Dealing with image shifting in 3D ToF-SIMS depth profiles

Acquiring a traditional dual beam depth profile involves sequential analysis and sputter cycles of a surface of interest. Each sputter cycle removes a layer of material, while each analysis cycle generates an image of the exposed surface. By stacking these images, one can create a 3D volume representation of the analyzed area. Conceptually, this process is simple; however, in practice, there are many factors that make reconstruction of accurate time-of-flight secondary ion mass spectrometry (ToF-SIMS) three-dimensional (3D) depth profiles challenging. These issues include surface topography, changes or differences in sputter rates of materials, matrix effects, surface damage, and shifting of images due to changes in the sample height from material removal.

Studies have reported on methods to understand and compensate for many of the issues that have made accurate representation of ToF-SIMS depth profiles difficult. Seah reported on a theoretical framework and analysis of the effects of topography on sputtering yields for monatomic primary ions.¹ Work has also been done using ToF-SIMS in combination with AFM data to accurately reconstruct 3D depth profiles while properly accounting for sample topography.^2,3 In studies of multilayer polymer films, it has been shown that the sputter rate of individual components can change when they are in a layered system, and that the layering order can affect the sputter rate of each material.^4,5 The presence of ion yield suppression and enhancement has been reported for a binary system of mixtures of poly(lactide) and codeine. It was found that suppression of the yield of poly(lactide) ions was related to the enhancement of codeine ions.⁶ Taylor et al. reported on a method to correct 2D ToF-SIMS depth profiles by accounting for the sputter rate of each material in a bilayer system.⁷ Several studies have looked at the effect of damage induced by the analysis or sputter source on the resulting depth profiles.^8–11 The influence of matrix effects has been shown for depth profiling cells¹² and for measuring compositions in organic multilayers.¹³ In the study of the organic multilayers, Shard et al. demonstrated how one could correctly quantify the composition of an organic multilayer using ToF-SIMS by accounting for matrix effects throughout the depth profile. These studies have provided key information for acquiring and displaying ToF-SIMS depth profiles. However, these studies were limited to depths of a few microns or less.

ToF-SIMS depth profiles of greater than 1 μm in depth have been made achievable for a wide range of organic materials with the advent of gas cluster ion beams (GCIB) due to their low damage accumulation and efficient sputtering.^14,15 With these new ion sources, thick coatings and buried interfaces that were previously inaccessible without cross-sectioning can now be analyzed via depth profiling. However, with increasing depth of analysis come new challenges. If one profiles deeper (>1 μm) into a sample, the image shifts can become significant. Since the sputter and analysis beams are typically located in fixed positions, and at fixed angles to the sample surface (often 45°), the actual position sampled by the imaging beam shifts as material is removed. For an analysis beam oriented at 45° to the sample surface normal, each micron of material removed will shift the image laterally by 1 μm. For shallower depth profiles (tens to hundreds of nanometers), this shift is typically insignificant and can generally be ignored. However, for deeper depth profiles (e.g., 20 μm into the surface), the imaged position would shift laterally by 20 μm, significantly distorting a reconstructed 3D image. There are few reported ToF-SIMS studies that have profiled to depths where shifting could be an issue. Bailey et al. reported on depth profiling a multilayer polymer system to depths up to 15 μm; however, since the layers were uniform no lateral correction was required to visualize the data.¹⁶ Fletcher et al. reported on depth profiling up to approximately 175 μm into a Xenopus laevis oocyte, but did not comment on lateral shifting of the images as it was not the main focus of the paper.¹⁷ Tyler et al. mention aligning layers of a 3D depth profile to account for image shifting and to improve the resolution at the material interface.¹⁸ Vanbellingen et al. reported on using automated image shifting to correct for mechanical drift in 2D images using tools built into the iontof software.¹⁹ However, to our knowledge, no one has reported on methods to correct for lateral shifting in ToF-SIMS depth profiles with depths above a few microns.

Depth profiling with ToF-SIMS to depths greater than 1 μm into a sample produces three key problems: (1) the reconstructed 3D image will be skewed if no corrections are made, (2) the resulting 3D image will have to be significantly cropped if post-acquisition shifting is used, or a much larger area would need to be analyzed to compensate for the lateral shifting during data collection, and (3) the sputter and analysis areas could diverge since the two beams are typically oriented 180° from each other, meaning the profiled region and the analysis region would be moving away from each other as the depth of the sputter crater increases.

One potential solution to more accurately reconstruct ToF-SIM 3D depth profile images would be to image a much larger area than the size of the features of interest and then shift the images to align the features after acquiring the data. The limitation of this solution is that one would have to significantly increase the imaging area to accommodate for large image shifts over depths greater than a few microns which would greatly increase the analysis time and the size of the resulting data. Furthermore, one would also have to significantly increase the sputtered area in order to avoid artifacts in the profiles from sampling the crater edges.

In this paper, we explore three solutions to the issue of image shifting in depth profiles that do not require increased data acquisition time or sputtered area. The first involves using software to shift and crop the images. The second involves adjusting the stage height during each sputter cycle to accommodate for the loss of material and reposition the sample surface at approximately the same height for the next analysis cycle. The third involves systematically changing the X raster position of the analysis beam to accommodate for the image shift during the data acquisition. Examples of each method, using model polymer samples, are presented along with a discussion of the advantages and limitations of each. These methods provide readily applied solutions to keep the images nominally aligned during depth profiles. Improved 3D images can be reconstructed using a combination of hardware and software solutions to optimize image alignment.

Five micron polymethyl methacrylate (PMMA) beads were mixed with polyethylene glycol (PEG) MW 2000 (Fluka) by mixing 1 ml of a 1% PMMA microsphere suspension (Phosphorex, Inc., Hopkinton, MA) with 9 ml of a 3% wt./vol. solution of PEG 2000 in ultrapure water. 100 μl of this mixture was pipetted onto a clean 1 cm × 1 cm silicon wafer piece and allowed to air dry. After this first drop dried, a 50 μl drop of 3% wt./vol. PEG 2000 was placed on top of the sample and allowed to air dry.

A mixture of polyvinyl alcohol (PVA) MW 85–124K and PEG 2000 was made by dissolving 9 g of PVA (Aldrich, St. Louis, MO) in 100 ml of ultrapure water by eight cycles of heating the solution in a microwave on high (Sharp R-230EW, 800 W) for 30 s and then swirling the solution for 90 s. After the final heating/mixing cycle, 3 g of PEG 2000 (Fluka) was mixed into the PVA solution. Once this solution was well mixed, 1 ml was placed in a clean glass scintillation vial and 1 small scoop with a spatula of 20–27 μm PMMA beads (Cospheric LLC, Santa Barbara, CA) was added and mixed thoroughly.

A poly 2-hydroxyethyl methacrylate (PHEMA) porous scaffold was generously provided by the lab of professor Buddy Ratner.²⁰ The scaffold consisted of an approximately 5 mm disk that was around 1 mm thick. A drop of optimal cutting temperature (Tissue-Tek OCT) compound (VWR, Radnor, PA) was placed onto a clean piece of silicon. The scaffold was then placed into this drop and allowed to sit for 1 h, after which it was removed, flipped over, and placed into a new drop of OCT for 1 h. The scaffold was then removed, placed onto a clean 1 cm × 1 cm piece of silicon wafer, and allowed to dry. Once dried, the sample was flipped over, and the surface previously facing the silicon wafer was analyzed.

A concentrated solution of PEG 2000 (Fluka) was prepared by adding PEG 2000 flakes into 2 ml of ultrapure water until they did not readily dissolve. Then the solution was heated 2× for 5 s in a microwave (Sharp R-230EW, 800 W) to force the remaining material to dissolve. To this solution, a scoop with a spatula of 20–27 μm PMMA beads (Cospheric LLC, Santa Barbara, CA) was added, and the solution was mixed thoroughly. A drop of this solution was placed onto a clean 1 cm × 1 cm silicon wafer piece so that it covered the entire surface. The drop was then allowed to air dry.

ToF-SIMS data were acquired using an iontof TOF.SIMS 5 using the liquid metal ion gun (LMIG) Bi₃⁺ for imaging and an argon GCIB for sputtering. Both beams were oriented at 45° relative to the sample surface and 180° from each other. All data presented are from positive ion data. Specific settings for the data sets presented herein are stated below. All depth profiles were acquired using the noninterlaced mode with sequential imaging and sputter cycles. A flood gun for charge neutralization was used for all samples. All data were acquired using primary ion currents low enough to avoid peak saturation.

Imaging was carried out in the high spatial resolution mode using a 0.02 pA Bi₃⁺ beam with a random raster for nine frames over a 100 μm × 100 μm area at 256 pixels × 256 pixels for a dose per image of 1.5 × 10¹¹ ions/cm². Sputtering was carried out using a 10 keV, 5 nA beam of argon 1000 clusters over a 500 μm × 500 μm area for 60 s resulting in a sputter dose of 7.5 × 10¹⁴ ions/cm² per sputter cycle. A total of 20 image/sputter cycles were performed.

Imaging was carried out in the high spatial resolution mode using a Bi₃⁺ beam rastered randomly over a 150 μm × 150 μm area at 256 pixels × 256 pixels for four frames. The current could not be measured for this experiment due to an instrument issue; however, based on previous work we would estimate that the current was approximately 0.03 pA. Sputtering was carried out using a 10 keV, 7.8 nA beam of argon 1000 clusters over a 450 μm × 450 μm area for 60 s resulting in a sputter dose of 1.2 × 10¹⁵ ions/cm² per sputter cycle. A total of 100 image/sputter cycles were performed.

Imaging was carried out in the high spatial resolution mode using a 0.025 pA Bi₃⁺ beam rastered randomly for 10 frames over a 200 μm × 200 μm area at 256 pixels × 256 pixels for a dose per image of 5.1 × 10¹¹ ions/cm². Sputtering was carried out using a 20 keV, 7 nA beam of argon 1000 clusters over a 500 μm × 500 μm area for 60 s resulting in a sputter dose of 1 × 10¹⁵ ions/cm² per sputter cycle. A total of 25 image/sputter cycles were performed.

Imaging was carried out in the high spatial resolution mode using a 0.027 pA Bi₃⁺ beam rastered randomly for five frames over a 100 μm × 100 μm area at 256 pixels × 256 pixels for a dose per image of 1 × 10¹¹ ions/cm². Sputtering was carried out using a 20 keV, 3.4 nA beam of argon 1000 clusters over a 500 μm × 500 μm area for 60 s resulting in a sputter dose of 5 × 10¹⁴ ions/cm² per sputter cycle. A total of 31 image/sputter cycles were performed.

ToF-SIMS dual beam depth profiles involve sequential analysis and sputter cycles to generate a series of images that, when properly stacked, can be used to reconstruct a 3D volume representation of the analyzed area. ToF-SIMS depth profiling to depths of more than 1 μm into a sample can result in lateral shifting of the images. For ToF-SIMS instruments where the analysis beam is oriented at 45° to the sample surface normal, each micron of material removed will shift the image laterally by 1 μm. An example of image shifting is shown in Fig. 1, where 3D and 2D “summed depth profile” data [Figs. 1(a) and 1(b), respectively] are shown from unshifted images of 5 μm PMMA beads in PEG 2000. While the beads are spherical, they are shown as oblong features in these unshifted images. Due to the fixed location of the guns in our (and most) SIMS system, these depth profile shifts are isolated to one direction (the X direction on these images). Three different methods to compensate for this type of lateral shifting during depth profiles are shown and discussed: (A) image shifting through software; (B) adjusting stage Z height; and (C) adjusting LMIG X raster position.

Software postprocessing of the data was used to accommodate for image shifting due to depth profiling. The software postprocessing shifted versions of the 5 μm PMMA beads in PEG 2000 are shown in Figs. 1(c) and 1(d). These data were manually shifted by 1 pixel at a time in the X direction. To accomplish this manual data shifting process, a graphical user interface (GUI) panel was added to the nbtoolbox imagegui (https://www.nb.uw.edu/mvsa/nbtoolbox). These data were loaded into the imagegui and shifted manually layer by layer to align the sequential images. This was facilitated by the fact that the spherical (round) shape of the beads could be clearly identified and used as orientation markers in each image. These data were then cropped using an in house custom matlab (Mathworks, Natick, MA) GUI that automatically crops a series of images by a chosen amount (imageshiftcropgui). The resulting data were then formatted so it could be imported into the nbtoolbox zcorrectorgui to produce 3D images.²¹ See supplementary material²² for a copy of the scripts used and a tutorial on their use.

The 3D reconstructed images show the data at an angle so that the elongation of the unshifted data [Fig. 1(a)] can be clearly seen when compared to the shifted data [Fig. 1(c)]. The 2D data are from a summed depth profile where all layers of the depth profile were summed into a single image. Without correction, the shift in the images can be clearly seen in the 2D summed data as the beads appear to be ellipsoidal with a long axis of around 10 μm. Manual shifting of the image layers adequately accounts for the image shifting as seen by the spheroidal shape of the particles in the 3D image [Fig. 1(c)] and the rounded features in the 2D image [Fig. 1(d)].

Since the previously discussed process of manual shifting and aligning is tedious and time consuming, we developed an automated version of the imageshiftcropgui (see supplementary material²²) that takes advantage of the uniform shifting that occurs with materials where the sputter rate does not change over time or with depth. This modified GUI enables shifting of a sequence of images by a given number of pixels per image inputted by the user (generally determined by a known sputter rate of the material). The images can then be cropped using the same GUI.

Figure 2 shows data from the 20 μm PMMA beads in the PVA/PEG sample before and after software shifting of the image layers. These data were loaded into Matlab, shifted, and cropped using the imageshiftcropgui. A shift of 1 pixel per layer was used in the X direction. These data were cropped to 156 pixels × 156 pixels to create a square matrix for 3D display. As seen in Fig. 2(a), before accounting for the image shift, the 3D data show elongated ellipsoidal features from the PMMA signal [m/z = 69 (C₄H₅O⁺)]. Elongated ellipsoidal features are also seen in the PMMA features of the 2D data [Fig. 2(b)] along with blurring of the image due to the image shifting. In contrast, in the autoshifted data [Figs. 2(c) and 2(d)], the PMMA features appear more spherical in the 3D image and rounded and crisp in the 2D image. As a second validation of the automated shifting process the manually shifted data from Fig. 1 were processed with the updated imageshiftcropgui (see supplementary material, Fig. S1).²² As seen in the supplementary material,²² the resulting autoshifted images look almost identical to the manually shifted data in Figs. 1(c) and 1(d).

A hardware based solution is another option to compensate for image shifting in depth profiles. This is particularly useful when the depth of analysis results in a shift that would require greatly increasing the analysis and sputter area in order to keep the features of interest within the field of view when using software shifting. As long as the sputter rate remains constant throughout a depth profile, the amount of material eroded per sputter cycle will remain constant. This means that the height of the sample will change by a constant amount each sputter cycle. The Z height of the sample can be readjusted by moving the sample stage by the same amount as the thickness of material removed. For this to work, the sputter rate must be known beforehand so the thickness of material removed per sputter cycle can be determined. The sample can be raised by this thickness during each sputter cycle, assuming that the stage Z height adjustment is accurate enough to change by the required increments. Figure 3 shows 3D depth profile data from a PHEMA scaffold filled with OCT acquired without any adjustments [Figs. 3(a) and 3(b)] and with adjusting the Z height during each sputter cycle [Figs. 3(c) and 3(d)]. From previous work with a PHEMA scaffold filled with OCT, it was estimated that the dose used for sputtering should erode around 1 μm of material each sputter cycle. Therefore, the stage height was manually adjusted by raising the stage 1 μm after each sputter cycle. The stage height was adjusted using the system's joystick control and monitored using the IONTOF stage control software.

As seen in Figs. 3(a) and 3(b), the scaffold pores show a slant from back to front on the left front face due to shifts in the beam position after the material was eroded from the surface. In contrast, the pores seen in Figs. 3(c) and 3(d) are more vertical and rounded and do not show the same slanted edges. Nonuniformities in the exposed pores in the shifted data may be due to irregularities in the scaffold shape or incomplete filling of the pores with OCT.

Adjusting the stage height to compensate for image shifting involves either using a manual control such as a joystick or a software interface to make changes. It is possible that this could be automated; however, most stage control software includes a backlash check that increases the time of the adjustment and could be problematic for shorter sputter cycles. An alternative method for using hardware to adjust for image shifting is to correct the beam location using the X raster position. For this to work properly, the X raster control must be capable of making adjustments within the range of shifts seen for a given sample, and the amount of shift in the image per step change in the X raster control must be calibrated for the given instrument. The shift in the image position based on the smallest possible change in the X raster position was calibrated using a mesh grid. This was done by imaging a mesh grid using several different X raster position settings (changing X Target in our iontof 5 system) while keeping the stage position constant and then determining the change in microns per change in X raster position based on the scan size and grid dimensions. It was determined that in our system an X Target change of 0.244% corresponds to a movement in the X direction of approximately 1 μm. However, the software only allows shifts of ±0.1% increments for the X Target control. The sputter rate for the 20 μm beads in PEG was estimated to be around 1 μm/cycle and so X raster shift of 0.2% per cycle was used. This shift was applied manually at the start of each sputter cycle.

This method was tested on a depth profile for the sample with 20 μm beads in PEG. Figure 4(a) shows the PMMA signal [m/z = 69 (C₄H₅O⁺)] for a bead depth profiled without any adjustments or X raster shift. As seen in the resulting profile, the bead appears to spread to the right of the image with initial signal (top of the bead) starting at the 25 μm mark and the center of the final signal at the ∼75 μm mark. Figure 4(b) shows a bead profiled using an LMIG X raster shift of 0.2% during each sputter cycle. As seen in the figure, the bead appears slightly more rounded, but it is clear that the raster shift was overcompensating for the shift in the sample height resulting in an image that slants to the left. This suggests that (1) our estimate for the sputter rate was incorrect and (2) more precise control of the X raster shift would be required to use this method to account for image shifting in depth profiles. As a comparison, Figs. 4(c) and 4(d) show the data from Figs. 4(a) and 4(b) after applying a post-acquisition (software) shift to the data using the imageshiftcropgui software described previously. For both of these images, the bead data look more spherical though it was noted that the X raster shifted data showed some artifacts on the right side of the bead. It is unclear whether this was due to a sample defect or an error in the image shifting.

The ability to depth profile through large amounts of organic materials could enable or enhance research in many areas, including characterization of buried interfaces and tissue depth profiling. However, there are many issues that still need to be solved before this type of analysis becomes routine. One of the key factors for successful 3D reconstruction of a depth profile is properly accounting for the image shift that occurs due to the fixed location of the analysis and sputter beam. As demonstrated in the data presented here, there are several ways one can correct for this shift. Each method presents advantages and disadvantages which must be weighed when choosing an appropriate data collection and processing method. Regardless of the method chosen, it is clear that one must have a good idea of the sputter rate of the material(s) used in each study in order to properly correct for any image shifts since the amount of shifting will be directly affected by the sputter rate of the material.

If the features being profiled have a relatively small shift in the lateral direction, one can typically apply post-acquisition image shifting using software that can properly align the images for 3D reconstruction. One advantage of software correction is that if the features being depth profiled are distinct, the images can be aligned without any knowledge of the sputter rate since the features can simply be lined up. This was demonstrated with the 5 μm beads in PEG sample where a manual adjustment was carried out by aligning the beads in each image in order to correct the 3D profile. After applying this manual shift correction, the resulting 3D image clearly showed [Figs. 1(c) and 1(d)] spherical features that were of the correct dimensions (∼5 μm in diameter). If no clear features are present or if the features are discontinuous between layers, then manual alignment without knowledge of the sputter rate would be difficult or sometimes impossible.

If the sputter rate is constant throughout the material, then automatic shifting and alignment of the images can be done using software. This was demonstrated using the 20 μm beads in the PVA/PEG mixture. For that data, it was shown that an automatic shift of each image by 1 pixel resulted in an accurate reconstruction of the 3D data [Figs. 2(c) and 2(d)]. One potential limitation of this method, and the above manual shifting method, is that the required shift could turn out to be a fraction of a pixel. Limiting the shift to whole numbers could potentially over or under shift the images. One could implement a shifting method that could interpolate between pixels; however, we have found that in most cases this type of accuracy is not required and adequate shift values can be found. Also, if one finds that a shift of 1 pixel per layer is too large, one could try down binning the data in the Z direction and then carry out the image shift. This would obviously reduce the resolution in the Z direction, but unless one is looking for small delta layers, this is usually not a problem with depth profiles.

Compensating for image shifting using the stage height or the X raster shift can be accomplished as long as the sputter rate is known. The main limitations of these methods include the accuracy and precision of moving the stage or X raster shift and the current lack of automation. It seems feasible that instrument manufacturers could include options in the software to adjust the stage height or raster shifts during the sputter cycle of a depth profile and eliminate the need for manual manipulation. As noted in the data presented here, adjustment of the stage height and X raster shift may not always completely account for the image shifts. This is particularly true if the sputter rate estimation or the adjustment precision is not accurate (as seen in Fig. 4 with the 20 μm PMMA beads in PEG). However, even reducing the shifts can be useful when profiling to larger depths since final adjustments can be made using software during data processing.

GCIBs have enabled a wide range of organic depth profiling experiments. The low damage and high efficiency of these sources allow profiling to large depths (10–100s of microns) in reasonable times. With these new opportunities come new challenges such as the large shifts that can occur when profiling to deeper depths. Herein, we presented three methods that can be used to account for these shifts and create data matrices that can be used to more accurately display 3D data from depth profiles. Each method has advantages and disadvantages, but all of them produce data sets capable of displaying a 3D profile that eliminates the large shifts caused by profiling to large depths. The scripts used in this work are also provided in the supplementary material²² along with tutorials on their use.

The methods presented include (1) software shifting (either manually or automatically) to correctly align sequential images. (2) Adjusting the stage height during the sputter cycle to account for the removal of material and align the sample surface to approximately the same height before each analysis cycle. (3) Adjusting the X raster shift during the sputter cycle to move the beam back to approximately the same location before each analysis cycle. It was shown that all three methods were capable of correcting for image shifts that occurred during depth profiling. After correction, the features in the 3D images were no longer elongated. Hardware adjustments during the sputtering step of the depth profile were shown to be able to compensate for large shifts during depth profiles, though it was noted that to properly use these methods, one must know the sputter rate of the material being analyzed and must be able to adjust the hardware in increments that match the sputter rate. The hardware correction methods are recommended when depth profiling deeper than a few microns as they keep the features of interest within the imaged area throughout the analysis regardless of the sampled depth. Hardware correction methods can be tedious as adjustments have to be made during each sputter cycle; however, working with hardware manufacturers should make it possible to automate these methods and simplify the data acquisition. A combination of hardware and software shifting is likely the best method to fully account for image shifts during depth profiles as the hardware adjustments can keep the features of interest within the field of view and then software can be used to fine tune the 3D image stack.

The authors gratefully acknowledge the support of the research by NIH NIBIB Grant No. EB-002027 to NESAC/BIO and thank Yusupha Jallow for preparation of the PVA/PEG used for the PEG/PVA mixture with the 20 μm PMMA bead sample. Part of this work was conducted at the Molecular Analysis Facility, a National Nanotechnology Coordinated Infrastructure site at the University of Washington, which is supported in part by the National Science Foundation (Grant No. NNCI-1542101), the University of Washington, the Molecular Engineering & Sciences Institute, and the Clean Energy Institute.