WO2003077283A2

WO2003077283A2 - Drift-dominated detector

Info

Publication number: WO2003077283A2
Application number: PCT/US2003/006506
Authority: WO
Inventors: Yanning Sun; Jerry M. Woodall; Eric S. Harmon; David B. Salzman
Original assignee: Yale University
Priority date: 2002-03-04
Filing date: 2003-03-04
Publication date: 2003-09-18
Also published as: AU2003230597A8; WO2003077284A3; AU2003225650A8; WO2003077284A2; AU2003230597A1; AU2003225650A1; WO2003077283A3

Abstract

A conventional pin photodetector structure consisting of an n-type semiconductor substrate(101) of thickness (121). On top of the substrate, a n-type semiconductor buffer layer (102) of thickness (122), followed by an intrinsic semiconductor layer (103) of thickness (123), and p-type semiconductor layer (105) of thickness (125), are all grown epitaxially on the substrate (101).

Description

DRIFT-DOMINATED DETECTOR

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims priority from U.S. Patent application S/N 60/361,665, filed March 4, 2002, which is incorporated by reference herein.

FIELD OF THE INVENTION

The invention relates generally to semiconductor materials, and especially to devices formed from such materials and used for converting photons into mobile charge carriers, such as photodetectors, avalanche photodiodes and solar cell devices. It teaches methods and structures for reducing recombination losses and generation noise at semiconductor surfaces. These methods/structures can be used to improve ultraviolet photoconversion efficiency near an illuminated surface, as well as to improve the speed for collecting minority carriers from the region near an illuminated surface, thus improving device durability under ultraviolet irradiation, extending photoconversion efficiency across the visible and into the near infrared band, and providing tolerance to high defect levels in this surface region. More broadly, any minority carrier semiconductor device with an appreciable rate of recombination near a surface or interface can potentially benefit from the invention.

LIMITATIONS OF THE PRIOR ART

It is well-known in the prior art of semiconductor materials and devices that having a low surface or interface state density is advantageous in its own right. However, most semiconductors do not exhibit a low surface state density, because the dangling bonds and broken crystalline symmetry at the semiconductor surface give rise to a high density of mid-gap states. Likewise, many semiconductor interfaces, such as heterojunctions, exhibit an appreciable surface state density, even though this interface state density is usually significantly lower than that of an exposed surface.

In some semiconductors, prior art techniques can be used to significantly reduce the density of surface states, such as incorporating a wide band-gap semiconducting or insulating layer on top of the semiconductor surface that reacts with the dangling bonds or continues the crystal symmetry on the surface and thereby eliminates a significant fraction of the surface states. For example, under certain growth conditions, thermally growth SiO_≥ on silicon provides such a low surface state density, whereas bare (i.e., etched) silicon would exhibit a significantly higher surface state density.

Likewise, a similar prior art reduction in surface state density can be achieved for a GaAs surface passivated by AlGaAs. The case of AlGaAs on top of GaAs causes most, but not all of the GaAs surface states to be passivated. The remaining surface states that have not been passivated can have a detrimental effect on some devices. The upper (exposed) AlGaAs surface is not passivated, so most surface states occur there instead of at the AlGaAs/GaAs interface. Fortunately, the band gap difference between the GaAs and AlGaAs generally prevents minority carriers in the GaAs from reaching the AlGaAs surface states. Unfortunately, any minority carriers in the AlGaAs would still experience a strong surface recombination, hence loss, particularly for ultraviolet (UV) photons, which are absorbed efficiently in the AlGaAs layer. Furthermore, it is possible that thermal generation at the AlGaAs surface can inject minority carriers into the GaAs layer and thereby introduce an additional noise current into the GaAs device.

A low density of surface states makes it easier for doping to fill the surface states, and thereby reduce the extent of band bending, undesirable surface recombination or generation effects, and field crowding. Avoiding or reducing the density of surface states is thus a well-recognized challenge in the art. For instance, reducing surface state densities has been central to progress in miniaturizing silicon MOSFET devices.

This is why high performance prior art photodetectors have generally been limited to materials with (a) low crystallographic defect density in the bulk, (b) inherently low surface state density, (c) low surface recombination velocity, or (d) suitable surface passivation, such as Si0₂/Si or AlGaAs/GaAs that reduces the- surface state density at the semiconductor surface. Cases (a) and (b) reduce recombination, while cases (c) and (d) avoid it. Cases (b) , (c) and (d) may offer certain advantages: greater field uniformity, reduced field crowding at the edges of device (hence a higher voltage at breakdown) , less aging at the wavelengths of operation, and lower noise due to generation at the surface.

For applications of UN photodetectors, short wavelength light generates carriers very near the first illuminated (a/k/a top) surface, where the carriers often recombine and are lost due to the presence of surface states. Surface states create a dead layer region of the semiconductor where minority carriers recombine efficiently, preventing most minority carriers generated in the dead layer from participating in the photocurrent . This dead layer makes photodetectors blind to UV light. In some semiconductors, surface states may also be created by UV light of sufficient energy to break bonds at the surface of the semiconductor, leading to a rapid, irreversible increase in the dead region, destroying of the photodetector' s quantum efficiency.

A notable prior art approach uses a transparent Schottky diode as a top contact to a photodetector . Use of the Schottky diode allows direct biasing of the surface depletion region in order to drive carriers toward the appropriate contacts. However, there can be problems with absorption, scattering and reflection in or at the semi-transparent contact. These and other effects generally reduce the device' s quantum efficiency to a level where noise sources and spectral limitations matter greatly. See, for instance, Necmi Biyikli, Tolga Kartaloglu, Orhan Aytur, Ibrahim Kimukin and Ek el Ozbay, "High-speed visible-blind GaN-based indium-tin-oxide Schottky photodiodes." Appl . Phys . Lett . , Vol. 79, No. 17, 22 October 2001.

In the prior art, good structures and means are not known for reducing surface recombination in order to enable enhanced UV performance, the relation between materials parameters and device design for this purpose, and optimizations for performance, size, cost, noise, and other constraints. Furthermore, there is a need for structures and means to reduce the detrimental effects of recombination sites and defects at or near this surface, including point defects, crystalline defects, and grain boundaries. The invention addresses these needs.

MOTIVATION AND SUMMARY OF THE INVENTION

Reverse band bending at a surface is caused by a non-negligible density of surface states, such as frustrated chemical bonds, crystal imperfections, or surface defects caused by a flux of UV or gamma photons, X-rays, alpha or beta particles, nuclear fragments, or molecular fragments. Surface band bending occurs because of surface Fermi level pinning, an effect whereby surface states drive the Fermi level towards a specific energy level within the forbidden band gap of the semiconductor. Surface band bending usually causes an enhanced surface recombination of photo-generated carriers, because carriers generated near the surface are accelerated by the bent band toward the surface (where surface states act as efficient recombination centers) , instead of toward the bulk material (where they would be collected by the intended electrical junction and create a photo- current) . Furthermore, surface band bending acts as a very effective sink for minority carriers, so even carriers generated beyond the surface band bending region experience a strong diffusion force towards the surface. Surface band bending gives rise to an effective dead layer from which few minority carriers can escape into the active region of the device. Since short wavelength light generates photocarriers near the surface, the surface band bending effect, together with the diffusion force effect, jointly result in poor efficiency in detecting UV and blue light.

In addition to recombination due to reverse band bending, surface or interface states may also act as efficient thermal generation sites, creating minority carriers. These minority carriers will have a certain probability of escaping from the surface region and reaching the active junction of a device, where they can be collected and generate a dark noise current. The generation rate depends on the generation cross section and density of the surface or interface states, and the probability of escape depends on the details of minority carrier transport from the surface. For the case of surface or interface generation noise, reverse band bending may advantageously reduce noise because the reverse band bending reduces the escape probability.

In general, the present invention succeeds by avoiding the deleterious effects of surface states, especially preventing recombination and generation at these surface states from interacting with the bulk active region of a device. One aspect of the invention relates to a technique to reduce the extent of reverse band bending. Reducing the effects of surface recombination and generation, in accordance with the invention, extends UV sensitivity, increases dynamic range, improves gain characteristics for avalanche devices, and makes aging under bombardment less severe.

Any minority carrier semiconductor device with an appreciable rate of recombination at a surface or interface can potentially benefit from the invention. The invention also enables devices to be tolerant of defects near the surface region by decreasing the collection time and thereby decreasing the probability of recombination losses at any defects, including point defects, crystalline defects and grain boundaries. Devices using the invention can therefore be produced in lower quality material than prior art devices.

For good performance, a drift dominated detector ("D3") typically requires both (a) that the photo-generated minority carriers will be drifted by a strong electric field away from the surface dead layer and toward the depletion region where they are collected, and (b) that the extent of the dead layer be reduced. This is in essence what the invention accomplishes. It is known in the prior art that drift through a thin region, with a high electric field results in short transit times, resulting in very fast collection of minority carriers in the region. If a thin, high electric field region can be generated near the surface of a photodetector, then collection from the surface region of the photodetector can be enhanced, even in the presence of a high density of surface states that normally result in a high recombination rate. The inventors also recognize that in addition to developing a built-in drift field that drives minority carriers away from the surface dead layer, steps should preferably be taken to suppress the otherwise dominant effects of reverse band bending due to surface states. By controlling the extent of the dead layer and providing a strong built-in drift field to accelerate minority carriers away from the dead layer, the invention reduces minority carrier losses due to surface recombination.

In accordance with the invention, noise due to generation at surface states can also be suppressed by designing an appropriate, reverse band bending region that confines minority carriers generated at the surface and keeps such carriers from reaching the bulk active region of the device. Indeed, by making this surface band bending region thin, the magnitude of the reverse band bending electrical field is increased, resulting in a very effective force for reducing the probability of escape for minority carriers generated at surface states. Six ways to accomplish this, individually or in combination, are as follows:

First, heavy doping at the surface can be used to reduce the thickness of the dead layer by lessening the penetration of the surface depletion region into the active region of the semiconductor. However, at sufficiently high doping levels, there is usually a tradeoff between the advantages of higher doping and the disadvantages from bulk recombination effects at higher doping. At some high enough level of doping, such as p+-GaAs doped beyond 2xl0²⁰ cm^-3, Auger, radiative, and other non- radiative effects will cause non-negligible ^λbulk" recombination regardless of the surface recombination effects and begin to degrade device performance. The inventors and others have disclosed in another patent application techniques for achieving very high doping densities without significantly increasing non- radiative recombination due to defects, while simultaneously avoiding dopant diffusion. (See U.S. Patent Application S/N 60/341,051, entitled "Methods Of Hyperdoping Semiconductor Materials And Hyperdoped Semiconductor Materials And Devices," filed 10/22/02, incorporated herein by reference) . In particular, this "hyperdoping" technique enables stable, high doping densities to be achieved, while retaining relatively long minority carrier lifetimes (usually dominated by radiative and/or Auger effects) . Very high doping concentrations at the top surface of a device significantly reduce the extent of the surface depletion region.

Second, very heavy doping can be exploited to saturate the surface states, driving the surface Fermi level closer to the band edge and thereby further reducing the band bending caused by surface Fermi level pinning. This occurs by transferring sufficient charge from the bulk semiconductor near the surface to the surface states to fill up a significant fraction of these surface states. One consequence of achieving heavier doping near the surface is an increase in the charge transferred to the surface states, making it possible to fill a larger fraction of the surface states, reducing the Fermi level pinning effect. In p-type materials, forcing the surface Fermi level towards the valence band edge reduces band bending and the surface depletion region thickness. In n-type materials, forcing the surface Fermi level towards the conduction band edge enables reduced band bending and reduced surface depletion region thickness.

Third, very heavy doping makes it possible to degenerately dope the surface region, which may be advantageous because the Fermi level can be pushed into the conduction band for n-type materials (or valence band for p-type materials) , which results in a widening of the effective band gap (the Moss-Burstein shift) . A wider band gap lowers surface recombination (which is a function of the carrier density, n ) , and enables a band gap gradient to be produced by grading the doping from a high degeneracy level to a low degeneracy or non-degenerate level . Such a band gap gradient provides an effective electrical field, which can be used to drift photo-generated minority carriers away from the surface depletion region and toward the active junction of a device. An effective widening of the band gap is also very beneficial for counteracting band gap shrinkage, in which heavy doping can produce an effective reduction of the band gap as the doping level is increased, that would normally drive minority carriers towards the surface. Thus, while band gap widening is counteracted by band gap shrinkage, at sufficient levels of degeneracy, the band gap widening effect is stronger than the band gap shrinkage effect, enabling a highe fective electrical field to be produced that accelerates minority carriers away from the surface and towards the active junction of a device. Therefore, an optimal design should preferably consider the drift field component due to band gap shrinkage and counter it with the some combination of the effective electrical fields due to doping gradients, band gap widening at degenerate doping densities, and alloy compositional grading (see below) .

Fourth, incorporation of a high doping gradient that extends beyond the surface band bending region to provide a strong drift field that accelerates minority carriers away from the surface and toward the active junction of a device will enable the device to quickly and efficiently collect minority carriers generated near the surface of the device. The effective electrical field can advantageously be engineered to match or counter the diffusion fields, which would normally drive minority carriers to the surface depletion region. This effective electrical field can also improve the efficiency of minority carrier collection even in the presence of a high density of defects such as point defects, crystalline defects and grain boundaries. A properly designed drift field reduces the time carriers spend at or near the reverse band bending region.

A high effective electrical field can be produced within a material by incorporating a very rapid grading of the doping from a very high level at the surface to a lower level in the bulk region. By building in a high enough electrical field, the transit time of minority carriers through this region can be made short enough to reduce the amount of recombination due to heavy doping, point defects, crystalline defects, grain boundaries, and surface states and therefore increase the collection efficiency of a device. As noted above, when incorporating a doping gradient into a device, effects of band gap shrinkage due to heavy doping and band gap widening due to degeneracy should be taken into account. The high doping densities and good control of doping gradients achieved with epitaxial techniques such as MBE or MOCVD will result in a strong electrical field, and hence a strong drift region.

In particular, it is advantageous to make the dopant density gradation sufficiently smooth and high that dE_c / dx = E is approximately constant between the surface and the depletion region of an active pn (or pin) junction. The value of E should be chosen to insure that the minority carriers are accelerated to their saturation drift velocity limit for their entire transit through the device. This results in collection of the minority carriers by the junction optimally, since minority carriers drift efficiently, with high sensitivity to blue and UV wavelengths, and carriers are not lost or slowed near the surface. The high doping density and good control of doping gradients give a high gradient drift region, which accelerates carriers to saturation drift velocity, and thereby produce efficient collection of minority carriers.

Fifth, alloy compositional gradients can be used to decrease the band gap away from the surface, which increases the effective drift force on minority carriers for driving them away from the surface. Indeed, an alloy compositional gradient may be used to provide an effective drift field that accelerates minority carriers away from a surface. Alloy compositional grading is most useful when combined with either uniform heavy doping or with doping gradients, to minimize the penetration of reverse band bending into the semiconductor.

Sixth, using one or more of the above-described techniques, materials with non-negligible densities of surface recombination sites can be used more effectively as photon detectors because the electrical fields built into the device will mitigate the reverse band bending and surface recombination in these devices. By optimizing the reverse band bending region at the surface of the device to confine minority carriers generated at the surface of the device, one reduces the probability that these minority carriers will reach the active junction of the device, where they would add to the dark noise of the device. Thus, the invention enables lower noise to be achieved in properly designed devices. Beyond improving the detection limit, lower noise detectors are useful for increasing the dynamic range, which has important implications for analog systems.

There are numerous commercially-interesting applications of the invention. These include photodetectors with: enhanced sensitivity to blue and UV light, wide sensitivity across the UV/visible/NIR, avalanche gain, Geiger mode avalanche gain, , low dark current and noise, low dark count rate, high duty cycle, fast response time, fast recovery time, high tolerance of defects (including surface states, point defects, crystalline defects, and grain boundaries), and combinations of the above. A notable application of the invention is as a replacement for certain kinds of photomultiplier tubes or time-of-flight detectors, in which case the invention provides a solid state device with high quantum efficiency, comparative immunity to magnetic fields, robustness, small size, small mass, low voltage, and potentially low cost. Detectors deployed in linear or area arrays may also be notably useful embodiments. Combined with optics (such as refractive, diffractive, reflective, holographic, or other optical components) and/or optical paths (such as drift spaces, optical fibers, waveguides, photonic crystals, and related elements) an array of the photodetectors enabled by the present invention can use well-known techniques to achieve unprecedented capabilities in spectroscopy, imaging, and other optical systems. Solar cells, which are quasi-direct current, may also be built. These take advantage of further features, including high quantum efficiency, radiation hardness due to reduced recombination at both the surface and in the bulk, long operational lifetime, space-flight qualifiable durability, low mass, and gravimetric efficiency.

To achieve maximum advantage from the invention, doping profiles, materials, and combinations of them should be chosen such that a photodetector' s efficiency in detecting short wavelength light is dominated by bulk semiconducting properties, rather than surface or edge effects. Materials and doping profiles should reduce surface band bending and/or provide an internal drift field to counteract the diffusion field directly. Surface chemistry should be designed to be stable against a high integrated flux of soft UV light.

Highly doped n-type semiconductor material is particularly useful for this purpose. Since holes have a higher effective mass than electrons, balanced designs favor reducing the distance traveled for hole transport, and therefore holes should be collected from a thinner region while electrons are collected from .a thicker region (i.e., the n-type region should be thinner than the p-type region) . For example, InP can be stably doped n-type to 10²⁰ cm^-3 for this purpose, giving excellent n-type contacts and doping, n- type InP can also be graded smoothly downward starting at 10²⁰ cm^" ³ at the surface to below 10¹⁶ cm^-3 near the junction to produce a high intrinsic drift field that accelerate holes away from surface, where they would otherwise recombine. InP can also be stably doped p-type to about 10¹⁹ cm^-3, for good p-type contacts and low resistivity, p-type InP can also be graded smoothly downward from 10¹⁶ cm^-3 at the junction to 10¹⁹ cm^-3 deeper into the device, providing a drift field on the p-type side of the device to sweep electrons generated deep in the device toward the active junction of the device. It is well-known in the field that surface Fermi level pinning in InP occurs near the conduction band, so the reverse band-bending occurs over a smaller distance for any given level of n-type doping than in materials whose surface Fermi level pins nearer to mid gap. Even if the reduced band-bending provided by the invention is not as important in InP as it is in many other materials, the invention nevertheless enables improved InP-based detectors, by virtue of using grading from a high doping densities to a lower doping densities over small distances. While this specific example applies to InP, those skilled in the art will recognize that it can be readily extended to any semiconductor that can be heavily doped n-type.

In some embodiments, the dopant type may be switched to p-type and similar devices may be built using high p-type doping densities at the surface graded down to a lower doping density near the active junction of the device.

Grading both sides of the junction can be useful for other embodiments. For instance, a singly graded homojunction device (i.e., grade only the n-type doping at the surface of the device) can be built, but a double graded (grade both the n-type doping at the surface and the p-type doping on the other side of the active junction) homojunction has superior properties. The double graded homojunction has all the advantages of single graded device, and it can be engineered using well-known optimizations to collect electrons and holes more efficiently, eliminate the slow tail from holes by balancing electron and hole collection, achieve a wider or narrower spectral response (especially if very thin devices are used to reduce sensitivity to longer wavelengths) , or achieve improved breakdown characteristics useful for operation in the avalanche gain regime.

It is also possible to apply the teachings of the invention to a heterojunction. For instance, a heterojunction of p-+AlGaAs graded to p+- GaAs (or p-type GaAs) has advantageous properties as an embodiment .

Thus, to briefly summarize, the invention teaches techniques for: reducing recombination and carrier generation at surface states; reducing recombination near the surface due defects such as point defects, crystalline defects, and grain boundaries; relating materials parameters and device design for this purpose; optimizing performance, size, cost, and other constraints; avoiding deleterious effects from the reverse band bending caused by a non-negligible density of surface states. In accordance with the invention, these techniques can be used to produce photodetectors with (i) increased sensitivity into the ultraviolet (UV) , including devices with sensitivity extended from the UV into the visible and into the near infrared;

(ii) high-speed operation, avalanche gain, proportional mode operation, Geiger mode avalanche operation, a high duty cycle capability, array capability, and low dark current; and

(iii) lower noise, and hence higher linearity, higher maximum gain, and lower dark counts.

BRIEF DESCRIPTION OF THE FIGURES

Figure 1A depicts a prior art pin photodetector. Figure IB depicts the band diagram of the pin photodetector shown in Figure 1A.

Figure 2A depicts a prior art approach to reducing the surface dead region using a passivation window.

Figure 2B depicts the band diagram of the windowed pin photodetector shown in Figure 2A.

Figure 3A depicts a prior art approach to reducing the surface dead region using a very heavily doped cap layer.

Figure 3B depicts the band diagram of the pin photodetector with cap layer shown in Figure 3A.

Figure 4A depicts a preferred embodiment of the invention capable of extended response in the UV.

Figure 4B depicts a doping profile as a function of depth for the embodiment of Figure 4A.

Figure 4C depicts a band diagram for the embodiment of Figure 4A.

Figure 5A depicts a first alternative embodiment using a modified p-type doping.

Figure 5B depicts a doping profile as a function of depth for the first alternative embodiment of Figure 5A.

Figure 5C depicts a band diagram for the first alternative embodiment. Figure 6A depicts a second alternative embodiment using p-type doping at the surface.

Figure 6B depicts a doping profile as a function of depth for the second alternative embodiment of Figure 6A.

Figure 6C depicts a band diagram for the second alternative embodiment of Figure 6A.

Figure 7A depicts a third alternative embodiment incorporating InGaAs to extend the response of the detector for longer wavelengths .

Figure 7B depicts a doping profile as a function of depth for the third alternative embodiment of Figure 7A.

Figure 7C depicts a band diagram for the third alternative embodiment of Figure 7A.

Figure 8A depicts a forth alternative embodiment incorporating AlGaAs at the surface, graded down to GaAs, to provide a built-in quasi electrical field that accelerates electrons away from the exposed AlGaAs surface.

Figure 8B depicts a doping profile as a function of depth for the forth alternative embodiment of Figure 8A.

Figure 8C depicts a band diagram for the fourth alternative embodiment of Figure 8A.

Figure 9A depicts a device structure of an experimental realization of the invention. Figure 9B depicts a experimental internal quantum efficiency results from the experimental realization of the invention.

Figure 9C shows theoretical analysis of the internal quantum efficiency of materials as a function of absorption depth.

DETAILED DESCRIPTION OF THE ILLUSTRATIVE EMBODIMENTS

Reference is now made to Figures 1A - B, which depict a prior-art approach photodetector. Shown in Figure 1A is a conventional pin photodetector structure consisting of an n-type semiconductor substrate 101 of thickness 121. On top of the substrate, a n-type semiconductor buffer layer 102 of thickness 122, followed by an intrinsic semiconductor layer 103 of thickness 123, and p-type semiconductor layer 105 of thickness 125, are all grown epitaxially on the substrate 101. In general, the thickness 125 is kept small in order to reduce the absorption of incident photons in layer 105, because a significant fraction of the photo electrons generated in this layer are lost to recombination at the top surface 135. While the above description of the photodiode employs epitaxial techniques such as molecular beam epitaxy (MBE) to develop the layer structure, the layer structure may be developed by other techniques, such as ion implantation and dopant diffusion, as is often done when the semiconducting material is silicon.

Figure IB shows the approximate band diagram of the pin photodiode structure depicted in Figure 1A. The Figure shows the conduction band edge E_c 161, the valence band edge E 162, and the Fermi level E_F 163 as a function of energy. The energy axis is represented by 198 while the depth (along the growth direction) is represented by axis 199. Region 173 is the depletion region between the p-type region and the n-type region, and includes all of layer 103, plus the portions of layer 102 and 105 that are depleted due to the combination of the built-in voltage of the pin diode and the externally applied bias. Region 171 is the quasi-neutral n-type side of the device and includes a portion of layer 102 and all of 101, and region 175 is the quasi neutral portion of layer 105. Layer 105 includes a surface depletion region 176 whereby surface states 145 act to deplete free carriers from the region near the top surface 135 of the pin photodiode .

Included in the diagram are representations of photo-generated electron-hole pairs. For example, the generation of electron 155 and hole 154 by the absorption of a photon in region 176 does not result in any external photo-current, because the electrical field caused by the surface depletion region forces the electron 155 to be accelerated to the surface 135 and become trapped in one of the surface states 145. Subsequent to the trapping of the electron at a surface state, a free hole in region 176 (not necessarily the photo-generated hole 154) will be captured by the surface state and annihilate the electron 155.

Similarly, the reverse process, thermal excitation of electrons from surface states 145 to the conduction band 161 results in the generation of free minority electrons. There is a certain probability that some of these thermally generated minority electrons will surmount the potential barrier in layer 105, enabling them to be swept into the depletion region 173 where they will add to the dark noise current of the device. The probability that thermal generation at the surface will add to the dark noise of the device depends on the details of the surface band bending, including the height of the potential barrier at surface 135, and the magnitude and direction of the electrical fields near surface 135. Absorption of a photon in region 173 is schematically represented by the photo-generated electron 152 and photo-generated hole 153. The electrical field in region 173 causes the electron 152 to be accelerated and collected by the n-type side of the device, resulting in an external photo-current. Likewise, the electrical field in region 173 causes the hole 153 to be accelerated and collected by the p- type side of the device, resulting in an additional photo-current component .

Reference is now made to Figures 2A-B, which depict a prior-art approach to reducing the effects of surface state recombination on detector performance. These figures are nearly identical to Figures 1A-B, with the addition of a *window" layer 107 with a thickness 127. The window layer 127 is incorporated to reduce the surface state density of layer 105, and is generally formed in a semiconductor with a larger band gap that that used in layer 105 (e.g., layer 107 could be crystalline Alo.4Gao.6As deposited on top of layer GaAs layer 105) . Alternatively, layer 107 could be a dielectric such as SiO^ grown on top of a silicon layer 105 with properties such that surface recombination at the top surface 135 of layer 105 is reduced.

The band diagram for this structure is depicted in Figure 2B. Because of the incorporation of the window layer 107, there is no longer any surface depletion region in layer 105. The surface depletion region 177 (if any exists) is now confined to layer 107, which is caused by surface states 147. Photogeneration in the window layer 107 creates an electron-hole pair, with the electron represented by 155 and the hole represented by 156. The presence of surface states in layer 107 will increase the probability that electron 155 will be trapped at surface 137, and subsequently recombine with a hole from regions 177 or 175. Similarly, the reverse process, thermal excitation of electrons from surface states 147 to the conduction band 161 results in the generation of free minority electrons. There is a certain probability that some of these thermally generated minority electrons will surmount the potential barrier in layer 107, enabling them to be swept through layer 105 into the depletion region 173 where they will add to the dark noise current of the device.

The probability that thermal generation at the surface will add to the dark noise of the device depends on the details of the surface band bending, including the height of the potential barrier at surface 137, and the magnitude and direction of the electrical fields near surface 137. Because the band-gap of layer 107 is larger than that of 105, there will be a range of wavelengths where absorption in the window layer 107 is transparent or semi-transparent, resulting in an increase in the quantum efficiency of the device compared with the structure presented in Figure 1A. For example, absorption of a photon in layer 105 causes electron 157 and hole 154 to be generated. The wide band-gap window layer 107 introduces a potential barrier 187 that blocks the transport of electrons from layer 105 to layer 107, thereby preventing recombination of electrons generated in layer 105. The photogenerated electron will diffuse toward the depletion region 173, which will accelerate the electron toward the n-type side of the device, creating an external photo- current. Similarly, electron 152 and hole 153 generated in the intrinsic layer 103 will also create -an external photo-current.

Reference is now made to Figures 3A-B, which illustrate yet another prior-art technique to reduce the effects of surface recombination by tailoring the doping profile to reduce the thickness of the surface depletion region and provide an additional potential barrier to reduce the transport of photo- generated holes to the surface states. This structure is identical to that presented in Figures 1A-B, with the addition of a p++ cap layer 109 of thickness 129 deposited on top of layer 105. The heavily doped cap layer 109 can, in some cases, be effective in reducing recombination due to surface states 149 at the top surface 139. The addition of very heavy doping has two effects. The first effect is to reduce the extent of the surface depletion region 179, and hence reduce the capability of the surface depletion region to capture photo-generated electrons that are generated deeper than the surface depletion region 179. In addition, the junction between the very heavily doped p++ cap layer 109 and the lighter doped p-type semiconductor layer 105 results in a potential barrier 189 that reduces the flow of minority electrons to the surface 139. In many semiconductors, band gap shrinkage due to heavy doping of layer 109 acts to reduce the potential barrier 189, resulting in a less effective barrier against electron transport toward surface 139, and therefore the doping levels in layers 109 and 105 must be chosen to give sufficiently large potential barrier 189 in the presence of band gap shrinkage.

Region 182 is the depletion region between layers 109 and 105, region 181 is the quasi-neutral portion of layer 109, and region 179 is the surface depletion region of layer 109. In layer 109, photon absorption results in the generation of electron 155 and hole 156. Due to the surface depletion region 179, electron 155 has a high probability of becoming trapped at a surface state 149. Subsequently, the trapped electron will recombine with a hole from regions 179, 181, 182 or 175. Similarly, the reverse process, thermal excitation of electrons from surface states 149 to the conduction band 161, results in the generation of free minority electrons. There is a probability that some of these thermally generated minority electrons will surmount the potential barrier in layer 109, enabling them to be swept through layer 105 into the depletion region 173 where they will add to the dark noise current of the device. That probability depends on the details of the surface band bending, including the height of the potential barrier at the surface 139, and the magnitude and direction of the electrical fields near the surface 139. Photo- generation that occurs deeper than the potential barrier 189 has an enhanced probability of being swept into the i-region and generating external photo-current. For example, photo-generation creates an electron 157 and corresponding hole 154 as shown in the figure. Electron 157 has a reduced probability of recombining at surface 139 due to potential barrier 189. Electron 157 will therefore diffuse toward i-layer 103, where it will be swept toward n-type side of the device, creating a photo-current. Likewise, photo-generation in region 173 creates electron 152, which is swept toward n-type side of the device and hole 153 which is swept toward p-type side of the device, both of which add to the external photo-current.

Reference is now made to Figures 4A - C, showing a preferred embodiment of the present invention. As indicated by Figure 4, this embodiment is grown on a p-type InP substrate 401 with a thickness 421. The backside (bottom) of 401 may be used in some embodiments which trans-illuminate from the backside or put devices or patterns on the backside. Preferably, MBE is used to grow the layer stack, although alternative epitaxial techniques can be used. First, a p-lnP buffer layer 402 with thickness 422 is grown on top of the substrate 401 to provide a high quality superstrate for the growth of the subsequent layers. Next, layer 403 of thickness 423 includes a doping gradient that smoothly transitions between the heavily doped p+InP buffer layer 402 and the intrinsic layer 405. Preferably, layer 403 is p-InP doped with Be. The intrinsic InP layer 405 (either not intentionally doped, or doped with a dopant concentration sufficient to compensate the background doping in order to produce as low a free carrier concentration in the semiconductor as possible) is then grown to a thickness 425. Finally, the n-InP layer 407 is grown where the doping concentration is graded from near the intrinsic doping level to a maximum doping level, which is determined by MBE growth capabilities, device performance, and reliability concerns.

The doping gradient is formed over a thickness 427, and is designed to build in a strong drift field that will accelerate photo-generated minority holes away from the surface 437 toward the intrinsic region, where the minority holes will be collected and produce an external photo-current. The approximate doping profile is shown in Figure 4B, where the abscissa 451 is the depth from the top surface 437 along the growth direction, and the ordinate 452 is a log scale representation of the doping density. Preferably, the n-type doping in layer 407 is silicon and the doping gradient is shown in region 467 and line segment 477. Likewise, layer 405 is not intentionally doped, so no dopant is introduced during the growth of layer 405, as illustrated in region 465 of the graph. Region 463 and line segment 473 of the graph illustrate the Be doping density that will be used during the growth of layer 403. The Be doping profile for the buffer . layer 402 is shown in region 462 and line segment 472. Finally, the doping of the InP substrate is shown in region 461 and line segment 471 (nominally a Zn doped substrate, but other p-type dopants may be suitable for the InP substrate) .

Preferably, electric field is present and continuous throughout the active region of the device, where the active region is defined functionally as the region from which photogenerated carriers are collected. However, it would be acceptable for the field in 461 to have the wrong sign or be discontinuous, if the design did not require the collection of all the photogenerated minority carriers in this region.

Figure 4C shows an approximate band diagram of the pin photodiode structure depicted in Figure 4A. The Figure shows the conduction band edge E,- 461, the valence band edge E_v 462, and the Fermi level E_F 463 as a function of energy. The energy axis is represented by 498, while the depth (along the growth direction) is represented by axis 499. Region 473 is the quasi-neutral p-InP drift region that is used to accelerate minority electrons toward the n-InP side of the device. Region 475 is the depletion region between the p-InP region and the n-InP region, and includes all of layer 405, plus the portions of layers 403 and 407 that are depleted due to the combination of the built-in voltage of the diode and the externally applied bias. Region 477 is the quasi- neutral n-InP drift region that is used to accelerate minority holes toward the p-InP side of the device. Layer 407 includes a surface depletion region 479 whereby surface states 447 act to deplete free carriers from the region near the top surface 437 of the pin photodiode.

Included in the diagram are representations of photo-generated electron-hole pairs. For example, the generation of electron 455 and hole 456 by the absorption of a photon in layer 407 does not result in any external photo-current, because the electrical field caused by the surface depletion region forces the hole 456 to be accelerated to the surface 437 and become trapped in one of the surface states 447. Subsequent to the trapping of the hole at a surface state, a free electron in region 479 (not necessarily the photo-generated electron 455) will be captured by the surface state and annihilate the captured hole 456. Similarly, the reverse process, thermal excitation of holes from surface states 447 to the valence band 462 results in the generation of free minority holes. There is a certain probability that some of these thermally generated minority holes will surmount the potential barrier in region 479, enabling them to be swept through region 477 into the depletion region 475, where they will add to the dark noise current of the device. The probability that thermal generation at the surface will add to the dark noise of the device depends on the details of the surface band bending, including the height of the potential barrier at the surface 437, and the magnitude and direction of the electrical fields near the surface 437.

Absorption of a photon in region 477 is schematically represented by the photo-generated electron 457 and photo-generated hole 454. The drift field in region 477 causes hole 454 to be accelerated toward the p-side of the device, away from surface 437, resulting in an external photo-current. Likewise, absorption of a photon in region 473 is schematically represented by the photo-generated electron 452 and the photo-generated hole 453. The drift field in region 473 causes electron 452 to be accelerated toward the n- side of the device, resulting in an additional photo-current component .

Finally, an example of absorption of a photon in the intrinsic layer 405 results in the generation of hole 458 and electron 459. In this case, the electrical field in intrinsic region 475 accelerates the electron toward the n-side of the device and the hole toward the p-side of the device, generating an additional components of the photo-current.

The embodiment depicted in Figure 4A has several advantages over the prior art, including a reduction in the thickness of the surface depletion region 479 due to the use of hyperdoping techniques that enable the doping at the surface region to be increased relative to prior art approaches. Furthermore, the use of InP materials and n-type doping is advantageous for three reasons: first, the surface state density in InP is relatively small compared to a semiconductor such as GaAs; second, the surface states that do exist in InP tend to pin the Fermi level near the conduction band, which will minimize the amount of surface band bending for n-InP; and finally, it is possible to achieve strongly degenerate doping in n-InP, which enables larger drift regions to be achieved because the effects of band gap shrinkage are offset at strongly degenerate doping levels, due to a band gap widening effect caused by the Fermi level pushing into the conduction band of the device.

Reference is now made to Figures 5A-C, showing an alternative embodiment of the present invention. This alternative embodiment is grown on a p-type InP substrate 501 with a thickness 521. First, a p+InP buffer layer 502 with thickness 522 is grown on top of the substrate 501 to provide a high quality superstrate for the growth of the subsequent layers. The intrinsic InP layer 505 (not intentionally doped) is then grown to a thickness 525. Finally, the n-InP layer 507 is grown, and the doping concentration is graded from near the intrinsic doping level to a maximum doping level, which is determined by MBE growth capabilities, device performance, and reliability concerns. This doping gradient is formed over a thickness 527, and is designed to build in a strong drift field that will accelerate photogenerated minority holes away from surface 537 toward the intrinsic region, where the minority holes will be collected to produce an external photo-current.

An approximate doping profile is shown in Figure 5B, where the abscissa 551 is the depth from the top surface 537 along the growth direction, and the ordinate 552 is a log scale representation of the doping density. In this embodiment, the n- type doping in layer 507 is preferably silicon, and the doping gradient is shown in region 567 and line segment 577. Layer 505 is not intentionally doped, so no dopant is introduced during the growth of layer 505, as illustrated in region 565 of the graph. Next, the Be doping profile for the buffer region is shown in region 562 and line segment 572. Finally, the doping of the InP substrate is shown in region 561 and line segment 571 (nominally a Zn doped substrate, but other p-type dopants may be suitable for the InP substrate) . This structure differs from the previous structure shown in Figures 4A-C only by the deletion of layer 403, which may simplify epitaxy and allow different design tradeoffs to be achieved. Figure 5C shows an approximate band diagram of the pin photodiode structure depicted in Figure 5A. Figure 5C shows the conduction band edge E 561, the valence band edge E_v 562, and the Fermi level E_F 563 as a function of energy. The energy axis is represented by 598, while the depth (along the growth direction) is represented by axis 599. Region 573 is the quasi-neutral portion of the p-InP buffer. Region 575 is the depletion region between the p-InP region and the n-InP regions, and includes all of layer 505, plus the portions of layer 502 and 507 that are depleted due to the combination of the built-in voltage of the diode and the externally applied bias. Region 577 is the quasi- neutral n-InP drift region that is used to accelerate minority holes toward the p-InP side of the device. Layer 507 includes a surface depletion region 579, whereby surface states 547 act to deplete free carriers from the region near the top surface 537 of the pin photodiode.

Included in the diagram are representations of photo-generated electron-hole pairs. For example, the generation of electron 555 and hole 556 by the absorption of a photon in region 579 does not result in any external photo-current, because the electrical field caused by the surface depletion region forces the hole 556 to be accelerated to the surface 537 and become trapped in one of the surface states 547. Subsequent to the trapping of the hole at a surface state, a free electron in layer 507 (not necessarily the photo-generated electron 555) will be captured by the surface state and annihilate the captured hole 556. Similarly, the reverse process, thermal excitation of holes from surface states 547 to the valence band 562 results in the generation of free minority holes. There is a certain probability that some of these thermally generated minority holes will surmount the potential barrier in region 579, enabling them to be swept through region 577 into the depletion region 575, where they will add to the dark noise current of the device. This probability depends on the details of the surface band bending, including the height of the potential barrier at surface 537, and the magnitude and direction of the electrical fields near surface 537.

Absorption of a photon in region 577 is schematically represented by the photo-generated electron 557 and photo-generated hole 554. The drift field in region 577 causes hole 554 to be accelerated toward the p-side of the device and away from surface 537, resulting in an external photo-current. An example of absorption of a photon in the intrinsic layer 505 results in the generation of hole 558 and electron 559. The electrical field in intrinsic region 575 accelerates the electron toward the n-side of the device and the hole toward the p-side of the device, generating an additional component of the photo-current. Finally, for generation in region 573, some of the photo-generated electrons will diffuse toward depletion region 575, for example, photoelectron 552 may diffuse toward the n-side of the device, resulting in an additional photo-current component, and associated photo-generated hole 553 remains in region 573.

Reference is now made to Figures 6A-C, showing an alternative embodiment of the present invention. This alternative embodiment is grown on a semi-insulating GaAs substrate 601 with a thickness 621. First, a n-GaAs buffer layer 602 with thickness 622 is grown on top of substrate 601 to provide a high quality superstrate for the growth of the subsequent layers. The intrinsic GaAs layer 605 (not intentionally doped) is then grown to a thickness 625. Finally, the p-GaAs layer 607 is grown where the doping concentration is graded from near the intrinsic doping level to a maximum doping level which is determined by MBE growth capabilities, device performance, and reliability concerns. This doping gradient is formed over a thickness 627, and is designed to build in a strong drift field that will accelerate photogenerated minority electrons away from surface 637 toward the intrinsic layer, where the minority electrons will be collected and produce an external photo-current.

An approximate doping profile is shown in Figure 6B, where the abscissa 651 is the depth from top surface 637 along the growth direction, and the ordinate 652 is a log scale representation of the doping density. In this embodiment, the p-type doping in layer 607 is preferably Be and the doping gradient is shown in region 667 and line segment 677. Likewise, layer 605 is not intentionally doped, so no dopant is introduced during the growth of layer 605, as illustrated in region 665 of the graph. Next, the Si doping profile for the buffer layer is shown in region 662 and line segment 672. Finally, the GaAs substrate is shown in region 661 and is undoped, so no corresponding doping is shown in Figure 6B.

Figure 6C shows an approximate band diagram of the pin photodiode structure depicted in Figure 6A. The figure shows the conduction band edge E_c 661, the valence band edge E_v 662, and the Fermi level E_F 663 as a function of energy. The energy axis is represented by 698, while the depth (along the growth direction) is represented by axis 699. Region 671 shows the depletion region between the substrate 601 and the n-contacting layer 602. Region 673 is the quasi-neutral portion of the n-GaAs contact layer. Region 675 is the depletion region between the n-GaAs region and the p-GaAs regions, and includes all of layer 605, plus the portions of layers 602 and 607 that are depleted due to the combination of the built-in voltage of the diode and the externally applied bias. Region 677 is the quasi-neutral p-GaAs drift region that is used to accelerate minority holes toward the n-GaAs side of the device. Layer 607 includes a surface depletion region 679, whereby surface states 647 act to deplete free carriers from the region near the top surface 637 of the pin photodiode.

Included in the diagram are representations of photo-generated electron-hole pairs. For example, the generation of electron 655 and hole 656 by the absorption of a photon in region 679 does not result in any external photo-current, because the electrical field caused by the surface depletion region forces the electron 655 to be accelerated to the surface 637 and become trapped in one of the surface states 647. Subsequent to the trapping of the electron at the surface states, a free hole in layer 607 (not necessarily the photo-generated hole 656) will be captured by the surface state and annihilate the captured electron 655. Similarly, the reverse process, thermal excitation of electrons from surface states 647 to the conduction band 661 results in the generation of free minority electrons. There is a certain probability that some of these thermally generated minority electrons will surmount the potential barrier in region 679, enabling them to be swept through region 677 into the depletion region 675 where they will add to the dark noise current of the device. The probability that thermal generation at the surface will add to the dark noise of the device depends on the details of the surface band bending, including the height of the potential barrier at the surface 637, and the magnitude and direction of the electrical fields near the surface 637.

Absorption of a photon in region 677 is schematically represented by the photo-generated electron 657 and photo-generated hole 654. The drift field in region 677 causes electron 657 to be accelerated toward the n-side of the device and away from surface 637, resulting in an external photo-current. Likewise, generation in region 673 is schematically represented by photo-generated electron 652 and photo-generated hole 653. Some of the photogenerated holes will diffuse toward the depletion region 675, for example hole 653 may diffuse toward the p-side of the device, resulting in an additional photo-current component, and associated photo-generated electron 652 remains in region 673. Finally, an example of absorption of a photon in the intrinsic layer 605 results in the generation of hole 658 and electron 659. The electrical field in intrinsic region 675 accelerates the electron 659 towards the n-side of the device and the hole 658 towards the p-side of the device, generating additional photo- current components.

Reference is now made to Figures 7A-C, showing another alternative embodiment of the present invention. This alternative embodiment is grown on a p-type InP substrate 701 with a thickness 721. First, a p+InP buffer layer 702 with thickness 722 is grown on top of the substrate 701 to provide a high quality superstrate for the growth of subsequent layers. Next, a p+InGaAs layer 703 is grown to a thickness 723. The intrinsic InGaAs layer 705A (not intentionally doped) is then grown to a thickness 725A. Next, layer 705B is grown to a thickness 725B which provides a smooth grading of the band gap from InGaAs to InP by smoothly grading InGaAsP from InGaAs to InP. Finally, the n-InP layer 707 is grown where the doping concentration is graded from near the intrinsic doping level to a maximum doping level which is determined by MBE growth capabilities, device performance, and reliability concerns. This doping gradient is formed over a thickness 727, and is designed to build in a strong drift field that will accelerate photo-generated minority holes away from the surface 737 toward the intrinsic layer, where the minority holes will be collected and produce an external photo-current.

An approximate doping profile is shown in Figure 7B, where the abscissa 751 is the depth from the top surface 737 along the growth direction, and the ordinate 752 is a log scale representation of the doping density. In this embodiment, the n- type doping in layer 707 is preferably silicon and the doping gradient is shown in region 767 and line segment 777. Likewise, layers 705A and 705B are not intentionally doped, so no dopant is introduced during the growth of these layers, as illustrated in region 765 of the graph. Next, the Be doping profile for layers 702 and 703 is shown in region 762 and line segment 772. Finally, the doping of the InP substrate is shown in region 761 and line segment 771 (nominally a Zn doped substrate, but other p-type dopants may be suitable for the InP substrate) . This structure differs from the previous structure shown in Figure 5A-C by the addition of InGaAs the intrinsic and p-type layers of the device, enabling the device to be sensitive to photons with energies smaller than the band gap of InP, thus extending the wavelength performance of the device to 1.7 μm and beyond. This produces a truly broadband detector capable of detecting photons with wavelengths less than 200 nm to beyond 1700 nm.

Figure 7C shows an approximate band diagram of the pin photodiode structure depicted in Figure 7A. The figure shows the conduction band edge E_c 761, the valence band edge E_v 762, and the Fermi level E_F 763 as a function of energy. The energy axis is represented by 798, while the depth (along the growth direction) is represented by axis 799. Region 773 is the quasi-neutral portion of the p-InGaAs layer 703. Region 775 is the depletion region between the p-InP region and the n-InP regions, and includes all of layers 705A and 705B, plus the portions of layer 703 and 707 that are depleted due to the combination of the built-in voltage of the diode and the externally applied bias. Region 777 is the quasi-neutral n-InP drift region that is used to accelerate minority holes toward the p-InP side of the device. Layer 707 includes a surface depletion region 779, whereby surface states 747 act to deplete free carriers from the region near the top surface 737 of the pin photodiode.

Included in the diagram are representations of photo-generated electron-hole pairs. For example, the generation of electron 755 and hole 756 by the absorption of a photon in region 779 does not result in any external photo-current, because the electrical field caused by the surface depletion region forces the hole 756 to be accelerated to the surface 737 and become trapped in one of the surface states 747. Subsequent to the trapping of the hole at the surface states, a free electron in layer 707 (not necessarily the photo-generated electron 755) will be captured by the surface state and annihilate the captured hole 756. Similarly, the reverse process, thermal excitation of holes from surface states 747 to the valence band 762 results in the generation of free minority holes. There is a certain probability that some of these thermally generated minority holes will surmount the potential barrier in region 779, enabling them to be swept through region 777 into the depletion region 775, where they will add to the dark noise current of the device. The probability that thermal generation at the surface will add to the dark noise of the device depends on the details of the surface band bending, including the height of the potential barrier at the surface 737, and the magnitude and direction of the electrical fields near the surface 737. Absorption of a photon in region 777 is schematically represented by photo-generated electron 757 and photo-generated hole 754. The drift field in region 777 causes hole 754 to be accelerated toward the p-side of the device and away from surface 737, resulting in an external photo-current. An example of absorption of a photon in the depletion region 775 results in the generation of hole 758 and electron 759. The electrical field in the depletion region 775 accelerates the electron toward the n-side of the device and the hole toward the p-side of the device, generating an additional component of the photo-current. Finally, absorption in region 773 is schematically represented by photo-generated electron 752 and photo-generated hole 753. Most of the photo-generated electrons will diffuse toward the depletion region 775 and away from the p-InP buffer 702, because the heterojunction barrier 789 prevents transport of electrons toward the substrate. For example, photoelectron 752 may diffuse toward the n-side of the device, resulting in an additional photo-current component, and associated photogenerated hole 753 remains in region 773.

Reference is now made to Figures 8A-C, showing another alternative embodiment of the present invention. This alternative embodiment is grown on a semi-insulating GaAs substrate 801 with a thickness 821. First, a n--GaAs buffer layer 802 with thickness 822 is grown on top of the substrate 801 to provide a high quality superstrate for the growth of the subsequent layers. The intrinsic GaAs layer 805 (not intentionally doped) is then grown to a thickness 825. Finally, a p-type AlGaAs layer 807 is grown, where the Al mole fraction is graded from p+GaAs on the side nearest layer 805 to p- Alo._--₅Gao.75As at surface 837. This compositional gradient is formed over a thickness 827, and is designed to build in a strong effective drift field that will accelerate photo-generated minority electrons away from the surface 837 toward the intrinsic layer, where the minority electrons will be collected and produce an external photo- current .

An approximate doping profile is shown in Figure 8B, where the abscissa 851 is the depth from top surface 837 along the growth direction, and the ordinate 852 is a log scale representation of the doping density. In this embodiment, the p-type doping in layer 807 is preferably Be and the doping profile is shown in region 867 and line segment 877. Likewise, layer 805 is not intentionally doped, so no dopant is introduced during the growth of layer 805, as illustrated in region 865 of the graph. Next, the Si doping profile for the buffer layer is shown in region 862 and line segment 872. Finally, the GaAs substrate is shown in region 861 and is undoped, so no doping is shown in Figure 8B.

Figure 8C shows an approximate band diagram of the pin photodiode structure depicted in Figure 8A. The figure shows the conduction band edge E_c 861, the valence band edge E_v 862, and the Fermi level E_F 863 as a function of energy. The energy axis is represented by 898, while the depth (along the growth direction) is represented by axis 899. Region 871 shows the depletion region between the substrate 801 and the n-contacting layer 802. Region 873 is the quasi-neutral portion of the n-GaAs contact layer. Region 875 is the depletion region between the n-GaAs region and the p-GaAs regions, and includes all of layer 805, plus the portions of layers 802 and 807 that are depleted due to the combination of the built-in voltage of the diode and the externally applied bias. Region 877 is the quasi-neutral p-GaAs drift region that is used to accelerate minority holes toward the n-GaAs side of the device. Layer 807 includes a surface depletion region 879, whereby surface states 847 act to deplete free carriers from the region near the top surface 837 of the pin photodiode .

Included in the diagram are representations of photo-generated electron-hole pairs. For example, the generation of electron 855 and hole 856 by the absorption of a photon in region 879 does not result in any external photo-current, because the electrical field caused by the surface depletion region forces the electron 855 to be accelerated to the surface 837 and become trapped in one of the surface states 847. Subsequent to the trapping of the electron at the surface states, a free hole in layer 807 (not necessarily the photo-generated hole 856) will be captured by the surface state and annihilate the captured electron 855. Similarly, the reverse process, thermal excitation of electrons from surface states 847 to the conduction band 861 results in the generation of free minority electrons. There is a certain probability that some of these thermally generated minority electrons will surmount the potential barrier in region 879, enabling them to be swept through region 877 into the depletion region 875, where they will add to the dark noise current of the device. The probability that thermal generation at the surface will add to the dark noise of the device depends on the details of the surface band bending, including the height of the potential barrier at the surface 837, and the magnitude and direction of the electrical fields near the surface 837. Absorption of a photon in region 877 is schematically represented by the photo-generated electron 857 and photo-generated hole 854. The drift field in region 877 causes the electron 857 to be accelerated toward the n-side of the device and away from surface 837, resulting in an external photo-current. For generation in region 873, some of the photo-generated holes will diffuse toward depletion region 875, for example, hole 853 may diffuse toward the p-side of the device, resulting in an additional photo- current component, and associated photo-generated electron 852 remains in region 873. Finally, an example of absorption of a photon in intrinsic layer 805 results in the generation of hole 858 and electron 859. The electrical field in intrinsic region 875 accelerates- the electron toward the n-side of the device and the hole toward the p-side of the device, generating additional photo-current components.

Note that while the above-described embodiment (s) may incorporate smooth grading of the alloy composition (and hence smooth grading of the band gap) , those skilled in the art will recognize that the alloy composition (s) could be changed in a stepwise manner. Furthermore, alternative grading techniques, such as digital superlattice grading, may also be used to effectively grade the alloy compositions.

This discussion underscores the fact that the invention has application to other materials. Indeed, numerous commercially or scientifically important materials can be optimized for drift- dominated detectors with distinctive, useful characteristics, utilizing the teachings of this disclosure, along with the knowledge and experience that a person skilled in the art would possess. Several illustrative examples follow.

InP is fast, and has a Fermi level pinned just below the conduction band edge. Therefore, n-type InP can be designed to have efficient hole collection, enabling detection of wavelengths from below 200 nm to 950 nm, with low dark current.

InP/InGaAs with n-type InP at the surface also is capable of fast drift-dominated transport of holes, and has its Fermi level pinned near the conduction band. The inclusion of InGaAs ( e . g. , within the depletion region of the device) would enable improved long-wavelength sensitivity, enabling a device to detect wavelengths from below 200 nm to around 1700 nm, with slightly increased dark current due to the higher generation rate in the smaller band gap InGaAs regions.

Polycrystalline InP on silicon is also capable of fast drift- dominated transport of holes, with the Fermi level pinned near the conduction band. Using fast, drift-dominated transport mitigates the effects of the crystalline defects and grain boundaries present in polycrystalline material, resulting in better detector performance. Such polycrystalline InP on silicon drift dominated detectors would provide a path to high-speed monolithic optoelectronic detector integration, enabling a single chip solution to an optical communications reciever.

InAs is fast, and has its Fermi level pinned in the conduction band, like InP. Therefore, a detector using n-InAs at the surface would be expected to collect holes efficiently from UN wavelengths out to several μm, but with a somewhat higher dark current due to the narrow band-gap InAs.

GaSb is extremely fast, and has its Fermi level pinned near the valence band. Therefore, it is potentially useful as a p-type detector, where minority electrons are collected for UV to - 1.5 μm wavelengths. Because such a device uses electron transport, even faster response might be expected. Furthermore, such a device may be advantageous for APDs that favor electron multiplication over hole multiplication, and thereby achieve low noise . AlGaAs, GalnP, GalnAsP, and related GaAs-based materials can be fast or slow, depending upon particulars. Because these are wider band-gap materials, they will be transparent to the longer (more red) wavelengths, and therefore detect a narrower gamut, but may also be lower noise, because wider band-gap materials generally exhibit lower thermal generation rates . p++GaAs can be doped to saturate its surface states, and heavily doped or hyperdoped, graded p-GaAs can be used to create a drift field. There are commercial advantages to a GaAs-based UV detector, such as better substrates, lower cost, and higher integration levels than InP- based detectors. The use of GalnP or GalnAsP may also be advantageous, because these materials exhibit a wider band gap, which should result in reduced dark currents, and therefore yield higher sensitivity photodetectors.

GaP appears to have a low density of surface states, and also is expected to exhibit an extremely low dark current, due to its wide band-gap. GaP is nearly lattice-matched to silicon, simplifying integration with silicon substrates and circuitry. GaP is an indirect band-gap semiconductor, with very low absorption coefficients for most visible wavelengths, making it possible to build (particularly with double grading) devices capable of detecting a very restricted range of wavelengths. The hydrophobicity of GaP makes it suitable for non-hermetic applications .

In all of these materials, hyperdoping can be combined with band- gap grading to achieve better performance. Examples include p++ AlGaAs graded to p++ GaAs, where the doping is held constant, or p++ AlGaAs graded to p+ GaAs, where both the doping and the compositional band-gap are graded. Reference is now made to Figures 9A-B. Figure 9A shows a layer stack of a structure grown in accordance with the invention. This device structure was grown by MBE, using Be for the p-type doping and Si for the n-type doping. First, a p+InP buffer layer 902 was grown on a p+ InP substrate 901. For this device, layer 902 is p+InP doped to IxlO¹⁹ cm^-3, with a thickness 922 of 0.5 μm. Next, layer 905 is a lightly doped p-InP layer, doped 5xl0¹⁷ cm^"3 with a thickness 925 of 1.0 μm. Note that layer 905 takes the place of the intrinsic layer 505 in Figure 5, and was not optimized to achieve optimal performance. Finally, layer 907 was grown on top of layer 905 using n-InP, graded from IxlO¹⁶ cm^-3 on the substrate 901 side to 5xl0¹⁹ cm^-3 at surface 937. Layer 907 was grown to a thickness 927 of 0.2 μm.

This structure was fabricated into large area photodiodes by using a backside contact to substrate 901 and a topside contact to surface 937 using metalization. The internal quantum efficiency (IQE) of the device was determined using lasers at wavelengths of 632 nm, 532 nm, 325 nm, and 253 nm, and the incident and reflected power were measured, as well as the detected photocurrent . The IQE was determined by estimating the number of photons absorbed (calculated from the reflected power subtracted from the incident power) , and the detected photocurrent. The IQE is the ratio of detected electrons to the absorbed photons. The IQE is plotted in Figure 9B, with the wavelength along^' the abscissa axis 952 and IQE along the ordinate axis 951. Curve 961 is the measured IQE for devices fabricated from the layer structure shown in Figure 9A.

Figure 9C shows a theoretical analysis of the IQE of the device shown in Figure 9A, using the known absorption coefficient of InP. The calculations assumes that all photo-generation inside the active region of the device (including layers 907, 905, and a substantial fraction layer 902) is collected, except for any photo-generation in the dead layer near the surface. The wavelength is shown along the abscissa axis 999 and the IQE along the ordinate axis 998. Curve 981 is the calculated IQE for a conventional pin photodiode in InP with a dead layer thickness of ~500 A. Curve 982 is the calculated IQE for InP D3 device shown in Figure 9A with a dead layer thickness of 80 A. The experimental data 961 shows reasonable agreement with the theoretical curve 982, which is excellent considering that the theoretical curve 982 is only an approximation of the expected performance. The heavy doping near the surface may cause the absorption coefficient to change, and some photo-generated holes in the dead layer may be collected by the active region of the device, which would modify the theoretical curve 982.

As those skilled in the art will appreciate, a significant effort has been made to provide much more than just process recipes in this disclosure. Instead, with the hope of facilitating a complete appreciation of the full scope the invention and the numerous applications of these teachings to the design and fabrication of drift dominated detectors, the inventors have attempted to explain the underlying physics in detail. The reader should understand, however, that the drift dominated detector is a new device structure, and there exists a possibility that certain of the inventors' theories of operation may not be 100% correct. It is important to note that errors or gaps in the understanding of how or why an invention works do not undermine the validity of a patent. See Newman v. Quigg, 720 F.2d 1565, 1570 (Fed. Cir. 1983) (^xx[I]t is axiomatic that an inventor need not comprehend the scientific principles on which the practical effectiveness of his invention rests.")

Furthermore, notice is hereby given that the applicants intend to seek, and ultimately receive, claims to all aspects, features and applications of the current invention, both through the present application and through continuing applications, as permitted by 35 U.S.C. §120. Accordingly, no inference should be drawn that applicants have surrendered, or intend to surrender, any potentially patentable subject matter disclosed in this application, but not presently claimed. In this regard, potential infringers should specifically understand that applicants may have one or more additional applications pending, that such additional applications may contain similar, different, narrower or broader claims, and that one or more of such additional applications may be designated as ^xxnot for publication prior to grant."

Claims

We claim:

1) A method for making a drift-dominated detector, comprising: incorporating one or more compound semiconductor material (s), wherein at least one of said material (s) has a doping concentration of at least 1 x 10¹⁸ cm^"3 and a Fermi level pinned near its majority-carrier band edge; and forming one or more pn or pin junction (s) in an active region of said semiconductor material (s), the active region representing that portion of the detector from which photogenerated minority carriers are collected, and an electric field being present at substantially all the portions of said active region.

2) A method as in Claim 1, wherein incorporating includes hyperdoping.

3) A method as in Claim 1, wherein said doping is degenerate.

4) A method as in Claim 1, wherein said doping concentration is above 5 x 10¹⁸ cm^-3.

5) A method as in Claim 1, wherein said doping concentration is above 1 x 10¹⁹ cm^-3.

6) A method as in Claim 1, wherein said doping concentration is above 5 x 10¹⁹ cm^"3.

7) A method as in Claim 1, wherein said doping concentration is above 1 x 10²⁰ cm^"3. 8) A method as in Claim 1, wherein said semiconductor material (s) include InP.

9) A method as in claim 1, wherein said semiconductor material (s) include GaP.

10) A method as in claim 1, wherein said material (s) include one or more composition (s) of InGaAlAs, InGaAs, AlGaP, SiC or InGaAsP that is/are substantially lattice-matched to InP or InAs.

11) A method as in claim 8, wherein said material (s) include one or more compositions of AlN, GaN, AlGaN, or AlGalnN, substantially lattice-matched to GaN.

12) A method as in Claim 1, wherein said photogenerated minority carriers are electrons.

13) A method as in Claim 12, wherein said material (s) include heavily doped p-type material (s) .

14) A method as in claim 12, wherein said material (s) include GaAs and one or more compositions of SiC, GaSb, or Al_xGaι__xAs or Al_xGaι-_xInP, where 0 < x < 1.

15) A method as in claim 12, wherein said one or more of 'said material (s) is/are substantially lattice-matched to GaAs.

16) A method as in Claim 1, wherein said electric field is continuous and has a sign suitable for drifting said minority carriers across said pn or pin junction (s) . 17) A method as in Claim 16, wherein said electric field has sufficient strength to accelerate said minority carriers to their saturated drift velocity.

18) A method as in Claim 16, wherein said field has a strength between 100 and 1000 V/cm.

19) A method as in Claim 1, wherein one or more of said compound semiconductor material (s) is/are grown on a substrate formed of a crystallographically distinct material.

20) A method as in Claim 19, wherein said crystallographically distinct material is silicon.

21) A method for reducing minority carrier recombination by surface or interface states in a semiconductor device, said device having a dead layer near a surface or interface and a depletion region in said device, said method comprising: reducing said dead layer extent, reducing band bending, and drifting minority carriers away from said dead layer and toward said depletion region.

22) A method as in Claim 21, wherein said minority carriers are photogenerated.

23) A method as in Claim 22, wherein more than 10% of said minority carriers are photogenerated within 50 nm of said surface or interface. 24) A method as in Claim 22, wherein more than 50% of said minority carriers are photogenerated within 50 nm of said surface or interface.

25) A method for constructing a semiconductor device having minority carrier recombination at surface or interface states, comprising: reducing the extent of a surface or interface dead layer, reducing band bending, and/or drifting minority carriers away from said dead layer and toward a depletion region.

26) A method as in Claim 25, wherein said semiconductor device is heavily doped at its surface in order to lessen the penetration of the surface depletion region into the active region.

27) A method as in Claim 25, wherein reducing band bending involves heavy doping at the surface or interface to occupy a significant fraction of the surface states and thereby drive the Fermi level closer to the majority carrier band edge.

28) A method as in Claim 25, wherein drifting includes widening the effective band-gap or countering band-gap shrinkage by highly doping the surface or interface until it is degenerately doped.

29) A method as in Claim 25, wherein drifting entails engineering an effective electrical field to match or counter diffusion fields using a high doping gradient that extends beyond the surface or interface band bending region.

30) A method as in Claim 25, wherein drifting entails grading dopant concentration in order to produce an intrinsic field. 31) A method as in Claim 25, wherein an electric field between the surface or interface and the depletion region accelerates minority carriers to their saturation drift velocity.

32) A method as in Claim 25, wherein drifting entails compositional grading of alloy (s) in order to produce an intrinsic field.

33) A method as in Claim 25, further comprising exploiting materials with non-negligible densities of surface recombination sites, wherein band bending is reduced, or surface or interface recombination effects are reduced, by building electric fields into the device.