METHOD FOR THE DETECTION AND CORRECTION OF LENS DISTORTIONS IN AN ELECTRON DIFFRACTION SYSTEM

Abstract

A method for correcting distortion in a coherent electron diffraction imaging (CEDI) image induced by a projection lens makes use of a known secondary material that is imaged together with a sample of interest. Reflections generated from the secondary material are located in the image, and these observed reflections are used to approximate a beam center location. Using a known lattice structure of the secondary material, Friedel pairs are located in the image and unit cell vectors are identified. Predicted positions for each of the secondary material reflections are then determined, and the position differences between the observed reflections and the predicted reflections are used to construct a relocation function applicable to the overall image. The relocation function is then used to adjust the position of image components so as to correct for the distortion.

Claims

1. A method of correcting a distortion imparted to a diffraction image of a sample of interest by a projection lens in a coherent electron diffraction imaging (CEDI) system, the method comprising: a) obtaining a CEDI image of the sample of interest together with a secondary material having a known lattice structure; b) locating within the image the observed position of reflections corresponding to the secondary material; c) identifying unit cell vectors for the secondary material; d) determining a predicted position for each secondary material reflection corresponding to a respective Bravais-Miller index assigned thereto based on the unit cell vectors and the known lattice structure of the secondary material; and e) using differences between observed positions and predicted positions of secondary material reflections to relocate image components within the CEDI image.

2. A method according to claim 1 wherein the secondary material comprises graphene.

3. A method according to claim 1 wherein identifying unit cell vectors comprises finding difference vectors for pairs of the CEDI reflections corresponding to the secondary material and grouping the difference vectors according to length and direction, averaging the vectors of each group and selecting the shortest averaged group vectors having a similar length as the unit cell vectors.

4. A method according to claim 1 wherein determining a predicted position for each secondary material reflection comprises finding a linear combination of the unit cell vectors that corresponds to a position in the image proximate to the observed position of that reflection.

5. A method according to claim 1 wherein said differences between observed positions and predicted positions of secondary material reflections are found relative to a beam center location in the image.

6. A method according to claim 5 wherein said beam center location is found by averaging centroids of Friedel pairs using the observed positions of the secondary material reflections.

7. A method according to claim 5 wherein, for each secondary material reflection, a ratio is found between a distance of an observed reflection to the beam center location and a distance of a predicted reflection to the beam center location, and said ratios are used to characterize the distortion within the image relative to the beam center location.

8. A method according to claim 7 wherein said ratios and observed secondary material reflection positions are used to determine coefficients a, b, c, d, c.sub.x and c.sub.y that best satisfy the equation: $\begin{array}{cc}f\left(x,y\right)=a+b*\sqrt{{\left(x-c_x\right)}^2+{\left(y-c_y\right)}^2}+c*\sin \left(2*{\tan }^{-1}\left(\frac{y-c_y}{x-c_x}\right)+d\right)& \left(4\right)\end{array}$ for each of said ratios and observed secondary material reflection positions, where c.sub.x and c.sub.y represent the x and y coordinates of the beam center location, said equation with said determined coefficients being used in said relocating of image components.

9. A method according to claim 8, wherein a least squares refinement is used to fit said equation to said ratios and observed secondary material reflection positions.

10. A method according to claim 8 wherein revised x and y values for a pixel location in the image are determined as follows:
x.sub.rev=(x−c.sub.x)/f(x, y)+c.sub.x  (5)
y.sub.rev=(y−c.sub.y)/f(x, y)+c.sub.y  (6) where x.sub.rev and y.sub.rev are, respectively, the revised x and y pixel values, x and y are, respectively, the original pixel values, and f(x,y) is a calculated version of Equation (4) using the determined coefficients.

11. A method according to claim 1 wherein the CEDI image is a first CEDI image, and wherein the method further comprises obtaining a plurality of CEDI images of the sample of interest together with the secondary material at a plurality of incidence angles, reflections from each of the images being used in the method after being corrected for position variations due to angle of incidence.

12. A method according to claim 1 wherein the secondary material comprises a plurality of layers each of which generates separate reflections that are used in the method.

13. A method according to claim 1 wherein the secondary material comprises a substrate on which the sample is located.

14. A method according to claim 1 wherein the secondary material is part of the sample.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0014] FIG. 1A shows a simulated CEDI image of a double-walled carbon nanotube (DWCNT) along the tube axis.

[0015] FIG. 1B depicts a successful reconstruction of the DWCNT of FIG. 1A from the simulated data.

[0016] FIG. 1C depicts a first failed reconstruction of the DWCNT of FIG. 1A using the simulated data after applying a projection lens distortion from experimental carbon nanotube (CNT) data.

[0017] FIG. 1D depicts a second failed reconstruction of the DWCNT of FIG. 1A using the simulated data after applying a projection lens distortion from experimental CNT data.

[0018] FIG. 1E depicts a third failed reconstruction of the DWCNT of FIG. 1A using the simulated data after applying a projection lens distortion from experimental CNT data.

[0019] FIG. 1F depicts a fourth failed reconstruction of the DWCNT of FIG. 1A using the simulated data after applying a projection lens distortion from experimental CNT data.

[0020] FIG. 2A is a schematic representation of an undistorted image of an object according to the prior art.

[0021] FIG. 2B is a schematic representation of the object shown in FIG. 2A that has been subjected to a pincushion-type distortion.

[0022] FIG. 2C is a schematic representation of the object shown in FIG. 2A that has been subjected to a barrel-type distortion.

[0023] FIG. 2D is a schematic representation of the object shown in FIG. 2A that has been subjected to a spiral distortion.

[0024] FIG. 2E is a schematic representation of the object shown in FIG. 2A that has been subjected to shows an elliptical distortion.

[0025] FIG. 3 is a schematic representation of a hexagonal lattice of graphene, with a rhombic unit cell ABCD.

[0026] FIG. 4 is an example CEDI image of graphene.

[0027] FIG. 5A is a perspective view of flat graphene in real space.

[0028] FIG. 5B is a perspective view of corrugated graphene in real space.

[0029] FIG. 5C is a schematic depiction of the reciprocal space for a flat graphene sheet.

[0030] FIG. 5D is a schematic depiction of the reciprocal space for a corrugated graphene sheet.

[0031] FIG. 5E is a schematic depiction indicating how, for a corrugated graphene sheet, the diffraction spots become blurred at large angles.

[0032] FIG. 6A is a first section of a flow diagram of a method according to the invention.

[0033] FIG. 6B is the second section of the flow diagram of FIG. 6A.

[0034] FIG. 7A depicts a CEDI image of graphene like that of FIG. 4, but showing the material lattice structure.

[0035] FIG. 7B is a table of the measured reflection data for the image of FIG. 7A.

[0036] FIG. 7C is a table of the predicted reflections for the image of FIG. 7A.

[0037] FIG. 7D is a table of the distance ratios for the reflections in the image of FIG. 7A.

[0038] FIG. 7E shows the coefficient fitting for a distortion characterization equation associated with the image of FIG. 7A.

[0039] FIG. 7F is a table showing the fitted distance ratios for the reflections in the image of FIG. 7A.

[0040] FIG. 8 is the CEDI image of FIG. 7A on which is overlayed arrows indicative of the direction and distance that each of the reflections shown in the image would need to be displaced to be at its correct location in the absence of the lens distortion.

[0041] FIG. 9 is a CEDI image of a gold nanocrystal supported by two layers of graphene.

[0042] FIG. 10 is a CEDI image of a freestanding carbon nanotube.

DETAILED DESCRIPTION

[0043] The present invention makes use of Bragg diffraction peaks from a secondary material such as graphene for identifying and correcting for CEDI lens distortion errors. In many CEDI experiments, graphene is used as a support for the specimen under test, as it is highly translucent to low energy electrons. Graphene is an allotrope of carbon in the form of a single layer of atoms in a 2D hexagonal lattice, in which one atom forms each vertex. An example of the hexagonal lattice of graphene is shown in FIG. 3, with the rhombic unit cell ABCD and four neighboring cells being denoted by the numbers 1, 2, 3 and 4 in circles. Graphene is also the basic structural element of other allotropes, including carbon nanotubes. The rhombic unit cell of graphene contains two chemically identical carbon atoms, labeled P1 and P2 in FIG. 3, and has unit cell dimensions b1=b2=2.46 Å. Graphene has lattice symmetry p6m. An example CEDI image of graphene is shown in FIG. 4.

[0044] From the hexagonal symmetry, the relative distances of the reflections from the beam center can be derived from their Bravais-Miller indices, as shown below in Equation (3). The ratio between the observed distances and the distances predicted from the symmetry provide input for determining the distortion parameters.

d.sub.rel,hexagonal(h, k)=√{square root over (h.sup.2+k.sup.2+h*k)}  (3)

As a 2D crystal, the Bragg reflections of graphene in reciprocal space, instead of spheroids as for 3D crystals, are cones (rods for idealized flat graphene). Consequently, the Bragg condition is fulfilled for all reflections at any incidence angle of the electron beam on the graphene, obviating the need to carefully align the graphene with the beam, as is the case with 3D crystals.

[0045] For 3D reconstructions, where the incidence angle changes continuously during the tomographic scan, the graphene reflections move away from the scan axis with increasing acuteness of the angle of incidence. This spreads out the reflections on the detector surface and allows for the sampling of different regions of the distortion function, which in turn allows for a more precise determination of the distortion parameters. This is demonstrated in FIGS. 5A-5E.

[0046] FIGS. 5A-5E show the characteristics of flat and corrugated graphene sheets as described in J. C. Meyer et al., The Structure of Suspended Graphene Sheets, Nature 446 (2007) 60-63. FIG. 5A is a perspective view of flat graphene in real space, while FIG. 5B is a similar view of corrugated graphene. FIG. 5C shows the reciprocal space for a flat graphene sheet which, as indicated, is a set of rods oriented perpendicularly to the hexagonal reciprocal lattice of graphene. FIG. 5D shows how, for a corrugated graphene sheet, a superposition of diffracting beams from microscopic flat areas effectively turns the perpendicular rods of FIG. 5C into cone-shaped volumes. As a result, the diffraction spots from such a surface become blurred at large angles, as indicated by the dotted lines in FIG. 5E. Moreover, this effect is more pronounced further away from the tilt axis.

[0047] The present invention uses the measured CEDI images of graphene as part of a method for identifying and compensating for lens distortions such as those described above. An exemplary embodiment of this method is outlined in the steps shown in FIGS. 6A and 6B. This embodiment may be applied, for example, as part of a CEDI measurement for which one or more samples of interest are mounted on a graphene support surface.

[0048] The method begins with the acquisition of a CEDI image in step 600. In step 602, the Bragg reflection positions of the graphene are located in the CEDI image. This may be done manually by a user visually identifying the reflections, and noting their pixel positions in the overall image. Alternatively, an automated procedure, such as a blob-detection method (e.g., connected-component labeling) may be applied. In general, a manual approach works best for images that also contain diffraction data from the specimen, or if multiple graphene domains are present.

[0049] In step 604, a correction for the angle of incidence tilt of the electron beam with respect to the graphene sheet is applied. This may be done by modifying the image so that the reflection positions are closer by a factor of 1/cosφ to the projection of the scan axis onto the image, with φ=0° for perpendicular incidence. Although the invention may make use of only one image, it is also possible to use a set of images taken using different angles of incidence. As discussed above with regards to FIGS. 5A-5E, the different angles of incidence allow for sampling of different regions of the distortion function, and can improve the determination of the distortion parameters. Thus, the method shown in FIGS. 6A and 6B includes the possibility of repeating Steps 602 and 604 for each of the images that might be used, as indicated by decision branch (step 606) that accounts for all of the different images that may be desired.

[0050] Once the graphene reflection positions for each image are located and corrected as described above, the difference vectors for all reflection pairs of each image are calculated (step 608). The difference vectors are grouped according to their length and direction using threshold criteria specified for limiting the maximum fractional deviation of vector lengths for grouping and the maximum angular deviation from collinearity for vectors within a group. An average difference vector for each group is then calculated.

[0051] In step 610, the three shortest group difference vectors are found, and those vectors will form the hexagonal lattice. The vectors should be similar in length, and oriented at angles of 60° relative to one another. The two vectors that are most closely similar in length are selected, and those vectors will be the unit cell vectors in reciprocal space. The beam center position (c.sub.x, c.sub.y) is then estimated by calculating the center of gravity of all of the reflection positions (step 612).

[0052] A Bravais-Miller index is assigned to each reflection (step 614) by finding the best linear combination of the two unit cell vectors from the beam center to the observed reflection position. In the preferred embodiment of the invention, the Ewald sphere curvature is neglected and approximated as a plane. This is a valid approximation given that the de Broglie wavelength of the electrons is very small, and the diffraction angles of the observed reflections are very low. An improved beam center position is then calculated by averaging over the centroids of all Friedel pairs of the observed reflections (step 616).

[0053] For each reflection, the ratio of the observed and calculated distance to the beam center is then determined (step 618). Using Equation (3) above, the distance between the observed position of each reflection and the beam center is determined, as is the distance between the calculated position of each reflection and the beam center. For each reflection, the ratio between these two values is then found and recorded.

[0054] In step 620, Equation (4) shown below is fit to the calculated ratios and observed reflection pixel positions (x,y) to determine a set of coefficients a, b, c, d, c.sub.x and c.sub.y.

[00003]$\begin{array}{cc}\begin{array}{c}f\left(x,y\right)=f_{\mathrm{radial}}\left(x,y\right)+f_{\mathrm{ellipse}}\left(x,y\right)\\ =a+b*\sqrt{{\left(x-c_x\right)}^2+{\left(y-c_y\right)}^2}+c*\sin \\ \left(2*{\tan }^{-1}\left(\frac{y-c_y}{x-c_x}\right)+d\right)\end{array}& \left(4\right)\end{array}$

The coefficients define the type and magnitude of the distortion indicated by the errors in the distances of the reflections from the beam center. The values c.sub.x and c.sub.y represent the coordinates of the beam center, and the values a, b, c, d provide different respective information regarding the nature of the distortion itself. These are defined herein as follows: 1) “a” represents an average magnification factor; 2) “b” represents a radial distortion coefficient; 3) “c” represents an elliptical distortion coefficient; and 4) “d” represents an ellipse inclination angle. Thus, by finding the magnitudes of these coefficients, an indication of the type of distortion being created by the projection lens may be determined. Moreover, knowing these values allows for a correction to be applied to the image data of a sample that has been shown to significantly improve the data quality.

[0055] The fitting process using Equation (4) may be according to a known fitting technique and, in the present embodiment, is a least squares refinement. The values for c.sub.x and c.sub.y found in step 616 are used, and starting values for the coefficients a, b, c and d are chosen empirically. In this embodiment, starting values of a=1, b=0 and d=0 are chosen. The starting value of c is the largest distance ratio found in step 618. Those skilled in the art will understand that, although values for the beam center coordinates c.sub.x, c.sub.y were obtained previously, these values are refined together with the coefficients a, b, c, d during the fitting process, which improves the overall result.

[0056] Once the coefficients are found, the equation may be used as a general tool for correcting the distortion in a CEDI image, affecting both image components related to the secondary material, e.g., graphene, and image components related to a sample under test. In this context, Equation (4) may be used as a transform function to restructure the image on a pixel-by-pixel basis. In applying the equation, a, b, c, d, c.sub.x and c.sub.y are replaced with the actual values determined during the fitting process, and revised x and y values can be calculated for any pixel location as follows:

x.sub.rev=(x−c.sub.x)/f(x, y)+c.sub.x  (5)

y.sub.rev=(y−c.sub.y)/f(x, y)+c.sub.y  (6)

where (x.sub.rev, y.sub.rev) is the revised pixel position, (x, y) is the original pixel position, and the distortion factor f(x,y) is the evaluation of Equation (4) at the original pixel position using the coefficients determined in the fitting process.

[0057] FIG. 7A shows a CEDI image of graphene, similar to that of FIG. 4, but without the presence of a beam stop. The location of many or all of the different reflections shown in the figure may have a position error due to a projection lens distortion. The image was obtained using a perpendicular angle of incidence, and no correction for angle of incidence is therefore necessary. From the measured distance vectors of all pairs of reflections in the image, the shortest are found to be associated with the two unit cell vectors that define the hexagonal lattice and are also among the most frequently occurring. The unit cell vectors will be at an approximately 60° angle relative to each other. Of the reflections shown connected in the figure by dashed lines, the distance vectors from the beam center to any two neighboring reflections form a valid set of unit cell vectors.

[0058] The measured reflection data for the image shown in FIG. 7A is presented in the table of FIG. 7B. The table shows a list of all of the measured difference vectors and, at the bottom, the length (in pixels) of the shortest three vectors is indicated, along with their relative angles. As shown, the three vectors are similar in length, and have angles relative to each other of approximately 60° . Two of these have a closely similar length (283.0 and 283.3 pixels, respectively), and these will represent the unit vectors for the graphene material. Also shown in the table are the coordinates of the beam center (1030, 1020), determined by finding the center of gravity of all observed reflection positions.

[0059] Once the unit vectors are determined, the Bravais-Miller indices may be assigned to the observed reflections, as shown in the table of FIG. 7C. An improved value for the beam center is found by averaging over the centroids of all Friedel pairs of observed reflections found from the indexing step. For each reflection, a predicted position from the updated value for the beam center can then be found using Equation (3), which relies on the symmetry properties of the graphene crystal to make an accurate calculation that is less affected by distortion than the observed reflection positions. The ratio between the calculated distances of observed reflections to the beam center and the distances of predicted reflections to the beam center is then determined, as shown in the table of FIG. 7D.

[0060] The fitting of Equation (4) to the calculated ratios from FIG. 7D and the observed pixel positions (x,y) is shown in FIG. 7E. As shown, the resulting values for the coefficients are a=1.167, b=−0.00009, c=−0.017 and d=−0.36. The resulting values for the beam center coefficients are c.sub.x=1021 and c.sub.y=1025. With these values, fitted ratios can then be determined for each of the reflections, as shown in FIG. 7F. Thus, FIG. 7D shows the observed ratios and FIG. 7F shows the fitted ratios. The Fit column in FIG. 7F shows the observed ratios divided by the fitted ratios and indicates the quality of the fit. These fitted ratios indicate for each reflection the magnitude of the displacement of that reflection away from or towards the beam center due to lens distortion. This is shown visually in the version of the FIG. 7A CEDI image shown in FIG. 8, on which are overlayed arrows indicative of the direction and distance that each of the reflections shown in the image would need to be displaced to be at its correct location in the absence of the lens distortion. That is, the corrected positions of the graphene reflections are indicated by circles, with the arrows pointing from what was their distorted (observed) positions. As coefficient b is close to zero, it can be seen that this particular example represents primarily an elliptical distortion.

[0061] Once the distortion correction is known, it may be applied to the entire image, and therefore provides correction of image components related to other materials examined together with the graphene, such as a sample of interest mounted on a graphene substrate. Although both the sample and the secondary material are typically imaged together, it may also be possible to collect an image of just the secondary material by moving the electron beam to a region in which no sample material is present, and then use this image to calibrate the distortion function to be used for an image collected of the sample and secondary material together. Once a correction for the lens distortion is obtained, it may be applied to an unknown material based on the characteristics of the distortion determined by the examination of the secondary material, in this case graphene.

[0062] Although graphene was the material chosen for the present embodiment, those skilled in the art will understand that other materials with known characteristics may also be used for the characterization of a lens distortion when they are examined together with a sample of interest. One alternative material, for example, is silicon nitride. Although it is a three-dimensional crystal, and not two-dimensional like graphene, a similar process of finding predicted reflection positions and fitting Equation (4) to determine the coefficients indicative of the nature of the distortion may be used. However, such a 3D crystal would need to be aligned with the symmetry axis parallel to the electron beam.

[0063] In another embodiment of the invention, more than one layer of secondary material may be used. Shown in FIG. 9 is a CEDI image of a gold nanocrystal supported by two layers of graphene. As shown, this creates two sets of graphene reflections, each associated with a different graphene layer. The reflections of these two graphene domains are distinguished in the figure by circles and squares, respectively, adjacent to which are shown their Bravais-Miller indices.

[0064] When using multiple layers of secondary material, as in FIG. 9, the method according to the invention proceeds in the same manner, with the additional layer providing an additional set of reflections for characterizing the distortion imparted by the projection lens. Those skilled in the art will recognize that the two layers must be evaluated separately with regard to establishing the predicted positions of the reflections, but both sets of reflections may be used for finding the beam center and fitting Equation (4) for finding the coefficients a, b, c, d, c.sub.x and c.sub.y.

[0065] In another embodiment, characterization of a lens distortion may be done using reflections of a known material that may be intrinsic to a sample of interest. FIG. 10 shows a CEDI image of a freestanding carbon nanotube, which will inherently produce reflections corresponding to the graphene contained therein. In the figure, the graphene reflections are indicated by circles, adjacent to which are their respective Bravais-Miller indices. As in the foregoing embodiments, the predicted reflection positions can be used to determine the displacement of the reflections in the image due to lens distortion, and the fitting of Equation (4) provides the coefficients for characterizing the distortion. Correction of the overall image can then be performed accordingly.