Photodiodes with no excess noise

Abstract

Providing various methods, apparatus and systems for photodiodes. [Solution] A linear-mode avalanche photodiode has a superlattice multiplication region that allows only one current carrier type, such as electrons or holes, to accumulate sufficient kinetic energy for impact ionization when biased, and is constructed of a lattice-matched pair of a first semiconductor alloy and a second semiconductor alloy in the superlattice multiplication region, an absorption region, and a semiconductor substrate.

Description

[Technical Field]

[0001] (Related Applications) This application claims priority to and the benefit of International PCT Patent Application PCT / US18 / 41574, entitled "LINEAR MODE AVALANCHE PHOTODIODES WITHOUT EXCESS NOISE," filed July 11, 2018, which is incorporated by reference herein in its entirety as a continuation-in-part application.

[0002] FIELD OF THE DISCLOSURE Embodiments of the present disclosure generally relate to photodiodes. [Background technology]

[0003] The function of a photodetector is to sense incident light and output a current proportional to the incident light flux, i.e., a photocurrent proportional to the incident light flux. Ideally, a photodetector would generate one output electron or hole for every incident photon, with every output electron or hole being the result of an incident photon. In many applications where the incident flux is small, the output current needs to be amplified before it can be used by subsequent electronic devices. Some previous photodetectors to achieve photon counting would increase the photocurrent by 10 to detect a single photon. 4 Double or 10 6 A gain mechanism is required to double the amplitude of the current. In addition to increasing the amplitude of the current, the amplifier also has the disadvantage of adding noise to the current.

[0004] In fact, in many types of photodetectors, the noise current from the electronic amplifier at room temperature is much larger than the current required to detect a stream of single photons (photon counting). The amplifier noise in these types of photodetectors is driven by temperature, so cooling the amplifier to cryogenic temperatures reduces the noise, but this approach is acceptable for a limited number of applications.

[0005] One mechanism for current multiplication is impact ionization, and photodetectors based on this mechanism are known as avalanche photodiodes (APDs). Summary of the Invention [Problem to be solved by the invention]

[0006] Various photodiode methods, apparatus and systems are provided herein. [Means for solving the problem]

[0007] In embodiments, photodiodes, such as linear-mode avalanche photodiodes, can eliminate excess noise by having a superlattice multiplication region that allows only one current carrier type, such as electrons or holes, to accumulate sufficient kinetic energy for impact ionization when biased, where the layers are lattice-matched. The photodiode can be constructed with: i) a lattice-matched pair of first and second semiconductor alloys in the superlattice multiplication region; ii) an absorption region; and iii) a semiconductor substrate. Detectors having multiple photodiodes can be fabricated using these structural layers to have cutoff wavelengths ranging anywhere from 1.7 μm to 4.9 μm, and noise resulting from dark current at a level such that electromagnetic radiation signals having the desired minimum wavelength cutoff can be accurately detected by the photodiodes.

[0008] The photodiode can be constructed with i) a lattice-matched pair of a first semiconductor alloy and a second semiconductor alloy in a superlattice multiplication region, ii) an absorber region, and iii) a semiconductor substrate. The absorber region can also have a superlattice structure of multiple semiconductor alloys. A given photodiode can be constructed with a lattice-matched pair of a first semiconductor alloy and a second semiconductor alloy in a superlattice multiplication region, allowing only one current carrier type to be selected from i) electrons or ii) holes to accumulate sufficient kinetic energy for impact ionization when the photodiode is electrically biased to conduct current. The photodiode will have a third semiconductor alloy forming the absorber region and a semiconductor substrate. The multiplication region and absorber region can be lattice-matched to the semiconductor substrate. Alternatively, the multiplication region and absorber region are stacked on top of each other starting with the semiconductor substrate.

[0009] The alloys in the multiplication region, absorption region, and semiconductor substrate are matched to provide a photodiode having a minimum wavelength cutoff of 1.0 μm to 4.9 μm and noise resulting from dark current at a level such that electromagnetic radiation signals having the desired minimum wavelength cutoff can be accurately detected by the photodiode.

[0010] Many variations of this design are discussed. [Brief explanation of the drawings]

[0011] [Figure 1] FIG. 1 illustrates a block diagram embodiment of a linear-mode APD including a matched superlattice structure (hereinafter referred to as a “matched superlattice structure”) in which excess noise is substantially eliminated by having a superlattice in the valence band when the carriers are holes.

[0012] [Figure 2a] FIG. 2a shows an embodiment of a graph of holes in the valence band of a bulk semiconductor under bias storing kinetic energy with distance.

[0013] [Figure 2b] FIG. 2b shows an embodiment of a graph of a hole in the valence band hopping through a well in a biased superlattice and the energy levels of states in a Wannier-Stark ladder.

[0014] [Figure 3] FIG. 3 shows an embodiment of a graph illustrating the theoretical excess noise figure F as a function of the average gain M and the β / α ratio for electron injection.

[0015] [Figure 4] FIG. 4 shows a graph of a semiconductor band diagram with energy plotted vertically and distance plotted horizontally, and an embodiment of the amount of kinetic energy required to impact ionize and create a new electron-hole pair.

[0016] [Figure 5] FIG. 5 shows a block diagram embodiment of a linear-mode APD including a matched superlattice structure in which excess noise is substantially eliminated by having a superlattice in the conduction band when the carriers are electrons.

[0017] [Figure 6a] FIG. 6a shows an embodiment of a graph of electrons in a bulk semiconductor with an applied electric field being accelerated by the field and storing kinetic energy.

[0018] [Figure 6b] FIG. 6b shows a graphical embodiment of a commensurate superlattice that suppresses impact ionization by controlling localized Wannier-Stark states.

[0019] [Figure 7] FIG. 7 shows a graphical embodiment of a linear-mode APD including a matched superlattice design in which electron multiplication is suppressed but hole multiplication is not suppressed.

[0020] [Figure 8] FIG. 8 shows an embodiment of a diagram of a linear mode APD array including a matched superlattice structure with in-situ memory and parallel column readout.

[0021] [Figure 9] FIG. 9 illustrates an embodiment of a readout circuit schematic having a linear-mode APD including a matched superlattice structure.

[0022] [Figure 10] FIG. 10 shows a block diagram of an exemplary InGaAs-GaAsSb superlattice band diagram.

[0023] [Figure 11A] FIG. 11A shows a flow diagram of an embodiment for constructing a photodiode that eliminates excess noise by superlattice and lattice matching to have a minimum wavelength cutoff of 1.0 μm to 4.9 μm and noise resulting from dark current at a level such that electromagnetic radiation signals having the desired minimum wavelength cutoff can be accurately detected by the photodiode. [Figure 11B] FIG. 11B shows a flow diagram of an embodiment for constructing a photodiode that eliminates excess noise by superlattice and lattice matching to have a minimum wavelength cutoff of 1.0 μm to 4.9 μm and noise resulting from dark current at a level such that electromagnetic radiation signals having the desired minimum wavelength cutoff can be accurately detected by the photodiode.

[0024] While this design is subject to various modifications, equivalents, and alternatives, specific embodiments of this design are shown by way of example in the drawings and will now be described in detail. It is to be understood that this design is not limited to the specific embodiments disclosed, but on the contrary, it is intended to cover all modifications, equivalents, and alternatives using those specific embodiments. DETAILED DESCRIPTION OF THE INVENTION

[0025] In the following description, numerous specific details may be set forth, such as examples of specific data signals, referenced components, and frame numbers, to provide a thorough understanding of the design. However, it will be apparent to those skilled in the art that the design may be practiced without these specific details. In other instances, well-known components or methods are not described in detail but rather shown in block diagrams to avoid unnecessarily obscuring the design. Furthermore, specific references to numbers may be made, such as a first photodiode. However, these specific references to numbers should not be interpreted as a literal order, but rather that the first photodiode is different from the second photodiode. As such, specific details described may be merely examples. These specific details may deviate from the spirit and scope of the design, but are still intended to be included within the spirit and scope of the design. The term “coupled” is defined to mean either directly connected to the component or indirectly connected to the component through another component.

[0026] Figures 1-9 discuss a first set of exemplary photodiode structures and techniques. Figures 10-11B discuss a second set of additional photodiode structures and techniques that build on and incorporate concepts and techniques discussed in the first set.

[0027] Generally, methods, devices, and systems are discussed. Photodiodes, such as linear-mode avalanche photodiodes, can eliminate excess noise by having a superlattice multiplication region that allows only one current carrier type, such as electrons or holes, to accumulate sufficient kinetic energy for impact ionization when biased, where the layers are lattice-matched. The photodiode can be constructed with i) a lattice-matched pair of first and second semiconductor alloys in the superlattice multiplication region, ii) an absorber region, and iii) a semiconductor substrate. The absorber region can also have a superlattice structure of multiple semiconductor alloys. A given photodiode can be constructed with a lattice-matched pair of first and second semiconductor alloys in the superlattice multiplication region that allows only one current carrier type, i) electrons or ii) holes, to be selected to accumulate sufficient kinetic energy for impact ionization when the photodiode is electrically biased to conduct current. The photodiode will have a third semiconductor alloy forming the absorber region and a semiconductor substrate. The multiplication region and the absorber region can be lattice-matched to the semiconductor substrate. Alternatively, the multiplication region and the absorber region are layered together starting with the semiconductor substrate. The alloys in the multiplication region, the absorber region, and the semiconductor substrate are matched to provide a photodiode with a minimum wavelength cutoff between 1.0 μm and 4.9 μm and noise resulting from dark current at a level such that electromagnetic radiation signals having the desired minimum wavelength cutoff can be accurately detected by the photodiode.

[0028] A second set of exemplary photodiode structures and techniques that builds upon and incorporates the concepts and techniques discussed in the first set is discussed below.

[0029] Figure 10 shows a block diagram of an exemplary InGaAs-GaAsSb superlattice band diagram. The photodiode layers 1000 can be constructed so that electrons are largely confined within the InGaAs-GaAsSb layers and holes are largely confined within the GaAsSb layers. The diagram outlines a GaAs1-ySby alloy in which only one current carrier type, such as electrons or holes, interacts with the InxGa1-xAs alloy to accumulate enough kinetic energy to undergo impact ionization when biased.

[0030] Photodiodes such as linear-mode avalanche photodiodes (LM-APDs) can eliminate excess noise by having a superlattice gain region that allows only one current carrier type, such as electrons or holes, to accumulate enough kinetic energy for impact ionization when biased, where the different regions are lattice-matched.

[0031] The photodiode can be constructed of i) a lattice-matched pair of a first semiconductor alloy and a second semiconductor alloy in a superlattice multiplication region, ii) an absorption region, and iii) a semiconductor substrate.

[0032] Photodiodes using these lattice-matching techniques herein can be used to fabricate short-wavelength infrared (SWIR) linear-mode avalanche photodiodes (LM-APDs) with cutoff wavelengths >2.0 μm and operating temperatures >250 K.

[0033] Several exemplary photodiode structures herein can extend linear-mode avalanche photodiodes (LM-APDs) to wavelengths beyond 2 μm at operating temperatures >250K.

[0034] A.In a first exemplary photodiode, the first and second semiconductor alloys can form a lattice-matched pair of semiconductor alloys that make up the multiplication region. This particular multiplication region can be composed of indium gallium antimonide arsenide (InGaAsSb) and aluminum gallium antimonide arsenide (AlGaAsSb). The third semiconductor alloy that makes up the absorption region is InGaAsSb. The semiconductor substrate is composed of gallium antimonide (GaSb). Both regions are lattice-matched to the GaSb substrate.

[0035] The multiplication region is i) In 0.09 Ga 0.91 As 0.08 Sb 0.92 and ii) a first semiconductor alloy of Al 0.14 Ga 0.86 As 0.01 Sb 0.99 In another example, the lattice-matched pair of semiconductor alloys for the multiplication region can be, for example, In. 0.17 Ga 0.83 As 0.16 Sb 0.84 and Al 0.26 Ga 0.74 As 0.02 Sb 0.98 Alternatively, the photodiode structure may have an InGaAsSb absorption region lattice matched to GaSb.

[0036] The superlattice in the multiplication region is lattice matched to the GaSb substrate. Note that an InGaAsSb-AlGaAsSb superlattice in the multiplication region is of particular interest because its larger band offset can lead to higher gain without excess noise.

[0037] Detectors with multiple photodiodes can be fabricated using these structural layers to have cutoff wavelengths ranging anywhere from 1.7 μm to 4.9 μm.

[0038] B. The lattice-matched pair of semiconductor alloys comprising the multiplication region can also include a superlattice having first and second semiconductor alloys of indium gallium arsenide phosphide (InGaAsP) and indium aluminum arsenide (InAlAs). The third semiconductor alloy comprising the absorption region can be an indium gallium arsenide (InGaAs) and gallium antimonide arsenide (GaAsSb) superlattice lattice-matched to an indium phosphide (InP) substrate.

[0039] This photodiode structure replaces the InGaAs absorber region of the other exemplary structures with an InGaAs-GaAsSb Type II superlattice absorber region lattice-matched to an InP substrate. Detectors fabricated using this absorber region can have a cutoff wavelength of at least 2.4 μm. An InGaAsP-InAlAs superlattice multiplication region is matched to the absorber region.

[0040] Generally, a photodiode can be fabricated that is constructed of a lattice-matched pair of i) a first semiconductor alloy and ii) a second semiconductor alloy in a superlattice multiplication region, an absorption region, and iii) a semiconductor substrate.

[0041] The lattice-matched pair of the first and second semiconductor alloys in the superlattice multiplication region is matched to allow only one current carrier type to be selected: i) electrons or ii) holes, to accumulate sufficient kinetic energy for impact ionization when the photodiode is electrically biased to conduct current. The third semiconductor alloy can form the absorber region. The multiplication region and absorber region are also lattice-matched to the semiconductor substrate. Alternatively, the multiplication region and absorber region can be stacked on top of each other starting from the semiconductor substrate.

[0042] The alloys in the multiplication region, absorption region, and semiconductor substrate are lattice-matched (e.g., matched to minimize the amount of dark current) to provide a photodiode constructed with a minimum wavelength cutoff of 1.0 μm to 4.9 μm and noise resulting from dark current at a level such that electromagnetic radiation signals having the desired minimum wavelength cutoff can be accurately detected by the photodiode. The matching is set so that noise resulting from dark current is minimized so that electromagnetic radiation signals can be properly detected by the photodiode.

[0043] It should be noted that dark current can be a relatively small current that flows through a photosensitive device, such as a photodiode, even when no photons enter the device. In embodiments, the alloy in the multiplication region, the absorber region, and the semiconductor substrate are lattice-matched to provide a photodiode constructed with a minimum wavelength cutoff of 4.9 μm or less. In embodiments, the alloy in the multiplication region, the absorber region, and the semiconductor substrate are lattice-matched to provide a photodiode constructed with a minimum wavelength cutoff of at least 3.3 μm.

[0044] A photodiode comprising i) a substrate and ii) a semiconductor layer can be fabricated using a complementary metal-oxide semiconductor (CMOS) process. The photodiode can be fabricated using a CMOS-compatible process using substantially the same fabrication steps to grow both the indium gallium antimonide arsenide (InGaAsSb) layer in the multiplication region and the InGaAsSb layer in the absorption region. Superlattice matching with these semiconductors can achieve absorption at higher maximum cutoff wavelengths, such as 1.7 μm, 3.3 μm, and beyond.

[0045] In an embodiment, the photodiode comprises: i) an In superlattice set in the conduction band; 0.09 Ga 0.91 As 0.08 Sb 0.92and ii) an InGaAsSb absorber region with alternating multiplication and absorber regions grown on a gallium antimonide (GaSb) substrate.

[0046] A different set of semiconductors can be grown on a gallium antimonide substrate to provide photodiodes with maximum cutoff wavelengths such as 1.7 μm, 3.3 μm, and beyond. The gallium antimonide substrate allows for the growth of semiconductors on the substrate that have current carrier absorption at higher wavelengths.

[0047] One or more of the semiconductor regions comprises a superlattice. In one example, two semiconductors in the superlattice structure of the multiplication region are aluminum gallium antimonide (Al gallium arsenide). 0.14 Ga 0.86 As 0.01 Sb 0.99 ) and indium gallium antimonide (In ) with a superlattice set within the conduction band. 0.09 Ga 0.91 As 0.08 Sb 0.92 and a second semiconductor alloy of . Electrons can be the first current carrier set for multiplication. The superlattice structure can be used as the multiplication region in this example and as the absorption region in other exemplary photodiodes.

[0048] It should be noted that the composition of the absorber region is not critical, although it can be made from similar alloys.

[0049] In one example, the absorber region can be the same or similar to the multiplication region, with the indium gallium layer in the absorber region being a similar alloy. In addition, the other alloyed portion of the multiplication region is semiconductor-paired to match this change. Note that the reason for the change to the indium content in the alloy is that it is easier to grow the same layer for both the multiplication and absorber regions.

[0050] An exemplary constructed photodiode may have an InGaAs absorption region with a cutoff wavelength of at least 1.6 μm and an InGaAsP-InAlAs superlattice multiplication region, all lattice matched to an InP substrate.

[0051] The above embodiments of the photodiode can potentially achieve pixel pitches down to 15 μm for high resolution imaging applications.

[0052] Further details of an exemplary linear-mode APD fit to the SWIR band Various approaches can be used to design low-excess-noise LM-APDs with strong photosensitivity to wavelengths beyond 2.0 μm. Two exemplary approaches, i) an InGaAs-GaAsSb type II superlattice absorber in InP, and ii) an InGaAsSb absorber in GaSb, are discussed in more detail below. Both photodiode structures can meet the operating temperature constraints for wavelengths beyond 2 μm, with operating temperatures >250 K. A third photodiode approach, an InGaAs absorber in InP, is also discussed further below. This can be used for imagers sensitive to wavelengths from 1.6 μm to 2.5 μm.

[0053] Photodiodes, including APDs using InGaAs-GaAsSb type II superlattice absorber regions in InP, can absorb electrons and / or holes in the SWIR band. Because the InGaAs-GaAsSb type II superlattice absorber regions are composed of layers lattice-matched to InP, they do not have any of the mismatch problems associated with extended InGaAs.

[0054] An InGaAs-GaAsSb type II superlattice built on an InP substrate can have achievable cutoff wavelengths of 2.4 μm in the short term and 4.2 μm in the long term. Note that in this case, quantum efficiency (QE) decreases with increasing cutoff.

[0055] Photodiodes using InGaAsSb with a Type I superlattice on a GaSb substrate can have achievable cutoff wavelengths of 2.3 μm in the short term and 4.9 μm in the long term. The multiplication region superlattice has a larger band offset for higher gain.

[0056] alignment In the example, the effective bandgap of an InGaAs-GaAsSb superlattice is adjusted by changing the thickness of its two constituent layers. Increasing the thickness of the InGaAs layer lowers the electron miniband, while increasing the thickness of the GaAsSb layer raises the light and heavy hole minibands. Both of these changes decrease the effective bandgap. Because electrons are concentrated in the InGaAs well and holes are concentrated in the GaAsSb inversion well, the overlap between the electron and hole wavefunctions decreases as the layers become thicker and the wavefunctions become more localized within the well. This localization has two effects that reduce the QE. First, it reduces the transition probability that relies on wavefunction overlap, resulting in lower absorption efficiency. Second, it becomes more difficult for carriers to tunnel between wells, which slows carrier transport due to the recombination of many of the photogenerated electron-hole pairs before they can leave the absorption region. A superlattice with 5 nm per layer provided a reasonable compromise to this trade-off, resulting in a photodiode with a cutoff wavelength of 2.39 μm and a QE of 43% at 2.23 μm. Strain-compensated InGaAs-GaAsSb superlattices can be used to circumvent this trade-off, but with little success in obtaining a high QE at longer wavelengths.

[0057] The dark current of lattice-matched InGaAs-GaAsSb superlattice photodiodes is matched to a low bias setting. Here, recombination-generated current dominates and is similar to that of extended InGaAs photodiodes at a similar temperature with a similar bandgap. Lattice-matched InGaAs-GaAsSb superlattice photodiodes do not have the threading dislocations of extended InGaAs photodiodes, but they have high dark currents that may be due to unintentional intermixing at the interface, for example, forming InSb precipitates. Note that unintentional AIP formation in multiplication superlattices can be resolved by optimizing growth conditions, including the timing at which precursor gases are introduced into the growth chamber.

[0058] In an example, an InGaAs-GaAsSb superlattice absorber region can be used for a SWIR LM-APD, designed for a 2.4-μm cutoff and 43% QE, with the absorber superlattice optimized for growth conditions for low dark current.

[0059] InGaAsSb absorption region on GaSb substrate The InGaAsSb absorber region approach uses a spatially direct bandgap semiconductor lattice matched (for high absorption coefficient and high QE) to a GaSb substrate (for low defect density).

[0060] Lattice matching to a GaSb substrate instead of InP would also involve the use of a new multiplication superlattice. 0.17 Ga 0.83 As 0.16 Sb 0.84 and Al 0.26 Ga 0.74 As 0.02 Sb 0.98A multiplication superlattice constructed from InGaAsP-InAlAs has only a conduction band offset. The multiplication superlattice offset is 0.502 eV, compared to 0.399 eV for an exemplary InGaAsP-InAlAs superlattice. Better confinement of electrons in deeper wells allows a higher electric field to be applied to the superlattice before the electrons can escape the well, thereby achieving higher gain.

[0061] A concern with InGaAsSb absorbers is that alloy phase separation occurs at some alloy concentrations. This alloy phase separation, called the miscibility gap, occurs for alloys with bandgap wavelengths between approximately 2.3+ μm and 4.4 μm. InGaAsSb absorbers can use 2.3-μm alloys, which avoids phase separation. To obtain absorbers with direct transitions (in k-space and real space) at longer cutoff wavelengths, the photodiode structure can use several strategies, depending on the wavelength. a. The design can be modified to vary the growth temperature and growth rate to obtain a cutoff wavelength slightly longer than 2.3 μm. b. The design allows for growth of lattice-matched alloys at the long wavelength end of the miscibility gap, obtaining cutoff wavelengths between 4.4 μm and 4.9 μm. c. To obtain a cutoff between 2.3 μm and 4.9 μm, the design can use an absorbing layer consisting of a (Type I) superlattice of alloys at either end of the miscibility gap. The absorbing superlattice will consist of two alloys for which the design is stable, with an effective bandgap determined by the relative thicknesses of the two alloys that mimics any alloy in the miscibility gap.

[0062] In summary, InGaAsSb alloys have the strong absorption properties of direct-gap semiconductors and can be grown lattice-matched to a cutoff of almost 5 μm. The multiplication superlattice is lattice-matched to GaSb. Note that this superlattice has the potential to provide higher gain due to the larger band offset.

[0063] As discussed, several exemplary approaches can be used to extend LM-APDs to wavelengths beyond 2.0 μm at operating temperatures above 250 K. The first exemplary photodiode discussed above has a 2.4-μm wavelength cutoff, requiring only minor design modifications from the other LM-APD structures discussed herein. This photodiode has limited options for achieving a longer cutoff and, thus far, has high levels of dark current for lattice-matched absorption regions. The first exemplary photodiode discussed above has a 2.3 μm wavelength cutoff using a novel multiplication region superlattice, potentially providing higher gain without excess noise. The cutoff wavelength for the second exemplary photodiode approach can be extended to longer wavelengths (4.9 μm) without sacrificing QE.

[0064] It should be noted that pixel pitches down to 15 μm can also be achieved with the LM-APDs of the present invention. The lower sensitivity and temperature of biasing LM-APDs (compared to Geiger-mode APDs) are also important for achieving high yields in large arrays.

[0065] How to grow Photodiodes can be grown in a way that avoids a high density of threading dislocations and therefore very high dark current. All of these growth methods require a series of buffer layers that gradually change the substrate lattice parameters, for example, from those of InP to those of the extended InGaAs alloy, which prevents most threading dislocations from entering the extended InGaAs layer. The combined thickness of these buffer layers can be 5 μm to 10 μm. When grown correctly, the buffer layers significantly reduce the density of threading dislocations through the extended InGaAs, but do not completely eliminate them.

[0066] Before engaging wavelength-extended InGaAs with a superlattice gain region, the design must consider whether the superlattice is located above or below the absorption region. If the InGaAs is above, threading dislocations in the extended InGaAs also penetrate the superlattice, and these are likely to become nucleation centers for premature breakdown. If the superlattice is below, the photogenerated carriers must traverse a thick buffer layer, where many of the dislocations resulting from the lattice mismatch are concentrated. Depending on the details, (1) these dislocations can become generation centers causing large dark currents; (2) they can trap and annihilate photogenerated electrons and holes, preventing their detection; or (3) they can trap one type of photogenerated carrier and re-emit it at a later time, resulting in slow response times and memory effects.

[0067] 11A-11B show a flow diagram of an embodiment for constructing a photodiode that eliminates excess noise by superlattice and lattice matching to have a minimum wavelength cutoff between 1.0 μm and 4.9 μm and noise resulting from dark current at a level such that electromagnetic radiation signals having the desired minimum wavelength cutoff can be accurately detected by the photodiode. A method 1100 for constructing a photodiode can be performed. The exemplary steps can be performed in a different order, and not necessarily all steps, and are not limited to only these steps.

[0068] In step 1102, a photodiode constructed of a lattice-matched pair of a first semiconductor alloy and a second semiconductor alloy in a superlattice multiplication region, an absorption region, and a semiconductor region can be fabricated.

[0069] In step 1104, a lattice-matched pair of a first semiconductor alloy and a second semiconductor alloy in a superlattice multiplication region is formed, allowing only one current carrier type to be selected from i) electrons or ii) holes to accumulate sufficient kinetic energy for impact ionization when the photodiode is electrically biased to conduct current. A third semiconductor alloy can form the absorber region. The multiplication region and absorber region are lattice-matched to the semiconductor substrate; alternatively, the multiplication region and absorber region are layered on top of each other starting from the semiconductor substrate.

[0070] In step 1106, the alloys in the multiplication region, absorption region, and semiconductor substrate are matched (matched to minimize the amount of dark current) to provide a photodiode constructed to have a minimum wavelength cutoff between 1.0 μm and 4.9 μm and noise resulting from dark current at a level such that electromagnetic radiation signals having the desired minimum wavelength cutoff can be accurately detected by the photodiode. The photodiode can be a linear-mode avalanche photodiode with a wavelength cutoff of 3.3 μm or greater. These alloy configurations can have an operating range that responds to all lower wavelengths.

[0071] In step 1108, a photodiode including i) a substrate and ii) a semiconductor layer can be fabricated by a complementary metal-oxide-semiconductor (CMOS) process using substantially the same fabrication steps to grow both the InGaAsSb layer in the multiplication region and the InGaAsSb layer in the absorption region.

[0072] In step 1110, the first semiconductor alloy in the multiplication region can be InGaAsSb with a superlattice set in the conduction band for electron impact ionization, the third semiconductor alloy in the absorption region can be InGaAsSb, and the semiconductor substrate can be GaSb. 0.09 Ga 0.91 As 0.08 Sb 0.92It should be noted that the third semiconductor alloy comprising the absorption region is a similar alloy of InGaAsSb, but does not have the same composition as the multiplication region.

[0073] In step 1112, as discussed, the first semiconductor alloy is paired with a second semiconductor alloy of AlGaAsSb to form the multiplication region. The first semiconductor alloy and the second semiconductor alloy may form a lattice-matched pair of semiconductor alloys that form the multiplication region of InGaAsSb and AlGaAsSb. The second semiconductor alloy of the multiplication region may be Al 0.14 Ga 0.86 As 0.01 Sb 0.99 The alloy may be made of an alloy having the composition:

[0074] In step 1114, the lattice-matched pair of semiconductor alloys that make up the multiplication region is In 0.17 Ga 0.83 As 0.16 Sb 0.84 and Al 0.26 Ga 0.74 As 0.02 Sb 0.98 The third semiconductor that makes up the absorbing layer is InGaAsSb, and both regions are lattice-matched to the GaSb substrate. This structure can achieve an exemplary cutoff wavelength of 2.3 μm.

[0075] In step 1116, the semiconductor alloy comprising the absorption region can be an InGaAs-GaAsSb superlattice lattice-matched to an InP substrate, and the lattice-matched pair of semiconductor alloys comprising the multiplication region can be an InGaAsP-InAlAs superlattice.

[0076] A first set of exemplary photodiode structures and techniques are discussed below.

[0077] The system can employ an array of linear-mode avalanche photodiodes, each including a matched superlattice structure. Each linear-mode avalanche photodiode senses light and outputs a current by producing a gain of 1000 times or more while producing an excess noise figure less than three times the thermal noise present at non-cryogenic temperatures or higher due to the gain from the amplification. The linear-mode avalanche photodiode detects one or more photons of light by using a matched superlattice structure that suppresses impact ionization for a first carrier of the linear-mode avalanche photodiode while achieving at least one of: 1) increasing impact ionization, 2) substantially maintaining impact ionization, and 3) suppressing impact ionization to a lesser extent for a second carrier. The first carriers whose impact ionization is suppressed are either i) electrons or ii) holes, and the second carriers are electrons or holes, respectively. A power supply is used to provide power to the linear-mode avalanche photodiodes.

[0078] 1 and 5 show a linear-mode APD 100, 500 including a cap layer, an absorber layer, a matching superlattice structure in the superlattice multiplication region, and a collector layer. A voltage controller applies a reverse bias to the linear-mode APD 100, 500. The linear-mode APD 100, 500 is configured to produce a potentially exponential gain of 1000 times or more while producing an excess noise figure of less than three times the thermal noise present at non-cryogenic temperatures or higher due to the gain from the amplification, thereby sensing light and outputting a current. The superlattice structure is matched to suppress impact ionization for a first carrier in the linear-mode APD 100, 500 while maintaining impact ionization for the other carrier. The first carrier whose impact ionization is suppressed is either i) an electron or ii) a hole, and the other carrier whose impact ionization is maintained is either an electron or a hole. When the carriers are holes, a superlattice is formed in the valence band (see Figure 1), and when the carriers are electrons, a superlattice is formed in the conduction band (see Figure 5), thereby suppressing the impact ionization of carriers. Non-cryogenic temperatures can be, for example, temperatures higher than -50°C.

[0079] Thus, the linear-mode APD 100, 500 including the matched superlattice structure is configured to suppress impact ionization of the first carrier in the superlattice multiplication region by: i) implementing a superlattice in the valence band when the first carrier is a hole, and ii) implementing a superlattice in the conduction band when the first carrier is an electron. Impact ionization gain can be initiated through the use of a matched superlattice by only the other carrier with a kinetic energy (KE) greater than the bandgap energy (Eg), preventing the first carrier, either an electron or a hole, from accumulating sufficient kinetic energy under a large electric field for impact ionization.

[0080] The superlattice structure and the voltage controller cooperate to apply a fixed or variable electric field to the superlattice structure. The periodic structure of two or more layers of materials forming the commensurate superlattice structure is designed to have at least one quantum state in each well when the superlattice is biased by the voltage controller with an electric field that causes impact ionization to persist only in thick layers of well or barrier semiconductors. The periodic set of materials for the commensurate superlattice structure is further designed to have a large spatial overlap of the wave functions of the quantum states of adjacent wells when under an applied bias, such that a first carrier traverses the superlattice by hopping from a first well to a second well and ends up losing substantially less energy than the kinetic energy required to impact ionize and form a new electron-hole pair with each well-to-well hop, the amount of energy provided by the applied bias. A range of bias from the voltage controller facilitates these objectives.

[0081] The large spatial overlap encourages carriers in a localized state in a particular well to rapidly decay to a lower-energy localized state in a neighboring state. This process is known as hopping. This energy difference is lost as the emission of phonons. Carriers traverse the superlattice by hopping from one well to the next, losing the energy provided by the applied bias as phonons at each stage.

[0082] The matched superlattice structure in the multiplication region can provide band discontinuities (noise-free gain) in only one band by having matched materials that can have hundreds of pairs of nanometer-thick layers that are periodic to allow precise shaping of the band energy structure. The materials forming the matched superlattice structure in the multiplication region repeat periodically.

[0083] By using a periodic structure of two or more layers of materials grown on a substrate, which can be selected from lattice-matched pairs of semiconductors to form a superlattice structure, linear-mode APDs 100, 500 including the matched superlattice structure are matched to suppress only impact ionization of first carriers in the superlattice multiplication region. These lattice-matched pairs have a band offset that is set so that i) they are at least predominantly within the band corresponding to the suppressed carriers that suppress impact ionization, and ii) at most, they are completely within the band corresponding to the suppressed carriers that suppress impact ionization. The lattice-matched pairs of semiconductors that form one period of the superlattice structure consist of at least one pair of alternating layers of a first semiconductor material and a second semiconductor material. A superlattice consists of multiple repeated periods, and the corresponding thicknesses of each layer in a period are the same. All of the first semiconductor layers have substantially the same thickness unless doped. All of the second semiconductor layers have substantially the same thickness unless doped. However, the thicknesses of the first semiconductor layers and the second semiconductor layers can differ.

[0084] Note that at low energies, unsuppressed carriers in the other band do not enter the superlattice, so their transport proceeds qualitatively as if it were in a bulk semiconductor. Unsuppressed carriers in the other band undergo impact ionization, providing the necessary gain for linear-mode APDs. Because impact ionization in the opposite band of the first carrier is suppressed, excess noise is eliminated.

[0085] Linear mode APDs 100, 500 containing matched superlattices can operate over a wide range of light detection wavelengths, including UV, visible, and short-wave IR (approximately 2500 nm) lightwaves.

[0086] In one embodiment, a linear-mode APD 100, 500 can detect single photons with high quantum efficiency near room temperature. However, in previous photodetectors that do not include a matching superlattice, noise generated by the photodetector can overwhelm the detected signal due to the photon. A linear-mode APD 100, 500 that includes a matching superlattice can detect single photons without the penalty of a dead time after the detection of a single photon, in which the linear-mode avalanche photodiode cannot detect another photon until after a set period of time has elapsed. Therefore, there is no pixel "dead time" after the detection event (i.e., "real-time" data acquisition). A linear-mode avalanche photodiode that includes a matching superlattice amplifies the single-photon signal to a measurable voltage while multiplying the photocurrent by a value higher than the noise-equivalent input current of the electronic device, which adds only negligible noise in the multiplication process.

[0087] APDs 100, 500 with matched superlattices behave as linear amplifiers. Therefore, in contrast to Geiger-mode APDs, linear-mode APDs 100, 500 with matched superlattices can resolve multiple photons arriving nearly simultaneously and do not experience a dead time after detecting a photon. The APDs 100, 500 are InP-based, room-temperature linear-mode APDs with gains of 10 electrons per photon, low excess noise, and high quantum efficiency, achieving the ultimate in detector sensitivity. Linear-mode APDs 100, 500 with matched superlattices can simultaneously achieve single-photon sensitivity with minimal excess noise and no dead time, even at temperatures above room temperature.

[0088] Next, typical APDs typically have a large internal gain M, potentially enabling tremendous photocurrent multiplication. However, even at modest values ​​of M≈50, their performance has been disappointing because the noise added by the multiplication process, known as excess noise, overwhelms the amplified signal. The excess noise is caused by the simultaneous existence of electron-initiated and hole-initiated impact ionization, which together create a positive feedback loop that magnifies small fluctuations in the impact ionization process. Two parameters used to characterize APDs are the electron-initiated impact ionization coefficient α and the hole-initiated impact ionization coefficient β. When either α or β is zero (unipolar gain), the excess noise factor F is ≤ 2 for all values ​​of M, but for large M associated with bipolar gain, F is proportional to M (see Figure 3).

[0089] Figure 3 shows a graph 300 of the theoretical excess noise figure F as a function of the average gain M for electron injection and the β / α ratio. A β / α curve of 0.03 is the current best value for APDs at room temperature. The β / α curve shows that for M>50, F increases with increasing M. This commensurate superlattice exhibits a β / α<<10 -4 can have:

[0090] Linear-mode APDs containing matched superlattices have single-pole gain and behave like photomultiplier tubes (PMTs), but with the tremendous system advantages of the area occupied by the device and the ruggedness associated with semiconductor devices. By providing high gain along with low noise in the detector, single-pole gain APDs can count photons in linear mode, and their output can be coupled to a commercially available electronic amplifier to provide a waveform that can detect single photon arrivals above noise and distinguish between multiple photons arriving simultaneously.

[0091] If room-temperature APD technology that can provide sufficiently low excess noise gain for linear-mode photon counting is unavailable, an alternative is Geiger-mode photon counting. In this mode, the APD is biased above its breakdown voltage, and the APD is in a metastable state where no current flows until it is triggered by the arrival of a single photon or by dark current. Once triggered, the APD current rapidly increases to a value that can be easily detected. While Geiger-mode APDs can detect single photons, they have two serious drawbacks: (1) after being triggered, Geiger-mode APDs have a dead time of approximately 100 nanoseconds or more during which they are insensitive to light, and (2) after pulsing, Geiger-mode APDs have re-emission of carriers captured by trap states in avalanche, which results in a higher dark count rate for many microseconds after triggering. For applications such as LIDAR where the arrival of signal photons is concentrated in time, the first of these drawbacks increases the false negative rate, and the second increases the false positive rate.

[0092] Next, Figures 1 and 5 show linear-mode APDs 100, 500 that include a matched superlattice structure that virtually eliminates excess noise by creating a significant asymmetry between electron and hole transport. This superlattice structure has desirable unipolar impact ionization using components that do not have special impact ionization properties, and therefore potentially has a set of materials that can be used for the APDs 100, 500 and span a material system that can be collectively used to detect a wide range of wavelengths. The gain region is comprised of two or more lattice-matched semiconductor alloys. These matched two or more lattice-matched semiconductor alloys have the property that their entire band offset is contained solely in the valence band or solely in the conduction band. For band offsets contained solely in the valence band or solely in the conduction band, see Figures 1 and 5.

[0093] Figure 2a shows an embodiment of a graph of a hole in the valence band of a bulk semiconductor under bias, storing kinetic energy with distance. Figure 2b shows an embodiment of a graph of a hole in the valence band hopping through a well in a biased superlattice, and the energy levels of states in a Wannier-Stark ladder. Figures 1, 2a, and 2b show gain due to electron multiplication where the superlattice is only in the valence band. Similarly, Figures 5, 6a, 6b, and 7 show gain due to hole multiplication where the superlattice is only in the conduction band instead of the valence band.

[0094] Impact ionization in APDs Figure 4 shows a band graph 400 of a semiconductor, with energy plotted vertically and distance plotted horizontally, and the amount of kinetic energy required to create a new electron-hole pair through impact ionization. Free electrons reside in the conduction band, and free holes reside in the valence band. The conduction and valence bands are separated by a band gap of width Eg, within which no electron or hole states can exist. The presence of an applied electric field causes the bands to tilt proportionally to the magnitude of the field, as shown in Figure 4. The electron in the upper left corner moves to the right when accelerated by the applied electric field. The vertical distance between the electron's instantaneous position and the conduction band is the electron's instantaneous kinetic energy. When the electron's kinetic energy is greater than the band gap energy, the electron has enough energy to create an electron-hole pair through impact ionization. When an electron is impact ionized, most of the kinetic energy is used to create the electron-hole pair, resulting in two electrons and one hole with negligible kinetic energy. Electron-initiated impact ionization is characterized by a factor α, which is the average number of electron-initiated impact ionizations per centimeter (see Figure 1).

[0095] A similar sequence of events can occur in the valence band, where holes accelerate to create electron-hole pairs. In this case, hole-initiated impact ionization is followed by two holes and one electron (see Figure 5). This process is characterized by the hole-initiated impact ionization coefficient, β.

[0096] Figure 1 shows a graph of these two bands placed in the superlattice multiplication region. Both electrons and holes are accelerated by the electric field, and therefore both electrons and holes initiate the creation of new electron-hole pairs by impact ionization. The general case is discussed in the next section, but the case where β=0 is easy to understand. A single electron undergoes impact ionization, resulting in two electrons and one hole. These two electrons each accelerate, leaving four electrons after impact ionization. These four electrons grow to 8, 16, 32, ... so the number of electrons decreases with distance z, e αz The number of holes increases exponentially as β = 0. Whenever an electron is created, a hole is also created, so the number of holes also increases exponentially. When β = 0, these holes do not increase the number of electrons because they are not impact ionized. Therefore, when β = 0, the semiconductor has an exponential gain e αz where the only noise comes from variations in the distance the electron travels before impact ionization. Similarly, α=0 provides an exponential gain e βz Give.

[0097] Excess noise in APDs In some previous techniques, the values ​​of α and β are very similar for most semiconductors, so the simple case of exponential gain does not apply. Consider the case where a photogenerated electron is injected from an absorbing layer at z = 0 into a biased semiconductor layer, initiating the first impact ionization at z = z0.

[0098] This impact ionization generates a second electron and hole. The hole, having the opposite charge to the electron, drifts in the opposite direction. If it does not reach the edge of the semiconductor layer first, it will be impact ionized, creating a new electron-hole pair at z=z1 between z=0 and z=z0. Due to randomness in when and where the hole is scattered by crystal defects or phonons, there is some random variation in how far the hole drifts before being impact ionized. This variation can make a large difference near the edge, depending on whether the hole is impact ionized just before reaching the edge of the layer. The newly created electron at z=z1 drifts toward z=z0, possibly being impact ionized along the way, creating more electron-hole pairs. Even if we ignore the electron-hole pairs subsequently created by the two electrons emerging from the initial impact ionization, it should be clear that a single initial photoelectron impact ionized at z=z0 generates subsequent electrons that pass through z=z0. However, it is impossible to distinguish whether the second electron passing through z=z0 is a by-product of the initial photoelectron generated by the impact ionization sequence described above, or whether it is a second photoelectron injected into the layer. The signal in this first situation results from the injection of only the initial electron, while the signal in the second situation results from the injection of two electrons at slightly different times. Because impact ionization obscures the difference between these possible inputs, this process adds noise to the signal. This noise is known as excess noise. Excess noise in APDs has been well studied.

[0099] Figure 3 shows the excess noise figure F as a function of β / α for electron injection into the gain region and the average APD gain M. The curve for β / α=0 gives the smallest F for all M, which is the β=0 case discussed above. The result that F has a value of 2 rather than the expected value of 1 for this ideal case is due to the assumption made in this calculation that impact ionization is a Poisson process. As β / α increases, the excess noise figure remains F=2 for small M, but increases as M increases. A closer inspection of these curves gives the rule that to achieve a gain M at F=3, α / β≈M must be satisfied. A gain of 10, sufficient for almost excess noise-free detection of single photons under certain conditions, is needed. 4 To achieve this, β / α=10 -4 To date, no single-component semiconductors are known with such a large difference between α and β, so engineered materials, including alloys, are required to achieve this difference and to achieve gain initiated only for the first carrier type (electrons or holes).

[0100] In Figure 3, as F gets larger, M also gets larger unless α / β or β / α is very small. The current design is how to make α / β or β / α very small, and therefore how to make F small for large M.

[0101] Another consequence of having α ≠ 0 and β ≠ 0 simultaneously is avalanche breakdown. Breakdown occurs when the number of electrons or holes becomes arbitrarily large for a finite voltage and a finite size of the multiplication region. As discussed previously for α = 0 or β = 0, the number of electrons and holes grows exponentially with z. For finite voltages (which imply that α and β remain finite) and finite sizes (which imply that z remains finite), this number remains finite, albeit growing rapidly.

[0102] However, when α ≠ 0 and β ≠ 0, the situation changes. Suppose a single electron is injected into the multiplication region at z = 0 and begins impact ionization at z = z1. The hole created by this impact ionization drifts in the -z direction and may be impact ionized before reaching the edge of the multiplication region at z = 0. When this happens, the hole generates an electron at z between 0 and z1. This is exactly the initial situation that is repeated later. At the breakdown threshold, such a loop becomes self-sustaining, and as a result, the injected single electron generates an infinite number of subsequent electrons. In this simplified situation, which considers only the first impact ionization initiated by each carrier, the accumulation of carriers occurs exponentially over time. A more realistic situation, which considers all impact ionization, would give a much faster increase over time.

[0103] As discussed, linear mode avalanche photodiodes are 1000 times (10 3 ) amplification, while producing excess noise less than three times the thermal noise present at non-cryogenic temperatures or higher (β / α≦0.1). (See the shaded operating area in the lower right portion of FIG. 3.) In one embodiment, the smallest usable gain may be 200 times amplification or greater, while producing excess noise less than three times the thermal noise present.

[0104] In one embodiment, a linear mode avalanche photodiode comprising a matched superlattice structure in such a manner that amplification occurs in i) the conduction band only, or ii) the valence band only, can be used to achieve a 10,000x (10 4) is configured to produce a gain greater than or equal to the amplification, while generating excess noise less than three times the thermal noise present at temperatures above room temperature due to the amplification gain (β / α≦0.1). A linear-mode avalanche photodiode including a matched superlattice structure with a gain of 10,000 times amplification can detect single photons without the penalty of a dead time after the detection of a single photon, in which the linear-mode avalanche photodiode cannot detect another photon until after a set period of time has elapsed. (Thus, there is no pixel "dead time" after the detection event (= "real-time" data acquisition). With dead time, the waveform captured by the APD would have a time interval after the detection of each photon during which subsequent photons cannot be detected.

[0105] Figures 2a and 2b show a comparison of hole transport in a biased bulk semiconductor and a biased superlattice. Referring to graph 200 in Figure 2a, in a bulk semiconductor, an electric field accelerates holes in an extended state toward the left. As the holes accelerate, their kinetic energy increases until they have enough energy to undergo impact ionization. Referring to graph 250 in Figure 2b, the allowed states in a biased superlattice are Wannier-Stark states that are mostly within one well but extend several wells to the left. The ground state of well A is also the initially excited state of well B. Holes in this extended state can relax to the ground state of well B by emitting phonons or photons. In this way, holes can be transported to the left while remaining in their local ground state and not possessing enough energy to undergo impact ionization.

[0106] As shown in Figures 1, 2b, 5, 6b, and 7, a superlattice, such as a superlattice made of two matched alloys, has wells and barriers in one band but not in the other, so carrier transport within these two bands is very different. Transport within offset bands is due to hopping from states localized in one well to states localized in an adjacent well (also known as phonon-assisted tunneling). The energy imparted to carriers by an applied electric field dissipates via phonons or mid- or long-wave infrared photons, so the carriers do not accumulate enough kinetic energy for impact ionization. Therefore, the superlattice has β=0.

[0107] Carriers in unoffset bands do not have the localized states necessary for hopping. Therefore, as in normal semiconductors under bias, carriers are in extended states where they accumulate kinetic energy as they drift through an electric field, and they are impact ionized when their kinetic energy becomes greater than the band gap energy (see Figures 1, 4, and 5). Matched linear-mode APD structures have a large asymmetry between the ability of electrons to impact ionize and the ability of holes to impact ionize, resulting in high gain (>10 4 ) and low excess noise (F<2) at room temperature.

[0108] From the above discussion, it can be seen that excess noise-free exponential gain can occur when α = 0 or β = 0. To achieve this, note that only carriers with kinetic energy greater than the bandgap energy can initiate impact ionization. A commensurate superlattice structure prevents either electrons or holes from accumulating sufficient kinetic energy for impact ionization under a sufficiently large electric field that would allow carriers of the opposite polarity to accumulate the kinetic energy necessary for impact ionization. Thus, impact ionization gain can be initiated only by carriers with kinetic energy greater than the bandgap energy through the use of a commensurate superlattice, where the superlattice structure prevents either electrons or holes from accumulating sufficient kinetic energy under a sufficiently large electric field that would allow carriers of the opposite polarity to accumulate the kinetic energy necessary for impact ionization.

[0109] The matched linear-mode APD 100, 500 eliminates excess noise by suppressing impact ionization for one carrier while maintaining impact ionization for the other carrier. Attempts to create linear-mode APDs with large α / β or small α / β are based on enhancing impact ionization for one carrier while maintaining impact ionization for the other carrier. For a given small change in the impact ionization rate of the first carrier, using it for suppression results in a larger change in α / β. For example, increasing the impact ionization rate of a carrier by 50% increases α / β by a factor of 3 / 2, whereas decreasing the impact ionization rate by 50% doubles β / α.

[0110] Similar to the discussion of holes in Figures 2a and 2b, Figure 6a shows an embodiment of a graph 600 of electrons in a bulk semiconductor under an applied electric field, accelerating and storing kinetic energy. Figure 6b shows an embodiment of a graph 650 of a commensurate superlattice that suppresses impact ionization by controlling localized Wannier-Stark states. A superlattice can be a periodic structure of layers of two or more materials. Typically, the thicknesses of the material layers can be measured in nanometers. With appropriate band offsets and layer thicknesses, a superlattice under an electrical bias forms Wannier-Stark states that are largely localized in a single well. The energy levels of these states are indicated by dashed lines and their wave functions, which are represented by solid-shaded curves. Above the barrier energy, there exists a continuum of extended states into which electrons can accelerate in a field. It is essential to prevent electrons from occupying this continuum. If the superlattice is designed correctly, the Wannier-Stark states and the continuum are separated by a large energy kT, so it is unlikely that electrons in the Wannier-Stark states will be thermally excited into the continuum, where they can be accelerated by an electric field.

[0111] Note that for an electron in a Wannier-Stark state in well 0, the state in well 1 is lower energy and therefore more favorable. This electron can transition from well 0 to well 1 by emitting a phonon or photon and conserving energy. This process is known as hopping. The probability of hopping is proportional to the overlap of the wave functions in well 0 and well 1. Thinner wells and barriers, and smaller differences between the conduction band levels of the barrier and the well, lead to larger overlaps, which can cause the localized state to revert to the extended state it seeks to eliminate. Electron hopping from a well 0 state to a well 1 state accomplishes two useful functions. Electron hopping from a well 0 state to a well 1 state moves the electron one superlattice period "a" in the z direction, removing energy qEa from the electron, where q is the electron charge and E is the applied electric field. This energy is precisely the kinetic energy that the electric field imparts to the charge q over the distance "a" in the bulk semiconductor. An electron in the superlattice has traveled a distance a in the field without changing its kinetic energy. After entering well 1, the electron can hop to well 2, then to well 3, and so on, without gaining the kinetic energy necessary for impact ionization.

[0112] The thicknesses of the well and barrier layers are determined by the following design goals: Wannier-Stark states are formed at the applied electric field required to provide the desired carrier multiplication, the energy levels of the Wannier-Stark states are low enough within the well that carriers are not thermally excited and do not enter the continuum, and states in adjacent wells have a large overlap so that hopping can be rapid.

[0113] The above discussion demonstrates that electron-initiated impact ionization can be suppressed by a conduction band superlattice. Similarly, hole-initiated impact ionization can also be suppressed by a valence band superlattice (see Figures 2b and 1). However, it should be noted that suppressing impact ionization in one band by a superlattice is only half the solution to eliminating excess noise in APDs. Blindly selecting two semiconductors—one for the well and one for the barrier—will likely result in superlattices in both the valence and conduction bands. While impact ionization can be suppressed in one band by design, impact ionization in the other band may be suppressed, or at least reduced. Therefore, it would be advantageous to select a pair of semiconductors whose band offset completely encompasses the band that suppresses impact ionization.

[0114] The other band will have no offset and will behave more or less like a bulk semiconductor. However, there are two major differences. First, the impact ionization coefficients of the two materials will likely be different. The effective impact ionization coefficient will be somewhere between that of the individual bulk semiconductors. Second, the effective mass of the carriers will likely be different. An effective mass superlattice will have no wells or barriers for carriers near the Γ point in reciprocal space.

[0115] The unsuppressed band in a superlattice whose band offset of the semiconductor pair is completely contained within the suppressed band still has a different effective mass relative to these two layers. This results in an effective-mass superlattice where there is no well for carriers with little kinetic energy (near the gamma point in reciprocal space), but where a well gradually forms as carriers gain kinetic energy. Depending on the specific materials forming the superlattice, the conduction band well may be in the same layer as the valence band well (type I alignment) or in the opposite layer (type II alignment). In the worst case scenario for APDs, the emergence of the effective-mass superlattice may begin to suppress carriers in the unsuppressed band. This effect can be ameliorated by selecting the band offset such that the effective-mass superlattice for carriers with some intermediate kinetic energy disappears.

[0116] Next, to reiterate, Figure 1 shows a graph of these two bands placed in the superlattice multiplication region. The superlattice is configured so that the Ev valence band is completely contained (thus multiplying electrons but not holes). Therefore, the Ec conduction band is essentially a downward-sloping straight line. Figure 1 shows the epitaxial structure and energy band diagram of the proposed APD100 under reverse bias. A photon (red wavy arrow) is absorbed in the p-InGaAs absorber layer, which generates an electron (black circle)-hole (white circle) pair. The hole diffuses into the p-InP cap layer, and the electron diffuses back to the multiplier. The electron is accelerated by the electric field and eventually accumulates enough kinetic energy to impact-ionize a secondary electron-hole pair. Each electron then accelerates and eventually generates another electron-hole pair. The holes generated by impact ionization lose kinetic energy as they hop through the valence band superlattice, and therefore cannot be impact-ionized. Thus, a very large asymmetry between α_ and β_ is produced.

[0117] Because one-sided band alignment can appear unusual, several pairs exist that are lattice-matched to either III-V (III and V refer to the NIB and VB columns of the periodic table, respectively) semiconductor substrates, InP or GaSb. Promising pairs are InP and In 0.53 Al 0.30 Ga 0.17 The alloys are 53% In, 30% Al, and 17% Ga for the group III component, and entirely As for the group V component. Figure 1 shows an APD 100 built around an InP-InAlGaAs multiplication layer for electron multiplication by impact ionization without co-multiplication of holes. Electrons photogenerated with high quantum efficiency in the InGaAs absorber layer are injected into the multiplication layer. These electrons are impact ionized as they traverse the multiplication region. Secondary electrons can also be impact ionized, but secondary holes cannot. This is because secondary holes lose kinetic energy as they hop from one quantum well to the next. Electron multiplication by impact ionization without co-multiplication of holes results in an exponential growth of each initial photoelectron with a small excess noise.

[0118] A challenging systematic search was conducted to find matched semiconductor pairs with perfectly matched band offsets in either the valence or conduction band. A constraint on state-of-the-art performance is the requirement that all semiconductor layers in a device have the same crystal structure type and dimensions as the substrate. This constraint is known as lattice matching. Without lattice matching, crystal defects originating at the interfaces between layers degrade performance. Therefore, the search for suitable pairs had to be performed separately for each substrate of interest. For linear-mode APDs in the near-infrared and shortwave infrared (wavelengths approximately 0.8–3 μm), the most interesting substrates are InP and GaSb.

[0119] For these two substrates, InP and GaSb, the inventors estimated the conduction and valence band energies for all lattice-matched alloys that could be grown by current epitaxial growth techniques. For pairs with either equal conduction or equal valence band energies, the inventors tended to exclude pairs containing alloys of five or more elements, pairs containing alloys with indirect bandgaps, and pairs containing alloys known to phase separate. The exclusion of pairs containing alloys of five or more elements is due to the perceived difficulty in controlling the stoichiometry.

[0120] This search yielded cases where an alloy A1(x1), which depends on the parameter x1 characterizing the alloy's composition over a range of x1, has the same conduction or valence band energy as a second alloy A2(x2) over a range of x2. The inventors of the present invention found that this condition of equal conduction or valence bands is satisfied over a range of x1 that is synchronized with the range of x2. Within such a range, the endpoints of the range are of most interest, since there the band offset of the unequal bands is maximized.

[0121] Specific alloy pairs are listed below. In cases where the conduction or valence band energies can be the same over a range, the pair with the largest band offset relative to the unequal band for that range is listed.

[0122] The present inventors have discovered semiconductor pairs that are useful for linear-mode APD design, both in terms of having suitable properties and in terms of the practicality of performing epitaxial growth of the pairs.

[0123] InP substrate The superlattice structure has a semiconductor lattice-matched pair in the superlattice multiplication region, and the lattice-matched pair is also matched to the InP substrate, and the lattice-matched pair is

[0124] A first semiconductor alloy of InP and In, where the superlattice is set in the valence band and holes are the primary carriers. 0.53 Al 0.30 Ga 0.17 two semiconductors in a superlattice structure, including a second semiconductor alloy of As;

[0125] A first semiconductor alloy of InP and In, where the superlattice is set in the valence band and holes are the primary carriers. 0.19 Ga 0.81 As 0.69 Sb 0.31 two semiconductors in a superlattice structure, including a second semiconductor alloy of

[0126] A first semiconductor alloy of InP and GaAs, where the superlattice is set in the valence band and holes are the primary carriers. 0.12 Sb 0.61 P 0.27 two semiconductors in a superlattice structure, including a second semiconductor alloy of

[0127] The superlattice is set in the valence band, and holes are the primary carriers. 0.81 Ga 0.19 As 0.42 P 0.58 The first semiconductor alloy and In 0.37 Ga 0.63 As 0.85 Sb 0.15 two semiconductors in a superlattice structure, including a second semiconductor alloy of

[0128] The superlattice is set in the valence band, and holes are the primary carriers. 0.94 Ga 0.07 As 0.12 P 0.88 First semiconductor alloy and GaSb 0.65 P 0.35 two semiconductors in a superlattice structure, including a second semiconductor alloy of

[0129] The superlattice is set in the conduction band, and electrons are the primary carriers. 0.52 Al 0.48The first semiconductor alloy of As and In 0.79 Ga 0.21 As 0.46 P 0.54 two semiconductors in a superlattice structure, including a second semiconductor alloy of is selected from the group consisting of:

[0130] GaSb substrate The superlattice structure has a semiconductor lattice-matched pair in the superlattice multiplication region, and the lattice-matched pair is also matched to the GaSb substrate, and the lattice-matched pair is

[0131] A first semiconductor alloy of GaSb and In, where the superlattice is set in the valence band and holes are the primary carriers. 0.57 Al 0.43 As 0.55 Sb 0.45 two semiconductors in a superlattice structure, including a second semiconductor alloy of

[0132] The superlattice is set in the conduction band, and electrons are the primary carriers. 0.14 Ga 0.86 As 0.01 Sb 0.99 The first semiconductor alloy and In 0.09 Ga 0.91 As 0.08 Sb 0.92 two semiconductors in a superlattice structure, including a second semiconductor alloy of

[0133] The superlattice is set in the conduction band, and electrons are the primary carriers. 0.44 Ga 0.56 As 0.04 Sb 0.96 The first semiconductor alloy and In 0.28 Ga 0.72 As 0.26 Sb 0.74 two semiconductors in a superlattice structure, including a second semiconductor alloy of

[0134] The superlattice is set in the conduction band, and electrons are the primary carriers. 0.40 Al 0.60 As0.42 Sb 0.58 The first semiconductor alloy and In 0.79 Ga 0.21 As 0.72 Sb 0.28 two semiconductors in a superlattice structure, including a second semiconductor alloy of is selected from the group consisting of:

[0135] The properties of these alloys are based on measurements made by many laboratories with varying uncertainties, so these compositions are approximate.

[0136] Recognizing that these alloys can be difficult to grow due to unintentional intermixing of the two alloys at the interface, the present inventors consider ways to relax the constraints on the structure without adversely affecting the physics of the device.

[0137] The principle used by the inventors of the present invention is that it is not necessary to completely eliminate the offset of one of the bands. The offset need only be small enough for the designed layer thickness that there are no trapped states lower than kB away from the band continuum, where kB is Boltzmann's constant and T is the intended operating temperature. In the case of the valence band, it is sufficient to consider only the heavy hole states, since eliminating the heavy hole trapped states automatically eliminates the light hole trapped states as well.

[0138] 1. Including a spacer layer between the well and the barrier For clarity, the inventors of the present invention have, by way of example, 0.52 Al 0.48 As barrier and In 0.79 Ga 0.21 As 0.46 P 0.54An exemplary alloy layer consisting of a well was considered. The rough surface morphology and low photoluminescence may be the result of the unintended formation of AlP at the interface of the InGaAsP-InAlAs superlattice. A lattice-matched spacer layer of InGaAs(P) containing less P than 54% of the well layer inserted between the well and barrier layers will reduce AlP formation by juxtaposing the Al-containing layer with a P-containing layer that has a lower P content than the well. If the valence band offset of the spacer is small enough and the spacer is thin enough, the spacer has little effect on the valence band. Any trapped holes. In lattice-matched to InP 0.53 Ga 0.47 The As spacer is clearly the best spacer for preventing AlP formation, but is the worst for forming a valence band well.

[0139] Because growth defines a direction for the superlattice, placing a spacer immediately after the growth of an InAlAs barrier layer is not equivalent to placing a spacer immediately after the growth of an InGaAsP well layer. Starting with a baseline superlattice of a 3.5 nm InGaAsP well and a 1.5 nm InAlAs barrier, the inventors modeled three modified superlattice periods (with the last listed layer grown first): (a) InGaAs(P)-InGaAsP-InAlAs, (b) InGaAsP-InGaAs(P)-InAlAs, and (c) InGaAs(P)-InGaAsP-InGaAs(P)-InAlAs. The thickness of the InAlAs barrier layer was kept constant at the initial 1.5 nm, and the thicknesses of the remaining layers summed to the initial well thickness of 3.5 nm. For cases (a) and (b), the spacer thickness was 1.5 nm, while for case (c), the respective spacer thickness was 0.5 nm. InGaAs(P) spacer layers were modeled for P contents of 0%, 13%, 27%, or 40%, with the remaining components of the layer adjusted to maintain lattice match with InP. This modeling indicated that to prevent quantum confinement of holes, the spacer layer needed to contain at least 13% P for case (a), at least 27% for case (b), and at least 40% P for case (c). It is questionable whether a spacer layer containing more than 40% P would be effective in preventing AlP formation and whether a 0.5 nm layer could be grown, making cases (a) and (b) promising candidates for reducing intermixing at the interface. For this example, it would work best if intermixing occurred primarily at the InAlAs-on-InGaAsP interface or the InGaAsP-on-InAlAs interface, since only half of the interface has the spacer.

[0140] 2. All-arsenide superlattice structure The more atomic fractions that must change from one layer to the next, the more difficult the superlattice to grow. Superlattices in which both the well and the barrier are pure phosphide, pure arsenide, or pure antimonide will be easier to grow than any of the 10 superlattices listed above.

[0141] As an example, the inventors explored the possibility of growing an all-arsenide lattice-matched superlattice in which the wells have relaxed properties, with localized quantum states in one band but not the other. InGaAlAs wells and InAlAs barriers did not seem promising because both electrons and heavy holes were confined for practical growth thicknesses. However, a superlattice with coupled wells per period seemed to offer a means to separate the electron and heavy hole confinement. The electric field would push electrons, for example, into the right-coupled well, while heavy holes would be pushed into the left-coupled well. With coupled wells, the right-well parameters would have a greater effect on electrons, and the left-well parameters would have a greater effect on heavy holes. A superlattice with periods consisting of 1.5 nm Q-1.5 nm InAlAs-3.0 nm Q-1.5 nm InAlAs would have confined electron states, good well-well wavefunction overlap, and no confined heavy hole states. where Q=In 0.53 Ga 0.23 Al 0.24 As and InAlAs=In 0.52 Al0. 48 As.

[0142] The superlattice in which electrons or holes are suppressed consists of alternating layers of alloy "A" and alloy "B" selected from the pairs listed above. All the "A" layers have the same thickness, and all the "B" layers also have the same thickness, which may or may not be the same as the thickness of the "A" layers.

[0143] As an alternative to Figure 1, Figures 5, 6b, and 7 show a linear-mode APD 500 with a matched superlattice design in which electron multiplication is suppressed but hole multiplication is not. Electrons cannot accumulate enough kinetic energy to initiate multiplication. Only holes can initiate gain, potentially resulting in exponential multiplication. Excess noise in APDs results from simultaneous electron and hole multiplication. However, suppressing the multiplication of either electrons or holes minimizes noise while providing gain. This matched superlattice achieves carrier multiplication at non-cryogenic temperatures, such as room temperature (30-50°C), and even above this range. This superlattice is designed to have at least one quantum state in each well when the superlattice is biased with an electric field that sustains impact ionization only in thick layers of well or barrier semiconductors. In the case of a valence-band superlattice, there should be at least one quantum state for light holes and one quantum state for heavy holes.

[0144] Second, to suppress only the impact ionization of the first carriers, the design can select pairs of semiconductors whose band offset lies mostly within the band corresponding to the suppressed carriers. The band offset of the superlattice pair does not need to lie entirely within the band corresponding to the suppressed carriers. There can be a sufficient offset in the opposing bands, as long as the offset is not sufficient to support quantum states in the thickness of the superlattice wells and barriers. This allows for a tolerance in the superlattice composition.

[0145] Looking at Figure 5 graphically, when the superlattice is configured to be fully contained in the EC conduction band (thus multiplying holes but not electrons), the Ev valence band is essentially a downward-sloping straight line. However, the material's coefficients can be matched to place the majority of the material within one band, e.g., the conduction band, but potentially not entirely within that band. Visually, this means that the other band, i.e., the valence band, will not be a solid straight line, but rather a roughly straight line with occasional small, imperceptible pulses that are not deep enough to form a well. Thus, when the band offset of a superlattice pair is configured to be largely contained within the band corresponding to the suppressed carriers, the band of the other carrier can have a sufficient offset as long as it is not sufficient to support a quantum state within i) the well layer, ii) the barrier layer, or iii) the thickness of either the well or barrier layer of the superlattice.

[0146] Additional advantages of room temperature linear mode APDs containing commensurate superlattice structures. The proposed APD can operate as a low-noise photocurrent amplifier with sufficient gain to enable single-photon sensitive capture of the photocurrent waveform when its output is coupled to a commercially available electronic amplifier. This device achieves ultimate detector sensitivity at room temperature by detecting single photons with high quantum efficiency over a large band of wavelengths.

[0147] The proposed device does not require Geiger-mode operation to detect single photons. To demonstrate this, an array of SRI linear-mode APDs can obtain a complete 3D point cloud LIDAR image from a single shot of the laser (see Figures 8 and 9). In contrast, an array of Geiger-mode APDs requires a laser shot for each of N range bins, which increases acquisition time and laser energy by a factor of N.

[0148] Various embodiments for suppressing electron-initiated or hole-initiated impact ionization The choice that must be made is whether one wants to suppress electron-initiated impact ionization or hole-initiated impact ionization. For example, a conduction-band-only superlattice and / or a valence-band-only superlattice. The main factor influencing this choice is the possibility that impact ionization will generate suppressed carriers in the continuum state rather than the Wannier-Stark state. To estimate this possibility, the inventors of the present invention considered the kinetics of impact ionization in bulk InP. Under the effective mass approximation for electron-initiated impact ionization, the energy conservation equation and the two-dimensional momentum conservation equation for the following reaction were solved: e→e+e+H (I) e→e+e+h (II)

[0149] In the above equation, e is the electron, H is the heavy hole, and h is the light hole. The initial electron has kinetic energy E. We calculated the threshold E at which the reaction is allowed. Reaction I, which produces the heavy hole, has a threshold energy of 1.50 eV, and Reaction II, which produces the light hole, has a threshold energy of 1.99 eV. The minimum threshold energy is the bandgap energy of InP (1.34 eV). The threshold energy is increased by the amount necessary to impart final momentum to each final particle, thereby conserving energy and momentum. As the initial electron energy increases above the 1.50 eV threshold, the density of final states increases, increasing the rate at which electrons can be impact ionized. The electron energy is unlikely to reach the threshold for Reaction II. According to the kinetics, in Reaction I, when impact ionization occurs near the threshold energy, the heavy hole has an energy 0.10 eV below the valence band edge. This deposits heavy holes into the InP-InGaAlAs couplet well (0.23 eV deep) which is used for electron-initiated impact ionization.

[0150] For hole-initiated impact ionization, consider the following reaction: h → h + e + H (III) h→h+e+h (IV) H → H + e + H (V) H→H+e+h (VI)

[0151] The threshold is 1.52 eV for reaction III and 2.06 eV for reaction IV. The inventors of the present invention expect that light holes will generate heavy holes. When impact ionization occurs near the threshold, the electrons will have energy 0.03 eV above the conduction band edge. For heavy hole-initiated impact ionization, the threshold is 2.63 eV for reaction V and 6.07 eV for reaction VI. When impact ionization occurs near the threshold, the electrons will have energy 0.04 eV above the conduction band edge. In both cases, the electrons will approach the bottom of the InAlAs-InGaAsP couple well (0.40 eV deep) used for hole-initiated impact ionization.

[0152] Although hole-initiated impact ionization structures are by far superior for preventing unwanted carrier injection into the continuum states of the superlattice, electron-initiated structures may also be suitable.

[0153] This analysis points out another factor that needs to be considered when comparing these two structures. When holes are initiating impact ionization, the threshold for generating heavy holes is lower. The valence band of the APD will be filled with heavy holes. Comparing the 2.63 eV threshold for heavy hole-initiated impact ionization with the 1.50 eV threshold for electron-initiated impact ionization, the β for the hole-initiated structure is about 1.8 times lower than the α for the electron-initiated structure. Therefore, to obtain the same gain with both structures operating at the same electric field, the hole-initiated structure would require a 1.8 times thicker multiplication region and a 1.8 times higher voltage.

[0154] This design evaluates that a conduction band superlattice that suppresses electron avalanche gain is superior to a valence band superlattice that suppresses hole avalanche gain.

[0155] Another advantage of the hole-initiation structure is that holes are injected from the light-absorbing layer into the superlattice multiplication region. The holes are injected from below. This allows APDs to be fabricated with zinc diffusion for selective p-doping, which is commonly used to obtain low dark current and to eliminate surface breakdown in pin photodiodes and APDs.

[0156] Neither structure is clearly superior, and the choice will likely depend on the details of the intended application. [Industrial Applicability]

[0157] Purpose Various systems can use an array of multiple linear-mode avalanche photodiodes including a matched superlattice structure. Each linear-mode avalanche photodiode is configured to sense light and output a current by producing a gain of 1000 times or more while producing an excess noise figure less than three times the thermal noise present at non-cryogenic temperatures or higher due to the gain from the amplification. The linear-mode avalanche photodiode detects one or more photons of light by using a matched superlattice structure that suppresses impact ionization for a first carrier of the linear-mode avalanche photodiode while achieving at least one of: 1) increasing impact ionization, 2) substantially maintaining impact ionization, and 3) suppressing impact ionization to a lesser extent for a second carrier. The first carriers whose impact ionization is suppressed are either i) electrons or ii) holes, and the second carriers are either electrons or holes. A power supply is used to provide power to the multiple linear-mode avalanche photodiodes.

[0158] These systems may include i) LIDAR systems, ii) night vision goggle or headset systems, iii) optical communication systems, iv) spectroscopy systems, v) quantum key distribution systems, vi) systems using high breakdown voltage transistors, vii) low noise microwave generation systems, and viii) biomedical systems.

[0159] A linear-mode avalanche photodiode including a matched superlattice structure is configured to produce a gain of 10,000 times or more such that amplification occurs in i) the conduction band only, or ii) the valence band only, while producing excess noise due to the gain amplification that is less than 10 percent of the thermal noise present at temperatures above room temperature. A linear-mode avalanche photodiode including a matched superlattice structure having a gain of 10,000 times amplification is configured to detect a single photon without the penalty of a dead time after the detection of the single photon, in which the linear-mode avalanche photodiode cannot detect another photon until after a set period of time has elapsed.

[0160] Exemplary functions enabled by the linear-mode APDs described herein include the following:

[0161] FIG. 8 shows an embodiment of a diagram 800 of a linear-mode APD array including a matched superlattice structure with in-situ memory and parallel column readout. Each pixel of the array looks in a different x-y direction and captures the z-direction. The array is designed for in-situ storage of each time bin, which is sequentially read out between light pulses. Light pulses, such as laser pulses, can be captured in 100 time bins, each of which is a 50 μm × 50 μm pixel. Each time bin has a readout circuit including a linear-mode APD including a matched superlattice structure. FIG. 9 shows an embodiment of a diagram 900 of an exemplary readout circuit diagram for a linear-mode APD including a matched superlattice structure. Readout can be performed at 100 frames per second. A readout circuit with a 50 μm × 50 μm pixel structure can store, for example, 100 time bins.

[0162] A system using a linear mode APD array including a matched superlattice structure can include:

[0163] Single-shot light detection and ranging (LIDAR) - LIDAR illuminates a scene with a short burst of light from a laser and detects the reflection of this pulse. The time between firing the laser light and detecting the reflection provides the distance to the reflector. SRI linear-mode APDs have no dead time, allowing continuous acquisition of the reflected light waveform using a single APD. With the appropriate optics, each APD in a linear-mode APD array can detect light arriving from a different direction. Each APD in a linear-mode APD array can acquire the complete waveform of light reflected from an object located at a unique orientation determined by the optics. Thus, a linear-mode APD array can collect all reflections of a single laser shot across the entire field of view determined by the optics. The collected data represents a three-dimensional image of the scene, with the arrival time providing the z-direction and each APD providing the x- and y-directions.

[0164] The waveform representing the data along the z-direction can be broken down into time bins, and the APD's output current can be integrated over each time bin, stored as a charge on a separate capacitor in the readout circuitry. The inventors of the present invention have determined that a 50 μm×50 μm pixel can have 100 storage capacitors, in addition to the electronic circuitry required to integrate the current and read out the charge on the capacitors.

[0165] Due to the dead time of Geiger-mode APDs, two methods are used to capture all reflections. The first method uses an array of Geiger-mode APDs connected in parallel to form a superpixel to improve the dead time. The large size of the superpixel limits the ability to arrange the superpixels in a linear array. Therefore, imaging perpendicular to the linear array is performed by illuminating the entire scene and rotating a narrow-field-of-view APD array, or by illuminating a stripe of the scene collinear with the wide-field-of-view APD array and moving the stripe. This method requires multiple laser shots to capture a complete three-dimensional image.

[0166] The second method uses a two-dimensional array of independent Geiger-mode APDs. As with a linear-mode APD array, each APD collects reflected light from a different direction. At a predetermined delay after the laser shot, all Geiger-mode APDs are gated open by biasing them above breakdown for a predetermined time. The output of this array provides an x-y image of photon arrivals during the time interval the APDs are gated open. The output from a particular APD is the same whether one photon or many photons arrive during that interval. By varying the gate delay, multiple laser shots can be used to obtain a complete three-dimensional image.

[0167] The advantages of acquiring 3D imagery in a single shot are (1) faster acquisition time, and (2) the absence of distortion and blurring due to motion, both within the scene and between the LIDAR system and the scene. Faster acquisition enables 3D imaging. Insensitivity to motion gives sharper images and alleviates the need for a very stable platform for the LIDAR system.

[0168] Imaging through an obscurant - The principle is the same as that of single-shot LIDAR. The difference is that reflections from an obscurant, such as water droplets in fog or smoke particles, occur randomly. This is because these particles are randomly distributed and change rapidly over time due to Brownian motion, sedimentation, or convection. The object of interest is much larger than the obscurant and therefore moves more slowly and cannot change velocity as rapidly. Therefore, by retaining only reflections from the same point or a series of points in a straight line over several successive 3D images, many of the reflections from the obscurant can be filtered out.

[0169] Low-light-level imaging - In low-light-level imaging, it is often desirable to increase the photocurrent from the detector sufficiently so that it is above the noise floor of the electronic amplifier. To achieve this, a linear-mode APD can be operated with a modest gain, not enough to detect single photons. An alternative way to achieve this is to increase the collection aperture and lengthen the integration time. A larger aperture results in larger optical components and a bulkier, more expensive system. A longer integration time requires the imager to be mounted on a stable mechanical support, such as a tripod. Integration cannot be used on dynamic scenes that are changing faster than the integration time.

[0170] An important case for low-light level imaging is night vision. In the absence of sunlight or illuminators, the only sources of illumination are moonlight, airglow, and starlight. Airglow is a light emission in the visible, near-infrared, and shortwave infrared regions that occurs in the ionosphere due to the recombination of ions created by sunlight and cosmic rays earlier in the day. Imaging without gain requires full moonlight to obtain a usable image at standard video rates using the collection optics of a typical handheld camera. 10 4 or 10 6 Night vision goggles and image intensifiers with gain greater than 100 MHz can be used to image on moonless, overcast nights. The illumination sources are airglow and starlight scattered and attenuated by clouds. The gain provided by a linear-mode APD without excess noise is comparable to that of night vision goggles and image intensifiers, so this APD is expected to image on moonless, overcast nights. A key difference is that the size of the APD is much smaller. Another key difference is that the output of night vision goggles is an image on a phosphorescent screen that cannot be read out for archiving or image processing without a separate camera, whereas the output of an imager made from an array of linear-mode APDs is digitized.

[0171] Spectroscopy and Chemical and Biological Agent Detection - Operationally, spectroscopy is equivalent to low-light-level imaging, where the scene is the output of a dispersive element such as a diffraction grating. The gain of an array of linear-mode APDs can be adjusted to provide sufficient gain for signal photon detection but not too much, providing a large dynamic range. Additionally, similar to single-shot LIDAR applications, the waveforms seen by each APD in the array can also be stored. Similar to single-shot LIDAR, this capability allows for the acquisition of an entire time-resolved spectrum triggered by a single event.

[0172] The ability to perform spectroscopy at the single-photon detection level can be used to detect molecules at very low concentrations. Time-resolved spectroscopy adds the ability to monitor molecular dynamics, which can provide information about the molecule's environment. The ability to detect trace amounts of specific chemical and biological molecules is useful in the fields of security, where these molecules may be illegal drugs, chemical or biological warfare agents, or explosives; quality control, where these molecules may be impurities or unwanted by-products of manufacturing processes; and biomedical research, where these molecules may be new drugs under development.

[0173] Quantum Key Distribution - Quantum key distribution requires the detection of single photons with high quantum efficiency, preferably at temperatures near room temperature. Currently available room-temperature single-photon detection technologies are photomultiplier tubes and Geiger-mode APDs. Photomultiplier tubes have very low quantum efficiency, especially at wavelengths typically used in telecommunications. Geiger-mode APDs have effectively low quantum efficiency due to their dead time. Linear-mode APDs, without excess noise, are believed to be capable of detecting single photons with high quantum efficiency.

[0174] Optical Communications - The dominant noise in most, if not all, optical communication systems is thermal noise in the receiver electronics. Therefore, a fundamental challenge in optical communications is ensuring that the optical signal reaches the receiver with enough power to exceed the receiver's noise-equivalent input. Common methods for ensuring this include increasing the power produced by the laser, shortening the optical link length to reduce losses, and incorporating repeaters into the link. Linear-mode APDs without excess noise at the receiver provide a lower-cost, less disruptive method of boosting the incoming optical signal. On the other hand, lasers with high power output are expensive, increased laser power can introduce new problems, such as nonlinear effects in optical fiber or atmospheric transmission media, link lengths are often fixed by application or geography and cannot be shortened, and repeaters are prohibitively expensive, especially in areas such as the ocean floor where electrical power is not readily accessible.

[0175] High Breakdown Voltage Transistors - The output power of a transistor is limited by its breakdown voltage. As discussed above, single-carrier initiated impact ionization (α=0 or β=0) results in an exponentially increasing, but not arbitrarily large, carrier density. Furthermore, for superlattices with α=β=0, achieved by engineering localized Wannier-Stark states in both the conduction and valence bands, the exponential increase in carrier density is suppressed. Carrier transport through a superlattice is by hopping. Increasing the wavefunction overlap between adjacent Wannier-Stark states can shorten the hopping time, giving carriers effective velocities comparable to the saturation velocity.

[0176] The highest electric field in a bipolar transistor is in the collector, near the collector-base junction. Because the current flow in a bipolar transistor is usually perpendicular to the surface of the semiconductor, it would be simple to replace the collector of a bipolar transistor with a superlattice with α=β=0, where the normal of the superlattice is aligned with the current flow.

[0177] Low-Noise Microwave Generation—Impact ionization is a critical part of the operation of certain classes of microwave diodes, including Read diodes and impact ionization avalanche transit time (IMPATT) diodes. With impact ionization providing gain and carrier transit time providing the appropriate phase shift, they are used as negative differential resistance devices to generate microwaves. While these generators are low cost and can produce high power, the high phase noise of their output limits their applications. The source of the noise is excess noise in the impact ionization mechanism. Using the superlattices described herein as the impact ionization layer of these diodes suppresses the excess noise, allowing microwave generators built around these diodes to be low cost, high power, and low noise.

[0178] References herein to "embodiments," "examples," and the like indicate that the described embodiment or example may include a particular feature, structure, or characteristic, but not all embodiments include that particular feature, structure, or characteristic. Such phrases do not necessarily refer to the same embodiment. Furthermore, when a particular feature, structure, or characteristic is described with respect to one embodiment, it is believed that the impact of such feature, structure, or characteristic on other embodiments, whether or not explicitly stated, is within the knowledge of one of ordinary skill in the art.

[0179] While the above designs and embodiments thereof have been described in considerable detail, it is not the intention of the applicants of the present invention that the designs and embodiments described herein be limited to them. Additional adaptations and / or modifications are possible, and the broader aspects encompass these adaptations and / or modifications. Accordingly, departures may be made from the above designs and embodiments without departing from the scope given by the following claims, which scope is limited only by those claims when properly interpreted.

Claims

1. 1. An apparatus comprising a linear-mode avalanche photodiode, The linear mode avalanche photodiode comprises: a lattice-matched pair of first and second semiconductor alloys in a superlattice multiplication region that allows only one current carrier type to be selected from i) electrons or ii) holes to store sufficient kinetic energy for impact ionization when the photodiode is electrically biased to conduct current, the lattice-matched pair having a band offset that is set to include at least a majority of the band corresponding to inhibited carriers that inhibit impact ionization; The linear mode avalanche photodiode including the matched superlattice structure is configured so that amplification occurs i) in the conduction band only, or ii) in the valence band only, producing gain while generating sufficiently low excess noise from impact ionization so that single photons can be detected without the penalty of dead time, thereby allowing detection of subsequent photons substantially immediately after detection of the single photon.

2. 10. The apparatus of claim 1, wherein the linear-mode avalanche photodiode including the matched superlattice is configured to simultaneously achieve single-photon sensitive detection without dead time at temperatures above room temperature.

3. 10. The apparatus of claim 1, wherein the linear-mode avalanche photodiode including the matched superlattice is configured to resolve multiple photons arriving at approximately the same time, and wherein the linear-mode avalanche photodiode has no dead time after detecting an individual photon.

4. 10. The apparatus of claim 1, wherein the linear-mode avalanche photodiode is configured to be an InP-based room-temperature linear-mode avalanche photodiode having a gain of at least 10<4> electrons per photon with a controlled amount of excess noise.

5. 10. The apparatus of claim 1, wherein the linear-mode avalanche photodiode is configured to couple an output of the linear-mode avalanche photodiode to an electronic amplifier to provide a single-pole gain and generate a waveform that can detect the arrival of a single photon above noise and can distinguish between multiple photons arriving simultaneously.

6. 10. The apparatus of claim 1, wherein the linear-mode avalanche photodiode including a matched superlattice structure is configured such that amplification occurs i) in the conduction band only, or ii) in the valence band only, and produces a gain of 10,000 times (10) amplification or more, while generating excess noise due to the gain amplification that is less than three times the thermal noise present at temperatures above room temperature.

7. 2. The device of claim 1, wherein the lattice-matched pair of the first and second semiconductor alloys comprising the multiplication region is an InGaAsP-InAlAs superlattice.

8. 2. The device of claim 1, wherein said second semiconductor alloy is AlGaAsSb, and said first semiconductor alloy is paired with said second semiconductor alloy of AlGaAsSb to form said multiplication region.

9. 2. The device of claim 1, wherein the first semiconductor alloy comprising the multiplication region is InGaAsSb with a superlattice set in the conduction band for electron impact ionization.

10. 1. A method for a linear mode avalanche photodiode, comprising: forming a lattice-matched pair of first and second semiconductor alloys in a superlattice multiplication region, the lattice-matched pair having a band offset set to at least predominantly fall within a band corresponding to inhibited carriers that inhibit impact ionization, the lattice-matched pair allowing only one current carrier type to be selected from i) electrons or ii) holes to store sufficient kinetic energy for impact ionization when the photodiode is electrically biased to conduct current; configuring said linear mode avalanche photodiode including a matched superlattice structure so that amplification occurs in either i) the conduction band only, or ii) the valence band only to produce gain, while generating excess noise from impact ionization that is low enough to allow single photon detection without the penalty of dead time; A method that allows for the detection of a single photon to be performed substantially immediately after the detection of a subsequent photon by configuring the detection of the single photon without the penalty of dead time.

11. 11. The method of claim 10, further comprising configuring the linear-mode avalanche photodiode including the matched superlattice to simultaneously achieve single-photon sensitive detection without dead time at temperatures above room temperature.

12. 11. The method of claim 10, further comprising configuring the linear-mode avalanche photodiode including a matched superlattice to resolve multiple photons arriving at approximately the same time, wherein the linear-mode avalanche photodiode has no dead time after detecting an individual photon.

13. 11. The method of claim 10, further comprising configuring the linear-mode avalanche photodiode to be an InP-based room-temperature linear-mode avalanche photodiode having a gain of at least 10 electrons per photon with a controlled amount of excess noise.

14. 11. The method of claim 10, further comprising configuring the linear-mode avalanche photodiode by coupling an output of the linear-mode avalanche photodiode to an electronic amplifier to provide a single-pole gain and generate a waveform that is capable of detecting the arrival of a single photon above noise and distinguishing between multiple photons arriving simultaneously.

15. 11. The method of claim 10, further comprising configuring the linear-mode avalanche photodiode including the matched superlattice structure such that amplification occurs i) in the conduction band only, or ii) in the valence band only, and to produce a gain of 10,000 times (10) amplification or more, while generating excess noise due to the amplification gain that is less than three times the thermal noise present at temperatures above room temperature.

16. 11. The method of claim 10, further comprising configuring the lattice-matched pair of the first and second semiconductor alloys comprising the multiplication region to be an InGaAsP-InAlAs superlattice.

17. 11. The method of claim 10, wherein the second semiconductor alloy is AlGaAsSb, and the first semiconductor alloy is paired with the second semiconductor alloy of AlGaAsSb to form the multiplication region.

18. 11. The method of claim 10, wherein the first semiconductor alloy comprising the multiplication region is InGaAsSb with a superlattice set in the conduction band for electron impact ionization.

Images