Essential maintenance: Leak Detection

Faebian Bastiman

After performing routine maintenance (Essential Maintenance: The Maintenance Cycle) and pumping down your system to the mid-high 10-8 mBar range (Essential Maintenance: Pump down) you are almost ready to bake. Almost!

First of all you will need to fire up your trusty quadrupole mass spectrometer (QMS) or residual gas analyser (RGA) and sample the vacuum. Here I recommend the (Pfeiffer Prisma) though in this post I will use its predecessor the trail blazing Balzers Prisma. Set up your analyser as per the user manual and start a 100 atomic mass unit (amu) scan. You will (hopefully) obtain a scan similar to that shown in Figure 1.

 RGA spectra

Figure 1: Pre-bake out RGA scan of a III-As system

What are all those little peaks telling us? Well they represent the chemical fingerprint of the vacuum. Each element is easily recognisable by its individual mass number; however compounds require some knowledge of the prominent peaks and the relative intensity. To do that you can use the values in Table 1.

QMS/RGA cracking patterns
Name Formula Mass peak 1 peak 2 peak 3 peak 4 rel sens
m/z % m/z % m/z % m/z %
1 acetone C3H6O 58 43 100 58 33 15 20 42 6 3.6
2 Acetylene C2H2 26 26 100 25 20 24 6 13 6 1
3 air 28 100 32 27 14 6 16 3 1
4 ammonia NH3 17 17 100 16 80 15 8 14 2 1.3
5 argon Ar 40 40 100 20 16 36 1 38 1 1.2
6 benzene C6H6 78 78 100 77 19 52 16 51 15 5.9
7 boron trichloride BCl3 117 81 100 83 64.7 35 28.7 1
8 carbon dioxide CO2 44 44 100 16 9 28 8 12 7 1.4
9 carbon monoxide CO 28 28 100 12 5 16 2 14 1 1.05
10 carbon tetrafluoride CF4 88 69 100 50 12 19 7 31 5 1
11 diborane B2H6 27 26 100 27 97.4 24 89.6 1
12 DP oil DC 705 78 100 76 83 39 73 43 58 1
13 DP oil FOMBLIN 69 100 20 28 16 16 31 9 1
14 DP oil PPE 446 446 100 77 79 51 29 39 10 1
15 ethane C2H6 30 28 100 27 33 30 26 26 23 2.6
16 ethanol C2H5OH 46 31 100 45 39 27 24 29 24 3.6
17 ethylene C2H4 28 28 100 27 63 26 61 25 11 3.6
18 Fomblin oil 69 100 20 28 16 16 1
19 Freon 12 CCl2F2 104 85 100 87 32 50 16 2.7
20 Freon 13 CClF3 104 69 100 85 28 35 15 50 15 1
21 helium He 4 4 100 0.14
22 hydrogen H2 2 2 100 1 10 0.44
23 hydrogen chloride HCl 36 36 100 38 32 35 17 37 5 1.6
24 hydrogen sulphide H2S 34 34 100 32 44 33 42 1 5 2.2
25 isopropyl alcohol C3H7OH 60 45 100 43 17 27 16 29 10 1
26 krypton Kr 84 84 100 86 31 82 21 83 21 1.7
27 methane CH4 16 16 100 15 85.8 14 15.6 13 8 1.6
28 methanol CH3OH 32 31 100 32 66.7 29 64.7 28 6 1.8
29 neon Ne 20 20 100 22 9.9 21 0.3 0.23
30 nitrogen N2 28 28 100 14 5 29 1 1
31 oxygen O2 32 32 100 16 11.4  34 0.4 1
32 phosphine PH3 34 34 100 33 33.1 31 32.1 2 1 2.6
33 pump oil 500 57 100 43 73.3 55 72.7 41 33 1
34 silane SiH4 32 30 100 31 78 29 29 28 27 1
35 silicon tetrafluoride SiF4 104 85 100 86 5.2 28 4 33 4 1
36 sulphur dioxide SO2 64 64 100 48 50 32 10 16 5 2.1
37 Trichloroethane C2H3Cl3 132 97 100 99 64 61 58 26 31 1
38 Trichloroethylene C2HCl3 130 95 100 130 89 132 85 60 65 1
39 water H2O 18 18 100 17 21 16 2 1
40 xenon Xe 132 132 100 129 98 131 79 134 39 3

Table 1: QMS/RGA cracking patterns for common MBE vacuum species

Before we go into detail analysing the spectra, the first thing to do is to check the most prominent peak. The peak at 18 is almost an order of magnitude higher than the peaks around 2 and 28 and only the peaks around 16, 17 and 91 come close. From table 1 we can deduce that the family of peaks at 16, 17 and 18 are in fact water. Since we have just opened our system to atmosphere we would naturally expect water to be the most abundant species in the system. What about the other peaks?

Analysing the spectra in figure 1 from left to right, we can first of all see the prominent H and H2 peaks that are caused by the stainless steel (SS) outgassing into the vacuum. The exact height of these hydrogen peaks varies for system to system and largely relate to how the MBE system was heat treated when it came off the production line. Often this involves heating the system to 250°C for 400 hours then 350°C for 120 hours causing the SS to take on a “golden” tinge. After the water peaks discussed above, the next peak is at 28 and corresponds to N2, as is to be expected from the multiple N2 flushes we used during the pump down. Notice that the O2 peak at 32 amu is very low, but together with the peaks at 28 and 14 we can say there is some air in the vacuum chamber. But how much air do we have in the chamber? Do we have a leak? Well let me finish the spectra analysis first and then I will answer those questions… the peaks at 37.5 and 45.5. are probably double ionised As and AsO (thanks to Scott for his comment), and the peaks at 75 and 91 are the single ionised As and AsO (as to be expected on a III-As system). Some of the As has also formed arsine-type compounds AsH and AsH3 with masses of 76 and 78 respectively. The peak at 44 is CO2 which was soluable in the water vapour at atomspheric pressure and is now dissociating. CO2 has a corresponding peak at 12 and adds to the peaks at 16 and 28. The other small peaks at 15, 19, 20, 25, 26 and 30 are various simple hydrocarbons formed from the interaction of the numerous C and H ions. All in all a very clean pre-bake vacuum.

Back to the oxide: Well from table 1 we can see that air has an O2 peak that is 27% of the N2 peak in the “air” spectrum. Remember figure 1 is on a log scale so we would expect the O2 peak to be much higher (~2.4 x 10-10 A) for a N2 peak at 9 x 10-10 A to represent purely air. Clearly the O2 peak is at around 5 x 10-11 A and thus the N2 peak is formed predominantly from pure N2 (and 8% CO2) gas. From the relative peak heights we can estimate there is in fact roughly 5x as much N2 as there is air. The air is actually two orders of magnitude lower than the water vapour, so we can conclude that there is no significant leak, at least no leak above 10-10 mbar.

What if the situation were reversed? If the air peak had been dominant and the O2 peak had been in the mid 10-10 A range, we could only conclude we had a small leak in the 10-9 mbar range. If the background pressure had been in the low 10-7 mbar range and the “air” spectra was an order of magnitude higher, we could only conclude we had a more significant leak in the 10-8 mbar range. What to do?

Well first of all do some simple checks and work through a process of elimination. Check the pump exhaust fittings, particularly the turbo pump to roughing pump connections and the TSP ceramics. Close the various gate valves and try to determine on which chamber the leak is present. Spray some IPA onto any suspicious looking flanges (where the gasket looks unevenly compressed) and look for pressure and current spikes in the ion gauge and RGA spectra. You can find a guide that details the flange nut sequence required for leak free flange in MBE maintenance: how to tighten a CF flange. If none of these work you will need to utilise He leak detection. To do this set the RGA to 4 amu for He detection only, turn off your air conditioning so it does not waft the He around the MBE system and with a narrow aperture pipe and very low flow rate of He systematically spray every area of the machine. Once you have identified a likely origin of the leak wrap a plastic bag around it and test both inside and outside the bag to see if you can “enhance” and “supress” the He signature.

In summary, because the water vapour pressure is higher than the N2, and the N2 is higher than the “air” and because our pressure is in the high 10-8 mBar range we can conclude that this system has no significant leak. After baking to 140°C for 48 hours (plus 8 hour ramp up and ramp down times) we get the RGA spectra shown in Figure 2. Note this was immediately after baking, whilst the system was at around 40°C with no water or LN2 cooling attached. The ion gauge pressure was in the high 10-10 mBar range.

post bake crop

During the baking the water vapour rapidly evaporated and was swept out through the pumps with the other temperature assisted highly mobile species. The result is a spectra that is almost identical to Figure 1, except that each peak is reduced by 1-2 orders of magnitude. So we cannot conclude that  the hydro-carbons and oxygen are “gone”, they are simply beneath the sensitivity of the detector (that is to say they are now in very dilute quantities). What do we have left in the vacuum? Well the most dominant peak is now the H2 (2) closely followed by As (75, 37.5) and water (18, 17, 16) then N2 (28),  CO2 (44, 28, 16, 12) and AsO (91). On the one hand there is an argument for baking the system for longer, since every additional day at high temperature would see more of the contaminants removed from the system.  On the other hand since the rate of removal of contaminants falls exponentially with time the “time-purity trade off” results in a most efficient bake out time between 16 to 72 hours. The system will continue to “clean up” through the first week of operation and the contaminants will experience another 1-2 orders of magnitude reduction once the LN2 cooling is turned on. And so now it is time to perform the many little post-bake tasks that get the system ready to grow.

Essential maintenance: how to tighten a CF flange

Faebian Bastiman

Earlier this week I was asked how to tighten a CF-40 flange. The implied but unvoiced condition being: and ensure that the Cu gasket is evenly clamped thus creating a leak-free seal. I happily replied with the adage from my early MBE days: “do one miss one” in regard to the order to tighten the nuts, however realised as I said it that when taken literally one would only ever tighten 3 of the 6 nuts. The more correct but slightly longer adage would seem to be: “Do one miss one, until you land back where you started…when you do, do not do that one but instead the next one, then do one miss one in the opposite direction and try to ensure even loading by only increasing the tightness by ¼ a turn on each nut” Needless to say this is expressed more simply in a diagram, and so I direct you attention to figure 1. Once you reach nut 6, restart again from nut 1 and continue until the copper of the gasket is just visible between the flanges (~1.5 mm) or for those of you with a torque wrench set it to 20 Nm.


Figure 1: CF-40 flange nut tightening sequence

However that made me think: what is the correct sequence to use on other flanges? Well both a CF-16 and CF-40 flanges have 6 nuts so the sequence for both is shown in figure 1. Generalising the sequence you would come up with a do one miss (n/3) -1, where n is the total number of nuts on the flange. Or put a different way, tighten every (n/3)th nut. Great (!) but what about when the number of nuts is not divisible by 3? Like on a CF-63, 100 or 160 flange? Hmm. Well the whole point of the sequence is to evenly bite the gasket. So for any flange the “triangular-reverse-direction” loading would seem logical. Hence for a CF-63 flange you would use the sequence in figure 2 and for a CF-100 the sequence in figure 3. This same selection rules can be applied to CF-160 flanges.


Figure 2: CF-63 flange nut tightening sequence


Figure 3: CF-100 flange nut tightening sequence

Once you reach CF-200 you have 24 holes (n/3 = 8) and the sequence is a multi-pass version of the original CF-40 flange shown in figure 4. Here I have made the first half of the sequence green and the second half black so you can see the pattern is still essentially “do one miss one” at the halfway point.

flange tighten order

Figure 4: CF-100 flange nut tightening sequence

Armed with these sequences and a good set of spanners I wish you a leak free maintenance cycle.

Essential maintenance: Cryo pumps

Faebian Bastiman

Cryo pumps and ion pumps make an exceptionally good team for III-As MBE. Whilst the ion pump stoically and gradually maintains the UHV vacuum, cyro pumps provide that little extra pumping capacity to bring you from “good” high -10 mBar to “great” low -10 mbar. However the pumping capacity of a cryo pump degenerates with time, making them less and less effective. In fact from the moment they reach minimum temperature (~10K) every following moment sees them becoming less effective until their internal absorbing surfaces finally saturate. The efficient operating time will depend on the cryo pump volume and how aggressively you operate your MBE system. In a typical R&D III-As system the efficient operating cycle is usually 2-3 months.

But how to regenerate a cryo pump?

Well first of all you will need to shut the gate valve that connects the cryo pump to the growth chamber. After that you are free to turn off the compressor and power down the pump. However remember in so doing the pump will rapidly heat up and the internal pressure can exceed atmosphere by over 100 times. Clearly leaving the pump sealed on both ends is not a good idea. The pressure increase is of course not instantaneous, and happens more readily above 77K when N2 gas starts evaporating.

The next step involves connecting a dry scroll pump to the exhaust and opening the exhaust valve once the cyro pump reaches ~80K. The cryo pump can then be left in this state for several hours until the internal temperature equilibrates to the ambient. At this temperature most of the contaminants are highly volatile; however water and some of the other cryo absorbed matter may need a little more encouragement. The final step in regeneration is achieved by using heated dry N2 gas to raise the cryo pumps internal temperature to ~60°C. Indeed most cryo pumps have a double-exhaust to enable this function.

In order to perform this final step you will need (i) ultra-pure N2 from a cylinder (ii) a fully baked, ultra clean SS line from the cylinder to the cyro pump and (iii) a heat source. Since you are baking the line anyway, you can simply install heat wraps or tape on the line to heat the pipe to ~100°C and in so doing this will also heat the N2 passing through the pipe. A more elegant (and expensive) solution is to utilise an ultra-pure 316L SS inline N2 gas heater (from Heateflex). The N2 needs applying constantly during the ramp up and ramp down steps, with a hold time of about an hour.

Once the N2 heating cycle is completed the dry scroll pump can be operated to pull the cryo pump down to ~10-3 mBar. Whilst the cyro pump can be operated from this base pressure , you can significantly reduce early stage contamination by employing a turbo pumping station (like the Pfeiffer Hi-cube Classic N.B. avoid the Hi-cube Eco which does not have sufficient pump rate for this application) to bring the interior pressure down to ~10-7 mBar (N.B. It is the size of the roughing pump and not the turbo molecular pump that is the limiting factor).

A schematic of the regeneration system is shown below. Note the inclusion of the 3 way bypass valves for the pumping station to avoid flushing it with heated N2 gas during the final regeneration step. It is also possible to include an optional wide range gauge (WRG) on the cryo exhaust to monitor the internal pressure whilst refining the regeneration cycle’s timing.

Cryo pump regeration system

Then simply seal the exhaust, restart the compressor and wait for the internal temperature to reach ~10K. Once done you are free to open the gate valve and continue normal system operation. The entire regeneration cycle may take 5-6 hours. So rather than lose an entire day it is better to schedule it over night or over a weekend. How can you do this? Why by automating the process with the MBE control software of course. What is that you say? “Your software does not have this feature”. Why not?

Essential maintenance: Cell material regeneration

Faebian Bastiman

A simple means of ensuring the purity of your cells’ source material is to include a mini regeneration cycle at the start of each day. Under some circumstances you may operate your MBE system 24/7 and never actually cool down your sources. Indeed in the ideal world of batch processing and in situ growth monitoring with a dream machine (see MBE: Dream machine) an MBE system could be run non-stop for 3 months. However even under this utopia of operation you do not need all the sources to be hot all of the time, doing so is simply too wasteful.

When you cool down a cell, any impurities in the system can condense on the shutter, cell divider, water-cooling jacket and worse still alloy with the source material. Simply reheating your cell to operating temperature and growing will result in the cell outgassing onto your sample at the start of the growth. To avoid this you simply need to “over heat” the cell by ~25°C for half an hour at the start of each day. The impurities will outgas more readily at the higher temperature, and upon cooling to operating temperature the source will be ultra-pure once more. This extra step also has the added bonus of reducing the settling time between the cell’s thermocouple registering the operating temperature and the cell’s source material equilibrating at the desired flux.

One’s first thought is: “this is going to take up my valuable MBE time”. Quench that thought immediately! Consider instead the ability to set a “wake up” time for the MBE system at say 0700. The system will then execute a recipe to ramp all the cells to outgas temperature, outgas all the cells for 30 minutes and then ramp them down to operating temperature. At around 0800 the system can then start a flux check on each of the cells and tune the beam equivalent pressure (BEP) to the desired values. At around 0845 the auto flux tuning is complete and the machine sits waiting patiently for you to grow your first sample. My software does this, what you need to ask yourself is: why doesn’t yours?

Essential Maintenance: Pump down

Faebian Bastiman

After your maintenance cycle the time has come to pump the system down and start recovering the ultra-high vacuum (UHV) conditions necessary for MBE operation. In principle this simply involves turning on the pumps, but you can perform a few specific steps to greatly improve your pre-bake base pressure.

After maintenance your system has likely been exposed to atmospheric pressure for several hours. Water vapour contamination is unavoidable, though of course can be reduced by situating your MBE system in an air conditioned laboratory. Once the last flange is sealed (see Essential Maintenance: A how to guide for tips on MBE maintenance) the system interior is completely isolated with an air tight seal. Completely isolated from atmosphere, but sadly still full of atmosphere.

The first step is to pump that out. Dry scroll pumps should be used throughout on an MBE system. They have an excellent pumping rate and virtually zero possibility of contamination. The Edward’s nXDS is a good choice. Enable the ballast valve to enhance the pump’s handling of moist atmosphere and pump through until the Super Bee pressure gauge on the scroll pump’s inlet reaches 0.05 mBar. Now turn off the pump and vent the MBE system with N2.

What?! We have just pumped the system down and now you want me to vent it again? Are you sure? Yes I do. You just pumped out moist atmosphere from an air tight system, and now you are going to vent with dry, ultra pure N2. What will happen? Evaporation.

Evaporation is a wonderfully useful natural process for an MBE operator. Evaporation will transport moisture from the chamber moist inner surfaces into the dry N2 gas. Not all, but a significant fraction. The evaporation rate will depend on:

  1. The concentration of water in the N2: which is zero at the beginning
  2. Flow rate: initially high as the N2 penetrates the vacuum, but constatly dropping
  3. Pressure: evaporation happens faster at lower pressure
  4. Temperature: evaporation happens faster at higher temperature

Simply let the N2 in until the Super Bee reads 500 mBar (remember the process happens faster at lower pressure). Wait a minute. Then pump the chamber down with the scroll pump again. Once it has reached 0.05 mBar, turn the pump off and flush it with dry N2 again. On the second pump down, open the As and P cracker’s needle valve manually a whole turn. The bulk should have been protected from atmosphere when you shut the valve during Essential Maintenance: System Venting. It should therefore be free from moisture, but there is always a chance the valve has a little leak. To be safe, we will pump down the bulk with the scroll pump now.  To increase the effectiveness of the N2 flush you may consider heating the N2 gas to ~50°C. This can be highly regulated with an inline gas heater or poorly regulated by baking the SS N2 pipework to 50-70°C with heat wraps during the flushing phase.

Repeat whole the pump down and vent  process 5 times.

Finally once the system has been pumped down for the last time, start up your turbo pump. After 15 minutes, turn on your ion gauge to 0.1mA emission and check your pressure. It should be in the 10-6 mBar range. Having used the maintenance cycle as an opportunity to regenerate your cyro pump, it should now be operating and ready to take over too. Open the valve to the cryo and let it assist the turbo for a further hour. With the turbo and cyro pumping the growth chamber the pressure will quickly drop into the 10-7 mBar range. Then open the gate valve to the ion pump (which was never turned off) and leave the system over night.

What you should find is that in the morning the background pressure is ~5.0 x 10-8 mBar. That is a very impressive pressure before a bakeout. This is in fact at least 10x lower than without the repeated N2 purge step. If your system is not in the mid-high 10-8 mBar range you will need to leak test. Leak test either with an ex situ external leak tester or an in situ quadrupole mass analyser. Follow the instructions in Essential Maintenance: Leak Detection to identify the cause of the leak.

Once the pressure is <10-7 mBar you are ready to start powering up all the sub-systems you turned off in order to vent. First, turn on all your ion gauges to 0.1mA emission current. Next, double check all the PID controllers for your effusion sources are set to lower than ambient temperature, then turn all the PSUs on. Current limit all the PSUs to 2A to protect them during the effusion cell filament’s initial ramp up from cold. With a ramp rate of 0.1 °C /s ramp the sources (with the exception of the As bulk) to 100°C. This will drive off moisture. The pressure will inevitably rise for a few hours until the moisture is purged through the pumps. At the same time you can heat your substrate heater to 100°C too.

You are now ready to bake the system (see Essential Maintenance: Bakeout).

Essential maintenance: System venting

Faebian Bastiman When the time has come to perform maintenance on your MBE system you will need to raise the internal pressure to atmospheric pressure. You can vent the system at several points:

  1. Through the fast entry lock (FEL)
  2. Through a turbo molecular pump
  3. Through a cryopump
  4. Through a dedicated venting valve

Before you start to vent, take a moment to consider where you need to vent. Do you need to vent the growth chamber and the preparation? Or do you need to only vent the growth chamber? When venting through the FEL you need to vent the entire system in series with the FEL-preparation and preparation-growth chamber valves open? Alternatively you can vent through a pump or dedicated value on the growth chamber.  Most modern turbos support a venting feature. Pfeiffer’s certainly possess this feature. You need to selected the correct menu options, turn off the backing line scroll pump and once the speed drops below a pre-described value the vent process starts automatically. The cyro pump vent requires a little more effort: First you need to turned it off and allow it to reach room temperature, all the while pumping it down with a scroll pump (regeneration), and after that a small turbo or Pfeiffer Hi-cube to get a good base pressure. Only then can you open the gate valve that connects the pump to the main chamber and finally flush the cryo and chamber with N2 gas from the cryo’s exhaust valve. The third alternative is simply using a spare flange as a deidicated venting flange. It only needs to be a small flange, and you would have an adapter that drops to a ¼” Swagelok fitting. This is the simplest method. The next question is with what to vent. I would recommend hi purity N2 from a cylinder that is routed through a baked SS line (see MBE: N2/Gas system). Alternatively you can use LN2 boil off N2, but in that case it is more difficult to bake down the whole line to ensure the purity of the gas. Either way the N2 venting line needs regulating down to 1.2bar. Never use dry filtered air from a compressor to vent your system! It simply not pure enough! Venting then involves a generic, multi-step procedure:

  1. First ensure all your cells are at room temperature, except the Ga which should be at 50°C . Remember the As and P cracker valves need to be fractionally open whilst heating/cooling the cracker head. Always cool the bulk down first, then the cracker heat to avoid condensing material on the valve.
  2. Next current limit the Ga PSU to 2.00A and turn all the other PSUs off. This both saves power and gives peace of mind.
  3. Seal the As and P needle valves manually now to avoid contaminating the bulk.
  4. Next turn off the ion gauges and allow the filaments to cool: 30 minutes.
  5. Close the valves to the pumps.
    1. The ion pump can be left running (ion pumps should in fact never be turned off)
    2. Take the opportunity to regenerate your cryo
    3. Switch off the turbo pump and the backing scroll pump
    4. Close all valves to the pumps (unless you are venting through a pump)
    5. Vent
    6. Watch the Ga cell’s temperature and make sure it stays over 45°C

Venting can be dangerous. An uncontrolled vent can expose your system to dangerous pressures that cause damage to glass viewports, bellows and delicate internal items (like RHEED filaments). Which is why we regulate the N2 line to 1.2 bar. How many bars of pressure can an MBE system safely take? Well it is predominately a vacuum system so anywhere between 0 and 1. On the other hand you do not want to only partially vent it and then break a vent seal only to have atmosphere rush in at an alarming pace. The best way to vent is to monitor the internal pressure with a vacuum gauge. The Super Bee from Instrutech Inc  can monitor pressures from 0.001 to 1500 mbar. You can place one atop each of your backing scroll pumps in place of the more common pirani gauges. Then you can simply valve off the scroll pump, but leave the Super Bee open to the internal pressure via the Turbo’s gate valve. Let in the N2 and watch the Super Bee’s pressure reading slowly climb. Stop when you reach 1050 mBar, either manually or automatically using the Super Bee’s pressure trip digital output signal (the configuration is shown in MBE: Auto-FEL vent/pump down). What to do once the chamber is vented? There are 2 schools of thought:

  1. Keep the N2 flowing at all times to prevent excess moisture entering the system whilst you are conducting maintenance
  2. Turn off the N2 gas it has done its job

Both are valid, but remember you cannot stop moisture entering your system with 1.2 Bar N2. You would need to have a much greater rate of flow, particularly when opening a flange greater than a CF-38. That is a lot of N2 waste gas. It is therefore more economical and practical to simply turn off the N2 and leave it off. You can quickly get rid of the moisture with the pump down technique described in Essential maintenance: Pump down. Now the system is at atmosphere you can get started on all those maintenance jobs in your maintenance log. The general dos and don’ts are outlined in Essential Maintenance: A how to guide. Once done pump the system down using: Essential maintenance: Pump down and prepare for the bake out (see Essential Maintenance: Bakeout).

Essential Maintenance: Crucible cracking

Faebian Bastiman

Pyrolytic Boron Nitride (PBN) crucibles for MBE are actually grown in a process not too dissimilar from III-V MOCVD deposition. The pyrolytic process involves depositing crystalline ceramic boron nitride on a carbon mold (mandrel). The PBN crucibles are therefore very high purity, nonporous and possess smooth sidewall (no pores). The PBN is stable up to around 1400°C and is resistive to HF acid.

Each III, V and dopant element has its own particular relationship with PBN, some of them are neutral conscientious objectors, other treat PBN with open hostility! The following article highlights the necessary steps to protect your PBN crucibles from the common III-V MBE elements.

First we have the friendly elements: P, As, In and Be. None of these elements pose any risk to PBN and as such the sources can be operated without additional consideration for the PBN-element relationship and interactions. In is the only nice group III. Aside from passively “spitting” during operation it is essentially neutral. A cell operated with a large thermal gradient from top to bottom (hot-lip) helps reduce the buildup of In near the cell’s orifice and prevents spitting affecting the substrate surface.

In contrast Al, Ga, Si, Sb and Bi each require additional care when situated inside a PBN crucible. Al can readily shatter PBN. The situation is unique amongst III-V elements and pertains to Al’s ability to wet PBN. Liquid Al placed in the bottom of a PBN crucible gradually “creeps” up the sidewalls. The cell must be operated with a cold-lip to prevent Al climbing out of the cell and shorting the heater filaments. Ga and In remain at the bottom of the crucible, thus as the material is depleted the flux drops over time for a given temperature. The thin and complete Al coverage provides a large and constant surface area, such that Al cells do not suffer from flux drops for a given temperature. This can make it difficult to predict when an Al cell is running out. Finally Al has a very different thermal expansion rate to PBN. The wet PBN cannot contract as fast as its Al film, hence cooling Al too rapidly from standby (850°C) to solid (660°C) can readily shatter the crucible. It is common practice to utilize very, very slow temperature ramps when cooling through this range. For example 0.05 °C/min, equating to ~63 hours. Once the cell falls below ~600°C, and you can be certain the Al is solid, a faster ramp can be used. It is therefore a very good idea to fully deplete the Al cell before cooling it down. Moreover, it is good practice to replace the PBN crucible every time the cell is cooled. A double walled crucible or a crucible insert offer a buffer against the damaging effect of cooling Al too rapidly. You can further buffer against damage by having the Al cell on an uninterruptable power supply (UPS), having a backup DC power supply unit (PSU) that automatically takes over if the primary supply fails and by utilising a backup water cooling system rather than simply crashing the cell temperature upon primary water cooling failure (see MBE: water cooling system).

The next dangerous element is Ga. Ga, like Si, Sb and Bi (and the common compound water), possesses a lower density (they expand) upon freezing. Ga decreases in density by around 3.1% upon freezing. There are only 4 elements that expand upon freezing and in III-V MBE we encounter them all. What luck!  Liquid Ga sitting at the bottom of a crucible first freezes at its exposed upper surface creating a nice air tight cap against the PBN, then as the main Ga liquid freezes it expands sideways cracking the PBN like freezing water in a pipe. Luckily Ga freezes at around 29°C, this means that holding the cell at 50°C at all times (even during maintenance and air exposure) protects the PBN against the damaging effects of freezing Ga. Ga cells can be protected in an identical manner to Al cells, however you can also utilize a “water-heater-loop” on the Ga cell to maintain 50°C by supplying 50°C water through the cell’s water cooling element in the event of a DC PSU failure or other electrical power fault.

I avoided discussing Si along with Ga, because Si is more dangerous. Not only does it freeze and expand (decreasing in density by a massive 10%), it also etches PBN when liquid. The simple solution to both problems is to keep the Si solid. Si melts at 1414°C, which is inside the “danger zone” for PBN temperature operation. In some circumstances it is advised to melt Si on the first usage, for example when wanting to utilize a “downward facing” cell port, Si can first be melted in an “upward facing” position and upon freezing will remain rigidly fixed inside the PBN crucible even when facing vertically downwards and physically shaken. The etching effects of liquid Si on PBN will be minor for quick, one time melting, however this does preclude PBN crucibles from being utilized with high Si deposition rates. In those cases either an e-beam or a Si-filament source are required. The best advice I can offer is to only outgas the Si cell to 1275°C maximum, avoid melting the Si and hence avoid any dangerous effects.

Next we have Bi. Bi decreases in density by a moderate 2.8%. Not as dangerous as Ga, but still posing the risk of crucible cracking. It also freezes at 271°C, so there is no chance to keep it melted during maintenance. The best practice here is to be careful. Never over load your crucible. You will be amazed how long Bi actually lasts. A 10 – 20g charge at the bottom of a crucible out lasts all other III-V elements; largely because it is only used in dilute quantities in III-V epitaxy. It is good practice to fully deplete the Bi source before cooling. This means accurately predicting your Bi usage to coincide with maintenance periods. Using rounded bottomed crucibles is also good practice, as they seem less susceptible to cracking. Finally, like Al, go through the freezing point slowly. I use 0.05°C/min when cooling from 325 to 250°C.

Finally we have Sb. Sb presents the smallest density decrease upon freezing. A mere 2.6%. Luckily it has a relatedly large vapour pressure whilst solid. A valved Sb source produces a suitable Sb flux for III-Sb epitaxy with the Sb bulk held at 500°C. This is safely distant from the melting-freezing point of 630°C. The best advice for Sb, like Si, is therefore to avoid melting it altogether and thereby avoid any dangerous situation.

Essential Maintenance: Sample holders

Faebian Bastiman

In III-V MBE we direct a molecular beam at a heated substrate. The substrate is located within a sample holder (platen) for ease of mounting and manipulation within the MBE system. Hence whilst intentional epitaxial deposition is taking place upon the substrate, any exposed metal will partake in unintentional amorphous deposition. The shiny new platen soon becomes coated in a blue/purple film. This film readily adsorbs moisture and readily oxidizes in air, thus it is a significant source of contamination to an MBE system. The substrate and platen are typically outgassed to around 400°C in a buffer chamber. A clean platen and epiready substrate need only 30 minutes to reach the mid 10-9 mBar range. After some 30 growth runs, the same substrate and platen take 90 minutes. Clearly the epiready substrate still requires 30 minutes, but the contaminated platen needs 90 minutes. The thicker the film, the longer the outgas procedure.  It seems we are wasting time outgassing our platen and in so doing we are unnecessarily contaminating our vacuum system. Also the III-V layer significantly alters the platen’s thermal response and actually lowers the maximum temperature you can achieve! What to do?

Most platens are composed of Molybdenum (Mo). Occasionally stainless steel (SS) may be utilized for low temperature (<550°C) substrates (InAs, InP, InSb) to save on material and machining costs, but Mo is standard. Mo is a very robust element, which means it can be rather aggressively cleaned without worrying about damaging it with chemical etching. Outlined below is a procedure to remove all III-V deposits from Mo platens. Note full breathing apparatus is advised and work should only be performed under a fume hood!

Firstly you will need two relatively coarse grades of wet & dry abrasive sheets. Automotive wet and dry is sufficient. I use 80 and 120 grit. Using a container of DI-water, submerge the contaminated platen and then rub it with the 80 grit. During this phase you can touch the platen with your gloved hands. 80 grit cuts through the III-V material relatively swiftly, whilst only inflicting minor scratches on the Mo platen. Working with the platen submerged in water will minimize III-V dust, however if you prefer not to work underwater regularly wet the platen to reduce the depth and number of scratches. Once all the material is removed, sand down the scratches with the 120 grit. Repeat with 240 grit if you want to eliminate all scratches. Finally thoroughly rinse the platen with DI-water and dispose of the III-V contaminated water by your facilities established practices. From this point on only handle the platen with clean tweezers.

Next you will need to submerge the platen in undiluted (38%) HCl for several hours. The Mo is impervious to HCl, though any wet & dry residue will be etched away. I deliberately avoid using Aqua Regia since this tarnishes the Mo. After the etching, rinse again with DI-water and this time blow dry with dry N2 gas. Finally soak the platen in isopropanol (IPA) for 5 minutes and blow dry for the last time.

You should hold in your tweezers a very shiny, silvery Mo platen free from aqueous IPA residue. Before you place the platen in your MBE system you need to bake off the dried on IPA residue. To do this place the platen in a 120°C baking cupboard overnight. Once removed and cooled the platen can be stored in a desiccator until needed.

The final phase is to outgas the platen in vacuo. It is best to load an appropriately sized tantalum (Ta) disc in place of a wafer. A Si wafer is a viable alternative to Ta. The Ta/Si is required in order to outgas the platen to the maximum heater temperature, which is around 100-200°C hotter than one would typically take a III-V wafer in an MBE system. Outgas the platen in two stages. First heat it to the maximum you can reach in the system’s preparation chamber outgas stage (~600°C), then take the platen to the maximum operation temperature of the growth chamber’s sample heating stage (~800°C). The maximum temperature will vary from system to system. It should however be at least 100°C higher than you would otherwise normally grow. When outgassing the platen keep an eye on the background pressure and avoid letting it exceed the mid 10-7 mBar range. Aim to hold the platen at maximum temperature for an hour. The entire procedure may take 8 hours per platen to complete. I perform the procedure manually the first few times, then write a simple recipe of substrate temperature ramps to perform the outgas automatically overnight thereafter.

Each platen may need cleaning as often as once a month or as infrequent as once a year. It depends on the system usage and the cocktail of materials deposited. Keep an eye out for the telltale lengthening of normal outgassing times. It is good practice to purchase a number of spare platens to have a few clean ones in reserve whilst the contaminated ones are “out for cleaning”.

Essential Maintenance: The maintenance cycle

Faebian Bastiman

MBE operation should be a constant cycle of growth campaigns and regular, scheduled maintenance periods. I put the word should in italics because, unless you are careful, maintenance can enter the cycle of reactionary quick fixes that inevitably lead to extended downtime. So how do we avoid this undesirable maintenance cycle? We need to put in some extra effort and employ some good operation/maintenance practices: fault logging, planning and regular servicing and keep a stock of spare parts.

Fault logging is highly underrated, but in retrospect it is one of the fundamentally most useful aspects of maintenance. It can be as simple as a quick note in a logbook at the end of each day to (hopefully) say “everything is ok” or (sadly) to note the occurrence of a new fault. I like to keep a spread sheet log that lists the problem, the date it was detected, actions taken and the date it was fixed. The log is also a useful reference during maintenance periods, since it is essentially a task list comprising all the little jobs that need to be completed.

Some problems are minor irritations, some are ex situ and can be repaired at your leisure and (again sadly) some are serious and require the premature interruption of the growth campaign. It is the latter type of problem that we would like to avoid. That is where two crucial aspects of good maintenance practice come in: planning and regular servicing.

Every part of the system has a finite lifetime. The service intervals are (i) on-going, (ii) once per campaign (which can be 3 – 12 months depending on your system, configuration and usage), (iii) annual, (iv) biennial, (v) once every 5 years and (vi) once every 10 years.

The on-going maintenance tasks are those that do not require the system to be opened up, repaired and baked, but are nevertheless essential. This includes cleaning the sample holders (see Essential Maintenance: Sample holders), regenerating the cryo pumps (see Essential Maintenance: Cryo pumps) and outgassing the cells after the weekend off (see Essential Maintenance: Cell material regeneration).

The once-per-campaign maintenance tasks should be undertaken during regular, scheduled maintenance periods before each growth campaign. The tasks are:

  • Replace Al and Bi crucibles (or replacing the PBN crucible inserts)
  • Replenishing effusion cell source material
  • Cleaning shutters (depends on system)
  • Cleaning viewports (depends on system)

It is a good idea to try to anticipate and time your material usage so everything runs out at the same time. This is easier said than done and experience (as always) plays a big part. It is a good start to put as little Al and Bi as possible, since it is best to fully deplete this sources each campaign (see Essential Maintenance: Crucible cracking for more details). Of course your other choice is to simply switch your alloy composition until all your cells are down. For example once my Ga cells are depleted I grow some InAs/InAs(100) pins diode  detectors for a colleague.

Annual and biennial tasks are in some ways system dependent and hence it is up to the user to decide what good practice is for their own particular circumstances. The tasks are:

  • LN2 cryo-panel cleaning/regeneration
  • Cleaning the substrate heater/manipulator
  • Inspect RHEED screen As coverage
  • Clean MIG ion gauge head (see Essential Maintenance: MIG head)
  • Replace any broken TSP filaments  (as necessary)
  • Replace broken ion gauge filaments (as necessary)
  • Service scroll pumps

The LN2 cyro panel regeneration is a neglected service item. In an ideal world each III-V MBE machine would possess two panels, employed with a “one to wash against the other” philosophy. I personally cleaned my Omicron MBE-STM cyro-panel once a year. This is good practice because the As layer is still relatively thin and can be brushed away. If you are going to service your cyro-panel once every 5 years you will need to have the whole thing shipped and dipped (see Essential Maintenance: Cryo-panel).

Arsenic gets everywhere. A thin film starts to build up on the growth chamber interior, despite the presence of shutters, and that means it starts to grow where it is not wanted. In particular this is the RHEED screen, MIG and the manipulator metal work. The latter point can result in poor RHEED access to the sample and eventually no RHEED at all.

5 year service items are often forgotten simply because they are 5 year service items. A lot can happen in 5 years, the entire composition of a research group could well change. Here are some notes to leave for your successors:

  • Inspect, clean or replace rubber seals on load lock and gate valves
  • Service UHV pumps
  • Service Al and Si cells

Lengthier service items are:

  • Other effusion cell replacement or service
  • Ion pump replacement or service
  • Software upgrade
  • Rack hardware service/upgrade
  • Clean and service RHEED gun

The UHV pumps are probably the most neglected regular service items. Oil-and-ball-bearing based turbo pumps require a return to manufacturer and full service once every 5 years, those with magnetic bearings can be operated up to 10 years. Ion pumps too require regenerating/replacing every 8-10 years.

It is a good idea to have your Al and Si cells serviced once every 5 years, since they are high operating temperature cells. The other effusion  cells will typically need servicing once every 10-15 years. The best practice for effusion cells is to buy ones with separate water cooling jackets and shutters, and hence the cell itself is simple to replace or repair. As crackers can operate 20+ years, but they will need refilling once every 5-15 years depending on your usage. The cost of a full service on an As cracker is around 65% of the price of a new one.

Finally let’s consider our stock of spare parts. Obviously the safest way to operate is to have two of every item; though for financial or space reasons this is impractical. You can buffer against some problems by entering into a service contract with the manufacturer whereby effusion cells and pumps are serviced on a rolling-replacement basis. It is however good practice to standardise your parts as much as possible and keep in stock:

  • One of each kind of crucible
  • One of each kind of bellows
  • Numerous gaskets of various sizes
  • Blanking flanges of various sizes
  • Cell source material and an As charge
  • Ion gauge heads and filaments
  • TSP filaments
  • A RHEED filament
  • A RHEED screen
  • A gate valve seal kit for each type of gate valve on your system
  • Manipulator/outgas stage filament
  • Multiple shutter blades
  • Cell parts (should you decide to service them yourself)
  • DC power supplies
  • A water chiller pump
  • An ion gauge controller

Clearly maintenance should not be taken lightly. It is a large part of MBE operation; however by establishing good maintenance practice it need not be overwhelming or full of unexpected surprises. And most importantly, good maintenance practice forms the foundation upon which one can establish very fruitful growth campaigns with minimal disruption.

Let’s fix: Cleaning an Al cell shutter

Rob Richards

When we leave the lab we make sure that everything is in its standby state and we set up an automated outgas (or “morning initialization” as we call it). This usually means that the cells are clean and ready to use for the next day. However, one Friday morning we came in to find that the water cooling system had broken; water was not being pumped through the cells and was boiling in the pipes, causing the cells to heat up and outgas into the chamber. Having failed to restart the water cooler we crashed the cell temperatures to prevent them from being damaged. When we took the machine apart (it needed restocking with gallium and indium anyway) we found the aluminium shutter looking like this.

The sizable blob of aluminium which had accumulated on the shutter had clearly reacted with the outgassed material from the cells and now looked pretty disgusting!

The first step in the cleaning process was to remove as much of the aluminium as possible with a craft knife. After this the shutter looked significantly better but still not quite good enough to be used again. The remaining aluminium had to be etched away, so the shutter was placed in a beaker of room temperature HCl. The reaction was very quick and after ten minutes the shutter looked like this. Nice and clean and ready to be put back into the system!