EM Characterization of UV Mirror Thin Film Layers

Scott Paoni

Department of Environmental Medicine
University of Rochester Medical Center, Rochester, NY 14642 USA


The project goal was to characterize thin film layers in a broken UV mirror to determine the possibility of manufacturing a replacement in-house within the URNano facility. SEM, FIB, X-ray EDS and image analysis were primarily used to evaluate thin film composition.


Dielectric multi-layered materials coat UV mirrors to modify the reflective properties of the mirror with the intent to alter optical properties such as reflecting as much light as possible or functioning as an optical band width filter to restrict light transmission in a certain wavelength range. This UV mirror sample was housed inside a microscope system and fractured during use. Due to the cost of replacement, there is an interest in evaluating the feasibility of manufacturing the mirror in-house within the URNano facility.

A broken UV mirror sample was embedded in PMMA, polished and EM imaged. While producing images with good film layer resolution, edge artifacts appeared in the upper thin film layers complicating analysis. X-Ray EDS analysis yielded overall compositional information, but a goal of composition analysis per layer could not be met due to resolution issues.

While TEM analysis of a thin sample would have been the preferred method for further evaluation, there was not enough time in the project window to attempt to obtain an adequate sample. FIB was instead performed to notch a “wall” in the sample to remove sample material from where the interaction zone would be in hopes of improving resolution for SEM based X-Ray analysis. FIB produced higher quality EM micrographs, but failed to improve layer resolution for X-Ray analysis, although image quality improved for more reliable analysis of layer thickness. EDAX analysis produced elemental information indicating the thin film components to be oxygen, silicon and tantalum.


Sample Prep

Broken UV mirror samples were kindly provided by Brian McIntyre. Samples were sputter coated with 3nm of either Pt or Au, adhered to a stub with sticky tape and conductive graphite adhesive. The Pt coated sample was initially imaged “as is” in cross-section, then embedded in PMMA, polished with routine iterations of polishing paper (coarse through 0.5µ), recoated with Pt and re-imaged. An unpolished Au coated sample was prepared, as though imaging in the horizontal plane, focused ion beam lithography performed with the resulting cutout imaged in cross-section.

EM and X-Ray Analysis

SEM/FIB imaging and lithography was performed using a Zeiss-Auriga FESEM. FIB utilized a 30kv Gallium source and 4nA aperture to mill a rough trapezium pattern then 50pA to “polish” the area before analyzing. To create a “wall” for imaging and analysis, the first trapezium cut was performed, then the pattern flipped and positioned to mill on the backside of the first cut. X-Ray EDS was performed using EDAX Genesis Software.

Image Analysis

Image analysis techniques (Image-Pro Plus V7, Media Cybernetics) were used to measure thin film layers and enhance thin film column structures. Micrographs were processed using a flatten filter (to even out background variations by reducing intensity variations in background pixels), a special background filter to create a background image that was then subtracted from the original image using an arithmetic operation and a median filter applied to remove impulse noise. The highest magnification column micrograph (Fig. 7E/F) was contrast enhanced but not processed due to resolution concerns related to empty magnification. Multiple areas of interest were drawn on a thin film layer micrograph post image processing, size filtered to remove small artifacts and auto-bright counted to measure layer thickness. Micrographs were auto-contrast enhanced and colorized for presentation only. Colorization was performed using Image-Pro Plus and/or Photoshop.


Polished Samples

A polished UV mirror sample was SEM imaged in cross-section to evaluate thin film layer thickness as shown in Figure 1. Fig. 1A and 1B show SE2 and Inlens micrographs of a thin film stack with Fig. 1C and 1D representing image processed and colorized versions of A and B respectively. Of note is the increased signal intensity in the brighter top three layers, which is attributed to edge effect artifact. Qualitatively, there is a thicker top layer noticeable in the Inlens image (Fig.1B), a thinner bottom brighter layer, and thick and thin appearing layers throughout the stack. Due to artifacts and variations in signal contrast across the image, image analysis was inconsistent and difficult to resolve.

se2 polishedA  inlens polishedB

se2 polished processed+colorizedC  inlens processed+colorizedD

Figure 1: SE2 (Fig. 1A) and Inlens (Fig. 1B) micrographs of polished UV mirror. SE2 (Fig. 1C) and Inlens (Fig. 1D) processed and colorized images. Acquisition information: 10kv 4.1mm 33kx

Focused Ion Beam Lithography

The edge effects observed with the polished sample and contrast variations made for unreliable layer thickness and elemental composition analysis per film layer, which was a project goal. Ideally, a thinner sample would be either sectioned via ultramicrotome or milled out and captured for TEM analysis using focused ion beam lithography (FIB). Due to project time constraints, and the inherent difficulty in obtaining such a sample, an alternative FIB technique was employed. Another sample was prepared in a horizontal imaging plane and FIB used as described above to mill out a sample wall. The hypothesis was by removing sample from behind as well as in front we could create a pseudo-TEM grade sample through which the interaction volume could be driven in order to reduce noise as the region of sample generating secondary electrons and x-rays would be closer to the size of the incident electron beam, which serves to increase resolution.

Fig. 2A shows the FIB attempts at creating a wall with Fig.2B showing a colorized version of the resulting wall. The blue top represents a platinum coat laid down over the sample top prior to FIB milling to increase mechanical stability of the sample to prevent the wall collapsing or warping, whereas the yellow colorizes the film stack. Fig. 2C and 2D are another view of the original and colorized FIB wall respectively. In comparison to the polished samples, this alternative FIB technique produced a cleaner image and was used for subsequent image analysis.

FIB regions zoomed outA  FIB milled wallB

FIB inlensC  FIB inlens processed+colorizedD

Figure 2: FIB lithography with Fig. 2B representing the manufactured FIB “wall”. An FIB “wall” for analysis is shown in Fig. 2C and processed and colorized in Fig. 2D. Acquisition information: Fig.2A: 10kv SE2 5.1mm 298x; Fig. 2B: 20kv Inlens 2.3mm 57kx; Fig. 2C/D: 5kv Inlens 5.1mm 34kx


X-Ray EDS EDAX analysis indicated the overall sample elemental composition included silicon (Si), oxygen (O) and tantalum (Ta). The characteristic x-ray spectra seen in Fig. 3A also indicates the presence of gallium (Ga) and platinum (Pt). As this sample was sputter coated with 3nm of gold, the Pt signal is from FIB platinum deposition as described earlier, with the Ga signal coming from deposition of the Ga in the sample during FIB. The spectra shown is from a whole field x-ray analysis, but layer by layer analysis showed similar results, indicating the alternative FIB technique was unable to reduce noise and improve resolution for x-ray layer analysis. Fig. 3B shows the x-ray spectral map from a scan region of a few layers, indicating there is not enough resolution to determine elemental composition per film layer. Even though SEM based EDAX analysis can only analyze elemental composition and not bonding states, based on the current information and known thin film components, the make-up of the thin film stack is likely alternating darker SiO2 layers and brighter Ta2O5 layers.

characteristic x-ray spectra
Figure 3A: Characteristic X-Ray spectra.

x-ray spectral map
Figure 3B: X-Ray spectral map.

ZAF correction and standard-less analysis also yielded the following layer elemental ratios:

ZAF correction + standardless analysis
Figure 3C: ZAF correction and standardless analysis

Thin Film Layer Image Analysis

The FIB inlens image was processed as described above and analyzed over five areas of interest (aoi) as shown in Fig. 4B. Even though the alternative FIB technique and image processing resulted in a cleaner image with less background noise, the micrograph still contained signal variation. Using one whole region of interest would result in noise prohibitive for quantification, although smaller regions allowed for more reliable and reproducible analysis both qualitatively and quantitatively. However, thickness values for the second to last Ta2O5 layer (layer38) are likely inaccurate and lower than they should be as indicated from qualitative validation. Analysis with the smaller aoi’s as described resulted in smaller thickness counting than was obvious qualitatively. This was not corrected until an aoi was drawn over only two tantalum layers. A repeat of analysis drawing smaller areas of interest box by box yielded similar results to the larger aoi.

This situation exemplifies some pros and cons of image analysis and optimization of standardized processing steps and how they relate to the resolution of intended analysis. Preferably, standardized image processing steps should be applied consistently across images within the sample set within the experiment, but there is a need to consider the inherent diversity of samples (especially biological ones) and system dynamics as well that can influence consistent image processing results. When performing image processing, the computer will only do what the operator designs it to do and this must be recognized. Not paying qualitative attention to both the visual processing during analysis and quantitative results could result in missing some scientific diversity, or system limit to resolution, that results in incorrect data being obtained. In this case, the inconsistent application of image processing steps throughout all layers is likely an artifact of the imaging system but also the sample, as the layer in question, one the bottom most layer, is also qualitatively one the thinnest, thus presenting different signal values for the image processing to work on. If there were more project time, more effort should be expended to obtain optimal images with less signal variation as well as developing more precise image processing steps that can be applied reliably across the whole image. As such, the obtained values provide a reasonable first approximation of thin film layer thicknesses left to be validated through future TEM study and improved image processing development.

FIB inlens to analyzeA   image with aoi's analyzedB

original contrast enhanced onlyC   original processed+contrast enhancedD

Figure 4 A-D: Thin film analysis of FIB Inlens micrograph (Fig.4A; Acquisition information: 5kv Inlens 5.1mm 34kx). Fig. 4B shows multiple areas of interest analyzed for auto-bright counting. Fig. 4D represents the processed image for comparison to the auto-contrast enhanced original image.

film layer thickness dataE

Figure 4 E: Quantitative thin film layer thickness (nm). Image calibration ~3.5nm/pixel. Standard error values shown with n=5 of sub-regions analyzed.

Electron Flight Simulation

Electron flight simulation (Figure 5) models the trajectory of electrons through a sample and the resulting interaction volume, providing an indication of image resolution. Fig. 5A models the sample in horizontal imaging plane, with a characteristic tear drop interaction volume visible. Fig. 5B models a cross-sectional plane treating the alternating layers as a bulk homogenous sample. While the classic tear drop shape is less noticeable, what is observable is how quickly the interaction volume spreads out, in both models, and how this is indicative of the noise preventing reliable layer by layer x-ray analysis. Fig. 5C is a rendering of the homogenous simulation penetrating the FIB wall and spreading out decreasing the quantitative resolution.

original contrast enhanced onlyA   original processed+contrast enhancedB

Figure 5: Electron flight simulation of thin film stack (Fig. 5A) and approximation of homogenous material of film stack interacting with incident EM beam in cross section (Fig. 5B).

Tantalum Layer Column Structure

Columns are layed down during deposition process and molecules will stack up on each other like columns. Depending on the deposition process and temperatures involved, these column structures, and resulting pores around them in the layer filled in with the same composition, may change which could impact the optical properties of the layers. Measurement of the columns relative to the whole layer could be useful information in evaluating the manufacturing process of the layers. For the purposes of this project, the column to layer ratios were not calculated but instead qualitatively evaluated through image processing techniques for curiosity.

Figures 6 and 7 show enhancement of tantalum columns through image processing. Of interest in the Figure 7 panels is the “worm like” particles present in the Si layer. From the characteristic X-ray spectra we know that gallium was deposited in the sample so we hypothesize that these particles are some combination of gallium particles and re-deposited sample particles. The lower right region of Fig. 7A is clouded out with debris, which is suggestive of this interpretation. Fig. 7E+F were not processed as mentioned previously due to resolution concerns regarding empty magnification. Empty magnification occurs when increasing magnification only results in a magnified image with no gain in object resolution and information. In this particular case, the increased magnification resulted in decreased image resolution and quality when processed with the same set of procedures as the other images in the panel.

original contrast enhanced onlyA   original processed+contrast enhancedB

original contrast enhanced onlyC   original processed+contrast enhancedD   

original processed+contrast enhancedE

Figure 6: Thin Film Column structure (Fig.6A), processed and colorized (Fig. 6B). Zoomed-in sub-region of micrograph auto-contrast enhanced (Fig.6C) for presentation to see extent of image noise. The zoomed-in sub-region image is processed (Fig.6D), auto-bright and watershed counted (Fig.6E) with an overlay of quantified film columns shown (Fig.6E). Acquisition information: FIB 10kv Inlens 5.1mm 124kx.

original contrast enhanced onlyA   original processed+contrast enhancedB

original contrast enhanced onlyC   original processed+contrast enhancedD

original contrast enhanced onlyE   original processed+contrast enhancedF

Figure 7: High magnification images of thin film column structure in colorized originals with corresponding colorized processed images. 95kx (Fig.7A+B), 156kx (Fig.7C+D) and 256kx (Fig. 7E+F). Acquisition information: FIB 10kv Inlens 2.3mm.


X-Ray EDS analysis yielded elemental information that, based on known thin film layer materials, likely indicates the mirror is composed of alternating layers of SiO2 and Ta2O5. Image analysis techniques were able to determine thin film layer thicknesses with enough confidence that Brian McIntyre believes he can show the measurements to a lens designer for interpretation of likely standard thicknesses. Analysis also further validated the qualitative observations of thicker and thinner layers, including a much thicker initial silicon layer at the film stack top and a thinner tantalum layer at the bottom with varying thicknesses of alternating layers in between. Additionally, image processing enhanced the presence of tantalum layer column microstructures formed during thin film deposition manufacturing allowing for qualitative curiosity observations of the film stack.

If more project time was available, future directions should include characterization of the mirror through higher resolution thin sample TEM imaging and X-Ray analysis of layer thicknesses and elemental composition for confirmation of the suspected material bonding states and diffraction patterns.


I would like to express immense gratitude to Brian McIntyre for his infectious enthusiasm for course instruction, project mentorship and guidance and for performing FIB on my sample at an insanely short working distance with sharp angles.

I also thank my wife, Jennifer Paoni, for patiently providing me with Photoshop instruction and, as always, keeping me centered.

Please enter any comments, criticisms, questions, etc. below.

Your name:

Email address: