A Study in Fabricating Microstructures (Part 1)

(This article is based on our recent SIGGRAPH publication “Elastic Textures for Additive Fabrication.” The project is a joint collaboration with multiple authors from multiple institutions. Please see the project web page for full details.)

Test prints of microstructures on B9Creator

3D printing technologies have come a long way since Chuck Hull invented Stereolithography back in 1986. The first 3D printed object, a small blue eyecup, took about 30 minutes to finish [1]. Upon closer examination, one can clearly see layer marks caused by the coarse resolution. Today’s printers are better in both accuracy and speed. Layer thickness has been reduced to less than 30 microns, and objects with size of a baseball can be fabricated in minutes instead of hours.

One of the amazing properties of 3D printing technologies is that the complexity of the printed shape is practically uncorrelated to the time and cost of printing. In fact, it is often faster and cheaper to print something complex than a solid cube. According to Wikipedia, such phenomena is called the “Complexity Paradox.”

All prices are computed for printing with Frost Ultra Detail (FUD) material at shapeways.com

This “complexity paradox” has allowed artists to materialize their wildest imagination. However, 3D printing is not perfect, and fabricating intricate structures is still not easy. I hope to share my experience in 3D printing flexible microstructures, where the thickness of each strut ranges from 0.2mm to 0.5mm.

Sample printed with Autodesk Ember printer.

Part 1: Pick the right technology

There are more than a dozen different 3D printing technologies in the market nowadays. My collegues and I have tried Fused Deposition Modeling (FDM), Selective Laser Sintering (SLS), Stereolithography (SLA) and Direct Light Modeling (DLM). Here is a summary of our experiences.


The first FDM printer we tried is a Dimension Elite courtesy of FabLab Pisa. This printer has the desirable ability to print with soluble support structure The minimum layer thickness is 0.178mm [2].

Sample printed with Dimension Elite

As shown in the picture above, we have tried printing a sample with feature thickness about 2mm. There are visible artifacts such as staircase and dangling strands of extruded material. We did not measure the geometry for accuracy, but judging from the picture, the accuracy is unlikely to be good.

Another FDM printer we tried is Prusa i3, again courtesy of FabLab Pisa. The attractive property of this printer is that we can use Soft PLA, a flexible extruded material. We print our samples at layer thickness 0.3mm and without any support.

Samples printed with Prusa i3

From the image above, it is obvious that the resolution is not sufficient. Printing without support seems to exacerbate the situation by causing the structure to collapse slightly like melting ice-cream in a hot day.

Recently, Pavel Pavlov, a Thingiverse user, has better luck printing using a FDM printer with a different material:

Microstructure printed by Thingiverse user Pavel Pavlov

Still, broken struts are visible from the image above, and thicknesses of struts are not consistent across the shape.

So in conclusion, we have not found a FDM printer with high enough resolution to print microstructures. For curious FDM printer owners/manufacturers, you can find some of my models freely available on my microstructures page at Thingiverse. Feel free to give it a try, and let me know how it goes.


The best SLS printers I know are from EOS, and Shapeways offer this material as “White, Strong and Flexible.” This printer uses nylon powder as the raw material and the grains are sintered together using powerful laser beams. At the highest resolution, each layer is 60 microns thick [3]. A very attractive property of SLS printers in general is that support structures are not needed. This means one could simply send any design as is to the printer. Once finished, the non-sintered powder can be shake off. I ordered a few samples from Shapeways. (If you also use Shapeways’ service, make sure to check the “print anyway” option, otherwise they will almost certainly reject such orders.)

Both the input and the printed output are shown here. The thickness of each bar is ~0.5mm. Sample (A) and (B) simply fall apart after printing. This suggests the resolution of the printer is not high enough. Sample (C) and (D) seems to be mostly OK except minor breakages. However, (C) is severly distorted during shipping.

In my second attempt, I have increased the thickness of each strut to 1mm, and this time everything is fine, except the structure is not very “micro” and the shape is very rigid.

In conclusion, SLS is a promising technology, but at the current stage it does not provide sufficient resolution for printing microstructures, and shipping such models tends to introduce extra problems.

SL and DLP:

Stereolithography (SL) is known for its accuracy. It uses liquid resin as the raw material, and when exposed under ultraviolet light, liquid resin transformed into solid state. This is the technology invented by Chuck Hull almost 30 years ago [4].

A related technology is Direct Light Processing (DLP). Instead of using laser to solidify liquid resin one spot at a time, it uses a projector to solidify the entire layer all at once.

For this project, we have tried ProJet 7000 HD from 3D Systems, Form1 from FormLabs, B9Creator from B9Creations, and Ember from Autodesk.

The first printer we tested is ProJet 7000 HD at NYU Advanced Media Studio. Using its XHD mode, each layer is 50 micon thick [5]. According to the spec, the XY resolution is “0.001–0.002 inch per inch of part dimension” [4]. I had some hard time understanding what exactly the unit “inch per inch of part dimension” means. It seems to suggest the XY resolution is relative to the size of parts. Bigger parts will have coarse resolution and tiny parts will have very fine resolution (and the printer is infinitely accurate for infinitely tiny parts!? That is a very bold claim, 3D Systems).

As indicated from the image on the left, the first attempt did not go very well. 3D Systems’ Windows-only proprietary software is obsessive in adding support structures, and it is impossible to use your own supports. The automatically generated support structure is all over the place and completely altered the microstructure. Given the complexity of our models, these support structures are impossible to remove.

Furtunately, the kind folks at NYU Advanced Media Lab allowed me to access the printer software, and turns out it is possible, but tedious, to delete some of the automatically generated supports. This produced a much better outcome.

Samples printed on ProJet 7000 HD, feature thickness is around 0.5mm.

I am impressed by the accuracy and durability of the prints. Overall, ProJet 7000 HD is definitely a good candidate for our purposes, but the usability of the accompanied software needs a lot of improvement.

The next printer we tried is Form1. Its layer thickness can be as low as 25 microns, and the XY resolution (i.e. min feature size) is 300 microns [6]. Form1's software is much more user friendly, and it even has an option to avoid generating internal supports. Max Lobovsky from FormLabs helped us to print some test samples. Again, we have a lot of initial failures.

To be fair, these failures are not entirely due to the printer. The samples were printed at an angle, but the microstructures were not designed to be self-supporting at 45 degree rotation. Therefore, failure is kind of inevitable.

Max did eventually succeed with Form1 as indicated from the image above. Although some struts are missing, it could be understood because my design is not meant to be printed at an angle. Many thanks for Max Lobovsky and FormLabs for testing this!

The next printer we tested is B9Creator. This is what we used to print most of our samples.

Samples printed with B9Creator using red resin

All samples were printed at 50 micron XY resolution and 30 micron layer thickness. These early samples shown above were printed with the red resin. We later switched to cherry resin for better accuracy.

The advantage of B9Creator is that you can change every setting. To ensure the microstructures can be printed, careful calibration of the built platform and the projector is essential. The resolution is heavily influenced by the quality of the PDMS coating over transparent window. We recoated the vet with PDMS for every 5 to 8 prints. A cloudy PDMS coating could cause severe decrease in print quality as shown below.

Comparison of prints of the same structure on a clear PDMS vs on a cloudy PDMS.

A few more setting needs to be changed as well. I typically set “A base” time to below 20 seconds because the default setting tends to cause partially printed base to stuck on PDMS. We avoid changing per-layer exposure time fearing it would alter the material properties of the base material.

More samples printed using B9Creator

Overall, B9Creator is a good fit for printing microstructure due to its easy-to-adjust settings and its low cost. We used it to print the majority of samples. The drawback of B9Creator is that it require a lot of maintainess. During the deadline crunch, I have to recoat PDMS almost weekly.

Last but not least, we are able to test our samples on Autodesk Ember printer, curtesy of Duann Scott and the Autodesk Spark team. I used Autodesk’s standard clear resin to print all the samples. The layer thickness is 25 microns, and XY resolution is 50 microns [7].

Diamond pattern printed on Autodesk Ember printer. Each cell is of size 5mm, and each strut is 0.2mm in thickness.

By far, Ember gives the best quality prints. The resolution of Ember is amazing. I am able to print struts slightly thicker than a human hair (200 micron in diameter). The printed microstructures are amazingly flexible.

Sample printed with Ember. Feature thickness ranges from 0.5mm (left) to 0.2mm (right).

The most delicate structure I have printed on Ember has feature thickness at around 0.2mm in diameter. The default setting needs to be changed for this to work. Fortunately, multiple resources and DYI guides are available ([8] and [9]). In the end, I found changing just 2 parameters is sufficient: check the variable strength exposure checkbox and increase the per-layer exposure time to 4s or 6s.

Just like B9Creator, the clarity of the transparent window influences the achievable resolution by a lot. For Ember, the transparent window becomes cloudy at a much lower rate. I was able to print daily for over a week and the PDMS only became slightly cloudy.

Cloudy PDMS on Ember can cause broken struts.

Admittedly, Ember is not perfect. All of my prints failed for the first one or two weeks of getting access to an Ember. On the plus side, the failures are much less catastrophic comparing to B9Creator. I have not ripped any PDMS yet.

Overeall, Stereolithography and Direct Light Processing produce the highest quality prints. Among all the printers we tried, Autodesk Ember stands out for its high accuracy and usability. The B9Creator is a very good choice but one have to put in a lot of efforts to get good quality outcomes. ProJet 7000 HD is also capable of printing microstructures, but its high price and poor software support make it less desirable.

Stay tuned for the second part of this article where I will talk about how to design self-supporting and flexible microstructures. If you want to try to print some samples yourself or if you have any questions, please leave a comment.


[1] “Roots of a High-Tech Revolution” The Wall Street Journal, August 26, 2014 http://www.wsj.com/articles/roots-of-a-high-tech-revolution-1409066696

[2] “Dimension Elite Product Specs” http://www.stratasys.com/3d-printers/design-series/dimension-elite

[3] “System Data Sheet FORMIGA P 110” http://www.eos.info/systems_solutions/plastic/systems_equipment/formiga_p_110

[4] “Apparatus for production of three-dimensional objects by stereolithography” US Patent 4,575,330, http://www.google.com/patents/US4575330

[5] “ProJet 6000 & 7000 Brochures” http://www.3dsystems.com/sites/www.3dsystems.com/files/projet_6000_7000_0514_us_web.pdf

[6] “Form1+ Tech Specs” http://formlabs.com/products/form-1-plus/tech-specs/

[7] “Ember Specs” https://ember.autodesk.com/specs

[8] “Ember’s Print Settings” https://support.ember.autodesk.com/hc/en-us/articles/203693789-Ember-s-Print-Settings

[9] “How to tune Ember’s print settings for new resin” http://www.instructables.com/id/How-to-tune-Embers-print-settings-for-new-resins/?ALLSTEPS