This invention relates to a process for fabricating optical waveguides and optical detectors.
To build an optical signal distribution network within a semiconductor substrate, one needs to make good optical waveguides to distribute the optical signals, and one needs to fabricate elements that convert the optical signals to electrical signals in order to interface with other circuitry. Extracting the optical signals can be accomplished in two ways. Either the optical signal itself is extracted out of the waveguide and delivered to other circuitry that can convert it to the required form. Or the optical signal is converted into electrical form in the waveguide and the electrical signal is delivered to the other circuitry. Extracting the optical signal as an optical signal involves the use of mirrors, gratings or couplers within the waveguides, or other elements that function like these devices. The scientific literature has an increasing number of examples of technologies that can be used to construct such devices. Extracting the optical signal as an electrical signal involves the use of detectors within the waveguide, i.e., circuit elements that convert the optical signal to an electrical form. The scientific literature also has an increasing number of examples of detector designs that can be used to accomplish this.
The challenge in finding the combination of elements that produces an acceptable optical distribution network becomes greater, however, when one limits the space of solutions to particular optical signal distribution network designs and takes into account the practical reality that any such designs should be relatively easy to fabricate and financially economical.
The combination of silicon and SiGe alloys (e.g. SixGe1-x) has attracted attention as a useful combination of materials from which one might be able to easily and economically fabricate optical signal distribution networks. With SiGe alloys it is possible to fabricate waveguides in the silicon substrates. The index of refraction of a SiGe alloy is slightly higher than that of silicon. For example, a SiGe alloy with 5% Ge (i.e., Si0.95Ge0.05) has an index of refraction of about 3.52 while crystalline silicon has an index of refraction that is less than that, e.g. about 3.50. So, if a SiGe alloy core is formed in a silicon substrate, the difference in the indices of refraction is sufficient to enable the SiGe alloy core to contain an optical signal through internal reflections. Moreover, this particular combination of materials lends itself to the use of conventional silicon based semiconductor fabrication technologies to fabricate the optical circuitry, and therefore it does not interfere or prohibit the further building of electrical circuitry using the usual CMOS processing technology.
Of course, for such a system to work as an optical signal distribution network, the optical signal must have a wavelength to which both the Si and the SiGe alloy are transparent. Since the bandgap energy of these materials is approximately 1.1 eV, they appear transparent to optical wavelengths having a wavelength greater than 1150 nm. A further reduction in bandgap energy caused by use of a SiGe alloy rather than pure Silicon, and higher temperature operation as high as 125° C. may further require the wavelength be longer than 1200 nm or even 1250 nm for very low absorption loss (approximately 1 db/cm or less). But, the transparency of these materials to optical signals having those wavelengths brings with it another problem. These materials are generally not suitable for building detectors that can convert the optical signals to electrical form. To be a good detector, the materials must be able to absorb the light in a manner so as to create useful charge that can be detected electrically. That is, the optical signal must be capable of generating electron transitions from the valence band to the conduction band within the detector to produce an electrical output signal. But the wavelengths greater than 1150 nm are too long to produce useful absorption by electron transitions in silicon, or in Si0.95Ge0.05 alloys at room temperature. At a wavelength of 1300 nm, the corresponding photon energy is about 0.95 eV, well below the room temperature band gap of silicon and Si0.95Ge0.05 and consequently well below the amount necessary to cause transitions from the valence band into the conductor band.
One detector that meets at least some of the criteria mentioned above is the impurity-based embedded waveguide detector described in U.S. Ser. No. 10/856,127, filed May 28, 2004, entitled “Impurity-Based Waveguide Detector System,” incorporated herein by reference. The impurity-based detector described in that reference is fabricated in a trench that is formed in the substrate.
The embodiments described herein present alternative methods for making such a detector.
In general, in one aspect, the invention features method of fabricating an optical detector in an optical waveguide. The method involves: forming at least one layer on a surface of the substrate, said at least one layer comprising SiGe; implanting an impurity into the at least one layer over a first area to form a detector region for the optical detector; etching into the at least one layer in a first region and a second region to form a ridge between the first and second regions, said ridge defining the optical detector and the optical waveguide; filling the first and second regions with a dielectric material having a lower refractive index than SiGe; and after filling the first and second regions with the dielectric material, removing surface material to form a planarized upper surface.
Other embodiments include one or more of the following features. The first and second regions are first and second trenches, respectively. The dielectric material is Si. The method also involves depositing a SiN layer above the at least one layer, and wherein etching into the at least one layer in the first and second regions to form the ridge involves etching through the SiN layer. Removing surface material to form a planarized upper surface involves removing surface material down to the SiN layer. The method further involves growing a top layer of crystalline silicon on the planarized upper surface, the top layer of crystalline silicon providing a material of sufficiently good quality to fabricate semiconductor microelectronic circuitry within it. Forming the at least one layer involves: depositing a graded layer of SixGe1-x above the substrate; and depositing a uniform layer of SiyGe1-y on the graded layer of SixGe1-x. Alternatively, forming the at least one layer involves; depositing multiple layers of silicon on the substrate each having a different doping level; and after depositing the multiple layers of silicon, depositing said at least one layer comprising SiGe. Forming the at least one layer further involves: on the multiple layers of silicon, depositing a graded layer of SixGe1-x; and on the graded layer of SixGe1-x, depositing a uniform layer of SiyGe1-y, said uniform layer being said at least one layer of SiGe. The multiple layers of silicon include a lowermost layer and wherein each layer after the lowermost layer is characterized by a lower doping level than a preceding layer. Within the graded SixGe1-x layer, the value of x decreases in an upward direction within the layer.
Still other embodiments, include one or more of the following features. The method further involves: forming a first electrically conductive path extending down through the top layer of crystalline silicon and making electrical connection with the silicon in the trenches; and forming a second electrically conductive path extending down through the top layer of crystalline silicon and making electrical connection with an upper portion of the ridge. Forming the first electrically conductive path involves implanting p-type dopant in the top layer of crystalline silicon. Forming the second electrically conductive path involves implanting n-type dopant in the top layer of crystalline silicon. Forming the first electrically conductive path involves: forming a via through the top layer of crystalline silicon; and filling the via with electrically conductive material. The electrically conductive material is doped polysilicon or tungsten. Growing the layer of crystalline silicon on top of the planarized substrate involves epitaxially growing the layer of crystalline silicon. Removing surface material to form the planarized upper surface involves chemical mechanical polishing. The implanted impurity is selected from the group consisting of thallium and indium. Filing the first and second regions involves filling the first and second regions with doped silicon. The method further involves prior to forming at least one layer on a surface of the substrate, etching a wide trench in the substrate, and wherein the first and second regions are located within the wide trench.
In general, in another aspect, the invention features an optical structure including: a substrate; a ridge on the substrate, the ridge made up of at least one layer comprising SiGe, the at least one layer forming a light carrying core region of an optical waveguide, wherein the ridge includes impurities implanted in a region of the at least one layer comprising SiGe to form an optical detector region along the ridge; and dielectric-filled regions on either side of the ridge and defining opposing sidewalls of the ridge, the dielectric-filled regions forming conduction paths for electrically connecting to the optical detector, wherein the dielectric has a lower index of refraction than SiGe.
Other embodiments include one or more of the following features. The dielectric includes silicon. The dielectric-filled regions are dielectric-filled trenches. The optical structure further includes a crystalline silicon layer above the ridge and the dielectric-filled trenches and covering the substrate, the crystalline silicon layer including a first electrically conductive path extending down through the crystalline silicon layer and making electrical connection with the silicon in the trenches and a second electrically conductive path extending down through the crystalline silicon layer and making electrical connection with an upper portion of the ridge.
The details of one or more embodiments of the invention are set forth in the accompanying drawings and the description below. Other features, objects, and advantages of the invention will be apparent from the description and drawings, and from the claims.
The embodiments described herein involve a double trench method for fabricating waveguides with embedded optical detectors. According to this fabrication method, the materials from which the optical detectors and waveguides are formed are first fabricated on the substrate and then double trenches are etched into this material to define the regions (or ridges) that are to form the optical detectors and the waveguides. This fabrication method is in contrast to an alternative fabrication method according to which a single trench is first formed in a substrate and then the optical detectors and waveguide are fabricated within that single trench.
In the embodiments described below, the process starts with high quality material grown on crystalline silicon. More specifically, multiple epitaxial layers, including SiGe layers from which the waveguide and the detector will be fabricated, are first deposited on a silicon substrate. Then, impurities are implanted at multiple levels in the region that will form an optical detector of the type described in U.S. Ser. No. 10/856,127, filed May 28, 2004, entitled “Impurity-Based Waveguide Detector System,” incorporated herein by reference. After the impurities are implanted in the detector region, two trenches are etched into the substrate to define a ridge structure which defines the waveguide with the optical detector embedded therein. The trenches are subsequently filled with doped poly-silicon to make electrical connections to the sides of the optical detector. After planarizing the resulting structure, a high quality silicon layer is formed on top of the wafer to provide the material into which microelectronic circuitry will be subsequently fabricated by using conventional semiconductor fabrication techniques, as described in U.S. Ser. No. 10/623,666, filed Jul. 21, 2003, entitled “Optical Ready Substrates,” also incorporated herein by reference.
A more detailed description of the process will now be described with reference to
Referring to
The sequence of p-type silicon layers 102a-d forms a graded structure in which p-type doping level is reduced in each successive layer that is grown. In other words, the doping level of the silicon decreases as the distance from the waveguide region decreases; or conversely, the doping level of the silicon increases as the distance from the waveguide region increases. This is done to reduce the losses in the waveguide that is fabricated on the silicon. Optical signals will experience greater loss in regions of high doping as compared to regions of low doping. Thus, the design is such as to keep the high doping levels that are needed for electrical contact further away from the optical mode that propagates through the waveguide.
In the case of the graded SiGe layer, the grading serves to reduce the lattice mismatch that would result if the SiGe (6.5% Ge) was grown directly on the silicon material. Reducing the mismatch reduces defects in the waveguide layer and results in a higher quality SiGe core.
In the described embodiment, the actual structure that is grown on the p-type silicon substrate is as follows:
After the waveguide material has been grown on the substrate, the lower electrode region and the absorber region of the optical detector are formed. This is done by first depositing a thin oxide layer 108 (e.g. 100 Å) on top of the SiGe and then covering that with photoresist 110 (
Referring to
After the lower electrode implant is performed, the layer of photoresist is removed by using conventional techniques (e.g. an ashing step followed by a wet clean). Referring to
Referring to
Referring top
The thickness of the upper undoped silicon layer 124 depends on what metal will later be used to connect to the device. If it is aluminum the thickness can be about 500 nm. However, if it is tungsten, the thickness will need to be greater, e.g. 800 nm. If using silicide to reduce the contact resistance, it is preferred to manufacture the contact off to the side of the waveguide.
Referring to
With the lower electrode and absorber impurity implants in place, the wafer is ready for patterning the mesas or ridges that will define the optical waveguides and the embedded detectors. In essence, the epitaxially grown material is removed except in areas that will serve as the optical waveguides with the embedded detectors. In
To form the trenches 128a and 128b, photoresist 130 is again put on the surface and it is lithographically patterned to form openings 132a and 132b that define the trench regions. The openings expose the underlying SiN layer 126. The exposed SiN layer and the underlying SiGe and Si layers are etched away (e.g. by using SF6, C4F8 chemistry) to a depth just below the level of the lower electrode to form the trenches 128a and 128b. In the described embodiment the depth is about 2.8 μm. To withstand an etch to the desired depth, the photoresist protecting the other portions of the surface needs to be sufficiently thick. Thus, the deeper the etch is, the thicker the photoresist will need to be. However, thicker photoresist also degrades line edge roughness which determines the smoothness of the sidewalls of the optical waveguide. Rougher sidewalls lead to greater losses in the transmitted optical signal. All of these considerations need to be taken into account when designing the structure.
After the trenches 128a and 128b have been formed, the photoresist 130 is removed to expose the SiN protective layer 126 on top of the wafer.
Referring to
The doping concentration that is chosen for the silicon deposited in trenches needs to be compatible with the characteristics that are desired for the waveguide. If the doping level is too high, the trench material will bleed off energy from the optical modes that inevitably extend into the trench material. In the described embodiment, the silicon that is grown is 3 μm thick and has a doping of 1×1017 p-type. The amount that is grown needs to be sufficient to fill the trenches and the doping level that is chosen is based on the desired characteristics of the waveguide that is being fabricated.
Using epitaxially grown silicon in the trenches results in a small Δn between the optical waveguide and the trench material. The small Δn tends to permit the optical signal to extend out into the slightly doped and therefore more lossy silicon cladding region but it also reduces the more serious impact of any line edge roughness that might result from the etch process.
For subsequent processes during which a high quality crystalline silicon will be formed above the optical network that has been fabricated, the top of the wafer needs to be flat. So, after the trenches are filled, the surface of the wafer is planarized by using CMP to remove the silicon down to the SiN layer (see
Of course, instead of using CMP one could use other techniques to remove the excess material above the SiN layer. For example, one could alternatively use RIE (reactive ion etch).
After the surface has been adequately prepared, a silicon epitaxial layer is then grown on top of the entire wafer (see
In the described embodiment, electrical contact is made to the underlying silicon-filled trench regions by using ion implantation to form highly doped p-type wells 132 within the epi layer. To adequately dope the silicon down to the silicon-filled trenches typically requires performing multiple implantations each at a different energy level. For example, there is a high energy implant to drive the dopant down to just above the top of the silicon-filled trench, a moderate energy implant to penetrate to the middle of the epitaxial layer, and a lower energy implant to deposit dopant near the top of the epitaxial layer. An anneal is then used to diffuse the dopant out in the well region to make the doping of the well region more uniform.
A similar procedure is used to form highly doped n-type wells 134 extending down to the upper electrode of the impurity detector.
An alternative technique for making electrical contact to the trench regions is to etch vias that extend down through the epitaxial layer to the trench region and the upper electrode. These via are then filled with either a highly doped polysilicon material or a tungsten plug, both of which will provide a good electrical connection to the trench material.
Earlier in the process when the graded silicon layers are deposited (see
The details presented above for the described embodiments are meant to merely be illustrative. The selections of grading structure, depths, doping levels, etc. all represent tradeoffs in performance and design. In part, the selections might also be constrained by the equipment that is available and by circuit design considerations. It might also be desirable to use conventional fabrication processes that are fine tuned for the fabrication of the optical elements; that would result in a protocol that fits within existing fabrication protocols that have been developed for the microelectronic circuitry that is made at the facility.
An alternative method for fabricating an integrated ridge waveguide and detector is illustrated in
In addition, instead of using epitaxial growth to fill the trenches one could use other CVD processes.
In the embodiment described above, the ridge is formed by etching the material away to form two trenches which define the ridge. Those trenches can be narrow trenches, as described above, or they can be very wide trenches. Indeed, one approach is to etch away everything except the ridge thereby causing the trenches to extend to the edges of the chip or wafer (also referred to as the Arkadii method).
Also in the embodiments described above, the trench was filled with silicon. But it can be filled with any appropriate material that has lower refractive index, e.g. Si (doped or undoped) or dielectrics like SiN or SiOx, diamond-like carbon, spin-on-glass (SOG), where those filling materials are preferably CMOS compatible. Also, note that a SiN top layer (i.e., deposited before filling the trenches) can be used as a CMP stop layer. However, it may not be needed if the filling material is something other than silicon.
The field area (i.e., the area without any waveguide or trench) can be used to manufacture microelectronic circuitry (after planarization that follows the trench fill). If that microelectronic circuitry is fabricated away from the trenches and waveguides, one can probably use the original silicon that was protected by the SiN during the planarization process. This high quality silicon could also be grown on the whole wafer including the trench and waveguide area, i.e., covering them or it could also be applied using a layer transfer method. For lower quality electrical devices (e.g. having slower speed and/or more leakage current) an additional top layer or keep-out areas for avoiding the waveguide and trenches may not be required.
With respect to electrical connections, it might be desirable to dope the silicon and parts of the SiGe that forms the waveguide. Implant is a particularly good method to dope the bottom and top and active areas of the detector, to form conductive parts in the trench fill material, etc. In-situ doping of material during its deposition is also possible, but has to be done in such way as to keep the losses in the optical waveguide low. Connections using metal and/or silicide or very highly doped regions should be kept off to the side of the waveguide, as their absorption coefficients are typically too large for them to be located on top of the waveguide ridge, unless a thick top layer is used. In general, details of the geometry of the connections depend on the trade-off between resistance, capacity (i.e., detector speed), and non-productive optical losses incurred within the detector.
It should be noted that planarity is important for the subsequent manufacture of devices (to avoid lithography problems, resist problems, etc.). In addition, one might be able to manufacture some microelectronic circuitry before the manufacture of the waveguide.
Other embodiments are within the following claims.
This application claims the benefit of U.S. Provisional Application No. 60/713,497, filed Sep. 1, 2005, which is incorporated herein by reference.
Number | Name | Date | Kind |
---|---|---|---|
3217856 | Miles | Nov 1965 | A |
4006432 | Streifer et al. | Feb 1977 | A |
4100562 | Sugawara et al. | Jul 1978 | A |
4368481 | Ohashi et al. | Jan 1983 | A |
4438447 | Copeland, III et al. | Mar 1984 | A |
4486765 | Capasso | Dec 1984 | A |
4720468 | Menigaux et al. | Jan 1988 | A |
4734910 | Izadpanah | Mar 1988 | A |
4759023 | Yamaguchi et al. | Jul 1988 | A |
4835575 | Plihal | May 1989 | A |
4959540 | Fan et al. | Sep 1990 | A |
4966430 | Weidel et al. | Oct 1990 | A |
4980568 | Merrick et al. | Dec 1990 | A |
5098861 | Blackstone | Mar 1992 | A |
5122852 | Chang et al. | Jun 1992 | A |
5190883 | Menigaux et al. | Mar 1993 | A |
5193131 | Bruno et al. | Mar 1993 | A |
5195161 | Adar et al. | Mar 1993 | A |
5280189 | Schuppert et al. | Jan 1994 | A |
5291010 | Tsuji et al. | Mar 1994 | A |
5298787 | Bozler et al. | Mar 1994 | A |
5345557 | Wendt et al. | Sep 1994 | A |
5357122 | Okubora et al. | Oct 1994 | A |
5382810 | Isaksson et al. | Jan 1995 | A |
5463229 | Takase et al. | Oct 1995 | A |
5481515 | Kando et al. | Jan 1996 | A |
5485021 | Abe et al. | Jan 1996 | A |
5523557 | Bruno et al. | Jun 1996 | A |
5536974 | Nishiguchi et al. | Jul 1996 | A |
5548433 | Smith et al. | Aug 1996 | A |
5569934 | Fujii et al. | Oct 1996 | A |
5604361 | Isaksson et al. | Feb 1997 | A |
5633527 | Lear | May 1997 | A |
5663592 | Miyazawa et al. | Sep 1997 | A |
5682455 | Kovacic et al. | Oct 1997 | A |
5728605 | Mizutani et al. | Mar 1998 | A |
5767508 | Masui et al. | Jun 1998 | A |
5793913 | Kovacic | Aug 1998 | A |
5801872 | Tsuji et al. | Sep 1998 | A |
5812708 | Rao | Sep 1998 | A |
5818096 | Ishibashi et al. | Oct 1998 | A |
5825051 | Bauer et al. | Oct 1998 | A |
5828679 | Fisher et al. | Oct 1998 | A |
5854804 | Winer et al. | Dec 1998 | A |
5889903 | Rao | Mar 1999 | A |
5942789 | Morikawa et al. | Aug 1999 | A |
5945720 | Itatani et al. | Aug 1999 | A |
5946438 | Minot et al. | Aug 1999 | A |
5970081 | Hirayama et al. | Oct 1999 | A |
5987196 | Noble | Nov 1999 | A |
6011296 | Hassard et al. | Jan 2000 | A |
6031243 | Taylor | Feb 2000 | A |
6043515 | Kamiguchi et al. | Mar 2000 | A |
6066860 | Katayama et al. | May 2000 | A |
6075253 | Sugiyama et al. | Jun 2000 | A |
6093939 | Artigue et al. | Jul 2000 | A |
6125217 | Paniccia et al. | Sep 2000 | A |
6202165 | Pine | Mar 2001 | B1 |
6232142 | Yasukawa et al. | May 2001 | B1 |
6288410 | Miyazawa et al. | Sep 2001 | B1 |
6310372 | Katayama et al. | Oct 2001 | B1 |
6318901 | Heremans et al. | Nov 2001 | B1 |
6320204 | Hirabayashi et al. | Nov 2001 | B1 |
6343171 | Yoshimura et al. | Jan 2002 | B1 |
6346717 | Kawata | Feb 2002 | B1 |
6392342 | Parikka | May 2002 | B1 |
6393183 | Worley | May 2002 | B1 |
6403395 | Hirabayashi et al. | Jun 2002 | B2 |
6407438 | Severn | Jun 2002 | B1 |
6426522 | Kean et al. | Jul 2002 | B1 |
6449411 | Nakamura et al. | Sep 2002 | B1 |
6528779 | Franz et al. | Mar 2003 | B1 |
6645829 | Fitzergald | Nov 2003 | B2 |
6658173 | Delwala | Dec 2003 | B2 |
6677655 | Fitzergald | Jan 2004 | B2 |
6680495 | Fitzergald | Jan 2004 | B2 |
6690078 | Irissou et al. | Feb 2004 | B1 |
6753214 | Brinkmann et al. | Jun 2004 | B1 |
6770134 | Maydan et al. | Aug 2004 | B2 |
7110629 | Bjorkman et al. | Sep 2006 | B2 |
20020079427 | Xu et al. | Jun 2002 | A1 |
20020146865 | Hoel | Oct 2002 | A1 |
20020174826 | Maydan et al. | Nov 2002 | A1 |
20020174827 | Samoilov et al. | Nov 2002 | A1 |
20020181825 | Johnson et al. | Dec 2002 | A1 |
20030015720 | Lian et al. | Jan 2003 | A1 |
20030025118 | Yamazaki et al. | Feb 2003 | A1 |
20030026523 | Chua et al. | Feb 2003 | A1 |
20030052082 | Khan et al. | Mar 2003 | A1 |
20030052088 | Khan et al. | Mar 2003 | A1 |
20030072548 | Bhardwaj et al. | Apr 2003 | A1 |
20030110808 | M'Saad et al. | Jun 2003 | A1 |
20030113085 | M'Saad | Jun 2003 | A1 |
20030114006 | White | Jun 2003 | A1 |
20030176075 | Khan et al. | Sep 2003 | A1 |
20040012037 | Venkatesan et al. | Jan 2004 | A1 |
20040012041 | West et al. | Jan 2004 | A1 |
20040012401 | King et al. | Jan 2004 | A1 |
20040013338 | Bjorkman et al. | Jan 2004 | A1 |
20040037512 | Cho et al. | Apr 2004 | A1 |
20040155751 | Benzel et al. | Aug 2004 | A1 |
20040266145 | Morse | Dec 2004 | A1 |
20050053347 | West et al. | Mar 2005 | A1 |
20050072979 | West et al. | Apr 2005 | A1 |
20050212068 | Leon et al. | Sep 2005 | A1 |
20050214964 | West et al. | Sep 2005 | A1 |
20060039666 | Knights et al. | Feb 2006 | A1 |
Number | Date | Country |
---|---|---|
0661561 | Jul 1995 | EP |
WO-02095092 | Nov 2002 | WO |
WO-03036369 | May 2003 | WO |
WO-03060599 | Jul 2003 | WO |
Number | Date | Country | |
---|---|---|---|
20070053643 A1 | Mar 2007 | US |
Number | Date | Country | |
---|---|---|---|
60713497 | Sep 2005 | US |