I recently talked with Mike Willis of Pivotal Resources about MEMS. Why? Because what Mike doesn’t know about ink jet heads probably isn’t worth knowing. And because MEMS seems to be one of the key buzzwords of the moment for industrial print development.
Mike, what is MEMS?
There’s a lot of interest at the moment in silicon MEMS printheads, and another term encountered is thin-film piezo. These are actually two different things, although they go hand in hand.
MEMS stands for Micro Electrical Mechanical Systems, and is a way of making structures either on the surface or inside silicon layers. Thin-film piezo is a way of making the thin flexible actuators that are required for ink jet printheads made from silicon.
Silicon has of course been used for decades for manufacturing integrated circuits, and in our industry for making thermal ink jet printheads. But apart from through holes for ink feeds all of the processing has been flat and on the surface. In MEMS, silicon structures and components are fabricated. One of the largest applications is accelerometers used in smart phones to detect movement.
Silicon has been used for ink jet thermal printheads since the 1980s, requiring precise holes for the ink to come through. The channel structures for thermal ink jet are then formed in layers of photopolymers on the surface. However, for piezo printheads more elaborate processing is required, with more complex structures.
Silicon MEMS manufacturing technology is being used for piezo printheads to make the ‘channel’ plate, that is the ink channels both from the ink manifold to the actuator chamber, and then on to the nozzles. The silicon is etched to form the required structures. What you need are very narrow straight walls to make the ink chambers. Traditionally silicon has been etched with liquids, called wet etching, but you then get angled walls which are not always optimum.
The use of silicon for piezo printheads had to wait for a process called deep reactive ion etching (DRIE) – also called the Bosch process because Bosch engineers were one of the early developers. This process has been around a while. However, for many years it was very slow and it has only been in the past 5 to 10 years that the process has been speeded up to make it viable for production.
So why use silicon?
If you want to make printheads to print at high speed and at high resolution you need to be able to cram many channels and nozzles into printheads at a high density. With DRIE you can produce fine features and therefore manufacture on a much smaller scale than before. You have to be careful when you reduce the spacing between channels as a problem called crosstalk can occur – pressures in one chamber being transmitted to neighbours that will then cause variations in drop placement on the substrate. Because the silicon is stiff you can reduce the wall thickness between channels without increasing the crosstalk between them.
How does this work with thin-film piezo?
With silicon MEMS you have the ability to make fine channels and structures, but the problem is that the piezo material that flexes when a drive waveform is applied, creating pressures in the ink that ejects drops. Traditionally bulk piezo has been used, where it is added to the printhead as a separate piece and patterned into individual actuators. But a high-density printhead design has narrow ink chambers, so you need a much more flexible piezo material than before to flex in the narrow chambers. The answer is to make the piezo layer very thin – only one or two microns thick. You can’t make this from slices of piezo. What you have to do is deposit the material directly as a thin film – hence the other term of the moment: ‘thin film’ piezo.
Various techniques are used, the sol-gel process being the most popular. The material is put down as a solution and as the carrier liquid evaporates it gels. You then fire it to form the piezo ceramic structure. However, this is very difficult and it has taken years of development to get it right. The piezo material really must have the right properties and crystal structure after it is deposited and fired.
So what are the other issues?
One of the key challenges is that to be able to do this you have to be able to process wafers of silicon, which means you have to have a fab (a fabrication plant or foundry that processes silicon). These are expensive clean rooms filled with expensive equipment. Existing fabs that produce semiconductor chips are reluctant to process piezo materials due to the risk of contaminating the clean rooms – the lead in piezo materials is a poison for semiconductors as much as humans!
Fujifilm Dimatix has their own fab. By contrast, Xaar has decided not to build its own fab, intending instead to buy in processed silicon from elsewhere and assemble it into their own design. This is a common approach these days, for instance Cambridge company CSR is a fab-less chip company. Even HP outsources the majority of its silicon dies for thermal ink jet printheads.
As a result in the growth of MEMS fabs, applications for deploying MEMS has increased. There is also an increase in specific fabs dedicated to piezo and MEMS. This is not just for piezo ink jet printheads but for other applications as well.
So why MEMS? And why are they coming together at the same time?
The reality is that we all want small drop sizes. To achieve this all of the dimensions of the printhead must be smaller, and you need to find a way to manufacture the designs accurately. You need consistency in drop size across a wide single-pass printhead to avoid banding artefacts in the printed image and with this approach - silicon MEMS and thin-film piezo - you can achieve smaller drop sizes on a consistent basis.
For example, you can make modules like the Fujifilm Dimatix Samba head which is 45 mm wide, and you can stack up modules to make a full width printhead. Because the performance of each printhead module is similar then you can make a page array with them. 30 years ago you would have had to carefully select modules with similar performances from all of the ones you made to do the same job, but now the need to mix and match is unnecessary.
With smaller drop sizes you see higher drop frequencies – these silicon printheads can run up to 100 KHz, whereas 20-30 KHz has been the norm up to now.
Another thing is that you can integrate driver chips directly onto the silicon, reducing the number of external connections. With the new Xaar head, the number of external connections is minimal, and this goes for the Samba head too – they are simpler and as a result, more effective.
So why would you not want to make silicon MEMS heads?
There is a huge cost involved in the design of silicon MEMS printheads. You need expertise to design in silicon, which printhead manufacturers haven’t required in the past. There are yield issues due to the complexity of the structure. If your manufacturing yield is low each head is expensive. If you can achieve a high yield you achieve a lower unit cost.
In the past some heads have contained heaters to reduce the ink viscosity so it can be jetted, in particular with uv-curable inks, giving more freedom for ink formulation. Silicon has a relatively low thermal conductivity, so I doubt we will see silicon MEMs heads running at similar elevated temperatures as previous piezo printheads. This will make ink formulation harder.
Is all ink jet head design manufacture going to become silicon MEMS?
No, because not everyone needs or wants high-resolution high speed printing. There will be ink difficulties due to the temperature, and therefore viscosity limits. Also they may not be the cheapest printheads - they are optimised for high-speed, high-quality printing.
Who is using silicon MEMS?
Panasonic were one of the first to make silicon printheads, and these were used in printers by IMPIKA, now part of Xerox. The most well-known manufacturer is Fujifilm Dimatix with their Samba head. Epson has been making silicon MEMS heads with thin film piezo for their higher end products and label presses. Konica Minolta announced a silicon printhead a couple of years ago which will be available soon, and Ricoh and Xaar have also made announcements - they will both be selling printheads soon. And Xaar is jointly working with Ricoh on thin film piezo technology.
How did we get to this point with head development?
Silicon MEMS has really taken off in the past decade, with Memjet, Kodak’s Stream continuous ink jet printhead technology in their Prosper press, Dimatix and Seiko Epson. We see from patent filings that almost every printhead manufacturer has MEMS designs in the labs. Everybody is doing it because it really does provide some advantages.
If you want to build a fixed array printhead for high-speed high resolution printing, then up until now you’ve had to mount multiple printheads, not only across the print width, but in the process direction too, to achieve the print resolution. This has been because the base dpi of printheads has only been 360, or 600 dpi for many years. To be able to stitch a single row of 1,200 dpi modules together is much easier.
Will we begin to see these heads used in future with industrial printing?
Single pass printing is already there with many industrial applications, in particular textiles, packaging and décor. For markets where high resolution and therefore the highest print quality is required, then silicon MEMS heads offer a solution.
But there are issues, in particular ink related. Take packaging for example: silicon MEMS heads will no doubt feature in printing onto films and labels with aqueous inks, but using UV inks is not so easy. That said, Heidelberg/Gallus are running a label press with UV curable inks so it is possible.
I can also see where it is being used in 3D printing: you need small drops to print fine features, which is possible with this head technology and is seen as advantageous. And for economic manufacturing you need to run at high speed, so the requirements are similar to high speed commercial printing. I think that this could be a big application area and the only problem will be the viscosity of the material. This is where the high speed sintering process, that Neil Hopkinson is developing at Xaar, has a great advantage as it is putting down a low viscosity ink rather than the build material.
In the future we will undoubtedly see more applications for silicon MEMS!
Thanks to Mike Willis for this excellent explanation :)