Challenges In Printed And Disposable Chips

Printing inexpensive chips using technology developed for newspapers and magazines is gaining traction across a wide range of applications, from photovoltaic cells to sensors on a flexible substrate. But it’s also adding a slew of new challenges that are unique to this approach.

The world of flexible hybrid electronics (FHE) — printing integrated circuits on or attaching thin IC chips to a flexible substrate — is a long way from discussions about 3nm designs and manufacturing techniques, or even mixed-signal chips in a package. Printed chips probably will never replace a complex SoC or be able to handle the compute tasks that even older silicon process technology can handle with ease. But this also is no longer just a science experiment. Printed chips already are being used for everything from monitoring temperature and vibration in industrial operations to flexible sensors that can be applied on the skin, and they are being tested for applications where rigid chips don’t fit or where they are difficult to replace.

Exactly how far along printed chip technology has progressed varies greatly depending upon market segment and the type of printing technology being used. The next challenge is to add the same kinds processes and consistency that have enabled semiconductors to be used in mission- and safety-critical applications.

“There is this whole idea that every object is going to have a sensor,” said Matthew Dyson, technology analyst at IDTechEX. “You’re going to have sensors in factories tracking temperature or whatever you want, and these will be quite cheap. Let’s say you have a box of drug tablets and you want to track their temperature or consumption, so you put a little sensor on the packaging and it will record the temperature and whether you’ve taken your medicine. Now the question is how much processing capability do you put on product and how much do you put somewhere else. There’s a tradeoff, because if you do a lot of processing in the sensor, it costs more to make, but then you have to communicate less data. And if all you do is acquire data and don’t really do anything with it, you then have to send all the information via some antenna somewhere else. And every bit of information that you send requires energy and where are you getting the energy from?”

Printed chips are an increasingly important element in that discussion, and work is underway across the globe to figure out the best tradeoffs between localized processing versus processing at the edge or in the cloud.

Technology innovation
Many of these approaches rely on existing printing technologies. Inkjet is the most prevalent for low-volume applications, where it has been proven as a low-cost but relatively slow option. Alongside of that are technologies like offset and gravure printing, which have gained in popularity due to the ability to print precise layers in high volume. Gravure technology, which uses images etched into plates, dates back to 1800s, when it was used to make high-quality art prints. Offset printing has been used for newspapers for even longer. There also is some research into using existing chip lithography equipment for creating more complex structures on a flexible substrate.

But the real innovation is less about the printing processes than the chemistry and stability of the inks and how they are bonded to a substrate. As with all semiconductors, the emphasis is on repeatability, consistency and economies of scale, and that effort extends well beyond just printing the inks. It’s also about how these chips can be diced, bonded to a substrate, and ultimately calibrated for accuracy and reliability.

“Gravure and offset are fast and have excellent edge definition, and aerosol jet is good for prototyping,” said Nathan Pretorius, prototyping and automation engineer at NextFlex. “What we found, though, is that calibration must be done after die-bonding. If you print something, the printing itself may stress a device and keep it in that state, so you need to calibrate that chip. For an RF die, you have to re-characterize the transmission line, and you have to do that on an FHE (flexible hybrid electronics) substrate.”

Oxidation poses another problem. “You need a material that will not form an oxide layer,” said Pretorius. “So you cannot have vias under the die.”

All of this factors into how these devices will be manufactured and ultimately where they will be used. For example, the Air Force Research Laboratory has been working with UES Inc. to develop sensors for monitoring gases on a continuous basis rather than just measuring the cumulative exposure, as many sensors do today.

“The requirements were that this needed to be a metal oxide sensor that was flexible, with a conformal band,” said Michael Brothers, technical program manager at UES. “If you look at most sensors today, they’re large and power-hungry. There is a dire need for flexible, wearable sensors that are basically disposable sensors, where you can create a large batch of them in hours in disposable form factors. So these have to operate at about 100 millivolts, the dielectrics need to be robust, and they have to be able to detect organic solvents.”

UES has lots of company in this space. Arm’s plastic armpit sensor, which has been in development for a couple years, is another example of just how small these devices can get. Arm has developed an organic FET that responds to organic compounds, such as body odor. “If you put an array of these devices and add machine learning, it can classify smell as one of the features,” said John Biggs, Arm co-founder and R&D consultant engineer.

Just how small and thin this technology can get is only one of the myriad issues facing flexible technology. Another problem is how dense the lines and spaces can be printed. The more that can be packed onto a flexible substrate, the more useful these devices can become for a variety of applications, ranging from complex sensors with on-board processing to more disposable applications.

“If you’re looking at a 32-bit microprocessor, that’s still challenging for printed electronics,” said Biggs. “Using thin films on a substrate is still three or four decades behind silicon. So think about 1 micron chips and 3 to 4 volts. Even if plastic electronics follows the pattern Moore’s Law, it will not track quite the same way.”

For some applications, that doesn’t matter. “Right now you can put a physical tracker or a smart label on food or wine to sell by a certain date,” said Biggs. “And some food products go off more quickly, depending upon whether they are stored in the sun or shade.”

Making sure these devices work
That’s a much more complex measurement than just the date or price of a given product, which is what many of the chips are used for today. But it’s also a much less expensive way of ensuring quality over time. And that has ignited research into what are basically disposable chips.

“We try to borrow as much as we can from the printed circuit board community,” said Will Stone, director of printed electronics integrations and operations at Brewer Science. “We do physical and environmental stresses. There are two main challenges. With a printed circuit board, you’re trying to make sure the joints—whether those are conductive epoxy solder or something else—can withstand the motion, because obviously those are not designed for that. And second is the traces themselves. Just by flexing you can get micro-cracking. We run those through the usual gamut of testing, like environmental chambers and stressing and flexing thousands or millions of cycles, just to ensure the integrity of the device.”

Not everything needs to be flexible, though. Rather than printing everything onto a flexible substrate, chips can be made small enough and thin enough to achieve the same goals for many applications. This involves metal oxide lithography, and while the lines and spaces are relatively large compared to advanced silicon chips, that is good enough for many applications.

“The attraction is that it’s cheap,” said IDTechEX’s Dyson. “You can just vapor deposit this stuff. And you can end up with things that are more flexible than silicon. You don’t have to go through the whole process of making a high-purity wafer. With normal silicon you first have to mine the silicon and then convert it to really high purity, make logs, then cut it up. With deposition, you don’t have to do any of that. You just have to have enough metal oxide transistors to make an RFID tag. It is not very much. So for RFID tags, they have tiny silicon chips on them. They are not flexible, but it doesn’t really matter.”

This is already happening with MEMS devices. NextFlex’s Pretorius said that for high-G applications, thinning a device from 700 to 450 microns reduces the mass nearly in half, which makes it useful in applications where traditional chips are not. “They’re not exactly flex, but they are lighter weight,” he said. “What we’ve found, though, is the thinning cannot be a post-processing step for MEMS devices.”

NextFlex also has been experimenting with different ways of dicing those chips, from self-dicing to using lasers. There are problems with each of those. For a multi-layer devices, lasers can be blocked by different metal layers. Other options include saw-dicing, which is not ideal for MEMS chips with vents, and laser ablation, which works but is not a clean solution.

Thinned, semi-rigid chips also have the challenge of adhering to a flexible substrate. One option is stud-bumping, but getting a good interface between the FHE and the bump without causing damage isn’t easy.

Brewer’s Stone has seen similar problems. “The inks are usually pretty stable once you get the process down,” he said. “It’s the epoxy solders that we use. A lot of these weren’t designed to flex at all, particularly the solders. We’re playing in a rigid world. We’re trying to make a rigid world flexible. At the end of the day, that’s the challenge there. Any of those joints are where you have the greatest areas of concern.”

Standards, supply chain, testing
FHE standards are being hashed out now by industry standards groups. “There are no explicit standards for FHEs,” said IDTechEX’s Dyson. “If you want to put this device in an aircraft, yes, there are standards. And if I want to put this device in a medical device, medical application, satellite, whatever, yes there are loads of standards. But that’s for the system level.”

The FHE industry is just beginning, and it’s still in the VC/government funding stage. “The companies that make FHE don’t make any money. You can buy them, but they’re making tiny volumes as a showcase to eventually try and get it adopted. They’re all existing with VC money and grants and so on. The flexible ICs — you can’t just go and buy them. Will you be able to buy them in five years? Yeah, probably.”

This also makes it hard to create a supply chain that is reliable and secure, with multiple vendors supplying the same quality materials. “If you’re the device maker, it means you can’t set up a supply chain because your supply chain may be be one company,” said Dyson. “That company will say, ‘We made the product with X capability,’ and you say, ‘Great, we’ll optimize, we’ll set up our process for that.’ And then that company goes bankrupt. You’re stuck. Whereas if you have standards, you say ‘Oh, no problem, I’ll just buy from somebody else.’ If you want to manufacture in volume, and you’re going to invest millions in a huge production line, you need to know that you can buy your stuff from everybody. You want standardized stuff.”

Standards adjust by industry. “There are huge numbers of applications, and some will be very niche, some will be mass-produced,” said Dyson. “There’s a huge difference in putting that with sensors on my clothing or sensors on shelving. That’s a relatively low-value mass market with cheap little sensors. For some little flexible bit that’s going to go inside a weapon, the requirements are completely different.”

Even for the military, which buys a relatively small number of specialized products, the lack of standards is a problem. “An FHE’s reliability is very limited. A lot of the stuff that is out there is proprietary. That makes it very difficult,” said James Zunino, materials engineer, Armaments Engineering Analysis & Manufacturing Directorate, U.S. Army Combat Capabilities Development Command Armaments Center (CCDC AC). “There’s a lot of information needed — the materials, the inks, the substrates, the coatings, the actual devices, and the environmental applications, as well as packaging and interconnects. For a lot of that information, we try to share it, but it’s not shared very well yet.”

The U.S. Department of Defense likes to know where everything is coming from and who touches what. “For qualification and certifications, it is definitely lacking for flex hybrid electronics. But you have to qualify and certify the materials, the equipment, the process and the operator. And if any one of those isn’t trusted, repeatable and known, then your final component will never be at the quality that you need it to be at to be able to be used safely, reliably and repeatedly,” said Zunino. “As everyone here knows, DoD is these heavily investing in flex hybrid electronics, the manufacturing techniques and the systems. But again, the failure mechanism, failure rates, all the accelerated test protocols and all the data aren’t properly defined, so we’re not doing a good job of telling people what we really want because we’re not quite sure exactly what we want yet. This industry and these technologies haven’t been around as long as a lot of the other things we’ve been doing for the past 100, 150 years. So reliability models for these FHE devices are not fully developed yet. And we keep developing new materials, new technologies and adding more variabilities to these problems and the models and the software, and the tools.”

Of course the military has some tough requirements. “We have the extra military challenges in the DoD. The operating storage and transportation environments of the Department of Defense is significantly different to what industry is developing. We have extremely high G loads, so sometimes it’s 50,000 to 150,000 G shock. Now we’re bringing in a whole hypersonic world into our electronics, extreme temperature/humidity ranges. So we need stuff to work and -45°, -50° degrees and all the way up to where the soldiers have to operate which is 170° Fahrenheit. There are places our soldiers are operating where the daytime temperature is 174°F. And then you bring in the hypersonics, where you can now see temperatures up 3,500°F and all of electronics have to supply these environments at a high shock and vibration load. So if you are shooting something at like 50,000 Gs, and it hits something, that’s an extreme shock environment that’s very hard to test and predict any other way. And a lot of times we want our stuff to sit for 30 years before we then shoot it on to 50,000 Gs into something, and it has to work.”

NextFlex, Georgia Tech’s George W. Woodruff School of Mechanical Engineering, Dupont and the U.S. Airforce Research Lab are working on testing standards. “It’s really important to have standards for how you qualify and quantify the reliability of these types of materials,” said Benjamin Stewart, a mechanical engineering doctoral candidate with Georgia Tech, who worked making testing standards for FHEs and helped create a new testing machine for flex circuits.

“One person says, ‘We stretched this new productive ink to 30% and experienced this level of assistance increase.’ Another person could do other similar test and get wildly different results because the standards are really loose,’ said Stewart, noting loose standards are a sign of an immature field. “We’re in the early stage of maturity here. So one standard to talk and use in this space is called the IPC guideline on flexibility/stretchability testing. Right now this guideline says, ‘If you run a test, just tell us everything about your test. If possible, choose to describe, so what is your specimen look like? What was the size of your specimen? How many samples are in your study? What is the strain rate you used? Show us your electrical test data versus stretch percentage.’ So they just say, ‘Tell us everything because we don’t want to tell you what is important. So just say what you did, and we can take a look at it.’ What’s missing here, and what everybody’s trying to really understand — and this is part of the focus of our group — is we’re missing the actual recommended parameters for each of these things. So you want to be in this strain-rate range if you are interested in this realm of physics, if this is your application, to say you should use this strain rate, your sample should be this size, things like that.”

Georgia Tech is working to define how to test and what standards are important to meet, including how to create test samples layouts and stretch tests that won’t be polluted by connector issues. The school developed a biaxial stretch test machine that creates standard stretch and measure conductivity/resistance results.

Fig. 1: Testing approaches by Georgia Tech. Photo: Susan Rambo/Semiconductor Engineering

Fig. 2: Georgia Tech’s Benjamin Stewart (presenting at SEMI’s FLEX/MEMS & Sensors conference in San Jose, Calif. on Feb. 26, 2020) explains a new testing machine the university worked on with partners. The machine enables standardized testing flexible on uni- and bi-axial stretches of flexible circuits to measure the circuits’ conductivity while under stress. Georgia Tech found that uniaxial stretching did not reduce the conductivity so much as the biaxial stretching. Photo: Susan Rambo/Semiconductor Engineering

End of the line
And finally, just because chips are inexpensive doesn’t mean they are easily disposable. A big question is what to do with these devices once they have lived out their projected lifespan.

“This is toward the idea of putting sensors on everything,” said IDTechEx’s Dyson. “If I put a sensor on my coffee cup, it basically needs to be compostable. I’m not going to spend the time disassembling it and putting my silicon in a separate box. Clearly that’s stupid. So you’ve got to just be able to recycle it. But how do you do that, because you’ve got all these different materials? It’s very hard to recycle stuff with a lot of different materials. Your silicon isn’t going to compost. Metal oxide is not really going to compost, either. You want this stuff to be so cheap that you can throw it away. And then you get the ethical sustainability issues. Is it ethically okay to throw away very small amounts of silicon? People are worrying about biodegradable substrates. They’re not worrying about biodegradable chips. That’s miles away.”