By Lee Richmond, Makino

In an endless pursuit of the capabilities of micromachining, we often run across test cuts that are of interest. Recently, a medical staple mold demonstrated what is possible with current technologies.

A customer challenged a Makino user with a medical staple mold that would require cutting tools down to 0.20mm radius. The user was very capable and experienced in mold building, and was comfortable at 0.30mm, but anything smaller was uncharted territory.

This mold is a good example of a lot of current micromachining work. Innovative products need innovative molds that often have difficult feature details, which can be nearly impossible without the proper machinery and techniques. This is especially true when it comes to the medical market, where cost of production is typically secondary to protecting intellectual property, following stringent government guidelines, and making a superior product.

The mold steel is 420 stainless, hardened to 52 – 54 HRc (Rockwell hardness C scale). It has eight cavities and inside corner radii had to be held to 0.20mm. The part is so small that hand-polishing to finish any details or to clean up a rough finish is impossible, so the mold must be accurate and polish-free straight off the machine.

The original process was to CNC (computer numerical control) mill an electrode, EDM (electrical discharge machine) the cavity detail, finish mill and grind the rest of the details. This took a total of 67 hours. It’s an efficient method when direct milling is not possible, but there’s a lot of time tied up in prep work, and not actual production of the mold.

The new process was to hard mill the mold on a Makino V22 vertical machining center, skipping the EDM process. This alone saved the user a substantial amount of time. However, to challenge the production capabilities even further, the customer requested a unique test cut be preformed, cutting the corner radii in half to 0.10mm.

Micromachining bears great resemblance to its macroscale counterpart. Photo courtesy of MakinoClick here to enlarge image

The part required five tools with nine different programs to keep the cuts in different areas of the mold to match specifications. Tools from 2mm down to 0.10mm radius were used. Tool-to-tool blending is critical, so we used Makino’s hybrid Automatic Tool Length Measurement (ATLM) system, installed on the V22. This system ensures that tool growth and wear is controlled, since the margin of error is so minute and surface finish requirements are so critical.

The final result allowed us to complete the core in 21 hours and the cavity in 13 hours. The majority of the 21 hours on the cavity was to pick out the tiny corners. The total time to produce the mold was reduced from 67 to 34 hours, or approximately 50 percent.

After the success of the new hard milling process, we decided that it might be interesting to push the boundaries of this process even further, just to see what was possible.

We decided to tweak the part design a bit – shrinking the overall size of the part and tightening up the corner radiuses, now down to 0.076mm. To do this, we had to use a 0.05mm radius tool, as no standard tool in between 0.10mm and 0.05mm radius is available.

The smaller cuts increased the cut time from 13 hours to 32 hours, a 244 percent cycle time increase in the cavity. The core proved too long to continue the cut. Step-over in the core got down to 2.5 microns and step-down to 4 microns. Tool failures started to pile up.

In conclusion, we were able to get great results cutting corner radii to 0.20mm, then even better when we went down to a 0.10mm. We hit a wall when we attempted a 0.076mm corner radius with a 0.05mm radius cutter. Cycle times spiked dramatically and the tooling wore out quickly.

Without a tool between 0.10mm and 0.05mm radius, a 0.076mm corner radius is impractical in hardened 420 stainless steel.

Why does this matter? Because we learn what’s possible from our failures as much as we learn from our successes. Since we were able to hold 0.10mm corner radii while improving cycle times in a hardened steel mold, it shows that we’re going down the right path. All the ingredients aren’t there to go even smaller, but we’re getting very close.

The test cuts will continue, and we are confident we’ll get a 0.076mm corner radius, and even tighter, with some more tweaking. In the meantime, we need to examine our processes, tooling, and machine to make sure we’re doing everything we can to increase accuracies even further.

To get tolerances like these, there are a few essential components that must be considered. Typical machining applications allow for several microns or more of variation. When you’re dealing with micro-molds, a single micron of inaccuracy will often ruin your work piece.

First, the machine you choose must be stable, both thermally and for vibration. Any growth in the tool that isn’t addressed properly can devastate your accuracy, just as chatter can when milling hardened steel with a tiny tool.

The way Makino accomplishes thermal stability is with a core-cooled spindle and insulation of the work envelope. These advancements permit us to accurately control and predict how much the tool will grow, which allows the machine’s software to compensate, and therefore maintain our accuracies. Vibration is kept to an absolute minimum with heavy castings, a spindle designed to be utterly rigid, and software which compensates for the properties of metal being cut at high speeds.

Makino engineers found that they could get great results cutting corner radii to 0.20mm, then even better down to a 0.10mm radius, but that tooling wore out quickly when attempting a 0.076mm corner radius with a .05mm radius cutter. Image courtesy of MakinoClick here to enlarge image

Another aspect of the machine tool that must be considered is the servo control technology. Makino has implemented high resolution scale feedback that is able to recognize and achieve sub-micron movements, and therefore hold the tolerances of micro applications.

The second thing you’ll need to accomplish cuts this accurate and small is a programmer with a creative streak. Most programmers aren’t accustomed to working with step-overs and step-downs of only a few microns or less, or with tooling that will break if it rubs too long or hits a corner incorrectly, so it’s important to consult someone who’s done it before to program the cuts correctly.

To program a mold in hardened steel is hard enough, but when you factor in the accuracies needed and the unique tooling, it’s simply not something that just anybody can do. And that programmer needs the right tools – programming software that gives him the ability to run the tricks, like trichoidal roughing and arc fitting corners, needed to successfully mill hardened micro molds.

Finally, you need the right tooling. Everyone has their vendor preference, but without tools with predictable life and performance, that cut as they’re designed to, you’ll never hold the tolerances needed for micro work.

Most of the tooling needed in submicron work is so small you literally can’t touch it without breaking it – most machinists aren’t used to using it, much less buying it. And keep in mind that tooling is usually not an area where cutting corners saves you money. You’ll just end up paying on the back end with re-work, delayed deliveries, or even worse, outsourcing the part when you can’t produce it.

Nearly every micromachining process is a trial-and-error situation, especially when it comes to sub-micron cuts. This process can be shortened greatly with the aid of engineers experienced in micromachining, along with some fundamental ingredients like a capable machine tool, tooling, and programming tailored specifically for micromachining.

Lee Richmond is micro market manager at Makino (www.makino.com) in Auburn Heights, Mich. He can be reached at [email protected].

New array suggests throughput could eventually meet commercial needs

By Charles Q. Choi

Dip-pen nanolithography can literally draw structures only nanometers in scale, but was always limited in throughput by how many pens it could write with at the same time. Now the technique’s inventor has devised an array with hundreds of times more pens than before.

Dip-pen nanolithography (DPN) uses atomic force microscope (AFM) tips as pens and dips them into inks containing anything from DNA to semiconductors. The new array from Chad Mirkin’s group at Northwestern University in Evanston, Ill., has 55,000 pens – far more than the previous largest array, which had 250 pens.

The new array can draw 55,000 likenesses of Thomas Jefferson in a space the size of a nickel – images consisting in total of some 470 million 80-nanometer-wide dots – in less than 30 minutes.

“The throughput barrier is a grand challenge for the nanotechnology field, as lithography is the foundation for all that we do. And in this paper Mirkin and coworkers drive a truck through this challenging barrier,” said Joseph DeSimone, a chemist at the University of North Carolina at Chapel Hill and director of the new UNC Institute for Advanced Materials, Nanoscience and Technology. The researchers reported their new DPN array in the November 6 issue of Angewandte Chemie.

“One of the gating aspects of the technology is that, with a single-pen tool, you’re extremely limited in throughput. Seeing they’re able to take their research to a 55,000-tip array instrument would definitely address the issue of throughput,” said Vahe Mamikunian, an analyst with Lux Research.

NanoInk, the Chicago company Mirkin founded in 2002 which is commercializing dip-pen lithography, has licensed the new array and has exclusive rights. “It will be on the research market early next year,” Mirkin said.

The schematic shows the fabrication process for 2D cantilever arrays developed by the Mirkin group at Northwestern University and licensed by NanoInk. Image courtesy of Chad MirkinClick here to enlarge image

The array was made by modifying a microfabrication process developed for single AFM probes. Oxidized silicon wafers had 10-micron square openings lithographically patterned into them, and pyramidal pits were etched in those openings. These pits served as molds for pyramidal probes when films of silicon nitride were deposited. The silicon nitride on the front side of the wafers was then lithographically patterned to form arrays of cantilevers.

The probes are roughly 7.6 microns high, with tips about 60 nanometers across. “The tall tips keep the arm holding the tips from running into the surface,” Mirkin said.

The researchers also bend the cantilevers by coating them with gold and then annealing them. The resulting curvature is due to how gold and silicon nitride layers expand differently in heat, and how the different layers in the cantilever restructure due to annealing.

“The bent cantilevers give us more play in the vertical direction to get any misaligned tips in contact with the substrate. In other words, the tips in the array do not have to be perfectly in one plane to bring them all in contact with the surface to be patterned,” Mirkin explained.

The result is an array that can bring all the pens in contact with their substrate using merely gravity, as opposed to a complex set of feedback systems. “This makes the approach innovative, straightforward, inexpensive and extremely useful and versatile,” Mirkin said. Once all the tips of the array are in position, the array is locked in place by a rapidly curing epoxy resin on the tip holder.

“What is most surprising to me is the degree of fidelity achieved,” said Jim De Yoreo, a member of the scientific staff at Lawrence Livermore National Laboratory who conducts research into scanned probe nanolithography techniques. “Highly multiplexed cantilever arrays have been fabricated by a number of groups. But overcoming the challenges of adequately uniform inking of all the tips and obtaining registry between the substrate and all tips has never been achieved. I would have expected these to be daunting tasks, but Professor Mirkin’s group has dealt with both quite handily.

The most important implication of having overcome these two challenges is that high-throughput constructive patterning of a nearly unlimited set of functional chemistries at sub-100-nanometer length scales is now possible. This puts commercial use of constructive scanned probe lithography within our grasp.”

In initial experiments, Mirkin and his colleagues could generate 88 million dot features with their new array, each pen generating 1,600 dots in a 40 by 40 array, where the distance between each 80-to-120-nanometer-wide dot was 400 nanometers. In other experiments, they generated protein nanoarrays. “I was really excited to see 55,000 cantilevers working in unison – really impressive work,” said Thomas Thundat at Oak Ridge National Laboratory in Tennessee, an expert on nanomechanical devices who leads a nanoscale science and devices group at the lab.

The massively parallel DPN approach heralded by the new array “really opens the technique up for many applications, especially in the life sciences,” Mirkin said, “and perhaps some in other areas like integrated electronics and photonics.”

Nanoarrays that DPN can build can “allow us to study cells and the factors that control their behavior – adhesion, growth, motility, differentiation and apoptosis – in a way one could never achieve with conventional technologies. It will allow us to study how viruses work at the single particle level – how they bind and infect cells.”

Moreover, “they will provide important insight into many areas of cancer research – the factors that lead to metastasis events,” Mirkin added. The nanoarrays DPN can build “will lead to many new screening procedures for new therapeutics for a variety of diseases, including many forms of cancer.”

“I suspect the market opportunities for high-resolution, massively parallel DPN will be as diverse as semiconductors, large-area displays, pharmaceutical/packaging tagging for anti-counterfeiting and research tools,” DeSimone said.

The new 55,000-pen DPN array was used to draw 55,000 likenesses of Thomas Jefferson in a space the size of a nickel in less than half an hour. Image courtesy of Chad MirkinClick here to enlarge image

Still, it seems this array “is looking at some of the applications that nano-imprint lithography is targeting, and I’m not sure that’s a good place to be targeting,” Mamikunian said. “The reason for that is, with dip-pen nanolithography, you effectively have one supplier right now in the form of NanoInk. So there’s a limited supply base. With nano-imprint, five or six companies are really pushing the technology, from Molecular Imprints to Obducat to Nanonex.”

Mamikunian noted that some of the applications NanoInk is targeting for DPN are different from those of nano-imprint, such as pharmaceutical anti-counterfeiting and photomask and circuit line repair. And, in their paper, Mirkin and his colleagues noted that the advantages their massively parallel DPN approach has over nano-stamping include the distortion effects that plague nano-stamping and nano-stamping’s need to fabricate a mask each time a new design is required.

Another question Mamikunian had with the new array was how manufacturable these were on a large scale.

“If you were ordered to produce 50 or 60 a quarter, what would your ability be to produce such large array instruments predictably and reliably?” Mamikunian said. “It’s really great work, and there will be some applications coming out of it, but until they show how to make such a large array tool on a mass scale, (there are) more questions than answers.”

“These are made by conventional microfabrication processes,” Mirkin responded, adding that mass production was not a big hurdle. Integration is the real issue, he contended, including getting different inks to different tips on the fly. “Maintaining registration with the underlying substrate during the entire process needs to be worked out as well,” he added. “All doable, but it will take some time.”