The invention relates to a method for manufacturing a non-volatile memory array comprising at least one array of memory cells on a substrate of a semiconductor material, according to the preamble of the first claim.
Such non-volatile memory structures are e.g. known in the art as flash memory devices, whereby the memory function is implemented by storing electrical charges on memory cells known as floating gates. Two well known types of such flash memory devices are NAND flash and NOR flash. There is a constant pressure to make cheaper memory devices with a higher storage capacity, which is a.o. addressed by constant miniaturization, different memory topologies, better process and yield control. Particularly relevant for the present invention are memory cell structures with a floating gate having an inverted T shape.
US20060292802 describes a NAND flash memory device and a method for forming the same. The memory structure comprises elements with an inverted T shape, fabricated from box-shaped parts, the sides of which are narrowed by etching. A disadvantage of the method used in US20060292802 is that the process is complex and extra process steps are needed to prevent etching damage.
US20080074920 describes a 2-dimensional non-volatile memory array, wherein each memory cell comprises a floating gate, the floating gate having an inverted T shape in cross section. Methods are described to form these inverted T shapes, whereby a first floating gate layer is deposited on top of a dielectric layer, to form the lower (horizontal) side of the inverted T shape, and a second floating gate layer, forming the upright (vertical) part of the floating gate, is deposited on top of the first floating gate layer. In a first method, the upright part of the floating gate is formed by filling slots between sidewall spacers formed against masking portions deposited on top of the first floating gate layer. In a second method, the upright part of the floating gate is formed by filling slots between sidewall spacers formed against Shallow Trench Isolation (STI) structures, extending above the active areas. In a third method masking portions on a floating gate layer establish a pattern for partially etching the floating gate layer thereby forming the upright (vertical) part. The outer surface of the etched floating gate layer is then oxidized, and then the oxide layer is used as sidewall spacers to form STI trenches to define the individual memory cells. A disadvantage of the methods described in US20080074920 is that the methods to form inverted T shapes are complex, and require many process steps.
It is an aim of the present invention to present a less complex method to form a non-volatile memory structure having floating gates with an inverted T-shape on a substrate.
This aim is achieved according to the invention by a method comprising the steps of claim 1.
Thereto, the method according to the present invention comprises the following steps:
a) providing a silicon substrate,
b) providing an array of memory cells, each memory cell comprising a stack of a tunnel oxide layer on top of the substrate, and a floating gate layer on top of the tunnel oxide layer, the memory cells being self-aligned to and separated from each other by Shallow Trench Isolation (STI) structures extending into and above the substrate in height direction,
d) thinning an upper part of the floating gate layer by removing a predetermined thickness of its sidewalls, thereby creating floating gates with an inverted T shape in a cross section along the array of memory cell,
e) forming an interpoly dielectric layer on top of, and on the sides of the floating gates,
f) forming a control gate on top of the interpoly dielectric layer, the interpoly dielectric layer isolating the control gate from the floating gates,
wherein the method additionally comprises a step c) preceding step d), wherein an upper part of the sidewalls of the floating gates is exposed by partially etching back the STI structures to a height position above the tunnel oxide layer, above which height position the floating gates are to be thinned,
and wherein the step d) comprises the following steps:
i) oxidation of the exposed parts of the floating gates, thereby forming a sacrificial oxide,
j) removing the sacrificial oxide.
The inventor has discovered that a floating gate layer with an inverted T shape can be produced in a more efficient way by starting from a structure whereby the gaps between Shallow Trench Isolation (STI) structures, formed on and into a substrate, are filled with a tunnel oxide layer and a floating gate layer, then exposing an upper part of the floating gate layer by partially removing the STI structures to a first desired height, then oxidizing the exposed parts of the floating gate layer, and removing the sacrificial oxide, thereby effectively thinning the upper part of the floating gate layer.
It is an advantage of the method of the present invention that a floating gate with an inverted T shape can be made of a single piece of material, instead of a combination of a lower part and an upper part that are made in different process steps. A floating gate made of a single piece of material has better electrical and mechanical properties, by avoiding unwanted impurities between the lower and upper part of the floating gate.
By providing the floating gate material in a self-aligned way to the STI structures, a memory cell is obtained whereby the tunnel oxide and the floating gate layer are perfectly aligned to the STI structure, and whereby the STI structure can be used as a mask to deposit the tunnel oxide and the floating gate material, thus saving process steps.
By controlled oxidation of the exposed parts of the floating gate layer, the width of the upper part of the floating gate can be accurately determined by the process parameters, especially the oxidation time. An advantage of thinning by oxidation rather than by etching is that no etching damage can occur, and thus no extra steps such as using protection layers, are required to prevent etching damage. Another advantage is that oxidation of the sidewalls occurs in any direction, thus a small deviation of process parameters only has a small impact on the thickness of the upper part of the floating gate, as opposed to etching which is usually done in a direction parallel to the sidewalls, whereby thinning of the sidewalls is more difficult to control. Another advantage of oxidation over etching is that doping atoms such as Phosphor, which were present in the floating gate before the oxidation, will migrate into the floating gate material, thereby increasing the doping level and thus the conductivity of the floating gate material.
By controlling the initial height and width of the floating gate layer in step b), and the predetermined thickness of the sidewalls removed in step d), the dimensions of the inverted T shape can be accurately controlled. In addition, most of these dimensions are independent of the lithographic process used, and can thus be very accurately fine-tuned. This is particularly important for the thickness of the inverted T shape, which is the most important parameter.
In an embodiment of the method according to the present invention, in step b) optionally a hardmask is placed on top of the floating gate layer, self-aligned with the STI structures, to prevent thinning of the floating gate layer in height direction.
By providing a hardmask on top of the floating gate layer, reduction of the floating gates height can be avoided. A trade-off can be made between a method with or without using a hardmask. The method with a hardmask requiring more process steps to place and remove the hardmask, but requiring less time and material to deposit floating gate material. A hardmask can be applied in order to preserve the original height of the floating gate when thinning the upper part. Additionally a hardmask can help in controlling corner rounding effects occurring at the upper corners of the upper part of the floating gate.
In a preferred embodiment of the method according the invention, the floating gate layer is made of silicon and the STI structures are made of silicon oxide. This way the sacrificial oxide can be removed in the same step as further etching back of the STI structure.
In an alternative embodiment of the method according the invention, in step j) simultaneously with the removal of the sacrificial oxide, the STI structures are further removed to a second height above the tunnel oxide layer, to allow replacement of the further removed STI structure by interpoly dielectric. An Interpoly dielectric typically has better material properties, such as less leakage and less defects, thus providing a non-volatile memory with better data retention.
In an embodiment of the method according to the present invention, step d) optionally comprises a step k) following the step j), wherein the exposed parts of the floating gates are oxidized a second time to form a bottom layer of the interpoly dielectric layer between the floating gates and the control gate. The interpoly dielectric can be made of a single layer or multiple layers. Using oxidized floating gate material instead of deposited oxide as the bottom layer of the interpoly dielectric, results in a better quality of the interpoly dielectric and better cohesion of the interpoly dielectric layer to the floating gate layer.
The invention will be further elucidated by means of the following description and the appended figures.
a-b shows the control gate coupling ratio CR as function of the half pitch Wpitch/2 for various values of the height tFIN of the floating gate for (a) a 4 nm SiO2/4 nm Si3Ni4/4 nm SiO2 ONO stack and (b) a 4 nm SiO2/6 nm Al2O3.
The present invention will be described with respect to particular embodiments and with reference to certain drawings but the invention is not limited thereto but only by the claims. The drawings described are only schematic and are non-limiting. In the drawings, the size of some of the elements may be exaggerated and not drawn on scale for illustrative purposes. The dimensions and the relative dimensions do not necessarily correspond to actual reductions to practice of the invention.
Furthermore, the terms first, second, third and the like in the description and in the claims, are used for distinguishing between similar elements and not necessarily for describing a sequential or chronological order. The terms are interchangeable under appropriate circumstances and the embodiments of the invention can operate in other sequences than described or illustrated herein.
Moreover, the terms top, bottom, over, under and the like in the description and the claims are used for descriptive purposes and not necessarily for describing relative positions. The terms so used are interchangeable under appropriate circumstances and the embodiments of the invention described herein can operate in other orientations than described or illustrated herein.
The term “comprising”, used in the claims, should not be interpreted as being restricted to the means listed thereafter; it does not exclude other elements or steps. It needs to be interpreted as specifying the presence of the stated features, integers, steps or components as referred to, but does not preclude the presence or addition of one or more other features, integers, steps or components, or groups thereof. Thus, the scope of the expression “a device comprising means A and B” should not be limited to devices consisting only of components A and B. It means that with respect to the present invention, the only relevant components of the device are A and B.
The invention relates to a method for manufacturing a non-volatile memory comprising at least one array of memory cells on a substrate 1 of a semiconductor material, according to the preamble of the first claim.
In the drawings only a small number of memory cells are shown, for clarity of the explanation.
The method according the invention describes an improved way for manufacturing a memory device with floating gates 7 having an inverted T shape. A preferred embodiment of the method according to the present invention will be described, and variants thereof.
a shows the structure of
Optionally (
It is preferable that the tunnel oxide 3 is made of silicon oxide through thermal oxidation process, but a insulating layer with high dielectric constant may be used also. The tunnel oxide can be also a stack of different insulating layers, e.g. a combination of lower-k and higher-k dielectrics to control the distribution of the electrical field over the tunnel oxide during tunnelling of carriers to and from the floating gate. The thickness of the tunnel oxide layer 3 is typically 2-20 nm. It is preferable that the STI structures 2 are made of silicon oxide. According to the invention, the floating gate 7 is made of a conducting or semi-conducting material that can be oxidised, e.g. polysilicon, and the sacrificial oxide layer 9 is silicon oxide. The IPD layer can e.g. by deposited by means of Atomic Layer Chemical Vapour Deposition (ALCVD). The interpoly dielectric (IPD) layer 5 can consist of a single isolating layer, e.g. a silicon oxide layer having a thickness of typically 2-20 nm, but an insulating layer with high dielectric constant may also be used. The interpoly dielectric can be also a stack of different insulating layers, e.g. a combination of lower-k and higher-k dielectrics to control the distribution of the electrical field over the tunnel oxide to prevent tunnelling of carriers to and from the control gate. The IPD layer can also consist of a combination of multiple layers, for example ONO, which consists of 3 layers: a silicon oxide (SiO2) layer having a thickness of typically 3-8 nm, a silicon nitride (Si3N4) layer having a thickness of typically 5-15 nm, and a silicon oxide (SiO2) layer having a thickness of typically 3-10 nm. Another example of an IPD layer made of multiple layers is the combination of a silicon oxide layer having a thickness of 2-10 nm e.g. 5 nm, and an aluminum oxide layer (Al2O3) having a thickness of 2-10 nm, e.g. 7 nm. IPD layers consisting of a combination of multiple layers are described in EP1605517, EP1903602, EP1748472, which are hereby incorporated by reference. An IPD consisting of multiple layers provides a better isolation and less leakage than a single layer. The control gate layer 6 is made of a conducting material, e.g. polysilicon, a silicide, a metallic material, or a combination thereof. The control gate 6 can e.g. be deposited by CVD such as ALCVD, LPCVD.
The dimensions (W1, W2, H2, k1) of the resulting floating gate 4 can be determined by choosing proper dimensions of the starting structure (W1, H1), and process parameters according to the invention. For example, the height H1 of the floating gate layer 4 of
The material of the hardmask layer 10 is typically silicon nitride (Si3N4) and is applied on top of the floating gate layer 4 using conventional techniques known in the art. To obtain floating gates 7 with the same dimensions as those shown in
Optionally (not shown) in the same step as removing the hardmask 10 and the sacrificial oxide 9, the STI structure 2 can be removed further to a height k2b above the tunnel oxide 3, e.g. by increasing the etching time, but again without exposing the tunnel oxide layer 3. This allows replacement of the further removed STI structure 2 by interpoly dielectric 5 having better material properties, such as less leakage and less defects, thus providing a non-volatile memory with better data retention.
In a variant of the second method, the hardmask is not or not entirely removed, for tuning of the corner rounding at the top of the floating gates 7. This way field strengthening at the upper corners of the fin can be counteracted.
As described above, optionally also in this third method the predetermined height of the STI structure 2 can be reduced further.
A fourth method according to the present invention is a variant of the third method, whereby a hardmask 10 is used on top of the floating gate layer 4, to prevent thinning of the floating gate layer 4 in height direction during the first oxidation step similar to the second method illustrated in Example 2. Then the hardmask 10 and the sacrificial oxide 9 are removed in a similar way as described in the second method. Then a second oxidation is performed without using the hardmask 10, whereby the second oxide is used as a bottom layer of the Interpoly dielectric oxide 11.
The oxidation of the exposed parts as described in the methods above can be done using e.g. ISSG oxide (in-situ steam generated oxidation) or wet oxidation. The oxidation parameters determine the thickness of the oxide formed on the exposed parts and hence the reduction in width W1-W2 obtained.
Table 1 below lists different ISSG recipes for forming a predetermined silicon oxide on an undoped silicon layer. If the silicon layer is doped, the oxidation rate increases, and for the given recipe a thicker oxide is formed. If e.g. the silicon layer is doped with 5e20 cm-3 P, then typically 40% more oxide is formed. Depending on the technology generation, 5 to 15 nm of the floating gate needs to be oxidized in order to obtain the desired width reduction to form the inverted T-shape.
Table 2 below lists different wet oxidation recipes for forming a predetermined silicon oxide on an undoped silicon layer. If the silicon layer is doped, the oxidation rate increases, and for the given recipe a thicker oxide is formed. If e.g. the silicon layer is doped with 5e20 cm-3 P, then typically 40% more oxide is formed. Depending on the technology generation, 5 to 15 nm of the floating gate needs to be oxidized in order to obtain the desired width reduction to form the inverted T-shape.
The dimensions of the inverted T-shape floating gate can be determined as followed, whereby in
The coupling ratio CR of the control gate 4 to the channel 13 can be expressed as:
CR=CIPD/(CIPD+CTD) (1)
with CR being the coupling ratio, CIPD being the capacity between the control gate 6 and the floating gate 4, CTD being the capacity between the floating gate 4 and the channel 13. Typically the coupling ratio CR is equal to or larger than 0.6. Hence the ratio CIPD/CTD is equal to or greater than 1.5 or CIPD is equal to or greater than 1.5*CTD.
In a first order approximation one can define the capacitances as follows
CTD=W1*L*epsilonox/EOTTD (2)
CIPD=(W2+2·tFIN)*L*epsilonox/EOTIPD (3)
with W1 the width of the memory cell, L the length of the memory cell, W2 is the width of the fin, i.e. the upper part of the floating gate, tFIN=H2−k1 is the height of the fin, EOTTD is the equivalent electrical oxide thickness of the tunnel dielectric 3, EOTIPD is the equivalent electrical oxide thickness of the IPD dielectric 5 which is typically in the range of 7-9 nm, epsilonox is the relative dielectric constant of SiO2(epsilonox=3.9*epsilon0 with epsilon0 being the absolute dielectric constant of vacuum). It is assumed that silicon oxide is used a dielectric material to form the tunnel dielectric 3 and the interpoly dielectric 5. As stated above these two dielectrics 3 and 5 can be formed as a stack of layers of materials with different dielectric properties in which case of course the combination of thickness and dielectric constant of the respective layers is to be used in formulas (2) and (3). The minimal values of W1 and L depend on the technology generation. In case of a 32 nm technology, W1=L=32 nm as 32 nm is the minimal feature size targeted at in this technology.
Apart from having the desired coupling ratio CR, the memory cell must of course fit in the pitch along the W direction X. One can define the following ratio:
W2+2*TIPD+TCG=Wpitch (4)
with Wpitch being the pitch in the W direction, typically Wpitch=2*W1=2*feature size of the technology generation, TIPD being the physical thickness of the IPD dielectric 5, TCG being the minimal physical thickness of the control gate. In case of a 15 nm technology generation, Wpitch=15*2=30 nm, W2=TCG=5 nm and TIPD=10 nm as formed by a stack of 4 nmSiO2/6 nmAl2O3. More relaxed dimensions are Wpitch=20 nm*2=40 nm, W2=TCG=6 nm, TIPD=14 nm.
When determining the dimensions of the inverted T-shape floating gate memory cell, other electrical, physical and process considerations must be taken into account:
The width W2 of the fin is limited by conductivity as a too small fin can suffer from complete depletion, by wafer uniformity and by mechanical stability as a too small fin might collapse.
When etching the control gate 6, the complete stack of control gate 6, interpoly dielectric 5, floating gate 7 and tunnel dielectric 3 must be etched. Hence the thickness H2 of the floating gate 7 determines the thickness of the control gate 5 to be etched as the control gate 5 envelopes the floating gate 7 and fills the gaps in between adjacent control gates 7 as illustrated in
The thickness TIPD and composition of the interpoly dielectric 5 is selected in order to reduce leakage between the control gate 6 and the floating gate 7. An example of such an interpoly dielectric 5 is a stack of SiO2 AND Al2O3 whereby the thickness of SiO2 is about 4 to 5 nm while the thickness of the Al2O3 is about 6 to 10 nm resulting in a physical thickness is about 11 to 15 nm
For a more accurate determination of the coupling ratio CR the 3D structure of the memory cell is to be taking into account. One can use simulation data as represented in
Number | Date | Country | Kind |
---|---|---|---|
09167749 | Aug 2009 | EP | regional |
Number | Name | Date | Kind |
---|---|---|---|
7585730 | Song et al. | Sep 2009 | B1 |
20060292802 | Lee et al. | Dec 2006 | A1 |
20080074920 | Chien et al. | Mar 2008 | A1 |
Number | Date | Country |
---|---|---|
1605517 | Dec 2005 | EP |
1748472 | Jan 2007 | EP |
1903602 | Mar 2008 | EP |
Number | Date | Country | |
---|---|---|---|
20110039380 A1 | Feb 2011 | US |