US20010050387A1

US20010050387A1 - Semiconductor non-volatile memory device and corresponding fabrication process

Info

Publication number: US20010050387A1
Application number: US09/814,177
Authority: US
Inventors: Thomas Skotnicki; Didier Dutartre; Pascal Ribot; Maryse Paoli; Richard Fournel
Original assignee: STMicroelectronics SA
Current assignee: SEMANTECH Inc; STMicroelectronics SA
Priority date: 2000-03-29
Filing date: 2001-03-21
Publication date: 2001-12-13
Also published as: FR2807208B1; FR2807208A1; US6642108B2; US20030038315A1

Abstract

A non-volatile memory includes a floating gate extending in a substrate between source and drain regions. A channel region may be confined by two insulating layers. The invention is particularly applicable to EPROM, EEPROM, Flash and single-electron memories using CMOS technology.

Description

FIELD OF THE INVENTION

The present invention relates to the field of integrated circuits, and, more particularly, to semiconductor non-volatile memory devices including complementary metal oxide semiconductor (CMOS) technology and related fabrication processes.

BACKGROUND OF THE INVENTION

Non-volatile memories store charge on a floating gate, i.e., a gate not connected outside the device. The mode of injection or erasure of this charge depends on the type of memory in question. The control potential on the floating gate is induced by a capacitively coupled control gate. The charge may be injected or erased many times without appreciably damaging the structure. When the charge is stored on the floating gate, it can remain there a very long time (apart from leakage currents) without being impaired. This storage is said to be “non-volatile.”

Among the non-volatile memories, mention is made here of erasable and programmable read-only memories (EPROMs) and electrically erasable and programmable read-only memories (EEPROMs), as well as the so-called “Flash EPROMs” or “Flash EEPROMs” Depending upon the type of non-volatile memory, charge is introduced into the floating gate either by the injection of hot carriers from the channel of the transistor or by forcing conduction through the oxide. This modifies the threshold voltage of the transistor. To discharge the floating gate (i.e, to erase the contents of the memory), either illumination by ultraviolet radiation is used (e.g., in an EPROM memory), thereby making the oxide conducting, or a discharge by a tunnel effect is used.

A non-volatile point memory MMA according to the prior art is illustrated in FIG. 1. The memory device MMA has two polysilicon levels PL 1 and PL2 above the upper surface UPS of the substrate SB. These levels PL1, PL2 form the floating gate FG and the control gate CG, respectively, of the memory device. The floating gate FG is electrically isolated from the source region S, the drain region D, and the channel region CR by a first oxide layer OX1. The floating gate and the control gate are mutually electrically isolated by a second oxide layer OX2. Lateral isolating regions or spacers SPA complete the electrical isolation of the two gates, especially that of the floating gate FG. The use of two levels of polysilicon has drawbacks, especially from the standpoint of integrating a point memory therewith in a more complex structure using CMOS technology.

SUMMARY OF THE INVENTION

An object of the invention is to provide a non-volatile memory device using only a single polysilicon level above the surface of the substrate.

According to the invention, a semiconductor non-volatile memory device includes a silicon-based semiconductor substrate including source and drain regions, a control gate, and a floating gate. The floating gate extends between the source and drain regions formed in the substrate, and the control gate lies above the floating gate. In other words, whereas in the prior art the floating gate is above the source and drain regions, in this case it is “buried” in the substrate. Thus, only a single polysilicon level above the upper surface of the substrate is required to form the control gate.

More specifically, the substrate has a lower part or portion lying beneath the source and drain regions and a channel region lying above the lower part of the substrate between the source and drain regions. The floating gate may be formed from a semiconductor region overdoped with respect to the channel region. The semiconductor region lies above the channel region and is isolated therefrom and from the source and drain regions by an insulating layer.

The non-volatile memory device may also include an additional insulating layer between the channel region and the lower part of the substrate. Thus, the channel of the MOS transistor is bounded by two isolating regions, thereby making it possible to obtain narrow and confined channels. This gives the channels great robustness, as opposed to the drawbacks associated with the effects of short channels.

Apart from the fact that the floating gate is buried in the initial surface of the silicon wafer, it is advantageously a single-crystal silicon gate. The floating gate may be longer than the control gate in the channel direction. Further, it may be confined, in a perpendicular direction, to the width of the active region. This is not the case with the control gate.

A method aspect of the invention is for fabricating a semiconductor non-volatile memory device and includes forming a first layer of a material on an initial silicon substrate. The first layer is surrounded by an isolating peripheral region and may be removed selectively with respect to the silicon. The first layer may comprise a silicon-germanium alloy, for example. Further, a silicon second layer is formed on the first layer and overdoped (e.g., in situ or by implantation) with respect to the initial substrate, and a gate oxide layer is formed on the second layer. A control gate is formed on the gate oxide layer and contacts, at opposing ends thereof, the isolating peripheral region. Additionally, the gate oxide layer, the second layer, the first layer, and an upper part of the initial substrate are etched along two opposed sidewalls of the gate to form cavities. The first layer is selectively etched to form a tunnel between the second layer and the initial substrate, and an insulating layer is formed on the walls of the second layer and in at least part of the tunnel. Further, the cavities may be filled with silicon, and the source and drain regions may be formed in the filled cavities, on each side of the second layer, to form a floating gate.

More specifically, forming the insulating layer may include coating the walls of the tunnel with the insulating layer, where the inside of the tunnel is empty. Alternatively, forming the insulating layer may include completely filling the tunnel with the insulating layer. Either way, forming the insulating layer may also include forming the insulating layer in the bottom of the cavities. In this case, that part of the insulating layer formed in the bottom of the cavities is removed before filling the cavities with silicon. Further, that part of the insulating layer formed in the bottom of the cavities may be removed by anisotropic etching or by chemical etching in a wet bath.

It is also advantageous, before oxidizing the tunnel, to implant selected ions into the bottom of the cavities (e.g., nitrogen ions) to retard the oxidation of the silicon. Thus, a thinner insulating layer is obtained in the bottom of the cavities after the insulating step than inside the tunnel. In this way, it will be possible after the insulating step to maintain an insulating layer on the lower wall of the tunnel. This makes it possible to obtain a point memory whose channel is confined by two oxide layers. Additionally, the silicon may be grown by selective epitaxy during the cavity-filling step. The silicon may optionally fill the inside of the tunnel (if the tunnel has not otherwise been completely filled with oxide).

BRIEF DESCRIPTION OF THE DRAWINGS

Further advantages and characteristics of the invention will become apparent upon examination of a detailed description of embodiments and methods of implementation, which are given by way of non-limitative example, and the appended drawings, in which: [0013]
FIG. 1 (previously described) is a cross-sectional view of a non-volatile point memory according to the prior art; [0014]
FIG. 2 is a cross-sectional view of one embodiment of a non-volatile memory device according to the invention; [0015]
FIG. 3 is a cross-sectional view of another embodiment of a non-volatile memory device according to the invention; [0016]
FIG. 4 is a top plan view of a memory device according to the invention; [0017]
FIGS. 5[0018] a, 5 b, 6 a, 6 b, 7 a and 8 a are cross-sectional views illustrating making of the a memory device of FIG. 4;
FIGS. [0019] 9 to 11 are cross-sectional views illustrating the final steps of a first alternative method according to the invention;
FIGS. [0020] 12 to 14 are cross-sectional views illustrating a second alternative method according to the invention; and
FIGS. [0021] 15 to 17 are cross-sectional views illustrating a third alternative method according to the invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Turning now to FIG. 2, a first embodiment of a non-volatile memory device MMB has a floating gate FG “buried” in a substrate SB. This floating gate FG is formed from a silicon layer PSI and is isolated from the adjacent channel region CR and from the source and drain regions S and D by an insulating layer OX[0022] 1, for example, formed from SiO₂. This floating gate FG extends between the source and drain regions which are formed in the substrate and beneath the upper surface UPS of the substrate SB. Moreover, the floating gate FG is isolated from the control gate CG by a second gate oxide OX2. This control gate is flanked by two spacers SPA and is above the source and drain regions. A person skilled in the art will therefore have noticed that the structure according to the invention requires only a single polysilicon level PL above the substrate.
Referring to FIG. 3, another embodiment of a memory device MMC according to the invention is shown. The sole difference between the device MMC and the device MMB illustrated in FIG. 2 is that the channel region CR of the device MMC is not only confined in its upper part by the oxide OX[0023] 1 but is also confined in its lower part by an additional insulating layer OX3. A very narrow channel is therefore obtained which is particularly robust compared to short-channel effects.
Methods of fabricating the two embodiments illustrated in FIGS. 2 and 3 will now be described in greater detail with reference to FIG. 4 et seq. A partial top view of a memory device according to the invention is illustrated in FIG. 4. Within the semiconductor wafer PQ an isolating peripheral region BX has been produced in a conventional manner (e.g., by an shallow trench isolation (STI) process, which will be appreciated by those of skill in the art). This isolating peripheral region BX thus defines a silicon region forming a so-called “initial” substrate ISB within which the memory structure according to the invention will be formed. The control gate CG is also illustrated in FIG. 4. [0024]
As used in FIG. 5 et seq., the suffix “a” assigned to these figures denotes a section on the line A-A in FIG. 4, while the suffix “b” assigned to a figure denotes a longitudinal section on the line B-B in FIG. 4. Referring to FIGS. 5[0025] a and 5 b, the process starts by successively depositing a layer of a material 1 on the substrate ISB by selective epitaxy (e.g., by chemical vapor deposition). This layer 1 can be removed selectively with respect to the silicon and generally has a thickness of between 1 and 50 nm. Further, a silicon layer 2 is also formed by epitaxy and has a thickness of, for example, between 1 and 50 nm. The selective nature of the epitaxy means that the materials 1 and 2 grow only on the initial substrate ISB and not on the isolating peripheral region BX.
The [0026] material layer 1 that can be removed selectively with respect to the silicon may be any material which, preferably, ensures lattice cell continuity with the silicon of the substrate during the epitaxy. For example, a Si_1-xGex alloy where 0<x≦1 may be used. Si_1-xGe_xalloys are recommended since they can be easily removed selectively, either by a well-known oxidizing chemistry (such as a solution containing 40 ml of 70% HNO₃+20 ml of H₂O₂+5 ml of 0.5% HF) or by isotropic plasma etching.
Preferably, Si[0027] _1-xGe_xalloys (where 0<x≦1) having a high germanium content are used since the etching selectivity with respect to silicon increases with increasing content of germanium in the alloy. It is also possible to use Si_1-xGe_xC_yalloys (where 0<x≦0.95; 0<y≦0.05) which behave like the Si_1-xGe_xalloys with respect to selective removal but together with the silicon layers induce less strain.
As shown in FIGS. 6[0028] a and 6 b, the process continues in a conventional manner with the formation of a gate oxide layer 3 (e.g., SiO₂) and then a polysilicon layer PL on the gate oxide layer. The polysilicon layer PL is conventionally etched in a known manner to form the active region of the control gate CG. Spacers SPA (e.g., made of silicon nitride Si₃N₄) are also conventionally formed on the sidewalls of the polysilicon layer PL.
It should be noted here that, although the polysilicon region PL forms the control gate, the future floating gate of the memory structure according to the invention will be produced in the [0029] silicon layer 2. Thus, this silicon epilayer must be highly doped (either in situ or by implantation). By way of example, this doping will be carried out to achieve a dopant concentration on the order of 10²⁰atoms/cm³.
Starting from the structure illustrated in FIGS. 6[0030] a and 6 b, the process continues (FIG. 7a) with etching (e.g., with a plasma) of the gate oxide layer 3, the silicon layer 2, the layer of selectively removable material 1, and an upper part of the initial silicon substrate ISB on each side of the spacers SPA. This forms two cavities CAV. The remaining portion of the oxide layer 3 will form the gate oxide OX2 of the point memory. Also, the remaining portion of the silicon layer 2 forms a block of single-crystal silicon SCS which will form the future floating gate of the point memory. Further, the reference 10 denotes the remaining portion of the layer of selectively removable material, e.g., silicon-germanium.
Next, the material of the [0031] layer 10 is selectively removed, for example, by the above mentioned oxidizing chemistry. Thus, as illustrated in FIG. 8a, a tunnel TNL is formed instead of the layer 10. It should be noted here that the block of silicon SCS adheres to the gate oxide OX2 and to the lower surface of the polysilicon block PL of the control gate CG which, moreover, rests at its two ends on the isolating peripheral region BX.
FIG. 8[0032] a shows the cavities which have a bottom lying below the tunnel TNL. In this case, various depths of the cavities may be used in accordance with the invention as long as the depth is sufficient to allow the layer 10 to be exposed for the purpose of its subsequent removal. However, as seen in FIG. 9 and the following figures, the bottom of the cavities CAV have been shown for the sake of simplification in the extension of the lower wall of the tunnel TNL.
Referring more particularly to FIGS. [0033] 9 to 11, a first method of implementing the process according to the invention will now be described. The method provides a memory device structure as illustrated in FIG. 2. The process starts (FIG. 9) with the oxidation of the tunnel TNL. It is assumed that the thickness of the oxide layer and the height of the tunnel TNL have been chosen to completely fill the tunnel TNL with the insulator. This insulating layer, which may be silicon dioxide, for example, is obtained by thermal oxidation, for example. The insulating layer 20 thus formed has a part 202 lying in the tunnel TNL, a part 201 lying in the bottom of the cavities CAV, and a part 200 lying along the vertical sidewalls of the single-crystal silicon block SCS. In this regard, it should be noted that the thickness of the oxide layer 200 is greater than the thickness of the oxide layer 201 which grows on the bottom of the cavities. This greater thickness of the layer 200 is in fact due to growth of an oxide on a highly doped material.
The process then continues (FIG. 10), for example, with anisotropic plasma etching to remove the insulating oxide layer at the bottom of the cavities CAV. It should be noted here that this anisotropic plasma etching does not remove the [0034] oxide layer 202 lying in the tunnel since the inside of the tunnel is shielded from the plasma etching. Moreover, even if the thickness of the oxide layer 203 lying on the sidewalls of the block SCS has decreased after the anisotropic etching, some oxide still remains because of the initially greater thickness of the layer 200 compared with the layer 201. FIG. 10 therefore shows that the single-crystal silicon block SCS which will form the future floating gate FG is completely isolated from the substrate ISB, and also from the control gate, by the gate oxide OX2.
The bottoms of the cavities could be deoxidized by chemical etching in a conventional wet bath instead of anisotropic plasma etching. Here again, the configuration illustrated in FIG. 10 would be obtained because of the greater oxide thickness in contact with the overdoped single-crystal silicon block SCS. As illustrated in FIG. 11, the next step includes growing silicon in the cavities using a conventional selective epitaxy process. This epitaxial silicon ESR then rises up to the spacers and coats the single-crystal silicon block SCS. [0035]
The implantation of the source and drain junctions on each side of the spacers of the control gate then completes the fabrication process and allows a structure of the type illustrated in FIG. 2 to be obtained. The next steps of the fabrication process, such as the siliciding of the source, drain and control gate regions, are similar to those of a conventional CMOS process. It should be noted here that in FIG. 11 the substrate ISB and the epitaxial silicon ESR together form the substrate SB in which the floating gate SCS lies. The floating gate is isolated from the subjacent channel region and from the source and drain regions by the oxide OX[0036] 1.
FIG. 12 illustrates a step of oxidizing the tunnel TNL that results only in the encapsulation of the walls of the tunnel. The inner remaining part of the tunnel TNL remains empty. More specifically, the oxide layer grows thermally on the single-crystal block SCS and on the initial substrate ISB. As explained above, the thickness of the [0037] oxide layer 301 having grown on the initial substrate ISB is less than the thickness of the oxide layers 300 and 302 having grown in contact with the highly doped silicon SCS.
Deoxidation of the bottoms of the cavities CAV by a wet chemistry then results in the structure illustrated in FIG. 13. The silicon block SCS remains surrounded by insulating [0038] oxide layers 303 and 304, although these are thinner than the initial layers 300 and 302. On the other hand, the cavities CAV have been completely deoxidized together with the lower wall of the tunnel.
The next step is similar to that described with reference to FIG. 11. More specifically, as illustrated in FIG. 14, silicon is grown in the cavities CAV to obtain epitaxial silicon regions ESR. These regions ESR complete the substrate ISB to encapsulate, on the one hand, the silicon block SCS isolated by the oxide OX[0039] 1 and, on the other hand, to penetrate into the tunnel TNL.
Here again, as illustrated in FIG. 14, a point memory structure similar to that schematically illustrated in FIG. 2 is obtained. It should be noted that if, in the structure illustrated in FIG. 12, an anisotropic plasma etching operation had been carried out to deoxidize the bottoms of the cavities CAV instead of using wet chemistry, some oxide would have remained on the lower wall of the tunnel TNL. The final result would therefore be a memory structure as illustrated in FIG. 3 after the epitaxial silicon growth in the cavities and inside the tunnel. That is, a channel region would be confined in its upper part and in its lower part by two isolating regions. [0040]
As will now be seen in greater detail with reference to FIGS. [0041] 15 to 17, it is also possible to obtain a memory device structure according to that illustrated in FIG. 3 by deoxidizing the bottoms of the cavities CAV using a wet bath chemistry. However, it is then necessary, before carrying out the step of oxidizing the tunnel TNL, for the cavities CAV to undergo ion implantation. This may be done with nitrogen ions, for example, and the ion implantation retards the oxidation of the silicon. Consequently, during the subsequent tunnel oxidation step, the oxide 402 inside the tunnel will grow with a greater thickness than the oxide 401 at the bottom of the cavities. It should be noted here that the spacers (not shown in FIG. 15) lying on each side of the control gate CG serve as a mask for the nitrogen implantation.
Of course, as explained above, some [0042] oxide 400 and 404 also grows in contact with the block SCS to a greater thickness than that of the oxide 401. The process may then continue, as illustrated in FIG. 16, with deoxidation by wet chemistry resulting in an oxide layer 403 and 405 being left around the block SCS. Also, an oxide layer 406 results on the lower wall of the tunnel TNL. The process then continues, as illustrated in FIG. 17, with the growth of silicon in the cavities. This results in the formation of the epitaxial silicon region ESR penetrating the tunnel TNL and coating the block SCS. The channel region is then formed by the epitaxial silicon region ESR located between the oxide layer OX1 and the oxide layer OX3.
The operation of a memory device according to the invention, whether the one illustrated in FIG. 2 or the one illustrated in FIG. 3, does not differ from the conventional operation of a FLASH, EEPROM or EPROM point memory. Moreover, in the case of very small dimensions, such as for point memories having structures on the order of 10 nm×10 nm or 20 nm×20 nm, for example, the structure according to the invention can also operate as a one-electron memory. The passage of a single electron in the floating gate thus prevents the arrival of the following electrons by the Coulomb blocking mechanism. [0043]
More specifically, the electron stored in the floating gate creates a repulsive action with respect to the following electrons and, because of the very small dimensions of the low gate-channel capacitive coupling, the difference of one electron is sufficient to change the threshold voltage of the transistor considerably. Such a one-electron memory is less sensitive to aging since a smaller number of electrons cross the gate oxide. [0044]
Additionally, the process according to the invention makes it possible to easily obtain a thinner gate oxide between the floating gate and the channel region than the gate oxide separating the two gates. This is particularly advantageous for non-volatile memories. [0045]

Claims

That which is claimed is:

1. Semiconductor non-volatile memory device comprising a silicon-based semiconductor substrate (SB) containing a source region (S) and a drain region (D), a control gate (CG) and a floating gate (FG), characterized in that the floating gate (FG) extends between the source (S) and drain (D) regions formed in the substrate (SB) and the control gate (CG) lies above the floating gate (FG) so as to be proud of the source and drain regions.

2. Device according to

claim 1

, characterized in that the substrate (SB) has a lower part lying beneath the source and drain regions and a channel region (CR) lying above the lower part of the substrate between the source and drain regions and in that the floating gate (FG) is formed from a semiconductor region (SCS) overdoped with respect to the channel region, lying above the channel region (CR) and isolated from this channel region and from the source and drain regions by an insulating layer (OX1).

3. Device according to

claim 2

, characterized in that it includes an additional insulating layer (OX3) placed between the channel region (CR) and the lower part of the substrate (SB).

4. Process for fabricating a semiconductor non-volatile memory device, characterized in that it comprises the formation, on an initial silicon substrate (ISB) surrounded by an isolating peripheral region (BX), of a first layer (1) of a material that can be removed selectively with respect to the silicon; the formation on the said first layer (1) of a silicon second layer (2) overdoped with respect to the initial substrate; the formation of a gate oxide layer (3) on the second layer (2); the formation, on the gate oxide layer (3), of a control gate (CG) bearing at its two opposed ends on the isolating peripheral region (BX); the etching, along two opposed sidewalls of the control gate, of the gate oxide layer (3), of the second layer (2), of the first layer (1) and of an upper part of the initial substrate (ISB) so as to form cavities (CAV); selective etching of the first layer (1) so as to form a tunnel (TNL) between the second layer (2) and the initial substrate (ISB); a step of insulating the second layer (2), comprising the formation of an insulating layer on the walls of the second layer (2) and in at least part of the tunnel (TNL); the filling of the cavities with silicon (ESR); and the formation of the source and drain regions in the filled cavities, on each side of the second etched layer (SCS) forming a floating gate (FG).

5. Process according to

claim 4

, characterized in that the insulating step comprises coating the walls of the tunnel with the insulating layer (301, 302; 402, 404), the inside of the tunnel being empty.

6. Process according to

claim 4

, characterized in that the insulating step comprises completely filling the tunnel with the insulating layer (202).

7. Process according to one of

claims 4

to

6

, characterized in that the insulating step also includes the formation of the insulating layer in the bottom of the cavities (CAV) and in that part of the insulating layer formed in the bottom of the cavities is removed before the cavities are filled with silicon.

8. Process according to

claim 7

, characterized in that part of the insulating layer formed in the bottom of the cavities (CAV) is removed by anisotropic etching.

9. Process according to

claim 7

, characterized in that part of the insulating layer formed in the bottom of the cavities (CAV) is removed by chemical etching in a wet bath.

10. Process according to

claim 9

taken in combination with

claim 5

, characterized in that, before oxidizing the tunnel, selected ions are implanted into the bottom of the cavities so as to retard the oxidation of the silicon and to obtain, after the insulating step, a thinner insulating layer (401) in the bottom of the cavities than inside the tunnel.

11. Process according to one of

claims 4

to

10

, characterized in that the silicon is grown by selective epitaxy during the cavity-filling step, the silicon (ESR) also filling, optionally, the inside of the tunnel.