The invention relates to a method of forming a charge-trapping memory device, particularly an NROM or SONOS memory device. Such a device comprises a memory layer sequence of dielectric materials provided for programming by means of trapping charge carriers.
Non-volatile memory has become increasingly important for mass storage media, cards and code flash products. Non-volatile memory cells that are electrically programmable and erasable can be realized as charge-trapping memory cells, which comprise a memory layer sequence of dielectric materials with a memory layer confined between confinement layers of dielectric material, the latter having a larger energy band gap than the dielectric material of the memory layer.
The memory layer sequence is arranged between a channel region within a semiconductor body and a gate electrode of a transistor, the gate electrode being provided to control the channel by means of an applied electric voltage. Examples of charge-trapping memory cells are the SONOS memory cells, in which each confinement layer is an oxide and the memory layer is a nitride of the semiconductor material, commonly silicon (see e.g., U.S. Pat. No. 5,768,192 and U.S. Pat. No. 6,011,725, which are incorporated herein by reference).
Applying a huge enough positive voltage to the control gate of a SONOS device will generate an electrical field condition for tunnelling of electrons from an inversion channel below the gate through the confinement layer. These electrons can then be trapped in the charge trapping layer. The trapped charge carriers change the threshold voltage of the cell transistor structure. Different programming states can be read by applying the appropriate reading voltages.
According to a planar memory cell concept as shown, e.g., in Eitan, et al., “Can NROM, a 2-bit, trapping Storage NVM cell, give a real challenge to floating gate cells?”, Proc. Solid State Devices and Materials, pp. 1-3, Tokyo 1999, which is incorporated herein by reference, charge-trapping memory devices can be provided with, e.g., buried bit lines, which electrically connect the source/drain regions of the memory cells. A memory layer sequence, typically an oxide-nitride-oxide layer sequence, is located above the channel regions between corresponding source/drain regions. Above the buried bit lines, a thick oxide is formed as an electrically insulating region, especially to separate the gate electrode sufficiently from the buried bit lines. The memory layer sequence is confined to the channel region and overlaps the source/drain regions only slightly.
This fundamentally new concept called “NROM” provided by Eitan, et al. has the following important features:
(a) the programming of memory cells by means of hot channel electrons is combined with a delete operation by means of hot holes;
(b) a duplicated storage location per memory cell, i.e., on both junction sides of a transistor channel region; and
(c) a reverse read operation, which allows a separate detection of the storage content on both sides of the transistor; the term “reverse” refers to the direction of the voltage drop with respect to the source and drain regions and indicates that this direction is reverse to that applied during a programming operation.
Recently, a vertical memory cell design has been developed (Willer, J., et al. in “UMEM: a U-shape Non-Volatile-Memory Cell”, Proc. IEEE NVSMW, Monterey, pp. 42-43, 2003), which is incorporated herein by reference. According to this design, the word line is buried in the substrate with, e.g., 60 nm deep trenches. Advantageously, the channel length is folded around the bottom of the word line trenches resulting in an effective channel length of, e.g., 100 . . . 150 nm. Diffusion regions are arranged adjacent to the trenches connecting neighboring memory cells, which are arranged perpendicular to the word line direction.
In this design the ONO-layer sequence is deposited on the trench sidewalls. Accordingly, the storage region within the ONO-layer sequence, or more precisely within the memory layer typically provided by a nitride layer, is adjacent to the diffusion regions in a top region of the two opposite sidewalls of the trench.
With decreasing cell sizes, trenches forming the word lines and thus the regions wherein hot charge carriers such as electrons are generated move closer with respect to each other. The mean travel distance of hot charge carriers within the substrate may amount to 70 nm. Consequently, with trench distances approaching the 100 nm limit, the probability of trapping charge carriers generated during programming of a neighboring memory cell in the memory layer of a currently passive cell exponentially increases with decreasing cell sizes. The problem is schematically depicted in
In one aspect, the present invention provides a method of forming charge-trapping memory devices comprising a memory layer sequence and buried word lines, which removes the problem of a charge carrier exchange between neighboring memory cells during a programming operation.
A first embodiment of the method comprises forming a charge-trapping memory device having an array of memory cells, including a semiconductor substrate with a main surface and trenches formed within the substrate having sidewalls. The sidewalls are doped to form U-shaped channel regions. A sequence of layers comprising dielectric materials is deposited on the sidewalls of the trenches. Each layer sequence of dielectric materials includes a bottom confinement layer, a memory layer and a further top confinement layer. The trenches are filled with a conductive material that overlies the layer sequence to form gate electrodes and word lines. The main surface adjacent to the trenches is doped to form diffusion regions, which serve as the source and/or drain of each of the memory cells. Further trenches parallel to the first trenches are formed within the substrate thereby subdividing the diffusion regions into each a first diffusion region adjacent to a first trench of a first charge-trapping memory cell and a second diffusion region adjacent to a first trench of a second charge-trapping memory cell. The further trenches are filled with dielectric material in order to impede hot charge carrier exchange between adjacent charge-trapping memory cells.
A second embodiment of the method comprises forming a charge-trapping memory device having an array of memory cells, including a semiconductor substrate with a main surface and trenches formed within the substrate having sidewalls. The sidewalls are doped to form channel regions. A sequence of layers comprising dielectric materials is deposited on the sidewalls of the trenches. Each layer sequence of dielectric materials includes a bottom confinement layer, a memory layer and a further top confinement layer. The trenches are filled with a conductive material that overlies the layer sequence to form gate electrodes and word lines. The main surface adjacent to the trenches is doped to form diffusion regions, which serve as a source and/or drain of each of the memory cells. Further trenches parallel to the first trenches are formed within the substrate by means of etching thereby subdividing the diffusion regions into each a first diffusion region adjacent to a first trench of a first charge-trapping memory cell and a second diffusion region adjacent to a first trench of a second charge-trapping memory cell. The further trenches are filled with dielectric material in order to impede hot charge carrier exchange between adjacent charge-trapping memory cells and planarizing. Bit line contacts are formed, which each coincidently contact the first and second diffusion regions subdivided by the forming of the further trench.
In other embodiments, a charge-trapping memory device includes an array of memory cells, which are formed in a semiconductor substrate. Each of the memory cells comprises a trench formed within the substrate. The trench has sidewalls, which are covered with a layer sequence of dielectric materials. The trench is filled with a conductive material to form a gate electrode and a word line. First and second diffusion regions are disposed adjacent to the trench sidewalls at a main surface of the substrate. The diffusion regions serve as a source and/or drain of the memory cell. A channel region is arranged within the substrate and has a U-shape for extending adjacent to the trench sidewalls from the first diffusion region to the second diffusion region. The layer sequence of dielectric materials includes a memory layer the is confined by confinement layers and provides a gate dielectric for the gate electrode and a hot charge carrier storage region adjacent to each of the diffusion regions. A further trench formed in the substrate is filled with a dielectric material. The further trench is arranged parallel to the trench, which forms a word line. The further trench serves to shield hot charge carrier exchange between adjacent charge-trapping memory cells.
According to embodiments of the invention, a further trench is provided, which is filled with dielectric material, which may be an oxide, or flowing glass. This further trench is arranged parallel to a buried word line according to the vertical NROM-cell concept and thus separates each two word lines from each other. Thereby, the further trench filled with the dielectric material only consumes a portion of the substrate between each two word lines. In other words, substrate portions on both sides of the further trenches serving as source/drain regions adjacent to the gate electrodes of the word lines are retained.
One goal of the further trench is not to separate the source/drain regions of neighboring memory cells, as they are—according to the virtual ground array concept—in any way connected to the same bit line contact. Rather, the further trenches have a depth, such that portions of the memory layer of neighboring memory cells facing each other are shielded by means of the dielectric material within the further trench. In particular, a space within the substrate adjacent to a diffusion region and to the charge-trapping layer of a first memory that is able to generate high energy charge carriers is shielded from the charge-trapping layer of a neighbouring memory cell. Consequently, accumulation of charges in other cells than those, which are actually programmed, is considerably reduced, if not impeded.
As the hot charge carrier generation space is typically located beneath a diffusion region near the substrate surface, the further trench that blocks hot charge carriers due to its dielectric material should have a depth amounting to about half of that of the depth of the buried word line in order to completely shield the charge-trapping layers from each other. For example, a buried word line having a depth of 60 nm needs a hot charge carrier blocking a trench with a depth of at least 30 nm as measured from the depth of the p/n-junction, i.e., the location where hot charge carriers are generated. Therein, it is supposed that the trench is arranged in the middle between each two word lines.
The minimum width of the trench is merely limited by a thickness needed to sufficiently absorb high energy charge carriers. Rather, the width is limited by lithographic capabilities. As minimum widths, which can be achieved by lithographic projection, will currently exceed 70 nm, a spacer technique narrowing the manufacturable trench width is advantageously employed to achieve trench widths down to 15-25 nm.
Advantageous aspects and embodiments are evident from the appended claims. The invention will become clearer with regard to certain embodiments when taken in conjunction with the accompanying drawings.
Preferred embodiments and examples of the method according to this invention will be further described in detail in connection with the appended drawings.
a-16c show the further evolution of the memory cells of
The following list of reference symbols can be used in conjunction with the figures:
In
After cleaning, the memory layer 84 confined by the bottom confinement layer 82 and the top confinement layer 86 are deposited upon the substrate 6. In particular, an oxide layer of, e.g., 5 nm is deposited first, followed by a nitride layer having a thickness of 7 nm and then by a further oxide layer (e.g., 11 nm), resulting in an ONO layer sequence. The memory or charge-trapping layer is the nitride layer 84. After CMP- or etch removal of those portions of the ONO-layer not residing within trench 14, i.e., on a plane surface of the substrate, the trench 14 is filled with n+-doped poly silicon as a conductive filling 10 to yield a gate electrode and a buried word line. A planarization step follows, which emerges to the situation shown in
The spacers 24 have a lateral width of about 20 nm each. The trench-to-trench (word lines) distance is about 55 nm. Consequently, the width of further trenches 26 amounts to about 15-20 nm.
It may also be noted, with regard to
b shows a top view of the arrangement shown in
a shows a cross-section in the X-direction (bit line direction), which is offset with respect to that shown in
In this case no protection due to the hard mask 46 is active and an anisotropic silicon substrate etch results in a removal of the diffusion regions 28. Preferably, 30 nm deep shallow trenches 50 are formed thereby—as measured from the depth of the p/n-junction (junction between n-diffusion zone of the source/drain region and p-well beneath the channel region). The depth is ideally, but not necessarily, similar to that of further trenches 26.
b shows a top view with respect to slice level A′-A′ as indicated in
a shows the situation along section C-C after a thin oxide liner and a HDP oxide 34′ are deposited to fill trenches 50. In this embodiment, trenches 26 and trenches 50 are filled with the same isolating material (HDP-oxide 34, 34′) and have the same depth.
The situation after removal of the hard mask 46 is depicted in
Further, dot-dashed lines indicate the position of local interconnects 52 to be formed subsequently, which realize a virtual ground array architecture. That is, the diffusion regions 28 are finally electrically reconnected to provide a common bit line contact to both diffusion regions, which have previously been disconnected due the etching of trenches 26.
Each local interconnect is further provided with a bit line contact 54 to connect to a bit line (not shown in
Although the present invention and its advantages have been described in detail, it should be understood that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims.
Number | Name | Date | Kind |
---|---|---|---|
5768192 | Eitan | Jun 1998 | A |
6011725 | Eitan | Jan 2000 | A |
7060574 | Kim et al. | Jun 2006 | B2 |
20040180496 | Hsiao et al. | Sep 2004 | A1 |
20050032308 | Hsiao et al. | Feb 2005 | A1 |
20050082600 | Hsu et al. | Apr 2005 | A1 |
20060108646 | Hofmann et al. | May 2006 | A1 |
Number | Date | Country |
---|---|---|
100 39 441 A 1 | Feb 2002 | DE |
101 29 958 A 1 | Jan 2003 | DE |
102 04 873 C 1 | Oct 2003 | DE |
102 19 917 A 1 | Nov 2003 | DE |
WO2004107435 | Dec 2004 | DE |
09-014836 | Jan 1997 | JP |
Number | Date | Country | |
---|---|---|---|
20070007586 A1 | Jan 2007 | US |