D6.17: Public report on the layers grown on composite substrates and Si substrates with large diameter

Objectives: In many of the future high performance applications of the new composite substrates, epitaxial layers will have to be grown on the substrate to constitute the device active layer. One of the most promising family of applications involves GaN and related compound semiconductors (III-N), which can today be grown either on single crystal silicon or on single crystal Silicon Carbide both by MBE and MOCVD. After the succesfull growth of nitride layers on 2” composite substrates the development procedure was continoued on large substrates. The mains results of that phase is described in this report. We

• validate the epitaxy of III-N compounds on the composite substrates Si/pSiC and SiC/pSiC,

• evaluate the compound substrate crystal quality through III-N epilayers grown on them.

• evaluate the composite substrate quality at the user’s level, by providing complete GaAlN/GaN HEMT heterostructures to be processed into test devices, which can be directly compared to devices on bulk single crystal substrates. The objective of the present work is the structural characterization of GaN epitaxial

layers grown by MOCVD MBE on different composite substrates as well as on single crystal Si substrates. The subject of the reported work is the investigation of layers grown onto large wafers. Beside Makyoh topgraphy mainly transmission electron microscopy (TEM) was applied. TEM needs samples thinned to electron transparency. TEM sample preparation: Cross sectional TEM (Transmission Electron Microscopy) specimens were prepared from the samples using our standard procedure. (This includes cutting the wafers, embedding into a special Ti-holder, mechanical grinding and polishing, followed by 10 keV Ar ion milling. The final step was carried out at lower ion energy of 3 keV in order to decrease the surface damage of the thinned TEM specimens.). Although TEM is a local method, the thin regions were a few times 10 μm long and the layers were investigated at several regions in order to get a representative view. Both cross sectional and plan view specimens were prepared of the most important samples to determine the typical crystal defects and their density. Microscopy: A Philips CM20 TEM (working at 200 keV) equipped with a NORAN EDS analyser (with a high purity Ge detector and ultra-thin window) was used for taking images at relatively low magnification in order to get an overview. Also selected area diffraction patterns (SAED) were taken on that instrument.

Our high resolution JEOL 3010 TEM was used to take high resolution images. That microscope is equipped with a Gatan Image Filter (electron energy loss spectometer), more specifically with the new, Tridiem version. GIF was used to take elemental maps. Makyoh topography This method is probably not well known therefore we describe here its principle: Makyoh topography (Makyoh is a Japanese word, meaning 'magic mirror') is a powerful optical method. The principle of the method is the following: The mirror-like surface under test is illuminated by a parallel light beam, and the reflected beam is intersected by a screen. If the surface is perfectly flat, a uniform light spot appears on the screen. If the surface possesses deviations from the flatness, these deviations disturb the homogeneity of the reflected beam, and an image that is related somehow to the surface morphology appears on the screen (see the next figure). The main advantages of the method, as compared to other optical methods, are its simplicity, inexpensiveness, real-time operation and high sensitivity. For further details see: http://www.mfa.kfki.hu/~riesz/makyoh/

Operation principleof Makyoh topography.

MBE grown layers on SopSiC (Si on polycrystalline SiC) of large diameter L474 This sample was prepared by MBE deposition of Ga(Al)N HEMT structure on high purity Si of 3” diameter. It will be regarded as a reference to the samples prepared onto SopSiC composite substrate.

Overview TEM image (top left) of the cross sectional sample carefully aligned in edge-on direction allows the determination of individual layer thicknesses. The orientational relationship was determined from the double electron diffraction pattern of the substrate and epitaxial layers (like the top right image).

Bright field and dark field electron micrographs of the top layers of the HEMT structure show a smooth and uniform AlGaN film with sharp interface. On a specimen thinned in the plan-view geometry a dislocation density of (5.6 ± 0.5) x 109 cm-2 was measured. L810

This sample was prepared by MBE deposition of a Ga(Al)N structure on high purity Si of 4” diameter. It is also regarded as a reference to the samples prepared onto SopSiC composite substrate. The top structure is somewhat different: a multilayer stack of (ten times GaN 5 ML/AlN 2 ML) covered by a 2 nm cap layer of GaN. This deviation in the topmost layer sequence was considered not to influence too much the quality of epitaxy.

This image shows the overview of the grown structure. Here Picogiga buffer (and all the nitride layers) were grown onto thick Si wafer. Threading dislocations are present in the thick AlGaN layer. The orientational relationships observed are given beside the next selected area diffraction pattern.

(0001)GaN||(111)Si __ _ [2110]GaN||[011]Si

SAED pattern taken on sample L810. Orientational relationships. On this sample (after plan-view thinning) a dislocation density of (10 ± 1) x 109 cm-2 was measured. This was somewhat higher than expected that is why the sample on the 3” Si wafer was regarded as the reference to compare the samples grown on large diameter composite (SopSiC) substrates. Picogiga SopSiC L850 A, 4" This wafer was sent to MFA for non-destructive Makyoh analysis, only (after Makyoh that was sent back to PICOGIGA. The wafer is a 4” SopSiC overgrown with GaN by MBE method.

Morphology of that wafer shows better homogeneity, than the first SopSiC wafer measured during this project despite, that was 2” in diameter, only. Probably as a consequence of the large diameter the measured bow is relatively large, 70 μm.

L888 This sample is a Ga(Al)N HEMT structure MBE grown on a 4” SopSiC composite substrate. Although a full wafer was not available for characterization, the received sample (1/4 of a 4” size wafer) was investigated by Makyoh topography as well in order to learn homogeneity on large scale and to determine wafer bow.

Makyoh topography of the wafer showing homogeneous surface even at the edge of the large wafer. Exceptionally some round fringes can be seen, which are artifacts due to diffraction from the illumination source. On this image there is only one feature (gray spot a little bit on the left side from the center of the image) which belongs to a shallow pit on the surface. The long white stripe is definitely a scratch, while the small dark spots are dust.

Bow of the wafer was determined as 13 μm.

Overview TEM image of L888 In cross section. The epitaxial orientation relationship of the buffers (as well as the thick GaN film) to the Si (111) substrate was found to be the same as before, but in this case the sample was aligned and the diffraction pattern recorded in the [2110]GaN zone axis (see double diffraction pattern). At the “GaN” diffraction spots a splitting can be observed, too due to the presence of AlN phase (with lattice parameters a few percent smaller than GaN) in the buffer.

The fine structure of the buffer layers is revealed on the higher magnification images. Nearly all dislocations arise at these layers but the majority of them stops at the layer boundary. Their density decreases upon entering the next layer but still a large enough number of them arrives to the top of the thick GaN layer as threading dislocations causing a dislocation density of the order of 109 cm-2 which can be measured on plan-view specimens.

The HEMT structure at the top of the thick GaN layer (made visible by this two-beam bright field electron micrograph) shows a uniform thickness.

L1413 The characterization of this structure grown on 4”SopSiC started first by Makyoh topography, what was carried out on a large piece (1/4 of the wafer). Morphology was received and bow was determined first with that non-destructive method.

Makyoh topography of the wafer showing homogeneous surface even at the edge of the large wafer. Exceptionally some round fringes can be seen, which are artifacts due to diffraction from the illumination source. On this image there is only one feature (dark spot inside fringes) which belongs to a shallow pit on the surface.

Bow of the wafer was determined as 3.5 μm.

The overview of the near-substrate part of this sample is shown on the Figure below (This is also a Ga(Al)N HEMT structure MBE grown on a 4” SopSiC composite substrate):

Medium magnification micrograph of the substrate-buffer region of the sample shows mechanical strain partially relaxed with a dislocation network due to the large misfit. Elongated voids (denoted with asterisk) are regularly observed at the interface of buffers 2 and 3. The epitaxy, however, is perfect as shown on the diffraction pattern. The same orientation relationship is found as on bulk Si(111) substrates. (e.g. samples L474)

Bright field and dark field micrographs of the same area in the nucleation layer region reveal varying amount of threading dislocations in the individual buffer layers. The voids elongated to surface normal are clearly visible close to this crystallographic direction [2110]GaN. These pictures were taken in two-beam diffraction condition to enhance the contrast of crystal defects (mainly dislocations).

Bright field and dark field micrographs of the same area in the top region show a microstructure somewhat different of earlier MBE grown samples: a significant amount of threading dislocations arrive at the top surface forming a V-shaped pit (denoted with arrows). The observed broadening of defects and/or the resulting increased surface roughness is expected to have a deteriorating effect on the electrical properties. The presence of V-shaped pits at the emergence points of threading dislocations were confirmed on plan view specimens too in the form of more or less circular holes near the perforation of the specimen. (see Figure below). The mean dislocation density was found to be (4.7 ± 0.5) x 109 cm-2 while the density of holes (dislocations ended in funnels) is smaller by one order of magnitude, namely (6 ± 1) x 108 cm-2.

Bright field micrograph of the plan-view specimen showing part of the threading dislocations emerging in V-shaped pits (denoted by arrows). They are mainly visible near the perforation created during the thinning process in the wedge shaped specimen (bottom of the picture) and their density was found about one order of magnitude smaller than that of the usual “normal” threading dislocations.

A typical bright field micrograph used for the determination of density of threading dislocations with the plan-view specimen aligned in two-beam diffraction conditions (g=1010)

High magnification micrographs of the top region can be used to measure the thickness of some layers of the HEMT structure. The GaN cap is hardly visible, the measured 23 nm is the sum of the GaN cap (nominally 2 nm), the Al0.15GaN (4 nm) and Al0.25GaN (16 nm) layers. The next 20 nm thick GaN again cannot be separated from the thick Al0.025GaN film by this technique. The dislocation at the middle of the micrograph has no V-shaped termination.

L1037 with device structure and with metallization The epitaxial layer system was deposited by MBE (at PICOGIGA) onto 4” SopSiC similarly as in earlier samples (L888 or L1413), but later this sample was further processed (lithography, deposition of metallic contacts etc.) at UMS to devices and it was used for electrical test measurements.

Overview TEM images of the processed sample showing functionally different parts of the devices: Top and bottom left images show a region where the first metallization layer (about 240 nm thick) is covered by insulator film (SiO2) while on the bottom right image the thick metal contact (double gold layer separated by TiN) is directly deposited over the first metallization layer over the semiconducting structure.

Left hand images: fine structure of the interfaces between individual layers at insulated area. (Type I contact) The buffer layer structure (bottom left) and the interface (middle left) between the GaN and the complex metallization followed by an insulator (SiO2) film. No reaction occurs at this type of interface.

Right hand images: fine structure of the interfaces between individual layers at contacted area. (Type II contact) Top right image : the thick Au contact films are separated by a TiN layer and a very thin Pt layer. No trace of reaction is seen. Middle right image: the fine structure of the interface between the semiconductor and first metal layers suggests an interlayer reaction taking place between GaN and metal at this type of contact.

The above layers were labeled based on EDS (Energy Dispersive System) characterization in which the analysis of characteristic X-rays gave us the components of the individual layers. The same sample was investigated by energy filtered TEM as well in order to see the distribution and possible segregation of different elements. As an example some elemental maps are shown in the next table.

(a) TEM image (b) Au-map (d) Ti map

(c) Ni map (e) Al map

of a type II contact deposited onto the Ga(Al)N layer system. The first layer is Ti (15 nm) then a 230 nm thick metallization which consists of Ni, Au and Al, followed by 50 nm Ti layer again. Contrary to type I contact the following layers are all metallic: first a 50 nm Au followed by a thick Al-Au film. An interface reaction took place between the GaN and the metallization layers resulting in pockets in the GaN top region with enhanced Ti and Al content.

Elemental mappings confirmed not only the elements in individual layers of the metallization, but also showed that the layers are laterally homogeneous. Conclusions

• The results show, that dislocation densities are not higher at all in nitride layers grown on composite substrates than on high purity silicon. Typical dislocation density values are in the mid range of 109 cm-2.

• The layers are homogeneous laterally. • Device structures have been prepared on large diameter composite wafers

successfully.