Patent Grant
7972939
The invention generally concerns the manufacturing of multilayer wafer including materials selected from semiconductor materials, and used in applications in microelectronics, optics, optoelectronics and/or optonics.
More precisely, the invention concerns the processes for transferring from a “top” wafer (or donor wafer—these terms being herein understood as equivalent) a layer of a semiconductor material onto a “handle” (or receiving wafer, or “base” wafer—these terms being herein understood as equivalent) in order to make a multilayer wafer.
Several main types of such processes are known. Many of these processes imply a bonding step for bonding the handle and the top wafers, before a removal of the excess material from the bonded top wafer. We will refer to such processes which imply a bonding step as “bond and transfer processes” (or “BTP” processes).
It is specified that the definitions and corresponding acronyms givers in this document are provided for the purpose of clearly defining the invention which is the subject matter to be protected by this document. These definitional acronyms are therefore intended to be used only for the purpose of understanding this particular subject matter. They are not intended to be used for any other purpose, e.g., for the interpretation of other documents such as related patents or patent applications unless otherwise specified.
The known types of BTP processes include:
the BESOI-type process (the removal of material from the top wafer is performed through an etching of the top wafer),
the ELTRAN®-type process (the removal is performed through a detachment of the excess material with the detachment resulting from the attack of a porous region of the top wafer),
the SMART-CUT®-type process (the removal is performed through a detachment of the excess material with the detachment resulting from the splitting of the top wafer at an embrittlement or weakened region obtained by implantation of the top wafer with at least one species. It is specified that a general description of the main steps of a non-limiting embodiment of this type of method can be found in the text entitled SILICON-ON-INSULATOR TECHNOLOGY: Materials to VLSI, 2nd Edition (Jean-Pierre COLINGE).
The bonding step of such processes is generally performed without adding any adhesives (such as resins) between the top and handle wafers. We will refer to such bonding without adhesive as “direct bonding,” And a BTP comprising such a direct bonding will be referred to as “Direct bonding and transfer process” (or “DBTP”). For such direct bonding, the surfaces to be bonded must be very smooth (i.e., present a very low roughness). And it is possible to provide a smooth surface over a first wafer which is to be bonded with a second wafer through a direct bonding, e.g. by building a “bonding layer” (e.g., with an oxide layer) over the first wafer, and/or by submitting the surface of the first wafer to a specific treatment (e.g. polishing, . . . ).
A bonding of the surfaces of two wafers generates between the bonded surfaces a bonding interface associated with a given bonding energy with the higher this energy the stronger the bond. The bonding energy obtained by direct bonding two wafers without any additional treatment is sometimes not sufficiently high. This is the case in particular when it is desired to obtain with the bonding of two or more wafers a multilayer wafer (each initial wafer bringing a “layer” to the multilayer wafer thus obtained), with a high energy bond (i.e., a bonding energy typically greater than 500 mJ/m2) between the bonded wafers. In such cases, the bonded wafers are thus subjected to a “bonding heat treatment” (“BHT”) which improves bonding strength. More precisely, the BHT is generally carried out at high temperatures (i.e., at temperatures of 900° C. or higher). A high temperature BHT carried out at a temperature higher than 900° C. will be referred to herein as “HT BHT”.
A HT BHT reinforces the bonding interface and can raise its bonding energy to a desired level. As an illustration, the bonding energy between a Si wafer and a Si or SiO2 wafer is maximized after a HT BHT at temperatures on the order of 1100-1200° C. The bonding energy obtained after a HT BHT can e.g. be 2 to 2.5 J/m2 for a bonding of Si/Si, SiO2/SiO2, or Si/Si02. Such HT BHT can therefore allow the manufacturing of multilayer wafers with adequate bonding energy between the bonded layers of the wafer.
However, it can be undesirable to carry out a HT BHT on a multilayer wafer. This is the case e.g. when the layers of the multilayer wafer have respective coefficients of thermal expansion (CTE) that are significantly different. In such cases, the layers having the different CTEs will undergo expansions of different magnitudes when exposed to a HT BHT. This can lead to negative effects such as a warp of the wafers, or even the generation of defects such as dislocations in the crystalline structure of the wafers.
It can also be undesirable to carry out a HT BHT on a bonded multilayer wafer for other reasons, e.g. the wafer already comprises components totally or partially elaborated (such wafer is usually called “structured substrate”), or the material of the wafer is likely to be altered by a high temperature treatment (e.g. the case of a wafer that has a sharp doping profile with boron or phosphorous, or comprising metastable layers that would be altered by HT BHT, like low k materials, high k materials or layers with metallic elements like Al2O3, TiN, etc.).
In order to avoid such drawbacks and nevertheless obtain a high bonding energy between bonded layers of a multilayer wafer, techniques of high bonding energy at low temperatures have been developed. In one of these techniques, the surfaces to be bonded are “activated” by their exposure to a plasma, before being contacted together. Such technique will be referred to as “plasma activation” technique. Plasma activation is advantageous since it allows obtaining high bonding energy with a BHT which is carried out after contacting the surfaces to be bonded, but which remains limited to a low temperature BHT (“LT BHT”). And plasma activation has been proposed for DBTPs as a solution for obtaining a high energy bonding without requiring carrying out a HT BHT on a bonded multilayer wafer.
A general illustration of such known DBTP is given in the article entitled “Ultra-thin strained-silicon-on-insulator and SiGe-on-insulator created using low temperature wafer bonding and metastable layers”, Taraschi et al., J. Electrochem. Soc. Vol. 151, No. 1, p. 47 (2004). This article discloses a SMART-CUT®-type process in which prior to the wafer bonding step (between a top and a handle wafers) a plasma activation can be carried out on the surfaces to be bonded. The general principle of such process is illustrated in
b shows the activation of the front surfaces 100 and 200 by a plasma, in a plasma chamber. As illustrated in
In commercially available equipment for microelectronic application, the wafer is disposed on a chuck in the plasma chamber. The chuck is adapted to perform the following functions:
receive and hold the wafer very tightly in place during exposure to the plasma,
regulate the wafer temperature during plasma exposure, by temperature diffusion,
act as an electrode to transfer polarization (bias voltage) to the wafer.
This set of functions that have to be performed by the chuck leads to selecting materials for making the chuck that contain elements that are likely to contaminate the surface of the wafer in contact with the chuck (i.e. the back surface of the wafer, since it is the front surface which is directly exposed to the plasma). This material can typically be Al203 based sintered ceramics.
d illustrates the direct bonding of top wafer 10 and handle wafer 20, these two wafers being bonded by their front surfaces. This bonding generates a multilayer 30 wafer (also referred to as an “intermediate” multilayer wafer) which comprises in particular the embrittlement or weakened zone 11.
e then illustrates the multilayer wafer 35 obtained after a detachment carried out at the embrittlement zone. This wafer 35 can be a SOI (Silicon On Insulator), SGOI (Silicon-Germanium On Insulator), sSOI (strained Silicon On Insulator), GeOI (Germanium On Insulator), SiCOI (Silicon Carbide On Insulator), SOQ (Silicon On Quartz), or any other type.
This type of process illustrates that plasma activation could be advantageous for making multilayer wafers through a DBTP. But plasma activation is in itself associated with another drawback, one that is related to the contaminants generated by the activation. Indeed, as mentioned above plasma activation generates a contamination of the wafer which is exposed to the plasma, in particular on the back surface of the wafer. This “contamination” corresponds to a deposit of metallic elements (e.g. Al, Fe, Ni, Zn, Cr, Ti, Ca, Mg, Y, . . . ) that may be incorporated in fine particles, on the surface of the wafer, or absorbed in the surface in the case of isolated atoms, molecules or ions. This deposit of metallic elements can be observed with levels of contamination in the order of 5×1011 to 5×1012 atoms/cm2 on at least some areas of the surface of the exposed wafer. Such concentrations are too high for a multilayer wafer which is to be subject to further thermal treatments (e.g. a high temperature: annealing at temperatures above 1000° C. for smoothing the surface of the wafer, or further treatments carried out on the multilayer wafer for integrating 30 components into the wafer). Indeed, the contaminants of a multilayer wafer which would be exposed to such high temperatures treatments would undergo a diffusion in the volume of the wafer, and significantly alter the properties (in particular the electrical properties) of the wafer and of its components. And contaminant concentrations such as those mentioned above and resulting from a plasma activation are thus significantly higher than the specifications of maximum contamination level generally in use in the industry of manufacturing the multilayer wafers and their associated components (these specifications being in the order of 5×1010 atoms/cm2, or even less).
Thus, carrying out a direct bonding transfer process (such as e.g. a SMART-CUT®-type process) with an activation of the surfaces to be bonded as in the known state of the art, in order to make multilayer wafers, is associated with some minor drawbacks. And it is specified that if these drawbacks have been exposed hereabove in reference to a plasma activation, similar drawbacks can also be associated to the activation of a surface which is different from a plasma activation (e.g. an oxidation and/or a chemical activation, etc.). Accordingly, the present invention seeks to overcome the previously noted drawbacks and disadvantages.
The invention relates to a method for minimizing or avoiding contamination of a final multilayer structure obtained by transfer of a thin layer from a donor wafer. This method includes one step of providing a donor wafer and a receiving handle wafer, each having a first surface prepared for bonding and a second surface, with the donor layer including a zone of weakness that defines a thin layer of donor wafer material to be transferred to the receiving handle wafer. Next, at least one of the first surfaces is treated to provide increased bonding energy when the first surfaces are bonded together; the surfaces are then bonded together to form an intermediate multilayer structure; and the thin layer is transferred to the receiving handle wafer to form a final multilayer structure by detachment at the zone of weakness and removal of remaining material of the donor wafer. This method avoids or minimizes wherein contamination of the final multilayer structure is minimized or avoided by treating only the first bonding surface of the donor wafer prior to bonding, or by cleaning contamination from the second surfaces of the intermediate multilayer structure prior to detachment of the thin layer.
The treating step preferably comprises activating that surface by exposure to a plasma, preferably one that generates more than 1011 atoms/cm2 of metallic contamination of the second surface of the donor wafer, and the method further comprises removing this contamination by cleaning of that second surface. The contamination is preferably cleaned from the second surface of the donor wafer when present in the intermediate multilayer structure prior to detachment of the thin layer.
The treating step may also comprise activating the first surfaces of both the donor and receiving handle wafers by exposure to a plasma, and then cleaning the second surfaces of both the donor and receiving handle wafers. This cleaning step can be conducted when the donor and receiving handle wafers are present in the intermediate multilayer structure by immersion of the intermediate multilayer structure into a cleaning solution.
Alternatively, the treating step can comprise subjecting the first bonding surface of the handle receiving wafer to surface activation prior to bonding followed by cleaning contamination from the second surface of the handle receiving wafer when present in the intermediate multilayer structure prior to detachment of the thin layer, preferably by spraying a cleaning solution upon that second surface.
Also, the treating step can comprise subjecting the first bonding surfaces of both the donor and handle receiving wafers to surface activation prior to bonding followed by cleaning contamination from the second surfaces of the donor and handle receiving wafers when present in the intermediate multilayer structure prior to detachment of the thin layer, preferably by immersing the intermediate multilayer wafer into a cleaning solution immediately after the bonding step. The cleaning solution can be a SC2 solution or a hydrofluoric acid solution.
In advantageous embodiments, at least one first surface comprises an insulating material, and the zone of weakness is made by introducing at least one species into the donor wafer prior to bonding. Preferably, the zone of weakness is made by implanting one or two different species into the donor wafer prior to bonding. Detachment can be carried out by annealing the intermediate multilayer structure at a temperatures of between 200 and 350° C., and the second surface is cleaned utilizing monoface cleaning equipment. Also, the method further comprises subjecting the final multilayer structure to a high temperature thermal treatment at a temperature of greater than 900° C. Preferred materials for the wafers include a donor wafer made of Si, SiGe, SiC or Ge and a receiving handle layer made of Si, fused silica, SiC or glass.
Other aspects and advantages of the invention shall appear when reading the following description made in reference to the accompanying drawings, in which:
a-1e which have already been commented on in the background make reference to a SMART-CUT®-type process which is generally known in the art;
a-2e are a diagrammatical flowchart of a process according to a first embodiment of the invention, and
a-3f are a diagrammatical flowchart of a process according to a second embodiment of the invention.
As noted, the present method is conducted to avoid or minimize contamination of the second surface of the receiving handle wafer. This method includes transferring from a top donor wafer a layer of material onto a receiving handle wafer, the top and receiving wafers comprising respective surfaces to be bonded. The method steps comprise a treatment step for treating at least a surface to be bonded, the treatment of a given surface of a wafer generating on the opposed surface of the wafer a contamination, a bonding step for direct bonding the surfaces to be bonded of the top and handle wafers, in order to form an intermediate multilayer wafer, and a removal step for removing excess material from the top wafer. The method reduces contamination by treating only the first bonding surface of the donor wafer prior to bonding, or by cleaning contamination from the second surface of the handle receiving wafer when present in the intermediate multilayer structure prior to detachment of the thin layer.
Preferred but non-limiting aspects of such a method are the following:
the treatment step is a surface activation step,
the contamination of the wafer surface which is opposed to the activated surface corresponds to a metallic contaminants concentration above than 1011 atoms/cm2,
the treatment is carried out through the exposure of the surface(s) to be treated to a plasma,
the removal step is carried out by detaching the excess material from the top wafer,
the detachment is carried out at an embrittlement zone which has been made by introduction of at least one species in the top wafer,
the introduction of species is carried out through an implantation of one or more species,
the implantation is a co-implantation of at least two species,
the implantation is carried out with a global implantation dose of a few 1016 atoms/cm2,
the removal step is carried out through a splitting annealing carried out at temperatures remaining under 350° C., and preferably between 200° C. and 300° C.
during the treatment step only a surface of the top wafer is treated,
the cleaning step is carried out immediately after the bonding step,
the cleaning step is carried out before any high temperature thermal treatment,
the cleaning step is a wet cleaning,
the cleaning step is carried out with a SC2 solution and/or a HF solution,
the cleaning step is carried out with a mono face cleaning equipment,
the top wafer is made of Si, or SiGe, SiC, or Ge,
the handle wafer is made of Si, fused silica, SiC, or glass,
the method further comprises a high temperature thermal treatment of the structure which comprises the handle wafer and the part of the top wafer which has not been removed,
the high temperature thermal treatment is carried out at a temperature greater than 900° C.
It is specified that the two embodiments which shall be presented hereunder can be combined. It is also specified that on the Figures which shall be commented hereunder, the elements and features which are identical or similar to those commented with reference to
The embrittlement zone 11 can have been made by introduction of species in the top wafer (this “introduction being carried out e.g. as an implantation and/or by diffusion, or by any other known technique). In the case where the “introduction” of species is an implantation of species, such implantation can be an implantation of only one species (such as e.g. H or He). It can also have been made by implantation of several species (i.e. co-implantation), for example H and He. Here again, the front surface 100 of the top wafer is the surface of an oxide layer (e.g. a deposit of Si02, or an oxidized region of the underlying material).
b illustrates an activation step, during which the front surface 100 of the top is exposed to a plasma. This activation step—like all activation steps mentioned in the different embodiments described in this text—more generally corresponds to a treatment step of a surface to be bonded, with a generation of a contamination on the surface of the wafer which is opposite to the surface treated. The back surface 110 of top 10 is therefore covered by a region 111 which is contaminated with metallic contaminants such as those mentioned above (see
It shall be noted that during this activation step, only the top wafer is exposed to a plasma in order to be activated. The handle wafer 20 is not exposed to any plasma. Therefore, no contamination is deposited on the surfaces of the handle wafer. This selective exposure of the donor top wafer only is specific to this first embodiment of the invention.
In the known experiments for using a plasma activation in a DBTP, both front surfaces of the top and of the handle wafers were exposed for activation by a plasma. In such experiments, the drawbacks associated to the contamination of a multilayer structure which could undergo further high temperature treatments had not been pointed out.
It is to be noted that a reports on one of such experiments (see the article entitled “Ultra-thin strained-silicon-on-insulator and SiGe-on-insulator created using low-temperature wafer bonding and metastable layers”, Taraschi et W., J. Electrochem. Soc. Vol. 151, No 1, p. 47 (2004)) mentions the plasma activation of only one of the wafers to be bonded. But this mention makes reference to the activation of the handle wafer, and not of the top wafer. And not surprisingly, there is no mention in this article of any problem associated to the contamination resulting from the plasma activation.
The following step is the direct bonding of the top wafer and the handle wafer, for making an intermediate multilayer wafer 30 (
The removal can also be carried out according to any method known for BTPs. This removal can therefore be carried out by etching the back face of the top wafer 10 (i.e., the face of the top wafer which is opposed to the front surface 100), on the intermediate wafer 30. In such case (which corresponds to a BESOI-type method), it is not necessary that any implantation has been carried out on the top wafer.
The removal step can also be performed in accordance with the ELTRAN® method. In such case, the embrittlement zone has previously been made by porous material, and here again no implantation is necessary.
In the case of a SMART-CUT®-type or ELTRAN®-type method, a remainder 15 shall be obtained. In every case (i.e. whatever the method used for removing the excess material of the top), the multilayer wafer 35 obtained does not comprise a heavy contamination generated on a back surface of a wafer whose front surface would have been exposed to a plasma for its activation. Indeed, in all cases the heavy contamination region 111 which is on the back surface of the top wafer has been removed from the multilayer wafer 35. This is particularly advantageous since this multilayer wafer shall then undergo additional thermal treatments, some of these thermal treatments being as mentioned before carried out at high temperatures (e.g. smoothing annealing at temperatures above 1000° C.). And exposing a wafer comprising heavy concentrations of contaminants (e.g. concentrations above 1011 atoms/cm2) would generate an unacceptable and irreversible diffusion of the contaminants into the wafer.
In the case of a method generating a remainder 15, the only heavy ion contamination (region 111) is on the remainder, the multilayer wafer 35 being totally free of such heavy contamination. Therefore, this first embodiment of the invention allows to make a multilayer wafer with a DBTP, without having the multilayer wafer polluted by a heavy concentration of contaminants. It is specified that the contaminant concentrations which can be observed on the front face of the exposed top after the activation step are in the order of 2×1010 atoms/cm2, which is a generally acceptable concentration level.
The removal step illustrated in
A co-implantation (e.g. with H+He) therefore constitutes a favorable feature for performing a splitting annealing at moderated temperatures as mentioned above. And such co-implantation even allows a further reduction of the temperature for the splitting annealing. For a H+He co-implantation in a Si top wafer with a global implantation dose of a few 1016 atoms/cm2, this temperature can remain under 350° C. and in particular can be in a range as low as 200-300° C.
b illustrates the step of activation through exposure to a plasma. During this step, the front surface of the top wafer 10 and/or of the handle wafer 20 can be exposed. If only the top wafer 10 is exposed to a plasma for activation, the first embodiment described above can be combined with the second embodiment.
b-3d illustrate the particular case where both top and handle wafers are activated by a plasma.
d illustrates the direct bonding of the top and handle wafers, with their front surfaces contacted. In this second embodiment, the bonding is followed (preferably immediately followed) by a cleaning step (
The cleaning step is carried out before the wafer undergoes any high temperature treatment (i.e. any thermal treatment above 900° C.). The cleaning step illustrated in
Each cleaning of the cleaning step can be carried out as a wet cleaning, i.e. by immersing the wafer to be cleaned in a bath made of the cleaning solution. It is however also possible to perform such cleaning with a monoface cleaning equipment, which projects a controlled spray or jet of cleaning solution selectively, only on the surface to be cleaned. Concerning the cleaning step it is to be noted that the method of the invention is contrary to what one might have expected, in that the use of a cleaning solution on a multilayer wafer 30 whose layers have already been bonded together produces very satisfactory results. Indeed, the prior art tends to indicate that such cleaning should be avoided because the cleaning solution might be infiltrated at the bonding interface (see e.g. “Semiconductor on Wafer Bonding: Science and Technology’—Tong and Gosele, sponsored by the Electrochemical Society, Inc., a Wiley-Interscience Publication, 1999). Such infiltrations would be associated to risks of degrading the quality of the bonding, or even undo the bond between the bonded wafers.
In addition, when considering the prior art one would consider that the manipulations associated to a cleaning step would favor debonding of the bonded wafers. Moreover, one would also consider that the cleaning step would most probably lower the bonding energy at the edge of the bonding interface of the bonded wafer. Therefore, a cleaning step carried out as in this second embodiment goes against the teachings of the prior art.
And the specific process provided by this second embodiment, which has the cleaning performed after the bonding, further allows the inventive process to use a solution such as a SC2 or HF solution, which is known to efficiently remove metallic contaminants. A SC2 or HF solution would indeed have been associated with problems if used for cleaning the top and/or handle wafer before bonding them together: in such case, the SC2 or HF solution would have removed the undesirable contaminants, but would also have deactivated the surfaces to be bonded—thus losing the benefit of the plasma activation. It therefore appears that the specific method proposed in this second embodiment also corresponds to an advantageous process which allows the manufacturing of a multilayer wafer through a DBTP, with a LT BHT.
f illustrates the removal step of the excess material of the top wafer. Here again, all comments made above about the first embodiment are applicable to this embodiment as well. An advantage of this second embodiment is that in the case of a removal carried out in a way which generates a remainder (as illustrated in the left part of
In all cases the activation is
intended to enhance the bonding energy associated to a bonding of the surface which is activated (the surface can generally be referred to as a “front” surface),
associated to the generation of a heavy contamination on the “back” surface which is opposed to the “front” surface to be activated (with a “heavy” contamination being understood as corresponding to a contaminant concentration above 1011 atoms/cm2).
And in all embodiments the invention ensures that no wafer surfaces can undergo high temperature thermal treatments if the surface is heavily contaminated.
This application is a division of application Ser. No. 11/138,926 filed May 25, 2005, now U.S. Pat. No. 7,615,464 which is a continuation of International application no. PCT/IB2004/003275 filed Sep. 21, 2004. The entire content of each prior application is expressly incorporated herein by reference thereto.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5395788 | Abe et al. | Mar 1995 | A |
| 5877070 | Goesele et al. | Mar 1999 | A |
| 6180496 | Farrens et al. | Jan 2001 | B1 |
| 6368938 | Usenko | Apr 2002 | B1 |
| 6468923 | Yonehara et al. | Oct 2002 | B1 |
| 6548382 | Henley et al. | Apr 2003 | B1 |
| 6605518 | Ohmi et al. | Aug 2003 | B1 |
| 6911376 | Yoo | Jun 2005 | B2 |
| 6995075 | Usenko | Feb 2006 | B1 |
| 7176108 | Cayrefourcq et al. | Feb 2007 | B2 |
| 7176528 | Couillard et al. | Feb 2007 | B2 |
| 20030227057 | Lochtefeld et al. | Dec 2003 | A1 |
| 20040035525 | Yokokawa et al. | Feb 2004 | A1 |
| 20040097055 | Henley et al. | May 2004 | A1 |
| Number | Date | Country |
|---|---|---|
| 1 039 513 | Sep 2000 | EP |
| 1 050 901 | Nov 2000 | EP |
| 1 381 086 | Jan 2004 | EP |
| 11-087668 | Mar 1999 | JP |
| WO 2004008514 | Jan 2004 | WO |
| WO 2004042779 | May 2004 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 20100015780 A1 | Jan 2010 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 11138926 | May 2005 | US |
| Child | 12566036 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | PCT/IB2004/003275 | Sep 2004 | US |
| Child | 11138926 | US |
