1. Introduction

The single photon detection process typically requires a source of large gain with low noise to overcome the thermal noise of subsequent readout electronics. A number of single photon detectors take advantage of an avalanche mechanism to amplify the signal generated from a single photon to a size large enough to be detected by the readout electronics [1, 2]. For example, the avalanche photodiode (APD) is a solid state device where a single electron hole pair, created by the absorption of a single photon, is accelerated in a large applied field to impact ionize additional carriers across the bandgap of the material. These secondary electron hole pairs then undergo a similar process, producing a large current pulse.

An important performance metric for single photon detectors is quantum efficiency (QE), which is defined as the ratio of the number of photons resulting in a detection signal to the number of photons incident on the detector. The QE of a solid state single photon detector typically has a strong dependence on the wavelength of the photons, due to the variation in the absorption coefficient of the absorbing material used in the device.

One of the highest QE detectors available is the visible light photon counter (VLPC) [3,4,5]. The VLPC is a Si based solid state single photon detector that features an impact ionization based gain mechanism. Compared to an APD, the VLPC has impact ionization that occurs over an impurity bandgap rather than the semiconductor bandgap. As a result, the VLPC features a much lower gain dispersion than an APD [6] and photon number discrimination capability [7]. QE of ~ 88% at 694 nm has been demonstrated with its internal QE estimated at ~ 95% [5]. While the VLPC exhibits very high QE in the visible wavelengths (400 – 700 nm), the QE in the ultraviolet (UV, < 400 nm) is several orders of magnitude lower. A number of applications in quantum information science and spectroscopy would benefit greatly if the operating range of the VLPC was extended into the UV [8, 9].

One application that would benefit from extending the operating range of the VLPC into the UV is the state detection of trapped ions in the ion trap implementation of the quantum computer [10]. The state detection process involves initiating a cycling transition in an ion, resulting in scattering of ~ 5 × 10⁷ photons per second via spontaneous emission when the ion is in the bright state, and practically zero photons when the ion is in the dark state. Typically, low F/# optics are used to collect these scattered photons (~ 5%) and direct them towards a detector. To achieve a high signal to noise ratio (SNR) in the detection process, a detector with high QE, low dark counts, and high internal gain is necessary [8]. The SNR of the detection process can be increased by increasing integration time, but longer integration time is expected to be a major bottleneck in the quantum computation process.

Because the absorption coefficient of Si greatly increases in the UV wavelengths, almost all of the incident UV photons are absorbed within the degenerate Si front contact of the VLPC. The electric field in the front Ohmic contact is close to zero, so the photogenerated electron hole pair will recombine before it can be separated and initiate an avalanche in the gain layer of the detector. In order to extend the operating range of the VLPC into the UV, absorption within the front contact must be greatly reduced. We accomplish this by removing some of the front contact and replacing it with a Ti metal Schottky contact. This modified version of the VLPC is referred to as the ultraviolet photon counter (UVPC).

The organization of this paper is as follows: Section 2 discusses the operating principles and performance of the VLPC; Section 3 discusses the design principles of the UVPC; Section 4 details the fabrication of the UVPC and presents and discusses the experimental results of the UVPC; conclusions are presented in Section 5.

2. VLPC Operation and Spectral QE

The VLPC is a Si based solid state single photon detector utilizing impurity band impact ionization as the gain mechanism. The structure of the VLPC is shown in Fig. 1(a). The VLPC consists of a number of Si eptiaxial layers grown on a degenerately doped Si substrate. The contact, spacer, and gain layers are doped with arsenic and compensated with acceptors to precisely control the electric field profile under operating bias conditions. These layers are followed by a thick, high purity intrinsic Si absorption layer. The topmost layers are a thin, degenerately doped Si front contact and a single layer antireflection (AR) coating optimized for transmission at 550 nm.

 

Fig. 1. (a) Schematic device structure of the VLPC. The VLPC consists of a number of Si epitaxial layers grown on a degenerate Si substrate. (b) Schematic device structure of the UVPC. The degenerately doped front contact of the VLPC is thinned and replaced with a Ti Schottky metal contact and a MgF₂ AR coating. (Note: thicknesses not drawn to scale.)

Download Full Size | PDF

The VLPC is operated at ~ 7 K, where the arsenic impurity ions in the gain layer form an impurity band 54 meV beneath the conduction band. When a photon is absorbed within the absorption layer, the applied field separates the photogenerated electron hole pair causing the hole to drift into the gain layer. As the hole moves through the gain layer, it will impact ionize one or more electrons from the impurity band into the conduction band. The generated electron(s) will impact ionize additional electrons, resulting in an avalanche of electrons, which flows toward the front contact. During this process, several tens of thousands of electrons are generated over a very short timescale (< 1 ns) with minimal gain dispersion [6].

The measured QE of the VLPC system as a function of wavelength is shown in Fig. 2. The VLPC system QE is discussed in further detail in Section 4. Also plotted is the absorption coefficient of Si at 10 K [11]. The QE is very high in the 400 – 650 nm wavelength range but drops sharply below 400 nm. The high QE of the VLPC in the visible wavelength range is attributed to the VLPC’s near unity absorption efficiency and the high probability for the impact ionization process to occur and initiate an avalanche event [4]. The absorption efficiency of the VLPC, η_abs can be described by η_abs = (1 – R) e ^-αt_c [1 – e ^{−α(t_abs+t_gain)} , where R is the reflectance, α is the absorption coefficient of Si, t_c is the thickness of the contact layer, t_abs is the thickness of the absorption layer, and t_gain is the thickness of the gain layer [4]. In the visible wavelengths, where α is ~ 10³ - 10⁴ cm^-1, absorption within the front contact layer(~ 250 nm) is small and the contact is transparent. The large combined thickness of the intrinsic and gain layers (~ 30 μm) leads to a near unity probability of absorption. For wavelengths < 400 nm, a rises rapidly due to a direct bandgap transition of Si, and nearly all of the incident photons are absorbed within the front contact layer, causing the QE of the VLPC to decrease dramatically.

 

Fig. 2. The measured system QE of the VLPC and Si absorption coefficient at 10 K in the wavelength range of 300 – 650 nm is shown [11]. The QE drops sharply below 400 nm, which corresponds to a rapid increase in the absorption coefficient of Si. For wavelengths ≤ 370 nm, the input light level was increased to improve the SNR.

Download Full Size | PDF

3. Design of the UVPC

To increase the QE of the VLPC in the UV, it is necessary to reduce absorption within the VLPC front contact layer by reducing the thickness to below 10 nm. Potential methods for achieving this goal include Schottky metal contacts, laser annealed contacts [12, 13], and ultra-thin contacts grown with molecular beam epitaxy [14]. This section discusses the strategy used for increasing the QE of the VLPC in the UV by using the Schottky contact approach to design the UVPC.

The structure of the UVPC is shown in Fig. 1(b). The degenerately doped Si front contact of the VLPC is removed and replaced with a thin Ti contact layer (~ 10 nm thick) with a single layer MgF₂ AR coating that both protects the Ti from oxidation, as well as reduces the reflectance of the contact.

The design of the UVPC resembles the Schottky barrier photodiode (SBP), an ordinary p-i-n diode with p material replaced by metal. The QE of a photodiode is maximized by ensuring that incident light is absorbed within the depletion region. For a SBP, the depletion region begins immediately after the metal; for this reason, SBPs are particularly useful in the UV wavelengths where the absorption coefficient of Si is very high. For the same reasons, a Schottky contact will increase the QE of the VLPC in the UV. Photons transmitted through the Schottky contact will be absorbed within the VLPC’s intrinsic absorption layer where the field is strong enough to separate the carriers and drive the hole into the gain layer. Important design considerations for high QE SBPs are the transmission of the Schottky metal contact and the Schottky barrier height. These same considerations are important to the design of the UVPC.

Maximizing the transmission of a Schottky metal contact involves minimizing the absorption and reflectance of the metal contact. To minimize the absorption in the contact, a metal with a small absorption coefficient should be selected and made as thin as possible. There is a tradeoff between the electrical and optical properties of the metal film: thinner metal decreases the conductivity while thicker metal increases the absorption. Reducing the reflectance of the metal contact can be achieved through the use of AR coatings.

The Schottky barrier height also plays a very important role on device performance. For the UVPC, we want to ensure that there is no barrier to electron collection and a high barrier to hole injection. Direct injection of holes from the metal to Si will result in an avalanche identical to that from a photogenerated hole, leading to increased dark counts. For an ideal Schottky barrier, the barrier height is determined by the difference between the electron affinity of the semiconductor and the work function of the metal. For a real Schottky barrier, interface states have a major impact on the barrier height and generally reduce the p type barrier. For this reason, as well as to reduce recombination near the interface, it is important to keep the metal Si interface clean and to minimize surface states.

We selected titanium (Ti) for use as a thin Schottky metal contact in the UVPC due to its relatively low absorption in the near UV (300 – 400 nm) and its high p type barrier. Calculated values for the transmission of a 10 nm thick Ti film are > 40% with a single layer MgF₂ coating, and can be made > 70% with multilayer coatings for wavelengths > 300 nm. High p type barrier diodes have been fabricated on Si with Ti through careful control of the Ti/Si interface [15, 16]. Important features of these devices include a very clean Si surface and removal of hydrogen from the interface. Native oxide removal, crucial in preventing any electrical barrier, is typically accomplished with the use of hydrofluoric acid (HF), which leaves the Si surface terminated with hydrogen. We use in situ Ar ion RF sputtering to both remove any hydrogen from the interface and ensure a very clean surface prior to Ti deposition. Using this technique, we were able to fabricate p type Schottky diodes with a very high p type barrier height of 0.60 ± 0.04 eV and an ideality factor of 1.10, indicating a high quality interface. The p type barrier was determined by measuring current-voltage relationships and applying thermionic emission-diffusion theory [17] (See Fig. 3).

4. UVPC Experimental Results and Discussion

Fabrication of the UVPC starts with a 8 pixel VLPC chip. The first step was to remove the single layer SiO₂ AR coating with a 90 sec buffered oxide etch (BOE, 10:1 NH₄:HF). The degenerate front contact was etched 180 nm using a Cl₂ based reactive ion etching process. The etched VLPC was mounted in a RF sputter system after a 10 sec BOE etch was used to remove the native oxide. After pumping the chamber to a pressure of ~ 2 × 10^-7 Torr, the sample was cleaned by Ar ion RF sputtering for 5 min at 70 W and a pressure of 5 mTorr. Immediately following the RF sputter clean, a 10 nm Ti film was deposited at an RF power of 250 W and a pressure of 5 mTorr. A single layer AR coating of MgF₂ was then deposited at an RF power of 140 W and a pressure of 5 mTorr. The thickness of the MgF₂ layer was controlled by the use of an optical monitor at the wavelength of interest and a Si monitor sample. The device was then packaged and wire bonded for testing.

 

Fig. 3. Current density-voltage (J-V) relationship of a 1000 μm diameter Ti Schottky diode with a MgF₂ coating. Thermionic emission-diffusion theory was used to determine the p type barrier height. For clarity, data for all temperatures measured and used in the analysis are not shown.

Download Full Size | PDF

The UVPC was tested in a cold finger cryostat, which had a 200 μm core multimode UV enhanced fiber aligned to the UVPC. A UV lamp combined with a monochromator was used to control the wavelength of the input light. A series of variable attenuators was used to attenuate the photon flux, which was focused into a patch cable that could then be connected to the fiber input of the cryostat. The photon flux input into the cryostat was attenuated to ~ 2 × 10⁵ photons per second at each wavelength measured. Uncertainty in the photon number was ±0.4 dB, which was dominated by fiber connector loss uncertainty at the system input.

The system QE of the UVPC in the 300 – 400 nm wavelength range is shown in Fig. 4. The system QE is defined as the ratio of the number of photons detected (resulting in a electrical pulse) to the number of photons incident on the system. The input point of our system is defined to be the UV enhanced multimode fiber connector at the room temperature end of our cryostat. The system QE of the UVPC, which had a single layer MgF₂ AR coating optimized for ~ 370 nm, was 5.3% ± 0.5% at 300 nm and 11% ± 1% at 370 nm. The UVPC was operated at a device temperature of 7.1 K and a device bias of 7.25 V. The dark counts at these operating conditions were ~ 16,000 counts per second.

The internal QE of the UVPC was estimated by accounting for several sources of loss in the system. The internal QE is defined as the ratio of the number of photons detected (resulting in a electrical pulse) to the number of photons transmitted through the front contact. The attenuation of the 1.7 m long UV enhanced multimode fiber that runs down the cryostat has wavelength dependent attenuation ranging from 0.2 dB/m to 0.05 dB/m in the 300 – 400 nm wavelength range. Misalignment of the fiber to the detector reduces the output coupling efficiency of the fiber to an estimated 90%. The transmissive, absorptive, and reflective characteristics of the Ti/MgF₂ contact were simulated using ideal index values for the real and imaginary parts of the refractive index of the materials involved. Reflectance measurements show that the deposited films are more absorptive than the predicted values of the ideal films. The transmission values used in estimating the internal QE were the simulated values and therefore we expect that the reported internal QE values shown in Fig. 4 provide a lower bound on the actual internal QE of the UVPC. With these corrections, we estimate a lower bound of the actual internal QE of the UVPC to be 16.5% at 300 nm and 27% at 370 nm.

 

Fig. 4. The system QE and estimated lower bound for the internal QE of the UVPC and VLPC from 300 - 400 nm. The internal QE is estimated by accounting for losses in the fiber, output coupling efficiency, and detector coatings.

Download Full Size | PDF

The estimated internal QE of the UVPC is far below the internal QE of the VLPC (~ 95% at 694 nm). This is an indication that there is still significant absorption within the contact area of the UVPC. The thickness of the degenerate Si front contact of the VLPC is ~ 250 nm. The highest QE for the UVPC was observed when ~ 160 nm of the front contact was etched, leaving ~ 90 nm of the original contact layer intact prior to deposition of the Ti/MgF₂ Schottky contact. Shallower etching (< 160 nm) resulted in reduced QE due to increased absorption within the remaining contact while deeper etching (> 160 nm) also resulted in reduced QE. We expect that the impact of etch related defects becomes greater and decreases the QE as more of the front contact is etched (resulting in fewer dopants). More research is necessary to confirm this hypothesis.

5. Conclusion

The VLPC is a single photon detector with a number of desirable properties including high QE in the visible wavelengths but the QE of the VLPC drops to almost zero in the UV wavelengths. We introduced the UVPC, a modified version of the VLPC with increased QE in the near UV wavelengths (300 – 400 nm). System QE of 5.3% ± 0.5% at 300 nm and 11% ± 1% at 370 nm was demonstrated, with a lower bound on the internal QE estimated at 16.5% at 300 nm and 27% at 370 nm. Future efforts into better multilayer coatings, as well as the use of laser annealed or epitaxially grown contacts, should increase the QE of the UVPC even further.

This work was supported by the National Science Foundation under CCF-0546068.