High Speed Double Polycide Bipolar/Cmos Integrated Circuit Process

This invention relates to a mixed bipolar-MOS process for fabricating both bipolar and field effect transistors on a common integrated circuit chip.

Chip makers have long known that mixed bipolar-MOS processes, referred to in the art as BiMOS and BiCMOS, can produce faster and denser integrated circuits than can either technology alone. However, it has taken a long time to master the complex processing that these integrated circuits required. The advantages of mixed bipolar-MOS have been apparent since the early 1970's. At that time, however, bipolar and MOS structures required entirely different techniques and tools to fabricate. To design a mixed process device meant considerable expense. The present invention is a process technology that makes it very easy to fabricate mixed bipolar MOS devices in a common integrated circuit, obtaining the advantages of both bipolar and MOS processes and devices without significant additional expense.

Several integrated circuit manufacturers have been investigating methods of merging or mixing bipolar and MOS structures on the same chip and have disclosed or marketed BiMOS or BiCMOS devices. A summary of developments in this area of technology is presented in B. C. Cole, "Mixed-Process Chips Are About To Hit The Big Time," Electronics, Mar. 3, 1986, pages 27-31. This article discloses the wide range of processes that designers are trying in attempting to make satisfactory hybrid bipolar-MOS processes. Most begin with essentially a MOS process and add steps as needed to include a bipolar subprocess in the overall fabrication process. A few take the opposite approach.

A mixed bipolar-CMOS process is described, and the bipolar process is illustrated, in "A Bipolar Process That's Repelling CMOS," Electronics, Dec. 23, 1985, pages 45-47. One example of a mixed bipolar-CMOS circuit is disclosed in J. Miyamoto, et al., "A 28ns CMOS SRAM with Bipolar Sense Amplifiers," 1984 IEEE National Solid-State Circuits Conference Digest of Technical Papers, pages 224-225, 344 (1984). Mixed processes are briefly discussed, and additional references cited, in Brown, D. M., et al., "Trends in Advanced Process Technology--Submicrometer CMOS Device Design and Process Requirements," Proc. IEEE, Vol. 74, No. 12, pp. 1678-1702, Dec. 1986, at p. 1694.

The principal drawback, however, of all of these technologies has been an inability to produce very fast bipolar transistors in a mixed bipolar-MOS process. The mixed-process chips reported in the above-cited Cole article generally exhibit bipolar speeds in the range of 2-3 GHz. Only one process, of Hughes Aircraft Co.'s Semiconductor Division, is reported to produce NPN transistors with cutoff frequencies in excess of 5 GHz and PNP devices of over 2.5 GHz. The Hughes process is described as somewhat more complex and expensive than most other mixed processes, with 20 masking steps.

Another mixed process is disclosed in Ogive, K., et al., "13-ns, 500-mW, 64-kbit ECL RAM using HI-BICMOS Technology," IEEE Journal of Solid-State Circuits, Vol. SC-21, No. 5, pp. 681-685, Oct. 1986, which reports a cutoff frequency of 4 GHz.

These bipolar performance characteristics would be considered rather slow in terms of the current state of bipolar-only art. Most bipolar processes in current commercial use produce bipolar speeds in the range of 5-7 GHz. Recently, a new generation of bipolar devices has emerged which exhibit speeds in the range of 9-17 GHz. All of these devices are produced by processes that are substantially more complex and critical than those used to produce mixed bipolar-MOS devices or bipolar-only devices in the 5-7 GHz range.

Advanced bipolar technologies generally use a self-aligned process technology to attain very high speed device performance. Examples of these processes are disclosed in U.S. Pat. No. 4,381,953 to Ho et al., U.S. Pat. No. 4,483,726 to Isaac et al.; T. Ning et al. "Bipolar Trends," Proc. IEEE, Vol. 74, No. 12, pp. 1669-1677, Dec. 1986; and in S. Konaka, "A 30 ps Si Bipolar IC Using Super Self-Aligned Process Technology," Extended Abstracts 16th Conference on Solid State Devices and Materials, pages 209-212 (1984). A circuit fabricated using the Konaka et al. process is disclosed in "A 9-GHz Frequency Divider Using Si Bipolar Super Self-Aligned Process Technology," IEEE Electron Device Letters, Vol. EDL-6, No. 4, pages 181-183 (Apr. 1985).

A metal-oxide-semiconductor field-effect transistor (MOSFET) of conventional form is shown in FIG. 1. The conventional p- channel MOSFET 20 comprises a substrate 22 of p+ silicon having source and drain regions 24 and 26 of n+ material formed therein immediately beneath the upper surface 28 of the substrate. The source and drain regions are bounded by field oxide layer 29, and are separated by a channel region 30, in which the p+ material of the substrate extends to the upper surface 28. A PN junction is formed between the P+ silicon and each of the N+ regions. The spacing between the source and drain regions defines a channel width of region 30. A thin layer 32 of thermally-formed or chemical vapor deposited (CVD) silicon dioxide overlies the channel region and the immediately adjacent margins of the source and drain regions 24 and 26, and a gate 34 of polysilicon overlies the layer 32. A layer 36 of CVD silicon dioxide is applied over the source and drain regions and over the gate 34. Holes 38 and 40 are formed in the layer 36 to expose the source and drain regions 24 and 26, respectively. A third hole (not shown) is formed out of the plane of the drawing so as to expose the gate 34. Metal, e.g., aluminum, is deposited over the layer 36 and enters the holes in the layer 36. The metal is selectively removed so as to form distinct source and drain contact electrodes 42 and 44 and a gate contact electrode (not shown). When the electrodes 42 and 44 are connected to ground and a positive voltage respectively, and the gate is at approximately ground or negative potential (depending on the threshold voltage V.sub.t, which may be above or below "ground"), the source and drain regions are electrically isolated by the PN junction between the drain region 26 and the channel region 30. When the voltage at the gate is increased, an inversion layer is formed in the channel region immediately beneath the gate. Electrons can flow from the source region 24 to the drain region 26 through this inversion layer.

In the conventional MOSFET shown in FIG. 1, the holes 38 and 40 are formed by a photoprocessing operation that involves aligning a mask relative to the gate 34, and the electrodes 42 and 44 make contact directly to the source and drain regions, respectively. To allow for errors in locating and aligning the mask, the source and drain regions must be made sufficiently large to ensure that the electrodes 42 and 44 will make contact to the source and drain regions. The capacitance of the PN junctions between the bulk semiconductor material and the source and drain regions, respectively, depends on the area of the interfaces between the bulk material and the source and drain regions, respectively. This area in turn depends on the area of the surface 28 exposed on each side of the gate oxide 32. To obtain high speed operation, it is necessary to minimize the capacitance of the PN junctions. Therefore, it is necessary to minimize the exposed area of the surface 28. For a given channel width, this implies that the distance between the gate oxide and the field oxide must be minimized.

C. S. Oh and C. K. Kim "A New MOSFET Structure with Self-Aligned Polysilicon Source and Drain Electrode," IEEE Electron Device Letters, Vol. EDL-5, 1984, pages 400-402, disclose a MOSFET process in which the source and drain contacts are self-aligned with the source and drain regions. The source and drain electrodes are positioned over the field oxide. Therefore, the distance between the gate oxide and the field oxide can be made much smaller than in the case of the FIG. 1 device, with the result that higher speed operation is possible.

Another MOS process having similar objects is disclosed in Huang, T. Y. et al, "A MOS Transistor with Self-Aligned Polysilicon Source-Drain," IEEE Electron Device Letters, Vol. EDL-7, No. 5, May 1986, pp. 314-316. This process inverts many of the steps of Oh and Kim. The polysilicon and overlying insulative layers are first applied. Then a reverse-gate mask and etch procedure is used to open the active channel region. A layer of gate oxide is thermally grown and a second polysilicon layer is deposited to form the gate contact.

In order to minimize the channel length (and therefore maximize the operating speed), the regions of the PN junctions that are immediately adjacent the channel region must be lightly doped. Also, to maximize the source/drain breakdown voltage of a MOSFET, it is desirable that the drain region be lightly doped immediately adjacent the channel region. The source region and the major part of the drain region should be more heavily doped to minimize the source/drain resistance when the MOSFET is in its conductive state. Thus, the drain region is preferably formed in two zones: a lightly doped zone that extends at least partially under the gate, and a more heavily doped zone that connects the lightly doped zone to the drain electrode.

A known method of providing different doping levels within the drain region of a MOSFET is illustrated in FIG. 2. As shown in FIG. 2(a), the gate structure 32/34 is formed on the substrate 22. Then, a first ion implantation operation is carried out, providing a relatively low concentration of charge carriers in regions 40 on each side of the gate structure. A silicon dioxide layer 44 of uniform thickness is deposited over the gate structure and the adjacent areas of the upper surface of the substrate (FIG. 2(b)). The silicon dioxide of the layer 44 is selectively removed by reactive ion etching to a depth equal to the thickness of the layer 44, so as to expose substrate surface 28 but leave walls 46 extending along the sides of the gate structure 32/34 (FIG. 2(c)). A second ion implantation is then carried out, providing a higher concentration of charge carriers, sidewalls 46 acting as an implantation mask. Therefore, the source and drain regions each have two zones 48 and 50, of high and low conductivity, respectively. For ease of processing, the same operations are carried out on both sides of the gate.

The method described by Oh and Kim does not lend itself to the establishment of different doping levels within the drain region. Nor does the process of Huang et al. Processes disclosed in D. M. Brown et al., supra, p. 1682, FIG. 7, and p. 1690, FIG. 22, can produce such different doping levels, but employ metal source and drain contacts to the silicon substrate and added process steps.

In S. S. Wong et al., "Elevated Source/Drain MOSFET," IEDM Tech. Dig., 1984, pages 634-637, there is described a method for providing a high conductivity layer over the source and drain regions of a MOSFET by forming an epitaxial layer of silicon over the source and drain regions, but not over the gate structure, and implanting ions into the epitaxial layer and the gate structure. However, this method is subject to a disadvantage in that deposition of an epitaxial layer is normally carried out at high temperature, and high temperature processing is considered undesirable in fabrication of a MOSFET having a lightly-doped drain.

S. S. Wong, "Contact Technologies for Submicron CMOS," Cornell Program on Submicrometer Structures, 1985, discusses the use of a metal silicide contact layer over the source and drain regions to minimize the source/drain resistance. The silicide layer is formed by first depositing a layer of polysilicon over the source and drain regions and then depositing of a layer of refractory metal over the polysilicon. The device is then annealed, and the metal silicide layer is formed.

It is also known to use an oxide sidewall structure in making bipolar transistors, to define a spacing between polysilicon emitter and base contacts. This is shown by S. F. Chu et al. in "A Self-Aligned Bipolar Transistor," VLSI Science and Technology/1982 Proceedings Vol. 82-7, pp. 306-314. The process shown also uses successive implant and diffusion steps to form the intrinsic base and emitter from the layer of polysilicon forming the emitter contact. This procedure is detailed in another paper in the same Proceedings, presented by F. Barson et al., entitled "Shallow Bipolar Transistor Profiles by Diffusion from Implanted Polysilicon," pp. 282-287. It produces a very high speed bipolar transistor, with a cutoff frequency demonstrated to be as high as 15 GHz, but no way is suggested that makes this high speed available together with the high density and low power consumption of CMOS.

A need remains for a process that overcomes the drawbacks of the foregoing MOS and bipolar processes. Additionally, it would be desirable to have a process that can readily be used to produce mixed bipolar-MOS circuit.

In accordance with the invention, a bipolar transistor and a metal-oxide semiconductor field-effect transistor (MOSFET) are fabricated on a common substrate of semiconductor material having a main face and containing impurity ions that provide charge carriers of a first polarity. The substrate can be bulk crystalline silicon material but is preferably provided with a uniformly-doped epitaxial layer in which the bipolar, PMOS and NMOS devices can be formed in suitably-doped active device regions. The term "substrate" is intended to encompass both kinds of substrate, except where one kind is specified, as in the preferred embodiment. The dopant type or polarity of the active device regions depends on the type of device. Preferably, the epitaxial layer is N-type, for NPN bipolar and PMOS devices, with a P-well formed for the NMOS devices. The MOS and bipolar devices are simultaneously formed in parallel steps. For clarity, however, the process of forming the MOS devices is first summarized.

An elongated insulated gate structure is formed on the main face of the substrate in each of the MOS regions. First and second areas of the main face are exposed along first and second opposite sides respectively of the gate structure. Impurity atoms that provide charge carriers of a second polarity, opposite the first polarity, are introduced into the substrate by way of at least the first area of the main face, to achieve a predetermined concentration of charge carriers of the second polarity in a drain region of the substrate that is subjacent the first area of the main face. This is preferably done by ion implantation, controlled to produce lightly-doped drain and source regions. The gate structure, preferably of polysilicon, is opaque to the impurity atoms, and therefore the drain region does not extend fully under the gate structure. Nevertheless, the drain region has a channel-side zone that lies under the margin of the gate at the first side thereof. A sidewall of a selected material, such as silicon dioxide, is formed along the first side of the structure, whereby a strip of the first area of the main face is covered by the sidewall and other parts of the first area remain exposed adjacent the sidewall. Atoms of a selected impurity, to which the gate structure and the sidewall are opaque, are introduced into the substrate by way of the portion of the first area that is exposed adjacent the sidewall. The atoms of the selected impurity provide charge carriers of the second polarity. A zone that extends partially within the drain region, but does not extend fully under the sidewall, and in which the conductivity is provided by charge carriers of the second polarity and is higher than that in the channel-side zone of the drain region, is created. Self-aligned source and drain contacts are then formed, preferably by the following procedure. A layer of conductive material, preferably polysilicon, is deposited over the portion of the first area of the main face that is exposed adjacent the sidewall. This layer extends up the sidewall and over the gate structure. A layer of a selected material, such as a planarizable synthetic polymer, is formed over the layer of conductive material to a substantially uniform height over the main face. The height of the free surface of the layer of selected material is at least as great as the maximum height of the layer of conductive material over the gate structure. The selected material is uniformly thinned to expose the layer of conductive material over the gate only, which coincides with its region of maximum height. The conductive material at a height that is at least as great as that of the gate structure is then removed. Some degree of additional removal of conductive material may be used to optimize this structure. A discrete drain contact is thereby formed. The second side and an adjacent strip of the second area can be likewise covered and doped, and a discrete source contact formed in the foregoing procedure. The source and drain contacts are preferably provided by a polysilicon layer, implanted with appropriate impurity atoms for each type of MOS device and subjected to a drive-in treatment to form increased doping concentration regions beneath the contacts.

The bipolar devices are formed in parallel with the MOS devices, in accordance with essentially the same process steps but with patterning of the steps in a way suited to forming a bipolar structure. A base contact structure is formed when the gate structure is formed, in two portions spaced apart about a third exposed area of the substrate surface, preferably in the shape of a rectilinear annulus. Preferably, the substrate in the third area is lightly doped by ion implantation with atoms of the second impurity when the drain of the like-dopant type of MOS device is doped (for an NPN transistor, this is when the PMOS source/drain are doped). The same doping concentrations and profiles can used for both devices, for simplicity of fabrication, but different implants can be used to optimize respective device characteristics. Then, the sidewalls are formed on the sides of the base contact portions when they are formed on the MOS gate structures, with the effect of reducing the third area and shielding a marginal strip thereof. This is followed by a second doping step, which coincides with the second doping of the first and second areas of the MOS devices. An emitter and emitter contact are also formed. These steps are preferably combined in a second polysilicon deposition step, the same as that used to form the source/drain contacts, followed by implantation and drive-in steps. Implantation of the polysilicon is preferably done in two steps: first, of the second polarity impurity type, to dope the base to a higher concentration but lesser width and no greater depth than those of the lightly doped base region formed in the ion implantation step; and secondly, of the first polarity type to form an emitter of lesser width and depth of the base. This procedure forms a lightly-doped base connection within the substrate, similar to the above-mentioned lightly-doped drain but with different effect in the bipolar transistor. The lightly-doped base connection provides a relatively low breakdown voltage which has little variance from one bipolar device to the next.

The foregoing and other objects, features and advantages of the invention will become more readily apparent from the following detailed description of a preferred embodiment which proceeds with reference to the accompanying drawings.

FIG. 1 is a schematic sectional view of a conventional MOSFET structure.

FIGS. 2a-d is a series of views illustrating formation of an oxide box about the gate of a MOSFET.

FIGS. 3-12 are a series of views illustrating fabrication of a mixed bipolar-CMOS integrated circuit by a process embodying the present invention.

FIGS. 13-16 are a series of views of a portion of the circuit at the intermediate stage of FIG. 7, illustrating a subprocess for diffusion of both base and emitter of the bipolar transistor.

FIG. 17 is an enlarged view of a portion of the bipolar transistor formed by the process of FIGS. 3-16.

TABLE 1, appended to this specification, is a process-flow listing of the photolithographic and major process steps, and preferred process techniques and conditions, for an example of the method of FIGS. 3-12.

In the different figures, like reference numerals denote like elements. For a better understanding, reference will also be made, by way of example, to the photolithography and process steps in TABLE 1.

A wafer 122 of P-type, (100) oriented silicon of 40-100 ohm-cm resistivity is processed in accordance with process steps 1-34 of TABLE 1 to produce the substrate as shown in FIG. 3. These process steps are discussed in conjunction with describing the masking steps used to produce the intermediate mixed bipolar-CMOS structure of FIG. 3. As indicated in FIGS. 3 and 12, an NPN bipolar transistor is formed in region 114, a resistor in region 116, a PMOSFET in region 118, and an NMOSFET in region 120.

A layer of oxide (not shown) is thermally grown on the wafer, patterned and etched to define a mask for a buried layer ion implant. This procedure produces N+ (10.sup.16 concentration) buried layers 124 and 125 beneath the intended bipolar and PMOS device regions. This is followed by an annealing step.

Next, in step 7, the oxide layer is removed and an epitaxial layer 123 of 1.3 to 1.6 um. thickness (nominally 1.5 um.) is deposited on the wafer surface. This layer is uniformly N-type, in a concentration sufficient to produce a resistivity of about 0.2 to 0.3 ohm-cm. at a thickness of 1.3 um. This is followed by thermal oxidation and plasma oxide deposition steps to form another oxide masking layer (not shown). This layer serves as a P-well implant mask, which is patterned and etched in step 10 to expose the epitaxial silicon surface in NMOS region 120. In steps 11 and 12, P-type ions are implanted and diffused, by a heat treatment process, to form P-type region 131 extending through the epitaxial layer into the bulk of the substrate 122. This provides a P-well for the NMOS device to be formed in later steps in region 120. A PMOS field effect transistor (region 118) will be formed, in succeeding steps, on the substrate surface over buried layer 124. An NPN bipolar transistor (region 114) will be formed in subsequent steps over buried layer 125.

Following P-well drive-in, in steps 13 and 14, a second masking step is performed to pattern and etch silicon islands upon which the active devices will be formed, surrounded by recessed areas of the substrate. The recesses extend through the epitaxial layer 123 to or into the bulk substrate material 122. In steps 17 and 18, these recessed areas are filled with silicon dioxide to electrically isolate the active silicon islands. First, however, a P-type field implant step is performed to define channel stops 126 within the bulk substrate material 123 between the active device regions (step 15). After stripping the mask oxide (step 16), the entire exposed silicon surface is thermally oxidized to a depth of about 1,000 angstroms. This is followed by a low temperature chemical vapor deposition step to deposit a layer of silicon dioxide over the entire substrate surface. The thermal oxidation step produces an oxide layer 127a in the recessed regions. The deposition step fills the remaining space of these recesses with silicon dioxide 127b. By reactive ion etching (RIE), the oxide is planarized and reduced to a surface elevation 128 where epitaxial silicon in the active regions is exposed (step 19). This procedure produces planarized, recessed, bird's-beak-free field oxide regions 129.

Next are thermal oxidation, masking, P-type implant and drive-in steps (steps 20-23) for implanting active P-type region 131. An NMOS field effect transistor (region 120) will be formed in subsequent steps on the substrate over this region. An oxidation step produces a 200-300 angstrom thick gate oxide layer 132 on the substrate surface in each of the PMOS and NMOS regions 118, 120 (step 24). Separate threshold voltage masking and implant steps are performed in regions 118 and 120 for determining the gate threshold voltages for each type of MOS device (step 25). Then, a deep collector masking and implant steps are performed to implant collector contact region 130 to an increased, N+ concentration (steps 26 and 27). Next are clear oxide photolithography and etching steps (step 28), which clears the portions of oxide layer 132 in region 114 to selectively expose silicon surface 133 in the bipolar device region 114 for performing succeeding steps.

Next, a first polysilicon layer 134 is chemical vapor deposited to a thickness of 4000-5000 angstroms (step 29). This layer is then masked and implanted, first, to implant the polysilicon layer to N+ concentrations over the PMOS and NMOS active device regions and over the collector contact region 130 (steps 30 and 31). The N+ implant mask is stripped and the substrate is again masked and implanted to dope the polysilicon layer overlying the NPN bipolar active region to P+ concentrations (steps 32 and 33). The photoresist is then stripped.

Optionally, a layer of suitable metal is deposited on the polysilicon and reacted and removed to form a polysilicide layer 136. This is done by depositing a layer of refractory metal such as titanium over the polysilicon layer and then annealing the refractory metal to form a low-resistivity silicide (TiSi.sub.2), as shown in S. S. Wong, "Contact Technologies for Submicron CMOS", Cornell Program on Submicrometer Structures, 1985. In combination with ion implantation, siliciding the polysilicon can reduce its resistivity by two orders of magnitude.

Next, in step 34, a layer 138 of silicon dioxide is chemical vapor deposited to a thickness of about 5,000 angstroms.

Referring next to FIG. 4, gate/base/collector masking and etching steps 35 and 36 are performed to define base contacts 140, collector contact 142 and resistor 144, and MOSFET gate electrodes 146, 148. Masking and boron implant steps (steps 37 and 38) provide P-type base region 150 for the bipolar transistor in region 114 and lightly-doped P-type source and drain regions 152, 154 for the PMOSFET in region 118. These steps are further discussed below in connection with FIGS. 13-17. Mask and implant steps 39 and 40 similarly provide N-type source and drain regions 156, 158 for the NMOSFET in region 120.

FIGS. 5 and 6 illustrates process steps 41-44. A 0.5 um to 1.0 um layer 160 of silicon dioxide is deposited and annealed. This layer is then reactive ion etched to remove preferentially those horizontal portions lying parallel to the silicon surface while leaving vertical portions or sidewalls 162. An oxide box structure is thus formed by oxide layer 138 and sidewalls 162. The box structure encloses each of the underlying polysilicon electrodes, specifically base contact portions 140 and gate electrodes 146, 148. This step also exposes silicon surface 164 on both sides of the gate electrodes and surface 133 between the base contact portions 140.

Proceeding to FIG. 7 and process steps 45-51, a second layer of polysilicon is deposited over the substrate, the oxide box structure 138, 162 and the field oxide regions 129. This layer is silicided and selectively doped in successive mask and implant steps, and patterned and etched, to form an N+ emitter contact 166, a P+ source/drain contact structure 168 and an N+ source/drain contact structure 170. These steps are further discussed with reference to FIGS. 13-17, below.

FIGS. 8-10 illustrate process steps 52-54, by which the source/drain contact structures are separated to form separate self-aligned source and drain contacts. A layer 172 of synthetic polymer material is applied to the FIG. 7 structure. The layer of polymer material is applied by a method known as polymer planarization, which is described in A. C. Adams and C. D. Capio, "Planarization of Phosphorus Doped Silicon Dioxide", J. Electrochem. Soc., vol. 128(2), 1981, page 423; L. K. White, "Planarization Phenomena in Multilayer Resist Processing", J. Vac. Sci. Technol., vol. B1(4), 1983, page 1235; and L. K. White, "Planarization Properties of Resist and Polyimide Coatings", J. Electrochem. Soc., vol. 130(7), 1983, page 1543. The free surface of the resultant layer 172 is planar and is parallel to the upper surface of the wafer. The initial maximum depth 174 of the layer 172 is sufficient that the polysilicon contact structures 166, 168, 170 are fully covered.

The polymer material is then exposed to an etchant which etches the polymer anisotropically, that is, primarily in a direction perpendicular to the substrate surface 133. This etching continues until the contact structures 168, 170 are exposed over the oxide boxes 138, 162 at surface 174. An etchant may alternatively be chosen which etches both the polymer and polysilicon at substantially the same rate. This step is preferably followed by a masking step to protect emitter contact 166 with a photoresist layer 176, as shown in FIG. 9. A second etching operation which demonstrates a controlled degree of anisotropy is then carried out as shown in FIG. 10. This step employs an etchant that attacks polysilicon but does not attack either the polymer material or silicon dioxide. The second etching operation is carried out in the direction perpendicular to the upper surface 133 of the substrate until the free surface of the polysilicon is below the top of the oxide boxes 138, 162 surrounding gate structures 146, 148. In this manner, the polysilicon source/drain contact structures 168, 170 are each divided into two portions. This step forms separate source and drain contacts 168A and 168B overlying the source and drain regions 152 and 154, respectively, and source and drain contacts 170A and 170B overlying regions 156 and 158, respectively. The polymer material 172 is then removed.

Next, in FIG. 11, a layer 179 of silicon dioxide is deposited over the entire substrate, including over contacts 168A, 168B and 170A, 170B and over the oxide box structures 138, 162 (step 55). The substrate is then annealled in step 56, to diffuse dopant ions from the polysilicon layers into the underlying substrate silicon. This step produces an N+ emitter 178 and P+ base contact regions 180 in the bipolar transistor (region 114), P+ source and drain contact regions 182, 184 in the PMOSFET (region 118), and N+ source and drain contact regions 186, 188 in the NMOSFET (region 120). The base and gate contact diffusions lower base resistance in the bipolar and MOS transistors. Diffused emitter junctions formed in this manner are very shallow, typically less than 0.1 um for arsenic and less than 0.2 um for boron, which enables very high speed bipolar operation. Diffusion of source and drain contact regions is similarly limited so as to produce a gradation of dopant concentration that decreases proceeding downward and laterally from regions 182, 184 and 186, 188 in regions 152, 154 and 156, 158, respectively. This doping profile yields a narrow, high-speed FET gate with enhanced breakdown voltage while keeping source and drain resistance low.

In FIG. 12, holes are then formed in the layer 179 in conventional manner to expose the base (B), emitter (E) and collector (C) contacts of the bipolar transistor, a resistor contact (R), and source (S) and drain (D) contacts, respectively, of the FET devices. A third hole (not shown) is formed to expose the gate of each of the FET devices. Metal (not shown) is deposited over the layer 179 in conventional fashion and enters the holes in the oxide layer. The metal is selectively removed to form discrete base, emitter and collector electrodes, a connection to resistor 116, source, and drain gate electrodes, and metal interconnections as desired by the designer.

It will be appreciated that the present invention is not restricted to the particular method and structure described above. Instead of an epitaxial substrate with N-type buried layers, with some sacrifice in device speed, a p+ substrate with N-wells for the NMOS and bipolar transistors can be used. One or both silicide steps can be omitted, or applied selectively (e.g., omitted from resistor 144) by appropriate masking, to vary sheet resistances. In another variation, collector contact 142 and resistor 144 can be formed as part of the second, rather than first, polysilicon layer.

In yet another, preferable variation, the bipolar transistor base region 150 is not formed in toto by direct implantation in steps 37 and 38. Although it will yield operative bipolar transistors, it is difficult to control final base depth and width. Preferably, the region 150 is formed as lightly-doped connection between an intrinsic base 177 (FIG. 14) to be formed in step 46 and the base contact diffusions 180, which are more highly doped. For simplicity, all of the regions 150, 152, 154 can be simultaneously implanted to a concentration of about 10.sup.13 and a depth not more than the maximum desired depth of the intrinsic base 177, preferably about 2000 angstroms. If it is desired to vary the implant concentrations between the bipolar and PMOS devices, this can be done by using two masking and P-type implant steps, with minimum requirements for critical alignment. Steps 46 and 47 are performed after steps 48 and 49, with an intervening base drive-in step. This procedure is designed to diffuse both the base 177 and emitter 178 from successive implants of boron and arsenic into emitter contact 166. It is carried out without materially increasing complexity of the mixed bipolar-CMOS process of the invention.

In this technique, as shown in FIGS. 13 and 14, boron ions are implanted in the undoped-polysilicon emitter 166, and diffused to form a shallow intrinsic base 177. Then, as shown in FIGS. 15 and 16, arsenic ions are implanted to the same emitter polysilicon, and diffused to form a shallow emitter 178. The result of implanting a lightly-doped region 150, followed by the foregoing successive implants, is shown in FIG. 17. The resultant bipolar structure has very good characteristics, as described below.

A self-aligned TiSi.sub.2 layer is formed on top of the emitter polysilicon and this emitter polycide reduces the sheet resistance of the emitter polysilicon (As+ implanted and 2500 angstrom thick) from 300 ohm/sq. to 3-10 ohm/sq. with an arsenic implant dose of 3E15-5E15 ions/cm.sup.2. The resultant bipolar transistor is compared with two alternative transistors similarly fabricated, but with the emitter and base both directly implanted and the emitter diffused and base implanted. Table 2 compares the junction depths of each transistor and Table 3 compares their respective device parameters.

In an NPN transistor having both the base 177 and emitter 178 diffused from the polysilicon emitter contact, the breakdown voltage characteristic BVceo with implanted region 150 also compares favorably with the same device structure without such region. At Irev=1E-06, the former shows a very tight distribution with an average BVceo of 6.016 and standard deviation of 0.525, compared to a widely dispersed distribution over a range from 0.4 to 4.8 for the latter. The h.sub.FE characteristic is similarly distributed; at Ic=1E-04 and Vcb=2, an average of 105.3 and standard deviation of 12.06, with a range of 85 to 125 for the former, compared to a range of about 50 to 130 with most occurrences clustered between 60 and 90 for the latter.

Having illustrated and described the principles of our invention in a preferred embodiment thereof, it should be readily apparent to those skilled in the art that the invention can be modified in arrangement and detail without departing from such principles. We claim all modifications coming within the spirit and scope of the accompanying claims.