LIBRARY Mlchlgan State Unlversity PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE DUE DATE DUE use Macs-p.14 ESTABLISHMENT AND APPLICATIONS OF ELECTRON CHANNELING CONTRAST IMAGING By Benjamin Andrew Simkz'n A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Materials Science and Mechanics Michigan State University East Lansing, MI 48824 1997 ABSTRACT ESTABLISHMENT AND APPLICATIONS OF ELECTRON CHANNELING CONTRAST IMAGING By Benjamin Andrew Simkz'n Electron channeling contrast imaging (ECCI) is a scanning electron microscopy (SEM) technique that enables microscopic examination of crystalline features and defects in the near-surface region of bulk crystalline samples. Starting from the methodology as presented by Czernuszka, et. al.,1 the author has successfully sought to establish an experimental setup that would enable routine application of ECCI to Materials Science problems at Michigan State University using a CamScan44FE scanning electron microscope. The new experimental setup employs a simpler detector setup that simultaneously reduces costs and relaxes experimental constraints. This work has established a novel experimental setup allowing routine in—situ ob- servation via ECCI of samples during testing in the SEM, as well as a solid grounding of information concerning the interrelationship between experimental parameters, the microscopic sample, and the resulting electron channeling contrast images. Sample surface preparation ECCI artifacts and their elimination are also presented. 1J.T. Czernuszka, NJ. Long, E.D. Boyes, P.B. Hirsch:“Imaging of Dislocations using Backseattered Electrons in a Scanning Electron Microsc0pe”, Philosophical Magazine Letters, 62, p.227 (1990) Copyright by Benjamin Andrew Simkin 1997 ACKNOWLEDGEMENTS The CamScan44F E employed in this work was purchased using funds from the National Science Foundation (Grant No.DMR#9302040) and Michigan State Univer- sity. The author would also like to acknowledge the support for this work from the National Science Foundation through Grant No.DMR#9257826. In addition, the au- thor would like to acknowledge the support of his advisor, Dr. Martin Crimp. Thanks also to Bob Harris for assistance using the geological sample preperation saw. iii TABLE OF CONTENTS LIST OF TABLES vi LIST OF FIGURES ' vii 1 Introduction 1 1.1 Background ................................ 1 1.2 Intent ................................... 3 1.3 Electron Channeling ........................... 3 1.3.1 Historical Overview ........................ 3 1.3.2 Comparison of ECG to Other Signal Types and Explanation of the Usage of this Contrast Mechanism ............. 5 1.3.2.1 Microsc0pic Parameters ................ 10 1.3.2.2 Imaging via Channeling Contrast ........... 11 1.3.3 Choice of Microsc0pic Imageing Parameters .......... 13 1.4 The Typical ECC Image ......................... 19 2 Experimental Procedure 21 2.1 Sample Preparation ............................ 21 2.1.1 Silicon ............................... 21 2.1.2 Single Crystal FeAl ........................ 23 2.1.3 Single Crystal NiAl ........................ 24 2.2 Microscopy ................................ 25 2.2.1 Detector Configurations for ECCI ................ 25 2.2.2 Establishing ECCI ........................ 27 2.2.3 Establishment of Microscope Conditions for ECCI ....... 27 2.2.4 Alignment of the Sample for ECCI ............... 30 2.3 Evaluation of Microsc0pe and Sample Conditions Required for ECCI, and their Effect on resulting Images ................... 32 2.4 Application of ECCI ........................... 32 2.4.1 Analysis of Frank Loops in FeAl ................. 33 iv 2.4.2 Dislocation Detection Depth Measurements .......... 34 2.4.3 Effects of Channeling Conditions and Deviation Parameter on the ECC Image .......................... 37 3 Results and Discussion 40 3.1 Evaluation of ECCI SystemSetup .................... 40 3.1.1 Initial ECCI Experiments .................... 41 3.1.2 Typical Working Parameters ................... 42 3.1.3 Other Topics Concerning ECCI with the CamScan44FE . . . 54 3.1.3.1 Rotation Between the Image and Channeling Pattern 54 3.1.3.2 Use of Independent Image Processing ......... 57 3.2 Evaluation of Sample Conditions .................... 57 3.2.1 Sample Preparation ........................ 57 3.2.1.1 NiAl ........................... 57 3.2.1.2 FeAl ........................... 79 3.2.2 Si .................................. 92 3.2.3 Vacuum Quality .......................... 93 3.3 Topics Examined Using ECCI ...................... 94 3.3.1 Dislocation Loops in FeAl .................... 94 3.3.2 Dislocation Detection Depth Determination in Silicon ..... 95 3.3.3 Effect of Deviation From Bragg Condition on Defect Image Characteristics .......................... 102 4 Conclusions 106 5 References 108 LIST OF TABLES 3.1 Representative relationships between relevant probe parameters for the CamScan44FE at 10mm working distance with a 50pm aperture. . . . 47 vi 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 1.10 LIST OF FIGURES Electron channeling contrast between different grains in polycrystalline FeAl ..................................... Electron Channeling Pattern (ECP) configurations: (a) Conventional ECP; (b) Selected Area ECP (SACP). ................. Configuration for producing Selected Area Channeling Patterns: (a) Beam rocking geometry; (b) Example of a SACP. ........... Example of a conventional ECP from a Si single crystal. ....... Example of a SACP from FeAl. ..................... Plane bending model of diffraction contrast for a dislocation in a TEM. (a) side view of an edge dislocation parallel to the surface, represented as distorted crystal planes. (b) resultant image from electron beam passing through the crystal. ....................... Effects of local crystal rotation on local SACP, and effects of beam con- vergence angle. (a) Bulk crystal, small convergence angle; (b) Rotated crystal, small convergence angle; (c) Bulk crystal, large convergence angle; (d) Rotated crystal, large convergence angle. .......... Channeling contrast mechanism for a pure edge dislocation parallel to the surface, showing the SACPs expected from each region and the resultant ECC image ............................ Channeling contrast mechanism for a pure screw dislocation parallel to the surface. ECPs are represented for small volumes of crystal near the dislocation core to show the expected sense of contrast in the ECC image. ................................... Expected channeling contrast mechanism for a pure edge dislocation normal to the surface. ECPs are represented for small volumes of crystal near the dislocation core to show the expected sense of contrast in the ECC image. Note that contrast is obtained in this case by change in the Bragg angle for, rather than tilting of, the crystal volume. . . . vii 10 12 14 16 17 1.11 Showing schematically the changes that take place in an ECP with 2.1 2.2 .2.3 2.4 2.5 2.6 2.7 3.1 3.2 3.3 surface tilt. (a) No or low surface tilt. (b) High surface tilt ....... Layout of the two different BSE detectors used for ECCI. (a) the side- mount detector; (b) the polepiece-mounted Si diode detector ...... Schematic representation of the steps involved in estimating the conver- gence angle, a. (a) side view of the scanning electron beam intersecting the sample edge in either the in-focus condition (lower), or the out-of- focus condition (upper). (b) representation of the signal intensity from a horizontal line scan for each of the two conditions. (c) representation of the image resulting from each of the two conditions. The apparent 20 26 width of the feature in each image is taken as the apparent feature size. 29 Schematic of the components of the Faraday cup used for measure- ments of probe current ........................... Schematic showing how a particular crystal in the sample is aligned for ECCI. (a) channeling band used for contrast is aligned the center of the SACP; (b) the conventional ECP at low magnification is aligned to the center of the microscope scan field; (c) the same condition as (b), except with the microscope at somewhat higher magnification. Explanation of the terms used for stereo pair measurements. ..... Top view of the dislocation geometry symbolized in Fig. 2.5 ...... Illistration of the definitions of g and s in terms of selected area chan- neling patterns. (a) 8:0, g set as positive; (b) s>0, g still positive; (c) s=0, g negative; g positive, s<0. g is essentially the diffraction vector pointing towards the channeling plane. ................. ECC image of dislocation clusters in Si+10%Ge epitaxially grown over (100) Si at 25kV accelerating voltage using the polepiece—mounted BSE detector. 20mm working distance, 570x magnification, 9° tilt from surface normal. .............................. ECC image of SiGe epitaxially grown over (100) Si at 25kV acceler- ating voltage using the side-mounted BSE detector. The detector is located towards the bottom of the figure. 20mm working distance, 570x magnification, 9° tilt from surface normal. ............ ECC image of SiGe epitaxially grown over (100) Si at 20kV accelerating voltage using the polepiece—mounted BSE detector. 20mm working distance, 570x magnification, 9° tilt from surface normal ........ viii 30 31 35 36 39 43 44 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 ECC image of SiGe epitaxially grown over (100) Si at 20kV acceler- ating voltage using the side—mounted BSE detector. The detector is located towards the bottom of the figure. 20mm working distance, 570x magnification, 9° tilt from surface normal. ............ ECC image of SiGe epitaxially grown over (100) Si at 14kV accelerating voltage using the polepiece—mounted BSE detector, 570x magnifica- tion, 9° tilt from surface normal ...................... ECC image of SiGe epitaxially grown over (100) Si at 14kV accelerat- ing voltage using the side—mounted BSE detector, located towards the bottom of the figure. 570x magnification, 9° tilt from surface normal. ECC image of dislocation clusters in Si+10%Ge epitaxially grown over (100) Si at 7kV. Note the difference in image contrast between the areas with greater and lesser amounts of beam-fixed contamination. Pole- piece mounted BSE detector, 200 second frame time, 23 mm working distance, 570x magnification, 9° tilt from surface normal ........ ECC image of SiGe epitaxially grown over (100) Si at 7kV accelerating voltage using the side—mounted BSE detector. The detector is located towards the bottom of the figure. Note the contrast differences due to beam-fixed contamination. 23mm working distance, 570x magnifica- tion, 9° tilt from surface normal ...................... 12 mm working distance BSE image using the polepiece detector, show- ing the 110 ECP for NiAl. Surface is normal to microscope axis. 15x magnification ................................ 10 mm working distance BSE image using the polepiece BSE detector, showing the 110 ECP for NiAl. 15x magnification. .......... 9 mm working distance BSE image using the polepiece BSE detector, showing the 110 ECP for MA]. 16x magnification. .......... 8 mm working distance BSE image using the polepiece BSE detector, showing the 110 ECP for MA]. Note that this image and 3.11 em- ploy the same zero level and gain settings for the BSE signal. 17x magnification ................................ ECC image of dislocations in N iAl taken using the polepiece mounted BSE detector at ~7.5° tilt from horizontal. Channeling contrast from (200) channeling band. 25kV accelerating voltage, 12mm working dis- tance, recursively filtered 32 frame average, 14600x magnification. . . ix 44 45 45 46 46 49 50 51 3.14 3.15 3.16 3.17 3.18 3.19 3.20 3.21 ECC image of approximately the same field of view as in Fig. 3.13 of dislocations in NiAl, taken using the polepiece mounted BSE de- tector at ~12.5° tilt from horizontal. Channeling contrast from (200) channeling band. 25kV accelerating voltage, 12mm working distance, recursively filtered 32 frame average, 14600x magnification. ..... ECC image of approximately the same field of view as in Fig. 3.13 of dislocations in NiAl taken using the polepiece mounted BSE detector at ~23.5° tilt from horizontal. Channeling contrast from (200) channeling band. 25kV accelerating voltage, 12mm working distance, recursively filtered 32 frame average, 14600x magnification ............. ECC image of approximately the same field of view as in Fig. 3.13 of dislocations in NiAl taken using the polepiece mounted BSE detector at ~29.5° tilt from horizontal. Channeling contrast from (200) channeling band. 25kV accelerating voltage, 12mm working distance, recursively filtered 32 frame average, 14600x magnification ............. ECC image taken using the side mounted BSE detector. Channeling contrast from (211) channeling band. 25kV accelerating voltage, re- cursively filtered 32 frame average, 14600x magnification. Note that for this sample and holder geometry, it was not possible to move the 52 52 53 detector close to the sample, leading to a small solid angle of collection. 54 ECC image taken using the polepiece mounted BSE detector. Chan- neling contrast from (211) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. ..... Differances between the final lens strength between the regular imaging condition, (a), and the SACP condition, (b). Note that the final lens is only used to deflect the electron beam in the SACP case, and is therefore much weaker ........................... Image of dislocations in Si taken using external beam control and image processing via the Semper6 software package. Signal from the side- mount BSE detector, 6 frame Kalman (recursive) filter, 10.9 second frame time (26.8asec. dwell time). ~14400x magnification ....... Image of dislocations in Si taken using external beam control and image processing via the Semper6 software package. Signal from the side- mount BSE detector, 6 frame Kalman (recursive) filter, 10.9 second frame time (26.8psec. dwell time). ~14400x magnification ....... 55 56 58 59 3.22 3.23 3.24 3.25 3.26 3.27 3.28 3.29 3.30 3.31 3.32 3.33 3.34 Image of dislocations in Si taken using external beam control and image processing via the Semper6 software package. Signal from the side- mount BSE detector, 60 frame Kalman (recursive) filter, 1.136 second frame time (2.8usec. dwell time). ~14400x magnification. ...... E—T (Everheart-Thornley detector) image of electrode discharge ma- chined NiAl surface. 25 second frame time, 300x magnification. . . . E-T image of the as electrode discharge machined (EDMed) NiAl sur- face. 14700x magnification. ....................... BSE image of as-EDMed NiAl surface. 14700x magnification. E—T image of EDMed NiAl after ~50um material removal by electropolish. White feature is an inclusion unaffected by the electropolish.300x magnification. .................... E-T image of electropolished NiAl surface after EDM. 14700>< magni- fication. .................................. Image of the same area shown in Fig. 3.27 in NiAl showing the dom- inance of ECG features over topographic features after ~30 minutes of electropolishing. 32 frame average ECCI using the (110) band for channeling contrast. 14700x magnification ................ BSE image of EDMed, then electrOpolished NiAl after ~100am re- moval. 32 frame average ECCI using the (110) band for channeling contrast. 14600x magnification ...................... BSE image of EDMed, then electropolished NiAl after ~100/1m re- moval. 32 frame average ECCI using the (110) band for channeling contrast. 14600x magnification ...................... E—T image of same area as Fig. 3.32 in NiAl. 25kV accelerating voltage, 60 61 62 62 63 64 65 65 recursively filtered 32 frame average BSE image. 14600x magnification. 66 BSE image of NiAl under channeling conditions as given by Fig. 3.33. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. . SACP from as-mechanically polished NiAl. 25kV accelerating voltage, recursively filtered 32 frame average BSE image ............. SACP from mechanically polished, 5 minute electropolished N iAl sur- face showing the electron channeling conditions for Figs. 3.35— 3.38. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. “Mottling” near center and edges due to sweeping of beam across sample during rocking to form SACP. .............. xi 67 67 68 3.35 3.36 3.37 3.38 3.39 3.40 3.41 3.42 14600x magnification electron channeling contrast image from 5 min. electropolished NiAl surface. Channeling contrast from (110) chan- neling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image ............................. 7300x magnification electron channeling contrast image centered on Fig. 3.35 of 5 min. electropolished NiAl surface. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. Note variations in dislocation density. .................................. 3650x magnification electron channeling contrast image centered on Fig. 3.36 of 5 min. electropolished NiAl surface. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. .................. 1200x magnification electron channeling contrast image centered on Fig. 3.37 of 5 min. electropolished NiAl surface. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. Note that the dominant features are now due to the combined strain fields from many dislocations, and begin to take on the appearance of the “scratches” of Fig. 3.39. 80x magnification BSE image showing a convolution of conventional ECP and electron channeling contrast from “scratch” features after 5 minutes electrOpolish in NiAl. Channeling contrast from (110) chan- neling band. 25kV accelerating voltage, recursively filtered 16 frame average. .................................. 1200x magnification, 25kV accelerating voltage, recursively filtered 32 frame average E—T image of the 5 min. electropolished NiAl sur- face. Same area and conditions as Fig. 3.38; Note scan squares from Figs. 3.35— 3.38 ............................... 1200x magnification, 25kV accelerating voltage, recursively filtered 32 frame average. Same area and conditions as Fig. 3.38. BSE image in “AB” difference mode ........................... BSE image showing a convolution of conventional ECP and electron channeling contrast from “scratch” features after 10 minutes electropol- ish. Channeling contrast from (110) channeling band. 25kV accelerat- ing voltage, recursively filtered 32 frame average, 80x magnification. . xii 68 70 70 71 72 73 3.43 3.44 3.45 3.46 3.47 3.48 3.49 3.50 Channeling contrast image using contrast from the (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. Note scan square artifact in center of image. 7300x magnification ................................ BSE image of MA] showing a convolution of conventional ECP and electron channeling contrast from “scratch” features after 17.5 min- utes electropolish. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 80x magnification ................................ ECC image of NiAl mechanically polished then electropolished for 29 minutes showing “scratch” features. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 130x magnification. Note that the oval “bright-dark” features are ECC features consisting principally of clus- ters of end-on dislocations at either edge, separated by a region of low dislocation density. ............................ ECC image of NiAl mechanically polished then electrOpolished for 39 minutes showing “scratch” features. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 130x magnification .............. ECC image of MA] mechanically polished then electropolished for 47 minutes showing “scratch” features. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 220x magnification .............. ECC image of MA] mechanically polished then electropolished for 17.5 minutes. Channeling contrast from (110) channeling band. 25kV ac- celerating voltage, recursively filtered 32 frame average BSE image, 14600x magnification. .......................... ECC image of MA] mechanically polished then electropolished for 29 minutes, on a “scratch” feature. Channeling contrast from (110) chan- neling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 14600x magnification ................ ECCI of MA] mechanically polished then electrOpolished for 39 min- utes, on a “scratch” feature. Channeling contrast from (110) chan- neling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 14600x magnification ................ xiii 73 74 74 75 75 76 77 77 3.51 3.52 3.53 3.54 3.57 3.58 3.59 3.60 3.61 3.62 ECCI of NiAl mechanically polished then electropolished for 47 min- utes, on a “scratch” feature. Channeling contrast from (110) chan- neling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 14600x magnification ................ ECCI of MA] mechanically polished then electropolished for 29 min- utes, away from any specific “scratch” features. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 14600x magnification. ..... E-T image of as-mechanically polished FeAl surface. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. . BSE image of as-mechanically polished FeAl under channeling contrast conditions of the same area as shown in Fig. 3.53. Channeling contrast from (200) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification ............. SACP corresponding to the channeling conditions for Fig. 3.56. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. BSE image of FeAl under channeling conditions as given by Fig. 3.55. Channeling contrast from (200) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. . SACP from mechanically polished FeAl after 10 minutes electropolish. 25kV accelerating voltage, recursively filtered 32 frame average. BSE image of mechanically polished, 10 minute electropolished FeAl. Channeling contrast from (200) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. . BSE image of mechanically polished, 22% minute electropolished FeAl. Channeling contrast from (200) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. . E-T image of electrode discharge machined FeAl surface. 25 second frame time, 300x magnification ...................... BSE image of electrode discharge machined FeAl surface after 6 minutes electropolish. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame aver- age, 14700x magnification ......................... E—T image of electrode discharge machined FeAl surface after 6 min- utes electropolish. Recursively filtered 32 frame television scan-rate average, 300x magnification ........................ xiv 78 78 79 80 80 81 82 82 83 83 84 85 3.63 3.64 3.65 3.66 3.67 3.68 3.69 3.70 3.71 3.72 3.73 BSE image of electrode discharge machined FeAl surface after 6 min- utes electropolish. Recursively filtered 32 frame average, 300x magni- fication. .................................. SACP of electrode discharge machined, 6 minute electropolished NiAl surface. .................................. BSE image of electrode discharge machined FeAl surface after 12 min- utes electropolish. Recursively filtered 32 frame average, 300x magni- fication. .................................. BSE image of region of incomplete surface electropolish after 12 min- utes electropolish. 200x magnification. ................. SACP from electrode discharge machined FeAl after 12 minutes elec- tropolish. 25kV accelerating voltage, recursively filtered 32 frame av- erage. ................................... BSE image of EDMed FeAl surface after 12 minutes electropolish. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. Bright features are not topographic contrast ............... BSE image of EDMed FeAl surface after 12 minutes electropolish, ~20um away from Fig. 3.68. Channeling contrast from (110) chan- neling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. ..................... BSE image of EDMed FeAl surface after 12 minutes electropolish, ~150am away from Fig. 3.68. Channeling contrast from (110) chan- neling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. ..................... BSE image of EDMed FeAl surface after 22 minutes electropolish. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. . BSE image of EDMed FeAl surface after 22 minutes electropolish, ~140pm away from Fig. 3.71. Channeling contrast from (110) chan- neling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. ..................... BSE image of EDMed FeAl surface after 22 minutes electropolish, ~140pm away from Fig. 3.72 and ~140pm away from Fig. 3.71. Chan- neling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. ..... XV 86 87 88 90 90 3.74 3.75 3.76 3.77 3.78 3.79 3.80 3.81 3.82 3.83 SACP corresponding to the channeling conditions for Figs. 3.71—3.73. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. ................................... ECC image taken using the polepiece mounted BSE detector from mechanically polished silicon. 25kV accelerating voltage, recursively filtered 32 frame average, 11mm working distance. White “speck” fea- tures are residual polishing media. 15400x magnification ........ ECC image taken using the polepiece mounted BSE detector from me- chanically polished silicon. 25kV accelerating voltage, single 25 second frame, 11mm working distance. White “speck” features are residual polishing media. 15400x magnification .................. Image of prismatic loops in FeAl using channeling contrast from the (110) channeling band. 14600x magnification. ............. Image of prismatic loops in FeAl showing contrast loss for [100] ori- ented segments; note segment (a). Channeling contrast from the (200) channeling band. 14600x magnification. ................ Image of prismatic loops in FeAl using channeling contrast from the (110) channeling band. 14600x magnification. ............. Image of prismatic loops in FeAl showing contrast loss for [010] ori- ented segments; note segment (b). Channeling contrast from the (020) channeling band. 14600x magnification. ................ Dislocation image at +8° tilt relative to sample normal in Si. Note that by employing the (220) band for imaging contrast, the 220 g- vector becomes the tilt axis. Letter “e” marks the point of exit from the sample. ................................ Dislocation imaged at -10.5° from sample normal in Si. Arrowed dis- location same as in Fig. 3.81. ...................... Side view demonstrating the difference in depth of penetration required to reach the same point due to differences in tilt to either side of the surface normal. (a) positive tilt; (b) negative tilt. ........... xvi 91 92 93 96 97 100 3.84 3.85 3.86 Top view demonstrating the error introduced to the measured parallax between micrographs due to a difference in perception of the contrast loss point along a dislocation line between micrographs. The error introduced in the parallax measurement per error in length along the dislocation line increases proportionately to the sine of the projected angle between the rotation axis and the dislocation line, 6. Thus, condition 2 produces a larger parallax error per error e along the length of the dislocation line than does condition 1. .............. Individual and clusters of dislocations in Si displaying bright-dark con- trast reversal for Opposite g; compare to Fig. 3.86. g=113, s<0. Dislocations in Si displaying bright-dark contrast reversal for opposite g; compare to Fig. 3.85. g=11-3, s0 ............ Dislocation images displaying opposite senses of dislocation contrast for positive and negative values of s. g=022, s<0 ............ xvii 101 103 104 105 105 Acronyms and Common Abbreviations TEM 'Ii‘ansmission Electron Microscope or Transmission Electron Microscopy STEM Scanning Transmission Electron Microscope or Scanning Transmission Elec- tron Microsc0py ECCI Electron Channelling Contrast Imaging or Electron Channeling Contrast Im- age ECC Electron Channeling Contrast EDM Electrode Discharge Machining EDMed Electrode Discharge Machined SEM Scanning Electron Microsc0pe SACP Selected Area Channeling Pattern ECP Electron Channeling Pattern; includes both conventional ECPs and SACPs a Convergence half-angle for the scanning electron beam beam current 5 Electron gun brightness: unit area xsolid angle 2', Electron probe current E0 Accelerating voltage ,u Backscattered electron yield; the fraction of electrons backseattered from a sample relative to the scanning beam BSE BackScattered Electron (1,, Electron probe diameter xviii CHAPTER 1 Introduction 1.1 Background Historically, crystalline defects such as stacking faults, dislocations, phase boundaries, and the like have been studied on a microscopic level almost exclusively through the use of the Transmission Electron Microscope (TEM) [see for example:1,2]. This has been due to both the higher resolutions typically obtainable with these instruments (sub-nanometer resolutions are typical for most 200kV TEMs) as well as the great ease with which contrast can be obtained for crystal defects through the use of diffraction contrast. An added advantage of using diffraction contrast in a TEM is that im- portant defect parameters can be determined based on which diffraction conditions produce a certain type of contrast. Formation of diffraction contrast in a TEM can be understood at an intuitive level by considering diffraction from only one set of atomic planes in a crystal with a defect such as a dislocation (see [2], pp.404 or 428). This contrast mechanism is explained in some detail in the following chapter. Although the TEM is a very useful tool for the study of crystalline samples, signif- icant problems are also inherent to crystal defect studies using the TEM. These arise mainly from the requirement that TEM samples must be thin enough to be consid- ered “electron transparent”, and are frequently related to the extreme thinness of the samples. For example, crystal defect structures such as dislocations or deformation twins may be introduced or altered merely by careless manipulation of an electron transparent sample. Alternately, for a material with a sufficiently low dislocation flow stress, the mere presence of a free surface is sufficient to draw dislocations out of the near-surface region [3]. For a sample that is electron-transparent, there is generally little material that can not be considered near-surface. Most TEM samples experience significant distortions due to beam heating of the sample, and for samples with low strength nd a low thermal conductivity, this heating can distroy the sample. Other problems arise if any in-situ testing of samples is to be performed, as TEM samples rarely possess a shape both simple and well-defined so that stress states within the area of observation can be related to the observed phenomena. In addition to these, the mere process of making an acceptable TEM sample is often very difficult and time-consuming. Another method that can be used to image crystal defects in a Scanning Electron Microscope (SEM), referred to as Electron Channeling Contrast Imaging (ECCI), has recently received more attention. Long regarded as more a curiosity than a viable method, it has recently received interest as an alternative to TEM for certain near- surface defect studies in bulk samples. The impetus for an SEM technique to study dislocations seems obvious, in com- parison to the limitations on both the data and the samples imposed by the TEM thin foil configuration. SEM samples are typically large enough so that both easy sample manipulation and easy measurement of sample geometries are possible, so that more elaborate testing geometries are possible, while the surface stress states are more easily quantified. Additionally, preparation of the larger SEM sample is frequently easier than for a TEM foil. It is for these several reasons that advancing the usage of ECCI in general and at Michigan State University in particular improves the range and scope of both t0pics amenable to study as well as the information attainable from these studies for a wide range of materials science problems. 1 .2 Intent This thesis work was initiated with the intention of establishing a working instrumen- tal setup for, and making use of, Electron Channeling Contrast Imaging (ECCI) in materials science studies at Michigan State University, with an emphasis on maximiz- ing the versatility of the system by: maximizing the ease of information acquisition; maximizing the possible range of operational configurations; and easing interpretation of the resulting images through formulation of a more intuitive Electron Channeling Contrast (ECC) model and reducing foreshortening of ECC images. 1 .3 Electron Channeling 1.3.1 Historical Overview Electron channeling contrast in its most basic form is frequently seen by scanning electron microscopists without conscious recognition of it as a contrast mechanism in such images as in Figure 1.1. This micrograph shows a polycrystalline sample of FeAl with randomly oriented grains, each of whose “brightness” depends principally on the orientation of that grain relative to the electron beam. This dependence of the backseattered electron yield ()1) on beam-crystal orientation is the basis for all ECC. Although somewhat dated, an overview of ECC and some of its applications can be found in [4]. Electron channeling was first noted as a phenomenon in 1967 [5] during the early days of scanning electron microscopy as patterns formed on the images of large single crystals at low magnifications. Quickly identified as being related to the crystallinity Figure 1.1. Electron channeling contrast between different grains in polycrystalline FeAl. of the specimens and the wavelengths of the incident electrons [5,6], the patterns were first analyzed by the relatively new Kinematical Theory for electron interactions in a periodic potential, where it was demonstrated that the backseattered intensity from a crystal was dependent on the incidence vector of the electrons relative to the crystal structure. For an image formed by scanning the incident electrons through two perpendicular rotation axes, as is done with a Scanning Electron Microscope (SEM), the resultant pattern would take the form of bands whose edges separated high intensity from low intensity, with the band width equivalent to twice the Bragg angle, and whose band centers correspond to the trace of the plane in question. This preliminary analysis [6] immediately identified this as a potential means of imaging near—surface crystalline defects in a bulk sample. Specific predictions of the required microscope parameters and the expected image contrast were made [7-9]. All of the earliest experiments in the early 19705 involved thin foils [10, 11, 12, others], and it was not until significantly later that bulk ECCI was performed. The first imaging of dislocations in a bulk sample [13] employed a rather arduous experimental setup, which subsequent researchers have relaxed somewhat [14-15], enabling greater use of this method. 1.3.2 Comparison of ECC to Other Signal Types and Expla- nation of the Usage of this Contrast Mechanism There are a number of signals used for forming scanned images in an SEM; by far the most frequently used type of signal is the Secondary Electron (SE) signal, which is quite sensitive to fine topography and easy to measure [16]. The next most commonly collected signal for image formation is the BackScattered Electron (BSE) signal, which is the relevant signal displaying ECC. The BSE signal is formed from primary beam electrons that have been scattered through a sufficiently large angle in one or more scattering events to re—emerge from the sample surface, and typically posess a significant fraction of the primary beam energy. Because of the high energy of BSEs, detector systems are generally limited to line-of-sight approaches, with the signal strengths dependant on the solid angle of collection for the detector. The SE signal consists of mostly low energy (typically under 10eV) electrons that have been emitted from the sample atoms under the influience of the primary beam electrons. Due to their low energy, the escape depth of secondary electrons is quite small, so this signal is surface sensitive. In addition, the low energies also allow for easy collection of these electrons by electrostatic means, which is how most detectors sensitive to this signal operate. It is worth noting that SEs are produced from the sample by beam electrons that are also escaping the sample, so that the SE yield is linked to the BSE yield. When an electron beam strikes a crystalline sample, the probability that any given electron will backscatter is principally a function of the average local atomic species, the tilt of the electron trajectory relative to the sample surface, and the orientation of the electron trajectory relative to the crystal lattice of the sample. The atomic number dependence gives rise to what is referred to as “Z contrast”, the local tilt dependence gives rise to topographic contrast, and the electron/lattice dependence gives rise to electron channeling contrast. The difference in backscatter intensity due to electron channeling is almost al- ways less than 10%, and typically 5% or less, so this contrast mechanism is usually dominated by Z contrast and topographic contrast. In addition, as SEMs typically employ a relatively highly convergent beam to scan the sample with, a wide range of electron trajectories is used to form the backscatter intensity at each point, and therefore the channeling contrast tends to cancel itself out. Electron channeling in SEM is therefore often neglected as an important contrast mechanism. Electron channeling contrast may be used to gain crystallographic information on the crystal being observed by plotting the observed Backscattered Electron (BSE) intensity (p) as a function of the probe angle relative to the sample, rather than as a function of probe position on the sample as is most common. The patterns produced in this manner are referred to as Electron Channeling Patterns (ECP). In an SEM this rocking can be achieved either by the normal scanning action of the electron beam at low magnification (the “conventional” ECP, Figure 1.2(a)), or by a somewhat different configuration that rocks the electron beam at a point on the sample surface (the Selected Area Channeling Pattern, or SACP, Figures 1.2(b) and 1.3). ECPs take the form of agglomerations of bands corresponding to crystal planes within the sample, with interband angles corresponding to the related interplanar angles, and whose widths are inversely related to the interplanar spacings by Bragg’s law. The distinctive banded pattern of ECPs is due to the abrupt change in the BSE yield from one side of the Bragg condition to the other. Indexing of ECPs can be active scan coils \ inactive \ \ can coils ‘ \ \ H-. r_‘ \ r_ ) : I i ' l I l / ...‘ L" \L._.| / ) «lens / / / / ’ / Figure 1.2. Electron Channeling Pattern (ECP) configurations: (a) Conventional ECP; (b) Selected Area ECP (SACP). rocking angle \y sample surface (a) \I’ (b) Figure 1.3. Configuration for producing Selected Area Channeling Patterns: (a) Beam rocking geometry; (b) Example of a SACP. carried out in essentially the same way as for Kikuchi lines in TEM by matching the band width ratios and interband angles to the corresponding interplanar spacings and angles. ECPs formed by viewing at low magnification are commonly referred to as con- ventional ECPs, while those formed by rocking the beam are usually referred to as Selected Area Channeling Patterns (SACP). As the contrast levels involved with elec— tron channeling are fairly small, both methods require fairly fiat, polished samples to form an image so the electron channeling contrast is not overwhelmed by topograph- ical contrast. Figure 1.4 displays a conventional ECP from a silicon single crystal Figure 1.4. Example of a conventional ECP from a Si single crystal. formed by focusing the electron beam at ~144mm working distance for a sample that is at ~11mm working distance; the width of the crystal is ~5mm. Figure 1.5 shows a SACP from FeAl at 12mm working distance. 10 Figure 1.5. Example of a SACP from FeAl. Alignment of samples relative to the electron beam for the conditions required for ECCI is usually conducted by observation of the ECPs from the sample. 1.3.2.1 Microscopic Parameters Use of a scanning beam with low enough convergence angle to minimize the angu- lar range within the beam will maximize the channeling contrast to the point where practical contrast levels can be achieved. As the contrast achievable from electron channeling tends to be fairly low, the probe current necessary to achieve a high enough signal-to-noise ratio for clear imaging must be fairly high. As a consequence, the smaller the scale of the features being imaged, the higher the required source brightness (5) due to the requirement to maintain a small convergence angle (a) and a spot size on the order of or smaller than the characteristic feature size. Bright- ness is defined as the probe current per probe area per solid angle. Joy [14] cites a convergence half angle, a < 10mrad as a practical limit for viewing dislocations, although there is in fact no sharp drop off of contrast. The limitations imposed on 11 beam current by this convergence angle typically make a high brightness source (108 (A cmz/str) or greater, corresponding to a field emission gun) desirable to overcome signal-to-noise restrictions, although this is far from a concrete limit, as individual dislocations have been imaged with as low as 10°(Acm2/str)[17]. 1.3.2.2 Imaging via Channeling Contrast The imaging of individual dislocations via ECCI is made possible by the localized distortions to the crystal lattice surrounding any dislocation structure. This is es- sentially the same cause of contrast as in the TEM, where local strains cause local changes in plane orientation relative to the electron beam. For high-energy elec- trons, the diffracted intensity is strongly dependent on the diffracting plane angle relative to the electron beam, and as a consequence, there is a difference in the frac- tion of the incident electrons diffracted out of the beam used to form the image, based on which region around the dislocation they pass through. Figure 1.6 illis- trates this phenomenon. The principal differences between diffraction contrast in the TEM and channeling contrast in the SEM are that: (1) channeling contrast involves the backscattered component; and (2) an ECC image is formed point-by-point, rather than as a whole image, as occurs with a TEM. This formation of an ECG image on a point-by-point basis may be understood by considering the total BSE yield (a) for each point on the sample. The total a at each point can be broken down into the contribution of Au from each small volume of crystal along the beam trajectory. This local contribution An in turn can be seen to be related to the ECP that might be produced from this particular small volume of crystal, where Au is related to the backscattered intensity represented by the center point of the ECP in question. Small local crystal tilts would thus appear in the corresponding ECPs as small shifts of the entire pattern, and if the shift were to cross a Bragg condition, then there would be a change in Au intensity for that volume electron beam ] [ weaker diffraction (a) stronger diffraction (b) Figure 1.6. Plane bending model of diffraction contrast for a dislocation in a TEM. (a) side view of an edge dislocation parallel to the surface, represented as distorted crystal planes. (b) resultant image from electron beam passing through the crystal. 13 relative to the rest of the sample. This analogy may be carried further by considering not just the center point of the ECP for An, but rather the average over some circle whose radius corresponds to the convergence angle of the beam (see Figure 1.7). Note that while the average BSE intensity within the circle representing the convergence angle of the beam remains the same in Figures 1.7(a) and (c), the average BSE intensity differs for the case of the rotated crystal (Figures 1.7 (b) and (d)) due to a larger convergence angle in (d). This demonstrates the reason for the reduction in channeling contrast with increasing beam convergence angle. Figures 1.8, 1.9, and 1.10 are some examples of how this model of channeling contrast may be viewed for Specific crystal defects. 1.3.3 Choice of Microscopic Imageing Parameters Almost all previous ECCI experimentation / work has been performed at large beam / sample surface tilt angles, principally to increase the total “backscatter” yield from the sample surface and increase the relative incidence of high-energy backscat- tered electrons to improve the signal-to-noise in the ECCI signal [see for example: 13, 15, 17, 24]. Skewing the energy distribution of the BSEs tends to increase the ECCI signal strength because channeling contrast is displayed most strongly by electrons that have undergone few inelastic interactions [9]. This may be understood intuitively in light of the above channeling contrast model, as low energy loss electrons have a path direction that is closer to the incident beam direction and their lower energy loss leads to a smaller change in the Bragg angle being used for contrast, so the diffracting condition remains close to the electron path, keeping the contrast for the high angle elastic events that cause backscatter. The principal reasons historically that an increase in total BSE yield is desired have included inherently low detector acquisition rates, as well as inherently low total probe currents under the desired probe size and convergence angle requirements. The reason 14 Figure 1.7. Effects of local crystal rotation on local SACP, and effects of beam con- vergence angle. (a) Bulk crystal, small convergence angle; (b) Rotated crystal, small convergence angle; (c) Bulk crystal, large convergence angle; (d) Rotated crystal, large convergence angle. scanning electron beam Side view ‘x Resultant image Figure 1.8. Channeling contrast mechanism for a pure edge dislocation parallel to the surface, showing the SACPs expected from each region and the resultant ECC hnage. ;]],:,=jl~]‘ 'I.’ scanning ',]-lr:; 3,; , ',’: electron E,£J,__j_ ”a. ;. beam -.' (-’.- ir’ -- -—Center line of screw Side View bright Resultant image near core, front half plane dark Figure 1.9. Channeling contrast mechanism for a pure screw dislocation parallel to the surface. ECPs are represented for small volumes of crystal near the dislocation core to show the expected sense of contrast in the ECC image. 17 Image Dark Figure 1.10. Expected channeling contrast mechanism for a pure edge dislocation normal to the surface. ECPs are represented for small volumes of crystal near the dislocation core to show the expected sense of contrast in the ECC image. Note that contrast is obtained in this case by change in the Bragg angle for, rather than tilting of, the crystal volume. 18 for a low total probe current is that previous work has been conducted using electron sources with brightnesses (d) on the order of 109A/cm2str or less (representing cold field emission guns), and employing probe diameters on the order of ~10nm. Due to the tortuous experimental setups required to observe a sample under sev- eral different channeling conditions when the sample is highly tilted, ECCI used for analysis of crystallographic defects is more readily performed at low tilt angles, pro- vided there is sufficient contrast to form an image over the noise from the detector. Due to environmental factors inherent to the location of the microscope, limita- tion of the probe size to lenm, as has been done by previous researchers, was not considered critical as the microsc0pe resolution was not dominated by the probe size, but rather by the magnitude of the beam wander. In the present work, probe sizes were allowed to vary, and are estimated to have been in <40nm range (typically 10- 20nm). Finer probe sizes, on the order of ~10nm or less, might provide additional information, although the quoted dislocation image widths in the literature of 15nm- 35mm [12, 25, 13] (including both experimental and theoretical values) might argue in favor of a relaxed probe size. Conversely, theory [13] (as well as simple intuition) would expect differing dislocation image widths depending on both the material as well as the channeling condition, so smaller dp might prove benifical in some instances, provided site conditions can be improved. A combination of higher source brightness (measured very approximatly to be in the low 109A/str-cm2 range) due to the Schottkey thermal field emission gun em- ployed, as well as relaxation of the probe size has allowed sufficient probe current (~2.4nA typically) to be obtained so as to allow a sufficiently strong BSE signal to be produced to offset the gains produced by tilting the sample. Using the local channeling model of defect contrast formation, ECC images taken at low sample tilts will not necessarily display the same form of contrast as images taken at high sample tilts. This is a result of the loss of symmetry in ECPs when 19 taken at large sample tilts. This assymmetry consists of two seperate components: in the first, channeling bands at high angles to the tilt axis lose contrast, then invert contrast with increasing tilt; in the second, channeling bands at low angles to the tilt axis become asymmetric in intensity about the plane trace, with :i:g taking opposite senses of contrast (Figure 1.11) [18,19]. These effects are in part a function of detector position. More comprehensive channeling contrast models based on wave mechanics (see for example [20]) give the same sense of expected contrast. 1.4 The Typical ECC Image Electron channeling contrast images frequently bear some similarity to TEM diffrac- tion contrast images, the principal difference being that images of dislocations taken using ECC are frequently broader. When examining micrographs taken using ECC, it is useful to remember that what is in fact seen is a two dimensional projection of the near surface region of the crystal under examination, with contrast fading out for dislocation lines traveling into the crystal. Dislocations seen end-on appear as points at the center of contrast, while dislocations lying nearly parallel to the surface appear more like lines. Additionally, the width of a dislocation image is also typically a function of its depth, as the image at the surface is narrow, while the image line widens out at greater depths in the crystal. 20 Figure 1.11. Showing schematically the changes that take place in an ECP with surface tilt. (a) No or low surface tilt. (b) High surface tilt. CHAPTER 2 Experimental Procedure This chapter will address the topics of sample preparation, for ECCI and the gen- eral mechanics of the ECC imaging of these samples, then examine more closely the conditions and procedures followed for some specific experiments carried out using ECCI. 2. 1 Sample Preparation Samples for TEM and ECG observation in the SEM were prepared by a variety of methods. In general, samples were prepared from materials in the as-received condition, with little or no effort to control the dislocation morphologies present before preparation. Unless otherwise specified, the only controls over the defect structures present in the material are those inherent to the material. (Semiconductor grade single crystal silicon, for example, is effectively grown dislocation-free.) 2.1.1 Silicon Silicon compression samples were prepared from a single crystal sample constituting an end of a boole of semiconductor grade silicon (P doped, 1.0 W/cm, provided by Virginia Semiconductor) oriented with the compression axis along the (123) direction. 21 22 All samples were prepared with approximately the same orientation relative to the compression axis by orientation of the parent crystal through the use of Lane X-ray back-diffraction. Compression samples in the form of rectangular prisms approximately 5mm x 5mm x 20mm were cut from the parent crystal by first trimming the parent crystal using high speed diamond saws (Accutom—2, saw for preparation of geological sam- ples), then cutting the final sample bars by low speed diamond saw (Buhler), and subsequently grinding to a 600 grit surface using SiC metallographic paper. Some samples were further polished to a 0.3,um surface using SiC and A1203 polishing media. Dislocations were introduced into these samples by uniaxial compression at ap- proximately 750°C to approximately 2% total deformation, then cooling to 450°C and deforming slightly, then cooling under load. All deformation processing was conducted at environmental pressures between 2x10’5 to 1x10"6 torr on an MTS 810 servo-hydraulic load frame with load and displacement controls, equipped with a Centorr high temperature vacuum furnace (model# s60-3x8-WD-0454-A-20, serial# 860-2653). Transmission electron microscopy foils were prepared by cutting the compression samples either in a plane perpendicular to the compression axis or approximately parallel to a principal slip plane, which was identified by slip traces on the crystal surface. Wafers approximately 250-500pm thick were cut from the parent crystal by low speed diamond saw (Buhler) and ground to thicknesses of 100—200,um through the use of successive grades of SiC metallographic paper to a final 600 grit surface finish. Final thinning was conducted by masking the wafers between thin sheets of Teflon (polytetrafluoroethylene) with holes cut through them and repeatedly dipping the assembly in an acid mixture consisting of either a 3:2:1 solution of nitric, acetic, and 23 hydrofluoric acids (for more rapid thinning), or a 10% hydrofluoric acid+90% nitric acid solution. The compressed samples for bulk ECC observation were prepared by slicing ap- proximately parallel to the slip planes (as located by their slip traces) using a diamond saw; mechanically grinding to a 600 grit finish by successive application of SiC pa- pers; mechanically polishing using 600grit, 5pm, 0.3nm, and 0.05pm abrasive slurries; and briefly etching using the same 3:2:1 hydrofluoric/ nitric/ acetic acid mixture men- tioned previously. Additional samples were prepared by aggressively grinding and mechanically polishing undeformed single crystal silicon blocks using the same pol- ishing schedule without subsequent etching; the dislocation flow stress for silicon at room temperature is high enough that the resultant surface dislocation density was low for these samples. 2.1.2 Single Crystal FeAl Single crystal FeAl TEM foils were prepared by first sectioning from a parent crystal using a high speed diamond saw (Accutom-5) and grinding the resultant wafers to approximately 100—250nm thick by sequential application of silicon carbide paper to an ultimate 600 grit finish. After mechanical thinning, samples were annealed in air at BOO—850°C for 1-2 hours to reduce the near surface dislocation. density, and subsequently lightly sanded with 600 grit SiC paper to remove the resultant oxide film. After heat treatment, the samples were thinned to electron transparency by electropolishing at 10V, ~-40°C in a 90% methanol/10% HN03 mixture using a twin jet electropolisher (Tenupol-3). Initial specimens for bulk sample ECC imaging of dislocations were prepared by electropolishing a previously deformed single crystal FeAl compression sample for approximately 2 hours at 10V between -45°C and -30°C in the same methanol/10% HN03 mixture. Some additional samples were prepared by sanding the observation 24 surface to a 600 grit finish through subsequent applications of SiC paper, mechanically polishing down to a final media size of 50mm, and electropolishing at 12V and —50 - -60°C in a methanol/10%HN03 solution for ~30 minutes. As the TEM foil studies and bulk crystal ECCI studies were intended merely to establish the ECCI method, differences in dislocation morphology due to preparation differences for these samples were not considered a point of concern. Samples for surface preparation artifact studies were prepared by either mechan- ically polishing to a final media size of ~300nm, as mentioned previously, or by cutting with a wire electrode using Electrode Discharge Machining (EDM) with an electrode potential of 150V, then electropolishing at -80:t10°C and 15V in the same methanol+10%HN03 solution, interspersed with microscopic observation. The sam- ple for the EDM preparation study was annealed in air at ~1000°C for ~4 hours followed by furnace cooling prior to surface preparation. 2.1.3 Single Crystal NiAl Two different types of N iAl samples were prepared for bulk near-surface dislocation observations. Initial single crystal N iAl samples for bulk dislocation observation were prepared by electropolishing in a methanol+10%HNO3 electrolyte at 17V for 30-40 minutes at -78~60°C, subsequent to mechanical polishing in the same fashion as for FeAl above. NiAl bulk samples for sample preparation studies were prepared by either EDMing as for the FeAl samples above, or by diamond sawing followed by mechan- ical polishing. Samples were subsequently electropolished using an electrolyte of methanol+10%HN03 at -80°C:i: ~8°C and 15V over a number of steps until undis- turbed crystal was reached. 2.2 Microscopy Transmission electron microscopy (TEM) was performed on a Hitachi H-800 (serial #800-17-02) using a tungsten hairpin filament Operating at 200kV, employing both axial brightfield and axial darkfield techniques. Weakbeam techniques were employed where conditions were favorable and the sample foil had sufficiently low curvature that large changes in diffraction intensity were not observed over the areas of interest. Scanning electron microscopy (SEM) was performed on a CamScan 44FE scanning electron microscope, equipped with oil gas diffusion pump and liquid nitrogen cooled cold trap, operating at 20-25kV, employing signals individually and in combination from: a polepiece-mounted annular silicon diode backscattered electron detector; a retractable, side-mounted Yittrium-Aluminum Garnet (YAG) scintillator backscat- tered electron detector, designed for a large solid angle of collection; and a position- able Everheart-Thornley typelbackscattered+secondary electron detector. Both BSE detectors are described in greater detail below. Processing of SEM images was performed both internally on the microscope (re- cursive filtering, frame capture), and externally using the Semper6 software package (Synoptics Ltd.), using a Synoptics Synergy framestore (Synoptics Ltd.) running on a PC. All images processed externally had the scan rate and size externally controlled. 2.2.1 Detector Configurations for ECCI Two different types of detectors were used for imaging samples for ECCI. The first type was a retractable, side-mounted BSE detector similar to the type described in [15], consisting of a scintillator crystal held close to the sample joined to a photomul- tiplier tube by a glass light pipe (Figure 2.1(a)). This detector was custom-built for 1The Everheart-Thornley (or E—T) style electron detector is frequently regarded as a pure secondary-electron detector, although there is a significant sensitivity to backscattered electrons as well. It is for this reason that the author distinguishes between the two. 26 microscope polepiece 4— retractable glass light pipe to photomultiplier I YAG scintilator—l w (a) microscope polepiece ._.__K F Ihole for 7 [electron beam 1 I l \\\“’detector built into (b) Figure 2.1. Layout of the two different BSE detectors used for ECCI. (a) the side- mount detector; (b) the polepiece-mounted Si diode detector. 27 the microsc0pe by the author. The second detector type used was a commercially available, conventional Si diode-type polepiece mounted four- quadrant BSE detector of approximately 25mm overall diameter (Figure 2.1(b)) which sat directly above the sample. 2.2.2 Establishing ECCI Initial experiments to demonstrate channeling contrast imaging of features known to be dislocations were performed by comparing images of prepared TEM foils under theoretical ECCI conditions in the SEM with images taken of dislocations in the TEM from these same areas. The FeAl TEM foils for these experiments were prepared by altering the electropolishing conditions such that multiple thin areas were present on a typical foil. This was done by reducing the electrolyte flow rate. Typically, foils were removed prior to perforation, or subsequent to initial perforation at some point other than the center of the foil. Location of the thin areas in these foils in the TEM was accomplished by scanning the TEM stage in a raster pattern at low magnification until regions thin enough for dislocation observation were found. Further tests involved the less rigorous requirements for imaging the far-field strains of epitaxial misfit dislocation clusters of Si+10%Ge epitaxially grown over a (100) Si substrate [17]. Imaging of these dislocation clusters was performed on a sample provided by the authors of [17], having a 1.2nm thick epitaxial layer. For both tests, ECCI was performed as described in the next section. 2.2.3 Establishment of Microscope Conditions for ECCI A number of steps were performed in order to establish conditions suitable for ECCI imaging of dislocations in the materials examined; the steps common to all materials are presented here. 28 Subsequent to initial alignment of the electron-optical column suitable for routine imaging of sample topography, further adjustment of the SEM electron optics was performed to maximize the probe current (ip) under the conditions chosen for sample viewing (accelerating voltage, E0; working distance, WD), provided that certain min- imal conditions were met. The minimum criteria used were: 1) spot size, dp g the expected feature size (typically on the order of <40nm); 2) the probe convergence half angle, a S 10 mrad. The restriction on (1,, was the result of the desire to maximize the clarity of the features being imaged and was set as an arbitrary maximum; the limit on a was established to promote more efficient channeling contrast from the probe electrons. Generally, the initial step was to maximize probe current for a particular set of working conditions, then check a and tip to determine that conditions suflicient for ECCI were met by the methods following. Approximate measurements of beam convergence angles (a) were performed by focusing on a sharp feature (such as a cleavage edge) at high magnification, then, without changing focus, raising or lowering the sample in the microscope by some known distance (=AZ). The change in working distance due to the lowering of the sample caused the electron beam to intersect the feature of interest over a length z AZx2a+dp (treating a as a small angle), where AZ is the vertical motion of the sample relative to the lens plane and tip is the spot size at best focus. Further, if the spot size is assumed to be negligible when the microscope is in focus, (a fair approximation for most conditions where the change in resolution is large compared to the original spot size), then 2a z(apparent feature size) / (AZ), where the apparent feature size can be obtained by comparison to the microsc0pe scale bar. See Figure 2.2 for a pictorial representation of this procedure. Measurement of the probe current was conducted by measuring the stage current when the electron beam was directed into a Faraday cup attached to the microscope stage. The Faraday cup was produced by drilling a hole partway through a small slab 29 /—-—>-‘ 2a .fi >1 —» , 49: l‘ ' m c: 1dt -. l ‘3 m —> [straws _. II: * ,8 -> \ : g . g ObjeCt edge 2 . U —-> Width a) '6' — —> a \ I: g I con 1 ion Hz] 0 rd :3 m c m X 3’2 Resultant —> mg images -> (a) (b) (C) Figure 2.2. Schematic representation of the steps involved in estimating the conver- gence angle, a. (a) side view of the scanning electron beam intersecting the sample edge in either the in-focus condition (lower), or the out-of-focus condition (upper). (b) representation of the signal intensity from a horizontal line scan for each of the two conditions. (c) representation of the image resulting from each of the two con- ditions. The apparent width of the feature in each image is taken as the apparent feature size. of graphite, using a pin to form an entry hole from the other side, and backing up the drilled slab with another slab of graphite (see Figure 2.3). Currents were measured using a nanoammeter (Hewlett Packard, model 34703A DCV/DCA/Q, serial numbers 1213A06262, 1251A01800). Typical convergence angles used were 20 ~7mrad, and typical probe currents were ip ~2.4 nA. These values varied due to evolution of the Schottkey thermal field emitter tip over its lifetime, and the consequent alterations in the electron-optical system required to achieve a constant emission current from the electron gun. Field 30 hOle\ drilled graphite block graphite Al sample holder to current meter Figure 2.3. Schematic of the components of the Faraday cup used for measurements of probe current. emission tip evolution may be attributed to evaporation of tungsten from the heated tip, as well as electromigration of tungsten on the tip under influence of the applied electric field. 2.2.4 Alignment of the Sample for ECCI Achieving ECC conditions for defect imaging was done by manipulating the orienta- tion of the sample crystal through the use of the ti1t+rotate stage controls. This was done while observing the SACP from the sample using the selected area channeling mode on the microscope. Channeling contrast conditions typically employed involved aligning the center of the SACP to the Bragg condition for the channeling band em- ployed for contrast (see Figure 2.4). As the apparent center of the SACP does not necessarily correspond exactly to the orientation of the scanning beam in the unde- flected condition, observation of the conventional channeling pattern at moderate to high magnifications (in the range of ~80x to 15000x; see for example Figure 2.4(c)) was sometimes used to adjust the sample orientation to the desired channeling condi- tions. The difference between the two conditions is a consequence of slight misalign- 31 Figure 2.4. Schematic showing how a particular crystal in the sample is aligned for ECCI. (a) channeling band used for contrast is aligned the center of the SACP; (b) the conventional ECP at low magnification is aligned to the center of the microscope scan field; (c) the same condition as (b), except with the microscope at somewhat higher magnification. 32 ments of the electron beam and the electron-optical components themselves, which will cause small differences in the angle of the beam emerging from the final lens when no scanning deflection is applied. This final adjustment often involved merely observing the field of interest and tilting the sample to maximize defect contrast. Channeling bands used for contrast were typically of low index. 2.3 Evaluation of Microscope and Sample Condi- tions Required for ECCI, and their Effect on resulting Images Operational conditions corresponding to conditions habitually used for ECCI were es- tablished over numerous microscopy sessions. Initial systemic variation of microscopic parameters (working distance, condenser lens strengths, detector geometry, alignment relative to the Bragg angle for the channeling band used, etc.) was conducted in order to get a working understanding of how the various parameters effected the relevant operational variables (a, ip, signal strength). Later sessions focused on refining im- age quality using this information in a series of trial-and-error tests. Unintentional variations in sample conditions, processing methods, and the like, occasionally also provided insight on the effects of these differences on ECCI. Subsequently, systemic studies of some of these sample processing effects were conducted using electropolish- ing interspersed with ECC observation. 2.4 Application of ECCI The procedures followed for the specific experiments performed to examine topics using ECCI are detailed in the following sections. 33 2.4.1 Analysis of Frank Loops in FeAl An examination of the Frank loops formed in FeAl due to the quenching-in of thermal vacancies was performed through contrast-loss experiments, where contrast of the defect in question is used to determine the distortion that the defect causes to the atomic planes used to form contrast. By systematic determination of which imaging conditions cause a loss in contrast of a defect, characteristic information that describes the defect can be deduced through the same methods as employed for g-b analysis in the TEM. Single crystal F eAl was annealed in air for ~4 hours at 1000°C, followed by furnace cooling at >100°C / hour to ~350°C, followed by air cooling. Due to the comparatively rapid cooling rate [21], this sort of heat treatment would be expected to cause a substantial supersaturation of thermal vacancies. These excess vacancies tend to form dislocation structures of (100) (010) (where the first term represents the Burgers vector of the dislocation, and the second represents the dislocation line direction) dislocations after long annealing times, as described by Fourdeux and Lesbats[22]. Thus for this heat treatment schedule, significant numbers of (100) (010) dislocations would be expected to form. After heat treatment, the sample was then cut by wire EDM at 150V so that the surface plane was close to the (100) plane. All alignment was done using Laue X-ray back-diffraction with a Cu target at an accelerating voltage of 40kV and a target current of 30mA. ElectrOpolishing was conducted using the same methanol+10%HN03 electrolyte as used previously, at -80:l:10°C and 15V for ~22 minutes. ECCI examination was performed using the low-index channeling bands about the (100) pole for channeling contrast. 34 2.4.2 Dislocation Detection Depth Measurements The depth to which contrast can be seen for a single dislocation in silicon was another t0pic examined, and was investigated through the use of stereo images taken of dis— locations. Imaging was performed using ip=2.3nA, dp z20nm, and a=4mrad. BSE signal detection employed the polepiece-mounted four-quadrant solid state detector with a solid angle of detection of ~0.67r str. The (220) channeling band was used for imaging contrast, employing a negative deviation parameter. Images were recorded as single 25 second scans to minimize contamination accumulation. The basic model used to describe the dislocations in real space and their projec- tions as images is laid out in Figures 2.5 and 2.6, which shows all measurements in projection along the axis of rotation for the tilt experiment. In this situation, the dislocation line is modeled as the line connecting the point where contrast is lost in the crystal to the point of emergence from the crystal surface, irrespective of the actual path followed by the dislocation line (Figure 2.6).. Defining the following terms (see Figures 2.5, 2.6): 1 Length of line connecting disappearance point and emergence points of dislocation when projected along the tilt axis. ‘l/J Angle between line described by l and sample surface. (1 Depth of disappearance point when sample is tilted; the detection depth. t Depth of disappearance point measured normal to the surface. 0 Total tilt angle between images, so 9/2 is the angle of tilt to either side of the surface normal. 3 Longer projected distance of I from one micrograph. b Shorter projected distance of 1 from the other micrograph. 35 free surface r ‘1’ . . i ’ ,. vanishing ' , point __,.1; 1 / ’ ’ untilted 59/2. l V----------- a , positive tilt ', 6/2 negative tilt Figure 2.5. Explanation of the terms used for stereo pair measurements. 36 dislocation line connector vanishing line point / / / / / / dislocation line / beyond visible depth Figure 2.6. Top view of the dislocation geometry symbolized in Fig. 2.5. 37 From this, it can be seen that: a=1cos #, b=lcos 343:9, t=dcos % _ t _ dcosg— (2 1) — sing!) — sinip ' COSM . . . and therefore, g 2 55¢. Usmg this last equation, ’t/) can be found from the exper- imental measurements a and b, which in turn allows for the calculation of d from Equation 2.1. 2.4.3 Effects of Channeling Conditions and Deviation Pa- rameter on the ECC Image In order to expand the range of information available to later microscopists for in- terpretation of ECC micrographs, attempts were made to correlate the appearance of dislocation images to the channeling conditions used to obtain them. All imag- ing was performed using ip=2.3nA, dpz20nm, and az4mrad. Backscattered electron (BSE) signal detection was achieved by the use of a conventional polepiece-mounted four-quadrant solid state detector with a solid angle of detection of ~0.67r str. Single crystal semiconductor grade silicon was prepared for imaging by mechan- ically grinding to introduce near surface dislocations, then mechanically polishing using silicon carbide and alumina down to a final media size of 50nm. Images were taken employing a number of low index channeling bands for con- trast, by tilting the sample through a range of about 22° about a single axis. This simulates the conditions likely to be encountered in many in-situ deformation exper- iments, where some constraint on orientation manipulation of the sample is imposed by the sample loading frame. Where possible, images were acquired using positive and negative values of g with :ts, where g and s are used in the same sense as for 38 TEM electron diffraction notation (Figure 2.7). Images were acquired by time av- eraging for approximately 32 seconds using the internal recursive filtering algorithm with which the microscope is equipped. 39 centerline centerline of (b): s>0, a positive (c): 3:0, 9 negative (d): l<0, 9‘ positive Figure 2.7. Illistration of the definitions of g and s in terms of selected area channeling patterns. (a) s=0, g set as positive; (b) s>0, g still positive; (c) s=0, g negative; g positive, s<0. g is essentially the diffraction vector pointing towards the channeling plane. CHAPTER 3 Results and Discussion 3.1 Evaluation of ECCI System Setup The equipment initially purchased for ECCI (CamScan44F E, Semper6 image pro- cessing software package with provisions for direct microscope beam control, forward- scattered BSE detector) was found to be adequate for ECCI under certain Operating conditions. The principal limitations on attaining contrast were due to site problems: mechanical vibration, due to placement of the instruments on the top floor of the building; AC magnetic fields (principally 60 Hz), due to numerous sources, including lighting, power supplies, and one or more strong unidentified sources; and ground loops, due to differences between the microsc0pe ground potential and the ground potentials for other instruments electrically connected at ground to the microsc0pe. Partial mitigation of the vibration problem was achieved by installation of an additional vibration damping table underneath the microscope column and vacuum system, while AC magnetic field deflections of the microscope beam were reduced by installation of an active-feedback AC field compensation system (Linear Research Associates). Ground loops were reduced by re-routing the ground connections for all instruments connected to the microscope to the microsc0pe itself. Due to these external sources of noise, the resolution under the appropriate conditions for ECCI 40 41 is limited by some combination of AC magnetic field deflections, vibration, and spot size. 3.1.1 Initial ECCI Experiments Demonstration of ECCI of individual dislocations was done by viewing the same T EM foil both in the TEM and in the SEM via ECCI. This was done using regions of the foil that had thinned but not perforated, employing approximately the same g for diffraction contrast (TEM) and channeling contrast (SEM). Comparison of images of dislocations taken via TEM to channeling contrast images of those same regions yielded features common to both images which were identified as dislocations in the TEM images. Due to the strong dependence of the backscattered signal on sample thickness, a strong variation in background intensity in the regions of interest made it impractical to record the ECC images. The smallest rate of change in sample thickness over the regions of interest in these TEM foils was found to occur at thin regions away from any perforated region, but even for these regions, images typically had a very narrow band of non-saturated contrast. These regions of acceptable signal level were typically only a few microns wide under detector gain settings sufficient to image the channeling contrast. As demonstrated in [17], clusters of epitaxial misfit dislocations from Si+10%Ge over (100)Si produce sufficient near-surface strain to give clear channeling contrast. At suitable epilayer thicknesses, this strain can be approximated as belonging to an edge dislocation with a Burgers vector of some multiple of <100>, which can be viewed via ECC with substantially looser electron-optical requirements [17]. The ECC from the far-field strain of epitaxial misfit dislocation clusters can be seen in Figures 3.1—3.8 as lines parallel to the <100> directions. These images cor- respond to a range of accelerating voltages from 25kV to 7kV and were formed by employing alternately the polepiece BSE detector signal or the side-mount BSE de- 42 tector signal. As can be seen, channeling contrast of these features was achieved at accelerating voltages of 7kV at low surface normal tilt angles (<10°), using both the the side~mounted and polepiece—mounted BSE detectors, with contrast comparable to operations at higher accelerating voltages. The appearance of Figures 3.2, 3.4, and 3.6, for example, at 25kV, 20kV, and 14kV, respectively, is substantively identical, despite the differences in accelerating voltage. Only at an accelerating voltage of 7kV (Figure 3.8) does the contrast appear different from the previous images, and examination of the image reveals that this may be attributed largely to contamina- tion build-up on the surface significantly obscuring the underlying crystallinity; note that the shifted image field reveals significant contrast differences between the areas with longer and shorter beam exposure times. ECC image differences due to detector location are much more substantial. Examination of Figures 3.1 (polepiece mount de- tector) and 3.2 (side mount detector) reveals two characteristic features common to all of the images obtained using each detector: firstly, that ECC was more pronounced when using the side-mount detector for dislocation clusters running approximately perpendicular to the plane containing the microscope axis and the detector axis than for dislocation clusters within this plane; and secondly, that contrast for these same features was somewhat less pronounced when viewed with the polepiece mounted detector. (Comparison of other pairs shows these same characteristics.) Finally, it may be noted that the images obtained using the polepiece mounted detector showed a greater loss of sensitivity at an accelerating voltage of 7kV (Figure 3.7 as Fig- ures 3.1, 3.3, 3.5) than for those obtained using the side- mount detector (Figure 3.8 vs Figures 3.2, 3.4, 3.6). 3.1.2 Typical Working Parameters The following list outlines the typical conditions employed when doing ECCI on the MSU microscopic setup, including comments on variations from these conditions. 43 Figure 3.1. ECC image of dislocation clusters in Si+10%Ge epitaxially grown over (100) Si at 25kV accelerating voltage using the polepiece-mounted BSE de- tector. 20mm working distance, 570x magnification, 9° tilt from surface normal. ."3' ' {1% Figure 3.2. ECC image of SiGe epitaxially grown over (100) Si at 25kV accelerating voltage using the side-mounted BSE detector. The detector is located towards the bottom of the figure. 20mm working distance, 570x magnification, 9° tilt from surface normal. 30pm 44 Figure 3.3. ECC image of SiGe epitaxially grown over (100) Si at 20kV accelerating voltage using the polepiece-mounted BSE detector. 20mm working distance, 570x magnification, 9° tilt from surface normal. 30pm Figure 3.4. ECC image of SiGe epitaxially grown over (100) Si at 20kV accelerating voltage using the side—mounted BSE detector. The detector is located towards the bottom Of the figure. 20mm working distance, 570x magnification, 9° tilt from surface normal. — 30pm 45 a @ [ll-I Figure 3.5. ECC image of SiGe epitaxially grown over (100) Si at 14kV accelerating voltage using the polepiece—mounted BSE detector, 570x magnification, 9° tilt from surface normal. 30pm Figure 3.6. ECC image of SiGe epitaxially grown over (100) Si at 14kV accelerating voltage using the side—mounted BSE detector, located towards the bottom of the figure. 570x magnification, 9° tilt from surface normal. — 30am 46 23433?” Figure 3.7. ECC image of dislocation clusters in Si+10%Ge epitaxially grown over (100) Si at 7kV. Note the difference in image contrast between the areas with greater and lesser amounts of beam-fixed contamination. Polepiece mounted BSE detector, 200 second frame time, 23 mm working distance, 570x magnification, 9° tilt from surface normal. _ 30pm W0 [[1111 Figure 3.8. ECC image of SiGe epitaxially grown over (100) Si at 7kV accelerat- ing voltage using the side—mounted BSE detector. The detector is located towards the bottom of the figure. Note the contrast differences due to beam-fixed contam- ination. 23mm working distance, 570x magnification, 9° tilt from surface normal. _ 30pm 47 These values approximate conditions which were found to give fairly consistent and reproducible imaging conditions for ECCI as a result of attempts to Optimize ECCI quality while working with the microscope. Accelerating voltage: 25kV. Accelerating voltages as low as 7kV were used for ECCI under some conditions, although lower accelerating voltages tended to produce lower signal levels, as lower accelerating voltages were found to yield a substantially lower maximum probe current (see Table 3.1 for the measured relationship between accelerating voltage, E0, probe current, 1,,, and probe convergence angle, 20). Table 3.1. Representative relationships between relevant probe parameters for the CamScan44FE at 10mm working distance with a 50pm aperture. | Re I in l 20 l 25kV ~3.6nA ~7mrad 22.5kV ~1.7nA ~9mrad 20kV ~1.7nA ~12mrad Aperture: 50nm. Generally, smaller apertures produced better results, as a smaller aperture produces a smaller probe convergence angle. As the probe convergence an- gle strongly affects ECC, the smaller aperture in turn tends to produce better contrast. For some conditions where convergence angle restrictions were not arduous, such as at longer (~12mm) working distances, a 70pm aperture was also found to be adequate. 48 Working Distance: 10mm-15mm. Independent of BSE detector position, ECC from individual dislocations was found to show up poorly (if at all) at working distances much greater than ~15mm (as reported by the SEM console, based on objective lens currents required for focus). The reasons for this are not entirely clear, although it is hypothesized that the greater deflections of the scanned beam by stray magnetic fields at longer working distances suffices to drive the signal to noise on the resulting scanned images too low. In addition, when viewing samples at shorter working distances (~8mm or less) with the polepiece mounted BSE detector, a significant loss of signal strength was found to occur due to the large solid angle occupied by the dead space and hole in the center of the the annular BSE detector. Figures 3.9, 3.10, 3.11, 3.12 show the same low-magnification image of a crystal of silicon as the working distance is reduced from 12mm (Figure 3.9) to 8mm (Figure 3.12). Note how for the center of the image, where the signal used for high- magnification ECC images will be produced, the relative signal strength as compared to the outer region of the image decreases as the working distance is reduced. This loss of signal strength can be seen most dramatically by comparing Figures 3.11 (9mm) and 3.12 (8mm), which were recorded using the same contrast and brightness settings for both micrographs. Note how the center region of Figure 3.12 has an apparently lower signal level, even though the sample is closer to the detector. For channeling contrast features requiring less stringent beam control, longer working distances (over ~20mm working distance) were found to be adequate. Some examples of this can be seen in Figures 3.1-3.8, where the nature of the features examined make for easier collection a ECC signal sufficiently strong to form an image. 49 Figure 3.9. 12 mm working distance BSE image using the polepiece detector, showing the [110] ECP for NiAl. Surface is normal to microscope axis. 15x magnification. — 1mm Figure 3.10. 10 mm working distance BSE image using the polepiece BSE detector, showing the 110 ECP for NiAl. 15x magnification. — 1mm 5O Figure 3.11. 9 mm working distance BSE image using the polepiece BSE detector, showing the 110 ECP for MA]. 16x magnification. _ 1mm Figure 3.12. 8 mm working distance BSE image using the polepiece BSE detector, showing the 110 ECP for NiAl. Note that this image and 3.11 employ the same zero level and gain settings for the BSE signal. 17x magnification. 1mm 51 Sample normal: Typically ~25° or less. When imaging using the polepiece mounted BSE detector, surface normal tilts of less than ~25° relative to the microscope axis were commonly used, although greater tilts are possible. In general a loss of signal strength and reduction in contrast were observed for the higher tilts. This is shown for a series of images in Figures 3.13, 3.14, 3.15, 3.16, which show the same region of a crystal of NiAl from a relatively minor surface tilt (~7.5° for Figure 3.13) up to almost 30° tilt for Figure 3.16. Notice that as the tilt increases, the dislocation contrast as well as the overall signal strength both decrease. Figure 3.13. ECC image of dislocations in NiAl taken using the polepiece mounted BSE detector at ~7.5° tilt from horizontal. Channeling contrast from (200) channel- ing band. 25kV accelerating voltage, 12mm working distance, recursively filtered 32 frame average, 14600x magnification. 1pm 52 Figure 3.14. ECC image of approximately the same field of view as in Fig. 3.13 of dislocations in NiAl, taken using the polepiece mounted BSE detector at ~12.5° tilt from horizontal. Channeling contrast from (200) channeling band. 25kV accelerat- ing voltage, 12mm working distance, recursively filtered 32 frame average, 14600x 1pm magnification. Figure 3.15. ECC image of approximately the same field of view as in Fig. 3.13 of dislocations in MA] taken using the polepiece mounted BSE detector at ~23.5° tilt from horizontal. Channeling contrast from (200) channeling band. 25kV accelerat- ing voltage, 12mm working distance, recursively filtered 32 frame average, 14600x 1pm magnification. 53 Figure 3.16. ECC image of approximately the same field of view as in Fig. 3.13 of dislocations in NiAl taken using the polepiece mounted BSE detector at ~29.5° tilt from horizontal. Channeling contrast from (200) channeling band. 25kV accelerat- ing voltage, 12mm working distance, recursively filtered 32 frame average, 14600x magnification. 1pm Convergence Semi-Angle: 10mrad or less. Probe convergence semi-angles of 10mrad or less were found to exist for all practical working distances tried using the 50am aperture. If an inverse scaling between working distance and convergence angle is assumed, a S 10mrad is expected to extend down to around 4mm working distance. Choice of Detector: Typically the polepiece-mounted BSE detector. ECCI images acquired using the low angle scattered electrons (collected by the side-mounted BSE detector; Figure 3.17) were found to provide ECC images of quality comparable to images formed using high angle scattered electrons (collected by the polepiece mounted detector; Figure 3.18) while at low sample 54 Figure 3.17. ECC image taken using the side mounted BSE detector. Channeling contrast from (211) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. Note that for this sample and holder geometry, it was not possible to move the detector close to the sample, leading to a small solid angle of collection. —— 1pm tilts. Note that the principal difference between the images formed using the two detectors is the signal strength, which is most strongly dependent on the solid angle of collection. For Figures 3.17 and 3.18, the side mount BSE detector was partially obscured by the sample stage, so the resultant image quality (3.17) is worse than for the polepiece-mounted detector (3.18). 3.1.3 Other Topics Concerning ECCI with the Cam- Scan44FE 3.1.3.1 Rotation Between the Image and Channeling Pattern Some other points were found to be worth mention regarding ECCI using the Cam- Scan44F E, either because of differences between the present work and other work reported in the literature, or for practical reasons to assist others in following the 55 Figure 3.18. ECC image taken using the polepiece mounted BSE detector. Chan- neling contrast from (211) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. 1am present work. Due to the different ways in which the electron optics are used to form both electron channeling patterns and real-space images of the sample, a rotation between the two might be expected, and was in fact found to exist. This rotation was found to be strongest for shorter working distances, and is a result of the large differences in objective lens strengths required to form the SACP condition as the regular imaging condition. The amount of rotation between the two conditions is a function of working dis- tance, and is larger for shorter working distances. At 11mm working distance, the image—)SACP rotation is ~36°clockwise. This rotation is due to two factors. The first factor is that electrons are rotated about the lens axis of a magnetic lens, and the amount of rotation is directly related to the strength of the lens. The second factor is due to the fact that the final lens for a SACP is used only to redirect the electron beam to a specific place on the sample surface, and is therefore much weaker than 56 when it is used to focus the beam (Figure 3.19). Conventional ECPs formed from a large crystal while focused on the crystal surface obviously display no rotation, but appear “unfocused” due to the relatively large convergence angle of the beam; cor— recting for this apparent “defocus” by weakening the Objective lens and focusing far below the crystal surface effectively employs the same lens strength as in the SACP case, and so will rotate the ECP relative to the focused image. electron beam \\\‘ final lens electron Visa/m (a) strong lens (b) weak lens Figure 3.19. Differances between the final lens strength between the regular imaging condition, (a), and the SACP condition, (b). Note that the final lens is only used to deflect the electron beam in the SACP case, and is therefore much weaker. In addition to a simple relative rotation, there is also an inversion between an image and its SACP for diffraction and channeling information. This is due to the differant way in which SAC mode uses beam deflection for scanning, so that a deflec- tion that causes a positive angular deflection off the microscope axis at the sample in imageing mode will cause a negaitve deflection off the microscope axis in SAC mode. Thus, when determining the direction of a plane normal from a SACP Of an image, 57 the plane normal from the SACP would undergo a counterclockwise rotation and an inversion to reach the correct dircetion for the image. 3.1.3.2 Use Of Independent Image Processing Use of the Semper6 software package for external beam control and image processing was found to be unnecessary, as images produced using the Semper6 package (F ig- ures 3.20, 3.21, 3.22) were found to be comparable in quality to images produced using the inherent image processing functions of the CamScan44FE (for example Figures 313- 3.18). 3.2 Evaluation of Sample Conditions 3.2.1 Sample Preparation As sample preperation artifacts were one of the potential disadvantages to using the TEM to study crystalline defects as opposed to ECCI, some examination of the general form of preperation artifacts as they appear for ECCI and their of rectifying measures was preformed for a variety of samples. A number of different sample materials and surface preperation routes were examined to gain perspective on how these artifacts were dependent on both the material and the preperation route taken, and their effect on ECCI. None too surprisingly [23], ECCI sample surface preparation was found to strongly affect the observed near-surface crystal defect concentrations and geometries. 3.2.1.1 NiAl Samples of single crystal NiAl were prepared by two different methods: a more conven- tional metallographic polish followed by electropolishing, and electropolishing after electrode discharge machining. 58 Figure 3.20. Image of dislocations in Si taken using external beam control and im- age processing via the Semper6 software package. Signal from the side-mount BSE detector, 6 frame Kalman (recursive) filter, 10.9 second frame time (26.8psec. dwell time). ~14400x magnification. — 1pm 59 Figure 3.21. Image of dislocations in Si taken using external beam control and im- age processing via the Semper6 software package. Signal from the side-mount BSE detector, 6 frame Kalman (recursive) filter, 10.9 second frame time (26.8psec. dwell time). ~14400x magnification. _— 1pm 60 Figure 3.22. Image of dislocations in Si taken using external beam control and im- age processing via the Semper6 software package. Signal from the side-mount BSE detector, 60 frame Kalman (recursive) filter, 1.136 second frame time (2.8psec. dwell time). ~14400x magnification. 1pm 61 Electrode Discharge Machining (EDM) The as—EDMed NiAl sample surface displayed extensive surface damage (Figure 3.23), consisting principally of material removed by spark erosion re—deposited on the machined surface, as well as cracking of Figure 3.23. E—T image of electrode discharge machined NiAl surface." 25 second frame time, 300x magnification. _ 100nm the substrate, presumably due to thermal shock from local discharge heating. Higher magnification views (Figure 3.24) of the flatter “plateau” areas shows significant topography on the surface. Not surprisingly, ECCI examination of these areas (of which Figure 3.25 is typical, showing the same area as Figure 3.24) yields little useful information, as the dominant contrast mechanism is topographical. After ~50pm of material removal by electropolishing, very little fine-scale to- pography remained (Figure 3.26), the dominant features being residual topography following the as-electrode discharge machined (EDMed) surface features. These fea- tures in the as-EDMed surface took the form of ridges perpendicular to the cutting direction. The characteristic scale of these features was on the order of the EDM wire 62 Figure 3.24. E—T image of the as electrode discharge machined (EDMed) N iAl surface. 14700x magnification. _ 1pm Figure 3.25. BSE image of as-EDMed NiAl surface. l4700x magnification. 63 Figure 3.26. E-T image of EDMed NiAl after ~50pm material removal by electropol- ish. White feature is an inclusion unaffected by the electropolish. 300x magnification. 100nm gauge. At magnifications appropriate for ECCI, topographic contrast was observed to be effectively nonexistent (Figure 3.27), and channeling contrast of dislocations to be dominant over topographical contrast in backscatter images (Figure 3.28). For the electropolishing conditions given, this amount of material removal corresponded roughly to ~30 minutes of electropolishing. Further material removal by electropolishing (for ~100pm total removed) did not significantly alter the residual surface topography from wire EDM, but does appear to have yielded a slightly lower surface dislocation density than for the ~50pm removal case. Comparison of the apparent dislocation density in Figures 3.29 and 3.30, with ~100um removed, to the apparent dislocation density in Figure 3.28, with ~50pm removed, appears to show a slightly lower surface dislocation density in the ~100,um removed condition. 64 Figure 3.27. E—T image of electropolished NiAl surface after EDM. 14700x magnifi- cation. 1pm Figure 3.28. Image of the same area shown in Fig. 3.27 in N iAl showing the dominance of ECC features over topographic features after ~30 minutes of electropolishing. 32 frame average ECCI using the (110) band for channeling contrast. 14700x magnifi- cation. —— 1pm 65 Figure 3.29. BSE image of EDMed, then electropolished NiAl after ~100pm re- moval. 32 frame average ECCI using the (110) band for channeling contrast. 14600x magnification. 1pm Figure 3.30. BSE image of EDMed, then electropolished NiAl after ~100pm re- moval. 32 frame average ECCI using the (110) band for channeling contrast. 14600x magnification. 1pm 66 Mechanical Polishing The as—mechanically polished NiAl surface (Figure 3.31) retained a fair amount of polishing media, as is typical for many metallagraphically Figure 3.31. E-T image of same area as Fig. 3.32 in NiAl. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. 14600x magnification. _ 1pm prepared, unetched specimens. In addition to the foreign decorating particles, fine gouges on the order of the residual media size can also be seen in both the E-T signal image (Figure 3.31) and the backscattered electron image (Figure 3.32). Both images show the same area at the same electron channeling conditions, shown in Figure 3.33. As can be inferred from the extremely poor band contrast and poorly-defined band edges in this SACP, the quality of the surface crystallinity is low. After five minutes material removal via electropolishing, the SACP quality ob- tained from the surface improved significantly (Figure 3.34). Channeling contrast images taken under these surface conditions are dominated at low magnifications by “scratch” features. Imaging of these features at higher magnification (Figure 3.35) shows them to be composed of varying dislocation 67 Figure 3.32. BSE image of NiAl under channeling conditions as given by Fig. 3.33. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recur- sively filtered 32 frame average, 14600x magnification. 1pm Figure 3.33. SACP from as-mechanically polished NiAl. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. 68 Figure 3.34. SACP from mechanically polished, 5 minute electropolished NiAl surface showing the electron channeling conditions for Figs. 3.35— 3.38. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. “Mottling” near center and edges due to sweeping of beam across sample during rocking to form SACP. Figure 3.35. 14600x magnification electron channeling contrast image from 5 min. electropolished NiAl surface. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. —_ 1pm 69 densities, presumably due to subsurface deformation of the sample by the grinding media used for sample preparation. Going from higher to lower magnification for the same region shows how the “scratch” features emerge from dislocation clusters as shown in Figures 3.35-3.38. Note how the cumulative strain fields from the individual Figure 3.36. 7300x magnification electron channeling contrast image centered on Fig. 3.35 of 5 min. electropolished NiAl surface. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. Note variations in dislocation density. 2pm dislocations become the dominant features at lower magnifications (Figure 3.38). TO highlight exactly which features might be attributable to topographic contrast, images using the E-T signal (Figure 3.40) and the BSE signal in difference mode (Figure 3.41) were taken of the same area as Figure 3.38. (The BSE signal used in difference mode uses the directional componant of the BSE signal to gain high sensitivity topographic information.) Note the scan squares in the E—T image showing the location of earlier images, and the faint indications of dust particles (upper left) and voids. Note that in Figure 3.41 the bright /dark highlighting of voids and surface dust particles, and the Figure 3.37. 3650x magnification electron channeling contrast image centered on Fig. 3.36 of 5 min. electropolished NiAl surface. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. _ 4pm Figure 3.38. 1200x magnification electron channeling contrast image centered on Fig. 3.37 of 5 min. electropolished NiAl surface. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. Note that the dominant features are now due to the combined strain fields from many dislocations, and begin to take on the appearance of the “scratches” of Fig. 3.39. — 10pm 71 Figure 3.39. 80x magnification BSE image showing a convolution of conventional ECP and electron channeling contrast from “scratch” features after 5 minutes elec- tropolish in NiAl. Channeling contrast from (110) channeling band. 25kV accelerat- ing voltage, recursively filtered 16 frame average. — 100nm Figure 3.40. 1200x magnification, 25kV accelerating voltage, recursively filtered 32 frame average E-T image of the 5 min. electropolished NiAl surface. Same area and conditions as Fig. 3.38; Note scan squares from Figs. 3.35— 3.38. — lOum 72 Figure 3.41. 1200x magnification, 25kV accelerating voltage, recursively filtered 32 frame average. Same area and conditions as Fig. 3.38. BSE image in “A-B” difference mode. —— 10pm minimization of contrast from electron channeling is the result of using the polepiece detector in difference mode. The principal feature of note in this figure are the low levels of “mottled” contrast most likely attributable to topography. (The gradual brightening of the image from left to right is attributable to the small angular change of the scanning electron beam altering the beam-sample-detector geometry, and thus affecting the relative numbers of BSEs being received by each half of the detector.) Further electropolishing for another 5 minutes (Figure 3.42), to a total of 10 minutes, yielded little change in the appearance of the “scratch” features. Closer examination, however seems to indicate a lower surface dislocation density from the longer electropolish time. This can be seen by comparing Figures 3.36 (5 minutes electropolish) and 3.43 (10 minutes electropolish). Electropolishing for further time increments (Figures 3.44, 3.45, 3.46, 3.47) reduces the apparent sheer numbers of the “scratch” features, as well as their visibility, although after 47 minutes total electropolishing time a few “scratch” marks were still 73 Figure 3.42. BSE image showing a convolution Of conventional ECP and electron channeling contrast from “scratch” features after 10 minutes electropolish. Chan- neling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 80x magnification. _ 100nm Figure 3.43. Channeling contrast image using contrast from the (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. Note scan square artifact in center of image. 7300X magnification. 2pm 74 200 I‘m Figure 3.44. BSE image of NiAl showing a convolution of conventional ECP and electron channeling contrast from “scratch” features after 17.5 minutes electropol- ish. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 80x magnification. — 100nm Figure 3.45. ECC image of NiAl mechanically polished then electropolished for 29 minutes showing “scratch” features. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 130x magnification. Note that the oval “bright-dark” features are ECC features consisting principally of clusters of end-on dislocations at either edge, separated by a region of low dislocation density. — 100nm 75 100 pm 0 Figure 3.46. ECC image of NiAl mechanically polished then electropolished for 39 minutes showing “scratch” features. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 130x magnification. _— 100nm Figure 3.47. ECC image of NiAl mechanically polished then electropolished for 47 minutes showing “scratch” features. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 220x magnification. — 100nm 76 visible. This reduction in “scratch” number and intensity was accompanied by a simultaneous reduction in dislocation density in the “scratch” features themselves. Comparing Figures 3.48 (17.5 min total), 3.49 (29 min total), 3.50 (39 min total), and 3.51 (47 min total) shows that the general trend is for lower dislocation densities Figure 3.48. ECC image of NiAl mechanically polished then electropolished for 17.5 minutes. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 14600x magnification. — 1pm and less clustering with greater electropolishing time, giving rise to less pronounced differences in local orientation leading to the “scratch” features. This dislocation clustering is exemplified by comparison of the dislocation densities at different points on the same sample. Figure 3.49 shows the dislocation density from a region chosen on one of the “scratch” features, while Figure 3.52 was taken from an arbitrarily chosen point not on a “scratch” feature; note the higher apparent dislocation density in Figure 3.49. (This differance can be seen by comparing the ~10 dislocation features exiting the surface in Figure 3.49 vs the one exiting dislocation 77 Figure 3.49. ECC image of NiAl mechanically polished then electropolished for 29 minutes, on a “scratch” feature. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 14600x magnification. — 1pm Figure 3.50. ECCI of NiAl mechanically polished then electropolished for 39 min- utes, on a “scratch” feature. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 14600x mag- nification. _— 1pm 78 Figure 3.51. ECCI of NiAl mechanically polished then electropolished for 47 min— utes, on a “scratch” feature. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 14600x mag- nification. — 1pm Figure 3.52. ECCI of NiAl mechanically polished then electropolished for 29 minutes, away from any specific “scratch” features. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average BSE image, 14600x magnification. — 1pm 79 in Figure 3.52.) 3.2.1.2 FeAl Samples of a nominal single crystal of FeAl were also prepared by the same steps as used for NiAl, namely a conventional metallographic polish followed by electropolish- ing, and electropolishing after electrode discharge machining. Mechanical Polishing The surface of the as-mechanically polished FeAl sample displayed strong topography (Figure 3.53), and images taken under ECCI conditions (Figure 3.54) display a dominance of topographic contrast. Figure 3.53. E-T image of as-mechanically polished FeAl surface. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. 1pm After 5 minutes of electropolishing, the SACP quality (Figure 3.55) from the pol- ished surface showed a great increase in band edge sharpness relative to SACPs from the as-mechanically-polished surface, while BSE images taken under ECCI conditions 80 Figure 3.54. BSE image of as-mechanically polished FeAl under channeling contrast conditions of the same area as shown in Fig. 3.53. Channeling contrast from (200) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. 1pm Figure 3.55. SACP corresponding to the channeling conditions for Fig. 3.56. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. 81 displayed good dislocation contrast without noticeable topographic effects (Figure 3.56). LL04 an Figure 3.56. BSE image of FeAl under channeling conditions as given by Fig. 3.55. Channeling contrast from (200) channeling band. 25kV accelerating voltage, recur- sively filtered 32 frame average, 14600x magnification. 1pm Further electropolishing, to a total of 10 minutes, produced no noticeable further improvement to SACP quality (Figure 3.57), as well as no significant change in dislocation morphology or density. (Figure 3.58; note the higher contrast levels used in comparison to Figure 3.56.) Electropolishing for a total of ~22% minutes again showed no noticeable change in dislocation morphology (Figure 3.59). Additional electropolishing to a total of ~48 minutes showed no further changes in dislocation structure, implying that polishing damage had been effectively removed. Electrode Discharge Machining (EDM) The as-EDMed FeAl surface, as shown in Figure 3.60, resembled closely the as—EDMed NiAl surface (Figure 3.23), with 82 Figure 3.57. SACP from mechanically polished FeAl after 10 minutes electropolish. 25kV accelerating voltage, recursively filtered 32 frame average. Figure 3.58. BSE image of mechanically polished, 10 minute electropolished FeAl. Channeling contrast from (200) channeling band. 25kV accelerating voltage, recur- sively filtered 32 frame average, 14600x magnification. 1pm 83 Figure 3.59. BSE image of mechanically polished, 22% minute electropolished FeAl. Channeling contrast from (200) channeling band. 25kV accelerating voltage, recur- sively filtered 32 frame average, 14600x magnification. 1pm Figure 3.60. E-T image of electrode discharge machined FeAl surface. 25 second frame time, 300x magnification. 100nm 84 re-solidified material deposited on a cracked substrate. Six minutes of electropolish was sufficient to remove enough topography to allow ECCI imaging of dislocations in a limited area (Figure 3.61), although extensive Figure 3.61. BSE image of electrode discharge machined FeAl surface after 6 minutes electropolish. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14700x magnification. 1pm large-scale topography and cracks were still evident (Figures 3.62, 3.63) over much of the sample. Selected area channeling patterns were reflective of this; Figure 3.64 shows moderately sharp channeling band edges, with “flutings” due to the rocking beam intersecting the sample surface at different points on the surface, where cracks cause different crystal orientations. A further six minutes of electropolishing (12 minutes total) was sufficient to remove most of the cracked material as well as much of the remaining topography. Figure 3.65 shows an area typical of much of the surface, while Figure 3.66 shows a region of the surface slightly depressed from the rest with less complete electropolishing. 85 Figure 3.62. E-T image of electrode discharge machined FeAl surface after 6 min- utes electropolish. Recursively filtered 32 frame television scan—rate average, 300x magnification. 100nm “J \ Figure 3.63. BSE image of electrode discharge machined FeAl surface after 6 minutes electropolish. Recursively filtered 32 frame average, 300x magnification. 100nm 86 Figure 3.64. SACP of electrode discharge machined, 6 minute electropolished NiAl surface. Figure 3.65. BSE image of electrode discharge machined FeAl surface after 12 minutes electropolish. Recursively filtered 32 frame average, 300x magnification. 100nm 87 5... 59pm Figure 3.66. BSE image of region of incomplete surface electropolish after 12 minutes electropolish. 200x magnification. —— 100nm This was presumably due to lateral motions of the EDM wire during cutting. SACPs from the majority of the surface were of higher quality, consistent with the smoother surface (Figure 3.67). Note the bright and dark “rings” in the SACP Figure 3.67 which are caused by the slow undulations in the sample surface seen in Figure 3.65. Although 12 minutes of electropolish was sufficient to remove most topography, differences in subsurface dislocation density remained. Figures 3.68, 3.69, and 3.70 show differing dislocation densities from regions ~20pm (Figure 3.69) and ~150pm (Figure 3.70) away from Figure 3.68. Note the differences in dislocation density, even over the same image — unexpected for an annealed single crystal. Twenty-two minutes total electropolish time appeared to significantly reduce re- gional differences in dislocation morphology (Figures 3.71, 3.72, and 3.73), as well as reduce further the magnitude of the surface undulations, as evidenced by Figure 3.74, where only the faintest contrast can be seen from surface undulations. It may be concluded from this that electropolishing times in excess of ~22 minutes under 88 Figure 3.67. SACP from electrode discharge machined FeAl after 12 minutes elec- tropolish. 25kV accelerating voltage, recursively filtered 32 frame average. Figure 3.68. BSE image of EDMed FeAl surface after 12 minutes electropolish. Chan- neling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. Bright features are not topographic contrast. — 1pm 89 Figure 3.69. BSE image of EDMed FeAl surface after 12 minutes electropolish, ~20nm away from Fig. 3.68. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. Figure 3.70. BSE image of EDMed FeAl surface after 12 minutes electropolish, ~150,um away from Fig. 3.68. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnifica- tion. — lflm 90 Figure 3.71. BSE image of EDMed FeAl surface after 22 minutes electropolish. Chan- neling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. 1pm Figure 3.72. BSE image of EDMed FeAl surface after 22 minutes electropolish, ~140pm away from Fig. 3.71. Channeling contrast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnifica- tion. _ lflm 91 Figure 3.73. BSE image Of EDMed FeAl surface after 22 minutes electropolish, ~140nm away from Fig. 3.72 and ~140pm away from Fig. 3.71. Channeling con- trast from (110) channeling band. 25kV accelerating voltage, recursively filtered 32 frame average, 14600x magnification. 1pm Figure 3.74. SACP corresponding to the channeling conditions for Figs. 3.71—3.73. 25kV accelerating voltage, recursively filtered 32 frame average BSE image. 92 the conditions used are sufficient to remove EDMing artifacts in annealed FeAl. 3.2.2 Si The easy cutting properties and chemical inertness of silicon necessitated different sample preperation routes than for the intermetallics FeAl and NiAl. Surface preper- ation methods consisted of either a long time immersion in a strong acid bath (which was assumed to yield no artifacts other than gross topological changes due to accel- erated attack of high strain—energy regions), or straight mechanical polishing. Standard metallographic preparation (mechanical polishing) was adequate for sil- icon samples, although dislocations and/or other defect structures are introduced in the near-surface region by this method, presumably during the earlier stages (that is, grinding with fixed-media paper); see for example Figures 3.75, 3.76. Defects introduced by this method tend to be clustered (Figure 3.76), and the overall de- Figure 3.75. ECC image taken using the polepiece mounted BSE detector from me- chanically polished silicon. 25kV accelerating voltage, recursively filtered 32 frame average, 11mm working distance. White “speck” features are residual polishing me— dia. 15400x magnification. 1pm 93 Figure 3.76. ECC image taken using the polepiece mounted BSE detector from me- chanically polished silicon. 25kV accelerating voltage, single 25 second frame, 11mm working distance. White “speck” features are residual polishing media. 15400x mag- nification. — 1pm fect density is low; note that Figures 3.75, 3.76 were produced by first searching the sample surface for defects. Due to the clustering tendency and the near-surface nature of these defects, short time regime etching of the surface to be examined is believed to be sufficient to remove these polishing defects. Note, however, that this solution will also attack grain boundaries and other high strain energy areas of the surface. 3.2.3 Vacuum Quality The “cleanliness” of the vacuum was also found to have some bearing on the ability to obtain ECC from samples. This is due in part to the frequently seen effect of accumulation on the sample of species resulting from the cracking of residual gasses in the chamber by the electron beam. Of greater concern was contamination apparently deposited on the sample surface from the vacuum upon warming of a heavily loaded 94 cold trap. A cold trap for a vacuum system is essentially nothing more than a cold surface which condenses out species that can be captured from the vacuum. Water, residual diffusion and roughing pump oils, residual cleaning solvents, and assorted other species (frequently hydrocarbons) are commonly collected. Samples prepared in a fashion as to give an acceptable surface for ECCI were found to yield poor or no channeling contrast due to accumulated layers of deposited material on the sample surface if it had been in the vacuum chamber during the warm-up of a heavily loaded cold trap. The degree of surface degradation was in general a function of the length of time that the trap had been collecting residual gasses from the vacuum system. Additionally, beam-fixed contamination was found to have a significant effect on channeling contrast. Figure 3.7 (Section 3.1.1) shows an example of the degredation of attainable channeling contrast due to significant amounts of beam time accumu- lating contamination on the sample surface; the darker region to the lower right has had sufficient beam time for several micrographs to be taken, while the rest of the micrograph had very little beam time. Note that the loss of contrast in this case is somewhat unrepresentative due to both the ease of obtaining ECC of these misfit dislocation clusters, as well as the low accelerating voltage (which makes the ECC signal more sensitive to the surface layers). 3.3 Topics Examined Using ECCI 3.3.1 Dislocation Loops in FeAl Dislocation structures closely resembling Frank 100ps form readily in FeAl from quenched-in vacancies [22], taking the form of dislocations along the <100> directions with <100> Burgers vectors. Even for furnace cooled FeAl, these features form for 95 all but the slowest cooling rates [21, 22] due to large changes in equilibrium vacancy concentration. Figure 3.77 shows the typical dislocation structure in the an- nealed+electropolished FeAl; note the large numbers of approximately perpendicular dislocation segments along (100) directions, typical of annealed structures from rapidly-cooled FeAl [22]. Examination of the same region under a number of different channeling conditions (Figures 3.77-3.80) shows channeling contrast loss only for the case of gllt, as would be expected for the edge dislocations of which these structures are formed. The exact form of these dislocations cannot be determined explicitly from the data set available due to the strong resudual contrast from edge dislocations in FeAl [26]; it is possible that the dislocations examined have b close to parallal to the beam direction. If indeed this is the case, than this would be a demonstration of the sensitivity of this technique due to the reasons laid out below. Because of the edge nature of these loops, the only undistorted crystal planes in the region of these prismatic loops are those that simultaneously meet the g-b=0 and g-(bxt)=0 criteria (see, for example, [1], p.261). (This is equivalent to gllt, or only for those planes normal to the line direction.) What this means is that if the vacancy IOOps in Figures 3.78-3.80 are parallal to the surface, than the strain causing contrast for those dislocations is the strain perpendicular to the slip plane, which typically is of lower magnitude than that in the direction of b. 3.3.2 Dislocation Detection Depth Determination in Silicon Previous work [13,24] to measure the depth of detection of dislocations via ECCI in silicon samples has employed substantially different imaging conditions from the ones used for the current work. The detection depths found by Morin, et al (~1000Afor g=220) employed a high-tilt geometry, energy filtering, and an accelerating voltage of 45KeV. Those of Czernuszka, et al (173—210nm at 100KeV and 95mm at 30KeV, 96 Figure 3.77. Image of prismatic loops in FeAl using channeling contrast from the (110) channeling band. 14600x magnification. — 1pm Figure 3.78. Image of prismatic loops in FeAl showing contrast loss for [100] oriented segments; note segment (a). Channeling contrast from the (200) channeling band. 14600x magnification. 1pm 97 Figure 3.79. Image of prismatic loops in FeAl using channeling contrast from the (110) channeling band. 14600x magnification. — 1am Figure 3.80. Image of prismatic loops in FeAl showing contrast loss for [010] oriented segments; note segment (b). Channeling contrast from the (020) channeling band. 14600x magnification. 1am 98 both with g-——400) also used a high-tilt geometry, but no energy filtering. The differ- ences between these measurements and the present ones are considerable, but may be attributable in part to the use of the low-tilt geometry for the current measurements. Dislocations imaged using ECCI have the sharpest line contrast and narrowest image width near the surface of the silicon crystal (marked ”e” in Figures 3.81 and 3.82), with the image widening out at greater depth. Determination of the point at which visual contrast is lost is thus somewhat subjective, and is considered one principal source of error. An approximate depth of maximum detection for the arrowed dislocation in Figures 3.81 and 3.82 following the method laid out in the experimental procedure is calculated to be approximately 250nm. Figure 3.81. Dislocation image at +8° tilt relative to sample normal in Si. Note that by employing the (220) band for imaging contrast, the 220 g-vector becomes the tilt axis. Letter “e” marks the point of exit from the sample. —— 1pm Geometrical errors associated with these measurements are believed to be the principal sources of error, and consist of: (1) different tilt to either side of the surface 99 Figure 3.82. Dislocation imaged at -10.5° from sample normalin Si. Arrowed dislo- cation same as in Fig. 3.81. — 1pm normal, leading to differences in depth for the same point on the same dislocation line, as illustrated in Figure 3.83; and (2) errors in measurement of parallax due to different perceptions of the disappearance point of the dislocation image between mi- crographs (Figure 3.84). Differences in penetration depths to reach the same point are neglected because the sample surface is assumed to be smooth, which is a reasonable approximation for most ECCI sample surfaces at the scale of interest. The error introduced due to the differences in tilt relative to the surface normal can be seen by inspection of Figure 3.83 to be: error=d1-d2=c= ). 01 and 02 are the actual tilts to either side of the surface normal. For the stereo pair shown here, the actual tilts to either side of the surface normal were nominally 0p,g_ 3_31=8.0° and 9pm 3.82:10.5°, giving an error e/dz0.7%. The error due to measurement of parallax due to the different perception of the exact placement of the contrast loss point relative to the dislocation line between tilt conditions is shown schematically in Figure 3.84. For some difference along the 100 Figure 3.83. Side view demonstrating the difference in depth of penetration required to reach the same point due to differences in tilt to either side of the surface normal. (a) positive tilt; (b) negative tilt. dislocation line between the two micrographs of the perceived point of contrast loss (designated e in Figure 3.84), the error in the measured parallax can be seen to be esin0 for that image. Thus, for the stereo pair with the dislocation image lying approximately along 02, there is a substantially larger error than for the stereo pair with the dislocation image along 01 for any given e. It can be seen from this that this source of error may be minimized by choice of a dislocation line close to the projection of the rotation axis. An alternate method would be to employ only dislocation lines with features (kinks, jogs, or sharp turns) for these measurements to minimise e. This would, however, tend to give only a lower limit for the maximum detection depth, as the features used as markers would tend to be clearly discernible for this application. 101 (5 rotation axis A e, (9 [r .___.l parallax error Figure 3.84. TOp view demonstrating the error introduced to the measured parallax between micrographs due to a difference in perception of the contrast loss point along a dislocation line between micrographs. The error introduced in the parallax measurement per error in length along the dislocation line increases proportionately to the sine of the projected angle between the rotation axis and the dislocation line, 0. Thus, condition 2 produces a larger parallax error per error e along the length of the dislocation line than does condition 1. 102 3.3.3 Effect of Deviation From Bragg Condition on Defect Image Characteristics In order to ease the interpertation of ECCI micrographs for future investigations, any systemic characteristics of defect images from ECC are presented here. For the range of tilt angles examined, where the surface normal did not vary from the microscope axis by more than ~18°, there was found to be no Obvious systematic variation in the contrast observed as a function of the sample tilt. In other words, except for effects such as foreshortening, the appearance of a dislocation image does not change significantly with changes in surface tilt; instead the dominant parameters are. the channeling band used for contrast (g), and the tilt angle away from the perfect Bragg condition (3). Section 3.1.2 demonstrates this image dependance on tilt in some detail, while Figure 2.7 shows the definitions of g and s in terms of SACPs. s is defined as positive if the incident beam intersects the crystal at an angle greater than 03, and negative for an intersection at less than 03. Examination of any arbitrary ECP (such as Figure 1.5,for example), reveals the high density of channeling band edges, each of which represents an available con- trast condition. This high density of available contrast bands essentially precludes obtaining contrast from only one band, requiring that multiple contrast sources be considered for each image. It is worth noting that as the BSE signal is derived from electrons with trajectories varying within the range of the probe convergence angle 20, reduction of 20 will reduce the difficulty of achieving a single contrast condi- tion. Note that the implication of this is that identification of Burgers vectors by image extinction is more difficult than in the TEM diffraction contrast case, due to the difficulty in obtaining only a single contrast condition. This can be viewed as an advantage, however, if the use for which channeling contrast is being employed is solely to image dislocations in contrast for viewing purposes, as might occur for some 103 in-situ testing. In agreement with previous reports [13,17], for the channeling conditions that pro- duced dark-bright dislocation images, the reversal of the sign of g generally reversed the sense of the contrast, with dark areas becoming bright, and vice-versa. Figures 3.85, 3.86 demonstrate this phenomenon. Exceptions to this general rule may be explained by overlying contrast from other channeling bands. Figure 3.85. Individual and clusters of dislocations in Si displaying bright-dark con— trast reversal for opposite g; compare to Fig. 3.86. g=113, s<0. — 1pm For the channeling conditions that do not produce the paired bright-dark images, the predominant sense of the channeling contrast was found to depend strongly on the sign of the deviation parameter s, which describes the angle of incidence of the scanning beam relative to the exact Bragg angle, 03. s is defined as positive if the incident beam intersects the crystal at an angle greater than 03, and negative for an intersection at less than 03. The dependence of the contrast on 3 may be explained in the following manner: for beam incidence angles 21/ close to normal incidence, a higher 104 Figure 3.86. Dislocations in Si displaying bright-dark contrast reversal for opposite g; compare to Fig. 3.85. g=11-3, s<0. 1pm ,u is observed at angles less than 03, where s<0. As a consequence of this, the bulk crystal (the background) will have a higher )1 than at least part of the area around the ECC feature, which will have a a characteristic of the crystal with s>0 due to local plane bending, giving dark contrast to the dislocation. The converse argument would therefore apply for incidence angles >03, giving bright dislocation contrast. An example of this is shown by Figures 3.87 and 3.88. As in the previous case, this rule obviously breaks down for a multiple channeling condition with multiple sources of contrast. 105 Figure 3.87. Dislocation images displaying opposite senses of dislocation contrast for positive and negative values of s. g=022, s>0. 1am Figure 3.88. Dislocation images displaying opposite senses of dislocation contrast for positive and negative values of s. g=022, s<0. _ 1pm CHAPTER 4 Conclusions This work has established a unique electron microscopic facility capable of performing ECCI at low sample tilt angles. Studies of competing surface preparation techniques indicate that mechanical polishing produces very deep damage layers in some mate- rials commonly thought of as hard, such as FeAl and NiAl. Specifically: 1. ECCI of dislocations and other near-surface crystallographic features can be performed using the equipment set up for that purpose (CamScan 44FE with accessories) at Michigan State University. 2. ECCI of crystallographic features can be performed at zero and low-angle tilts of up to at least ~30° of the surface normal from the microscope axis employ- ing either a retractable side-mount BSE detector or a polepiece-mounted BSE detector. 3. ECCI may be performed using accelerating voltages at least as low as 7kV with the present setup for some larger scale ECC features. 4. ECCI sample surface conditions effect ECC image quality and contrast. Samples that have oxide or other coatings have their crystallinity obscured by a foreign layer, reducing the coherency of the electron beam and thus reducing the ECC 106 107 obtainable from any given feature. Electron beam-fixed sample contamination falls into this category. Relatively thin foreign surface layers can completely obscure the underlying crystallinity. . Sample preparation can and often does produce crystal defect artifacts due to mechanical polishing and similar surface preparation methods; non-disruptive surface preparation methods are Often called for as a final step in sample prepa- ration. In particular, for the B2 intermetallic compounds FeAl and NiAl, it was found that cutting with a wire EDM produced a shallower damage layer than conventional metallographic preparation. Electropolishing was necessary for both materials to reach undamaged layers, but the damage left by mechani- cal polishing went much deeper into the crystal. In contrast, careful mechanical polishing of semiconductor-grade, dislocation-free silicon produced a substan- tially lower, although still significant, surface defect density. . Conditions suitable for ECCI of individual dislocations using the CamScan44FE in the Department of Materials Science and Mechanics at Michigan State University as currently configured may be summarized as follows: using polepiece backscatter detector: working distance: ~9mm-~15mm probe current(ip): >~1.5nA surface normal tilt from microscope axis: ~30° or less using side mounted backscatter detector: working distance: possibly as short as~4mm-~16mm or more probe current(z'p): >~1.5nA surface normal tilt from microscope axis: potentially quite large, but limited to where detector can achieve a large solid angle of collection on the surface being examined. Includes the zero tilt condition. CHAPTER 5 References [1] P. Hirsch, A. Howie, R. Nicholson, D.W. Pashley, M.J. Whelan: Electron Microscopy of Thin Crystals, GENERAL REFERANCE, Robert E. Krieger Publishing Co., Malabar, FL (1965,1977) [2] DB. Williams, C.B. Carter: Transmission Electron Microscopy: a Textbook for Materials Science, GENERAL REFERANCE, Plenum Press, New York (1996) [3] D. Hull, D.J. Bacon: Introduction to Dislocations, 3rd ed., pp. 88-90, Pergamon Press, Elmsford, New York (1984) [4] DE. Newbury, D.C. Joy, P. Echlin, C. F iori, J.I. Goldstein: Advanced Scanning Electron Microscopy and X-Ray Microanalysis, pp.87-145, Plenum Press, New York (1986) [5] DC. Coates: “Kikuchi-like Reflection Patterns obtained with the Scanning Elec- tron Microscope”, Philosophical Magazine, 16, p.1179 (1967) [6] GR. Booker, A.M.B. Shaw, M.J. Whelan, P.B. Hirsch: “Some Comments on the Interpertation of the ’Kikuchi-like Reflection Patterns’ observed by Scanning Electron Microscopy”, Philosophical Magazine, 16, p.1185 (1967) 108 109 [7] DR. Clarke, A. Howie: “Calculations of Lattice Defect Images for Scanning Electron Microscopy”, Philosophical Magazine, 24, p.959 (1971) [8] J .P. Spencer, C.J. Humphreys, P.B. Hirsch: “A Dynamical Theory for the Con- trast of Perfect and Imperfect Crystals in the Scanning Electron Microscope Using Backscattered Electrons”, Philosophical Magazine, 26, p.193 (1972) [9] R. SandstrOm, J.F. Spencer, C.J. Humphreys: “A Theoretical Model for the Energy Dependence of Electron Channelling Patterns in Scanning Electron Mi- croscopy”, Journal of Physics D: Applied Physics, 7, p.1030 (1974) [10] DR. Clarke:“Observation of Crystal Defects using the Scanning Electron Mi- croscope”, Philisophical Magazine, vol.24, p. 973 (1971) [11] DC. Joy, M.N. Thompson, G.R. Booker, W.H.J. Anderson:“Scanning Reflection Electron Micrographs of Stacking Faults in a C0-4wt% Ti Alloy”, Physica Status Solidi (a), vol.21, p. K1 (1974) [12] RM. Stern, T. Ichinokawa, S. Takashima, H. Hashimoto, S. Kimoto: “Dislocation Images in the High Resolution Scanning Electron Microscope”, Philisophical Magazine, vol.26, p. 1495 (1972) [13] P. Morin, M. Pitaval, D. Besnard, G. F ontaine: “Electron-channelling imaging in Scanning Electron Microsc0py”, Philosophical Magazine A, 40, 4, p.511 (1979) [14] DC. Joy: “Direct Defect Imaging in the High Resolution SEM”, Mat. Res. Soc. Symp. Proc., 183, p.199 (1990) [15] J .T. Czernuszka, N .J. Long, E.D. Boyes, P.B. Hirsch: “Imaging of Dislocations using Backscattered Electrons in a Scanning Electron Microscope”, Philosoph- ical Magazine Letters, 62, p.227 (1990) 110 [16] J.I. Goldstein, DE. Newbury, P. Echlin, D.C. Joy, A.D. Romig, Jr., C.E. Lyman, C. Fiori: Scanning Electron Microsc0py and X-Ray Microanalysis, 2nd ed., pp.107-116,176-181, Plenum Press, New York (1992) [17] A.J. Wilkinson, G.R. Anstis, J.T. Czernuszka, N.J. Long, P.B. Hirsch: “Elec- tron Channelling Contrast Imaging of Interfacial Defects in Strained Silicon- Germanium Layers on Silicon”, Philosophical Magazine A, 68, 1, p.59 (1993) [18] T. Ichinokawa, M. Nishimura, H. Wada: “Contrast Reversals of Psudo—Kikuchi Band and Lines Due to Detector Position in Scanning Electron Microscopy”, Journal of The Physical Society of Japan, 36, l, p.221 (1974) [19] T. Yamamoto: “Contrast Mechanisms in Electron-Channelling Patterns and Lattice-Defect Images Obtained with a Scanning Electron Microscope”, Philo- sophical Magazine A, 38, 4, p.439 (1978) [20] A.J. Wilkinson, P.B. Hirsch: “The Effects of Surface Stress Relaxation on Elec- tron Channelling Contrast Images of Dislocations”, Philosophical Magazine A, 72, 1, p.81 (1995) [21] P. Nagpal, 1. Baker: “Effect of Cooling Rate on Hardness of FeAl and NiAl”, Metallurgical Transactions A, 21A, p. 2281 (1990) [22] A. Fourdeux, P. Lesbats:“Annealing out of quenched-in vacancies in an ordered B2 type Fe-Al single crystal”, Philosophical Magazine A, 45,1, p. 81 (1982) [23] P. Hirsch, A. Howie, R. Nicholson, D.W. Pashley, M.J. Whelan: Electron Microscopy of Thin Crystals, pp. 50—56, Robert E. Krieger Publish- ing Co., Malabar, FL (1965,1977) 111 [24] J.T. Czernuszka, N.J. Long, E.D. Boyes, P.B. Hirsch:“Electron Channelling Con- trast Imaging (ECCI) of Dislocations in Bulk Specimens”, Mat. Res. Soc. Symp. Proceedings, vol.209, p. 289 (1991) [25] J .T. Czernuszka, N .J . Long, P.B. Hirsch:“Electron Channelling Contrast Imaging of Dislocations”, Proceedings of the XIIth International Congress for Electron Microscopy, p. 410, San Francisco Press (1990) [26] MA. Crimp: Doctoral Dissertation "‘will]lint