Patent Grant
11869942
This application is the U.S. National Phase of PCT Appln. No. PCT/EP2018/072229 filed Aug. 16, 2018, which claims priority to European Application No. 17188124.6 filed Aug. 28, 2017, the disclosures of which are incorporated in their entirety by reference herein.
The invention relates to a heteroepitaxial wafer comprising in the following order: a silicon substrate, an AlN nucleation layer, an AlGaN buffer and a GaN layer. The invention further relates to a method for producing such a heteroepitaxial wafer.
Lattice mismatch between the substrate and the layers deposited thereon and differences with respect to their thermal coefficient of expansion will lead to warping of the multi-layer structure during growth of the layers and during cooling of the layered substrate to room temperature. Warping of the wafer may induce the formation of cracks.
US 2014/0374771 A1 discloses a semiconductor multi-layer substrate, a semiconductor device and a method for manufacturing the same. The multi-layer substrate comprises a substrate made of silicon, a first warp control layer giving a predetermined warp to the substrate and a second warp control layer. An amount of increase in warp per unit thickness of the second warp control layer is smaller than an amount of increase in warp per unit thickness of the first warp control layer. An effect is obtained, namely that current leakage and current collapse can be restrained without increasing a crack density, and that production costs can be reduced.
Rowena et al. report that increasing the thickness of the buffer tends to suppress vertical leakage current and improves the breakdown performance of AlGaN/GaN high-electon-mobility (HEMT) devices (“Buffer Thickness Contribution to Suppress Vertical Leakage Current With High Breakdown Field (2.3 MV/cm) for GaN on Si” IEEE Electron Device Letters, VOL. 32, NO. 11, November 2011).
It is also known that the curvature increases when the thickness of the substrate decreases, and it was concluded accordingly that the growth of GaN on 150 mm and 200 mm substrates requires much larger compressive strain in the epitaxial films (Y.Cao et al., “GaN-on-Silicon Growth by MOCVD: A Mechanistic Approach to Wafer Scaling”; CS MANTECH Conference, May 19th-22nd 2014, Colorado, USA).
The inventors of the present invention found that there is still the need to solve various problems.
If the thickness of the AlGaN buffer is increased in order to improve the breakdown characteristics, wafer warp will increase, as well, all the more as when silicon substrates having comparatively large diameters of 150 mm or 200 mm or 300 mm and having thicknesses in accordance with the SEMI standard are used.
Large wafer warp during epi deposition causes temperature non-uniformity on the surface that ultimately results in layer thickness variation and material alloy fluctuation across the wafer diameter.
The present invention was made in order to address these problems.
The present invention is directed to a heteroepitaxial wafer comprising, in the following order:
The present invention is further directed to a method for producing a heteroepitaxial wafer, the method comprising steps in the following order:
Preferably, the previously mentioned nitride containing layers are deposited by metalorganic chemical vapor deposition (MOCVD). As precursor compounds trimethyl gallium, trimethylaluminum and ammonia can be used. Hydrogen or nitrogen is a suitable carrier gas. The growth temperature of the nitride containing layers is preferably not less than 700° C. and not more than 1160° C.
The silicon substrate is preferably a silicon single crystal wafer having a diameter of 150 mm and a thickness of not more than 1000 μm, preferably not more than 675 μm, or having a diameter of 200 mm and a thickness of not more than 1200 μm, preferably not more than 725 μm, or having a diameter of 300 mm and a thickness of not more than 1500 μm, preferably not more than 775 μm. Preferably, the silicon substrate has {111}-orientation and a crystal offcut of less than or equal to 2°. Alternatively, the silicon substrate can have the mentioned orientation and offcut and be an SOI wafer (silicon-on-insulator wafer) or a layered wafer having a silicon single crystal top layer.
The AlN nucleation layer has a thickness which is preferably not less than 100 nm and not more than 300 nm. The curvature of the substrate becomes more negative, i.e. more concave with increasing thickness of the AlN nucleation layer.
The first strain building layer is an AlGaN layer having a first average Al content z, i.e. the first strain building layer is an AlzGa1-zN layer, wherein 0<z. Alternatively, the first strain building layer can comprise more than one AlGaN layer with decreasing Al content in the growth direction. Preferably, the value of z is not less than 40% and not more than 60%. The first strain building layer forms an interface with the top surface of the AlN nucleation layer. The first strain building layer reduces the concave curvature so that the layered substrate becomes less concave under the conditions of growth of the first strain building layer. The concave curvature decreases the hinner the substrate is. Preferably, the curvature of the layered substrate remains concave in the course of growth of the first strain building layer. The thickness of the first strain building layer is preferably not less than 30 nm and not more than 200 nm.
The first strain preserving block comprises not less than 5 and not more than 50 units of a first sequence of layers. The first sequence of layers comprises an AlN layer and at least two AlGaN layers differing with regard to their Al content. Preferably, the Al content of the top surface of the first strain building layer is equal to or more than the Al content of the closest AlGaN layer of the first strain preserving block. Moreover, it is preferable that the AlN layer of the first unit forms an interface with the top surface of the first strain building layer. The Al content decreases in the first sequence of layers with increasing distance from the substrate. Preferably, one AlGaN layer of the first sequence of layers is a graded composition AlGaN layer. The grading of the Al composition can be stepwise or continuous.
According to a preferred embodiment of the invention, the first sequence of layers in the growth direction consists of an AlN layer, a thick constant composition AlGaN layer, a continuously graded composition AlGaN layer, and, optionally, a thin constant composition AlGaN layer. The thickness of the thick constant composition AlGaN layer is at least twice the thickness of the continuously graded composition AlGaN layer and at least twice the thickness of the thin constant composition AlGaN layer. The Al content of the thin constant composition AlGaN layer is equal to or similar to the Al content of the top surface of the continuously graded AlGaN layer. The curvature of the layered substrate more rapidly increases during the growth of the continuously graded composition AlGaN layer than during the growth of the thick and thin constant composition layers. The curvature of the layered substrate can be readily controlled just by adapting the Al content and the thickness of the continuously graded composition AlGaN layer.
The first strain preserving block has a second average Al content which is y, wherein y>z. Preferably, the first strain preserving block is deposited until the curvature of the layered substrate becomes zero, or until the curvature of the layered substrate becomes less concave than at the beginning of the deposition of the first strain preserving block. While depositing the first strain preserving block, the change in curvature ΔC with respect to the change of the thickness Δth of the first strain preserving block (ΔC/Δth) preferably becomes positive in the beginning of the deposition and then becomes almost ‘zero’, i.e. remains almost constant or constant or even becomes negative. The thickness of the strain preserving block does not much influence the curvature of the layered substrate. Preferably, the growth of the first strain preserving block is stopped when the value of ΔC/Δth becomes less than −5 km−1/μm.
The second strain building layer is an AlGaN layer having a third average Al content x, i.e. the second strain building layer is an AlxGa1-xN layer, wherein 0≤x<y. It is preferable that the second strain building layer forms an interface with the top surface of the first strain preserving block, wherein the third average Al content is less than the Al content of the top surface of the first strain preserving block. The second strain building layer builds compressive strain, and the thickness and the Al content of the second strain building layer are chosen such that the curvature of the layered substrate becomes close to zero, i.e. preferably less than |3|km−1 at the end of the deposition of the second strain building layer. Then, the layered substrate is almost flat after the second strain building layer has been grown. The thickness of the second strain building layer is preferably not less than 30 nm and not more than 200 nm.
The second strain preserving block comprises not less than 5 and not more than 50 units of a second sequence of layers. The second sequence of layers comprises an AlN layer and at least one AlGaN layer. It is preferable that the AlN layer of the first unit forms an interface with the top surface of the second strain building layer. The AlGaN layer of the second sequence of layers can be a constant composition AlGaN layer or a graded composition AlGaN layer, the Al decreasing with increasing distance from the substrate. The grading of the Al composition can be stepwise or continuous. According to a preferred embodiment of the invention, the AlGaN layer of the second sequence of layers is a constant composition AlGaN layer and its Al content is equal to or less than x. The second strain preserving block has a fourth average Al content which is w, wherein x<w<y.
Preferably, the second strain preserving block is deposited until the difference in curvature of the layered substrate to the beginning of the deposition of the second strain preserving block becomes negative.
The second strain preserving block does not substantially alter the curvature of the layered substrate. Preferably, the growth of the second strain preserving block is stopped when the value of ΔC/Δth is less than −1 km−1/μm, and the layered substrate stays almost flat after the second strain preserving block has been grown.
The GaN layer forms an interface with the top surface of the second strain preserving block or, alternatively, the GaN layer forms an interface with the top surface of an optional third strain building layer or, alternatively, the GaN layer forms an interface with the top surface of an optional superlattice block.
If present, the third strain building layer is an AlGaN layer having a fifth average Al content v, i.e. the third strain building layer is an AlvGa1-vN layer, wherein 0≤v<w. The third strain building layer forms an interface with the top surface of the second strain preserving block, wherein the fifth average Al content is preferably less than the Al content of the top surface of the second strain preserving block. The third strain building layer further builds up compressive strain and further increases the convex curvature of the layered substrate. The thickness of the second strain building layer is preferably not less than 30 nm and not more than 200 nm.
The implementation of strain building layers alternating with strain preserving blocks results in the provision of a comparatively thick AlGaN buffer while keeping the layered substrate almost flat, the wafer curvature not exceeding 25 km−1 and without plastic deformation occurring.
In case that the superlattice block is present, the superlattice block forms an interface with either the top surface of the second strain preserving block (i.e. the third strain building layer is not present) or with the top surface of the third strain building layer.
The superlattice block preferably comprises not less than 10 and not more than 100 layer units, each comprising an AlN layer and an AlGaN layer. The AlN layer of the first unit of the superlattice block forms an interface with either the top surface of the second strain preserving block (i.e. the third strain building layer is not present) or with the top surface of the third strain building layer. The AlN layer of the unit of the superlattice block has a thickness which is preferably not less than 5 nm and not more than 10 nm. The AlGaN layer of the unit of the superlattice block has a thickness which is preferably 2 to 10 times the thickness of the AlN layer of the unit of the superlattice block. The AlGaN layer of the unit of the superlattice block has a content of Al which is equal to or less than the Al content of the AlGaN layer which is next underneath the superlattice block. The superlattice block has a sixth average Al content which is u, wherein v<u<w. The superlattice block allows moderately increasing the convex curvature of the layered substrate.
Regardless of whether the third strain building layer and/or the superlattice block is present or not, the AlGaN buffer has an average Al content which is preferably not less than 30%. Moreover, the AlGaN buffer has a thickness which is preferably not less than 1.0 μm and not more than 5 μm.
The GaN layer can be a high purity GaN layer which is not intentionally doped and has a carbon concentration which is preferably less than 1×1018 atoms/cm3 and has a thickness which is preferably not less than 50 nm and not more than 500 nm.
Preferably, the GaN layer comprises an insulating GaN buffer containing carbon and the high purity GaN layer. In case that the insulating GaN buffer is present, the high purity GaN layer forms an interface with the insulating GaN buffer, and the insulating GaN buffer forms an interface with the second strain preserving block or the third strain building layer or the superlattice block.
The concentration of carbon in the insulating GaN buffer is preferably not less than 1×1018 atoms/cm3 and not more than 1×1020 atoms/cm3. The insulating GaN buffer has a thickness which is preferably not less than 500 nm and not more than 2000 nm.
The deposition of the GaN layer is preferably started at a point in time, when the curvature of the layered substrate is convex and not more than 50 km−1.
Regardless whether the insulating GaN buffer is present or not, the top surface of the GaN layer has a root mean square (RMS) roughness which is preferably less than 0.5 nm on 5 μm×5 μm atomic force microscope (AFM) scan area, and the full width at half maximum (FWHM) of X-ray diffraction (XRD) rocking curve for 002 and 102 reflections of the GaN layer is preferably less than 650 arcsec and 1200 arcsec, respectively.
The heteroepitaxial wafer according to the invention has excellent material properties regarding warp. For example, with an epilayer having a thickness of 4.5 μm, the epilayer is crack-free (3 mm edge exclusion considered), and the warp of the heteroepitaxial wafer is less than |50|μm.
Due to a high breakdown field which is more than 2.5 MV/cm for a vertical leakage current of 1 mA/mm2 on (111)-oriented silicon having a diameter of at least 150 mm, the heteroepitaxial wafer according to the invention can be used to build 600 V HEMT (high electron mobility transistor) devices thereon.
For this purpose, a device layer may be formed on top of the GaN layer. The device layer may comprise, for example, pGaN/AlGaN layers or SiN/AlGaN layers or SiN/AlGaN/AlN layers or GaN/AlGaN/AlN layers or GaN/AlGaN layers or an AlGaN layer.
Definitions
Layered substrate means the substrate and at least one layer deposited thereon. Epilayer means the total of the layers grown on the substrate.
AlGaN buffer means the total of layers situated above the AlN nucleation layer and below the GaN layer.
Heteroepitaxial wafer means the entity comprising the substrate, the AlN nucleation layer, the AlGaN buffer and the GaN layer.
The Al content or average Al content of a layer can be specified as percentage of Al atoms present relative of the total sum of group III atoms present.
Curvature means the reciprocal of a radius of curvature of the layered substrate or the heteroepitaxial wafer, respectively.
Warp means the difference between the maximum and minimum distances of the median surface of a wafer from a reference plane.
The invention is further explained with reference to drawings.
The silicon substrate 1 is heated from the backside (
Regardless whether the curvature is concave or convex, the temperature at the top surface varies along the radial position and the deposition conditions vary, as well. As a consequence, the material characteristics of a group-III nitride layer deposited on the top surface becomes radially inhomogeneous depending on the grade of curvature.
The present invention avoids excessive curvature of the layered substrate resulting in homogeneous material characteristics of the layers and, at the same time, enables high breakdown field.
Due to this structure, the second average Al content y of the first strain preserving block is dominated by the Al contents of the AlN layer 20 and the thick constant composition AlGaN layer 30, and consequently, is comparatively high. This results in high breakdown field, helps reducing internal stress within the GaN layer and keeps the thermal expansion rate of the constant composition AlGaN layer 30 closer to that of the silicon substrate.
A heteroepitaxial wafer according to
In principle, a p-doped AlGaN (Al content ranging from 0 to 30%) can be deposited on top of the AlGaN barrier layer 10a to fabricate e-mode HEMT devices.
A heteroepitaxial wafer with a structure according to
The dispersion performance of the AlGaN buffer was measured using the methodology as described by S. Stoffels et al. (The physical mechanism of dispersion caused by AlGaN/GaN buffers on Si and optimization for low dispersion, 2015 IEEE International Electron Devices Meeting (IEDM), 2015, p. 35.4.1-35.4.4). Metal pads 11 with a spacing of 10 μm have been deposited as shown in
Next, the breakdown performance was examined. For this purpose, a further heteroepitaxial wafer was produced having the AlGaN buffer structure according to
The heteroepitaxial wafer had the following structure:
The total thickness of the epi layer was 4.5 μm (200 nm AlN layer 2, 1800 nm layers 3, 4, 5 and 6, 1200 nm superlattice block 8, 1000 nm carbon doped GaN layer 9a, and 300 nm unintentionally doped GaN layer 9b.
For the purpose of comparison, a heteroepitaxial wafer not representing the invention was grown which differed to the further heteroepitaxial wafer merely with regard to the first and second strain preserving blocks 4 and 6 which had been omitted. To compensate for the first and second strain preserving blocks 4 and 6, the thickness of the superlattice block 8 had been enlarged, respectively. The total thickness of the epi layer was also 4.5 μm (200 nm AlN layer 2, 100 nm AlGaN 3, 2900 nm superlattice block 8, 1000 nm carbon doped GaN layer 9a, and 300 nm unintentionally doped GaN layer 9b.
At 600 V bias, the leakage current obtained from the further heteroepitaxial wafer was almost an order less than the leakage current obtained from the heteroepitaxial wafer used as a comparative example. The further heteroepitaxial wafer suffered a hard breakdown of the AlGaN buffer not up to 1500 V.
| Number | Date | Country | Kind |
|---|---|---|---|
| 17188124 | Aug 2017 | EP | regional |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/EP2018/072229 | 8/16/2018 | WO |
| Publishing Document | Publishing Date | Country | Kind |
|---|---|---|---|
| WO2019/042782 | 3/7/2019 | WO | A |
| Number | Name | Date | Kind |
|---|---|---|---|
| 7115896 | Guo | Oct 2006 | B2 |
| 7652282 | Yanagihara | Jan 2010 | B2 |
| 8247796 | Makabe | Aug 2012 | B2 |
| 8247842 | Sato | Aug 2012 | B2 |
| 8264001 | Sato | Sep 2012 | B2 |
| 8742396 | Ooshika | Jun 2014 | B2 |
| 8748866 | Lee | Jun 2014 | B2 |
| 8946773 | Tak | Feb 2015 | B2 |
| 8963164 | Shimizu | Feb 2015 | B2 |
| 8981382 | Gao | Mar 2015 | B2 |
| 9112077 | Liao | Aug 2015 | B1 |
| 9184242 | Kaneko | Nov 2015 | B2 |
| 9202873 | Harada | Dec 2015 | B2 |
| 9679974 | Hikosaka | Jun 2017 | B2 |
| 10211297 | Liu | Feb 2019 | B2 |
| 10256368 | Eum | Apr 2019 | B2 |
| 10418240 | Lin | Sep 2019 | B2 |
| 10586701 | Shikaguchi | Mar 2020 | B2 |
| 10727303 | Ikuta | Jul 2020 | B2 |
| 20080203382 | Yanagihara | Aug 2008 | A1 |
| 20110272665 | Yamaguchi et al. | Nov 2011 | A1 |
| 20120161152 | Miyoshi et al. | Jun 2012 | A1 |
| 20130092953 | Miyoshi et al. | Apr 2013 | A1 |
| 20140374771 | Umeno et al. | Dec 2014 | A1 |
| 20170154986 | Su et al. | Jun 2017 | A1 |
| 20170323960 | Sato et al. | Nov 2017 | A1 |
| Number | Date | Country |
|---|---|---|
| 2819152 | Dec 2014 | EP |
| 201631635 | Sep 2016 | TW |
| Entry |
|---|
| Y. Cao et al., “GaN-on-Silicon Growth by MOCVD: A Mechanistic Approach to Wafer Scaling”, CS Mantech Conference, May 19-22, 2014, Colorado, USA, pp. 261-264. |
| S. Stoffels et al., “The physical mechanism of dispersion caused by AlGaN/GaN buffers on Si and optimization for low dispersion,” 2015 IEEE international Electron Devices Meeting (IEDM), 2015. 4 pages. |
| Rowena et al., IEEE Electron Device Letters, vol. 32, No. 11, Nov. 2011. 1 page. |
| Number | Date | Country | |
|---|---|---|---|
| 20200203485 A1 | Jun 2020 | US |
