Introduction: Window Clouding in the Ember Printer
One of the major pain points to using both SLA and DLP printers is that the optical transmission of the PDMS windows change over time. This "clouding" or "fogging" is very clear to the naked eye, as the two photos of a used windows beside a new one show. There are two causes for the clouding. First, the UV absorbing species in the resin soak into the PDMS decreasing its transmission over time. In some resins this appears as a pale white glow over the whole window due to the fluorescence of the UV blocker. This is particularly obvious in the lower angle left photo. In resins like Autodesk Standard Clear, the uptake of resins by the PDMS is small, and the amount of UV blocker absorbed it even smaller. However, this small amount is still sufficient to greatly reduce the optical transmission. This process takes weeks and ultimately reduces the transmission from 90% to ~ 65% even when they are not used for printing.
The second mechanism occurs only in areas used for printing. In this case the surface becomes rough and scatters light. If the same print is performed over and over a very clear image will form on the PDMS. In the photos above this is the more opaque clouding in the upper right of the window.
This Instructable illustrates what happens when the same exact image is printed over 30,000 times (33 inches of 25 micron layers!) on the Ember printer. Throughout the course of this experiment the optical transmission of the window was measured using UV-VIS spectroscopy, and performance of the printer evaluated via a test print. Afterwards, a microscope and FTIR spectrometer were used to examine the window. This experiment demonstrates that it is possible to keep printing long after the PDMS window clouds. The Autodesk Standard Clear was observed to cloud the window at 5000 layers; however, there was no change in printing performance until 10,000 layers. Between 10,000 and 30,000 there was a small loss of performance. Somewhat surprising, the exposure times had to be reduced as the window clouded. A general recommendation would be that the windows need to be replaced when divots and tears to PDMS can be seen with the naked eye.
Step 1: Resin Soak Tests
When PDMS windows are soaked in resin they absorb a small amount of the resin. While this amount is small, 0.7 % wt./w.t for Autodesk Standard clear, the photoinitiator(s) and UV blockers present in the resin can also be taken up by the window. Because these species strongly absorb light, the optical transmission of the window is reduced. This process is not fast, and it requires about a month for the window to reach equilibrium.
Step 2: Test Geometries
I started off using two prints for this experiment. The first is a 30 mm diameter cylinder which is offset to leave half of the print area untouched. The second print is a series of 50, 100, 150, 200, 250, 300, 350, 400, and 450 microns posts at angles of 90, 60, 45, 30, and 10 degrees. This print tests the ability of the printer to produce delicate high aspect ratio features. Both files have small little features set off to the side so that when the slicer re-centers the .stl, the objects are still off-center. Sliced at 25 microns the cylinder prints in 4799 layers. I would recommend printing somewhere between 1000 and 3000 layers. Unpacking the tarball and deleting the extra layers is the fastest way to change the number of layers printed.
After, 24,000 layers the 30 mm cylinder became very difficult to print without jamming, so I switched to a 20 mm x 20 mm x 3000 layer block that culminated in the post test print. It was possible to use two different sets of conditions by treating the base 3000 layers as "burn in" layers, and the rest as model layers. This print was less prone to jamming, and I would recommend using it for future tests. I have also included a 1000 layer version for resins that are more prone to clouding and need to be measured more frequently.
Step 3: UV-Vis Measurements
A jig was fabricated to mount a tray on an optical post, and the tray was then positioned between the two fiber optic cables of the UV-VIS (Ocean Optics USB4000). The height was adjusted by simply adding a second post to the bottom.
One draw back of this setup is that the fiber optic is located more than an inch from the end of the fiber so the scattering losses are greatly exaggerated. This results in increased sensitivity, but a smaller dynamic range. As such, the measured transmission does not directly correspond to the actual transmission at the resin-PDMS interface in the printer.
Step 4: Printing and Measurements
Getting a 30 mm cylinder to print for thousands of layers can be a bit of a hassle.
Jamming of the printer is problematic because of the visco-elastic nature of the resin. As the build head lowers the arm holding the resin tray deflects. The arm only returns to its original position after the resin has time to flow out between the PDMS and printed part. If the light is turned on before the resin can flow out, the printed layer is thicker than the set value, and after tens to hundreds of cycles enough of an error accumulates in the z direction to cause the printer to bind and jam. To avoid this I would recommend adding a pause after the build head lowers to the PDMS by setting the "ModelApproachWait" to 2.25 s for a 300 mPas resin.
After I terminated the cylinder print, I would then print the posts.stl and record the number or posts printed. The tray was then drained and carefully cleaned using an acetone wetted wipe used to remove the final traces of resin. Unfortunately the Kim-wipes I used scratched the surface. This reduced the light transmission on the unprinted side of the window compared to the window that was just soaked in resin. I would strongly recommend using microfiber cloths for cleaning. After cleaning a UV-Vis measurement and a quick snap shot was then recorded.
Step 5: Photos
Despite the low quality of the photos a gradual increase in opacity and heterogenatity of the window can be inferred. At 30K layers several types of damage are apparent to the naked eye.
1) A general white haze in the printed area which began appearing at ~ 5K layers.
2) More opaque regions throughout the printed areas, which area to be more frequently encountered in areas further from the drive shaft. These began to appear somewhere around 7-10 K layers
3) Areas where large (~1 mm ) chunks of PDMS have been removed from the surface. I believe this is what finally "killed" the ability of the window to print without jamming. When these appeared at 24 K layers the 30 mm cylinder stopped printing. When they appeared at 30 K layers with the smaller solid the printing again became nearly impossible. Again, this damage appear at the furthers points of the print from the point of rotation. These are also the areas that experience the greatest force from the separation process.
4) Scratches from the print being rotated over the PDMS. Somewhat surprisingly these do not appear to correspond to the highly damaged areas, and are perhaps due to a marred surface on the build head.
5) Scratches from cleaning out side these regions.
Step 6: Microscope Images
A closer examination of the surface at 30K layers with microscope shows just how damaged the PDMS surface really is. One possible explanation for this damage is that first a finely featured rough surface forms on the PDMS which causing the clouding. Subsequent printed layers are then able to adhere to these asperities and the separation process during printing causes larger pieces of PDMS to rip off the surface forming the millimeter scale damage.
Step 7: UV-Vis Spectroscopy
The UV-Vis confirms the rapid increase in opacity that is visually observed in the areas used for printing. This approach is much more sensitive that visual observations, as the the transmission decreases from 85 % to less than 5% by 10 K layer; however, this sensitivity also eliminates any insight to changes past 10 K layers. A further drawback of this method is that the spot sized is focused, while the clouding results have been observed to be heterogeneous over the entire printed area. A better approach for future work may be to use a light source with a broader spot size and a radiometer positioned close to the window.
The extreme sensitivity of the UV-Vis experiments may be useful if damage to the window in non-print areas needs to be accurately measured, as transmission in the non-printed areas is clearly lower than that observed in windows only soaked in resin.
Step 8: Printing Performance
The most surprising results of this experiment is how robust the printing process is. While the finest features produced in the test print did increase in size on aggregate, the data is very noisy, and the change is slight at best. Unlike the cylinder print, no changes were made to the exposure, or other conditions used for the test print. It may be possible to preserve the printing resolution for fine features long after the window has clouded by increasing the light intensity to compensate for the uptake of the UV blocker and supply a similar dose to the print.
Step 9: ATR-FTIR
Fourier transform infrared spectroscopy (FTIR) is an analytical techniques that excels at identifying the chemical composition of a material. Attenuated total reflection (ATR)-FTIR is particularly useful variant as it only samples the first couple of microns into a sample. The FTIR spectrum of a virgin window is initially dominated by the large symmetric stretch of the methyl groups in the PDMS at 1250 1/cm. After, the resin swells into the PDMS in the unprinted parts of the window a very weak peak from the carbonyl groups in the resin is apparent at 1720 1/cm. This implies that some small amount of resin is present in the unprinted parts of the window. However, in the printed part of the window the peak from the resin is now comparable in size to the peak from PDMS. This suggests that the composition of the window at the surface now includes a substantial amount of polymer from the resin in addition to PDMS.
Step 10: The Change in Cure Time
To get the 30 mm cylinder to print it was necessary to continually change the exposure time. Unexpectedly, it was necessary to decrease the exposure time to prevent jamming. Initially 2.0 seconds would print with out jamming, but by the end only a 1.62 s exposure would print without jamming. Why would this be necessary? One possibility is that it is easier for the print to adhere to roughened surface. Another possibility is that oxygen permeability of the surface decreases as it clouds. If the oxygen permeability is reduced, then the layer being printed would cure faster. This result may explain why the resolution did not appear to decrease as the window clouds. It may be that the reduction in light due to the the UV blocker is more than balanced by the decrease in oxygen inhibition of the polymerization.
Step 11: Final Thoughts
The results of this experiment were very surprising. The clouded window performed much better than expected, and the window lasted considerably longer than any window used to date. I had expected the exposure time would need to be increased as the window clouded to compensate for UV blocker uptake over time. However, it appears that the changes occurring at the PDMS surface alter the polymerization rendering such an adjustment unnecessary. As such, rather than measuring the light intensity at the PDMS surface to adjust the printing parameters for clouding, a better approach may be to design a performance test that directly evaluates the printing process, and provides guidance on when to replace the window.
So whats going on at the surface to cause this clouding? These results seem to suggest that first the resin soaks into the PDMS. Subsequently, it polymerizes forming a hybrid acrylate-PDMS surface. This material is going to have a lower oxygen permeability, and lower surface energy. Buckling of the surface may also be occurring at this stage, contributing to the clouded appearance. All of these effects increase the chance that the printing layer will stick to the the surface, resulting in large chunks of PDMS being removed during the separation process, and the window being ruined for printing.