PLC: Risk Analysis

Hello again everyone.

This represents the fourth and final chapter in my brief series on PLC for MBE. I am going to end with talking about risk analysis. Every single commercially available hardware device, software package or customer service in the world has been scrutinised by some kind of risk analysis. It is basically impossible NOT to do a risk analysis. Whether it is a formal FMEA (Failure Mode and Effect Analysis) or listening to that feeling in your gut that something is off, human beings ascribe inherent risk to almost everything. And with good reason, knowledge of risk can prevent disaster.

FMEA is just a tool to quantify and track risks, much like a GANTT chart is a tool to plan and track a project. FMEA has 3 basic parts.

Firstly you gather ALL the risks and I mean all. This is usually pretty fun and involves many Brainstorming sessions. For a simple device like a toaster you might get 50 risks. For a complex device like an MBE system you might easily reach 1000.

In the second stage you ascribe each risk a value for level of severity (S), likeliness of occurrence (O) and difficulty of detection (D). A number from 0 to 10 is ascribed to each. The three are then multiplied together to give a risk out of 1000. A ridiculous risk might be that the system kills the operator (S=10) every second (O=10) and there is no means to detect it (D=10) gives a risk of 1000. Any risk that has a severity greater than 7 or a combined risk great than 500 is termed a RED risk. RED risks must be mitigated. Severity (S) is not only in terms of user injury, there is also the risk to finances and timeline to consider.

The third and final stage is just that, reducing the risks by redesign or adaptation of the system. Often this involves either redesigning so the occurrence is less or adding some feature so the detection risk is less. And of course the MBE system should not be killing anyone, ever, so the system needs redesigning to reduce the severity of any risks to “minor injury”.

If our system in second hand, DIY or cobbled together from 3 older systems it probably invalidates any OEM’s (Original Equipment Manufacturer’s) FMEA. You do not need to do a formal and thorough FMEA, just set a threshold in terms of money or time that you are not willing to risk. You could say, any breakage that costs more than 1000 Euro to repair or results in more than a week down time is unacceptable. Then think of the things that can cause these breakages, for example water chiller failure, power cut, thermocouple sensor break or pump failure. Then have a backup in case these failures occur to prevent the breakage, for example:

(1) water chiller failure can be avoided with a secondary chiller or a mains bypass

(2) power failure can be avoided with an uninterruptable power supply (at least on critical components)

(3) thermocouple sensor break can be mitigated with detection and switching to a fixed output power

(4) pump failure can be mitigated with closing the gate valve and having a second pump take over

All these solutions require some kind of detection of the fault and then an action to avoid damage. If you are thorough you will come up with 1000 or so of these risks and probably need to mitigate 100 of them. Luckily, a good PLC system already does this for you.

The bottom line is that of you have even one risk that is a RED risk, the FMEA is yelling you took system is to risky to operate!

So in order to wrap up my exploration of PLC for MBE, I ask you: what is the risk of NOT having one?

PLC: Central vs Distributed

Hello again.

Yes, yes I am still going on about PLCs. This time, I am looking at them from a system architecture point of view. Take a look at the picture below of the typical parts of an MBE system. Can you guess what goes in the big black box? Stop and take a guess, there is no wrong answer.

Ok if you said “a cat” you are not wrong, however your mind works very differently to mine. Moving on…

The first and perhaps most obvious answer is “the deposition chamber”. This makes sense since the deposition chamber (and by extension effusion cells, ion gauges, shutter and LN2 shroud) is connected to all these parts. The deposition chamber is the “outward facing” connecting part. But what is the “inward facing” part?

The next most obvious answer is the PC. It is indeed common to have the MBE software that resides on the PC connected to most of these parts. Perhaps most people would not consider connecting the pumps, water chiller and RHEED gun controller to the PC since these parts do not need “regular interaction”. However at the same time these parts are not “regularly monitored”. In the interest of preventing MBE being Mostly Broken Equipment, a viable solution is to connect ALL hardware to the PC and run a “Watchdog” program that acts as a “soft-PLC“.

The actual answer kind of given away in the title is the PLC controller. In this configuration the PLC is at the heart of all the hardware and the actual PC is just another device connected to the PLC.

The advantage of this centralised configuration is that the PLC is at the centre of knowledge. It can directly interact with all the hardware and prioritise communication. The PLC holds the complete state of the whole system. The PC is simply makes read and write requests to the PLC and the PLC passes them onto the corresponding hardware.

In the case where the PLC is NOT in the centre the system becomes distributed and complicated. No one, not even the PC, has the complete picture of the system and if the PC is part of any ensuing problem there can be no solution. In this case the configuration requires either (1) a Watchdog running on the PC or (2) basic error states passing to the PLC via digital IOs from each piece of hardware. The problem being that not all the hardware reports “error states” via digital out.

This ultimately means that, when considering system architecture the PLC need to be at the centre. The PC based control software should ideally be designed to work with the PLC and the PLC alone. The PLC carries the heavy burden of needing to work with all the other hardware from all different manufacturers using all sorts of COMs interfaces and protocols.

To sum up, the solutions in order of preference are:

(1) Centralised PLC designed to work with PC software

(2) Distributed PLC overriding software via digital IOs

(3) A PC based watchdog integrated into the PC control software

Thanks for reading. Have a great day everyone.

PLC: Hard vs Soft

Hello again.

Yes I am continuing with the theme of Programmable Logic Controller, that much overlooked but so critical feature. This time around I am going to explore the age old question of whether a software or hardware based PLC controller is better. (Spoiler alert it is hardware).

Many people would argue that a software based solution is easier. Most MBE systems already have a basic software package that communicates with most of the system hardware in order to run recipes and grow samples. It is therefore fairly intuitive to add some kind of a “Watchdog” feature to that software. Intuitive and fairly easy.

Easy does not equate to better. Imagine if Michelangelo had told Pope Julius II that despite his idea to pain the ceiling of the Sistine Chapel, it would be must easier in fact to paint the floor, and in crayon. (Note I recently visited the Rome and it was amazing, hence the reference).

My point is that easy does not equate to better. A software watchdog is better than no watchdog at all, but bear in mind that Windows is inherently equally flexible and unstable. The all seeing watchtower monitoring for constant threats should not reside in the land of BSOD (that is Blue Screen of Death) and OIASWYITMOSINTR (that is Oops I am sorry were you in the middle of something I needed to restart).

The fact is PLCs are hardware based for a reason: A dedicated microprocessor with a single task monitoring and dealing with faults. Some of the more complex PLCs are very rugged computers, but for MBE with a few dozen inputs and outputs, a single microprocessor is plenty.

That is not to say a PLC cannot have a sleek, elegant user interface. I have to admit I do love a sweet UI. A PC based UI that can add, edit and delete “rules” for the PLC can only be a good thing. The actual critical task of monitoring and reacting should however be performed on a dedicated little blinking box that does not get distracted by services and kernel threads and spyware and viruses and… and… and…

Thanks for reading. See you all tomorrow.

Mostly Broken Equipment

Hello again everyone! I am delighted to be back with my first post since 2015.

Molecular Beam Epitaxy (MBE) has inherited the pseudonym Mostly Broken Equipment and in this post I intend to explore why.

On the one hand, yes, this makes sense since an MBE reactor is an incredibly complex assembly of discrete devices designed to achieve a coherent whole. I can still remember that feeling of quiet awe and sinking dread when I was first introduced to the shiny beast upon which hinged my hopes of gaining a PhD. Such a mass of tangled wires and hoses, blinking LEDs, and weird noises from pumps I had never before seen.

On the other hand, no, it doesn’t make sense. This incredibly complex system is incredibly expensive and almost indestructible, to the point that even the 25+ year old second hand systems can belt out world class samples. Right, right, I hear you say, I heard the story of an indestructible hammer: The owner replaced the handle 3 times, and the head only once. Regular maintenance and repairs are the topic of an older post, the focus of this post is how to keep the system running “almost every day” during its 25+ year life time.

When I visit an MBE lab, I often get the basic tour showing off the fancy ISO 3 cleanroom and the recirculating LN2 system. I grudgingly don my cleanroom (space suit) and continue onto the lab floor noting the fume cupboards and the ranks of blinking hardware racks. On the surface everything looks great. In those racks, there are usually 10+ PID controllers and their corresponding DC power supplies, 2-3 ion gauge controllers, a medley of pump controllers, a shutter controller and maybe a RHEED gun power supply.

“Are the racks backed up?” I hear myself ask.

“Oh, yes, we have an uninterruptable power supply running in the basement,” comes the proud answer.

I nod, and smile, “Does the chiller run on the same UPS?” I ask.

Hesitation follows. Sometimes even wringing of hands. “Well…erm…”

My smile vanishes, but I offer a chance to save face: “Well at least the water flow meter can protect the cells?”

My proud host starts silently mouthing the words “flow meter” over and over with a perplexed look on his face.

Here starts the problem. Here is a nice shiny (sometimes new) MBE reactor costing upwards of €0.5M but there are some gaping holes in the interlocks and watchdogs. There is often quite literally a gaping hole in the rack where the PLC controller should be. A PLC controller constantly monitors a set of inputs (like water flow rate, LN2 temperature, vacuum level, etc) and then makes decisions based on a custom program to control certain outputs (like ramping cells to standby, opening the gate valve to the turbo pump, etc). The PLC acts as a nexus and watchdog protecting the system on a second by second basis.

When I describe this functionality many people glance over at their PhD student as if to say “my PLC controller is sitting over there”. Well sorry no. Not good enough. Even a washing machine prevents you opening the door when the drum is filled with water.

In fact, whether you physically have a PLC controller or not, your implicitly have one. You might even say a PLC system represents your values toward your MBE system. So the in absence of a physical PLC controller your implicit PLC controller control flow looks like this:

IF “the water chiller fails” DO NOTHING

IF “the thermocouple breaks” IGNORE IT

IF “the vacuum integrity is lost” CARRY ON REGARDLESS

With these implicit values, is it any wonder the MBE reactor is Mostly Broken Equipment? That is simply karmic action. You may say you do not believe in karma. Well, karma believes in you, and believe me, it can be a bitch.

If you are reading this and staring at your own gaping whole in your rack whilst simultaneous suppressing your feelings of smouldering rage because the system is “down again” today, I urge you to fill that space with a PLC controller.

Thank you for reading. See you again soon. Oh and…

if you want to buy a PLC controller designed and built to protect MBE systems please contract us as Scitech Solution Ltd, we are in the processes of building one.

Facility Setup: A room with a view

Faebian Bastiman

One of the questions I get asked frequently is: does an MBE system need to be in a cleanroom? My answer is: no, but it needs to be in a clean room. What do I mean? Well the room must be designed in a way that is easy to clean, and must be kept clean. The ideas are summed up very nicely in this blog post from Hutchins and Hutchins.

So what is the difference between a clean room and a cleanroom? What does an ideal MBE lab look like? Well consider the classic laboratory (and I am not talking about Dr Frankenstein’s castle’s highest tower complete with lightning rod) I am talking about the classic white walls, floors and ceilings, plethora of air vents and cold fluorescent light. Ghastly!

The white surfaces and ventilation can be maintained, but one of the first things you should consider is to flood the room with natural light. This of course means windows. Not only is sunlight a free and available mood enhancer, it also increases productivity and assists is healthy sleep patterns. Everyone, even scientist chained to their MBE systems, deserve natural light. Furthermore strong sunlight is an excellent means of detecting airborne or surface dust, and prompting you to do some cleaning.

The next thing to consider is that the dirtiest thing in the room is the human being, so one should limit the room one wishes to keep clean to human exposure. The MBE lab is also generally unpleasantly noisy due to the various pumps maintaining UHV. The best thing to do is therefore to divide the floor space into an office and a lab area. I direct your attention to figure 1 which represents an example MBE layout in a corner building location.

Figure 1: An example MBE lab

Note the light blue rectangles are windows and run the length of the upper and right edges. The corridor at the lower edge of the image has a door shown in purple that leads into the yellow office area. The office area is equipped with a desk that houses the MBE system’s control PC and has a further internal window where one can view the MBE system’s racks. The office should perhaps be air-conditioned to a desired temperature (e.g. 22 °C) but need not be filtered or humidity controlled.

The dark green space directly above the office is the dressing area. Here the user removes their external footwear and replaces them with non-shedding, clean footwear. One should also wear a knee length clean room coat (and if one is going to load/unload samples a hair net and face mask). The floor to the dressing room should be coated in sticky mats and there should be a sink for washing hands on entry and exit.

The light green lab is entered primarily through the dressing room, though the lower windowed wall should be removable to enable the MBE system to be installed in the first instance. The lab is where one should focus their intentions to create cleanliness. The room should be air-conditioned (e.g. 22 °C) to counteract the heat load from the racks. The air conditioning could also consist of low level filtering to reduce the typical class 1,000,000 or ISO 9 in a typical office room to 100,000 or ISO 8 to reduce cleaning frequency, but this is not essential for MBE research. Additionally one could consider employing humidity control (down to 40%) to reduce moisture intake into system during sample loading and maintenance. Most importantly the floors, walls and ceiling must be constructed from non-shedding material that is easy to clean with a solvent wipe. This is liable to be the highest expense, especially the ceiling tiles. The MBE system itself should be situated on the centre of the lab with at least a metre free on all sides. Ideally, the rack should be installed next to the system to avoid passing cables between the two. All cables should feed down from the ceiling above the rack and not across the floor, the same goes for the services (process gas, cooling water, cryo pump He, LN2) all should be fed down from the ceiling. The floor around the system can then be easily cleaned and there are no obstacles in the way that prevent dust being drawn into the low level extracting ventilation on the lab’s walls.

The lab requires a laminar flow fume cupboard for exchanging samples or performing maintenance. The remainder of the available wall space can be used for essential and convenient storage of bake out panels, substrates, ultra-pure metals, gaskets, flanges, spare cells, etc. Spare cells could optionally be stored on an outgas rig if there is sufficient space.

The lab should be free from paper, pencils, books, ink etc. All of that should be stored in the office that houses the MBE system’s operator. Ultimately, there are only three reasons to enter the lab space:

1. Load or unload samples
2. Clean
3. Perform system maintenance

If the MBE system has automatic sample transfer and can execute batches of samples (see my Dream Machine blog post), one need only enter the room for a few minutes every other day. In doing so one need only clean the lab once per week. It should be vacuumed with a high efficiency particulate air (HEPA) filtered vacuum cleaner and all surfaces should be wiped down with appropriate solvents. In most cases the system maintenance should be of the scheduled kind and should be followed by the most exhaustive cleaning. Maintaining a slightly higher differential pressure in the lab compared to adjoining rooms will also reduce particle intake on ingress.

Note the LN2 phase separator to provide cryo cooling is not shown in the figure, but is intended to be installed in the ceiling space above the MBE system.

The final room to the left of the figure is the service corridor. This houses any of the services that create particles or dirt. Particularly one may wish to install the air compressor for pneumatic gas, the He compressor for the cryo pump, the water chiller for cell cooling and a pair of N2 cylinders for ultra clean process gas. One can also include an AC power distribution panel and a large roughing pump that augments any turbo pumps on the system. The service corridor need not be located on the same floor as the lab, and could be a full service area that includes the facility’s backup generator, UPS system and end line trap for the roughing pump’s exhaust.

The result is a very clean, easy to maintain lab space where cutting edge opto-electronic and electronic semiconductor research can take place that does not cost the earth.

Little known MBE facts: High Purity Material

Faebian Bastiman

I was asked recently what purity of material a user should use in their MBE system. Here the grower was referring to the number of “N’s”. The N of the material is a reference to its purity and is actually the number of nines: Material that is 99% pure and 1% impure is termed 2N. 99.5% is termed 2N5, the “2N5” tells you it is at least half way to being 3N. Of course this does not tell you what the impurities are specifically, however you can assume they are made up of all the things you do not want in your III-V thin film layers (Zn, Cu, Fe, Sn, Hg, Ca, Te, etc) and a few III-V elements that you do not need to worry too much about (you are, after all, growing III-Vs anyway).The purity of material will ultimately determine the quality of your thin films, since any impurities in your cell material will likely incorporate as unintentional dopants in your layers.

If you are growing metals (for example MnAs) you are not too worried about unintentional doping and you may use a maximum of 5N5 material (99.9995%). If you are growing electronic or opto-electronic grade semiconductors (for example GaAs) you will want higher purity. Ultimately the highest you can get commercially is 9N (99.9999999%), however I would not recommend you all rush to the shops and buy such expensive material for general research. 9N is exponentially more expensive than 8N, 8N is exponentially more expensive than 7N. Similarly the ultimate background doping you can achieve with 9N is an order of magnitude lower than with 8N, and the same applies when comparing 8N to 7N. Before I suggest the appropriate material quality, let’s do a thought exercise with GaAs…

GaAs has a lattice constant of 0.565338 nm, and therefore an atomic density of 4.42 x 10²² cm^-3. When opto-electronics people (specifically detector people) talk about background doping requirements they say that 10¹⁵ cm^-3 is already good.  This means they would like the unintentional doping level to be 5×10⁷ times lower than the atomic density. In this case 7N5 would be the appropriate choice for you group III and V material. One could argue that you should use 7N5 for all materials and that replacing a specific material with 8N is a waste of money. However this “all or nothing” philosophy is not really justified since unintentional doping is accumulative, and hence replacing your group IIIs with 8N might reduce your background doping from high 10¹⁵ to low 10¹⁵. Thus 7N5 to 8N is a good choice for opto-electronic research, and 9N is only for exceptional high mobility cases or world record attempts. Moreover, if you are simply experimenting with a new opto-electronic alloy and not so interested in device quality at this stage 6N5 is acceptable. Note that the source material is only one factor in your ultimate achievable material quality, for more information read my Optimum Quality post.

Finally consider the dopant material. When we dope a semiconductor we typically dope in the 10¹⁶ to 10¹⁹ cm^-3 range depending on the application. At 10¹⁹ cm^-3 the dopant is around 10⁴ times lower in atomic density than the III or V species. This means the doping fluxes are around 10⁴ times lower that the group III and V fluxes and hence any impurities introduced into your system from the dopant source will also be 10⁴ times lower. Add to this that when you dope the alloy you are “trying” to add impurities and you can say that 6N is already a very good grade for dopant material and 5N may even be suitable.

MBE Dreams: Dream Software

Faebian Bastiman

Returning to the theme of “Dream MBE”, I would like to spend a moment discussing molecular beam epitaxy software. Off the shelf MBE software is passable at best. Ok, so it can open and close shutters, perform temperature ramp ramps and execute a sequence of commands in some form of recipe: This is, when all is said and done, the bare minimum. When you spend some time actually considering the features you would like in MBE software, you quickly realise nothing on the market really hits the spot. So… I have spent the past six months writing my own MBE control software: EPIC™.

You cannot create the world’s greatest MBE software in six months, however you can implement an exciting list of features and lay the foundation for fantastic things to follow. Beyond the bare minimum functionality, the features of EPIC™ are:

1. Comprehensive, optimised software-hardware interface (SHI)
2. <1 millisecond to process a shutter request
3. Full device parameter explorer and editor
4. Ability to monitor PID loops, pyrometers, plasma sources, pressure gauges, pumps, valves, motors, services (LN2, water flow, gas pressures) status and other hardware (e.g. reflectivity data) all from a central, grouped, tabulated status screen
5. Associate any number of physical devices with an individual control loop
6. Ramp to temperature set point or output power for PID controllers
7. Flexible, simple script recipe with full control flow, local and global variables, unrestricted parallel editing and execution including “edit on the fly” and access to any parameter from any device
8. A colour co-ordinated message box
9. Wake-up and sleep scheduling
10. Substrate location and status tracking
11. Automatic flux scheduling, gathering and Arrhenius/polynomial fitting
12. Atomic flux, growth rate, doping and composition calculation tools
13. Material usage tracking and cell fill level estimation
14. Two data logging modes: SPAM (log at fixed interval) and SMART (log only on change)
15. Watchdogs warnings and actions
16. Integrated simulation mode with full hardware emulation

Ok let’s analyse that numbered list. (1) SHI is a fantastic interface with every device possessing its own read and write thread. SHI enables (2) and (3) by its very nature. Not every process requires less than 1 ms shutter response, though since it exists by default I thought I might as well include it in my list. The device explorer is also a nice feature, since the front panel menu on most devices can be a pain.

The status screen (4) is very comprehensive and the layout facilitates ease of reading. A quick glance can reassure you that every part of the MBE system is ok (or equally quickly warn you when it isn’t!). All necessary items are interactive, giving the ability to perform ramps and open/close single/multiple shutters. The ability to associate any number of devices with a control loop (5) is very useful for plasma sources when they may have: a mass flow controller, a RF power controller, a main shutter, several valves and an optical monitor. So too is performing power ramps (6) for certain control loops, like carbon sources.

The recipe script (7) is essentially its own programming language complete with interpreter. Each recipe executes on its own thread and all commands run to 1 ms precision. The control flow allows you to step through each line individually one-by-one or allows the recipe to run freely, you can pause any line, go to any line from any line, skip lines, stop the recipe at any point and edit the recipe on the fly (only when paused for safety) and the recipe checks every line for errors on edit. Local and global variables allow parallel recipes to control complicated processes. Whereas access to any device parameter gives the ability to do almost anything, including in situ reflectivity control (see Dream Machine).

The message box (9) simply keeps you informed with a variety of colour co-ordinated, time stamped status updates including “Recipe X completed”, “Ramp X started”, “Sample X moved to position Y” or the less pleasant but equally useful “device X failed to respond”.

Wake up and sleep is fantastic (9). One can sleep much better knowing the machine will wake up at 06:30, do all the flux checks for the day and wait for you to stroll in at 09:00. Equally, the sleep mode prevents that troublesome “did I ramp down the Gallium before I left?” since it sets the system to a predefined standby state.

When you have more than a handful of substrates and/or more than a single user it is very useful to have a substrate tracking screen (10). Each substrate can be individually tagged with a name, a colour co-ordinated status (new / degassed / grown) and its current location in the system. The substrate tracking paves the way for extended functionality such as automatic run logging, automatic growth sheet creation, automatic sample transfer and batch sample processing (all of which are on my to do list).

Any software that purports to convert effusion cell temperature into growth rate is just plain lying! Beam equivalent pressure (BEP) is typically what you measure and atomic flux in atoms/nm²/s is what you get. Growth rate is simply something you induce depending on the atomic density of your substrate. Software that gathers flux data for you (11) and plots it on a graph is a welcome start. A one click fitting option ranging from simple Arrhenius to 12^th order polynomial fitting is sufficient to handle all source and opens the door for much, much more. The first step is enabling conversion from BEP to atomic flux. Once done you can use a growth rate and doping calculator tool that updates to the correct doping density and growth rate for any substrate (12). You can then “ramp to doping density” or “ramp to growth rate” rather than “ramp to set point”. You can estimate the real time composition during growth and calculate the composition ahead of time for a given cell temperature combination. Perhaps my personal favourite is that you can integrate the flux with time every single second and track the cell material usage, display it on a simply bar chart and estimate the fill level of each of your sources to 1%. All of these flux related features are present in the current version.

Of course you want to log all relevant data values (14) and the option to log only on change prevents large data files being created when the system is idle (over a weekend). You also want to monitor the instantaneous data values and be warned when they are outside of an “OK” range. Moreover, for certain values you would like the software to take an action on your behalf to prevent damage (15). Finally you may want to run the software in “off line” mode to “try before you buy” or even to “test a recipe before you run it for real” and in these circumstances the simulation mode is very handy.

Q: What could possibly be missing?

A: Access to the source code.

The reasons for open source MBE software are almost as numerous as the feature list itself, so let’s stick to the two most important ones:

1. Add device drivers for any hardware
2. Add/edit/remove content to the user interface

Adding device drivers is somewhat essential. The system hardware changes over time as the system is upgraded or re-commissioned for a new purpose. The current EPIC™ has a generic driver template that can be used to create a new device driver in around 30 minutes. The next step is to make that driver template available to all and to use device driver DLLs. Fairly simple.

The user interface is a very personal thing. The software-hardware interface (SHI) is rather universal once done right, but the user interface is a matter of taste. Script recipes are very powerful but perhaps not for everyone. Tabulated data makes for fast referencing but perhaps diagrams are more your thing. Etc. Etc. Etc. There are three solutions beyond making the software open source:

1. Create a software development kit (SDK) for EPIC™ to allow everyone to customise it with their own desired features
2. Allow SHI to be a stand-alone data acquisition tool that can be used either via windows API or via a socket interface to allow the user to create their own graphic user interface or web application
3. Provide customisation on demand and provide a tailor made solution for each user

Happily EPIC™ will have all three.

One final question: Where can I try the latest EPIC™ beta demo version?

Here.

Little known MBE facts: RHEED oscillations (3)

Faebian Bastiman

I was recently reading a nanowire publication and I was reminded of another means of calibrating the group V flux that I used in my III-Sb days. This is the preferred method for III-Sb epitaxy, however  it is also applicable to general III-V growth. In the following I will use GaAs as an example.

The first step is to establish your Ga growth rate using the method in my RHEED oscillations (1) post. Then convert this into atoms/nm²/s using the method in my flux and growth rate post. Finally, you can then calculate the atomic flux of the Ga cell versus temperature using the method in my Arrhenius plot post.

Then set the As cracker to a value that you wish to determine in atomic flux, for example set it to 25 % open. Set the Ga cell temperature to give you a flux of 0.069 atoms/nm2/s [i.e. 0.1 ML/s growth rate on GaAs(100)]. Open the Ga shutter and record the RHEED oscillations in the usual manner. As long as the As flux is larger than the Ga flux you should obtain a growth rate of 0.1 ML/s ± an error of up to 5% depending on how well the RHEED intensity oscillated. Note the error gets larger the fewer oscillations you obtain. Hopefully you can get at least 10 oscillations.

Next double the Ga cell’s flux to obtain a growth rate to 0.2 ML/s and (importantly) leave the As flux set to the original value. You will need to leave the Ga cell to settle for 10 minutes after changing its temperature.  As long as the As flux is larger than the Ga flux you should be able to obtain a growth rate of 0.2 ML/s ± 5%. The magnitude of the As flux compared to the Ga flux is key to this method. Keep increasing your Ga flux until (eventually) the RHEED oscillations no longer yield the growth rate determined by the Ga cell. When this happens the growth rate is no longer dictated by the Ga flux, it is dictated by the As flux. You can check this by increasing the As flux and repeating the measurement that gave the lower than expected growth rate.

If you plot out all your data points you should obtain a graph like the one shown in Figure 1. You can see that starting at small Ga flux, the growth rate initially increases linearly until eventually it becomes As poor (Ga rich) and the growth rate is limited by the As flux. The growth rate you obtain under these As poor (Ga  rich) conditions indicates the As growth rate in ML/s. 0.5 ML/s in this example. You can then convert the As growth rate into an As flux using my flux and growth rate post once more.

How to grow your first sample: As capping

Faebian Bastiman

Having a metallic As source offers us the ability to create a very useful protective epilayer: an As cap. Once you have grown a layer you may want to remove the sample from your system and for example transfer it into a neighbouring system or send it to a collaborator for analysis. As soon as you remove the sample from vacuum the surface will of course oxidise. Oxidation cannot be avoided, however if you deposit an As cap of sufficient thickness before you remove the sample only the As will oxidise and your III-V epilayer surface will remain perfectly intact. Once the sample is safely bask inside a vacuum system, one need only heat it to ~300 °C and the As sublimes. A quick look with RHEED will show the original reconstruction is even preserved.

Amazing.

How does one deposit an As cap?

You can deposit an As cap on any surface. In this example I will discuss depositing an As cap on GaAs(100). At the end of your epilayer growth you will typically observe an As-rich (2×4) reconstruction. If you want to know what a reconstruction is see my What is a reconstruction? post. Essentially to deposit a cap all you need to do is fully open the As and turn off the sample heating. However there are a few things to consider:

Firstly, As₂ tends to create a better cap than As₄, so wherever possible you should use As₂.

Secondly, the As cap grows very slowing whilst the substrate is above 100°C and it is preferable to cool the substrate down to 0°C in order to rapidly form a cap.

Thirdly, not only As but every other background species will readily condense onto the substrate at 0 °C including the hydrocarbon, water vapour, oxides and other undesirables lurking inside your vacuum.

In order to ensure the highest purity of your surface, it is a good idea to pre-protect the GaAs by creating an even more As-rich reconstruction than the (2×4). The reconstruction of choice is called a c(4×4). The c(4×4) is simply created by holding the substrate at 500 – 540 °C  in a moderate As flux for several minutes.

Test it yourself. Align the RHEED along the [-110] azimuth so you can see the 4x of the (2×4) shown in Figure 1a and then lower your substrate temperature until the 2x pattern of the c(4×4) appear as shown in Figure 1b. You may be asking yourself why a c(4×4) reconstruction has a 2x pattern. Well if you look closely at the reconstruction in Figure 1b you will see it is not simply a 2x, it is a pair of 2xs. The 2x at the top is out of phase with the 2x at the bottom (i.e. the lines on the RHEED screen do not line up). The 2x at the bottom is in fact exactly half way between the 2x at the top.

Rotate the sample to the [110] azimuth. You will see the exact same pair of 2x reconstructions there too. The c(4×4) looks the same along both [110] and [-110] azimuths. What is going on? Well the reconstruction’s unit cell is not aligned to the same directions as the (2×4). It is in fact centred on the original [100] and [010] directions of the (001) surface, which means it is at 45° to the (2×4) reconstruction.

The c(4×4) is composed of 1.75 ML of As on Ga. That  is a full ML of As plus another ¾ of a ML of As on top of that. This is significantly more As than the ¾ of a ML of the (2×4) reconstruction. The As of the upper ¾ ML of the c(4×4) are back bonded onto the As of the lower full ML, meaning that the As on the upper layer “sees” no Ga at all. Which more importantly means that the Ga in the buried layer is protected from contamination. The As dimers of the upper ¾ ML  arrange  themselves 3 abreast in a “brickwork pattern” and it is this arrangement that gives the pair of 2xs on the RHEED screen.

The c(4×4) is therefore perfect to pre-protect the epilayer. Once it has formed, simply ramp the substrate down to 0 °C  (or as low as you can go) and watch the RHEED. To get a more even As cap you can start the substrate rotation again now.

Once the substrate gets down below 100 °C you should find that the 2x is replaced by the 1x pattern shown in Figure 1c. This is because a thin amorphous As layer now exists on top of the c(4×4) and all you can see are the bulk lattice rods. As the amorphous As layer gets thicker, the RHEED beam can no longer penetrate through to the substrate and the RHEED pattern becomes the amorphous haze shown in figure 1d. This is probably somewhere around 5 – 10 nm thick. It may take >30 minutes from the moment the ramp down was started to the time the amorphous pattern is completed, depending on the background sample heating and the rate of cooling you can create inside your particular MBE chamber.

The As growth rate is highly temperature dependent. You want a final thickness of about 15 nm and from experience this takes about an additional 5 minutes after the amorphous RHEED has appeared. In order to get a more accurate estimate of the As cap growth rate, you can first create the c(4×4), then ramp down to 400°C, then turn off the As flux and leave the sample to equilibriate at 0°C  for half an hour, then apply the As flux and monitor the time it takes to create the amorphous pattern, then triple it. If you want to be really accurate with your cap thickness I suggest you perform cross sectional SEM on the sample and extract the cap thickness.

The As cap is not only useful to protect your epilayer ex situ, but can be used to perform two temperature calibrations at ~300 and ~400 °C  (in order to do that see my Making a static reconstruction map post.

How to grow your first sample: (2×4)/(4×2) transition

Faebian Bastiman

The (2×4)/(4×2) RHEED transition is a very useful flux calibration point, since it tells you when your As and Ga fluxes are equal. I described what a (2×4) and a (4×2) reconstruction are in my What is a reconstruction post. In this post I will explain how to spot the transition with RHEED.

First of all you need to produce a good starting (2×4) surface. In order to do that follow the steps in my Oxide remove post. With the Ga shutter closed and an As overpressure incident on the sample you should see the As-rich (2×4) reconstruction on your RHEED pattern (Figure 1a). How can you be sure this is a (2×4) and not a (4×2) in disguise?

You can be 99% certain that if you are annealing at a temperature around 15 °C below the oxide remove temperature and the As flux that you are supplying is sufficient to grow GaAs that this is a (2×4) reconstruction. You can test very simply by stopping the rotation so you can see the 4x on the [‑110] azimuth and reducing the temperature. Reduce the temperature until a 2x appears. Rotate to 90° to the [110] azimuth. Still 2x?

If yes, then this a c(4×4) and the early reconstruction was a (2×4). Return to the original temperature and retrieve your (2×4). Note if you want to know what a c(4×4) is read my As cap post.

If no, and you are looking at a 4x, then this is now a (2×4). Remain at this temperature.

If no, and you are also not looking at a 4x, then something is very, very wrong with your sample. Are you sure it is GaAs(100)?

Assuming you have found your (2×4) make sure it looks like the one in Figure 1. If not change your temperature by 5 – 10 °C and your As flux by 10-20% and try to get the best (2×4) you can. You will note that in Figure 1a on the [‑110] azimuth the 4x rods are fairly short and bright even if the 2^nd order rods are not very clear. The 2x on the [110] azimuth is simply a 2x and cannot give you any quality information at this point.

Try opening the Ga shutter. Does your 4x change to look like Figure 1b? It should. The Ga flux breaks up the static (2×4) and creates some disorder. This disorder results in an elongation of the 4x rods and an overall decrease in the intensity. The 2x on the other hand does not really change. You can consider Figure 1b as the dynamic (2×4), where “dynamic” means “growing”.

We are about to search for the Ga-rich (4×2) pattern. In your first few attempts do not rotate the sample, but rather keep the RHEED pointing along the [110] azimuth and looking at the 2x pattern. During the search the 2x will change into a 4x on [110] and the 4x will change into a 2x on [-110]. You can see the Ga-rich reconstructions in Figure 1d.

The problem with performing the transition without rotation is that you are only seeing the transition at the point the RHEED beam is hitting with the Ga and As flux pattern present at that point. This As/Ga ratio can be very different from the one created whilst rotating, since when you rotate you create a more even flux across the wafer. On the other hand since both the starting and ending reconstructions have both a 2x and a 4x azimuth it is not so obvious when the transition has happened until you have some experience looking at it.

So with the rotation off, the Ga shutter open and the 2x on the [110] azimuth on the screen close the As shutter (or close your cracker valve if you do not have a shutter).

Am I serious?

Yes.

Close the As shutter for a few seconds and watch the RHEED, it should quite rapidly turn into a 4x like Figure 1d. Once it does open the As shutter again and make sure the 2x returns. When the As flux is greater than the Ga flux, you see the 2x of the (2×4) when the Ga flux is greater than the As flux you see the 4x of the (4×2). In the extreme case we just witnessed the Ga was of course practically infinitely bigger than the As flux.

Ok step 2 is a little less brutal… with the rotation off, the Ga shutter open and the 2x on the [110] azimuth on the screen… half the As flux. Wait 15s. Did the 4x appear? If yes return to the original As flux, if no half the As flux again. What you are doing is performing a binary search. You take the lowest As flux you know gives you a 2x and the highest As flux you know gives you a 4x and you apply the average of the two and see if you get a 2x or a 4x. Always wait the same 15s between flux changes for consistency.  

You may notice on your search that the 2x creates a rather dim 3x rather than the desired 4x (Figure 1c). This is a mixed reconstruction between the (2×4) and (4×2) that many people overlook. It is weak on both [110] and [-110] azimuths and hence probably lacks the long range order of the (2×4) and (4×2) reconstructions. The low intensity makes it difficult to say whether it is a (3×1) or a (3×4) or a (3×6) and indeed the physical surface of the GaAs could be made up of domains of each.

Once you have found the maximum As flux that gives you a 4x on the [110] in 15s, return to the original As flux that gave you the 2x on the [110]. In the third step gradually reduce the As flux from the start value toward the value you found and watch the RHEED pattern evolve. Double check the As flux you believe gives you a 4x on the [110] is reproducible and get familiar with the process.

Finally set the substrate rotation to around 0.2 revolutions per second and perform the third step again. The As flux required to create the 4x on the [110] azimuth may be a little different this time. With the rotation on you will see the 2x on the [-110] azimuth for the first time. Take a closer look at the reconstructions in Figure 1d.

The dynamic 4x you get when you are Ga rich is much sharper than the dynamic 4x you get in Figure 1b. This is because that whilst the Ga breaks up the As-rich (2×4), the Ga only ever improves the Ga-rich (4×2). Hence the dynamic 4x of Figure 1d is similar to the static 4x in Figure 1a. Note too that the 2x in Figure 1d is very different. There exist some extra bright spots at the bottom of the image that are much dimmer in the 2x of Figure 1a and cannot be seen in Figure 1b.

Hopefully you are now familiar with the process and can readily ascertain the maximum As flux required to create a (4×2) reconstruction in 15s. This is called the “1 to 1” point. The point at which the As flux is approximately equal the Ga flux. You could argue that the Ga flux is slightly higher than the As flux, since the surface turns Ga rich; however they are approximately equal. More importantly this transition is very repeatable. You need to make sure you always do the test in the same way. i.e. you must always wait 15s and always do the test at this temperature. Otherwise the flux will vary from check to check. It is worthwhile observing that variation yourself.

Once you have found this point you can calculate your fluxes in the system independent units of atoms/nm²/s (see Flux determination post) and state the atomic fluxes in any journal articles you write.