MANUFACTURING A COMPONENT ON A CARRIER SUBSTRATE

Description

FIELD OF INVENTION

[0001] The present invention relates broadly to a method of fabricating a device on a carrier substrate, and to a device on a carrier substrate.BACKGROUND

[0002] For the growth of III-N materials (i.e. GaN, AlN and InN and their alloys) on 50, 100, 150, 200 mm, or even larger Si substrates, it is often discovered that although the tensile strain due to thermal mismatch is carefully compensated by a strain engineered buffer, the wafer is fragile during further process handling. The fragility manifests itself e.g. in the GaN-on-Si wafers breaking into large pieces with fairly high frequency during steps involving thermal processing (e.g. anneals, high temperature film deposition / etching etc.) and mechanical handling (e.g. chemo-mechanical polishing, wafer bonding etc.).

[0003] For example, it was noticed that the fragility of 200 mm diameter 725 µm thick GaN-on-Si wafers was caused primarily by the formation of slip-lines in the Si substrate during the substrate annealing step e.g. before the Low Temperature (LT)-AlN deposition. This is believed to be due to the presence of vertical and radial temperature variations across the 200 mm Si substrate. The Si crystal slip takes place if the local stress exceeds the yield strength at the annealing temperature (1050 °C) prior to the LT-AlN growth. There are two possible major sources of stress on the Si substrate in MOCVD growth. They are the contact stresses between the wafer and the point at which it contacts the susceptor, and the thermal stress due to temperature non-uniformity in the vertical and radial directions across the wafer. The slip lines originate from the edge of the wafer and propagate toward the center of the wafer. Minimizing radial temperature differences across the 200 mm Si wafer during growth through the optimization of heater zone settings is one key way to reduce slip formation and wafer fragility, but it is not possible to fully eliminate vertical temperature differences through the wafer due to heat only being supplied to the back-side of the wafer. Thus, in almost all cases, wafer fragility remains an issue due to the high growth temperatures involved in III-nitride on Si epitaxy.

[0004] US 2014 / 284609 A1 describes a method of manufacturing an III-N substrate includes bonding a Si substrate to a support substrate, the Si substrate having a (111) growth surface facing away from the support substrate, thinning the Si substrate at the (111) growth surface to a thickness of 100 µm or less, and forming III-N material on the (111) growth surface of the Si substrate after the Si substrate is thinned. The support substrate has a coefficient of thermal expansion more closely matched to that of the III-N material than the Si substrate.

[0005] US 2010 / 127353 A1 describes composite substrates that include a strained III-nitride material seed layer on a support substrate. Methods of producing the composite substrate include developing a desired lattice strain in the III-nitride material to produce a lattice parameter substantially matching a lattice parameter of a device structure to be formed on the composite substrate. The III-nitride material may be formed with a Ga polarity or a N polarity. The desired lattice strain may be developed by forming a buffer layer between the III-nitride material and a growth substrate, implanting a dopant in the III-nitride material to modify its lattice parameter, or forming the III-nitride material with a coefficient of thermal expansion (CTE) on a growth substrate with a different CTE.

[0006] US 2008 / 217645 A1 describes a semiconductor structure including a substrate, a nucleation layer on the substrate, a compositionally graded layer on the nucleation layer, and a layer of a nitride semiconductor material on the compositionally graded layer. The layer of nitride semiconductor material includes a plurality of substantially relaxed nitride interlayers spaced apart within the layer of nitride semiconductor material. The substantially relaxed nitride interlayers include aluminum and gallium and are conductively doped with an n-type dopant, and the layer of nitride semiconductor material including the plurality of nitride interlayers has a total thickness of at least about 2.0 µm.

[0007] Embodiments of the present invention seek to address at least one of the above problems.SUMMARY

[0008] In accordance with a first aspect of the present invention, there is provided a method of fabricating a device on a Si

[100] substrate as defined in claim 1. Further features of example embodiments are defined in the dependent claims.BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Embodiments of the invention will be better understood and readily apparent to one of ordinary skill in the art from the following written description, by way of example only, and in conjunction with the drawings, in which: Figure 1 shows the steps of Si(111) substrate replacement for GaN / Si, according to an example embodiment. Figure 2 shows a cross-sectional SEM image of the GaN and its buffer layers on Si (100) substrate after the 1 st< bonding and Si (111) removal of Figure 1. Figure 3 shows a cross-sectional SEM image of the GaN-on-insulator (GaN-OI) after the AIN, AlGaN buffers and part of the GaN are removed, according to an example embodiment. Figure 4 shows experimental results of the analysis of the lattice constant of the GaN for a 5 graded AlGaN layers strain engineered example, at different steps of buffer layer / GaN removal. Figure 5 shows experimental results of the analysis of wafer bow for the 5 graded AlGaN layers strain engineered example, at different steps of buffer layer / GaN removal. Figure 6 shows a flow chart illustrating a method of fabricating a device on a Si

[100] substrate, according to an example embodiment. DETAILED DESCRIPTION

[0010] In example embodiments of graded AlGaN layers, for example 3 graded AlGaN layers with Al content of 80%, 50% and 20% (starting from e.g. a Si (111) substrate), were grown on top of a AlN / Si structure at a temperature of 1050 °C. Example embodiments use a strain compensation method in GaN-on-Si heteroepitaxy. The basic idea is to introduce compressive strain during epitaxial growth by employing the in-plane lattice mismatch in the AlGaN material system to compensate for the large tensile strain due to thermal mismatch generated during cooling from growth temperature to room temperature. The stress evolution of step-graded AlGaN is discussed in detail based on the basic strain engineering principle. While the difficulty in strain engineering on a 725 µm thick 200 mm diameter Si wafer can be reduced by employing a shaped susceptor that decouples the change in thermal conduction from wafer curvature change, example embodiments of the present invention apply Al 0.2 Ga 0.8 N layer thickness adjustment and preferably SiN x in situ masking to tune the final bow of the GaN-on-Si wafers. Usually, a more convex wafer is produced from an increase in Al 0.2 Ga 0.8 N layer thickness and Threading Dislocation Density (TDD), in general, improves as well. However, the composition of screw and edge Threading Dislocation (TD) changes with Al 0.2 Ga 0.8 N layer thickness. Usually, 1.2 µm of GaN can be deposited with a SiN x in situ masking layer inserted after 80 nm of GaN at 1000 °C. In the case of SiN x in situ masking, it decouples the compressive strain of the GaN from the Al 0.2 Ga 0.8 N layer, so the GaN layer is less compressive. TDD monotonically improves with the coverage of SiN x . In preferred embodiments, Al 0.2 Ga 0.8 N layer thickness and SiN x in situ masking duration are combined to advantageously produce a bow-free wafer with minimum TDD on 725 µm thick 200 mm diameter Si wafers.

[0011] It is noted that in various embodiments the other e.g. 50% and 80% Al content AlGaN buffer layers can advantageously also help in the strain engineering. However, it is expected that in the step-graded buffer system according to example embodiments, it is the layers with lowest e.g. Al content (i.e. 20% Al and 0% Al == GaN) that adds the most compressive strain to the system, and thus has the greatest effect on bow control.

[0012] Example embodiments of the present invention provide a method to replace the initial substrate, Si (111), that contains slips with a new, slip-free substrate, a Si (100) substrate. Through this method, the thick buffer layers that may have been used for strain engineering can also be removed and a thin device layer structure on Si can be realized even when buffer layers are used for strain engineering. In addition, the stress of e.g. GaN can be manipulated by adjusting the final GaN thickness. Additionally or alternatively, the Si (100) substrate can also have the advantage of being accepted and processed in commercial CMOS foundries easily.

[0013] Figures 1 a) and b) show a sequence of steps for substrate replacement according to an example embodiment. In this embodiment, the graded layer approach for strain management discussed above has been used, however, different strain engineering methods may be applied in the formation of a starting GaN on Si (111) substrate, containing e.g. one or more GaN device layers,in different embodiments,.

[0014] As shown in Figure 1 a), prior to the actual epitaxial growth, a Si (111) substrate 104 was first in situ annealed to remove native oxide. Then, a 20 nm low-temperature (LT) AlN nucleation layer 106a was grown at 980 °C. The temperature was increased to grow a high-temperature (HT)-AlN layer 106b, with AlN precursor flow being maintained during the temperature ramping (10 nm of AlN deposition took place during the temperature ramping). After 210 nm AlN (106a / b) growth, step-graded Al x Ga 1-x N layers (250 nm 80% Al layer 107, 310 nm 40% Al layer 108 and 440nm 20% Al layer 109) are grown to introduce compressive strain to compensate the tensile strain built up in the subsequent grown GaN layers when the wafer is cooled down after growth. Usually, the growth of 0.5 to a few µm u-GaN layer 110 with SiN x masking layer 111 is carried out before the growth of the device layer(s) 112.

[0015] It is noted that in this example embodiment, the epitaxy is Ga-polar / metal-polar, which is the more typical case as will be appreciated by a person skilled in the art. However, N-polar epitaxy is possible in different embodiments. The terms "N-polar" and "Ga-polar / metal-polar" refer to the atomic arrangement at the top (accessible) surface of the wafer, and the polarity / arrangement is generally preserved throughout the various material layers. With reference to Figure 1 a) the polarity of the top (accessible) surface of the GaN device layer(s) 112 is Ga-polar in this embodiment, and thus the wafer is considered to have a Ga-polar surface, and correspondingly the bottom of the GaN device layer(s) 112 will have an N-polarity. As the polarity is preserved across each layer, it should be noted that the top of the adjacent GaN layer 110 will have a Ga-polar surface again, while the bottom of the GaN layer 110 will have an N-polarity. The converse is true for all the polarities for a starting / growth wafer with an N-polar surface in different embodiments.

[0016] With reference to step 1 of Figure 1b, in this example embodiment, SiO 2 102 was then deposited on the wafer 103 (i.e. the Si (111) substrate 104, the AlN layer 106a / b, 3 graded AlGaN buffer layers 107-109 with Al content of 80%, 50% and 20% respectively, GaN layer 110 with SiN x masking (not shown), which is considered a buffer layer in example embodiments, followed by the GaN device layer(s) 112) and then densified at high temperature (e.g. ~ 600 °C, several hrs) under N 2 ambient. The SiO 2 layer 102 can be replaced by Si 3 N 4 , Al 2 O 3 , AlN (aluminum nitride), BN (boron nitride) and other dielectrics in different embodiments (as well as a combination of different dielectrics, e.g. SiO 2 +Si 3 N 4 ) to improve the thermal conductivity as well as bonding strength. To achieve a successful fusion bonding, the RMS roughness of the wafer surface is preferably < 1 nm. Hence, the wafer 103, after dielectric deposition, was polished using Chemical Mechanical Polishing (CMP) followed by RCA clean in example embodiments. Another Si (100) substrate / wafer 114 was used and served as a donor wafer.

[0017] Prior to bonding, both wafers (i.e. wafer 103, after dielectric deposition, and Si (100) substrate 114) were subjected to plasma exposure (e.g. O 2 , N 2 , H 2 , Ar, etc) for several seconds, rinsed with deionized water and then spin-dried in example embodiments. Plasma exposure can increase the surface hydrophilicity of the dielectric (e.g. SiO 2 layer 102). The rinsing step terminates the wafers' surfaces with hydroxyl (OH) groups at a sufficiently high density to initiate wafer bonding. After bonding, which is based on van-der-Waals forces between the hydrogen atoms in this example embodiment, the wafer pair 103 / 114 was annealed at 300 °C in an atmospheric N 2 ambient for 3 hrs to further enhance the bond strength. It is noted that the bonding can be done with any reasonable Si (111) wafer's 103 starting bow value, e.g. absolute bow < 150 µm, with the final wafer bow being advantageously optimized as described in more detail below, and in general a final wafer bow < 50 µm is desirable to improve the yield of subsequent fabrication processing.

[0018] With reference to step 2 of Figure 1b), grinding of the Si (111) 104 (to 50 µm in this example embodiment) was performed. After that, a protective layer from Brewer Science 116 (which is able to survive in acidic environment) was deposited, e.g. spin coated, on the backside of the Si (100) 112 donor wafer to act as a protection layer during the Si removal process from the Si (111) 104 substrate. The remaining Si (111) is removed in this example embodiment by submerging the wafer bonded pair into the HNA solution (e.g. HF : Nitric acid : Acetic acid = 1 : 1.5 : 3.75 in volume, noting that the ratio can be changed to achieve different etching rates as desired). The AlN 106 was used as an etch-stop layer in this example embodiment, since the etching selectivity of AlN over Si is high in the HNA solution used. The etching was carried out at room temperature and until the Si was completely removed, which can be determined by noting when effervescence within the etchant ceases. The protective layer was removed by acetone resulting in the structure as shown in step 3 of Figure 1.

[0019] The AlN 106, the 3 AlGaN buffers 107-109, and the GaN layer can then be removed, e.g. by inductive coupled plasma reactive ion etching (ICP-RIE) or CMP process. According to the claimed invention, one or more of the buffer layers remain, e.g. due to the bow / strain requirements. A GaN device layer 112 with an N-polar surface can thus be obtained, as shown in step 4 of Figure 1b). This is because the wafer has been vertically inverted in the intervening step 2, so what used to be the Ga-polar top (accessible) surface of GaN device layer 112 in step 1 is now (in step 4) the bottom layer (bonded to the SiO2 layer 102), and thus the new top (accessible) surface is the previous N-polar surface of GaN device layer 112 that was adjacent to GaN layer 110 (in step 1). To achieve good process yield, it may be preferred to have an appropriate etch-stop layer (not shown) in such embodiment so that the ICP-RIE or CMP process in step 4 ends at the proper depth to result in the desired N-polar surface.

[0020] To realize a GaN layer 112 with a Ga-polar surface from step 4 in a continuation of the process of this example embodiment, a SiO 2 layer (or other dielectric or a combination of dielectrics) 117 was deposited on the wafer of step 4 in Figure 1b and then densified. CMP process was carried out to smoothen the SiO 2 film 117 to achieve a successful bonding. After the CMP process, the wafer was RCA cleaned and bonded to another Si (100) handle substrate 118, as shown in step 5 of Figure 1. The bonding process was similar to the one described above with reference to step 1 of Figure 1b). After the bonding, the same grinding as described above for the removal of the Si (111) substrate 104 with reference to step 2, and tetramethylammonium hydroxide (TMAH) etching is used for the removal of the Si (100) substrate 114, resulting in the wafer depicted in step 6 of Figure 1b), with a GaN layer(s) 112 with a Ga-polar surface.

[0021] It is again noted that if a Ga-polar surface is desired at the end of the etching in step 4, without the 2 nd< bonding step described above, a way to do it would be to grow an inverted device epi-structure (compare Figure 1 a)) with an N-polar top (accessible) surface of the GaN device layer 112, according to a different embodiment, instead of the Ga-polar top (accessible) surface of GaN device layer 112. This effectively inverts the polarity of the exposed (layer 112 in step 4) GaN surface to be Ga-polar at that point, thereby removing the need for steps 5 (2 nd< bonding step) and 6 for obtaining the Ga-polar device layer in such an embodiment. To achieve good process yield, it may again be preferred to have an appropriate etch-stop layer in such embodiment so that the ICP-RIE or CMP process in step 4 ends at the proper depth to result in the desired Ga-polar surface.

[0022] Figure 2 shows a cross-sectional Scanning Electron Microscopy (SEM) image of the structure of step 3 in Figure 1, i.e. after the 1st bonding and Si (111) substrate 104 removal.

[0023] Figure 3 shows a cross-sectional SEM image of the structure of step 4 in Figure 1, i.e. after the removal of AlN 106, AlGaN buffers 107-109, GaN layer 110 and parts of GaN layer(s) 112 through ICP-RIE.

[0024] Figure 4 shows experimental results of the analysis of the lattice constant of the GaN for a 5 graded AlGaN layers strain engineered example , at different steps of buffer layer / GaN removal, i.e. with different "top" layers exposed to the ambient around the wafer.

[0025] As can be seen from the results shown in Figure 4, by controlling the final GaN thickness (compare t B and t C for samples B and C respectively), the stress of the GaN can be manipulated from 0.14 GPa to 0.3 GPa and even higher tensile stress is achievable with a much thinner GaN. Highly tensile strain GaN layer(s) may e.g. be desired to increase the electron mobility and hence improve the performance of High-Electron-Mobility Transistors (HEMTs).Advantageously, since the fragile Si(111) substrate was replaced by the Si(100) donor substrate in example embodiments, the high tensile strain of the GaN layer(s) does not cause breaking of the wafer.

[0026] Figure 5 shows experimental results of the analysis of wafer bow for the 5 graded AlGaN layers strain engineered example, at different steps of buffer layer / GaN removal. As can be seen from Figure 5, the final wafer bow (-20 µm, compare numeral 500) after removing the AIN, AlGaN buffers, u-GaN and parts of n-GaN layers is getting smaller compared to the wafer bow (-30 µm, compare numeral 502) with all layers, including the buffer layers, present. It is noted that the n-GaN layers as indicated in Figure 5 include the device layer(s). As can be seen from Figure 5, bow of the initial GaN / buffers / Si (compare sample A in Figure 4) is -30 um, the bow is increased to -110 um after removal of AlN buffer layer according to an example embodiment, and the bow is reduced to -85 um after removal of AlN+AlGaN1+AlGaN2+AlGaN 3 buffers according to another example embodiment. Finally, the bow of the GaN / SiO2 / Si (compare sample C in Figure 4) is reduced to -20 um after all the buffer layers are removed.

[0027] Figure 6 shows a flow chart 600 illustrating a method of fabricating a device on a carrier substrate, according to an example embodiment. At step 602, a first substrate is provided. At step 604, one or more device layers are formed on the first substrate. At step 606, a second substrate is bonded to the device layers on a side thereof opposite to the first substrate. At step 608 the first substrate is removed. The carrier substrate is a Si

[111] substrate and the first substrate is a Si

[100] substrate. The one or more device layers are one or more GaN device layers.

[0028] Forming the device layers comprises forming a plurality of AlGaN buffer layers with different compositions and a GaN buffer layer with a SiNx masking layer disposed therein on the plurality of AlGaN buffer layers prior to forming the one or more GaN device layers on the plurality of AlGaN buffer layers and the GaN buffer layer. The plurality of buffer layers and the one or more device layers are grown by epitaxial growth. The method further comprises removing one or more of the plurality of AlGaN buffer layers and the GaN buffer layer such that some of the plurality of AlGaN buffer layers and the GaN buffer layer remain on the first Si

[100] substrate together with the one or more GaN device layers.

[0029] The method may further comprise bonding a second substrate to at least a portion of the device layers on a side thereof opposite to the second substrate, and removing the first substrate.

[0030] The polarity of the device layers may be inverted as a result of the bonding to the third substrate and the removing of the second substrate. The second substrate is a Si

[100] substrate. The method may further comprise providing another etch-stop layer and using the other etch-stop layer to achieve a desired polarity of the device layers with high yield.

[0031] Embodiments of the present invention can have one or more of the following characteristics / advantages: (i) graded layers, e.g. 3 step graded AlGaN layers with Al content of 80%, 50% and 20%, are grown to introduce compressive strain during epitaxial growth. Although 3-step graded and 5 step graded AlGaN layers are elucidated as examples in the above description, other numbers of graded layer are also possible, typically about 1-10 numbers of graded layers. This is advantageous as tensile strain will be introduced during cooldown from the growth temperature to room temperature at the end of the growth process. Hence, both the compressive and tensile strain will preferably compensate each other and results to a minimum change of the wafer bow. For subsequent processing, such as wafer bonding, lithography, etc. tailoring the step-grading profile (i.e. number of steps, thickness of each step, alloy composition of each step) can preferably be performed, noting that depending on the final device heterostructure and thickness, the optimum stress profile desired within the buffer layers can vary significantly. (ii) SiN x in-situ mask is used to decouple the compressive strain and make the GaN layer relatively less compressively-strained after the SiN x in-situ mask. (iii) The fragile Si (111) substrate is replaced by a Si (100) wafer. (iv) The stress of GaN and final device layers can also be manipulated by controlling the final GaN thickness. Highly tensile strained GaN layer(s) can e.g. increase the electron mobility and hence improve HEMT's performance. (v) Thin device (HEMT or LED) layers are achievable by removing the thick buffer layers for a better device's performance (better heat dissipation, e.g. in HEMT or Light Emitting Diode (LED) devices). (vi) N-polar GaN or Ga-polar GaN can be obtained. (vii) The Si (100) wafer is easier to be accepted and processed by foundries compared to Si (111) substrates.

[0032] It will be appreciated by a person skilled in the art that numerous variations and / or modifications may be made to the present invention as shown in the specific embodiments.

[0033] For example, in different embodiments not claimed as such, the concept of the present invention can be applicable to other device layer(s), including to other semiconductor material systems. For example, the tensile thermal mismatch in heteroepitaxial InGaP (e.g. for a light emitting device (LED)) on Si with Ge and GaAs buffer cause the wafer to have a large concave bow. In such embodiments, the method to replace the original Si substrate and remove the Ge and GaAs buffer can also be applied to improve wafer stability.

[0034] Also, while the use of a graded structure for strain engineering / control has been described in the example embodiments herein, the concept of the present invention can be applied to wafers with different types of strain engineering (e.g. superlattice buffers etc.) in different embodiments not claimed as such.

[0035] The invention is defined by the appended claims.

Claims

1. A method of fabricating a device on a Si[100] substrate, the method comprising: providing a Si[111] substrate (104); forming one or more GaN device layers (112) on the Si[111] substrate (104); bonding a first Si[100] substrate (114) to the one or more GaN device layers (112) on a side thereof opposite to the Si[111] substrate (104); and removing the Si[111] substrate (104); wherein forming the one or more GaN device layers (112) comprises forming a plurality of AlGaN buffer layers (107, 108, 109) with different compositions and a GaN buffer layer (110) with a SiNx masking layer disposed therein on the plurality of AlGaN buffer layers (107, 108, 109) prior to forming the one or more GaN device layers (112) on the plurality of AlGaN buffer layers (107, 108, 109) and the GaN buffer layer (110); wherein the plurality of buffer layers (107, 108, 109, 110) and the one or more GaN device layers (112) are grown by epitaxial growth; the method further comprising removing one or more of the plurality of AlGaN buffer layers (107, 108, 109) and the GaN buffer layer (110) such that some of the plurality of AlGaN buffer layers (107, 108, 109) and the GaN buffer layer (110) remain on the first Si[100] substrate (114) together with the one or more GaN device layers (112).

2. The method of claim 1, further comprising bonding a second Si[100] substrate (118) to the one or more GaN device layers (112) on a side thereof opposite to the first Si[100] substrate (114), and removing the first Si[100] substrate (114).

3. The method of claim 2, wherein the polarity of the one or more GaN device layers (112) is inverted as a result of the bonding to the second Si[100] substrate (118) and the removing of the first Si[100] substrate (114).