Trench Schottky barrier diode with differential oxide thickness

[0001] This application is a continuation-in-part of and claims priority to U.S. application Ser. No. 10/193,783, filed Jul. 11, 2002, now U.S. Pat. No. 6,855,593 entitled, “Trench Schottky Barrier Diode,” by Kohji Andoh and Davide Chiola, the contents of which are herein incorporated by reference.

[0002] 1. Field of the Invention

[0003] The present invention relates to semiconductor devices and more particularly, to a trench Schottky diode with differential oxide thickness and a process for fabricating such a device.

[0004] 2. Discussion of the Related Art

[0005] Schottky diodes are well known and are made with different topologies, including a trench topology as typically shown in U.S. Pat. No. 5,612,567 by Bantval Baliga. The process for manufacturing a trench Schottky diode requires a large number of mask layers and manufacturing steps. However, U.S. patent application Ser. No. 10/193,783, by the present inventor, presented an inventive process for fabricating trench Schottky diodes in which a reduced number of steps and fewer mask layers are required.

[0006] Referring to the Figures, in which like reference numerals refer to like elements, there is shown in FIG. 1 an example trench Schottky diode fabricated according to the process of the aforementioned application, FIG. 1 being a reproduction of FIG. 11 of that application. As seen, the device of FIG. 1 includes a silicon wafer 10 with a plurality of mesas 54 separating a plurality of trenches 30, including a termination trench region 60. A thin uniform oxide layer (SiO₂) 44 lines the sidewall and bottom surfaces of each trench 30 and termination region 60, forming gate oxides and a termination oxide, respectively. A P type polysilicon 48 fills each trench 30 thereby forming a plurality of electrodes that under reverse bias, reduce reverse biased leakage currents and increase the reversed bias blocking voltage. A Schottky barrier metal 50 covers the active region, forming Schottky contacts 52 that extend over the tops of the mesa regions 54. Finally, anode electrode 56 and a cathode electrode (not shown) extend over the top and bottom surfaces, respectively, of the device.

[0007] In a trench Schottky device, such as that shown in the example device of FIG. 1, the oxide layer is formed simultaneously along the sidewall and bottom surfaces of each trench 30 and termination region 60 and thereby has the same uniform thickness throughout. However, the oxide layer along the sidewall and bottom surfaces of the trenches and termination region accomplish different purposes. More specifically, depending on the thickness of the oxide layer in each of these regions, the electrical and structural characteristics of the resulting device are affected in different ways.

[0008] For example, the gate oxide on the bottom surface of each trench 30 serves to screen the high electric field regions located at the trench bottom corners. In general, the higher the oxide thickness in this critical region, the higher the reverse breakdown voltage the device can sustain. Accordingly, a thicker gate oxide is desired along the bottom surface of the trenches.

[0009] On the contrary, the gate oxide grown on the sidewall of each trench 30 mainly affects the pinch-off characteristics of the reversed bias leakage current. Specifically, during reverse bias, the oxide on the sidewall of a trench serves to transfer the anode voltage from the trench electrode to the mesa region 54. However, part of the negative anode voltage is dropped in the oxide, with the balance of the voltage serving as the pinch-off voltage for the mesa conductive region.

[0010] Accordingly, a thin gate oxide thickness is desired along the trench sidewall in order to minimize the voltage dropped in the oxide and allow a reduced reverse biased leakage pinch-off voltage. In addition, a thin oxide thickness along the trench sidewall means less silicon is consumed during the oxidation process, thereby increasing the conduction area in the mesa regions 54. As is known, an increased mesa region means a higher active area, which is beneficial towards reducing the forward voltage drop of the device during forward conduction.

[0011] As for the termination oxide in termination region 60, it serves several purposes. From an electrical perspective, the oxide along the bottom surface of the region serves as a field oxide for the field plate termination. In general, it is desirable to have this field oxide thick enough to minimize the electric field crowding that occurs in the oxide underneath the edge 58 of the metal field plate and that is responsible for breakdown voltage walk-out. From a mechanical perspective, this portion of the field oxide underneath the edge of the metal field plate experiences high stress during temperature cycle reliability tests of the packaged device. Again, a thicker field oxide is desirable to avoid oxide rupture under the edge of the field plate during these tests.

[0012] As seen, it is disadvantageous to use the same oxide thickness across all regions of the trench Schottky device. In particular, it is desirable to have the oxide layer that lines the sidewalls of the trenches and termination region thinner than the oxide layer that lines the bottom surfaces of these trenches.

[0013] Nonetheless, it should also be noted that others have disclosed trench type devices that use a gate oxide of variable thickness; in particular, U.S. Pat. No. 6,236,099 by Milton Boden and U.S. Pat. No. 6,580,123 by Naresh Thapar disclose such devices. However, both patents are directed at MOSgated devices, including MOSFETs and IGBTS, and not at trench Schottky devices. In addition, these patents only teach the use of a variable gate oxide and not the use of a variable termination oxide.

[0014] The process by which Thapar forms the variable gate oxide should also be noted. In particular, this process includes first depositing a silicon nitride layer on the trench sidewall and bottom surfaces. The nitride layer at the bottom surface is then etched away and an oxide layer grown thereupon to a desired thickness. The nitride layer on the trench sidewall is then removed and a subsequent oxide layer is grown on the sidewall to a desired thickness.

[0015] Accordingly, it would be desirable to produce a trench Schottky diode that has differential or varying oxide thickness in different regions of the device, thereby overcoming the above and other disadvantages of the prior art. In accordance with the present invention, a novel fabrication process is employed for the manufacture of an inventive trench Schottky diode that has differential oxide thickness and more particularly, that has an oxide layer along the sidewall of each trench that is thinner than the oxide layer along the trench bottom surfaces.

[0016] Specifically, in accordance with the present invention, a first silicon nitride layer is initially deposited directly on a surface of a silicon substrate. During a trench mask and etching step, a plurality of mesas and trenches are then formed in the surface of the substrate. This step may also include the formation of a termination trench. Once the trenches and termination trench are formed, a sacrificial oxide is grown on the sidewall and bottom surfaces of these trenches. This sacrificial oxide is then stripped. Next, a gate and termination oxide are grown on the sidewall and bottom surfaces of the trenches and termination trench, respectively, such that the sacrificial oxide thickness to the gate/termination oxide thickness is preferably at a ratio of 2:1. A second nitride layer is then deposited over the sidewall and bottom surfaces of these trenches and over the first nitride layer that remains on the substrate surface. This second nitride layer is then dry etched, once again exposing the first nitride layer and more importantly, exposing the oxide layer along the bottom surface of the trenches. However, the oxide layer along the sidewall of the trenches remains covered. Accordingly, the bottom surface of the trenches are then once again oxidized. Following this second oxidation, the second nitride layer is then stripped during a wet etch from the sidewall of the trenches. As a result, in accordance with the present invention, an oxide layer is formed on the sidewall of each of the trenches and the termination trench that is thinner than the oxide layer formed on the bottom surfaces of these trenches.

[0017] Once the second nitride layer is removed, the trenches are next filled with a P type polysilicon. The remaining first nitride layer is then stripped in a wet etch. Next, a Schottky metal layer is deposited over the active surface of the substrate forming a Schottky rectifier contact with the plurality of mesas. Finally, anode and cathode electrodes are formed on opposite surfaces of the substrate.

[0018] As such, according to the present invention, a novel trench Schottky diode that has differential oxide thickness is formed, wherein the oxide layer on the sidewalls of the trenches and the termination trench is thinner than the oxide layer along the trench bottom surfaces.

[0033] Referring now to FIGS. 2-14 (note that the Figures are not drawn to scale), a fabrication process and a resulting device structure for a trench Schottky diode that has differential or varying oxide thickness along trench sidewall and bottom surfaces is illustrated. The process is a modification of the two mask process presented in application Ser. No. 10/193,783. However, by varying the thickness of the oxide layer along the trench sidewall and bottom surfaces according to the present invention, a device is formed that has improved reverse breakdown voltage and that is less susceptible to temperature cycles. Note that the process described here is for a low voltage Schottky device rated at 15-45V. However, one skilled in the art will recognize that with suitable changes, the described process applies to any device rating.

[0034] Beginning with FIG. 2, an initial silicon wafer 100 is shown having an N+ substrate 104 with an N− epitaxially grown upper surface 102 (epi layer). As an example, wafer 100 can have a silicon thickness ranging from 2.5 um to approximately 4 um with a resistivity between 0.3 and 0.5 ohm cm; however, other silicon thickness and resistivity can be used. Wafer 100 is initially cleaned using, for example, hydroflouric (HF) acid, which removes thermal and native oxides.

[0035] Once cleansed, a removable surface nitride layer (Si₃Ni₄) 106 is directly deposited at a thickness of approximately 600-800 Å on the surface 108 of the epi layer 102 using a low-pressure chemical vapor deposition technique (LPCVD). Note that nitride layer 106 is deposited without first forming a pad oxidation layer.

[0036] In order to form spaced trenches, a layer of photoresist and a mask with a desired trench pattern (step not shown) are next applied to the surface of nitride layer 106. The resulting structure is then patterned using a suitable photolithographic process.

[0037] Next, using a suitable etching process, such as plasma etching, nitride layer 106 and epi layer 102 are etched to form a plurality of trenches 110 and mesas 122, as shown in FIG. 3. The trenches extend through nitride layer 106 and extend downwardly from the top side 108 of epi layer 102 and into the epi layer. As an example, the trenches can be arranged in parallel spaced stripes and can have a width of about 0.5 microns, a spacing of about 0.5 microns, and a depth of 1.5 microns. Nonetheless, other trench formations and dimensions can be utilized.

[0038] As shown in FIG. 3, during this process of forming trenches 110, a termination trench 111 can also be formed, thereby forming a termination trench region, as shown by arrow 112. Nonetheless, one skilled in the art will recognize that the present invention also applies to trench Schottky devices that do not include a termination trench region (for discussion purposes, the formation of a termination region will be assumed).

[0039] Once trenches 110 and 111 are formed, the remaining portion of the photoresist mask is stripped, and the exposed surfaces of the trenches undergo a pre-diffusion cleaning step.

[0040] Turning to FIG. 4, a sacrificial oxide layer (SiO₂) 114 having a thickness of approximately 1000-1500 Å is next simultaneously grown on the sidewall and bottom surfaces of each trench 110 and 111. A sacrificial oxide etch at a target depth of approximately 1500-2000 Å is then performed to completely remove this oxide layer, as shown in FIG. 5.

[0041] Next, according to the present invention and as shown in FIG. 6, a uniform oxide layer 116 is simultaneously grown to a thickness of approximately 500-750 Å on the sidewall and bottom surfaces of trenches 110 and 111 using a wet or dry process, thereby forming a gate oxide and termination oxide, respectively. Note that the preferred ratio of the sacrificial oxide thickness to the gate oxide thickness, grown at this step, is 2:1

[0042] $\left(i.e.,\frac{\mathrm{sacrificial}-\mathrm{oxide}-\mathrm{thickness}}{\mathrm{gate}-\mathrm{oxide}-\mathrm{thickness}}=2\right).$

[0043] As shown in FIG. 7, a removable trench nitride layer 118 is then deposited at a thickness of approximately 150-200 Å over surface nitride layer 106 and over oxide layer 116 along the sidewall and bottom surfaces of trenches 110 and 111 using for example, a low-pressure chemical vapor deposition technique (LPCVD). Using a dry nitride etch, trench nitride layer 118 is then removed from surface nitride layer 106, from bottom surface 110a of trenches 110, and from bottom surface or field region 111a of termination trench 111, thereby effectively applying the trench nitride layer to only the sidewall of each trench, as shown in FIG. 8. As such, oxide layer 116 is now exposed only along the bottom surfaces 110a and 111a of trenches 110 and 111, respectively.

[0044] Accordingly, exposed oxide layer 116 along trench bottom surfaces 110a and 111a is next grown to a total thickness of approximately 1000-5000 Å using a wet or dry process, as shown in FIG. 9. A wet nitride etch, for example phosphoric acid at 150° C., and at a target nitride removal of 200-250 Å is next performed to strip the remaining trench nitride layer 118 covering the sidewall of trenches 110 and 111. As such, according to the present invention and as shown in FIG. 10, oxide layer 116 lining trenches 110 and 111 now has a differential or varying thickness, with the oxide layer lining sidewalls 110b and 111b of trenches 110 and 111, respectively, being thinner than the oxide layer lining bottom surfaces 110a and 111a.

[0045] Again, note that the above oxide thicknesses are for example purposes only and one skilled in the art will recognize that other oxide thicknesses can be used to produce devices with different ratings without deviating from the present invention.

[0046] Turning to FIG. 11, a layer of an un-doped polysilicon 120 is next deposited on the surface of the above structure to a thickness of approximately 7500 Å to fill trenches 110 and 111. A boron implant, for example 1E14/cm² at 80 kev, is then applied to the surface of the structure. Nonetheless, one skilled in the art will recognize that any implant species that acts as a P-type dopant can be used. Among these dopants, Boron or BF₂ are the most common.

[0047] Then, after yet another pre-diffusion clean, the implanted species are activated and driven at 1050° C. for one hour to make the polysilicon P type conductive within the trenches. As a result, a plurality of electrodes are now formed within the trenches and between the mesas.

[0048] Next, a blanket poly etch is carried out by a suitable plasma etcher for a period of time lasting for at least five more seconds than would be necessary to remove the polysilicon on the device surface. A suitable mask is then applied and a portion of the polysilicon along the bottom surface 111a or field region of termination trench 111 is removed. Following these steps, the remaining surface nitride layer 106 covering mesa regions 122 is stripped by a wet etch in phosphoric acid at 150° C. The resulting structure is shown in FIG. 12.

[0049] Next, referring to FIG. 13, top face 108 of the structure is cleaned using any desired pre-metal clean. A barrier metal 126, such as titanium (Ti) or titanium-tungsten (TiW), etc. is next sputtered on surface 108. In general, the titanium layer can be applied at any thickness but preferably, the titanium layer is about 600 Å thick to maximize the thickness of the subsequent silicide layer and to minimize the thickness of the un-reacted titanium. Note also that any technique can be used for the titanium deposition, with sputtering and electron beam evaporation being the most common techniques.

[0050] Next, the layer of titanium is annealed at a high temperature in an inert atmosphere. As a consequence, the thin titanium layer reacts with the active-device region to form a titanium silicide layer over the active region, forming Schottky contacts 124 that extend over the tops of mesa regions 122. The non-reacted titanium layer extending along the termination trench 111 is then removed by etching the structure with any known Ammonium Hydroxide and Hydrogen Peroxide-based solution. In general, etch time can vary, but etch temperature should not exceed 80° C. to avoid excessively fast decomposition of the H₂O₂. The resulting structure is shown in FIG. 13.

[0051] As seen in FIG. 14, an anode contact metal layer 130, which may be a layer of aluminum or other conductive metal layer, is next deposited on the top of the structure. A metal mask (not shown) is then applied and contact metal layer 130 is partially etched within the termination trench 111 thereby leaving only a peripheral portion 128 of the metal in this region.

[0052] Finally, the entire wafer is attached to a frame of a grinding device (not shown) via an adhesive tape and a back-grind is carried out, thinning the wafer to 8 mils, after which the work piece is de-taped and etched to smooth out roughened surfaces as a result of grinding. A backside metal, such as a trimetal (Ti/Ni/Ag), is then sputtered on the bottom 132 of the wafer to form a cathode electrode on the bottom surface of the Schottky rectifier. The wafer may then be diced along scribe lines to separate a plurality of identical diodes.

[0053] Although the present invention has been described in relation to particular embodiments thereof, many other variations and modifications and other uses will become apparent to those skilled in the art. It is preferred, therefore, that the present invention be limited not by the specific disclosure herein.