Electron Channeling Contrast Imaging for Rapid III-V Heteroepitaxial Characterization

Summary

The use of electron channeling contrast imaging in a scanning electron microscope to characterize defects in III-V/Si heteroexpitaxial thin films is described. This method yields similar results to plan-view transmission electron microscopy, but in significantly less time due to lack of required sample preparation.

Abstract

Misfit dislocations in heteroepitaxial layers of GaP grown on Si(001) substrates are characterized through use of electron channeling contrast imaging (ECCI) in a scanning electron microscope (SEM). ECCI allows for imaging of defects and crystallographic features under specific diffraction conditions, similar to that possible via plan-view transmission electron microscopy (PV-TEM). A particular advantage of the ECCI technique is that it requires little to no sample preparation, and indeed can use large area, as-produced samples, making it a considerably higher throughput characterization method than TEM. Similar to TEM, different diffraction conditions can be obtained with ECCI by tilting and rotating the sample in the SEM. This capability enables the selective imaging of specific defects, such as misfit dislocations at the GaP/Si interface, with high contrast levels, which are determined by the standard invisibility criteria. An example application of this technique is described wherein ECCI imaging is used to determine the critical thickness for dislocation nucleation for GaP-on-Si by imaging a range of samples with various GaP epilayer thicknesses. Examples of ECCI micrographs of additional defect types, including threading dislocations and a stacking fault, are provided as demonstration of its broad, TEM-like applicability. Ultimately, the combination of TEM-like capabilities – high spatial resolution and richness of microstructural data – with the convenience and speed of SEM, position ECCI as a powerful tool for the rapid characterization of crystalline materials.

Introduction

Detailed characterization of crystalline defects and microstructure is a vitally important aspect of semiconductor materials and device research since such defects can have a significant, detrimental impact on device performance. Currently, transmission electron microscopy (TEM) is the most widely accepted and used technique for detailed characterization of extended defects – dislocations, stacking faults, twins, antiphase domains, etc. – because it enables the direct imaging of a wide variety of defects with ample spatial resolution. Unfortunately, TEM is a fundamentally low-throughput approach due to lengthy sample preparation times, which can lead to significant delays and bottlenecks in research and development cycles. Additionally, the integrity of the sample, such as in terms of the as-grown strain state, can be altered during sample preparation, leaving the opportunity for adulterated results.

Electron channeling contrast imaging (ECCI) is a complementary, and in some cases a potentially superior, technique to TEM as it provides an alternative, high-throughput approach for imaging the same extended defects. In the case of epitaxial materials, samples need little to no preparation, making ECCI much more time efficient. Additionally advantageous is the fact that ECCI requires only a field-emission scanning electron microscope (SEM) equipped with a standard annular pole-piece mounted backscatter electron (BSE) detector; forescatter geometry can also be used, but requires slightly more specialized equipment and is not discussed here. The ECCI signal is composed of electrons that have been inelastically scattered out of the in-going channeled beam (electron wave-front), and through multiple additional inelastic scattering events, are able escape the sample back through the surface.¹ Similar to two-beam TEM, it is possible to perform ECCI at specific diffraction conditions in the SEM by orienting the sample so that the incident electron beam satisfies a crystallographic Bragg condition (i.e., channeling), as determined using low-magnification electron channeling patterns (ECPs);^1,2 see Figure 1 for an example. Simply, ECPs provide an orientation-space representation of incident electron beam diffraction/channeling.³ Dark lines resulting from low backscatter signal indicate beam-sample orientations where Bragg conditions are met (i.e., Kikuchi lines), which yields strong channeling, whereas the bright regions indicate high backscatter, non-diffractive conditions. As opposed to Kikuchi patterns produced via electron backscatter diffraction (EBSD) or TEM, which are formed via outgoing electron diffraction, ECPs are a result of incident electron diffraction/channeling.

In practice, controlled diffraction conditions for ECCI are achieved by adjusting the sample orientation, via tilt and/or rotation under low magnification, such that the ECP feature representing the well-defined Bragg condition of interest – for example, a [400] or [220] Kikuchi band/line – is coincident with the optic axis of the SEM. Transitioning to high magnification then, because of the resultant restriction of the angular range of the incident electron beam, effectively selects for a BSE signal that ideally corresponds only to scattering from the chosen diffraction condition. In this manner it is possible to observe defects that provide diffraction contrast, such as dislocations. Just as in TEM, the imaging contrast presented by such defects is determined by the standard invisibility criteria, g · (b x u) = 0 and g · b = 0, where g represents the diffraction vector, b the Burgers vector, and u the line direction.⁴ This phenomenon occurs because only diffracted electrons from planes distorted by the defect will contain information about said defect.

To date, ECCI has predominantly been used to image features and defects near or at the sample surface for such functional materials as GaSb,⁵ SrTiO₃,⁵ GaN,^6-9 and SiC.^10,11 This limitation is the result of the surface-sensitive nature of the ECCI signal itself, wherein the BSE that make up the signal come from a depth range of about 10 – 100 nm. The most significant contribution to this depth resolution limit is that of broadening and damping of the in-going electron wave front (channeled electrons), as a function of depth into the crystal, due to the loss of electrons to scattering events, which reduces the maximum potential BSE signal.¹ Nonetheless, some degree of depth resolution has been reported in previous work on Si_1-xGe_x/Si and In_xGa_1-xAs/GaAs heterostructures,^12,13 as well as more recently (and herein) by the authors on GaP/Si heterostructures,¹⁴ where ECCI was used to image misfit dislocations buried at the lattice-mismatched heteroepitaxial interface at depths of up to 100 nm (with higher depths likely possible).

For the work detailed here, ECCI is used to study GaP epitaxially grown on Si(001), a complex materials integration system with application toward such areas as photovoltaics and optoelectronics. GaP/Si is of particular interest as a potential pathway for the integration of metamorphic (lattice-mismatched) III-V semiconductors onto cost-effective Si substrates. For many years efforts in this direction have been plagued by the uncontrolled generation of large numbers of heterovalent nucleation related defects, including antiphase domains, stacking faults, and microtwins. Such defects are detrimental to device performance, especially photovoltaics, due to the fact that they can be electrically active, acting as carrier recombination centers, and can also hinder interfacial dislocation glide, leading to higher dislocation densities.¹⁵ However, recent efforts by the authors and others have led to the successful development of epitaxial processes that can produce GaP-on-Si films free of these nucleation related defects,^16-19 thereby paving the way for continued progress.

Nonetheless, because of the small, but non-negligible, lattice mismatch between GaP and Si (0.37% at RT), the generation of misfit dislocations is unavoidable, and indeed necessary to produce fully relaxed epilayers. GaP, with its FCC-based zinc blende structure, tends to yield 60° type dislocations (mixed edge and screw) on the slip system, which are glissile and can relieve large amounts of strain through long net glide lengths. Additional complexity is also introduced by the mismatch in GaP and Si thermal expansion coefficients, which results in an increasing lattice mismatch with increasing temperature (i.e., ≥ 0.5% misfit at typical growth temperatures).²⁰ Because the threading dislocation segments that make up the remainder of the misfit dislocation loop (along with the interfacial misfit and the crystal surface) are well known for their associated non-radiative carrier recombination properties, and thus degraded device performance,²¹ it is important to fully understand their nature and evolution such that their numbers can be minimized. Detailed characterization of the interfacial misfit dislocations can thus provide a substantial amount of information about the dislocation dynamics of the system.

Here, we describe the protocol for using an SEM to perform ECCI and provide examples of its capabilities and strengths. An important distinction here is the use of ECCI to perform microstructural characterization of the sort typically performed via TEM, whereas ECCI provides the equivalent data but in a significantly shorter time frame due to the significantly reduced sample preparation needs; in the case for epitaxial samples with relatively smooth surfaces, there is effectively no sample preparation required at all. The use of ECCI for general characterization of defects and misfit dislocations is described, with some examples of observed crystalline defects provided. The impact of invisibility criteria on the observed imaging contrast of an array of interfacial misfit dislocations is then described. This is followed by a demonstration of how ECCI can be used to perform important modes of characterization – in this case a study to determine the GaP-on-Si critical thickness for dislocation nucleation – providing TEM-like data, but from the convenience of an SEM and in significantly reduced time frame.

Protocol

This protocol was written with an assumption that the reader will have a working understanding of standard SEM operation. Depending upon the manufacturer, model, and even software version, every SEM can have significantly different hardware and/or software interfaces. The same can be said with respect to the internal configuration of the instrument; the operator must be cautious and observant when following this protocol, as even relatively small changes in sample size/geometry, sample orientation (tilt, rotation), and working distance, can present a risk for making contact with the pole–piece, especially if not at eucentric height. The instructions provided here are for the instrument used to perform this work, a FEI Sirion SEM equipped with a field emission gun and a standard, pole-piece mounted, annular Si backscatter detector. Therefore, it is imperative that the reader understand how to perform the equivalent actions on their own specific equipment.

1. Sample Preparation

1. Cleave sample, GaP/Si for this study, into a suitable size depending on the size of the SEM sample mount that is to be used. Note: The sample can be as small as 5 mm x 5 mm or as large as a full wafer (4 inches long), depending on the internal geometry of the SEM used and the available chamber space.The sample surface should be very clean and free of contamination that could disturb the channeling (e.g., crystalline or amorphous native oxides).
2. Place the sample onto the SEM sample mount. Note: The mounting method may change depending on the type of SEM stub used, typically either a clip style or via some adhesive (e.g., carbon tape, silver paint). The method of placement must ensure that the sample will not move and that it is electrically grounded to prevent sample charging.

2. Load Sample

1. Vent the SEM by clicking the ‘Vent’ button in the software interface and insert the sample after reaching atmospheric pressure.
2. Before closing the SEM door, ensure that the sample is at an appropriate height so as to not strike the BSE detector upon moving into the SEM.
3. Pump down the SEM by clicking the ‘Pump’ button in the software interface. Wait until system indicates that the pressure is low enough to start the measurements.

3. Set Appropriate Working Conditions

1. Turn on the electron beam via the control button in the ‘Beam’ control area and set the accelerating voltage via the ‘Beam’ drop-down menu in the software interface. For the work presented here, 25 kV was used.
2. Set the beam current to an appropriate value via ‘Beam’ drop-down menu. This is determined within the system used here by way of the spot size setting, which was set to 5 (approximately 2.4 nA). Note: High beam current is typically necessary because the ECCI signal is generally weak and larger current allows for a more distinguishable image.
3. Using the secondary electron detector, adjust the image focus and stigmation via the software interface. Note: This is performed here by right-clicking and dragging the mouse on the software interface; vertical for focus, horizontal for stigmation. Also, it is usually helpful to find a small particle or surface feature on the sample to provide a clear subject for focus/stigmation adjustment.
4. Move the sample into the vertical working distance by incrementally changing the Z position of the stage and adjusting the focus and stigmation as needed. The Z position is changed through the ‘Z’ drop-down menu in the ‘Stage’ control area of the software interface. For the work described here, a working distance of 5 mm placed the same at eucentric height and provided for a strong ECCI signal.

4. Visualize Sample ECP

1. Switch into BSE mode through the ‘Detectors’ drop-down menu in the software interface.
2. Decrease magnification to its lowest setting (27x), which is done here via the computer keyboard minus (-) key, to visualize the ECP.
3. Adjust the scan rate, done here via the ‘Scan’ drop-down menu, to provide an image with sufficient signal-to-noise (e.g., slow scan rather than TV mode). Note: Averaging or integrating the image may be necessary to obtain a clearer, more discernable image.
4. Adjust the image contrast and brightness, accomplished here via the ‘Contrast’ and ‘Brightness’ sliders, to help enhance the visibility of the ECP, being careful not to oversaturate.
5. Adjust the sample rotation and tilt, using the ‘R’ and ‘T’ entries in the ‘Stage’ control area in the software interface, to help make features of the channeling pattern more apparent. Sample rotation will result in a rotation of the ECP (as shown in Figure 2) and tilting will result in a translation of the ECP (as shown in Figure 3).

5. Image Defects/Features

1. Adjust sample tilt and rotation, as described in step 4.5, to set the desired diffraction condition. Accomplish this by translating and/or rotating the ECP to align the target Kikuchi band edge (i.e., inflection point between the bright Kikuchi band and its associated dark Kikuchi line) with the SEM optic axis. While maximum channeling actually occurs at the Kikuchi line, aligning in the method described here provides visualization contrast for defects with both dark and bright contrast levels (see Figures 4 and 5).
2. Once the desired diffraction condition is achieved, increase magnification, done here via the keyboard plus (+) key.
3. Refocus image and adjust for stigmation, as described in step 3.2. Note: Here, the focus and stigmation is best adjusted with respect to the specific defect/feature being imaged.
4. Because small deviations from the edge of the band can make large differences in the appearance of the target defect or feature, optimize the diffraction condition by making small (no more than adjustments to the sample tilt orthogonally to the Kikuchi band/line of interest, while watching a specific feature for maximum contrast. Note that moving toward the inside of the Kikuchi band will typically reduce the relative contrast of “bright” features, while moving toward the outside of the band (toward the Kikuchi line) will typically reduce the relative contrast of “dark” features.
5. Once the desired contrast is obtained, decrease the magnification to verify that the same band is still on or very near the optic axis; too much tilt can change the diffraction condition altogether.

Results

The GaP/Si samples for this study were grown by metal-organic chemical vapor deposition (MOCVD) in an Aixtron 3×2 close-coupled showerhead reactor following the authors’ previously reported heteroepitaxial process.¹⁷ All growths were performed on 4 inch Si(001) substrates with intentional misorientation (offcut) of 6° toward [110]. All ECCI imaging was performed on as-grown samples with no further sample preparation whatsoever (aside from cleaving to yield approximately 1 cm x 1cm pieces for lo...

Discussion

An accelerating voltage of 25 kV was used for this study. The accelerating voltage will determine the electron beam penetration depth; with higher accelerating voltage, there will be BSE signal coming from greater depths in the sample. The high accelerating voltage was chosen for this system because it allows for visibility of dislocations that are far from the surface of the sample, buried at the interface. Other types of defects/features may be more or less visible at different accelerating voltages depending on the ty...

Materials

Name	Company	Catalog Number	Comments
Sirion Field Emission SEM	FEI/Phillips	516113	Field emission SEM with beam voltage range of 200 V - 30 kV, equipped with a backscattered electron detector
Sample of Interest	Internally produced		Synthesized/grown in-house via MOCVD
PELCO SEMClip	Ted Pella, Inc.	16119-10	Reusable, non-adhesive SEM sample stub (adhesive attachment will also work)