Semiconductor Structures

Abstract

A semiconductor device comprises a substrate, one or more first III-semiconductor layers, and a plurality of superlattice structures between the substrate and the one or more first layers. The plurality of superlattice structures comprises an initial superlattice structure and one or more further superlattice structures between the initial superlattice structure and the one or more first layers. The plurality of superlattice structures is configured such that a strain-thickness product of semiconductor layer pairs in each superlattice structure of the one or more further superlattice structures is greater than or equal to a strain-thickness product of semiconductor layer pairs in superlattice structure(s) of the plurality of superlattice structures between that superlattice structure and the substrate. The plurality of superlattice structures is also configured such that a strain-thickness product of semiconductor layer pairs in at least one of the one or more further superlattice structures is greater than a strain-thickness product of semiconductor layer pairs in the initial superlattice structure.

Claims

1. A semiconductor structure comprising: a substrate; one or more first semiconductor layers; and a plurality of superlattice structures between the substrate and the one or more first layers, wherein the plurality of superlattice structures comprises an initial superlattice structure and one or more further superlattice structures between the initial superlattice structure and the one or more first layers; wherein the plurality of superlattice structures is configured such that a strain-thickness product of semiconductor layer pairs in each superlattice structure of the one or more further superlattice structures is greater than or equal to a strain-thickness product of semiconductor layer pairs in superlattice structure(s) of the plurality of superlattice structures between that superlattice structure and the substrate; and wherein the plurality of superlattice structures is configured such that a strain-thickness product of semiconductor layer pairs in at least one of the one or more further superlattice structures is greater than a strain-thickness product of semiconductor layer pairs in the initial superlattice structure.

2. The semiconductor structure of claim 1, wherein the plurality of superlattice structures is configured such that a strain-thickness product of semiconductor layer pairs in at least one superlattice structure of the one or more further superlattice structures is equal to a strain-thickness product of semiconductor layer pairs in a superlattice structure of the plurality of superlattice structures between the at least one superlattice structure and the substrate.

3. The semiconductor structure of claim 1, wherein the plurality of superlattice structures is configured such that the strain-thickness product of each semiconductor layer in each superlattice structure of the plurality of superlattice structures is less than a limit as defined by Equation 1: $ϵ=\frac{b}{2h_0\cos \left(\lambda \right)}\left(\frac{1}{10}+\frac{1}{4\pi }\frac{1-\nu {\cos }^2\vartheta }{1-\nu }\ln \left(\frac{h_c}{b}\right)\right).$

4. The semiconductor structure of claim 1, wherein the plurality of superlattice structures is configured such that the strain-thickness product of each layer pair in each superlattice structure of the plurality of superlattice structures is less than a limit as defined by Equation 1: $ϵ=\frac{b}{2h_0\cos \left(\lambda \right)}\left(\frac{1}{10}+\frac{1}{4\pi }\frac{1-\nu {\cos }^2\vartheta }{1-\nu }\ln \left(\frac{h_c}{b}\right)\right).$

5. The semiconductor structure of claim 1, wherein the plurality of superlattice structures is configured such that the thickness of each semiconductor layer pair in each superlattice structure of the one or more further superlattice structures is greater than the thickness of each semiconductor layer pair in superlattice structure(s) of the plurality of superlattice structures between that superlattice structure and the substrate.

6. The semiconductor structure of claim 1, wherein the plurality of superlattice structures is configured such that the strain of semiconductor layer pairs in at least one superlattice structure of the one or more further superlattice structures is greater than or equal to the strain of semiconductor layer pairs in a superlattice structure of the plurality of superlattice structures between the at least one superlattice structure and the substrate.

7. The semiconductor structure of claim 1, wherein the plurality of superlattice structures is configured such that the strain of semiconductor layer pairs in at least one superlattice structure of the one or more further superlattice structures is less than the strain of semiconductor layer pairs in a superlattice structure of the plurality of superlattice structures between the at least one superlattice structure and the substrate.

8. The semiconductor structure of claim 1, wherein the plurality of superlattice structures is configured such that the strain of each semiconductor layer pair in each superlattice structure of the plurality of superlattice structures is less than about 2%.

9. The semiconductor structure of claim 1, wherein: each semiconductor layer pair comprises a first semiconductor layer and a second semiconductor layer; and the plurality of superlattice structures is configured such that the semiconductor material and/or the composition of the semiconductor material of each first layer of one or more superlattice structure(s) of the plurality of superlattice structures is different to the semiconductor material and/or the composition of the semiconductor material of each first layer of one or more other superlattice structure(s) of the plurality of superlattice structures; and/or the plurality of superlattice structures is configured such that the semiconductor material and/or the composition of the semiconductor material of each second layer of one or more superlattice structure(s) of the plurality of superlattice structures is different to the semiconductor material and/or the composition of the semiconductor material of each second layer of one or more other superlattice structure(s) of the plurality of superlattice structures.

10. The semiconductor structure of claim 1, wherein the plurality of superlattice structures is configured such that a strain-thickness product of each superlattice structure of the plurality of superlattice structures is less than about 0.8 nm.

11. The semiconductor structure of claim 1, wherein the plurality of superlattice structures is configured such that a strain-thickness product of at least one superlattice structure of the one or more further superlattice structures is greater than or equal to a strain-thickness product of a superlattice structure of the plurality of superlattice structures between the at least one superlattice structure and the substrate.

12. The semiconductor structure of claim 1, wherein the plurality of superlattice structures is configured such that a strain-thickness product of at least one superlattice structure of the one or more further superlattice structures is less than a strain-thickness product of a superlattice structure of the plurality of superlattice structures between the at least one superlattice structure and the substrate.

13. The semiconductor device of claim 1, wherein: the plurality of superlattice structures is configured such that the number of repeats in each superlattice structure of the one or more further superlattice structures is less than or equal to the number of repeats in superlattice structure(s) of the plurality of superlattice structures between that superlattice structure and the substrate; and the plurality of superlattice structures is configured such that the number of repeats in at least one of the one or more further superlattice structures is less than the number of repeats in the initial superlattice structure.

14. The semiconductor device of claim 1, wherein the lattice constant of the semiconductor material of the one or more first layers is different to the lattice constant of the semiconductor material of the substrate.

15. The semiconductor device of claim 1, wherein the substrate is formed from silicon (Si), Germanium (Ge) or Gallium Arsenide (GaAs).

16. The semiconductor device of claim 1, wherein the one or more first layers are formed from one or more III-V compound semiconductor materials such as Gallium Antimonide (GaSb), Gallium Arsenide (GaAs), Gallium Phosphide (GaP), Gallium Nitride (GaN), Gallium Arsenide Antimonide (GaAsSb), Gallium Indium Antimonide (GaInSb), Gallium Arsenide Phosphide (GaAsP), Gallium Indium Arsenide Antimonide (GaInAsSb), Indium Arsenide (InAs), Indium Phosphide (InP), Indium Arsenide Antimonide (InAsSb), Aluminium Antimonide (AlSb), or Aluminium Indium Antimonide (AlInSb).

17. A semiconductor device comprising the semiconductor structure of claim 1.

18. The semiconductor device of claim 17, wherein the semiconductor device comprises a light-emitting device, a detecting device, and/or an electronic device.

19. A method of forming a semiconductor structure, the method comprising: forming an initial set of semiconductor layers on a substrate; and forming one or more first semiconductor layers on the initial set of semiconductor layers; wherein forming the initial set of semiconductor layers comprises forming a plurality of superlattice structures comprising an initial superlattice structure and one or more further superlattice structures; wherein forming the plurality of superlattice structures comprises forming the plurality of superlattice structures such that a strain-thickness product of semiconductor layer pairs in each superlattice structure of the one or more further superlattice structures is greater than or equal to a strain-thickness product of semiconductor layer pairs in superlattice structure(s) between that superlattice structure and the substrate; and wherein forming the plurality of superlattice structures comprises forming the plurality of superlattice structures such that a strain-thickness product of semiconductor layer pairs in at least one of the one or more further superlattice structures is greater than a strain-thickness product of semiconductor layer pairs in the initial superlattice structure.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0113] Certain preferred embodiments of the present disclosure will now be described in greater detail, by way of example only and with reference to the following figures, in which:

[0114] FIG. 1 is a representation of a two-step GaSb buffer layer grown on a Si wafer;

[0115] FIG. 2 is an illustration of misfit strain as a function of layer thickness;

[0116] FIG. 3 is a schematic illustration of a AlSb/GaSb DFSL structure in accordance with various embodiments;

[0117] FIG. 4 is a schematic illustration of the strain-thickness characteristics of the AlSb layer and the GaSb/AlSb layer pair of the first superlattice of the structure of FIG. 3;

[0118] FIG. 5 is a schematic illustration of the strain-thickness characteristics of the first GaSb/AlSb superlattice of the structure of FIG. 3;

[0119] FIG. 6 is a schematic illustration of the strain-thickness characteristics of the AlSb layer, the AlSb/GaSb layer pair and the total superlattice of the second dislocation filter structure of the structure of FIG. 3;

[0120] FIG. 7 is a schematic illustration of the strain-thickness characteristics of the AlSb layer, the GaSb/AlSb layer pairs and the total superlattice of the third and fourth dislocation filter structures of the structure of FIG. 3;

[0121] FIG. 8 is a graphical representation of the strain-thickness characteristics of the AlSb layer, the AlSb/GaSb layer pairs and the SL of all the AlSb/GaSb dislocation filter structures of the structure of FIG. 3;

[0122] FIG. 9 is a schematic illustration of a InAs/AlSb DFSL structure in accordance with various embodiments;

[0123] FIG. 10 is a graphical representation of the strain-thickness characteristics of the AlSb layer, the InAs/AlSb layer pairs and the superlattice of each of the four InAs/AlSb dislocation filter structures of the structure of FIG. 9;

[0124] FIG. 11 is a schematic illustration of a GaSb/Ga.sub.0.8In.sub.0.2Sb DFSL structure in accordance with various embodiments;

[0125] FIG. 12 is a graphical representation of the strain-thickness characteristics of the Ga.sub.0.8In.sub.0.2Sb layers, the GaSb/Ga.sub.0.8In.sub.0.2Sb layer pairs and the four GaSb/Ga.sub.0.8In.sub.0.2Sb superlattices of the structure of FIG. 11;

[0126] FIG. 13 is a schematic illustration of a GaSb/Ga.sub.xIn.sub.(1−x)Sb DFSL structure in accordance with various embodiments;

[0127] FIG. 14 is a graphical representation of the strain-thickness characteristics of the varied Ga.sub.xIn.sub.(1−x)Sb layers, the GaSb/Ga.sub.xIn.sub.(1−x)Sb layer pairs and the four GaSb/Ga.sub.xIn.sub.(1−x)Sb superlattices of the structure of FIG. 13;

[0128] FIG. 15 is a schematic illustration of a AlSb/GaSb DFSL structure in accordance with various embodiments; and

[0129] FIG. 16 is a graphical representation of the strain-thickness characteristics of the layers, layer pairs and superlattices of the structure of FIG. 15.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0130] Heteroepitaxial growth of thin films with high lattice mismatch to the underlying substrate leads to the formation of threading dislocations, which deteriorate the crystalline quality of the epilayer and hinders the performance of electrical devices.

[0131] For example, direct integration of gallium antimonide (GaSb) on group IV wafers, such as silicon (Si), is an attractive root for reducing manufacturing costs and developing fully integrated lab-on-chip mid-infrared (MIR) Si photonic circuits. However, the large lattice mismatch (˜12%) is challenging and direct epitaxial growth results in a large density (≥10.sup.10 cm.sup.−2) of threading dislocations and planar defects.

[0132] It has been previously reported that the growth of GaSb on a Si wafer using a thin aluminium antimonide (AlSb) nucleation layer enables the confinement of dislocations at the interface via the formation of a network of interfacial misfit dislocation (IMF) arrays. However, a substantial number of defects remain in the epilayer, which propagate from the lower parts of the epitaxial layer to the top, leading to a surface dislocation density of the order of 10.sup.9 cm.sup.−2. This value is approximately four orders of magnitude higher than the estimated defect limit of ˜10.sup.5 cm.sup.−2 for GaSb integrated on Si.

[0133] Various embodiments provide a new method for growing high crystalline quality epilayers on a mismatched substrate using a series of strained dislocation filters. The dislocation filter layers act as blocking barriers to the vertical propagation of threading dislocations, leading to a surface defect density of the order of 10.sup.6 cm.sup.−2.

[0134] Previous work on the heteroepitaxial integration of high quality GaSb on Si was based on the growth of a thick GaSb layer on 4 degrees offcut Si wafers using a 17 monolayer (ML) thin AlSb IMF nucleation layer and a two-temperature step growth procedure. The use of misoriented Si wafers suppresses the formation of planar defects such as antiphase domains (APDs).

[0135] FIG. 1 is a representation of the two-step GaSb buffer layer. The two-step growth technique comprises the deposition of a 1.5 μm thick GaSb layer using a growth temperature of 487° C., followed by the growth of another 500 nm of GaSb while gradually increasing the growth temperature up to 515° C.

[0136] This procedure significantly improves the layer quality, resulting in a surface dislocation density of 2×10.sup.8 cm.sup.−2. A variety of antimonide-based semiconductor material systems can then be grown on top of the buffer layer, based on the application and e.g. the desired operation wavelength.

[0137] Initially, the dislocation density is expected to reduce with increasing buffer layer thickness. This is due to a reaction between threading dislocations as they approach an area where interaction is energetically favorable. However, previous reports have shown that for a constantly incremental change in the thickness of the buffer layer, the probability of two dislocations being placed in the same interaction area is significantly decreased. Therefore, reducing the dislocation density lower than the order of 10.sup.8 cm.sup.−2 using a simple buffer layer is very unlikely.

[0138] Thus, it has been recognized that lateral overgrowth of dislocation filter superlattices (SL) comprising alternating semiconductor strained layers is important to further decrease the number of threading dislocations reaching the surface of the structure and/or the active region of a device. The additional interfacial misfit strain promotes movement and glide of threading dislocations, which triggers defect recombination. This is considered to be the principal method to achieve defects densities of the order of 10.sup.6 cm.sup.−2 or lower, greatly reduced compared to what is feasible using only thick buffer layers.

[0139] Strained layer dislocation filter structures usually comprise a number of superlattices separated by spacer layers. Each superlattice comprises a pair of layers (a layer pair), layer 1 of thickness h1 and layer 2 of thickness h2, which is repeated t times, where t=1, 2, 3 etc. Spacer layers can be grown after each superlattice structure to help relieve the total residual strain introduced by the alternative strained layers of the underlying superlattice. A key advantage of a superlattice over a bulk strained layer is that the interfacial misfit strain values are significantly lower than that present at the interface with the substrate and can be used repeatedly; thus increasing the effectiveness of dislocation blocking.

[0140] The interfacial misfit strain generated between layers of alternative materials can be used to force moving and bowing of threading dislocations. A misfit strain of up to 1.5-2% can be accommodated by uniform elastic strain of the epilayer. However, if the strain introduced during the epitaxial growth is sufficiently large or if the epilayer exceeds a critical thickness, further defects can be generated or start to move by glide in order to relieve the strain.

[0141] The concept of the critical thickness is described by the Matthews equation:

[00003]$\begin{array}{cc}ϵ=\frac{b}{2h_0\cos \left(\lambda \right)}\left(\frac{1}{10}+\frac{1}{4\pi }\frac{1-\nu {\cos }^2\vartheta }{1-\nu }\ln \left(\frac{h_c}{b}\right)\right),& \left(1\right)\end{array}$

[0142] where ε is the misfit strain, h.sub.c is the critical thickness, λ is the angle between the slip direction and the direction of the layer plane which is perpendicular to the intersection of the slip plane and the surface, b is the magnitude of the Burger vector, ν is the Poisson ratio, and ϑ is the angle between the dislocation line and its Burger vector.

[0143] The misfit strain of the layer is given by the equation:

[00004]$\begin{array}{cc}\epsilon =\frac{{\alpha }_j-{\alpha }_s}{{\alpha }_s}\text{.100}\%,& \left(2\right)\end{array}$

[0144] where α.sub.i is the lattice constant of the layer, and α.sub.s is the lattice constant of the substrate, the buffer layer or the spacer layer of a superlattice.

[0145] In general, the thickness of the layer is significantly lower than the underlying thick material layer (substrate, buffer or spacer). For a film grown on a (001) film plane, dislocations with Burger vector of the type

[00005]$\frac{a}{2}<110>$

are assumed. The magnitude of the Burger vector is given by the equation:

[00006]$\begin{array}{cc}.Math.b.Math.=\frac{a}{2}\sqrt{h^2+k^2+l^2}.& \left(3\right)\end{array}$

[0146] For antimonide (Sb) materials (GaSb, InAs, AlSb binary and their alloys), also known as the group of 6.1 Å III-V semiconductors, an average lattice constant of α=6.1 Å, and a Poisson ratio of ν˜0.33 is assumed. The magnitude of the Burger vector is approximately 0.43 nm, while

[00007]$\cos \left(\lambda \right)=\frac{1}{\sqrt{2}}\mathrm{and}\cos \left(\vartheta \right)=\frac{1}{2}.$

For arsenide (As), nitride (N) and phosphide (P) materials the lattice constant, the Poisson ratio and the Burger vector should be changed accordingly.

[0147] The critical thickness predictions from equation (1) for 6.1 Å semiconductor epitaxial layers grown on (001) films are shown in FIG. 2.

[0148] FIG. 2 is an illustration of the misfit strain as a function of layer thickness. The solid line shows the variation of the critical thickness with misfit strain predicted for antimonide semiconductors.

[0149] As a general rule, for layers with strain-thickness characteristics below the Matthews limit (black line), the elastic strain is less than that necessary to form defects, and new threading dislocations do not spontaneously advance. Thus, it is expected that for ε.Math.h placed below and up to the critical thickness line, the elastic misfit strain is easily accommodated via plastic deformation of the layer. However, for superlattice layers below this limit, blocking of pre-existent dislocations is still possible due to dislocation bending or interaction.

[0150] Increasing the layer thickness well beyond the critical value can lead to defect generation to accommodate the misfit strain. However, several reports have indicated that even for strain-thickness products higher than by up to 20% of the critical values calculated by the Matthews model, generation of new dislocations may not take place.

[0151] Hence, it is possible to grow epitaxial layers beyond the critical layer thickness using suitable growth conditions. In various embodiments, the values predicted by the Matthews model are taken as the minimum critical thickness values to be certain of ensuring no defect generation.

[0152] Eventually, as the critical thickness is well exceeded, it becomes energetically favourable for dislocation generation to become active. Initially, strain relaxation and defect generation is slow. However, for a thickness far greater than the critical value, the effective strain is sufficient that the density of dislocations increases exponentially until the effective strain is greatly decreased. For this defect multiplication mechanism to be activated, the thickness of the layer must be sufficiently thick for dislocation loops or circles to be generated. Thus, layer thicknesses several times that of the critical thickness are required.

[0153] Various embodiments use a dislocation multiplication limit of ε.Math.h˜0.8 nm (shown by the dashed-dot line in FIG. 2). If the strain-thickness characteristics of a layer are placed above the dashed-dot line of FIG. 2, then more dislocation sources are activated and the dislocation density is substantially increased. In other words, dislocation multiplication is activated for strain-thickness product values higher than 0.8 nm.

[0154] For strained layers with strain-thickness characteristics below and as close as possible to the ε.Math.h=0.8 nm limit line, high-quality layers can be obtained with low defect densities.

[0155] Based on the above considerations, it can be seen that the strain-thickness product provides an important role in the strain relaxation mechanisms, and can be used as an important parameter to design effective dislocation filter superlattices. The main goal is to remove threading dislocations as efficiently as possible whilst not generating new sources or multiplication.

[0156] According to various embodiments, the following design rules are provided to enable the formation of effective dislocation filter (DFSL) structures:

[0157] (i) The Matthews condition (Equation 1) presented earlier is defined as the thickness limit which completely ensures the suppression of defect generation when growing strained epitaxial layers, whilst allowing bending of defects at the interfaces. Thus, the individual thickness of the layers (h.sub.1 where i=1 or 2) should be lower than the corresponding Matthews critical thickness. The misfit strain of the layers is calculated using Equation 2.

[0158] (ii) Each superlattice (SL) comprises several repeats of a layer pair. It is assumed that the layer pair operates as a single layer with thickness:

h.sub.pl=h.sub.1+h.sub.2  (4)

The strain of the layer pair can be calculated using equation:

[00008]$\begin{array}{cc}{\epsilon }_{\mathrm{pl}}=\text{?},& \left(5\right)\end{array}$$\text{?}\text{indicates text missing or illegible when filed}$

where α.sub.i, i=1 or 2, is the lattice constant of the semiconductor layer. The thickness h.sub.pl of the layer pair must also follow the Matthew model for the calculated strain.

[0159] (iii) To avoid significantly increasing the total strain ε.sub.pl, which might lead to the growth of incoherently strained layers, the strain of the layer pair should also be lower than 2%.

[0160] (iv) The strain-thickness product of the layer pair (ε.sub.pl.Math.h.sub.pl) should increase for each SL structure when moving towards the top of the structure. Higher strain-thickness product values for the layer pair can be achieved by increasing the strain (ε.sub.i) and/or the thickness (h.sub.i) of the layers when moving towards the top SLs of the structure.

[0161] (v) The dislocation filter superlattices are considered to behave as a single layer with total thickness given by:

h.sub.SL=h.sub.pl×t,  (5)

where t is the number of iterations, while the SL strain will be equal to ε.sub.pl. As such, the number of iterations is chosen so that the strain-thickness characteristics of the superlattice is placed below the multiplication limit to avoid multiplication of defects. As a result, the strain-thickness product of each SL should be lower than 0.8, i.e. ε.sub.SL.Math.h.sub.SL<0.8 nm.

[0162] (vi) The total thickness of each SL and the number of iterations should be chosen so that the strain-thickness product of the SL (ε.sub.SL.Math.h.sub.SL) increases when moving towards the top SLs in the structure.

[0163] (vii) If all the SLs are placed very close to the 0.8 nm limit, then the total net strain in the structure might be too high leading to generation of defects. Therefore, only the strain-thickness characteristics of the top SLs in the filter structure should be placed closer to the 0.8 nm limit.

[0164] In general, it is anticipated that moving towards the upper parts of the structure, the dislocation density is significantly reduced by the blocking effect of the preceding SL structures. As a result, the spatial separation between the dislocations will increase, thereby reducing the probability of defect interaction.

[0165] In order to further increase dislocation sweeping by the interfaces and enhance the motion of dislocations to promote their intersection and thus reduce their number, thicker layers and/or higher strain-thickness products are used when proceeding towards the top of the structure. This also implies that h.sub.plSL1<h.sub.plSL2< . . . <h.sub.plSLn, and ε.sub.SL1.Math.h.sub.SL1<ε.sub.SL2.Math.h.sub.SL2< . . . <ε.sub.SLn.Math.h.sub.SLn≤0.8 nm, where n=1,2,3 etc.

[0166] It is important to note that, apart from the demand of thicker layers and/or higher strain-thickness product as a means to significantly increase the effectiveness of the filter SL structure, the use of several superlattices with maximum ε.sub.SL.Math.h.sub.SL product values (ε.sub.SL.Math.h.sub.SL≈0.8 nm) will lead to failure of the filters and the total structure/device. Therefore, it is important that the strain-thickness product is built up gradually when proceeding with the growth.

[0167] Using the above limitations, a variety of possible dislocation filter structures can be designed using different III-V semiconductor material systems (such as GaSb/AlSb, InAs/AlSb and GaInSb/GaSb).

[0168] The following present in detail the characteristics of four Sb-based dislocation filter structures in accordance with various embodiments.

GaSb/AlSb Dislocation Filter Superlattice for Growing GaSb Buffer Layers: Case A

[0169] First, a GaSb/AlSb dislocation filter structure design is presented using a series of GaSb and AlSb layers with variable thicknesses. The misfit strain of an AlSb layer grown on GaSb is equal to 0.649%, which was calculated using the following equation (equation 2):

[00009]$\epsilon =\frac{{\alpha }_{\mathrm{AiSb}}-{\alpha }_{\mathrm{GaSb}}}{{\alpha }_{\mathrm{GaSb}}}\text{.100}\%,$

[0170] where α.sub.AlSb=0.61355 nm and α.sub.GaSb=0.609593 nm, the lattice constant of AlSb and GaSb respectively.

[0171] The strain is significantly lower than 2%, so that the thin AlSb layers can be elastically grown on GaSb. According to the Matthews critical thickness rule, for a strain of 0.649%, the critical thickness h.sub.c up to which the defects generation can be avoided is approximately 20 nm.

[0172] Following the rules described earlier, the GaSb/AlSb dislocation filter structure design comprises four SLs, each one with varied AlSb, GaSb and total SL thickness. As shown in FIG. 3, for each adjacent superlattice, the thickness of the layers and the strain-thickness product of the SLs increases.

[0173] Table I summarizes the characteristics of the four GaSb/AlSb superlattices used in the structure.

TABLE-US-00001 TABLE I Strain and thickness characteristics of the GaSb/AlSb dislocation filter superlattice structure comprising four individual superlattices. SL 1 SL2 SL3 SL4 Theoretical Limitations h.sub.GaSb (nm) 10 10 11 14 h.sub.AlSb (nm) 10 11 13 15 h.sub.c.sub.AlSb = 20 nm ε.sub.AlSb (%) 0.649 0.649 0.649 0.649 2% ε.sub.AlSb .Math. h.sub.AlSb (nm) 0.0649 0.0714 0.0844 0.0974 h.sub.lp (nm) 20 21 24 29 ε.sub.lp (%) 0.3248 0.3400 0.3516 0.3357 2% ε.sub.lp .Math. h.sub.lp (nm) 0.0650 0.0714 0.0844 0.09094 Iterations 10 10 9 8 h.sub.SL (nm) 200 210 216 232 ε.sub.SL (%) 0.3248 0.3400 0.3516 0.3357 2% ε.sub.SL .Math. h.sub.SL (nm) 0.6496 0.7140 0.7595 0.7788 Multiplication limit = 0.8

[0174] Based on the results presented in Table I, the four superlattices satisfy the rules described above. In further detail:

[0175] (i) SL1: GaSb 10 nm/AlSb 10 nm, 10 iterations.

[0176] The thickness of the AlSb layers is h.sub.AlSb=10 nm, while the total thickness of the GaSb/AlSb layer pair is:

h.sub.lp=h.sub.AlSb+h.sub.GaSb=10+10=20 nm.

[0177] The strain values of the AlSb layer and the GaSb/AlSb layer pair are:

[00010]$\epsilon =\frac{{\alpha }_{\mathrm{AiSb}}-{\alpha }_{\mathrm{GaSb}}}{{\alpha }_{\mathrm{GaSb}}}=\frac{0.61355-0.609593}{0.609593}=0.00649.fwdarw.0.649\%,$$\mathrm{and}$${\epsilon }_{\mathrm{lp}}=\frac{\left(\frac{{\alpha }_{\mathrm{AiSb}}.h_{\mathrm{AiSb}}+{\alpha }_{\mathrm{GaSb}}{h}_{\mathrm{GaSb}}}{h_{\mathrm{lp}}}\right)-{\alpha }_2}{{\alpha }_{\mathrm{GaSb}}}=\frac{\left(\text{?}\right)-0.609593}{0.609593}=0.00325.fwdarw.0.325\%,$$\text{?}\text{indicates text missing or illegible when filed}$

which are both significantly lower than 2%.

[0178] The strain-thickness product of the AlSb layer and the GaSb/AlSb layer pair is ε.sub.AlSb.Math.h.sub.AlSb=0.0649 nm, and ε.sub.lp.Math.h.sub.lp=0.065 nm, respectively.

[0179] The thickness of the AlSb layer and the total thickness of the GaSb/AlSb layer pair (misfit strain of 0.325%) are also placed lower than the corresponding Matthews critical thicknesses.

[0180] FIG. 4 shows an illustration of the strain-thickness characteristics of the AlSb layer and the GaSb/AlSb layer pair of the first GaSb/AlSb superlattice structure.

[0181] To build the first SL structure, the AlSb/GaSb layer pair is repeated ten times resulting in a total SL thickness of h.sub.SL=h.sub.lp 10=200 nm, and strain of ε.sub.SL=0.00325.fwdarw.0.325%<2%. The total strain-thickness product of the SL is ε.sub.SL.Math.h.sub.SL=0.6496 nm, which is significantly lower than the multiplication limit of 0.8 nm.

[0182] FIG. 5 is a schematic illustration of the strain-thickness characteristics of the first GaSb/AlSb superlattice comprising ten iterations. As shown in FIG. 5, the strain-thickness characteristic of the SL is placed below the theoretical multiplication line.

[0183] (ii) SL2: GaSb 10 nm/AlSb 11 nm, 10 iterations.

[0184] Following the same procedure described for SL1, the strain thickness characteristics of the AlSb layer, the GaSb (10 nm)/AlSb (11 nm) layer pair and the second AlSb/GaSb (10 iterations) superlattice are shown in FIG. 6.

[0185] (iii) The same procedure is repeated for the third GaSb 11 nm/AlSb 13 nm (9 iterations) and fourth SL: GaSb 14 nm/AlSb 15 nm (8 iterations) superlattice. The strain-thickness characteristics of these two superlattices are shown in FIG. 7.

[0186] FIG. 8 summarizes the strain-thickness characteristic data points obtained for all four GaSb/AlSb dislocation filter superlattices, which satisfy the design rules described above. The squares, the dots and the triangles represent the characteristics obtained for the AlSb layers, the varied AlSb/GaSb layer pairs, and the SLs respectively.

[0187] The thickness of the AlSb layers and the GaSb/AlSb layer pair increase when moving towards the top of the structure, while the strain (ε.sub.AlSb, ε.sub.lp) is kept lower than 2%. The strain-thickness characteristics of the AlSb layers and the GaSb/AlSb layer pairs are placed below the Matthews critical condition line for all four filter structures. As a result, the interfaces should demonstrate an increased defect blocking effect, while avoiding regeneration of threading dislocations due to high strain.

[0188] Furthermore, the total strain-thickness product values of the four superlattices increases while moving from the first to the fourth structure. Note that in each case, the strain-thickness characteristics of the SLs were placed below the ε.Math.h=0.8 nm limit line to avoid dislocation multiplication.

InAs/AlSb Dislocation Filter Superlattice for Growing InAs Buffers

[0189] The misfit strain for an AlSb layer grown on InAs is calculated using Equation 2:

[00011]$\epsilon =\frac{{\alpha }_{\mathrm{AiSb}}-{\alpha }_{\mathrm{InAs}}}{{\alpha }_{\mathrm{InAs}}}\text{.100}\%,$

[0190] where α.sub.AlSb=0.61355 nm and α.sub.InAs=0.60583 nm, the lattice constants of AlSb and InAs respectively. The strain was calculated as being equal to 1.274%, significantly lower than 2%. Following the Matthews critical thickness rule, the critical thickness of AlSb grown on InAs is approximately 9 nm.

[0191] The InAs/AlSb dislocation filter structure design comprises four SLs, each one with varied AlSb, InAs and total SL thickness, as shown in FIG. 9.

[0192] Table II summarizes the characteristics of the four superlattices of the InAs/AlSb filter structure.

TABLE-US-00002 TABLE II Strain and thickness characteristics of the InAs/AlSb dislocation filter superlattice structure consisted of four individual superlattices. SL 1 SL2 SL3 SL4 Theoretical Limitations h.sub.InAs (nm) 5 5 6 6 h.sub.AlSb (nm) 5 6 7 8 h.sub.c ≈ 9 nm ε.sub.AlSb (%) 1.274 1.274 1.274 1.274 2% ε.sub.AlSb .Math. h.sub.AlSb (nm) 0.0637 0.0764 0.0892 0.1019 h.sub.lp (nm) 10 11 13 14 ε.sub.lp (%) 0.6371 0.6951 0.6861 0.7282 2% ε.sub.lp .Math. h.sub.lp (nm) 0.0637 0.0765 0.0892 0.1019 Iterations 10 9 8 7 h.sub.SL (nm) 100 99 104 98 ε.sub.SL (%) 0.6371 0.6951 0.6861 0.7282 2% ε.sub.SL .Math. h.sub.SL (nm) 0.6371 0.6525 0.7135 0.7136 Multiplication limit = 0.8

[0193] The InAs/AlSb superlattices satisfy the rules described above, as shown in Table II.

[0194] FIG. 10 summarizes the strain-thickness characteristics obtained for the four InAs/AlSb superlattices of the structure. The thickness of the AlSb layers and the InAs/AlSb layer pairs increase when moving toward the top of the structure, while the strain of the layer pair is below 2%.

[0195] The strain-thickness characteristics of the AlSb layers and of every InAs/AlSb layer pair are placed below the Matthews critical thickness condition line for all four filter structures.

[0196] The total strain-thickness product values of the four InAs/AlSb superlattices increases while moving toward the top of the structure. Furthermore, the strain-thickness characteristics of the SLs is placed below and as close as possible to the 641=0.8 nm dislocation multiplication limit line.

GaSb/GaInSb Dislocation Filter Superlattice: Case A

[0197] According to this embodiment, the thickness of the GaInSb layer is increased when moving towards the top SL structures, while keeping the composition and strain of the GaInSb layers stable.

[0198] The Ga content in the GaInSb layers is 80%. The lattice constant of Ga.sub.0.8In.sub.0.2Sb was calculated from Vegard's law using the following equation:

α.sub.Ga.sub.x.sub.In.sub.1−x.sub.Sb=x.Math.α.sub.GaSb+(1−x).Math.α.sub.InSb  (6)

where x=0.8, 1−x=0.2, α.sub.GaSb=0.609593 nm and α.sub.InSb=0.6479 nm.

α.sub.Ga.sub.0.8.sub.In.sub.0.2.sub.Sb=0.8.Math.α.sub.GaSb+0.2.Math.α.sub.InSb=0.61725 nm.

[0199] As such, the misfit strain calculated using Equation 2 is ε.sub.Ga0.8In0.2Sb=1.2568%. The Matthews critical thickness for Ga.sub.0.8In.sub.0.2Sb layers grown on GaSb is approximately 9 nm.

[0200] The GaSb/Ga.sub.0.8In.sub.0.2Sb dislocation filter structure design comprises varied thickness GaSb and Ga.sub.0.8In.sub.0.2Sb layers and GaSb/Ga.sub.0.8In.sub.0.2Sb layer pairs, as shown in FIG. 11.

[0201] Table III summarizes the characteristics of the GaSb/Ga.sub.0.8In.sub.0.2Sb superlattices.

TABLE-US-00003 TABLE III Strain and thickness characteristics of the GaSb/Ga.sub.0.8In.sub.0.2Sb dislocation filter superlattice structure comprising four individual superlattices. SL 1 SL2 SL3 SL4 Theoretical Limitations h.sub.GaSb (nm) 5 5 6 6 h.sub.Ga.sub.0.8.sub.InSb (nm) 4 5 6 7 h.sub.c ≈ 9 nm ε.sub.Ga.sub.0.8.sub.InSb (%) 1.2568 1.2568 1.2568 1.2568 2% ε.sub.Ga.sub.0.8.sub.InSb .Math. h.sub.Ga.sub.0.8.sub.InSb (nm) 0.0502 0.0628 0.0754 0.0880 h.sub.lp (nm) 9 10 12 13 ε.sub.lp (%) 0.5582 0.6280 0.6280 0.6764 2% ε.sub.lp .Math. h.sub.lp (nm) 0.0502 0.0628 0.0754 0.0880 Iterations 13 11 10 9 h.sub.SL (nm) 117 110 120 117 ε.sub.SL (%) 0.5582 0.6280 0.6280 0.6764 2% ε.sub.SL .Math. h.sub.SL (nm) 0.6531 0.6908 0.7536 0.7914 Multiplication limit = 0.8

[0202] As for the GaSb/AlSb and InAs/AlSb DFSL structure described above, the strain-thickness characteristics of the GaSb/Ga.sub.0.8In.sub.0.2Sb superlattices satisfy the design rules described above.

[0203] The strain-thickness characteristics of the Ga.sub.0.8In.sub.0.2Sb layers and the GaSb/Ga.sub.0.8In.sub.0.2Sb layer pairs are placed below the Matthews critical thickness line for all four filter structures, while the strain-thickness characteristics of the GaSb/Ga.sub.0.8In.sub.0.2Sb SLs is placed below the ε.Math.h=0.8 nm dislocation multiplication limit line, as shown in FIG. 12.

GaSb/GaInSb Dislocation Filter Superlattice: Case B

[0204] According to this embodiment, the thickness and/or the strain (composition) of the GaInSb layer is increased when moving towards the top of the filter structure in order to increase the filtering effect.

[0205] So far, for all three dislocation filter designs presented above, increase of the strain-thickness product was achieved by using thicker layers. However, for ternary, quaternary, etc. III-V semiconductor alloys, it is possible to increase the strain as well as the strain-thickness product by altering the material composition, not just the thickness of the layer.

[0206] As such, an alternative GaSb/Ga.sub.xIn.sub.1−xSb DFSL structure was designed comprising Ga.sub.xIn.sub.(1−x)Sb layers of varied thickness and composition. Three different compositions were used for the Ga.sub.xIn.sub.1−xSb layers, namely Ga.sub.0.85In.sub.0.15Sb, Ga.sub.0.82In.sub.0.18Sb and Ga.sub.0.8In.sub.0.2Sb, as shown in Table III.

[0207] The lattice constant and the strain for the three types were calculated using Equations 6 and 3 respectively, with the following results:

α.sub.Ga0.85In0.15Sb=0.615339 nm, ε.sub.Ga0.85In0.15Sb=0.9426%.fwdarw.h.sub.cGa0.85In0.15Sb≈13 nm

α.sub.Ga0.82In0.18Sb=0.616488 nm, ε.sub.Ga0.82In0.18Sb=1.1311%.fwdarw.h.sub.cGa0.82In0.18Sb≈10 nm

α.sub.Ga0.80In0.20Sb0.617254 nm, ε.sub.Ga0.80In0.20Sb1.257%.fwdarw.h.sub.cGa0.80In0.20Sb≈9 nm

[0208] The critical thickness values were calculated as h.sub.cSL1≈13 nm, h.sub.cSL2 and SL3≈10 nm, and h.sub.cSL4≈9 nm for α.sub.Ga0.85In0.15Sb, α.sub.Ga0.82In0.18Sb and α.sub.Ga0.80In0.20Sb respectively using the Matthews rule.

[0209] As shown in FIG. 13, the dislocation filter structure comprises four GaSb/Ga.sub.xIn.sub.(1−x)Sb SLs and varied thicknesses as well as composition (GaSb, Ga.sub.0.85In.sub.0.15Sb, Ga.sub.0.82In.sub.0.18Sb and Ga.sub.0.8In.sub.0.2Sb) GaInSb layers. Table IV summarizes the strain thickness characteristics of the structure.

[0210] The strain-thickness characteristics of the varied Ga.sub.xIn.sub.(1−x)Sb layers and the GaSb/Ga.sub.xIn.sub.(1−x)Sb layer pairs are placed below the Matthews line for all four filter structures, while the strain-thickness characteristics of the GaSb/Ga.sub.xIn.sub.(1−x)Sb SLs are placed below the ε.Math.h=0.8 nm dislocation multiplication line, as shown in FIG. 14.

TABLE-US-00004 TABLE IV Composition, strain and thickness characteristics of the GaSb/Ga.sub.xIn.sub.(1−x)Sb dislocation filter superlattice structure comprising four individual superlattices. SL 1 SL2 SL3 SL4 Theoretical Limitations h.sub.GaSb (nm) 5 6 6 7 Composition of Ga: 85%/ Ga: 82%/ Ga: 82%/ Ga: 80%/ GaInSb In: 15% In: 18% In: 18% In: 20% h.sub.GaInSb (nm) 6 6 7 7 h.sub.c.sub.SL1 ≈ 13 nm, h.sub.c.sub.SL2 and SL3 ≈ 10 nm and h.sub.c.sub.SL4 ≈ 9 nm ε.sub.GaInSb (%) 0.9426 1.1311 1.1311 1.2568 2% ε.sub.GaInSb .Math. h.sub.GaInSb (nm) 0.0566 0.0679 0.0792 0.0880 h.sub.lp (nm) 11 12 13 14 ε.sub.lp (%) 0.5141 0.5656 0.6091 0.6284 2% ε.sub.lp .Math. h.sub.lp (nm) 0.0566 0.0679 0.0792 0.0880 Iterations 11 10 9 9 h.sub.SL (nm) 121 120 117 126 ε.sub.SL (%) 0.5141 0.5656 0.6091 0.6284 2% ε.sub.SL .Math. h.sub.SL (nm) 0.6221 0.6787 0.7126 0.7918 Multiplication limit = 0.8

[0211] The dislocation filter structures described above can be used as an advanced buffer layer to enable the direct integration of a variety of semiconductor electrical devices, such as light emitting diodes, diode lasers, vertical cavity surface emitting lasers (VCSELs), detector arrays, transistors and memories, on large and low cost wafers such as GaAs and Si.

[0212] In the case of Si, this opens up the possibility for integration of compound semiconductors for Si photonic applications. Depending on the active region of the device, the operation wavelength could potentially be tuned across a wide range of the electromagnetic spectrum. For GaSb materials, this includes the 1.3-15 μm range, which also covers the mid-infrared spectral region from 2-5 μm.

[0213] A GaSb/AlSb DFSL structure has been grown based on the design rules described above, resulting in a surface dislocation density of the order of 10.sup.6 cm.sup.−2. This value is the lowest reported so far for Sb-based materials integrated on Si.

[0214] A variety of other material combinations could be used to design effective dislocation filter structures with increasing strain and/or strain-thickness product, such as InAsSb/InAs, GaInAsSb/GaSb, GaAsSb/GaAs, GaAsP/GaAs, InP/InAsP, etc.

[0215] For ternary, or higher order alloys, the strain and/or the strain-thickness product can be increased by changing the layer composition alongside increasing the layer thicknesses when moving towards the top of the structures. However, as the alloy composition is very sensitive to temperature and growth rate variations, it is believed that control of the inserted strain in such filters will be more difficult as compared to a filter structure comprising only binary III-V semiconductor layers.

[0216] Depending on the material system and the starting defect density, a number of dislocation filters structures can be used to create an effective filter structure. However, it should be noted that the use of a very high number of superlattice repetitions is not practical since the net strain will be too high in the structure, resulting in defect or crack generation. Furthermore, the resultant buffer layers will be too thick and expensive to manufacture.

[0217] Furthermore, the spacer layers placed between the filters may have a variety of thicknesses, depending on the type and net strain of the underlying filter structure.

[0218] A high order of defect densities are observed in a variety of lattice mismatched systems, such as GaAs/Si, GaAs/Ge, GaP/Si, InAs/Si, GaN/Si etc. The dislocation filter design rules presented herein can be used to create a compatible dislocation filter structures for all of these lattice mismatched systems in order to reduce the number of threading dislocations reaching the surface of the epitaxial layers.

[0219] Tensile stain present at interfaces such as GaN/Si and GaP/Si require careful attention as high strain can lead to the generation of cracks which can interfere with the dislocation filtering mechanism. It is also important to note that in such systems, reaction of threading dislocations lead to the formation of immobile dislocations which makes the design of effective dislocation filters structures more difficult.

[0220] Furthermore, in general, integration of III-V semiconductors is possible using a variety of growth techniques such as molecular beam epitaxy (MBE) and metalorganic vapor-phase epitaxy (MOCVD). MBE is considered as being one of the best options to grow high crystalline quality dislocation filters as it enables precise and accurate control of the composition and thickness of the layers of the superlattices. On the other hand, MOCVD is preferred for high volume applications.

[0221] It will be appreciated that various embodiments provide a semiconductor device comprising a semiconductor substrate and one or more mismatched epitaxial layers. A series of strained superlattice dislocation filter layers are used to remove the threading dislocations arising from the lattice mismatch between the substrate and epitaxial layers.

[0222] The superlattice dislocation filter layers are designed so that the strain-thickness product of each SL is lower than 0.8 nm. The strain-thickness product of the pair of layers increases towards the top of the SL structure. The thickness and/or the strain of the layers increases when moving towards the top of the structure. Increasing the strain of the layers can be achieved by changing the material composition. The strain-thickness product of the superlattice filter structure can be controlled through the choice of a binary, ternary or higher order compound semiconductor alloy.

[0223] Although the above generally describes embodiments in which all of design rules (i) to (vii) described above are always followed, in other embodiments some of design rules (i) to (vii) may not be followed. For example, the strain-thickness product of the layer pair (ε.sub.lp.Math.h.sub.lp) can stay the same for some adjacent SL structures when moving towards the top of the structure. Similarly, the strain-thickness product of the SL (ε.sub.SL.Math.h.sub.SL) can stay the same, or decrease, when moving towards the top SLs in the structure. Correspondingly, the SL closest to the 0.8 nm limit may not be the topmost SL in the filter structure.

GaSb/AlSb Dislocation Filter Superlattice for Growing GaSb Buffer Layers: Case B

[0224] FIG. 15 illustrates a structure according to this embodiment, and Table V summarizes the characteristics of the four superlattices.

[0225] In this embodiment, starting from the SL closest to the substrate and moving towards the top of the structure, the thickness of the layers and/or the thickness of the AlSb/GaSb layers pairs increases, while staying below the Matthews critical thickness. Furthermore, the number of iterations for each SL is such that the strain thickness product of the SLs is always lower than 0.8 nm.

[0226] In this embodiment, the strain-thickness product of the SL (ε.sub.SL.Math.h.sub.SL) increases when moving from SL1 to SL2. However, the strain-thickness product (ε.sub.SL.Math.h.sub.SL) of the third SL, SL3, is less than the strain-thickness product of the second SL, SL2. The strain-thickness product (ε.sub.SL.Math.h.sub.SL) is then the same for the final two SL structures, SL3 and SL4. This means that the second SL structure, SL2, has the strain-thickness product closest to the multiplication limit of 0.8 nm, and not the final two SL structures, SL3 and SL4.

[0227] In this embodiment, the strain-thickness product of the layer pair (ε.sub.lp.Math.h.sub.lp) increases when moving from SL1 to SL3. The strain-thickness product of the layer pair (ε.sub.lp.Math.h.sub.lp) is then the same for the final two SL structures, SL3 and SL4.

[0228] The thickness of the AlSb layer increases when moving from SL1 to SL3 but is constant for the final two SL structures, SL3 and SL4, and the thickness of the GaSb layer is constant for the first three SL structures, SL1 to SL3, and then increases for SL4. This means that the thickness of the layer pair increases when moving from SL1 to SL4.

TABLE-US-00005 TABLE V Strain and thickness characteristics of the GaSb/AlSb dislocation filter superlattice structure comprising four individual superlattices. SL1 SL2 SL3 SL4 Theoretical Limitations h.sub.GaSb (nm) 10.5 10.5 10.5 16 h.sub.AlSb (nm) 9 10.5 12.5 12.5 h.sub.c.sub.AlSb = 20 nm ε.sub.AlSb (%) 0.649 0.649 0.649 0.649 2% ε.sub.AlSb .Math. h.sub.AlSb (nm) 0.0584 0.0682 0.0811 0.0811 h.sub.pl (nm) 19.5 21 23 28.5 ε.sub.pl (%) 0.2996 0.3246 0.3528 0.2847 2% ε.sub.pl .Math. h.sub.pl (nm) 0.0584 0.0682 0.0811 0.0811 Iterations 10 10 8 8 h.sub.SL (nm) 195 210 184 228 ε.sub.SL (%) 0.2996 0.3246 0.3528 0.2847 2% ε.sub.SL .Math. h.sub.SL (nm) 0.5842 0.6816 0.6491 0.6491 Multiplication limit = 0.8 nm

[0229] The thickness characteristics of the AlSb layers and AlSb/GaSb layer pairs are placed below the Matthews line for all four filter structures, while the strain-thickness characteristics of the SLs are placed below the ε.Math.h=0.8 nm dislocation multiplication line, as shown in FIG. 16.

[0230] The Applicant has found that the propagation of threading dislocations can be successfully blocked even when some of the design rules described above are not followed.

[0231] Three samples were grown using MBE: (i) a 2 μm thick two-step GaSb buffer (substantially as illustrated in FIG. 1); (ii) a simple filter structure of total thickness of 2.2 μm comprising five identical GaSb/AlSb superlattices with the same strain-thickness characteristics, each SL consisting of five repeats of AlSb (10 nm)/GaSb (10 nm) each separated by a 300 nm thick GaSb spacer; and (ii) a filter structure according to the embodiment of FIG. 15, i.e. comprising four GaSb/AlSb superlattices, each having different thickness and strain characteristics as shown in Table V. In this superlattice filter structure embodiment, the four SLs are separated by 200 nm thick GaSb spacer layers, and the top GaSb layer has a thickness of 50 nm. The SL structure has a total thickness of about 1.7 μm.

[0232] Surface imaging indicated a surface threading dislocation density of 2×10.sup.8 cm.sup.−2 for the two-step GaSb buffer, and the simple filter structure was found to reduce surface defect density by approximately an order of magnitude down to 3×10.sup.7 cm.sup.−2. A significantly lower surface defect density of 6×10.sup.6 cm.sup.−2 was found for the FIG. 15 embodiment. Furthermore, surface roughness analysis indicated that the FIG. 15 embodiment provided the lowest root mean square (rms) surface roughness, thus providing a smoother surface for subsequent growth.