“This demonstration that our 200mm MEMS platform is now compatible with 300mm wafer fabrication shows a significant opportunity to cut MEMS production costs,” said Leti CEO Marie Semeria. “This will be especially important with the expansion of the Internet of things and growing demand for MEMS in mobile devices.”
Leti has a 30 year history in making MEMS, spinning out MEMS companies, and selling MEMS processes to chip firms. It currently has a team of 200 people working on MEMS.
The transferred process is its ‘M&NEMS’ technology – micro and nano MEMS, which came out of a programme to make many types of micro-mechanical structures using a single straightforward process.
What emerged a way to make any mixture of the following sensors on the same chip: accelerometers (x, y and z axis), gyroscopes (x, y and z axis), magnetometers (x, y and z axis), pressure transducers and microphones.
Capacitive (electrostatic) sensing has been avoided. Instead, for simplicity, all sensing is piezoresitive (piezo: Greek for squeeze).
“Piezoresistor sensing is robust, and importantly it is compact. We have miniature sensors, half the size of other peoples,” Jean-Rene Lequepeys, head of the Si component division at Leti, told Electronics Weekly.
Not to be confused with piezoelectric sensing, detection is through measuring the change in resistance of silicon nano-wires (220x220nm2 cross-section) as the tension in them changes. This is the ‘NEMS’ part of the process name. In lithography terms, 220nm wire geometry means relaxed constraints on available lithography.
These wires are just visible close to the substrate in the photo, which is of the measurement end of a single axis accelerometer. The ‘proof mass’ whose movement is measured, and extends far to the right of the photo (see accelerometer diagram) is hewn from 20µm-thick epitaxial silicon – the ‘M’icro part of the process.
Construction starts with a Si-on-insulator wafer, prepared with a 220nm silicon top layer – which will yield the piezoresistors and the foundations of other structures. “We can develop the same technology on bulk silicon, but it is easier on SoI,” said Lequepeys.
Piezoresistors are patterned and etched from this 200nm layer, then protected by a temporary oxide covering.
Following this, 5-25µm of silicon is grown on top, from which masses and other MEMS structures are then etched – the buried oxide layer acting as an etch-stop.
NEMS and MEMS components are finally released by removing the oxide, including SoI wafer oxide, from around the piezoresistors and masses.
All the sensors mentioned earlier have been designed to be made from the same basic parts – a 220nm thick layer for piezoresitive gauges and the full later-grown thickness for masses and other parts.
Magnetometers, for example (which are sensitive enough to be used as compasses), are moving structures with a ferromagnetic layer deposited on top.
The firm has transferred its technology Tronics Microsystems, which is making three accelerometers and three gyros within a <4mm2 footprint.
According to Lequepeys, while Leti can put triple accelerometers alongside triple gyros and triple magnetometers, competitors have to put the magnetometers on separate silicon. Type pressure monitors are a target for the process, as they need to combine an accelerometer and a pressure sensor.
300mm
According to Lequepeys, there are 97 process steps in M&NEMS, and a recipe for each has been adapted for 300mm production.
Moving the M&NEMS process to 300mm equipment is all about reducing the cost of MEMS production for consumer and automotive markets – at which M&NEMS is aimed. Ex-fab device costs are 30% cheaper in 300mm compared with 200mm, said Lequepeys.
300mm is also better for larger MEMS. “Auto-focus mechanisms are quite big devices, and so are digital loudspeakers and image-based sensors – for example for ultrasonic or micro-bolometer imaging,” he said.
Another reason for moving to 300mm is CMOS.
M&NEMS always uses a separate read-out chip to carry associated signal processing circuits. “We do MEMS and CMOS on the same wafer for other MEMS processes, but not M&NEMS,” said Lequepeys. “There are no major obstacles, but we would need to see an advantage in terms of cost.”
Both 200mm and 300mm processes at Leti offer 3d operations such as through-silicon vias (TSV), wafer thinning, wafer stacking and copper pillars which allow M&NEMS chips to be stacked with read-out chips, but while the 200mm process offers 130nm CMOS, 300mm opens the doors to 65, 45 and 28nm read-out processing.
Starting in a similar way, with an SoI wafer with thin top silicon, Leti has developed another process called ‘NEMS’.
Aimed at biological and chemical sensing, it runs only on Leti’s 200mm line. “The size of market not large enough to justify 300mm today,” said Lequepeys.
Sensing is once again though piezo-resistors, but there is no thick epitaxial layer. Instead, sensing structures are thin cantilevers made from the SoI wafer silicon (see photo right).
These cantilevers, which each have two piezoresistors, are vibrated by an oscillating electric field from a nearby electrode. The frequency at which they resonate is dependent on the mass of the cantilever and any mass on its surface.
NEMS has been used to develop a gas chromatograph in conjunction Caltech – the design of which is being exploited by a joint spin-out called Apix.
During the journey different molecules in the gas travel at different speeds, and by the time they reach the vibrating cantilever they can be sensed separately. Heating cleans out the sensor for repeated use.
Parts per million of many vapours can be can be detected, and sensitivity reaches parts per billion with heavy molecule like ethylbenzene.
Details of M&NEMS were presented at the European MEMS Summit earlier this month.
from News http://ift.tt/1WsE2lw
via Yuichun
沒有留言:
張貼留言