes for the near-infrared range: detector and circuit issues. J. Mod. Opt. 56, 299–308 (2009).

45. 45.

Statz, H., deMars, G., Davis, L. Jr. & Adams, A. Jr. Surface states on silicon and germanium surfaces. Phys. Rev. 106, 455–464 (1957).

46. 46.

Zhang, R., Iwasaki, T., Taoka, N., Takenaka, M. & Takagi, S. Al2O3/GeOx/Ge gate stacks with low trap density fabricated by electron cyclotron resonance plasma postoxidation. Appl. Phys. Lett. 98, 112902 (2011).

47. 47.

Paul, D. J. 8-band k·p modelling of mid-infrared intersubband absorption in Ge quantum wells. J. Appl. Phys. 120, 043103 (2016).

48. 48.

Shevchenko, S. V. Electrical conductivity of germanium with dislocation grids. J. Exp. Theor. Phys. 88, 66–71 (1999).

49. 49.

Sze, S. Physics of Semiconductor Devices 2nd edn (Wiley, New York, 1981).

50. 50.

Binetti, S. et al. Optical properties of oxygen precipitates and dislocations in silicon. J. Appl. Phys. 92, 2437–2445 (2002).

51. 51.

LeGoues, F. K., Meyerson, B. S., Morar, J. F. & Kirchner, P. D. Mechanism and conditions for anomalous strain relaxation in graded thin films and superlattices. J. Appl. Phys. 71, 4230–3243 (1992).

52. 52.

Sauer, R. et al. Dislocation-related photoluminescence in silicon. Appl. Phys. A 36, 1–13 (1985).

53. 53.

Noble, D. B., Hoyt, J. L., King, C. A. & Gibbons, J. F. Reduction in misfit dislocation density by the selective growth of Si1-xGex/Si in small areas. Appl. Phys. Lett. 56, 51–53 (1990).

54. 54.

Vines, P. et al., Dataset for “High performance planar Ge-on-Si single-photon avalanche diode detectors”, Heriot-Watt University https://doi.org/10.17861/5117a021-fc29-4577-ba3d-e53e86b9a4d7 (2018).

## Acknowledgements

The authors wish to acknowledge the support of UK EPSRC projects EP/N003446/1, EP/K015338/1, EP/L024020/1, EP/M01326X/1, EPN003225/1 and the DSTL Ph.D. scholarship DSTLX-1000092774.

## Author information

Authors

### Contributions

The planar Ge-on-Si SPAD device was proposed by R.W.M. and D.J.P., with later significant input from G.S.B. and P.V. The design was subsequently developed by all the authors. The devices were fabricated by J.K., D.C.S.D. and M.M.M. using a process developed by J.K., D.C.S.D., R.W.M., M.M.M. and D.J.P. P.V. and K.K. undertook the Silvaco TCAD simulations, and the experimental characterisation and data analysis, under the supervision of G.S.B. All authors discussed the results and approved the final manuscript.

### Corresponding authors

Correspondence to Peter Vines or Gerald S. Buller.

## Ethics declarations

### Competing interests

The authors declare no competing interests.

Journal peer review information: Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

## Rights and permissions

Reprints and Permissions

Vines, P., Kuzmenko, K., Kirdoda, J. et al. High performance planar germanium-on-silicon single-photon avalanche diode detectors. Nat Commun 10, 1086 (2019). https://doi.org/10.1038/s41467-019-08830-w

• Accepted:

• Published:

• ### Integrated photonic quantum technologies

• Jianwei Wang
• , Fabio Sciarrino
• , Anthony Laing
•  & Mark G. Thompson

Nature Photonics (2020)

• ### High concentration factor diffractive microlenses integrated with CMOS single-photon avalanche diode detector arrays for fill-factor improvement

• Peter W. R. Connolly
• , Ximing Ren
• , Aongus McCarthy
• , Hanning Mai
• , Federica Villa
• , Alberto Tosi
• , Franco Zappa
• , Robert K. Henderson
•  & Gerald S. Buller

Applied Optics (2020)

• ### Enhancing the absorption of a thin germanium slab with periodical patterning

• Xiyuan Cao
• , Yijin Zhang
• , Ziyang Han
• , Wenfei Li
• , Yi Jin
•  & Aimin Wu

Journal of Materials Science: Materials in Electronics (2020)

• ### On-chip nanophotonics and future challenges

• Alina Karabchevsky
• , Angeleene S. Ang

Nanophotonics (2020)

• ### Full Waveform LiDAR for Adverse Weather Conditions

• Andrew M. Wallace
• , Abderrahim Halimi
•  & Gerald S. Buller

IEEE Transactions on Vehicular Technology (2020)