Planar dielectrically isolated high speed photodiode

Abstract

A method is shown for producing a PIN photodiode wherein an intrinsic layer of the photodiode can be made arbitrarily thin. A fabrication substrate is lightly doped to have a first conductivity type in order to form the intrinsic layer of the photodiode. A first active region of the photodiode having the first conductivity type is formed on a first surface of the fabrication substrate. An oxide layer is also formed upon the first surface of the fabrication substrate. A handling substrate is bonded to the first surface of the fabrication substrate. A second surface of the fabrication is then lapped to a obtain a preselected thickness of the intrinsic layer. And a second active region of the photodiode having a second conductivity type is formed on the second surface of the fabrication substrate.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention is directed toward a structure and improved manufacturing process for a planar dielectrically isolated high speed PIN photodiode.

2. Description of the Related Art

Photodiodes are diodes in which charge carriers are generated responsive to light incident upon the photodiode. Any PN junction diode which admits light can function as a photodiode. A photodiode outputs voltage or current when absorbing light in a photodiode which is intended for high speed communication systems, it is important to optimize the performance for light conversion efficiency, speed (minimal transit time delay), minimum RC time constant, ability to operate at low reverse bias voltage, and cost in the application in which the photodiode will be employed.

FIG. 1 illustrates the structure of a conventional PIN photodiode 10. A wafer 12 is lightly doped with N dopant in order to produce an intrinsic region 16. A P+ region 14 is formed on one surface of the wafer and an N+ region 18 is formed on the opposing surface of wafer 12 with intrinsic region 16 interposed P+ region 14 and N+ region 18. A reflective layer 20 is disposed on the surface containing N+ region 18 with reflective layer 20 also acting as the electrical contact to N+ region 18. A metal contact ring 22 is disposed on the surface containing P+ region 14 to provide the electrical connection to the P+ region.

Typically, one power supply potential is applied to the reflective layer and another power supply voltage is applied to contact 22 to reverse bias the PN junction formed by P+ region 14 and N+ region 18. This forms a depletion region 17 within the intrinsic region 16 wherein electron and hole charge carrier pairs generated by light photons incident upon the intrinsic region 16 are rapidly accelerated toward the P+ and N+ regions respectively by the electric field of the reverse bias voltage. Charge carrier pairs are also typically generated outside the depletion region 17 in non-depletion regions 24A and 24B of intrinsic region 16 and diffuse, due to random thermal motion of the carriers, at a much slower velocity until they reach either depletion region 17 or the junction formed by P+ region 14 and intrinsic region 16 of photodiode 10.

A conventional photodiode that is designed for high quantum, i.e. light conversion, efficiency requires that the light path within the photo current collection zone, the non-depletion 24 and depletion 17 regions, be sufficient in length so that most of the light photons of the incident light signal are absorbed and converted into electron-hole pairs that are collectable at the P.sup.+ and N.sup.+ regions. Usually, this requires that the width W2 of the intrinsic region 16, which is the primary collection region between the P.sup.+ 14 and N.sup.+ 18 regions, be several times the length required for light absorption. If diode 10 has an efficient back-side reflector, such as reflective layer 20, which effectively doubles the light path within diode 10, then the intrinsic region 16 of the photodiode can be made narrower. For a typical near infrared silicon photodiode, the nominal absorption path length is about 15-25 microns. The path length should be at least two to three times the nominal absorption path length to obtain good light conversion efficiency.

On the other hand, a photodiode designed for high frequency response requires that the photo current pairs generated by the light signal be collected rapidly and that the diode RC time constant is fast. Rapid photo current pair collection usually requires that most of the photo current pairs generated by the light signal be generated within the depletion region 17 which has a high drift velocity when reversed biased. Otherwise, the photo generated charge carrier pairs produced in the non-depletion regions 24A and 24B outside the depletion region 17, but within the diffusion distance of the collection electrodes 14 and 18, will have a diffusion velocity which is several hundred times slower than the velocity of the pairs generated within the depletion region 17. The photo generated charge carrier pairs in non-depletion regions 24A and 24B will slowly migrate for collection at P+ region 14 and N+ region 18 resulting in a tail on the trailing edge of the electrical signal corresponding to the light signal. The diffusion distance of the charge carriers is determined by the carrier mean free path before re-combination and may exceed 150 microns.

A fast RC time constant for photodiode 10 requires minimal capacitance and low series resistance between the electrical contacts 20 and 22 and the photo current pair collection sites at the margin between the depletion region 17 and P+ region 14 and the margin between depletion region 17 and N+ region 18. The greater the width W2 of the depletion layer 16, then the lower will be the capacitance per unit area of photodiode 10. Since the depletion width of the depletion region formed between P+ region 14 and N+ region 18 increases with the level of the reverse bias voltage, it is typical for high speed photodiodes to have a relatively high reverse voltage applied to them.

The inclusion of the separate lightly doped intrinsic region 16 between the P.sup.+ and N.sup.+ regions 14 and 18 results in a PIN diode with a wider depletion region 17, depending on reverse bias voltage, which improves the light collection efficiency, speed, and reduces capacitance over that of a simple PN photodiode structure. Tailoring the width of the intrinsic region 16 allows for enhanced performance and tradeoffs for photodiode light conversion efficiency, response speed, and capacitance.

For example, a near infrared photodiode intended for use in a high speed, low cost IrDA data receiver operating from a 2.7V-5V power supply should ideally have an intrinsic layer width W2 of about 20-40 microns wide to allow for good light absorption efficiency and to allow for full depletion of the intrinsic region 16 with a low 1-3V averse bias, since typically not all of the power supply voltage is available for reverse bias. Such a diode will achieve minimal transit delay, less than 1 nanoseconds (ns), a minimal RC time constant, and optimal high current frequency response with low resistance in the intrinsic region 16.

Although a PIN photodiode outperforms a standard PN diode, it cannot be easily manufactured by standard semiconductor processes wherein fabrication is typically performed on only one side of the semiconductor wafer 12.

A PIN photodiode is typically produced by diffusing the N.sup.+ diffusion region 18 on the back side of the lightly doped (N) wafer 12, diffusing the P.sup.+ diffusion region 14 on the topside of wafer 12, and then adding metal contacts to each side of the wafer. Typically, the backside contact area connected to N+ region 18 is reflective layer 20 and is made of gold. The topside contact area 22 is an aluminum collector ring that is connected to P+ diffusion region 14. The intrinsic or depletion layer depth W2 is determined by the wafer starting thickness W1 less the thickness of the N.sup.+ 18 and P.sup.+ 14 diffusion regions. Since standard silicon wafers are 350-500 microns in thickness and N+ and P+ diffusions are only a few microns thick, this typically results in an intrinsic layer width W2 of 345-495 microns.

An improvement to the PIN manufacturing process described above is to lap the width W1 of the starting wafer 12 to as thin as 100 microns, which will reduce the intrinsic layer width W2 to about 95 microns. However, it is generally not practical to thin wafers beyond this limit without an excessive level of wafer breakage along with severe wafer handling and processing problems.

A PIN diode with an intrinsic region width W2 of 95 microns typically requires more than a 5V reverse bias to be applied to the P+ and N+ regions 14 and 18 in order to completely deplete the intrinsic region 16 and achieve optimal frequency response for the photodiode. Consequently, the use of such a PIN photodiode in a high speed data receiver operating from a standard power supply voltage level of 2.7V-5V results in degraded speed performance.

Another problem with the structure of PIN diode 10 is that the connection of reflective layer 20 on the backside of wafer 12 requires the conductive bonding of the die of photodiode to a conductive substrate which may not be at the desired power supply potential or may not be a convenient electrical connection in the receiver design. In addition, if this conductive substrate is the active output of a photodiode it may undesirably act as an antenna for noise pickup.

Another method which has been attempted to produce a low cost PIN diode is to grow an epitaxial intrinsic layer on top of an N.sup.+ or P.sup.+ diffusion. However, because such an epitaxial layer is grown it high temperature it has high auto-doping levels due to the diffusion of the dopants of the underlying diffusion region which consequently prevents the formation of a lightly doped intrinsic layer 16.

Another method for producing a PIN diode 10 that is more successful is to grow an insulating oxide layer on top of the N.sup.+ or P.sup.+ diffusion and then to grow a thick polysilicon layer of several hundred microns to act as a handling layer so that the wafer 12 may be lapped as thin as needed to obtain the desired intrinsic region width W2. Following the lapping operation, the PIN diode can be processed in the standard way. Although this method is proven effective, growing a thick polysilicon layer is an expensive processing step.

One problem with the structure of PIN photodiode 10 is that the junction between the P+ region 14 and intrinsic region 16 does not extend to the edge of the photodiode. This is done in order to prevent shorting of the junction to the intrinsic region when cutting wafer 12 to obtain the individual die for each photodiode. Consequently, the depletion region 17 produced by the reverse bias voltage does not fully extend to the edge of the die resulting in non-depletion regions 24A and 24B at the edge of photodiode 10 which can be significant sources of slow diffusing photo generated carriers. A metal light guard ring is typically placed over the edge region of photodiode 10 to minimize generation of carriers in drift regions 24A and 24B, but this does not fully suppress the formation of slow carriers.

Accordingly, it is an object of the present invention to reliably produce a PIN photodiode having high speed operation at low operating voltages.

SUMMARY OF PRESENT INVENTION

The present invention describes two improved methods utilizing standard IC processing technology, V-groove or trench etching technology, lapping technology, combined with the novel use of a handling wafer to produce a PIN diode featuring 1) an arbitrarily thin intrinsic region which could be depleted at a low operating voltage, 2) complete dielectric isolation of the backside of the photodiode from the mounting substrate, and 3) fill dielectric isolation of the side walls which ensures the complete depletion of the active region and totally preventing the generation of slowly diffusing carriers.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 describes the structure of a conventional PIN photodiode.

FIGS. 2A-L illustrate a series of steps for an embodiment of a process for producing a PIN photodiode according to the present invention.

FIGS. 3A-L illustrate a series of steps for another embodiment of a process for producing a PIN photodiode according to the present invention.

DETAILED DESCRIPTION OF THE PRESENT INVENTION

FIGS. 2A-L describe an improved PIN photodiode fabrication method which utilizes a standard thickness handling wafer with an isolating oxide layer grown on its surface that is bonded to the N.sup.+ diffusion side of a fabrication wafer. The handling wafer allows the intrinsic region of the fabrication wafer to be lapped to a thin thickness, as required for optimum performance of the PIN photodiode, while the handling wafer provides mechanical rigidity needed for processing. N.sup.+ and P.sup.+ regions are then diffused into the fabrication wafer. Next, diode trenches are etched around the periphery of each photodiode down to the N+ region. The sidewalls of the trenches are then N+ diffused before being covered with an oxide layer and then filled with polysilicon. Finally, coplanar metal contacts to the N+ and P+ regions are formed.

The process begins with fabrication wafer 212 of FIG. 2A, which is lightly N doped to form an intrinsic layer 216 which will later become the intrinsic region of a PIN diode. An N+ diffusion layer 218 is then formed on fabrication wafer 212, as shown in FIG. 2B.

Optionally, an oxide layer 220 can be formed on the N+ layer 218 on fabrication wafer 212 in order to form a dielectric interference reflector which improves the light absorption efficiency of the PIN photodiode, as shown in FIG. 2C. Alternatively, the dielectric interference reflector can be formed by producing an oxide layer on the handling wafer 230.

A handling wafer 230 is then bonded to a surface 232 of fabrication wafer 212 that includes the N+ diffusion 218, as shown in FIG. 2D. Bonding is accomplished by polishing the opposing faces of the handling wafer 230 and fabrication wafer 212 to a fine level of planarity. When the resulting highly planar surfaces of the handling wafer 230 and fabrication wafer 212 are subsequently pressed together, the polished surfaces will bind, to some degree, to one another through Van der Waal's cohesive forces. The handling wafer 230 and fabrication wafer 212 will become further bonded during the heating steps associated with subsequent processing steps, such as during the diffusion of dopants. The oxide layer 220 becomes important during these subsequent heating cycles as a barrier between the N+ diffusion 218 and the contaminants present in handling wafer 230. Also, it is important that the handling wafer 230 and fabrication wafer 212 have the same coefficient of thermal expansion to prevent warping or separation of the two wafers during the subsequent heating cycles.

The structure formed by the fabrication wafer 212 bonded to the handling wafer 230 is then flipped, as shown in FIG. 2E, so that further fabrication processing can be performed on fabrication wafer 212. The intrinsic layer 216 can then be lapped to the desired thickness for optimal performance of the PIN diode by lapping fabrication wafer 212. In the process of the present invention, the thickness of intrinsic layer 216 is not constrained by requirements for mechanical integrity of the fabrication wafer 212 needed for subsequent processing steps, as is the conventional PIN photodiode and process. The handling wafer 230 provides the mechanical integrity needed for the handling required for further processing of the fabrication wafer 212 while the intrinsic layer 216 can be lapped to the optimal thickness for the operation of the PIN photodiode.

After the intrinsic layer 216 is lapped, P+ regions 214 are diffused into fabrication wafer 212 along with N+ regions 240. Oxide regions 250 are fabricated between P+ regions 214 and N+ regions 240 with the resulting structure shown in FIG. 2G.

Isolation trenches 242A and 242B are then formed in N+ regions 240 and intrinsic layer 216 down to N+ layer 218, as shown in FIG. 2H. The sidewalls of trenches 242A and 242B are then N+ diffused to form a contiguous N+ region from layer 218 through trenches 242A and 242B to the surface of fabrication wafer 212, as shown in FIG. 2I. An oxide fill is then performed on fabrication wafer 212 that forms an oxide layer 252 that fills trenches 242A and 242B and covers the P+ regions 214 with the resulting structure shown in FIG. 2J.

As shown in FIG. 2K, oxide layer 252 is then etched in the area of the N+ wall of trench 242A and the P+ region 214. Contact metal is added to these etched areas to form the cathode contact 244 to N+ layer 216 via trench 242A and the anode contact 222 to P+ region 216. The semiconductor die are then cut from fabrication wafer 212 and handling wafer 230 to form the individual PIN photodiode 270 shown in FIG. 2L.

The resulting PIN photodiode 270 has a thinner intrinsic region 216 than can be obtained with conventional processes so that the thinner depletion region formed in the diode by lower reverse bias voltage levels will still permit high speed operation. In addition, isolation trenches 242A and 242B isolate the P+ region 214 and N+ region 218 of photodiode 270 from photo generated charge carrier pairs formed outside of intrinsic region 216, which again allows for higher speed operation of the photodiode.

Furthermore, photodiode 270 can be formed using standard semiconductor fabrication processes, rather than photodiode specific processes. And cathode contact 244 and anode connection 222 are coplanar, which allows for standard same-side bonding methods to be used to connect the photodiode to a receiver circuit and allows for additional flexibility in the polarity-of the power connections to the photodiode.

FIGS. 3A-L illustrate a process for forming another PIN photodiode according to the present invention. In the process of FIGS. 3A-L, V-grooves are etched in a lightly doped fabrication wafer and then an N+ diffusion is performed around the periphery of what will ultimately become the photodiode. The N+ diffusion is overlaid with an oxide layer and the V-grooves are filled with polysilicon. The fabrication wafer is then polished and bonded to a handling wafer that provides mechanical strength. The intrinsic layer formed in the fabrication wafer can then be lapped to the thickness desired for optimal performance of the photodiode while the handling wafer provides the mechanical support for further processing of the fabrication wafer into a finished photodiode.

FIG. 3A shows a fabrication wafer 312 that is lightly N doped to form intrinsic layer 316. Fabrication wafer 312 is then etched to form V-grooves 342A and 342B, as shown in FIG. 3B. An N+ region 318 is then diffused into the etched surface of fabrication wafer 312 including the sidewalls of V-grooves 342A and 342B, as shown in FIG. 3C. A field oxide layer 320 is then formed upon fabrication wafer 312 which covers the N+ region 318 and fills V-grooves 342A and 342b with the resulting structure shown in FIG. 3D. Optionally, the thickness of field oxide layer 320 can be controlled so as to form a dielectric interference reflector for improved light absorption efficiency of the resulting photodiode.

A polysilicon layer 321 is formed upon the field oxide layer 320. A surface 332 of the polysilicon layer is polished and fabrication wafer 312 is flipped, as shown in FIG. 3E. A handling wafer 330 is then bonded to the fabrication wafer at surface 332 to obtain the structure shown in FIG. 3F.

The polysilicon layer 321 has a thermal coefficient which is the same or approximately the same as the thermal coefficient of the silicon of fabrication wafer 312 and handling wafer 330. The thermal coefficient of the buried field oxide layer 320 is significantly different from the thermal coefficient of silicon. The oxide layer 320 is therefore kept thin, approximately 2 microns or less, and the polysilicon layer 321 is formed to prevent cracking of the fabrication wafer 312 as the fabrication wafer and handling wafer 330 are subjected to heating and cooling cycles during further processing.

The intrinsic region 316 of fabrication wafer 312 can now be lapped to the desired thickness, in order to obtain the desired characteristics from the photodiode that is ultimately produced by this process, because handling wafer 330 provides the mechanical support required for the further handling needed to complete the fabrication process on the photodiode of fabrication wafer 312. Lapping fabrication wafer 312 results in the structure shown in FIG. 3G. Note that V-grooves 342A and 342B are of sufficient depth that, after lapping, the grooves reach the lapped surface and field oxide layer 320 isolates the N+ region 318 of each diode from the N+ region of other diodes. Thus, separate intrinsic regions 316A-C and N+ regions 318A-C are formed for each photodiode produced by the process and each photodiode is dielectrically isolated.

P+ regions 314A, 314B and 314C are then diffused into fabrication wafer 312 to form the PIN structure of the photodiodes, as shown in FIG. 3H. The N+ regions 318A-C are then further extended on the surface of fabrication wafer 312 by N+ diffusing additional surface contact area, as shown in FIG. 3I. Regions of field oxide 352A-E are then formed on the surface of fabrication wafer 312 where the regions of field oxide 352A-E leave contact openings to P+ regions 314A-C and N+ regions 318A-C, as shown in FIG. 3J. Contact metal is then added in the contact openings formed by the regions of field oxide 352A-E to form contacts 322A and 322B to P+ regions 314A and 314B, respectively, as well as contacts 344A and 344B to N+ regions 318B and 318C, respectively, as shown in FIG. 3K. Contacts will be formed to all the photodiodes formed on fabrication wafer 312, even though only a subset of the total number of photodiodes is illustrated in the drawings.

Finally, fabrication wafer 312 and handling wafer 330 are diced to obtain the individual photodiode die. A resulting individual photodiode die is shown in FIG. 3L.

The processing method described above has several advantages over conventional processing techniques. One advantage is that the rigidity of the handling wafer 330 allows the fabrication wafer 312 to be lapped as thin as desired such that the intrinsic region 318 of the resulting PIN photodiode obtains optimum performance without the danger of breakage during further processing due to the thin nature of the fabrication wafer. Another advantage is that V-grooves 342A and 342B allow planar topside connections for the anode 322B and cathode 344A of the PIN photodiode which avoids the necessity of conductive die bonding for the backside diode connection experienced with conventional PIN photodiodes. Furthermore, the V-grooves isolate the depleted active region, formed in intrinsic region 316B when the photodiode is reverse biased, from the non-depleted regions at the periphery of the photodiode which blocks the flow toward P+ region 314B and N+ region 318B of slow diffusing photogenerated charge carrier pairs arising in drift regions. This improves the photodiode speed of response.

Also, controlling the thickness of the buried field oxide layer 320 allows the creation of a dielectric reflector for improved light absorption efficiency. Finally, it is advantageous that the handling wafer at the backside of the photodiode is dielectrically isolated from the diode and therefore can be biased as required or grounded as a shield to minimize signal interference problems.

Claims

1. A PIN photodiode, the photodiode comprising:

a first semiconductor substrate having first and second parallel planar surfaces;

a first doped region having a first dopant type, wherein the first doped region extends from the first planar surface into the first semiconductor substrate;

a second doped region having a second dopant type, wherein the second doped region extends from the first planar surface into the first semiconductor substrate and is spaced apart from the first doped region such that the second doped region surrounds the first doped region within the first semiconductor substrate;

an intrinsic region being lightly doped with the second dopant type and disposed between the first and second doped regions such that the intrinsic region surrounds the first doped region and separates the first and second doped regions;

a first contact coupled to the first doped region;

a second contact coupled to the second doped region; and

an oxide layer disposed adjacent the second doped region and extending from the second planar surface of the first semiconductor substrate, wherein a thickness of the oxide layer is controlled such that the oxide layer functions as a dielectric interference reflector.

2. A PIN photodiode, the photodiode comprising:

a first semiconductor substrate having first and second parallel planar surfaces;

a first doped region having a first dopant type and extending from the first planar surface into the first semiconductor substrate;

an intrinsic region being lightly doped with the first dopant type and extending from the first planar surface into the first doped region such that the first doped region encloses the intrinsic region within the first semiconductor substrate;

a second doped region having a second dopant type and extending from the first planar surface into the intrinsic region, the second doped region being separated from the first doped region by the intrinsic region;

a first contact coupled to the first doped region;

a second contact coupled to the first doped region;

an oxide layer extending from the second planar surface into the first semiconductor substrate, wherein a thickness of the oxide layer is preselected such that the oxide layer is configured to operate as a dielectric interference reflector; and

a second semiconductor substrate having a first planar surface, the first planar surface of the second semiconductor substrate being bonded to the second planar surface of the first semiconductor substrate.