DETAILED DESCRIPTION OF THE INVENTION

[0044] With reference to the accompanying drawings, a semiconductor device embodying the invention will be described in terms of a MIS field effect transistor (MISFET) of the three-dimensional structure.

[0045] FIG. 6 is a perspective view of a semiconductor device according to a first embodiment of the invention.

[0046] As shown in FIG. 6, a p-type silicon semiconductor substrate 11 is processed to have a substrate projection 11 A. The projection 11 A forms a device region. Portions of the substrate 11 on opposite sides of the projection 11 A form device isolation regions. A device isolation insulating film 12 is formed on semiconductor substrate 11 in the device isolation regions. Here, the projection 11 A is 0.1 μm or less in thickness (corresponding to the thickness in the direction normal to the direction of the channel length) and 0.5 to 1.0 μm in height from the surface of the substrate 11 . The height of the projection need not be limited to 1.0 μm or less and may be more than 1.0 μm if the manufacturing technology permits.

[0047] A gate insulating film 13 is formed on top and side with the substrate projection 11 A. That is, the gate insulating film 13 is formed to cover the substrate projection 11 A. The gate insulating film 13 consists of, for example, silicon oxide formed by thermal oxidation.

[0048] A gate electrode 14 is formed on a portion of the gate insulating film 13 that covers the projection 11 A. Also, the gate electrode 14 is formed on a portion of the device isolation insulating film 12 . After the gate electrode was formed, source diffusion layers 15 and drain diffusion layers 16 , which are opposite in conductivity type to the semiconductor substrate 11 , are formed in those portions of the side surfaces of the projection 11 A which are located on opposite sides of the gate electrode 14 . In the step of forming the source and drain regions, phosphorous (P) or arsenic (As) is introduced into the side surfaces of the projection except below the gate electrode by means of ion implantation. The ion implantation is carried out in a self-aligned manner using the gate electrode 14 as a mask.

[0049] Although, in the above description, the source and drain regions 15 and 16 are formed only in the side surfaces of the projection 11 A, they may also be formed in the top surface of the projection as required. In that case, contact can be made to interconnect layers via the drain and source regions formed in the top surface.

[0050] Heavily doped regions 17 of p-type conductivity are formed in the substrate below the device isolation insulating film 12 and the substrate projection 11 A. The heavily doped regions 17 are of the same conductivity type as the substrate. That is, the p ⁺ -type regions 17 formed by ion implantation in the semiconductor substrate 11 below the device isolation insulating film 12 on opposite sides of the substrate projection 11 A are joined together in the substrate below the projection.

[0051] A contact plug 18 is formed in the device isolation insulating film 12 to establish electrical connection between the semiconductor substrate 11 and an interconnect layer not shown. Since the p ⁺ -type regions 17 have been formed on top of the substrate 11 , ohmic contact can be obtained between the contact plug 18 and the substrate 11 without performing ion implantation anew at the time of forming the contact plug.

[0052] The semiconductor device shown in FIG. 6 serves the same function as the previously described conventional SOI substrate-based MISFET through the use of the bulk substrate. The gate electrode 14 needs a contact area (not shown) for connection to an interconnect layer. For this reason, the gate electrode 14 is formed to overlap the substrate 11 in the device isolation region.

[0053] With the semiconductor device, in order to maintain device-to-device isolation, it is required to control the short channel effect caused by a parasitic MISFET and to cause the parasitic MISFET in the overlapping portion of the gate electrode and the semiconductor substrate in the device isolation region to turn off at all times under actually used operating voltages.

[0054] In this embodiment, therefore, the p ⁺ -type regions 17 is formed in the substrate in the device isolation region including portions just below the gate electrode 14 by doping p-type impurities of opposite conductivity type to carriers in the channel. Further, the device isolation insulating film 12 on the substrate in the device isolation region is formed to be large in thickness, so that the insulating film that acts effectively as a gate insulating film in the overlapping portion of the gate electrode and the semiconductor substrate is made large in thickness. For this reason, the threshold voltage of the parasitic MISFET formed in the device isolation region can be increased, allowing the parasitic MISFET to be in the off state at all times. For example, to form an n-channel MISFET, boron (B) is introduced into the device isolation region to form the p ⁺ -type impurity regions 17 .

[0055] If the thickness of the substrate projection 11 A (the thickness in the direction normal to the direction of the channel length) is smaller than the maximum width Wd of the depletion layer formed in the substrate projection 11 A when a voltage is applied to the gate electrode, then the entire substrate projection will be filled with the depletion layer at the time of operation. Thus, the MISFET of this embodiment performs the same operation as a fully depleted SOI device. In this case, even if the doping level of the substrate projection 11 A is low, since potential control in the channel is performed by the gate electrode on the sidewalls of the projection, it is possible to control the short channel effect with ease in comparison with a planar MISFET of the conventional structure.

[0056] For the same operation as the SOI device, the doping concentration of the substrate projection 11 A can be set low, thus allowing the magnitude of electric field in the vertical direction produced by impurities in the substrate to be smaller than in normal planar MISFETs and the carrier mobility represented by a function of the vertical electric field to be larger than in the planar devices. Therefore, the MISFET of this embodiment provides higher current drivability than the planar devices even with the same operating voltage and the same gate width.

[0057] Decreasing the thickness of the substrate projection 11 A causes lateral diffusion of impurities which are ion implanted vertically into the semiconductor substrate 11 in the device isolation region, resulting in the substrate region below the projection 11 A being doped. That is, the p ⁺ -type doped regions 17 on opposite sides of the projection are joined together in the substrate region below the projection, so that the P ⁺ -type doped region is also formed below the projection. With the semiconductor device of this embodiment, therefore, the device isolation characteristic can be improved because not only the substrate region in the device isolation region but also the substrate region below the substrate projection 11 A can be doped with impurities. That is, the occurrence of punch-through can be prevented and the isolation between adjacent devices can be maintained.

[0058] In the semiconductor device of this embodiment which performs the same operation as fully depleted SOI devices, since its channel region is not isolated from the underlying semiconductor substrate by an insulating film low in thermal conductivity, the heat conduction characteristic can be improved, allowing a degradation in current caused by heat due to the Joule effect to be minimized.

[0059] Second Embodiment

[0060] In general, in the case of MISFETs using the SOI substrate, when a device that is very short in channel length is formed with the doping concentration of the substrate reduced, there is the possibility that punch-through occurs between source and drain as a result of extension of the depletion layer from the drain side. In order to control the short channel effect, the extension of the depletion layer must be controlled.

[0061] With conventional fully-depleted planar MISFETs using the SOI substrate, the occurrence of punch-through is controlled by making the channel forming silicon layer very thin. When the gate length is reduced to 100 nm or less, the silicon layer must be made much smaller in thickness than the gate length, which increases the difficulty in fabricating devices.

[0062] Consider here the case where, in the device of FIG. 6 , the substrate projection is made large in height and the channel width is made large so as to increase effectively the area in which current flows.

[0063] In this case, the transistor portion formed in the side surfaces of the substrate projection, as a thin-film device that performs the same operation as SOI devices, is relatively effective in controlling the short channel effect. The reason is that the thickness of the channel forming silicon layer is defined by the width of the substrate projection and the double-gate structure functions favorably.

[0064] The transistor formed in the top surface of the substrate projection is affected by the drain impurity diffusion layers formed in the side surfaces of the projection, so that the drain junction depth increases and the effective SOI film thickness in the vertical direction looks large.

[0065] As a result, depending on the source/drain structure, the extension of the depletion layer on the drain side increases, making punch-through between source and drain easy to occur. This is prominent particularly in the case where the substrate projection is 0.1 μm or more in height. For this reason, trying to increase the apparent current drivability by increasing the height of the substrate projection for compensation of the channel width results in increased possibility of punch-through.

[0066] It is the prior art shown in FIG. 4 that is intended to remedy this drawback by optimizing the substrate doping profile. In this case as well, however, increasing the gate width, i.e., increasing the height of the substrate projection, results in difficulty in doping the entire region in which punch-through can occur with impurities. The effective upper limit of the gate width is determined by the range of depth to which impurities can be doped by the ion implantation technology.

[0067] The second embodiment is intended to prevent the occurrence of punch-through between source and drain even with a very fine gate electrode formed by preventing the channel from being formed in the top surface of the substrate projection, i.e., by using only the side surfaces of the projection as the channel. Unlike the prior art, the semiconductor device according to the second embodiment of the present invention is basically characterized in that the top portion of the substrate projection is not used as the channel.

[0068] Examples of structures such that no channel is formed in the top surface of the substrate projection are illustrated in FIGS. 7, 8 and 9 . In the example of FIG. 7 , the top surface is doped with impurities of opposite conductivity type to carriers in the channel. In the example of FIG. 8 , the gate oxide film formed on the top of the substrate projection is made sufficiently large in thickness so as not to allow the channel to be formed under actually used operating voltages. In the example of FIG. 9 , the gate electrode is formed so that it is not located on the top surface of the projection. Any combination of these three structures may be used.

[0069] The structures shown in FIGS. 7, 8 and 9 will be described below in detail.

[0070] FIG. 7 is a perspective view of a semiconductor device according to the second embodiment.

[0071] A region 21 is formed in the upper portion of the substrate projection 11 A, as shown in FIG. 7 . The region 21 is doped with p-type impurities of opposite conductivity type to carriers in the channel. Otherwise, the device of FIG. 7 remains unchanged from the device of FIG. 6 .

[0072] The semiconductor device of such a structure is fabricated as follows: First, in forming the substrate projection 11 A, an insulating film as a cap film, such as a film of silicon nitride, is formed over the surface of the semiconductor substrate 11 and then patterned in the form of a strip. The semiconductor substrate is then subjected to reactive ion etching (RIE) using the silicon nitride film as a mask, whereby the substrate projection 11 A of a given width (thickness) and height is formed.

[0073] After the formation of the projection, impurities are introduced vertically into the top surface of the projection by means of ion implantation. For an n-channel MISFET as in this embodiment, boron (B) is implanted at 15 keV and at a dose of 5×10 ¹³ cm ⁻² or more.

[0074] Subsequently, the gate insulating film 13 is formed on the top surface and the side surfaces of the projection by means of thermal oxidation. A film of polysilicon is deposited onto the gate insulating film 13 and then patterned to form the gate electrode 14 .

[0075] Moreover, impurities of, for example, P or As are introduced into the side surfaces of the projection 11 A except their portions below the gate electrode 14 by means of ion implantation, thereby forming the source and drain regions 15 and 16 .

[0076] FIG. 8 is a perspective view of a semiconductor device according to a modification of the second embodiment.

[0077] An insulating film 22 is formed on the substrate projection 11 A as shown in FIG. 8 . It is recommended that as the insulating film 22 the cap film (for example, silicon nitride film) used in forming the substrate projection be used as it is. Alternatively, a silicon oxide film may be separately formed as the insulating film 22 . Otherwise, the device of FIG. 9 remains unchanged from the device of the first embodiment.

[0078] With the semiconductor device thus constructed, the insulating film present between the gate electrode 14 and the top surface of the substrate projection 11 A can be made sufficiently thick so as not to allow the channel to be formed under usually used operating voltages.

[0079] The semiconductor device of such a structure is fabricated as follows: First, in forming the substrate projection 11 A, an insulating film as a cap film, such as a film of silicon nitride, is formed over the surface of the semiconductor substrate 11 and then patterned in the form of a strip. The semiconductor substrate is then subjected to reactive ion etching (RIE) using the silicon nitride film as a mask, whereby the substrate projection 11 A is formed.

[0080] Subsequently, the gate insulating film 13 consisting of silicon oxide is formed on the side surfaces of the projection by means of thermal oxidation with the insulating film 22 left. A film of polysilicon is deposited onto the gate insulating film 13 and the insulating film 22 and then patterned to form the gate electrode 14 .

[0081] Moreover, impurities, such as P or As, are introduced into the side surfaces of the projection 11 A except their respective portions below the gate electrode 14 by means of ion implantation, thereby forming the source and drain regions 15 and 16 .

[0082] FIG. 9 is a perspective view of a semiconductor device according to another modification of the second embodiment.

[0083] As shown in FIG. 9 , no gate electrode is formed on top of the projection 11 A, but gate electrodes 14 A and 14 B are formed only on the sidewalls of the projection. That is, two gate electrodes 14 A and 14 B are formed on the sidewalls of the projection so as to sandwich the projection and self-align to each other. The two gate electrodes are arranged on a straight line perpendicular to the direction of the channel length. With this semiconductor device, the gate electrode is separated into two; thus, it is required to provide each of the two electrodes with a separate contact. Otherwise, the device of FIG. 9 remains unchanged from the device of the first embodiment.

[0084] The semiconductor device thus constructed can be used in the double-gate FET mode in which the same bias voltage is applied to the two gate electrodes 14 A and 14 B. It is also possible to apply a different voltage to each of the gate electrodes 14 A and 14 B.

[0085] An example of applying a different voltage to each of the two gate electrodes 14 A and 14 B is the backgate FET mode in which one of the two electrodes is supplied with a gate voltage for channel control and the other is supplied with a different voltage as the substrate potential. Although such a fully depleted device as shown in FIG. 7 is not allowed to change the threshold voltage after fabrication, the device shown in FIG. 9 is allowed to control the threshold voltage when used in the backgate FET mode.

[0086] A plurality of semiconductor devices shown in FIG. 9 can be used simultaneously, in which case modifications to the interconnections and the power supply system would allow some devices to operate in the double-gate FET mode and other devices to operate in the backgate FET mode.

[0087] The semiconductor device of such a structure as shown in FIG. 9 is fabricated as follows: First, in forming the substrate projection 11 A, an insulating film as a cap film, such as a film of silicon nitride, is formed over the surface of the semiconductor substrate 11 and then patterned in the form of a strip. The semiconductor substrate is then subjected to reactive ion etching using the silicon nitride film as a mask, whereby the substrate projection 11 A is formed.

[0088] Subsequently, the gate insulating film 13 consisting of silicon oxide is formed on the side surfaces of the projection by means of thermal oxidation with the silicon nitride film left. A film of polysilicon is deposited onto the gate insulating film 13 and the insulating film 22 and then patterned to form the gate electrode 14 .

[0089] After that, the polysilicon film present on the top of the projection is removed by means of CMP or RIE. Moreover, the insulating film present on the top of the projection is removed. Furthermore, impurities, such as P or As, are introduced into the side surfaces of the projection 11 A except their respective portions below the gate electrode 14 by means of ion implantation, thereby forming the source and drain regions 15 and 16 .

[0090] In this case, it is possible to form a self-aligned double-gate MISFET that performs the same operation as an SOI device.

[0091] Each of the semiconductor devices of the second embodiment is allowed to perform the same operation as fully depleted MISFETs using the SOI substrate.

[0092] The structure to prevent the channel from being formed in the upper portion of the substrate projection 11 A will not provide a very serious damage under conditions that the width of the substrate projection 11 A must be made small and the height of the projection must be set to 1 μm or more to obtain current drivability. If anything, it is apparent that the feature of the present invention of not using the top surface (upper portion) of the substrate projection for the channel provides effective means for controlling the short channel effect.

[0093] In the case of transistor operation using only the side surfaces of the substrate projection, in order to increase carrier mobility, it is required in the case of silicon that the side surface of the projection be parallel to the (100 ) plane and the channel be oriented in the direction of [100].

[0094] Third Embodiment

[0095] With the conventional three-dimensional MISFET shown in FIGS. 1A to 1 C and FIG. 2 , in order to implement a fully depleted SOI device, it is required to make the thickness of the channel region small. If the thickness is 50 nm or less, it is advantageous to the formation of shallow junctions for source and drain regions. On the other hand, the source and drain regions becomes very thin in comparison with conventional planar MISFETs. For this reason, parasitic resistance associated with the source and drain regions becomes high. As a result, there will be a degradation in current drivability.

[0096] The third embodiment is therefore intended to provide a MISFET which uses a thick substrate projection in which, as shown in FIGS. 10 and 11 A to 11 D, the channel region and its neighborhood are made small in thickness, whereas the region in which the source and drain are to be formed are made large in thickness, thereby allowing the increase of the parasitic resistance to be minimized.

[0097] FIG. 10 is a perspective view of a semiconductor device according to the third embodiment. FIG. 11A is a plan view of the semiconductor device. FIG. 11B is a side view of the semiconductor device. FIG. 11C is a cross-sectional view of the semiconductor device taken along line 11 C- 11 C of FIG. 11A . FIG. 11D is a cross-sectional view of the semiconductor device taken along line 11 D- 11 D of FIG. 11A .

[0098] The semiconductor device shown in FIGS. 10 and 11 A to 11 D are fabricated in the following way. The device geometry somewhat varies according to the fabrication method.

[0099] First, portions of a thick silicon substrate 11 around the region which is to form a projection are etched away to form the substrate projection of 0.15-0.20 μm in width. The direction of the width of the projection is parallel to the surface of the substrate 11 and normal to the direction from source to drain (i.e., the direction of the channel length). Next, an insulating film of, for example, silicon nitride is deposited which is to be used as a mask in forming the gate electrode. The silicon nitride film is patterned using lithographic techniques to form a groove in that film.

[0100] The semiconductor substrate 11 is subjected to oxidation to form an oxide film of 50-100 nm in thickness, the oxide film comes to have a region which resembles a bird's beak as is well known in the LOCOS process. After that, by selectively removing the oxide film, the substrate projection 31 A is formed which is thick in the source and drain portions and thin in the channel portion and part of the diffusion portion near the channel.

[0101] After that, the gate insulating film 13 is formed on the top surface and the side surfaces of that portion of the projection 31 A which is made thin in the gate electrode forming mask. A polysilicon film is formed on the gate insulating film 13 in the gate electrode forming mask and the excess polysilicon film is polished by CMP to form the gate electrode 14 .

[0102] Next, the gate electrode forming mask material (the silicon nitride film) is removed. After that, ion implantation or vapor phase doping is performed on the side surfaces (source and drain formed portions) of the substrate projection except the channel region, thereby forming deep, low-resistivity source and drain diffused layers 15 and 16 . At the same time, shallow junctions are formed in the diffused portions near the channel region. When the adjustment of doping conditions is required for the diffused portions and the deep junction portions, it is possible to form the deep junctions by forming the gate sidewall after the formation of the shallow junctions in the diffused portions as in the case of conventional planar MISFETs. the same structure can also be formed by applying the elevated source/drain structure using epitaxial techniques after the formation of a transistor in the substrate projection.

[0103] Although, in this embodiment, the source diffused layers 15 and the drain diffused layers 16 are formed only in the side surfaces of the substrate projection 31 A, they may be formed in the top surface of the projection as required. In that case, contacts may be made to interconnect layers via the source and drain diffused layers 15 and 16 formed in the top surface.

[0104] Fourth Embodiment

[0105] As described previously, in order to make a MISFET that performs the same operation as SOI devices having a very fine gate regardless of whether it is of the conventional type or the three-dimension type, it is required to form the channel portion from a very thin silicon layer. Depending on circumstances, however, the structures described so far are expected to make it very difficult to process semiconductor substrates through lithographic and RIE techniques in particular.

[0106] The fourth embodiment is therefore intended to provide a three-dimensional MISFET that performs the same operation as SOI devices by forming a substrate projection of a relatively large thickness of, for example, the order to 0.5 to 1.0 μm. The fourth embodiment is characterized by forming an intrinsic pillar as a substrate projection and then forming a stacked channel structure of stacked p ⁺ /n ⁻ /p ⁻ layers.

[0107] FIG. 12 is a perspective view of a portion of a semiconductor device of the fourth embodiment. FIG. 12 shows particularly the channel portion, and the projection in which the source and drain diffused layers are formed is made large in thickness as illustrated in FIGS. 10 and 11 A to 11 D.

[0108] First, to form a substrate projection 41 A, an insulating film as a cap film, such as a film of silicon nitride, is formed over the surface of a p-type semiconductor substrate 11 and then patterned into the shape of a strip. The semiconductor substrate is then subjected to reactive ion etching using the silicon nitride film as a mask, whereby an intrinsic pillar of a given width (thickness) and height is formed. A p ⁺ -type layer 42 , an n ⁻ -type layer 43 and a p ⁻ -type layer 44 are formed in this order on the top surface and the side surfaces of the pillar, i.e., so as to surround the pillar, to form the substrate projection 41 A.

[0109] Next, a gate oxide film 13 consisting of silicon oxide is formed on the side surfaces and the top surface of the substrate projection 41 A by means of thermal oxidation. A polysilicon film is deposited on the gate insulating film 13 and then patterned to form a gate electrode 14 .

[0110] FIG. 13 is a perspective view of a portion of a semiconductor device according to a modification of the fourth embodiment. FIG. 13 shows particularly the channel portion, and the projection in which the source and drain diffused layers are formed is made large in thickness as illustrated in FIGS. 10 and 11 A to 11 D.

[0111] First, to form a substrate projection 41 B, an insulating film as a cap film, such as a film of silicon nitride, is formed over the surface of a p-type semiconductor substrate 11 and then patterned into the shape of a strip. The semiconductor substrate is then subjected to reactive ion etching using the silicon nitride film as a mask, whereby an intrinsic pillar of a given width and height is formed.

[0112] Subsequently, impurities of, for example, boron are introduced into the device isolation region of the semiconductor substrate 11 through ion implantation to form p ⁺ -type doped regions 17 . At the same time, the impurities are also introduced into the pillar to form a p ⁺ -type layer 42 . After that, a device isolation insulating film 12 of, for example, silicon oxide is formed over the device isolation region of the semiconductor substrate 11 .

[0113] An n ⁻ -type layer 43 and a p ⁻ -type layer (channel layer) 44 are grown over the p ⁺ -type layer 42 , thereby forming the substrate projection 41 B.

[0114] With the semiconductor device shown in FIG. 13 , the formation of the p ⁺ -type doped region 17 into the device isolation region of the semiconductor device 11 is carried out when the pillar is still thin, that is, immediately after the pillar has been formed. As a result, the p-type impurities diffuse to under the substrate projection, allowing the p ⁺ -type doped region 17 to come in contact with the innermost p ⁺ -type layer 42 of the pillar. Thus, the potential application to the p ⁺ -type doped region 17 results in the potential application to the p ⁺ -type layer 42 of the substrate projection 41 B, allowing the device to operate as a four-terminal device.

[0115] With the semiconductor device shown in FIG. 12 or 14 , setting the thickness and doping level of the n ⁻ -type layer 43 so that it is fully depleted without depending on gate voltage and optimizing the doping levels of the p ⁺ -type layer 42 and the p ⁻ -type layer 44 allow the depletion layer formed in the n-type layer 43 to provide isolation between the p ⁻ -type layer 44 in which the channel is formed and the semiconductor substrate 11 . It therefore becomes possible to realize a device structure equivalent to MISFETs using the SOI substrate. In this case, to implement a fully depleted device, the thickness of the channel layer (p ⁻ -type layer 44 ) is simply made small.

[0116] Moreover, the center pillar of the substrate projection 41 A or 41 B can be formed to any thickness. Thus, the thickness of the pillar can be set so that the substrate can be processed easily to form the substrate projection with the same structure as thin-film SOI devices. Thus, the difficulties involved in fabricating the device can be decreased.

[0117] FIGS. 12 and 13 depict, in enlarged view, only the channel portion just below the gate. In the source and drain, it is required that their respective diffused layers be not in contact with the n ⁻ -type layer 43 . For the source and drain structures, therefore, such elevated source/drain structures using selective epitaxial growth as used in the device shown in FIGS. 10 and 11 A to 11 D or halo structures (pocket structures) are recommended. The use of these structures allows easy prevention of the source and drain diffused layers (n ⁺ -type layers) from coming into contact with the n-type layer 43 and the channel structures shown in FIGS. 12 and 13 to be implemented.

[0118] Although the fourth embodiment has been described in terms of an n-channel field effect transistor, the principles of the invention can be applied to a p-channel field effect transistor by reversing the conductivity type of impurities. By performing doping of the well and the channel separately and optimizing the halo structure, devices equivalent to high-performance CMOS SOI devices can be implemented.

[0119] Fifth Embodiment

[0120] In the previously described three-dimensional MISFETs, increasing the channel width, i.e., the height of the substrate projection, makes large the difference in level between the gate contact area and the source and drain contact areas, which, in future, may cause a problem of difficulty in forming the contacts using the same process. For example, to set the channel width to about 2 μm, it is required to set the height of the substrate projection to about 1 μm. In this case, processing problems make it impossible to increase the thickness of polysilicon as the gate electrode to the level corresponding to the height of the substrate projection. When the polysilicon is formed at a thickness corresponding to the height of the substrate projection, the aspect ratio becomes so large as to make it impossible to etch polysilicon by means of RIE.

[0121] The actual thickness of polysilicon that can be deposited is at most 200 nm, and a step of about 800 nm is created between the top of the substrate projection and the top of the gate electrode. For example, to form a contact hole as small as 150×150 nm, it is required to form a hole of a very large aspect ratio (5.3 plus an aspect ratio of the insulating film 12 ) by means of RIE because the step is large. At present this is very difficult in view of the capabilities of lithographic and RIE techniques.

[0122] The fifth embodiment is therefore intended to provide a three-dimensional MISFET in which the gate electrode is shaped as shown in FIG. 14 and the difference in level between the source and drain contact areas on the substrate projection and the gate contact areas is less than 200 nm.

[0123] FIG. 14 is a perspective view of a semiconductor device of the fifth embodiment.

[0124] First, portions of the p-type silicon substrate 11 around the region which is to form a projection are etched to form the substrate projection of about 2 μm in width. The direction of the width of the projection is parallel to the surface of the substrate 11 and normal to the direction from source to drain (i.e., the direction of the channel length). Next, an insulating film of, for example, silicon nitride is deposited which is to be used as a mask in forming the gate electrode. The silicon nitride film is etched using lithographic and RIE techniques to form a groove in that film.

[0125] The semiconductor substrate 11 is subjected to oxidation to form an oxide film of 50-100 nm in thickness, in which case the oxide film comes to have a region which resembles a bird's beak as is well known in the LOCOS process. After that, by selectively removing the oxide film, a substrate projection 51 A is formed which is thick in the source and drain portions and thin in the channel portion and part of the diffusion portion near the channel.

[0126] After that, the gate insulating film 13 is formed on the top surface and the side surfaces of that portion of the projection 51 A within the gate electrode forming mask. A polysilicon film is buried in the gate electrode forming mask and the excess polysilicon film is polished by CMP to form the gate electrode 54 .

[0127] When the thickness of the polysilicon film as the gate electrode 54 is first made relatively large, the gate electrode 54 can be formed to such a height as shown in FIG. 14 . Thus, the difference in level between the contacts on the gate electrode 54 and the contacts on the source and drain diffusion layers 15 and 16 can be reduced.

[0128] The formation of a fine gate pattern on the device isolation insulating film 12 is relatively easy even when the level difference between the substrate projection 51 A and that insulating film 12 is large. This is because, unlike the formation of a contact hole, in the formation of a gate pattern there is space in the direction of the gate width.

[0129] By the structure of the semiconductor device shown in FIG. 14 and the manufacturing method, the aspect ratio of the contact holes can be reduced and the contact holes can be formed simultaneously on the gate electrode 54 and the source and drain diffused layers 15 and 16 . Also, the parasitic resistance associated with the gate electrode 54 can be reduced because the thickness of the polysilicon film is large.

[0130] Sixth Embodiment

[0131] The sixth embodiment provides a semiconductor device in which the substrate projection is formed by epitaxial growth and the formation of contacts to the source and drain diffused layers is easy.

[0132] FIGS. 15A to 15 D are sectional views, in the order of steps of manufacture, of the semiconductor device of the sixth embodiment. FIGS. 16A to 16 C are plan views, in the order of steps of manufacture, of the semiconductor device.

[0133] First, as shown in FIG. 15 A, an insulating film 62 is formed on a semiconductor substrate 61 . By means of lithographic and dry etching techniques, the insulating film 62 is patterned to form an opening 63 in the region where the channel of a MISFET is to be formed. The plan view of the structure at this stage is illustrated in FIG. 16A .

[0134] Subsequently, epitaxial growth of silicon is performed using the exposed silicon semiconductor substrate 61 as a seed. As a result, as shown in FIG. 15 B, an overgrown epitaxial layer 64 of silicon is formed over the insulating film 62 .

[0135] Next, the epitaxial layer 64 is patterned by means of lithographic and dry etching techniques to form a substrate projection 64 A where the source, the drain and the channel are to be formed. The plan view of the structure at this stage is illustrated in FIG. 16B . In this figure, the upper and lower portions where the source and drain diffused regions are to be formed are made larger in width than the central portion where the channel is to be formed.

[0136] Next, as shown in FIG. 15D, a gate insulating film 65 is formed on the top surface and side surfaces of the substrate projection 64 A. A gate electrode material is deposited on the gate insulating film 65 and then patterned by lithographic and dry etching techniques to form a gate electrode 66 . By self-aligned ion implantation using the gate electrode pattern as a mask, phosphorous (p) or arsenic (As) is introduced into the portions of the substrate projection 64 A except below the gate electrode to form the source and drain diffused layers 67 and 68 . The plan view of the structure at this point is illustrated in FIG. 16C . The perspective view of the semiconductor device thus fabricated is illustrated in FIG. 17 .

[0137] With the semiconductor device shown in FIG. 17 , insulation of the source and drain regions of a MISFET called the delta or fin type from the substrate can be obtained. As a result, even if the device-to-device spacing is reduced, there is little possibility of erroneous conduction between adjacent devices and hence the device isolation characteristic can be improved. In addition, the channel portion is formed of the epitaxial layer 64 grown from the semiconductor substrate 61 ; thus, the bias to the channel can be controlled from the side of the semiconductor substrate 61 .

[0138] As in the case of the structure shown in FIGS. 10 and 11 A to 11 D, the length in the direction perpendicular to the direction of the channel length of the portions of the substrate projection in which the source and drain regions are formed is made larger than the corresponding length of the central portion where the channel is formed. In other words, the length in the direction parallel to the surface of the substrate and perpendicular to the direction from source to drain of that portion of the substrate projection on which the gate electrode is located is made shorter than the length in the corresponding direction of those portions of the substrate projection where the source and drain regions are formed.

[0139] Thus, the same operation as SOI devices can be achieved with the thin channel portion and the resistivity of the source and drain portions can be reduced. In addition, the formation of contacts to the source and drain regions can be made easy.

[0140] With the semiconductor device of this embodiment, since the channel is electrically connected with the substrate, a bias can be applied to the substrate, allowing threshold voltage control to be performed and the substrate floating effect to be reduced. Moreover, the channel and the source and drain regions are formed from monocrystalline silicon, allowing the parasitic resistance to be reduced.

[0141] Although the present invention has been described in terms of n-channel MIS field effect transistors, p-channel MIS field effect transistors can also be fabricated simply by changing the conductivity type using suitable process conditions.

[0142] According to an aspect of the present invention, as described above, semiconductor devices can be provided which are allowed to operate as fully depleted devices even if the gate length is very short and can reduce heat resulting from the Joule effect and the substrate floating effect.

[0143] Additional advantages and modifications will readily occur to those skilled in the art. Therefore, the invention in its broader aspects is not limited to the specific details and representative embodiments shown and described herein. Accordingly, various modifications may be made without departing from the spirit or scope of the general inventive concept as defined by the appended claims and their equivalents.