WO1985002296A1

WO1985002296A1 - IMPROVED p-i-n AND AVALANCHE PHOTODIODES

Info

Publication number: WO1985002296A1
Application number: PCT/US1984/001761
Authority: WO
Inventors: Gee-Kung Chang; Adrian Ralph Hartman; Mcdonald Robinson
Original assignee: American Telephone & Telegraph Company
Priority date: 1983-11-10
Filing date: 1984-11-02
Publication date: 1985-05-23
Also published as: EP0142316A3; EP0142316A2; JPS61500396A; EP0163720A1; CA1233549A; US4616247A

Abstract

High-speed p-i-n and avalanche photodetectors (photodiodes) use a heavily doped buried layer (14) to greatly limit minority carriers generated by incident light in the buried layer (16) and the substrate (12) of the device from reaching the cathode (18) and thus enhances device response time while substantially decreasing dark current. A p-i-n diode of this type with a 1.1 square millimeter active area can operate with an unusually small (e.g. 5 volt) reverse bias and is capable of having edge rise and fall times in the 4 nanosecond range.

Description

IMPROVED p-i-n AND AVALANCHE PHOTODIODES

Technical Field

This invention relates to photodiodes and particularly to n+-π-p+ silicon photodiodes of the p-i-n type and to n+-p-τr-p+ silicon avalanche photodiodes. Background of the Invention

One common type of photodiode, which is illustrated in FIG. 3 of U. S. Patent No. 4,127,932, is a p-i-n type semiconductor photodiode with, a p+ substrate which acts as the anode, a p- type epitaxial layer, and an n+ type surface region which acts as the cathode. In order to achieve high-speed operation with short response time, it is desirable to have a relatively thick epitaxial layer and a large enough reverse bias voltage to deplete the epitaxial layer such that essentially all light-induced electrons result in electron- hole pairs in the epitaxial layer, with the electrons relatively quickly exiting the epitaxial layer into the ' cathode, and the holes relatively quickly exiting the epitaxial layer into the anode-substrate. Any electrons which are induced as a result of light which reach the substrate must first relatively slowly diffuse into the epitaxial layer before they move relatively quickly to the cathode. This type of photodiode is typically operated with the magnitude of the reverse bias being sufficient to essentially completely deplete the epitaxial layer. As the thickness of the epitaxial layer is increased so as to enhance response time, the magnitude of the needed applied reverse bias voltage increases as the square of the thickness of the epitaxial layer. This requires in many applications a reverse bias potential of typically 30 volts or more for a p-i-n diode. Many of today's photodiodes operate with solid-state integrated circuits that use a 5 volt power supply. The-requirement of a separate 30 volt supply places additional expense on the overall system, increases potential failure of lower voltage integrated circuits used with the photodiode, and thus is undesirable. Another problem with photodiodes is that the reverse bias leakage current with no light applied, the "dark current", limits operating ranges. It is desirable to have a silicon p-i-n photodiode which operates with a supply voltage of typically 5 volts or less, can achieve high-speed operation with relatively short response time, and has a relatively small dark current. It is also desirable to have a silicon avalanche photodiode which can be operated in the 100 volt range, can achieve high-speed operation with short response time, and has a relatively small dark current. Summary of the Invention

The present invention is directed to a p-i-n or avalanche type photodiode which uses a p++ type buried layer sandwiched between the conventional p- type epitaxial layer and the p+ type substrate. The p++ type buried layer limits slow diffusing minority carriers generated in the substrate from reaching the depletion layer, hence the epitaxial layer can be made much thinner without regard to the increased minority carriers thereby generated in the substrate. The thinner epitaxial layer allows lower voltage operation. Brief Description of the Drawings FIG. 1 is a cross-sectional view of a semiconductor photodiode structure in accordance with one embodiment of the present invention; and

FIG. 2 is a cross-sectional view of a semiconductor photodiode structure in accordance with another embodiment of the present invention. Detailed Description

Referring now to FIG. 1, there is illustrated a semiconductor photodiode structure 10 in accordance with one embodiment of the present invention*. Photodiode 10 comprises a semiconductor substrate 12 of one conductivity type and of relatively high impurity concentration (heavily doped), a deep buried semiconductor layer 14 (which may be denoted as a minority carrier "killer layer") of the one conductivity type and of higher impurity concentration than substrate 12. an epitaxial layer 16 of the one conductivity type and of relatively low impurity concentration (lightly doped), and a semiconductor region 18 of the opposite conductivity type and of relatively high impurity concentration. A semiconductor region 24 encircles region 18 and is separated from same by portions of epitaxial layer 16. Separate electrodes 22, 26 and 28 are electrically coupled to region 18, region 24, and substrate 12, respectively. Portions of regions 24 and 18 and layer 16 share a common top surface 20 of photodiode 10. An anti-reflective layer 30 exists on a portion of surface 20 above portions of region 18. A dielectric layer 34 exists on portions of surface 20 and separates electrodes on surface 20. Light 32 is illustrated incident upon layer 30.

In one illustrative embodiment, substrate 12, layer 14, layer 16, region 18, and region 24, are respectively of p+. p++, p-, n+, p+ type silicon, and layer 30 is of silicon nitride, respectively. Electrodes 22, 26, and 28 are aluminum. Region 24 is rectangular and acts as a guard ring. Region 18 is rectangular and acts as the cathode and substrate 12 acts as the anode. Photodiode 10 is typically operated in reverse bias with cathode region 18 held at a more positive voltage than anode substrate 12. The conductivity type of each of the regions can be reversed and region 18 and substrate 12 would then become the anode and cathode, respectively. Incident light 32 on layer 30, which is essentially transparent, reaches and causes charge carrier generation in region 18, layers 14 and 16 and substrate 12 which results in the generation within photodiode 10 of a photo-induced current. A substantial portion of layer 16 is depleted under region 18 by the applied reverse bias. Electrons generated in region 18 are quickly discharged

OMPI through electrode 22 to a power supply or through circuitry which are both not illustrated. Electrons generated in layer 16 rapidly drift through depleted epitaxial layer 16 and rapidly pass through cathode 18 and exit through electrode 22. Electrons generated in layer 14 have a very short lifetime and therefore recombine with holes very rapidly such that most of these electrons do not reach epitaxial layer 16 or electrode 22. Electrons generated in substrate 12 are repelled by the electrical field resulting from the high impurity concentration of layer.14 and thus never reach layer 16 or electrode 22. Holes generated in layers 14 and 16 and in substrate 12 rapidly move to electrode 28 and then out of structure 10 to circuitry or a power supply (both not illustrated) coupled to electrode 28. The central portion of region 18 is purposely kept thin such that few holes recombine there. As a direct result of layer 14, the response time of photodiode 10 is relatively fast since essentially all electrons generated in layer 14 and substrate 12 are reco bined and do not slow up the response time. Also, because of the presence of the layer 14, the layer 16 can be made thinner, e.g., 10 microns, instead of the prior art thickness of 30 microns. This follows because, while a thinner layer 16 allows greater penetration of light into the substrate 12, hence a greater generation of electrons therein, the presence of the layer 14 prevents (as just described) this greater number of electrons from slowing the device response time. The use of a thinner layer 16 allows the use of a lower bias voltage, as in often desired. Because the thinner layer 16 results in less capture of energy therein, hence less generation therein of photo-induced current, the device suffers from some loss of sensitivity, but not to such an extent as to seriously impair the usefulness of the inventive devices. Reverse bias leakage with no light (known as dark current) applied to photodiode 10 is significantly reduced because layer 14 causes the minority carriers thermally generated in layer 14 and substrate 12 to recombine and thus they never reach cathode electrode 22. In addition, the electric field at the boundary of layer 14 and substrate 12 repels electrons back into the substrate and thus limits the number of electrons which reach layer 16. Measured dark current was found to be greater than one order of magnitude lower when layer 14 is used as compared to when it is not used. Measured dark current at 25 degrees C is less than 20 picoamps with layer 14 present and is 3000 picoamps when layer 14 is absent. The corresponding figures at 70 degrees C are 100 picoamps and 100,000 picoamps. respectively.

A 0.825 micron wavelength laser is used, in one application of the device 10, to provide light signal 32. The pulse width of the applied light signal 32 is

18 nanoseconds. The 10 percent to 90 percent rise and fall times of the light signal 32 is approximately 0.5 nanoseconds. With photodiode 10 having a total photosensitive area of 1.1 square millimeters and with a reverse bias applied to photodiode 10 (electrode 28 connected to a -4 volt power supply and electrode 22 connected ground potential), the 10 percent to 90 percent rise and fall times of photo-current induced in photodiode 10 are both approximately 4 nanoseconds. With 30 volts of applied reverse bias the rise and fall times are approximately two nanoseconds each. If photodiode 10 is fabricated without layer 14, then with an applied reverse bias voltage of 4 volts the rise and fall times are both greater than approximately 16 nanoseconds. With 30 volts of applied reverse bias, the rise and fall times are two nanoseconds each when there is no layer 14 present.

In many optical fiber communications, the photosensitive area of photodiode 10 can be one to two orders of magnitude smaller than the 1.1 square millimeters of the fabricated device. The overall area of the photodiode is also correspondingly reduced. This much smaller device is expected to have rise and fall times in one (1) nanosecond range.

Referring now to FIG. 2, there is illustrated an avalanche type photodiode 100 which is very similar to photodiode 10 of FIG. 1 except that it includes an additional semiconductor region 36 which in one illustrative embodiment is of p type conductivity. All regions, layers, electrodes, surfaces and substrates of photodiode 100 which are similar to those of photodiode 10 of FIG. 1 have the same reference symbol with a "0" added at the end.

Buried layer 140 of photodiode 100 enhances the response time, significantly reduces the dark current, and allows lower reverse bias potentials in substantially the same manner as buried layer 14 of photodiode 10 of FIG. 1. Photodiode 100 is typically operated at a reverse bias near • the avalanche breakdown voltage and the operating potentials can be in the 100 volt range.

Other arrangements are possible. For example, layers 14 and 140 can be replaced by other recombination center mechanisms which also provide for recombination of minority carriers generated in substrate 12 or 120. Crystallographic defects can also act as recombination centers, especially when decorated with impurities. A high density of defects such as misfit dislocations or stacking faults at the boundary between regions 16 and 160 and substrates 12 and 120, respectively, could also serve as the desired recombination center for minority carriers. Misfit dislocations are generated by a high concentration diffusion of a dopant whose atomic radius differs from that of silicon. A germanium-doped silicon layer used between regions 16 and 160 and substrates 12 and 120, respectively, could provide the desired misfit dislocations. Electrodes 28 and 280 could be coupled through p+ type deep regions (not illustrated) to electrodes- making contact with^* top surfaces 20 and 200. A reverse bias potential can be applied in some applications to electrodes 26 and 260. Electrodes 26 and 260 can be directly connected to electrodes 28 and 280, respectively. Regions 18, 24, 18 and 240 can be circular. Photodiodes utilizing the present invention can be fabricated with materials such as Ge, compound III-V, and ternary and quaternary materials and heterojunctions. The guard rings can actually come in contact with the cathode regions. The thickness of the epitaxial layers can be varied to accommodate different light wavelengths and different materials.

Claims

1. A semiconductor photodiode having a semiconductor body (12) of a first conductivity type and of relatively high impurity concentration, a first semiconductor layer (16) of the same conductivity type and of relatively low impurity concentration, a first semiconductor region (18) of the opposite conductivity type and of relatively high impurity concentration forming a photosensitive semiconductor junction with said first layer,

CHARACTERIZED BY: a second semiconductor layer (14) of the first conductivity type and of higher impurity concentration than the semiconductor body sandwiched between the first layer and the body.

2. The semiconductor photodiode of claim 1 further characterized by: a second semiconductor region (FIG. 2, 36) of the first conductivity type having an impurity concentration greater than the impurity concentration of the first semiconductor layer (160) forming an extension of said first layer and providing said junction with said first region (180).