Three DARPA-funded research teams have completed a foundational study of chip-scale vacuum pumps by inventing three very different approaches to removing air from a sample chamber with a volume of one cubic millimeter, which is about one-tenth the volume of a grain of rice. These new technologies will enable many micro-scale devices which require a vacuum or a controlled flow of gas, such as Lab-on-a-Chip sensors, radio frequency MEMS switches and microscopic vacuum tubes.

Why is DARPA interested in tiny devices that operate in a vacuum? Over the past quarter of a century, a host of MEMS (Microelectromechanical Systems) and other devices have been developed. Many of these devices require microscale pumps for their operation, such as the various Lab-on-a-Chip analytical sensors, which must pump air or liquid samples through a series of testing stations on the chip. This requires real-time in-situ pumping capacity, as do the maintenance procedures performed between measurements.

There are also microscale components that require a vacuum environment for proper operation, such as radio frequency MEMS switches, microscopic vacuum tubes, and other parts that depend on electron or ion optics. Simply sealing the devices while in a vacuum chamber often does not give sufficient service life, as leaks and degassing are far more detrimental to vacuum devices on the microscale. Accordingly, such components may need a vacuum pump on the chip to maintain functionality.

There are also microscale devices which may work better in a vacuum, but don’t require that environment for simple operation. For example, a MEMS resonator, essentially a tiny tuning fork, will work fine at atmospheric pressure. However, some of the air sticks to the surface of the tuning fork, changing the effective mass of the resonator. In addition, the wind stirred up by the vibration of the resonator damps the vibrations. Together, these effects change the resonant frequency of the resonator and cause the resonance to be less sharp. Perhaps worse, these alterations will vary with the temperature in which the overall system is functioning even if the MEMS resonator is hermetically sealed in a gas.

These and other potential advantages of integrating effective and efficient vacuum pumping capabilities into microscale systems triggered DARPA’s interest in finding some practical approaches through which these desirable properties can be attained. In 2008, it issued a request for proposals for its Chip-Scale Vacuum Micro Pumps (CSVMP) program asking for the development of microscale vacuum systems satisfying (or at least addressing) the following points.

DARPA sought a vacuum system no bigger than a US penny that is able to produce a vacuum of less than 100 microPascals (uPa, a billionth of atmospheric pressure) in a vacuum chamber with at least a cubic millimeter of volume (about the size of a grain of coarse sand). The vacuum system was to use less than a quarter of a watt, and contain instruments to accurately measure the pressure in the chamber.

To help appreciate the magnitude of the task DARPA had posed, consider the Creare Engineering miniature turbo/drag pump, considered to be the world’s smallest practical turbo pump produced by conventional machining tools. The pump is some 3.8 cm in diameter with a length of 8 cm, making it slightly larger than a D-cell battery. It weighs about 150 g (5.3 oz), and can reach an ultimate pressure of a few uPa. The pumping speed is four liters/second at a rotor speed of 200,000 rpm, and consumes about 2 W in doing so. This pump needs to be followed by a roughing pump as it cannot pump against atmospheric pressure.

So while the Creare pump can meet DARPA’s vacuum pumping requirements (it is more than capable of pumping a mm-scale chamber down to 100 uPa), it is nearly 200 times too large in volume, and uses ten times too much power in doing so. And because it is a specially designed pump made in very small numbers for NASA applications including on the Curiosity rover, the price is likely out of proportion when used to make a US$10 MEMS chip work properly.

Based on their responses to DARPA’s request, three research teams from the University of Michigan (U-M), MIT, and Honeywell Corporation, were chosen to mount independent R&D efforts to meet DARPA’s requirements. They responded to this challenge by inventing a series of novel microfabricated chip-scale vacuum pumps. Let’s take a look at a few of these.

University of Michigan

The single pumping cell (two-stage) of the Scalable Michigan gas micropump (left), and pho...

U-M researchers developed three different pumps. The first is a high-frequency 24-stage resonant peristaltic roughing pump (PDF). They used a scalable, resonant design that produced a very large flow rate of 0.36 cc/minute and allowed it to reduce the pressure of a sample chamber to about 97 kPa from the atmospheric pressure of 101.3 kPa. The theoretical limit on the performance of such a pump is to reduce the pressure to around 1.5 kPa, The lower performance of the 24-stage pump is thought to result from leakage of the microvalves in the pump and non-ideal motions of the membrane that moves air around within the pump.

Simplified operation of a Knudsen pump that produces vacuum through thermal transpiration ...

The U-M team also developed a 48-stage Knudsen vacuum pump (PDF). The Knudsen pump is unusual in that it has no moving parts. Instead, it works on the principle of thermal transpiration. If you connect two gas-containing chambers through a channel, the gas molecules will move between the two chambers until the pressure in the system is constant. If one of the chambers is now heated while the other remains at its original temperature, the pressure in the hot chamber will increase.

Since the overall system still wants to be in mechanical equilibrium, some of the gas molecules move from the hot chamber into the cold chamber. If you then close a valve between the hot and cold chambers, and let the temperature of the hot chamber return to the original value, the pressure in the hot chamber will be less than before the heating-cooling cycle was performed. The pair of chambers and appropriate valves acts as a vacuum pump.

The U-M 48-stage Knudsen pump can reduce the pressure in a sample chamber from atmospheric pressure to less than 7 kPa, or to less than 1 kPa if followed by a backing pump. These compression ratios are about ten times larger than had been previously obtained. At 1.35 W, the power requirement is a bit larger than the DARPA requirement, but optimization of the thermal properties of the pump structure should significantly reduce the power required.

U-M’s high-vacuum pump was a microplasma sputter-ion pump, which embeds gas molecules in a material layer constantly being deposited during the operation of the pump. They were able to obtain vacuums down to 1 Pa, and showed evidence that the type of pump is capable of ultimate pressures smaller than 1 uPa.


The MIT engineering team also developed three-pumps, including a displacement roughing pump, a field ionization pump, and an electron impact ionization pump. The latter two pumps operate by using carbon nanotubes to provide large electric fields to ionize gas molecules in the pump, and then using electric fields to embed the ions into a surface, thereby removing them from the vacuum system.

A two-stage roughing pump with curved surfaces developed by a team of MIT researchers (MIT...

MIT’s displacement pump (PDF) uses curved pumping surfaces to obtain record compression ratios up to 4.6 per pumping stage. Given a system goal that the roughing pump provide a pressure of 4 kPa to support the high-vacuum pumps, this means that two stages of rough pumping should suffice, rather than the 24-stage pumps developed by the Michigan team.

Six operational steps of the MIT microfabricated displacement pump (Image: B. Dodson)

The visual above explains the operation of the MIT mechanical pump. It is essentially a piston-type pump, with active inlet and outlet valves. This pump was actuated and powered by pneumatic actuators, but in actual application would probably include electrostatic actuators for this purpose.


Honeywell may well have developed the most ambitious pump of the DARPA project, a micromachined version of a turbomolecular pump, which is like a turbine in reverse. Operating at pressures below about 1 kPa, this approach was probably chosen because of Honeywell’s past work on developing micro-scale gas turbine engines to provide electrical power.

A micro-scale turbomolecular pump developed by Honeywell can achieve high vacuum condition...

The heart of the Honeywell pump is an exquisitely detailed turbine rotor about 1.5 cm (0.6 in) in diameter. Very little quantitative information has emerged on this pump, but judging from the figure, the outer 2.5 mm (0.1 in) of the rotor is covered with tiny turbine blades, so angled that, as the rotor spins, the blades pick up air from the central region of the rotor, and throw it out the edge of the rotor. There appear to be about 2,000 rotor blades, each of which is about 60 microns wide, 60 microns tall, and 10 microns thick.

Although the Honeywell pump requires a roughing pump to reach operating levels of pressure, it has strengths and weaknesses that differ from those of other micro-scale vacuum pumps. It will probably find a solid position in the quiver of chip-scale vacuum systems.

“There have never been ionic or mechanical gas pumps at the microscale before,” says DARPA program manager Andrei Shkel. “The CSVMP program has demonstrated both and more. The smallest commercially available pumps are the size of a deck of cards, which dwarf the vacuum electronics and sensors we want to attach our pumps to. These pumps are not only 300 times smaller than off-the-shelf pumps and 20 times smaller than custom-built pumps, but they also consume approximately 10 times less power to evacuate from atmospheric pressure to milliTorr pressures.”

Although the CSVMP program was initially focused on developing microscale pumps for for better chemical and biological pathogen detection through small mass spectrometry gas analyzer applications, other potential applications for the technology became apparent.

“These microscale gas pumps may ultimately be required for laser-cooled atomic clocks, accelerometers and gyroscopes,” says Shkel. “Laser cooling systems require vacuums, but are often significantly smaller than the pumps themselves. It is possible that these pumps will help enable smaller, more accurate atomic clocks.”