Molecular Beam Epitaxy: Run costs

Faebian Bastiman

New and updated figures February 2014

The run costs of an MBE system need not be astronomical, however like any system time and effort must be spent tuning it to efficient, economical operation. The MBE system costs can be separated into set up costs (see Molecular Beam Epitaxy: Initial Outlay) and the run costs discussed here. I will discuss the prices in GBP (£) since my current system is based in the UK. The day-by-day, year-by-year operation requires a number of consumables:

  1. LN2 (or alternative cooling)
  2. Compressed N2 or air
  3. Substrates
  4. Cell materials
  5. Cell crucibles
  6. Copper gaskets
  7. Electricity
  8. Miscellaneous maintenance items

Liquid nitrogen cooling cost: £12.4k per annum

Some form of cryogenic cooling is essential to MBE, though it need not be LN2 specifically. The alternatives are discussed in Molecular Beam Epitaxy: Initial Outlay. Assuming you decided to follow the “spend more now, save more later” philosophy in Molecular Beam Epitaxy: LN2 system and bought an efficient LN2 system from VBS, decoupled your liquid and gas generation sub-systems and that your system is 2” capable MBE system with suitable heat shielding you will have a MBE system LN2 requirement of 200 L/day with an associated “heat loss” of 10%. At £0.11 per litre, LN2 is pretty cheap. Based on this you will use ~£24 a day or £8.8k per annum. However to obtain this low LN2 usage level you need to have a fully vacuum insulated line from the tank to the phase separator and from the phase separator to and from the MBE system. High integrity vacuum insulation requires constant regeneration, so you will need to run a dedicated pump all day every day, which will cost you an extra £1.6k per annum (pumping cost courtesy of Bart Limpens, VBS) . You will also probably be renting your main LN2 tank, so you need to add another £2k per annum. Note a poorly designed LN2 system can cost more than twice as much!  

Gas system: £1.4k per annum

Pressurised gas is necessary in two forms: a 5 bar pneumatic line and a 1.5 bar venting line. Again in Molecular Beam Epitaxy: Initial Outlay the pressurised gas choices were discussed. Assuming you decided to follow the example in Molecular Beam Epitaxy: N2/Gas system and divide your demand into 5 bar and 1.5 bar compressed air (from a compressor) and 1.5 bar N2 (from a convenient and efficient boil off system using a dedicated tank), you will have an electrical run cost of £1k per annum to run the compressor and 10L (£400) of LN2 per day in producing gas, sadly most of which is flash loss. The actual gas has a volume nearly x100 that of the liquid. So the actual gas used is very small compared to the cost of generating that gas. This is a strong argument for NOT using boil off. Of course boil off is very convenient, but a convenience that will cost you £1.4k per annum.

Substrates: £7k per annum

Substrates can be very cheap or very expensive depending on what you are growing. Si substrates are notoriously cheap, and whilst the industry is phasing out 2” lightly doped Si substrates, you can still buy them for around £8 each. 2” GaAs substrates are around £50 each. 2” InP or InAs substrates are decidedly more expensive: £125 each. I will use 2” GaAs from now on by way of example. Handily you can do two things to save some money. Firstly, III-V substrates can be repolished (35reclaim) for around £10 – £20 each depending on the material. My experience with these substrates is that they are equal in quality to “new” epiready ones. Secondly, you can ¼ your substrates to effectively get 4 growths out of a single substrate at the cost of reduced sample area. Unfortunately you cannot do both, i.e. you cannot repolish a quarter. You can however repolish a repolished wafer several times. The choice, in the end, is whether you want to grow on full wafers or not. But bear in mind that a ¼ of a 3” wafer is about the same area as a whole 2” wafer at a fraction of the price. Of course the annual cost depends on how many samples you grow a day. Two per day is typical for a 280 day working year, or £7 per annum assuming you decided to grow on ¼ of 2″ substrates.

Cell material: £2k per annum

Cell material is not actually as expensive as you may think; check out the latest prices SciTech Solutions has to offer. You will want 7N material (that is 99.99999% pure) as standard. Though in many research applications 6N5 (99.99995) is perfectly suitable. The table below lists some experimentally gathered usage versus cost values. All materials are 7N aside from the Al which is 6N5. 7N Al is about x10 the cost.

Usage Actual material
Material £/g g/micron £/micron microns grams cost (£)
Al (6N5) 3 0.33 1.0000 60 20 60
Ga 2.2 0.17 0.3667 6000 1000 2200
In 1 0.17 0.1667 6000 1000 1000
As 2.29 0.33 0.7639 7200 2400 5500
Bi 1.7 0.33 0.5667 3 1 1.7
Si 10 0.00 0.0007 14400 1 10
Be 3500 0.00 0.2431 14400 1 3500

To help you interpret the table: The far right column (“cost (£)”) is the cost for the amount of material listed in the adjacent “Grams” column. Note: minimum order quantities apply. The “microns” column is an estimate of the number of microns grown with that number of grams. The left-hand columns are calculations based on the values in the right-hand columns. Note that Be is very expensive but also a gram will last 20+ years. Si is cheap, since it is essentially smashed up substrates. As is a rather expensive one of cost for a 2.4 kg charge, but that single charge should last 5 ‑ 15 years depending on individual usage. This means for an initial outlay of £20k you can grow for around 10 years, or £2k per annum.

Crucibles: £3k per annum

The cell crucibles are actually only £600 each from SciTech Solutions. Handily most of them are also reusable. Once they have been soaked in AQUA REGIA for a few days to remove any material, they can be rinsed in DI water, then IPA, blown dry with N2 and baked to 100°C in a little desktop oven. Al and Bi can very easily shatter a crucible so you are better off using a crucible liner and employing “good filling policy” (see Molecular Beam Epitaxy: Crucible Cracking). All in all you may spend up to £3k per annum on crucibles/liners per year, and that is if you are very unfortunate.

Gaskets:  £250 per annum

Next we have gaskets and again SciTech Solutions have the lowest prices around. UHV systems use conflat (CF) flanges that comprise hard stainless steel (SS) blades that bite into a soft copper gasket. The gaskets are then effectively consumed once used and would need to be melted down and reforged. A standard system service involves replenishing depleted cells, cleaning/replacing shutter blades, cleaning the manipulator of excess As, and perhaps replacing an ion gauge or TSP filament. Every component requires a replacement gasket and it is very easy to get through a few kgs of copper per annum. This is around £50 of copper. However the actual gaskets cost around £250 per annum. It is good practice to recycle your copper to recuperate costs.

Electricity: £13k per annum

Electricity is something you may or may not have to pay yourself, depending on how your research is funded. The bottom line is someone, somewhere will have to pay. An MBE system typically runs off a 64A, 3 phase AC supply. You can find some more detailed calculations in Molecular Beam Epitaxy: AC power but mostly it runs at about ~50% capacity during operation, at 25% idling over-night, perhaps 70% when baking but never at 100%. Applying some calculation based on general 280 day usage your electric bill for an MBE machine is therefore: £13k per annum. Note the essential parts of the system should be backed up on a suitable uninterruptable power supply and generator team (£10k), this may seem an expensive initial outlay but I can guarantee it will save you a cell per power failure. Which means it will pay for itself after the first power cut and hence significantly reduce your “unexpected maintenance” bill (see below)

General rainy day fund: £10k per annum

Finally you have general maintenance. Sadly UHV metals are rather rare and very hard (which makes them difficult and expensive to machine). The typical metals are stainless steel in grades A2 (304) or A4 (316), molybdenum, tantalum, niobium, some aluminium (as long as it is kept away from anything approaching 660°C!), copper (which is slowly eaten by arsenic), silver and (possibly) gold. There are various parts than can fail. Commonly Ga or In can run into places where they are not wanted and cause short circuits to electrical feedthroughs or high friction to rotary feedthroughs. In addition anything electrical: cells, manipulator, ion gauge, TSP, pumps etc requires annual servicing and minor replacement of parts. It is good to keep around £10k per annum simply for unexpected repairs. Of course the cost of repairs can be reduced if you are confident and able to do it yourself.

Total: £50k

The small system I have described here would cost £50k per annum in run costs for 2 samples a day or £90 per sample (however this obviously does not include salaries). Of course if you grow 4 samples a day, your run cost goes up to £60k per annum, but your sample cost drops to £54 per sample. This probably represents the lowest annual run cost for a productive MBE system. Remember that in order to reduce your run costs you will need to buy more infrastructure during the initial outlay. It is always beneficial to do so, since the outlay is typically recuperated within 1 – 2 years.


Molecular beam epitaxy: Systems and sub-systems

Faebian Bastiman

In essence an MBE system comprises:

  1. A vacuum system
    1. Pumps
    2. Valves
    3. Gauges
    4. Mass spectroscopy
    5. LN2
    6. Pneumatics (5 bar N2)
    7. Venting (1.5 bar N2)
  2. A deposition system
    1. A substrate heating and manipulation system
    2. Cells
    3. Shutters
    4. Temperature monitors
    5. Power supplies
    6. Water cooling
  3. A control, monitoring and/or interlock system
  4. A RHEED system
  5. A bakeout system

The vacuum system is largely set at the time of manufacture and post-production customisation is only necessary during upgrading beyond the original specification. The control system is a matter of personal preference, and the author’s endorsement of the Epimax suite is borne from a desire to possess a system that can be implemented, expanded and adapted at will. Staib produce a very reliable and robust RHEED system and kSpace have developed a powerful data acquisition suite. The bakeout system also comes down to personal preference. Bakeout jackets, tents or panels (boxes) with an array of heaters are available. Jackets are highly desirable, as they can be employed with minimal decommissioning of the system, but they are notably expensive or nonexistent. Tents and panels utilising “filament-and-fan” style heaters are the standard means of baking, and aside from the lengthy pre-bake preparation are perfectly adequate.

The deposition system (and its integration into the control system) is by far the most complicated system. Like all systems, it is necessary to break it down into its sub-systems in order to understand the function and interaction of the separate elements. The substrate heater and cells are essential identical. They comprise a heater filament (Ta, W, SiC or C-track) and a thermocouple (typically C or K type). They can be powered with either AC (using a thyristor) or DC (using a suitable power supply) and are maintained stable to within <0.1°C by a proportional, intergral and derivative (PID) loop usually resident in the temperature controller. The P, I and D values need to be carefully tune in order to achieve stability and suitable dynamic response, luckily all modern controllers have suitable automatic tuning algorithms. Water cooling is used to protect and equilibrate the delicate innards of the cell/heater. The cooling can be either integral or auxiliary to the cell. A mixture of water and glycol similar to that utilised in a car’s coolant system is preferred. The shutters too are either integral to the cell or separate and hence integral to the MBE system. They are typically actuated by a pneumatic or stepper motor driven mechanism. For automatic growth recipe execution it is of course necessary to manipulate the shutter state through the computer resident control software.

The actual sample manipulation, from the epiready supplier boxes to the growth chamber’s heater stage, is an area of serious neglect in the MBE world. Manipulating a sample from the start to the end of the journey is often the most complicated task involved in MBE operation, particularly on older system. A number of mechanisms are available, and until recently only Vacuum Generators (VG) had engineered a user-friendly system. The VG system employed on V90 and larger systems utilises optical sensors to identify when the sample is in a valid transfer state. The sample transfer can be conducted manually using switches, automatically from computer designated site to site or even as part of a 6 sample batch without any operator interaction. The system was literally decades ahead of its time, though in many ways also long overdue. More recently the “cluster” tool has been engineered in favour of the “linear” MBE system. The cluster tool comprises a central distribution chamber with a computer controlled arm for sample transfer. Six to eight discrete chambers can be installed on the cluster chamber’s side flanges with the possibility for a FEL chamber, outgas stage, park/storage chamber and 3 independent growth/deposition chambers. The cluster tool employs either optical sensor or “intelligent” stepper motor control for position determination. The bells and whistles of the cluster tool, however, come with a hefty price tag.  In terms of economics and pure practicality the VG V90 is easily the best research/semi-production MBE reactor ever manufactured. Though, unfortunately, since the dissolution of Vacuum Generators the V90 systems are now discontinued and, in the author’s opinion, we are stuck with either very expensive or inferior alternatives.