Silicon photodiodes with high photoconductive gain are demonstrated. The photodiodes are fabricated in a complementary metal-oxide-semiconductor (CMOS)-compatible process. The typical room temperature responsivity at 940 nm is >20 A/W and the dark current density is ∼100 nA/cm2 at 5 V reverse bias, yielding a detectivity of ∼1014 Jones. These photodiodes are good candidates for applications that require high detection sensitivity and low bias operation.
© 2012 OSA
Photoconductors have long been attractive as photodetectors for their simple structure, relatively low operating bias and high photo responsivity. In particular, high responsivity from photoconductive gain has been achieved in compound semiconductor systems ranging from bulk detectors made with PbS [1, 2] and CdS  to recently reported CdTe-based nanoparticles  and PbS based colloidal quantum dots [5–8]. These material systems are attractive for their high gain and absorption characteristics, but suffer from relatively high dark current levels and cost, and manufacturing issues related to uniformity and scaling.
In contrast, silicon photovoltaic detectors are manufactured in tremendous scale and are highly reliable, but silicon-based photoconductive photodetectors are rarely reported . Those that are reported incorporate materials that are incompatible with silicon processing facilities. Creating a silicon-based detector with photoconductive gain using a CMOS-compatible process is very attractive for low-light (non shot-noise-limited), discrete component applications, especially when the downstream read noise needs to be overcome with higher input signal. In particular, due to the relatively long response time for photoconductive gain detectors, applications such as fluorescence detection, biological and medical imaging, and other slow response imaging would benefit greatly from a low dark current, high photoconductive gain silicon detector. In this paper we present characterization results from silicon photodiodes with high photoconductive gain, low dark current, and low noise levels fabricated on 200 mm silicon wafers using a CMOS-compatible process in a commercial wafer foundry.
2. Device fabrication and characterziation
The photodetectors we characterized were fabricated with and without a diode junction incorporated in the device. The latter is a simple resistive silicon photoconductor, consisting of a high resistivity (>100 Ω-cm) and long minority carrier lifetime (>200 μs) float-zone (FZ) silicon layer with ohmic metal-silicon contacts on each end. Figure 1(a) shows the dark current-voltage (I–V) curve of a representative photoconductor (in black dash). As expected, the photoconductor exhibits a high dark current density, typically on the order of 1 mA/cm2 at 5 V reverse bias. A diode junction was then incorporated into the photoconductor by generating a doped active area with opposite polarity to the FZ substrate (∼ 725 μm thickness; boron doped for p-type and phosphorus for n-type) using ion implantation (boron for p-type and phosphorus for n-type, the implant recipes yield the junction depth ∼ 0.3 μm with the surface doping concetration ∼ 1019 atoms/cm3) and thermal annealing followed by contact fabrication, thereby creating a photoconductive gain device with a rectifying junction. Its dark I–V curve is shown in Fig. 1(a) in blue solid. Figure 1(b) shows the I–V curves under illumination (in red solid) and in the dark (in black dash) from such a photodiode with 940 nm wavelength responsivity >10 A/W at 5 V reverse bias.
Compared to the photovoltaic photodiodes, the photodiodes with gain show significantly higher photo response at reverse bias as well as at forward bias. The photodiodes with gain also exhibit significantly lower dark current density than the simple photoconductor with gain at reverse bias. Operating at reverse bias, the photodiodes with gain offer a good combination of photo response and dark current performance, which manifests as high detectivity.
A refinement of the photodiode with photoconductive gain is achieved by including a biasable guard ring structure that prevents the carriers generated outside the active area from being collected and amplified by the gain mechanism. The long carrier diffusion length (>100 μm) in the FZ silicon layer would otherwise allow these carriers to diffuse into the active area and contribute to the dark current. Figure 1(c) shows the device schematic of such a photodiode with a 2.5 mm diameter active region. Figure 1(d) is a plot of the photodiode dark current density at 5 V reverse bias as a function of the guard ring bias. At ∼0.8 V guard ring bias, the photodiode has a dark current density ∼100 nA/cm2.
3. Measurements and discussions
The spectral photoresponsivity of the photodiodes was measured using a calibrated, monochromator filtered, white light source. The monochromator is then scanned through the spectral range of interest, and a measurement is taken at various wavelengths in order to generate the spectral photoresponsivity data. Figure 2 shows a representative responsivity spectrum from a photodiode operated at 5 V reverse bias. The photodiode demonstrates excellent photore-sponse over a broad spectrum that covers various applications from fluorescence detection to biology and medical imaging. One will note that over most of the spectrum the responsivity values indicate external quantum efficiency (EQE) values much higher than 100%. An EQE value exceeding 100% implies electronic gain from either a multiple electron-hole pair generation process from a single photon or extra charges being supplied by the external circuit. The former is unlikely because the corresponding photon energies are low. The latter explanation allows for low photon energies, and is consistent with a photoconductive gain mechanism.
Fundamentally, the photoconductive gain mechanism is achieved by preventing recombination, thereby allowing carriers to traverse the circuit multiple times. This can be achieved by ”trapping” one type of carrier in long-lived, localized states while the other carrier species is able to flow through the photoconductor freely. This leads to the photoconductive gain, the gain coefficient for which is : , where τ is the average time spent by the ”trapped” carrier in the localized state, and t is the average time required for the free flowing carrier to transit through the circuit, which also means the free flowing minority carrier lifetime needs to be longer than the average transit time t. The photocurrent from the photoconductor is then: , where P is the incoming optical power, η is the internal quantum efficiency (IQE), q is the electron charge and ν is the optical frequency of the photon. Therefore, the extra photocurrent through the extra charges supplied from the external circuit is: .
As stated, the photoconductive gain description is consistent with the observed responsivity increase with bias; the higher bias fields lead to higher drift velocities for the freely moving carrier and hence the device transit time t is reduced. Alternatively, to achieve higher photoconductive gain, the device transit time t can also be reduced by physically shrinking the distance over which the free flowing carriers need to transit through the circuit. In this case, the distance can be shortened by reducing the bulk silicon thickness. However, reducing the bulk silicon thickness will reduce light absorption, especially towards the longer wavelengths near the silicon bandgap energy. This will lower the IQE, which in turn will lower the photoresponsivity. Methods such as laser surface texturing [11, 12] to increase the light path length in the silicon can be used to help maintain IQE while achieving higher gain and resulting in further improved photoresponsivity.
An additional trait of photoconductive gain (as compared to photovoltaic detectors) is an additional noise component associated with the statistical nature of the gain mechanism. The severity of the impact of this additional noise component is mitigated by the inclusion of a diode and a guard ring to reduce dark current, and its associated noise.
With abundant light illumination, shot noise (photon shot noise and other generation-recombination shot noise) dominates the noise performance for both photodiodes and photoconductors. The shot noise for photoconductor is : , where I is the total current, B is the bandwidth, G is the photoconductive gain, ω is the modulation frequency and τ is the carrier lifetime. For photodiode it is: . Noise spectra were measured for the photodiodes with gain. The noise-equivalent power (NEP) measured at 100 Hz under illumination is yielding a detectivity of or Jones. Moreover, systems using photoconductive gain detectors can exhibit high overall signal to noise ratios (SNR) in cases where the downstream read noise is higher than the increased noise from gain. More specifically, the overall detection system SNR is defined as: , where , , is the system downstream read noise and is the thermal noise. So the photoconductive gain G helps to overcome the downstream read noise and hence exhibit high overall SNR. This makes systems using such photoconductive gain detectors suitable for low-light (non shot-noise-limited) measurement applications.
It is worthwhile to mention that non-gain photovoltaic photodiodes can also exhibit overall high SNR for low-light measurement applications when they are configured with op amps, feedback resistances and capacitances in so called ”transimpedence ampliflier” forms. In this case, however, the gain is provided by the analog circuitry which is independent from the non-gain photovoltaic photodidoes. A typical photovoltaic photodiode integrated with such ”transimpe-dence amplifier” has an overall NEP , which is similar to that of the photoconductive gain photodidoes. However, the photocondutive gain photodiodes are lower cost and radiation and electrostatic discharge (ESD) safer, and have less power consumption with the amplifying analog circuitry eliminated. These features make the photoconductive gain photodiodes particularly advantageous in certian applications such as computed tomography (CT). 
To investigate possible trap states that trap one type of carrier and lead to the photoconductive gain, the Arrhenius plot of photoresponsivity vs. temperature for the photodiode was generated and is shown in Fig. 3(a). The plot reveals a single energy state with an activation energy ∼0.32 eV. The ∼0.32 eV activation energy indicates that this trap state is either ∼0.32 eV below the conduction band (electron trap) or ∼0.32 eV above the valence band (hole trap). We attribute this thermal quenching of photoconductive gain to the thermal deactivation of the trap state. At a higher temperature, the thermal energy is sufficient enough to empty the carrier trap states at a faster rate. Trapping becomes less effective at preventing recombination, resulting in a reduction of both τ and G.
In general, the above conclusion requires that mobility be constant over the temperature range observed. For photoconductors such as PbS, the mobility depends significantly on temperature, and thus must be accounted for in the thermal quenching rate calculation. In the case of silicon, however, the mobility is relatively constant (varies by less than a factor of 2) over the temperature range in Fig. 3(a) ; therefore the ∼0.32 eV activation energy fitted from the Arrhenius plot in Fig. 3(a) should be a close approximation to the energy difference between the trap state and the nearest band edge, ΔE.
The thermal quenching rate of photoconductive gain is expected to be proportional to for an electron trap or for a hole trap, where Nc and Nv are the effective density of states of the conduction and the valence bands and Ns is the trap state density. Based on these expressions and the measured thermal quenching rate, the trap state density is estimated  to be ∼1015 – 1016cm−3.
An additional way to determine the trap state energy is to study the photoconductive gain transient response dependence on temperature, more specifically, the fall time, tfall, dependence on temperature [2,8]. The fall time is a characteristic of the detrapping process of carriers from the trap states and therefore inversely proportional to the carrier thermal emission rate, which is , where σn is the capture cross section of the trap and vth is the thermal velocity of the carriers. Hence, with , the Arrhenius plot of tfallT1/2 vs. T also reveals the trap state and the band edge energy difference ΔE and the T1/2 term is to factor out the vth dependence on temperature (vth ∝ (kT)1/2).
Figure 3(b) shows an oscilloscope trace of typical rise and fall tails of the photodiode with photoconductive gain when illuminated using a current modulated 660 nm LED. The typical rise and fall times (T90) at room temperature from these silicon photodiodes with high photoconductive gain are ∼550 μs and ∼300 μs. The activation energy fitted from the Arrhenius plot in Fig. 3(c) is ∼0.32 eV, which is in good agreement with the one acquired from the photo response vs. temperature plot in Fig. 3(a), and confirms that the mobility temperature dependence contribution to the photoconductive gain is negligible within the temperature range. A similar plot of triseT1/2 vs. T shown in Fig. 3(d), on the other hand, reveals an activation energy of ∼0.47 eV. This activation energy is a characteristic of the trapping process of carriers into trap states, but also incorporates characteristics of the net recombination process, and is thus not indicative of the trap state energy.
In summary, this work reports silicon photodiodes with high photoconductive gain. The energy level of the trap state that is responsible for high gain has been determined experimentally by two methods. These photodiodes operate at low bias, and offer a manufacturable, scalable, low cost solution to applications such as fluorescence detection, biological and medical imaging.
The work was supported in part by the U. S. Army Research Office under contract No. W911NF-09-C-0084 and the The Night Vision and Electronic Sensors Directorate under contract No. W15P7T-10-C-S603.
References and links
1. R. H. Bube, Photoconductivity of Solids (Krieger Publishing, 1960), pp 365–367.
2. S. Espevik, C. Wu, and R. H. Bube, “Mechanism of photoconductivity in chemically deposited lead sulfide layers,” J. Appl. Phys. 42, 3513–3529 (1971). [CrossRef]
3. P. H. Wendland, “New large area CdS photoconductor,” Rev. Sci. Instrum. 33, 337–339 (1962). [CrossRef]
4. H. Chen, M. K. F. Lo, G. Yang, H. G. Monbouquette, and Y. Yang, “Nanoparticle-assisted high photoconductive gain in composites of polymer and fullerene,” Nat. Nanotechnol. 3, 543–547 (2008). [CrossRef] [PubMed]
5. G. Konstantatos, I. Howard, A. Fischer, S. Hoogland, J. Clofford, E. Klem, L. Levina, and E. H. Sargent, “Ultra-sensitive solution-cast quantum dot photodetectors,” Nature 442, 180–183 (2006). [CrossRef] [PubMed]
6. G. Konstantatos and E. H. Sargent, “PbS colloidal quantum dot photoconductive photodetectors: Transport, traps, and gain,” Appl. Phys. Lett. 91, 173505 (2007). [CrossRef]
7. G. Konstantatos, J. Clifford, L. Levina, and E. H. Sargent, “Sensitive solution-processed visible-wavelength photodetectors,” Nat. Photon. 1, 531–534 (2007). [CrossRef]
8. G. Konstantatos, L. Levina, A. Fischer, and E. H. Sargent, “Engineering the temporal response of photoconductive photodetectors via selective introduction of surface trap states,” Nano Lett. 8, 1446–1450 (2008). [CrossRef] [PubMed]
9. A. F. Sklensky and R. H. Bube, “Photoelectronic properties of zinc impurity in silicon,” Phys. Rev. 6, 1328–1336 (1972). [CrossRef]
10. S. M. Sze and K. K. Ng, Physics of Semiconductor Devices (Wiley, New Jersey2007).
11. J. E. Carey, C. H. Crouch, M. Shen, and E. Mazur, “Visible and near-infrared responsivity of femtosecond-laser microstructured silicon photodiodes,” Opt. Lett. 30, 1773–1775 (2005). [CrossRef] [PubMed]
12. M. U. Pralle, J. E. Carey, H. Homayoon, S. Alie, J. Sickler, X. Li, J. Jiang, D. Miller, C. Palsule, and J. McKee, “Black silicon enhanced photodetectors: a path to IR CMOS,” Proc. SPIE 7660, 76600N (2010). [CrossRef]
14. R. Steadman, F. M. Serrano, G. Vogtmeier, A. Kemna, E. Oezkan, W. Brockherde, and B. J. Hosticka, “A CMOS photodiode array with in-pixel data acquisition system for computed tomography,” IEEE J. Solid-State Circuits 39, 1034–1043 (2004) [CrossRef]