MORPHOLOGY OF SELF-ORGANIZED InAs ISLANDS ON GaAs(001)

  We present detailed morphology of individual islands that form during the initial stages of InAs epitaxy on GaAs (001).  The islands structures where resolved directly using in-situ ultra high vacuum (UHV) plane view scanning tunneling microscopy (STM). Our data confirm that the 3D InAs islands demonstrate crystallographic faceting, and two distinct shapes are observed as a function of island volume.

Growthand STM Imaging of InAs on GaAs

Our samples were prepared using all solid sources MBE in a combined MBE-STM facility that is described in references and is shown in Figure 1. The V-element sources (solid arsenic and phosphorous) are equipped with valves to provide control over fluxes. The substrate temperature was measured using optical transmission thermometry for reproducibility and absolute measurement to within ±2 °C.

We used n-doped GaAs wafers previously degassed at 350°C in a separate chamber. The wafer is brought to 580°C under As flux to remove the surface oxide in the MBE chamber. A 0.5mm n-doped (1017cm-3) GaAs buffer layer is then grown at 580°C. An annealing procedure of the GaAs(001)-2x4 surface produces a smooth surface morphology with large terraces.The samples are then rapidly cooled below 250°C and transferred under UHV conditions to a surface analysis chamber for STM imaging. Samples are 1 cm2, and our STM tips probe the same sample region in which our optical fiber measures the temperature during growth. Each image is acquired from a different sample growth using the identical procedures, with the exception of the parameters under investigation. For InAs deposition, each growth was begun by another ~300nm GaAs buffer layer grown at 580ºC, followed by 2.1 monolayers (MLs) InAs deposited at 495 ºC. InAs growth was performed using an In beam equivalent pressure (BEP) of 1x10-7 Torr, and an As BEP of 2x10-6 Torr. RHEED oscillations and high-resolution x-ray diffraction measure this In flux to correspond to a 0.10ML/s deposition rate of InAs. An In shutter cycle of 3s open, 20s closed was repeated 7 times, while the As flux was maintained at a BEP of 2x10-6 Torr, to achieve a total of 2.1ML InAs deposition. After InAs deposition, no change in the RHEED pattern is observed as the sample is cooled, prior to its being transferred for room temperature STM imaging. STM imaging was performed using a 3V-gap voltage (filled state mode), and a constant tunneling current of 0.1nA.  

Island Shape

Figure 2 is a plane view STM image of the resulting surface that reveals distinct island shapes, which are labeled as types I, II, and hybrid. The color scale selected effectively reveals the intersections of the island side-walls with the (001) surface. The observation that these lines correspond precisely to [uv0] lattice directions and indicates that InAs 3D islands demonstrate crystallographic faceting. The hybrid islands have both type-I and type-II characteristics. The facets we observe reproduce with the same crystallographic orientation as we employ different tips. Because our tips possess a random angular orientation, tip convolution effects can be ruled out as the origin of the observed island shapes.

Figure 3 reveals structural details of the type-I and type-II islands. These images have not been processed other than a background subtraction of a curved surface to flatten the 2D regions of each image. The island facets intersect the (001) surface along precise [uv0] lattice directions to within ±2º. The bounding lines may be most accurately measured just above the island base, as they become disordered where the facets intersect the sample surface. This disordered and rounded region may be attributed to the dynamic conditions on the surface resulting from the island instability.  (At the growth temperature, islands are constantly exchanging material with their surroundings.)  For type-I, eight <130> lines bound the island base. For type-II islands, the base is bounded by four <110>, and four <010> lines.

Figure 4 illustrates the proposed structures for the type-I and type-II islands derived from these STM data. The facet angles given here are measured values taken from STM surface profiles normal to the [uv0] bounding lines. The experimental uncertainty in these angular values is ±3º. Line profiles of both island types reveal a peaked apex, rather than a truncated (001) surface as has been proposed from calculations. (Similar studies on different materials systems, show that our STM easily resolves truncated islands. For the type-I structure, it is straightforward to propose a structure using only the dominant {136} and {133} facets observed by STM.  However, it is not so simple to combine the dominant type-II facets into a structure composed of only those facets. It is beyond the capability of the STM to resolve intersections between the dominant planes. These regions may be either rounded, or composed of smaller and more elaborate facet combinations, which connect the larger facets.  Because of this, we separate the major type-II facets with a shaded region in Figure 4.

These STM data resolve a number of basic issues concerning the InAs island structure. The islands are clearly faceted as RHEED data had suggested, and they primarily form in only two nanocystalline structures.  However we do observe some islands that possess both type-I and type-II features, and examples are shown in Figures 2 and 5. As indicated by the lines superimposed in Figure 5, both type-I <130> lines, as well as type-II <100> and <110> lines bound hybrids. Another interesting observation is the C2v symmetry of both type-I & II islands. This is somewhat of a surprise, as C4v symmetry might be expected due to the non-symmorphic symmetry of the zincblende lattice. However, surface kinetics can impact island shape, and the As terminated surface can be highly anisotropic in this respect. Finally it is interesting that the type-II island retains a type-I apex, and a simple physical explanation for this is provided below.

Comparison with Previous Studies

These findings clarify several previous studies addressing the structure of InAs 3D islands, which had appeared to be in conflict with one another. The first RHEED reports concluded the islands possessed {113} facets. However, recent RHEED studies report evidence for an elongated pyramid structure bounded by four {136} facets, contradicting earlier studies and complicating the interpretation of RHEED patterns to determine island structure. The type-I shape we report is more complicated this. However the STM data strongly support these latter two reports, as RHEED data would most clearly reveal the four largest {136} facets which bound the islands’ apex. In addition, Saito et al also reports that for growth temperatures above 510º C, RHEED analysis evidenced a multifaceted island with lower {011} facets, and upper {136} facets. This interpretation of the RHEED patterns is confirmed by the STM image of the type-II island shape. In conflict with these RHEED studies, transmission electron microscopy (TEM) experiments had concluded the island shape to be a square based pyramid. More recently, C4v symmetry was reported based on grazing incidence x-ray diffraction (XRD) analysis of InAs islands. These TEM and XRD findings are also consistent with the type-II island shape we report. While it is true that the elongated apex gives the type-II structure C2v symmetry, both TEM and XRD are diffraction techniques, which may be dominated by the larger volume of the (square) island base. Thus, the nanocrystal structures revealed by STM clarify these previous studies, provided careful consideration is given to the limitations of the experimental probes used.

To better understand the change in island shape we measured the height, and the [110] and [1-10] lengths of one hundred individual islands. The results are shown in Figure 6, a histogram showing the island shape as a function of its volume. Volumes where calculated by approximating the island shape as an elliptical cone. These data reveal that a change in island shape begins as its volume exceeds 1600nm3. Recent STM and TEM studies have shown a very similar shape transition for Ge islands on Si(001). Combined RHEED and AFM studies, as a function of growth conditions, have reported indirect evidence for such transitions for InAs on GaAs(001). Because we examine island shapes from a single surface, our data confirm that a change in volume alone is sufficient to alter the island shape, (independent of surface kinetics). Also, the histogram reveals that the transition to type-II islands is not particularly abrupt, as both island types exists simultaneously on the same surface, and 16% of the islands are hybrid shaped. The average values for island parameters from the histogram sample set are listed in Table 1. It is notable that for type-I islands, the standard deviation in volume distribution is 29% while the deviation in aspect ratio is less than 7%. This observations confirms that the island shape demonstrates a true shape transition as a function of island volume.  

Theoretical Model

Several models have been developed in order to explain such shape transitions in epitaxially strained islands. The shape transition may be understood as a consequence of the strain and surface energies scaling differently as a function of island volume. The stain energy reduction offered to the InAs epilayer scales with island volume, while the surface energy cost scales with the islands surface area. Thus smaller aspect ratio islands form initially, because they cause the smallest increase in epilayer surface area. Because steeper side walls provide less lattice coupling between the island and substrate, as island volumes increase so does the strain energy, and at some point the strain energy reduction afforded by steeper side walls out weighs the increased surface area cost. At that point a shape transition occurs. From this it is simple to understand why the apex of the type-II islands retain the lower aspect ratio {136} facets. The top of the island is less strained, because the lattice constant increases as one moves higher up inside an island, and without sufficient stain energy accumulation, it is energetically favorable to maintain the lower aspect ratio shape in this region.

The idea that shape transitions might occur during the evolution of coherently strained nanocrystals was introduced by Tersoff and Tromp. Interest in the subject was rekindled by the recent observations of a shape transition for Ge islands on Si. It has been suggested that the different island shapes represented minimum energy configurations for the island shape as a function of volume. In contrast, Ross et al showed that if an abrupt change in island chemical potential accompanied the shape transition, a bimodal bifurcation of island shapes would result. That is, it is an unnecessary requirement that the observed island shapes represent quasi-stable minimum energy configurations. Most recently, it has been shown that a first order shape transition can be expected based on the competing energetic considerations of island formation, and that this shape change does indeed produce a discontinuity in the island chemical potential (validating a fundamental assumption in Ross's work).

Taken together with the recent similar observation for Ge on Si, our findings on InAs island shapes show that shape transitions are a general feature of strained nanocrystal epitaxy for semiconductor systems and thus highlight the importance of these recent theoretical studies. However, the significant number of hybrid shaped islands we observe contrasts with the observation of less than 1% of such intermediate Ge islands on Si. Our data show that the type-I to II shape transition is not necessarily an instantaneous process, where adding a few atoms causes a rearrangement of the entire island morphology. Rather, they indicate this transition is incrementally achieved during the kinetics associated with further increases in volume. In this case, one or more asymmetric island shapes may be kinetically preferred for volumes between the type-I and type-II structures. By kinetically preferred we mean that if an abrupt change in chemical potential is associated with intermediate shape transitions, then asymmetric islands may be observed in significant numbers, even if they do not represent lowest energy configurations. Such a condition is not a contradiction to the findings of Ross et al, where asymmetric island shapes were examined, but symmetric shapes were energetically preferred, as it is noted this 2D treatment can not readily be extended to 3D for asymmetric shapes.

[1][2][3][4][5]