Molecular Beam Epitaxy: Which Pump?

Faebian Bastiman

Invented in 1915 the diffusion pump is the king of UHV pumps. The principle is rather simple, a unidirectional jet of oil (yes oil, yes oil!) is sprayed inside the vacuum trapping particulates and sweeping them through the exhaust. Possessing low cost, no moving parts, a high pump rate for all gasses (up to 50kL/s) and capable of operating in the 10-2 to 10-10 mBar range they offer superior characteristics to other UHV pumps. Sadly they have two major disadvantages. Firstly, diffusion pumps passively back stream small quantities of oil (yes oil, yes oil!) into the vacuum chamber they are pumping. The amount of back stream oil is minimised by optimal pump design and employing traps, however for many users the idea of any oil vapour entering their vacuum system is tantamount to blasphemy. Secondly, diffusion pumps rely on fore pumping in the form of a rotary vane pump to create an outlet pressure of around 0.1 mBar. Failure of the fore pumping causes significant back streaming of oil (yes oil, yes oil!) into the vacuum system. Failure of water cooling results in a similar back streaming of oil. The absolute system wide contamination of several MBE reactors (due to failure of one of these systems) saw the popularity of the diffusion pump rapidly diminish in the early 1990s.

The cryogenic pump (cyro pump) is arguably the ultimate evolution of earlier sorption pumps. Both employ a cryogenically cooled interior coated with highly absorbing material to trap gasses. The sorption pump is cooled by submerging the pump’s housing in LN2. They were often employed as a roughing pump on earlier MBE system. Whilst cheap and reliable, sorption pumps are notoriously slow with a low pumping volume. Cyro pumps possess the same broad pumping range as the diffusion pump (10-2 to 10-10 mBar) however at severely reduced capacity (1-2k L/s depending on the material). They also possess significant mechanical displacement (usually by the means of a piston) that can induce vibrations within the MBE system. The cyro pumps cooling is achieved with a He compressor that doubles the cost of the cryo pumps. In multi system facilities, three individual cyro pumps can be run off a single compressor. Both cyro pump and compressor require a full service every 5-8 years. Furthermore, sorption and cyro pumps require regular regeneration owing to saturation of their absorbing material and hence neither can operate continuously. Compressor failure leads to a rapid increase in internal temperature and pressure, resulting in back streaming of molecules into the chamber. This is prevented via a dedicated gate valve, however with no exit the rapidly building internal pressure can cause a fracture of the pump housing and thoughtful implementation is needed to avoid damage.

The ion pump is perhaps the most iconic of all MBE pumps. Older ion pumps and their controllers required hefty initialisation currents (~18A) and lower but equally significant operation currents (~5A). Modern ion pumps and controllers require around 1A for continuous operation, meaning (i) they are very efficient and (ii) they can be backed up with a small, desktop uninterruptable power supply (UPS). Ion pumps can therefore safely hold a system’s vacuum in the event of a power failure. Uniquely ion pumps also operate as a pressure gauge (see MBE: Which Gauge?). Their pump capacity is however smaller than other UHV pumps (~200-500 L/s depending on the pump dimensions) and they require a start pressure around 10-5 mBar however they can create an ultimate pressure close to 10-11 mBar. Note that the low pump capacity means that this low pressure is not possible in the growth chamber, however can readily be achieved in the preparation or buffer chambers. Ion pump have the advantage that they are clean, need little maintenance, produce no vibrations and unlike the cryo and turbo pump they can effectively pump H2 and He. Hence they are perfect for standalone pumping of a preparation or buffer chamber, however they must be combined with a cyro pump and/or a turbo pump to achieve the UHV condition necessary for low background doping material.

Turbo molecular pumps (or simply turbo pumps) come in all shapes and sizes and can be employed for a variety of applications. Like the diffusion pump they require fore pumping in the form of a roughing pump on their exhaust in order to operate. The roughing pump can take various forms (rotary vane, diaphragm or dry scroll) though in the interest of maximum pump capacity and low contamination the scroll pump is the roughing pump of choice. The turbo and scroll combination handily spans the atmospheric pressure to HV to UHV range. Turbo pump capacities range from 70 L/s to 4.5k L/s depending on the pump dimensions and the bore of the flange they are connected through. Smaller capacity pumps are employed to regenerate cryo pumps and pump down fast entry locks (FELs), whereas the large capacity pumps help the ion and/or cyro pump on an MBE’s growth chamber. Turbo pumps are “through” pumps with similarities in operation to a jet engine, unlike ion and cryo pumps which are “trap” pumps. Some applications, like phosphorus, require the through pump capabilities of a turbo pump along with additional phosphorus trapping capabilities. Since the turbo pump operates by accelerating molecules the pump’s efficiency is directly proportional to the molecules mass.

Titanium sublimation pumps (TSPs) operate in a similar manner to cyro and sorption pumps. In the case of TSPs the adsorbing material is a thin film of highly reactive titanium that is evaporated by supplying 40-50A through a titanium filament and the condensing surface is the MBE reactors chamber wall. Once saturated, the film can be re applied with a further current heating cycle. Pumping effectiveness is also increased by cooling the chamber walls with LN2 or positioning the TSP near the main LN2 cyro panel. Since titanium is particularly reactive with CO, H2O and O2, it can significantly reduce background doping and hence it is recommended to fire the TSP particularly prior to oxide removal and during growth of high mobility materials.

The cyro panel is a compulsory constituent of all high quality MBE systems. LN2 (77K) is by far the most common cooling medium and MBE systems typically consume 150 – 200 L/day when the LN2 is delivered from an efficient combination of high vacuum lines and a phase separator (see MBE: LN2 system). However some users argue that a custom chiller until pumping silicone oil (193K) or antifreeze (233K) is perfectly viable in certain situations.

So which pump? The cyropanel and TSP form a very effective secondary pumping combination and should always be present. As for the main pump… honestly I would like to see some new development go into the design and auxiliary services of the diffusion pump. Redundant fore pump configurations employing a team of pumps and triple backing up the water cooling (see MBE: water cooling system) along with suitable interlocks could make diffusion pumps fool proof. Resurgence, of course, relies on consumer confidence. Sadly the alternative is to employ all 3 of the remaining pumps (cyro, ion and turbo) in order to achieve even close to the same pumping capacity. This triple combo easily increases the pumping cost by a factor of 5x and of course occupies a great footprint and imposes other run cost and maintenance overheads. Of course the more pumps the better the vacuum, however most R&D reactors can be adequately pumped with a combination of 2 of the 3 pumps on the growth chamber. The exact combination depends on what you are pumping: for As systems an ion pump and cryo pump make a good combination, for N plasma systems two cyropumps make a good combination, for P systems an ion pump and a turbo (with a P trap) are needed. For further information, a simple R&D pump configuration can be found in MBE Design and Build: Vacuum System.


2 thoughts on “Molecular Beam Epitaxy: Which Pump?

    • The ability to fire the TSP during growth depends on the machine configuration. Most systems (including the V80 and V90) aim the TSP filaments directly at the cryo panel with no clear path to the substrate. The titanium atoms are highly reactive and essentially it is impossible from them to undergo enough “bounces” to make it to the sample before they adsorb onto the chamber walls. And there they sit awaiting any O2, CO and H2O to come along and adsorb. The reactive film can last several hours before it is saturated in these background dopants and needs replenishing. Firing the TSP is therefore a very good idea after oxide remove or immediately prior to the growth of high mobility layers and indeed is essential common practice for ultra high mobility AlGaAs.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s