Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
METHOD OF REDUCING SEMICONDUCTOR SUBSTRATE SURFACE UNEVENNESS
Document Type and Number:
WIPO Patent Application WO/2019/164449
Kind Code:
A1
Abstract:
Disclosed is a method of reducing surface unevenness of a semiconductor wafer (100). In a preferred embodiment, the method comprises: removing a portion of a deposited layer and a protective layer thereon using a first slurry to provide an intermediate surface (1123). In the described embodiment, the deposited layer includes an epitaxial layer (112) and the protective layer includes a first dielectric layer (113). The first slurry includes particles with a hardness level the same as or exceeding that of the epitaxial layer (112). A slurry for use in wafer fabrication for reducing surface unevenness of a semiconductor wafer is also disclosed.

Inventors:
ZHANG, Li (1 CREATE Way #10-01 CREATE Tower, Singapore 2, 138602, SG)
LEE, Kwang Hong (1 CREATE Way #10-01 CREATE Tower, Singapore 2, 138602, SG)
NG, Keith Cheng Yeow (1 CREATE Way #10-01 CREATE Tower, Singapore 2, 138602, SG)
LEE, Kenneth Eng Kian (1 CREATE Way #10-01 CREATE Tower, Singapore 2, 138602, SG)
FITZGERALD, Eugene A. (77 Massachusetts Avenu, Cambridge Massachusetts, 02139, US)
CHUA, Soo Jin (Faculty of Engineering Department of Electrical and Computer Engineering, 21 Lower Kent Ridge Road, Singapore 7, 119077, SG)
TAN, Chuan Seng (50 Nanyang Avenue, Singapore 8, 639798, SG)
Application Number:
SG2019/050093
Publication Date:
August 29, 2019
Filing Date:
February 19, 2019
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
MASSACHUSETTS INSTITUTE OF TECHNOLOGY (77 Massachusetts Avenue, Cambridge, Massachusetts, 02139, US)
NANYANG TECHNOLOGICAL UNIVERSITY (50 Nanyang Avenue, Singapore 8, 639798, SG)
NATIONAL UNIVERSITY OF SINGAPORE (21 Lower Kent Ridge Road, Singapore 7, 119077, SG)
International Classes:
H01L21/302; B24B1/00; C09G1/02
Attorney, Agent or Firm:
POH, Chee Kian, Daniel (Marks & Clerk Singapore LLP, Tanjong Pagar,,P O Box 636, Singapore 6, 910816, SG)
Download PDF:
Claims:
Claims

1. A method of reducing surface unevenness of a semiconductor substrate, the method comprising:

removing a portion of a deposited layer and a protective layer thereon using a first slurry to provide an intermediate surface, the first slurry having particles with a hardness level the same as or exceeding that of the deposited layer.

2. The method of claim 1 , wherein the deposited layer is epitaxially deposited, or formed by physical vapour deposition or chemical vapour deposition.

3. The method of claim 1 or 2, comprising, after the portion of the deposited layer and the protective layer is removed, depositing more of the protective layer to the intermediate surface and removing a further portion of the deposited layer and the protective layer using the first slurry to form a further intermediate surface.

4. The method of any one of the preceding claims, wherein the method further comprises depositing a further protective layer on the intermediate surface or the further intermediate surface; and removing a portion of the further protective layer using a second slurry.

5. The method of claim 4, wherein the particles are not found in the second slurry.

6. The method of claim 4 or 5, wherein the protective layers have the same material.

7. The method of any one of the preceding claims, wherein the particles are diamond particles.

8. The method of any one of the preceding claims, wherein the particles have a concentration of more than about 2% and not more than about 10%.

9. The method of any one of the preceding claims, wherein the first and/or further protective layer is a dielectric layer.

10. The method of any one of the preceding claims, wherein the deposited layer is a semiconductor layer, metal layer, or dielectric layer.

1 1. The method of any one of the preceding claims, wherein the deposited layer includes one of gallium nitride (GaN), silicon dioxide (S1O2), silicon nitride (SiN), copper (Cu) and tungsten (W).

12. The method of any one of the preceding claims, wherein the intermediate surface is provided in part by a residual portion of the protective layer.

13. The method of any one of claims 1 to 1 1 , further comprising:

removing a residual portion of the protective layer, wherein a remaining portion of the deposited layer provides the intermediate surface.

14. The method of any one of the preceding claims, further comprising: forming a bonding surface from the intermediate surface or the further intermediate surface.

15. The method of any one of claims 1 to 13, further comprising etching the intermediate surface or further intermediate surface to transfer a resist pattern to the epitaxial layer.

16. A semiconductor substrate obtained by the method of any one of the preceding claims.

17. A slurry for use in substrate fabrication for reducing surface unevenness of a semiconductor substrate, the semiconductor substrate having a deposited layer and a protective layer, the slurry including particles with a hardness level the same as or exceeding that of the deposited layer to enable the slurry to remove a portion of the deposited layer and the protective layer thereon to provide an intermediate surface.

18. The slurry of claim 17, wherein the particles are diamond particles.

19. The slurry of claim 17 or 18, wherein the predetermined concentration is more than about 2% and not more than about 10%.

Description:
Method of reducing semiconductor substrate surface unevenness

Technical field

The present invention relates to a method of reducing surface unevenness, a semiconductor substrate obtained by the method, and a slurry for use in semiconductor substrate fabrication.

Background

For the growth of semiconductor devices, often it was discovered that surface imperfections such as voids and surface protrusions exist on the semiconductor materials. The size of the surface imperfections could range from nanometre scale to millimetre scale and they are generated or formed by different mechanisms. Pre existing particles from ex situ sources such as epitaxy tool, process environment and substrate manufacturing alter the local gas flow, temperature and chemistry during the epitaxy process. The growth imperfections around the particles manifest itself as surface imperfections. In the case of GaN epitaxial growth on Si substrate, pre existing particles on the Si substrate create openings in the AIN nucleation layer. In the subsequent growth of GaN, a eutectic reaction between Ga and Si takes place at the openings. This is commonly known as melt-back etching which is characterised by a void in the surface of the wafer (or the semiconductor substrate) and a large surrounding area of materials with surface protrusions containing polycrystalline Ill- nitride and Si eutectics. Particles could also be generated in situ. Although most epitaxy tools are designed to minimize gas phase nucleation during the epitaxy process, particles that are formed in the gas phase reaction are unavoidable. Usually, these particles acting as nucleus for polycrystalline nucleation are the sources of surface protrusions too. Furthermore, protrusions could also be created during the epitaxial growth without particles. In the epitaxy process of Ill-nitride, hillocks or hexagonal voids could be created due to crystallographic defects such inversion domain boundaries, stacking faults and threading dislocations.

In general, surface imperfections in semiconductors reduce the fabrication yield in the device processing and pose concerns on the reliability of these devices. For example, protrusions prevent the direct contact of the two wafers in a wafer bonding process and create unbonded areas which are order of magnitude greater than the protrusions themselves. Surface voids acting as traps of gases and impurities could create unbonded areas too. Usually, the voids can be passivated with deposition and planarization of insulating dielectric material such as S1O2 and SbN 4 . Flowever, the removal of surface protrusions is difficult.

A conventionally fabricated GaN-on-Si wafer typically includes surface protrusions including both melt-back sites and hillock sites. Figures 1 (a) and (b) are the height profiles of a melt-back site and a hillock site of such a GaN-on-Si wafer respectively. The total thickness of the Ill-nitride epitaxial layers on the Si substrate is 3.7 pm. A melt-back site consists of a central void region in the surface of the wafer and a surrounding region with rough protrusions. The hillock structure is characterized by a protrusion with a well-defined hexagon with a smooth top surface. Both protrusion types are about 4 pm higher than the surface of the GaN layer.

During a hydrophilic Si02-Si02 wafer bonding process, S1O2 is deposited conformably by a PECVD process on the GaN-on-Si wafer. The height difference caused by surface protrusions will remain after the PECVD process as illustrated in the schematic drawing in Figure 2(a). A CMP step which employs a normal S1O2 slurry is carried out to remove part of the S1O2 layer to reduce the surface roughness. However, the S1O2 slurry is not effective at removing the surface protrusions such as the hillock and melt-back sites which are made of Ill-nitride and Si based materials, as illustrated in Figure 2(b). In the subsequent wafer bonding process, large unbonded regions are created due to the surface protrusions as illustrated in Figure 2(c). In principle, the unbonded areas can be reduced if a thick S1O2 layer (more than a few times the thickness of the GaN layer) is deposited and/or many iterations of S1O2 deposition + CMP process are applied.

The prior-art bonding process suffers from the following shortcomings:

(1 ) The additional S1O2 deposition exerts more stress on the GaN wafer and may cause plastic deformation of the epitaxial layer and the substrate, especially at the edge of the GaN-on-Si wafer;

(2) The wafer bow change is increased by the additional S1O2 layer, so that the non-uniformity in the CMP process will induce non-uniformity in the final bonding oxide thickness;

(3) Many iterations of S1O2 deposition and CMP process are time consuming; and

(4) It is difficult to avoid unbonded area if the density of surface protrusions is high (a few hundred per 200 mm diameter wafer). It is desirable to provide a method, a semiconductor substrate and a slurry which addresses at least one of the drawbacks of the prior art and/or to provide the public with a useful choice.

Summary

According to one aspect, there is provided a method of reducing surface unevenness of a semiconductor substrate, the method comprising: removing a portion of a deposited layer and a protective layer thereon using a first slurry to provide an intermediate surface, the first slurry including particles with a hardness level the same as or exceeding that of the deposited layer.

In a described embodiment, a portion of the deposited layer and the protective layer is removed using the first slurry so that the intermediate surface has reduced surface imperfections, such as melt-back and hillock sites, for the deposition of a second dielectric layer thereon. The first slurry includes particles with a hardness level the same as or exceeding that of the deposited layer to enable the first slurry to mechanically remove the portion of the deposited layer and the protective layer to provide the intermediate surface.

It may well be that this removal step may not remove‘any’ surface imperfection fully. For example, if there were twenty surface imperfections (of any kind) on the substrate at the start, there may still be twenty after this first removal step, although these twenty surface imperfections should be reduced in height/relief, preferably to a point that a second protective (e.g. dielectric) layer deposition step may be used to fully contact and cover the reduced imperfections. It is also possible that if the relief of the twenty surface imperfections is still too high, another protective (e.g. dielectric) layer deposition plus the first slurry removal step may be carried out; this may be needed for a starting surface that has an especially high density of defects, and more so if they are very large. In this respect, it is desirable to ‘polish’ the intermediate surface so that the size of the imperfections may be reduced to achieve better bonding.

The deposited layer may be epitaxially deposited, or may be formed by physical vapour deposition or chemical vapour deposition.

If there is a need to repeat the deposition and polishing steps, the method may further comprise, after the portion of the deposited layer and the protective layer is removed, depositing more of the protective layer to the intermediate surface and removing a further portion of the deposited layer and the protective layer using the first slurry to form a further intermediate surface.

As discussed above, the method may include depositing a further protective layer in the form of a second dielectric layer on the intermediate surface and removing a portion of the second dielectric layer using a second slurry to provide an improved bonding surface with reduced surface imperfections, thereby achieving a relatively even and strong bond. Yield loss attributed to surface imperfections may thus be reduced. Preferably, the particles in the first slurry may not found in the second slurry. The protective layers may have the same material. Advantageously, the particles in the first slurry are diamond particles.

Specifically, the particles of the first slurry may have a concentration of more than about 2% and not more than about 10%. In a specific example the first and/or further protective layer may be a dielectric layer. Preferably, the deposited layer may be a semiconductor layer, metal layer, or dielectric layer.

More specifically, the deposited layer may include one of gallium nitride (GaN), silicon dioxide (S1O2), silicon nitride (SiN), copper (Cu) and tungsten (W).

The intermediate surface may be provided in part by a residual portion of the protective layer. For example, the intermediate surface may be provided by a small portion of the deposited layer and the residual portion of the protective layer. Alternatively, the method may include removing the residual portion of the protective layer, wherein a remaining portion of the deposited layer provides the intermediate surface. For example, the remaining portion of the deposited layer includes said small portion and other exposed portions of the deposited layer.

The method may further include forming a bonding surface from the intermediate surface. The bonding surface may, for example, be for bonding with another semiconductor layer to form a wafer. According to another aspect, there is provided a slurry for use in substrate fabrication for reducing surface unevenness of a semiconductor substrate, the semiconductor substrate having a deposited layer and a protective layer, the first slurry including particles with a hardness level the same as or exceeding that of the deposited layer to enable the first slurry to remove a portion of the deposited layer and the protective layer thereon to provide an intermediate surface.

In a described embodiment, the particles are in the form of diamond particles, and a portion of the protective layer and an underlying deposited layer may be removed to provide the intermediate surface for, for example, the deposition of another protective (e.g. dielectric) layer thereon. The method is applicable to a wafer bonding process, and may also be applicable to other semiconductor processes, to reduce yield loss caused by surface unevenness or imperfections. As a result, the intermediate surface may be used as a bonding surface or the intermediate surface may be etched to transfer a resist pattern to the deposited layer.

It is envisaged that features relating to one aspect may be applicable to the other aspects.

Brief Description of the drawings

Example embodiments will now be described hereinafter with reference to the accompanying drawings, wherein like parts are denoted by like reference numerals. Among the drawings:

Figures 1 (a) and 1 (b) are height profile graphs of a melt-back site and a hillock site of a conventionally fabricated GaN-on-Si wafer, respectively; Figures 2(a), 2(b) and 2(c) are schematic representations of a semiconductor wafer at respective stages of a conventional wafer bonding process;

Figures 3(a) to 3(d) are schematic representations of a semiconductor wafer at respective stages of a wafer bonding process according to an embodiment of the present invention; and

Figures 4(a) and 4(b) show height profile graphs of a melt-back site and a hillock site, respectively, on a wafer surface processed according to the embodiment of Figures 3(a) to 3(d).

Detailed Description

Figures 3(a) to 3(d) are schematic illustrations of respective stages of a method of reducing wafer surface unevenness (e.g., surface protrusions) or generally surface unevenness of a semiconductor substrate according to one example embodiment.

Figure 3(a) illustrates a semiconductor wafer 100 formed from a known PECVD process and including a substrate layer 11 1 of Si, a deposited layer in the form of an epitaxial layer 1 12 of GaN arranged on the substrate layer 1 1 1 , and a protective layer in the form of a first dielectric layer 1 13 of S1O2 deposited on the epitaxial layer

1 12.

As illustrated in Figure 3(b), in a first step of the method, a portion of the epitaxial layer 1 12 and the first dielectric layer 1 13 thereon are removed using a first slurry (not shown) to provide an even intermediate surface 1 123. That is, the epitaxial layer 1 12 and the first dielectric layer 1 13 thereon are partially removed using the first slurry. In other words, the first step includes removing the portion of the deposited layer (i.e., the epitaxial layer 1 12) and the protective layer (i.e., the first dielectric layer 1 13) thereon using the first slurry to provide the intermediate surface 1 123. In this step, which may be a chemical-mechanical polishing (CMP) or mechanical partial-removing step, the first slurry has a predetermined concentration of diamond particles for mechanically removing the portion of the epitaxial layer 1 12 and the first dielectric layer 1 13, thereby reducing protrusions or imperfections in the epitaxial layer 1 12 and the first dielectric layer 1 13. The first slurry includes a first slurry base material which may or may not have a contribution towards the partial removal of the first dielectric layer 113. The predetermined concentration of the diamond particles in this embodiment is more than about 2% and not more than about 10%. It should be noted that the first slurry may, in other embodiments, have a predetermined concentration of other suitable particles, as long as the particles have a hardness level the same as or exceeding that of the epitaxial layer 1 12. SiC particles are such other suitable particles, for example. The concentration of diamond particles in the first slurry has a substantially uniform removal rate with respect to the material of the epitaxial layer 112 (i.e., GaN).

Initially, the CMP removal rate (RR) is higher (i.e., faster removal) at the protrusion materials because they protrude from the first dielectric layer 1 13. Once the protrusion materials are level with the first dielectric layer 1 13, the CMP removal rate is about the same over the GaN or any protrusion materials as well as the first dielectric layer 1 13, so that after the CMP process the protrusion materials are at the same level as the first dielectric layer 1 13’ (see Figure 3(b)). In other words, starting at Figure 3(a), the polishing rate/material RR at the protrusions is higher (i.e., faster removal) than the RR in the remaining“field area” (i.e. protrusion-free regions, or regions where there are no protrusions). The protrusions that are“sticking out” will be abraded more by the first slurry particles as they would be in“closer” contact with the polishing or removing surface. This operation results in a surface similar to that depicted in Figure 3(b), as the RR rate starts to even out as the protrusions are gradually removed and the surface relief becomes reduced. This is because the diamond particles in the first slurry achieve their removal/polishing action or effect mostly through a physical process (i.e. mostly abrasion), with little to no chemical selectivity (i.e. preferential etching/dissolution of one material over another).

The melt-back and hillock sites may not be removed completely. They are substantially reduced to the same level as the first dielectric layer 1 13’ (see Figure 3(b)) after the CMP using the first slurry. This step can be repeated until the surface is flat and until a desired height of the protrusion materials is reached. In particular, the first dielectric layer 1 13 and the surface protrusions (melt-back sites and hillock sites, marked by“M” and“H” in Figure 3(a), respectively) are reduced in thickness or height from 4 pm to less than 100 nm as illustrated in Figures 4(a) and 4(b). As shown in Figure 3(b), small, residual portions 1 13’ of the first dielectric layer 113 remain and small portions 1 12’ of the epitaxial layer 1 12 are exposed. Together, the portions 1 12’, 113’ provide or define the even intermediate surface 1 123. Following the first step, any residual portions of the first slurry is rinsed off. Depending on the original heights of the reduced melt-back and hillock sites, the first step and the preceding PECVD oxide deposition process may be repeated. It should be appreciated that after the first step, substantial protrusion materials would have been removed resulting in a smoother intermediate surface 1123. Specifically, it might be possible to reduce 4 pm protrusions down to around 100 nm and this would result in the intermediate surface 1 123 usable for etching/deposition processes. However, the intermediate surface 1 123 may still not be sufficiently smooth (due to fast RR, high and non-optimized concentration of diamond particles) and unsuitable for bonding and in this case, the method could include further steps as explained below to smoothen the surface further.

In a second step of the method, referring to Figure 3(c), a further protective layer in the form of a second dielectric layer 114 of S1O2 is deposited on the intermediate surface 1 123 by means of a PECVD oxide deposition process. The first and second dielectric layers 1 13, 114 may have the same or different materials. That is, the second step includes depositing the further protective layer 114 to the intermediate surface 1 123.

In a third step of the method, which is a CMP step, a portion of the second dielectric layer 1 14 is removed using a second slurry, which, in this embodiment, does not contain the predetermined concentration of particles. As a result of the third step, surface roughness and/or unevenness (e.g., protrusions) of the deposited second dielectric layer 1 14 is reduced, thereby providing a further intermediate surface 1141 for bonding of the wafer 100 with another wafer 200, as depicted in Figure 3(d). That is, the third step includes removing the portion of the deposited further protective layer 114 to form the further intermediate surface 1141 , and the bonding surface is formed from the further intermediate surface 1 141 in this embodiment. It should be noted that in other embodiments, the third step may be a non-CMP step requiring no slurry for reducing the second dielectric layer 1 14. The second and third steps may optionally be repeated at least once, with the second step of each new iteration resulting in a yet further protective layer deposited on the further intermediate surface of the third step of the previous iteration, and with the third step of each new iteration resulting in the yet further protective layer partially removed using the second slurry or a different slurry to provide a yet further intermediate surface for either another iteration, bonding with the wafer 200, or any other processing.

In this embodiment, the second slurry includes a second slurry base material and includes silicon dioxide (S1O2) particles for removing partially the second dielectric layer 1 14. The first slurry in this embodiment is based on the second slurry (i.e., having the second slurry as the first slurry base material and hence containing S1O2 particles) and, as described hereinabove, further has the predetermined concentration of diamond particles which, in this embodiment, is not found in the second slurry.

The first slurry is thus effective in partially removing the first dielectric layer 1 13 and the epitaxial layer 112 to provide the intermediate surface 1 123, and the second slurry in partially removing the second dielectric layer 1 14 deposited on the intermediate surface 1 123 in order to polish the second dielectric layer 1 14. In other embodiments, however, the first and second slurries may have different bases, with the first slurry containing particles, diamond or non-diamond, having a hardness level the same as or exceeding that of the epitaxial layer 1 12.

To elaborate further, considerations for the two slurries in this embodiment are:

a) the first slurry - configured to be able to abrade the protrusion material (i.e. hardness criterion) present in the epitaxial layer 1 12, using predominantly physical/mechanical means (as opposed to chemical), namely the diamond particles, so as to ensure that the protrusions (as they stick out/protrude) are removed at a higher rate than the first dielectric layer 1 13 displaced over the regions of the wafer being protected, resulting in the surface obtained in Figure 3(b); and

b) the second slurry - configured to be able to polish the second dielectric layer 1 14 to provide the further intermediate surface 1 141 of sufficient evenness or flatness for, for example, bonding with another wafer 200 (typically involving both physical and chemical activities, hence the term chemical-mechanical polishing).

Thus, the first slurry may or may not be based on the second slurry.

The dielectric layers 1 13, 1 14 are made of S1O2 in this embodiment, and can be made of a different material (e.g., Si3N 4 ) in other embodiments (which may mean using a different second slurry for the polishing step or the third step). That is to say, the dielectric layers 1 13, 1 14 may be made of different, respective materials, as long as the first dielectric layer 1 13 can be removed by the first slurry, and the second dielectric layer 1 14 can be removed by the second slurry or other suitable means. For example, where the first dielectric layer 1 13 is made of SiN and the second dielectric layer 1 14 is made of Si02, the first and second slurries may have different base materials. A first slurry based on SiN particles may be used to smoothen the first dielectric layer 1 13 made of SiN, while a second slurry based on Si02 particles may be used to smoothen the second dielectric layer 1 14 made of Si02. The first slurry can be produced by weighing the desired concentration or amount of diamond particles, and mixing and stirring the diamond particles in the first slurry base material to form the first slurry. Thus, a skilled person would appreciate that one difference with respect to the prior art resides in the insertion of an additional CMP step (i.e., the first step) employing a slurry, i.e., the first slurry, with a concentration of diamond particles of, for example, more than about 2% not more than about 10% mixed in a S1O2 slurry. The concentration range and the type of particles used in the first slurry may be otherwise in other embodiments. The deposited layer may be a semiconductor layer, a metal layer, or a dielectric layer. The deposited layer may include one of gallium nitride (GaN), silicon dioxide (S1O2), silicon nitride (SiN), copper (Cu) and tungsten (W).

With the method, surface unevenness or imperfection attributed to melt-back and hillock sites of significant diameters (>1 mm in diameter) may be substantially reduced. Indeed, it may be observed through infrared image comparison, for example, that the method of Figure 3 is able to reduce significant surface unevenness to such a point that the bond quality may be as good as bonding wafers with negligible starting/initial defects.

In the embodiment of Figure 3, the intermediate surface 1 123 is provided in part by the residual portions 1 13’ of the first dielectric layer 1 13. In an alternative embodiment, the method may further include removing the residual portion 113’ of the first dielectric layer 1 13 completely by any suitable means so that the intermediate surface 1 123 is formed from or provided by a remaining portion of the epitaxial layer 112. For example, the residual portion 1 13’ may be removed by wet- etching or dry-etching, preferably with chemical selectivity between the residual portion 1 13’ and the deposited epitaxial layer 1 12. The remaining portion of the epitaxial layer 1 12 includes the small portions 1 12’ and other portions of the epitaxial layer 1 12 now exposed with the removal of the residual portion 1 13’.

In addition, in the embodiment of Figure 3, the bonding surface is formed from the further intermediate surface 1141 (or the yet further intermediate surface if the second and third steps are repeated). It is envisaged that it may be possible for the intermediate surface 1123 to be used as the bonding surface for IC fabrication after the first step (i.e. without the second and third steps) but the bonding yield is unlikely to be good. Indeed, it has been found that if the residual portions of 113’ are removed completely as discussed above, the presence of -100 nm protrusions might mean that the bonding yield would be pretty poor (although better than if the first step was not performed at all). On the other hand, with the residual portion 1 13’ in place on the intermediate surface 1 123, this might lead to better or more decent outcomes. To achieve better results, performing the second and third steps of the method to get to the further intermediate surface 1 141 as the bonding surface is preferred.

Indeed, it is envisaged that the described embodiment may not be restricted to using the intermediate surface 1 123 (or the further intermediate surface 1 141 ) for bonding but the surface may be subject to other types of processing (such as general etching or deposition processes), regardless of whether the residual portions 1 13’ of the first dielectric layer 1 13 are removed. The reduction of surface defects (e.g., the flattening of protrusions) enhances the uniformity (and hence the processing yield) of any subsequent semiconductor processes. Such processes include, for example: deposition of thin films on the intermediate surface 1 123 via methods such as chemical vapour deposition (CVD), physical vapour deposition (PVD) or atomic layer deposition (ALD); and etching of the intermediate surface 1123 via plasma etching (e.g. reactive ion etching, inductively-coupled plasma reactive ion etching (ICP-RIE), ion beam etching (IBE) etc.) to transfer a resist pattern to the epitaxial layer 112, or deposition of layers/films using spin-coating techniques (e.g. spin-coating photoresist, and spin-on glass).

For example, the residual portion 1 13’ of the first dielectric layer 1 13 can be removed completely after the first step (Figure 3(b)) if a process is to be performed with respect to the underlying epitaxial layer 1 12 (e.g., depositing a metal film directly on the epitaxial layer 112 to form a contact). This may also be done to characterize the surface of the epitaxial layer 1 12 after the first step. Flowever, the residual portion 1 13’ may not have to be removed in some circumstances, for example, where a step after the first step is to deposit a dielectric passivation layer or a lithography hardmask on the intermediate surface 1 123.

The method of reducing surface unevenness of a semiconductor wafer or substrate offers at least the following advantages:

(1 ) Excellent ability to reduce any surface protrusions on the wafer surface or the substrate surface as the first slurry is rate non-selective over a wide range of materials; (2) No damage is introduced to the wafer surface or the substrate surface except to the protrusions as the residual S1O2 layer (i.e., the portions 1 13’) serves as a surface protection layer to the underlying GaN surface; and

(3) The process is cost-effective as the process may involve as few as one additional reducing step (the first step) followed, in some embodiments, by one additional S1O2 deposition (the second step).