METHODS FOR FABRICATION OF 3-DIMENSIONAL NOR MEMORY ARRAYS

Abstract

Carbon has many advantageous uses as a sacrificial material in the fabricating thin-film storage transistors, such as those organized as NOR memory strings. In one implementation, the carbon layers are replaced by heavily doped n-type polysilicon source and drain regions at a late step during device fabrication. As a result, many high temperature steps within the fabrication process may now be carried out without concern for thermal diffusion from the heavily doped polysilicon, thus allowing phosphorus to be used as the n-type dopant.

Claims

1. A process suitable for use in fabricating storage transistors of a NOR memory string above a planar surface of a semiconductor substrate, comprising: above the planar surface, repeatedly depositing, alternately and one over another, a multi-layer and an isolation dielectric layers, each multilayer comprising first and second carbon layers and an interlayer dielectric layer that is provided between the first and the second carbon layers; cutting a plurality of trenches into the multilayers and the isolation dielectric layers, thereby exposing the carbon layers, the interlayer dielectric layers and the isolation dielectric layers at the sidewalls of the trenches, each trench having (i) a depth that extends along a first direction that is substantially normal to the planar surface, (ii) a length that extends along a second direction that is substantially parallel to the planar surface, (iii) a width that extends along a third direction that is substantially orthogonal to both the depth and the length, the length of the trench being substantially greater than its width; etching the interlayer dielectric layer of each multi-layer exposed in the trenches to create a recess between the first carbon layer and the second carbon layer; filling each recess with a first semiconductor layer of a first conductivity type; filling the trenches with a first sacrificial material; cutting a plurality of via openings in the first sacrificial material of each trench and filling the via openings with a second sacrificial material; replacing the remainder of the first sacrificial material in the trenches by a charge-storage layer and a conductor; removing the second sacrificial material from the via openings; and through the via openings, in place of each carbon layer of each multi-layer, providing a second semiconductor layer of a second conductivity type.

2. The process of claim 1, wherein replacing the remainder of the first sacrificial material in the trenches by the charge storage layer and the conductor comprises: removing the remainder of the first sacrificial material create excavated portions in each trench, the excavated portions being separated from each other by the second sacrificial material filling the via openings; in each excavated portion in each trench, depositing conformally the charge-storage layer on the sidewalls of the excavated portion; and filling the excavated portion by the conductor.

3. The process of claim 1, wherein the charge-storage layer comprises a tunneling layer, a charge-trapping layer and a blocking layer.

4. The process of claim 1, wherein the charge-storage layer comprises one or more of: silicon oxide, zirconium oxide, one or more multi-layers each comprising silicon oxide and silicon nitride, aluminum oxide, and any combination thereof.

5. The process of claim 1, wherein providing a second semiconductor layer of a second conductivity type in place of each carbon layer of each multi-layer comprises: removing the carbon layers by an ashing step that converts the carbon layers into carbon oxide gases, which are then expelled through the via openings, thereby creating a space in place of the carbon layers; and depositing the second semiconductor layer into the space.

6. The process of claim 5. wherein the second semiconductor layer comprises a conformal polysilicon liner.

7. The process of claim 6, further comprising filling the remainder of the space by a second conductive material.

8. The process of claim 7, wherein the second conductive material comprises titanium nitride-lined tungsten.

9. The process of claim 1, further comprising, after providing the second semiconductor layer, lining each via opening by one or more dielectric materials.

10. process of claim 1, wherein the interlayer dielectric layer and the isolation dielectric layer are dielectric materials of different etch characteristics.

11. The process of claim 1, wherein the isolation dielectric layer comprises silicon oxycarbon.

12. The process of claim 1, wherein each via opening is oval and has a major axis that is greater than the width of the trench, such that cutting the via opening also removes a portion of the multiplayers of each side of the trench.

13. The process of claim 1, wherein the trenches are cut in multiple phases wherein, in each phase, a portion of the trenches are cut and filled with the first sacrificial material.

14. The process of claim 1, wherein the second semiconductor layer that is in place of the first carbon layer, the second semiconductor layer that is in place of the second carbon layer, the first semiconductor layer, the charge storage layer, and the conductor form a storage transistor of the NOR memory string.

15. A process suitable for use in fabricating storage transistors of a NOR memory string above a planar surface of a semiconductor substrate, comprising: above the planar surface, repeatedly depositing, alternately and one over another, a multi-layer and an isolation dielectric layers, each multilayer comprising first and second semiconductor layers, each of a first conductivity type, a carbon layer between the first and the second semiconductor layers; cutting a plurality of trenches into the multilayers and the isolation dielectric layers, thereby (1) exposing the carbon layers, the interlayer dielectric layers and the isolation dielectric layers at the sidewalls of the trenches, each trench having (i) a depth that extends along a first direction that is substantially normal to the planar surface, (ii) a length that extends along a second direction that is substantially parallel to the planar surface, (iii) a width that extends along a third direction that is substantially orthogonal to the depth and the length, the length of the trench being substantially greater than its width; and (2) dividing the multi-layers into a plurality of stacks of multi-layer strips, each stack being separated from an adjacent stack by the width of one of the trenches, with each multi-layer strip being the portion of a multi-layer between adjacent trenches; filling the trenches with a dielectric filler material; cutting a plurality of via openings in the dielectric filler material of each trench; depositing into each via opening a charge-storage layer and a conductor; cutting a plurality of shafts at predetermined locations into each stack of multi-layer strips, thereby exposing the carbon layer of each multi-layer strip in each stack to the shaft; through the shafts, in place of each carbon layer of each multi-layer strip, providing a third semiconductor layer of a second conductivity type.

16. The process of claim 15 wherein the multi-layer further comprises a first sacrificial layer in contact with the first semiconductor layer or the second semiconductor layer.

17. The process of claim 16, wherein the first sacrificial layer is replaced by a conductive material after the trenches are cut.

18. The process of claim 15, wherein the charge-storage layer comprises a tunneling layer, a charge-trapping layer and a blocking layer.

19. The process of claim 15, wherein the charge-storage layer comprises one or more of: silicon oxide, zirconium oxide, one or more multi-layers each comprising silicon oxide and silicon nitride, aluminum oxide, and any combination thereof.

20. The process of claim 15, wherein providing a third semiconductor layer of a second conductivity type in place of each carbon layer of each multi-layer strip comprises: removing each carbon layer by an ashing step that converts the carbon layers into carbon oxide gases, which are then expelled through the shafts, thereby creating a space in place of the carbon layers; and depositing the third semiconductor layer into the space.

21. The process of claim 20, wherein the third semiconductor layer comprises a conformal polysilicon liner.

22. The process of claim 21, further comprising filling the remainder of the space by a second dielectric filler material.

23. The process of claim 15, wherein the interlayer dielectric layer and the isolation dielectric layer are dielectric materials of different etch characteristics.

24. The process of claim 15, wherein the isolation dielectric layer comprises silicon oxycarbon.

25. The process of claim 15, wherein each via opening is oval and has a major axis that is greater than the width of the trench, such that cutting the via opening also removes a portion of the multiplayers of each side of the trench.

26. The process of claim 15, wherein the trenches are cut in multiple phases wherein, in each phase, a portion of the trenches are cut and filled with the first sacrificial material.

27. The process of claim 15, wherein the first semiconductor layer, the second semiconductor layer, the third semiconductor layer, the charge-storage layer and the conductor provide a bit line, a source line, a channel region, a charge-storage region and a gate electrode, respectively, of a storage transistor of the NOR memory string.

28. The process of claim 27, wherein the storage transistors of the NOR memory string share the source line and the bit line.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0023] FIGS. 1(a), 1(b), 1(c), 1(d), 1(e), 1(f), 1(g), 1(h), 1(i), 1(j), 1(k), 1(1), 1(m), 1(n) and 1(o) illustrate a first process for fabricating an HNOR memory string, according to one embodiment of the present invention.

[0024] FIGS. 2(a), 2(b), 2(c), 2(d), 2(e), 2(f), 2(g), 2(h), 2(i), 2(j), 2(k), 2(1), 2(m), 2(n) and 2(o) illustrate a first process for fabricating an HNOR memory string, according to one embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0025] The present invention provides significantly more advantageous processes for fabricating HNOR memory strings than previously disclosed. For example, FIGS. 1(a) tol(o) illustrate a first process for fabricating an HNOR memory string without the need for a custom-designed “4-color strata” reaction chamber, according to one embodiment of the present invention. To facilitate reference to the figures, a Cartesian coordinate reference frame is used, in which the Z-direction is normal to the planar surface of the semiconductor surface and the X-direction and the Y-directions are orthogonal to the Z-direction and to each other.

[0026] Initially, as shown in FIG. 1(a), memory structure 100 is formed by successive depositions of (i) isolation dielectric layer 101 and (ii) multilayer 104, which comprises (a) first carbon layer 102a, (b) dielectric layer 103, and (c) second carbon layer 102b. When isolation dielectric layer 101 and dielectric layer 103 are provided by compatible dielectric materials of different etch characteristics (e.g., both of the family of silicon oxides, such as SiOC and SiO.sub.2), the successive depositions of the thin films can be very effectively and easily handled in an existing 2-color reaction chamber. Carbon layers 102a and 102b each may be deposited, for example, by chemical vapor deposition at 650° C., at 1.0 Ton. Isolation dielectric layer 101 may be, for example, a 30-nm thick SiOC layer. Carbon layer 102a and 102b may be each, for example, 15 nm thick, dielectric layer 103 may be 40-nm thick silicon dioxide (SiO.sub.2) or SiOC. Of course, the thicknesses and other dimensions disclosed herein are provided merely for illustrative purposes. In any implementation, any suitable variations may be adopted. For example, the thin films of FIG. 1(a) may each have a thickness that varies ±10 nm from the respective disclosed thickness above.

[0027] Note that carbon layers 102a and 102b are each much thinner than the semiconductor layers (e.g., the n.sup.+ polysilicon layers) in the active layers disclosed in Related Application II. For example, the total thickness of the multilayer 104 and isolation dielectric layer 103 is approximately ⅔ the total thickness of the active layer and an adjacent isolation dielectric layer disclosed in Related Application II. The thinner carbon layers are possible because of their greater mechanical strength, as the carbon so deposited has a lattice structure in the cubic form, which provides it significantly greater mechanical strength than polysilicon. Because of the thinner carbon layers, significantly greater number of thin films may be provided in the memory structure without increasing the aspect ratios of the vertical etches. At this time, while the manufacturing processes of Provisional Applications II and III disclose memory structures having up to eight HNOR memory strings along the vertical direction, a process of the present invention using carbon layers 102a and 102b may support memory structures having 12 or more HNOR memory strings along the vertical direction.

[0028] Trenches 106 (e.g., 60 nm wide, 160 nm apart) are then cut in memory structure 100 using, for example, an anisotropic etch set up by a photo-lithographical patterning step. FIG. 1(b) shows resulting memory structure 100. Because of the thinner carbon layers, the aspect ratio of this anisotropic etch is reduced significantly form the corresponding etch of the active layers in Related Application II.

[0029] Thereafter, dielectric layer 103 is recessed (e.g., by 10 nm) using, for example, an isotropic etch, followed by a conformal deposition of an in situ-doped p.sup.− amorphous silicon or polysilicon layer. An anisotropic etch removes the amorphous silicon or polysilicon from the sidewalls of trenches 106 and the top of memory structure 100, leaving separated amorphous silicon or polysilicon layer 104 in the recesses of dielectric layer 103. Resulting memory structure 100 is shown in FIG. 1(c). Amorphous silicon or polysilicon layer 104 (e.g., 10 nm thick) are provided to subsequently form the channel regions. Trenches 106 are then filled by dielectric material 107 (e.g., SiO.sub.2). Excess dielectric material may be removed from the top of memory structure 100 using, for example, a chemical-mechanical polishing (CMP) step. Resulting memory structure 100 is shown in FIG. 1(d).

[0030] With the mechanical support from dielectric material 107 in the first set of trenches 106, second set of trenches 106, each also approximately 60 nm wide. are then cut using substantially the same technique as illustrated in conjunction with FIG. 1(b) above. Resulting memory structure 100 is shown in FIG. 1(e). Each of the second set of trenches 106 is cut substantially equidistant between an adjacent pair of the first set trenches 106, such that the resulting stacks are each approximately 40 nm wide. Dielectric layer 103 exposed in the second set of trenches 106 is then recessed using, for example, substantially the same technique as discussed above in conjunction with FIG. 1(b) and followed by a conformal deposition of an in situ-doped p.sup.− amorphous silicon or polysilicon layer using substantially the same technique as illustrated above in conjunction with FIG. 1(c). Likewise, an anisotropic etch removes the amorphous silicon or polysilicon from the sidewalls of trenches 106 and the top of memory structure 100, leaving separated amorphous silicon or polysilicon layer 104. Thereafter, the second set of trenches 106 are filled with dielectric material 107, substantially as illustrated above in conjunction with FIG. 1(d). Resulting memory structure 100 is shown in FIG. 1(f). As discussed above, the relatively thinner carbon layers 102a and 102b significantly reduced the aspect ratios of anisotropic etches that form the first and the second sets of trenches 106, such that it may be possible to perform all the trench and channel region formations, filling of dielectric material 107 steps and subsequent CMP steps only once rather than twice: first in the steps of FIGS. 1(e) to 1(f), and second in the corresponding steps of FIGS. 1(b) to 1(d).

[0031] Thereafter, as shown in FIG. 1(g), vias 108 are patterned and cut in dielectric material 107 using, for example, an anisotropic etch. Vias 108 may be oval in horizontal cross-section, having a major axis of 100 nm along the X-direction and a minor axis of 60 nm along the Y-direction. Vias 108 are filled at a later time by a dielectric material so as to isolate adjacent storage transistors that are to be formed in an HNOR memory string. FIGS. 1(g) to 1(o)—which illustrate the remainder of the fabrication process steps according to the embodiment of the present invention illustrated in conjunction with FIGS. 1(a) to 1(f)—each include three views. View (i) is a horizontal cross-section (i.e., in an X-Y plane) through carbon layer 102a, view (ii) is a horizontal cross-section through dielectric layer 103, and view (iii) is a vertical cross-section (i.e., in an X-Z plane) along line A-A′ of view (i).

[0032] Vias 108 are then filled with sacrificial amorphous silicon material 109, as shown in FIG. 1(h). Excess amorphous silicon on top of memory structure 100 may be removed using, for example, CMP. FIG. 1(h) illustrates resulting memory structure 100. The remainder of dielectric material 107 in trenches 106 is then removed using, for example, a wet etch step, as illustrated in FIG. 1(i).

[0033] Charge-storage layer 110 is then conformally deposited on the exposed sidewalls of trenches 106. In this embodiment, charge-storage (“OZNOA”) layer 110 may be a multi-layer that includes, for example, (i) a 1-nm thick tunneling dielectric layer (e.g., silicon oxide (SiO.sub.2)); (ii) a 6-nm thick charge-trapping multilayer that includes zirconium oxide (ZrO), silicon nitride (SiN) and SiO.sub.2; and (iii) a 3-nm thick blocking layer (e.g., A1.sub.2O.sub.3). (These dimensions are provided merely for illustrative purposes; any suitable thicknesses or dimensions may be used.) As known to those of ordinary skill in the art, each of these materials need not be the stoichiometric compounds. For example, the silicon nitride in the charge-trapping layer may be silicon-rich nitride. Thereafter, trenches 106 are filled by conductive layer 111, which may include successively deposited titanium nitride (TiN) liner and tungsten (W). The TiN liner may be formed using, for example, an atomic layer deposition (ALD) technique. In each of trenches 106, between adjacent vias 108 (now filled by sacrificial amorphous material 109), conductive layer 111 provides a vertical local word line that serves as gate electrode for each of the storage transistors that are vertically aligned in the same active stack. Excess deposited materials may be removed from the top of memory structure 100 by CMP. FIG. 1(j) illustrate resulting memory structure 100.

[0034] Sacrificial amorphous silicon material 109 is then removed from vias 108 using, for example, a wet etch step. FIG. 1(k) illustrates resulting memory structure 100. Carbon layers 102a and 102b may then be removed and evacuated as a carbon oxide gas in an ashing step (e.g., 275° C., dry oxygen (O.sub.2) ambient), leaving cavities 112. The ashing step has the advantage that it does not damage p.sup.− polysilicon layer 104 (i.e., the channel regions). In general, being amenable to removal by ashing is a significant advantage that carbon has over other sacrificial materials, as the input reactant gases (e.g., oxygen) are readily available and safe to handle, and the resulting and byproduct gases (e.g., CO.sub.2) can also be easily handled and disposed, unlike the reactants of common wet etches, for example, which may include toxic, hazardous, or noxious chemicals or gases. FIG. 1(l) illustrates resulting memory structure 100.

[0035] Thereafter, phosphorus-doped n+polysilicon liner 113 (e.g., 20 nm thick) is then conformally deposited in cavities 112, as illustrated in FIG. 1(m). Phosphorus-doped n.sup.+polysilicon liner 113 provides the common source line and the common bit line for storage transistors in the HNOR memory strings to be formed. An etch-back step using, for example, a wet etch may then be carried out to remove the portions of n.sup.+ polysilicon liner 113 outside of cavities 112 (e.g., from the vertical sidewalls of trenches 106), As phosphorus-doped n.sup.+polysilicon liner 113 are deposited at this late stage of the fabrication process, thermal diffusion into the channel regions (i.e., p.sup.− polysilicon layer 104) is avoided, as no further high temperature process steps are expected. Thus, phosphorus, which has less onerous handling requirements than other n-type dopants, such as arsenic, may be used. FIG. 1(m) illustrates resulting memory structure 100.

[0036] Cavities 112 are then filled by deposition of conductive layer 114 over n.sup.+ polysilicon liner 113. Conductive layer 114 which includes successively deposited titanium nitride (TiN) liner and tungsten (W). The TiN liner may be formed using, for example, an atomic layer deposition (ALD) technique. Excess deposited materials may be removed from the sidewalls of trenches 106 and from the top of memory structure 100 using, for example, an anisotropic etch, followed by CMP. Conductive layer 114 reduces the resistivity encountered by a signal imposed on the common source line or the common bit line conductive layer 114 contacts. FIG. 1(n) illustrates resulting memory structure 100.

[0037] Conductive layer 114 may then be recessed from vias 108 using, for example, any suitable etching step (e.g., a wet etch). Thereafter, SiOC liner layer 115 may be deposited to provide electrical isolation between adjacent storage transistors in the HNOR memory strings. The remainder of vias 108 may be filled with a dielectric material (dielectric material 116; e.g., SiO.sub.2). FIG. 1(o) illustrates resulting memory structure 100.

[0038] Alternative, FIGS. 2(a) to 2(n) show a second process for fabricating HNOR memory strings that simplifies formation of the channel regions, according to one embodiment of the present invention. In a previous disclosed process, the channel regions are formed by depositing p.sup.− polysilicon into recesses formed by etching a sacrificial dielectric material provided between the common source line and the common bit line laterally (e.g., along the Y-direction). The process of FIGS. 2(a) to 2(n) take advantage of the significant advantages of using carbon as a sacrificial material, as discussed above.

[0039] Initially, as shown in FIG. 2(a), memory structure 200 is formed by successive depositions of (i) isolation dielectric layer 201 and (ii) multilayer 204, which includes (a) first silicon nitride layer 202a, (b) first phosphorus-doped amorphous silicon layer 203a, (c) carbon layer 204, (d) second phosphorus-doped amorphous silicon layer 203b, and (e) second silicon nitride layer 202b. Isolation dielectric layer 201 may be, for example, a 30-nm thick SiOC layer. First and second silicon nitride layers 202a and 202b each may be 30 nm thick, and first and second phosphorus-doped amorphous silicon layers 203a and 203b each may be 40 nm thick and carbon layer 204 may be, for example, also 40 nm thick. (These dimensions are provided merely for illustrative purposes; any suitable thicknesses or dimensions may be used.) Carbon layer 204 may be deposited, for example, by chemical vapor deposition at 650° C., at 1.0 Ton. FIG. 2(a) shows memory structure 200 after the depositions of the initial layers of thin films. Multi-layer 204 is also referred in this detailed description as an “active layer.”

[0040] Trenches 206 (e.g., 60 nm wide, 160 nm apart) are then formed in memory structure 200 using, for example, an anisotropic etch after a photo-lithographical patterning step. FIG. 2(b) shows resulting memory structure 100. Trenches 206 are then filled by dielectric material 207 (e.g., SiO.sub.2). Excess dielectric material may be removed from the top of memory structure 100 using, for example, CMP. Resulting memory structure 100 is shown in FIG. 2(c).

[0041] With the mechanical support from dielectric material 207, second set of trenches 206, each also approximately 60 nm wide, are then cut using substantially the same technique as discussed in conjunction with FIG. 2(b) above. Resulting memory structure 100 is shown in FIG. 2(d). Each of the second set of trenches 206 is cut between an adjacent pair of the first set of trenches 106, such that the resulting stacks—also referred to herein as “active stacks”—are each approximately 40 nm wide. The resulting narrow strips resulting from the active layers are also referred herein as “active strips.”

[0042] Thereafter, as shown in FIG. 2(e), first and second silicon nitride layers 202a and 202b are then removed using, for example, a wet etch, thereby creating cavities 215a and 215b. FIG. 2(e) shows resulting memory structure 200. Cavities 215a and 215b are filled by conductive layers 211a and 211b, respectively. Conductive layers 211a and 211b are each formed by successively deposited titanium nitride (TiN) liner and tungsten (W). The TiN liner may be formed using, for example, an atomic layer deposition (ALD) technique. An etching step removes the deposited material from the sidewalls of trenches 206 using, for example, a wet etch. This removal step also recesses conductive layer 211a and 211b to create recesses 216 by a small suitable distance (e.g., by 5 nm). Excess deposited materials may be removed from the top of memory structure 200 by CMP. FIG. 2(f) illustrates resulting memory structure 200.

[0043] First and second phosphorus-doped amorphous silicon layer 203a and 203b provide the common source line and the common bit line of an HNOR memory string to be formed. Conductive layers 211a and 211b reduce the resistivity of their adjacent common source line or common bit line. Polysilicon liner layer 217 is then deposited into recesses 216, followed by an etch-back step that removes excess polysilicon liner material from the sidewalls of second set of trenches 206. FIG. 2(g) shows resulting memory structure 200. Dielectric material 207 then fills second set of trenches 206. FIG. 2(h) shows resulting memory structure 200. Polysilicon liner layer 217 prevents conductive layers 211a and 211b from oxidation as being in contact with dielectric material 207.

[0044] Vias 208 are then patterned and cut in dielectric material 207 using, for example, an anisotropic etch. As the anisotropic etch is of a high aspect ratio, hard mask 218 is used, which provides extraordinary mechanical support. FIG. 2(i) shows resulting memory structure 200. FIG. 2(i), as well as FIGS. 2(j) and 2(k) in the following, each include three views. View (i) is a horizontal cross-section (i.e., in an X-Y plane) through carbon layer 204, view (ii) is a horizontal cross-section through first phosphorus-doped amorphous silicon layer 203b, and view (iii) is a vertical cross-section (i.e., in an X-Z plane) along line A-A′ of view (i). Vias 208 may each have an oval X-Y horizonal cross-section with suitable dimensions (e.g., substantially the same as vias 108 discussed above).

[0045] Because vias 208 exposes conductive layers 211a and 211b, to protect conductive layers 211a and 211b, the recess etch and deposition of polysilicon liner steps, described above with respect to FIGS. 2(f) and 2(g), are repeated. FIG. 2(j) shows resulting memory structure 200.

[0046] Charge-storage layer 210 is then conformally deposited on the exposed sidewalls of vias 208. In this embodiment, charge-storage (“OZNOA”) layer 210 may be a multi-layer including (i) a tunneling dielectric layer (e.g., 1 nm-thick silicon oxide (SiO.sub.2)); (ii) a 6-nm thick charge-trapping multilayer that includes zirconium oxide (ZrO), silicon nitride (SiN) and SiO.sub.2; and (iii) a 3-nm thick blocking layer (e.g., A1.sub.2O.sub.3). (These dimensions are provided merely for illustrative purposes; any suitable thicknesses or dimensions may be used.) As known to those of ordinary skill in the art, each of these materials need not be the stoichiometric compounds. For example, the silicon nitride in the charge-trapping layer may be silicon-rich nitride. Thereafter, vias 208 are filled by conductive layer 219, which includes successively deposited titanium nitride (TiN) liner and tungsten (W). The TiN liner may be formed using, for example, an atomic layer deposition (ALD) technique. Excess deposited materials may be removed from the top of memory structure 200 by CMP. FIG. 2(j) illustrates resulting memory structure 200.

[0047] In this embodiment, the channel regions for the storage transistors in a HNOR memory string is provided by depositing a p.sup.− polysilicon liner in a cavity created by removal of carbon layer 204. Removal of carbon layer 204 is accomplished by gaseous evacuation through vias or shafts to be cut in each active stack, as disclosed next

[0048] FIGS. 2(1) to 2(o) illustrate formation of the channel regions following removal of sacrificial carbon layer 204, according to one embodiment of the present invention. FIGS. 2(1) to 2(n) each include 3 views. View (i) is an X-Y cross-section through carbon layer 204 of an active strip, showing locations of shafts 220 cut in the active stack to facilitate gaseous removal of carbon layer 204. View (ii) shows the same X-Y cross-section through carbon layer 204 as in view (i), but at a higher resolution. View (iii) is an X-Z cross-section, taken in a plane through line B-B′ of view (i), which runs along the X-direction through two of shafts 220. FIG. 2(1) shows memory structure 200 after shafts 220 are cut using, for example, an anisotropic etch. As shown in view (i) of FIG. 2(k), shafts 220 are cut in each active stack at 3-pm intervals along the Y-direction. Providing shafts 220 at any suitable interval may also be used.

[0049] Carbon layer 204 may then be removed in an ashing step as gaseous carbon oxide, using substantially the same the same ashing step technique as illustrated above with respect to FIG. 1(e) (e.g., 275° C., dry oxygen (O.sub.2) ambient), leaving cavities 222. As described above, being amenable to removal by ashing is a significant advantage that carbon has over other sacrificial materials, as the input reactant gases (e.g., oxygen) are readily available, and the resulting and byproduct gases (e.g., CO.sub.2) are easily handled and disposed, unlike the reactants of common wet etches, for example, which may include toxic, hazardous, or noxious chemicals or gases that require additional treatment. FIG. 2(m) illustrates resulting memory structure 200.

[0050] P.sup.− polysilicon liner layer 227 is then deposited into cavities 222 (e.g., 20 nm thick), followed by an etch-back step that removes excess polysilicon from the sidewalls of shafts 220. FIG. 2(n) shows resulting memory structure 200. Along the X-direction, p.sup.− polysilicon liner layer 227 contacts both charge storage layer 210, deposited previously on the sidewalls of vias 208, and dielectric material 207, deposited previously on the sidewalls of trenches 206. A storage transistor is formed by the common source line (i.e., second phosphorus-doped amorphous silicon layer 203b), the common bit line (i.e., first phosphorus-doped amorphous silicon layer 203a), a channel region (i.e., p.sup.− polysilicon liner layer 227), the portion of charge storage layer 210 in contact with the channel region, and the gate electrode enclosed by that charge storage layer 210.

[0051] Dielectric material 228 (e.g., SiO.sub.2) then shafts 220. Excess material on top of memory structure 200 may be removed by CMP. FIG. 2(n) shows resulting memory structure 200. In FIG. 2(o) shows that, due to the length of cavities 222 along the Y-direction, dielectric material 228 may not completely fill cavities 222, and thus may leave voids 230.

[0052] The above detailed description is provided to illustrate specific embodiments of the present invention and is not intended to be limiting. Numerous modifications and variations within the scope of the present invention are possible. The present invention is set forth in the accompanying claims.