A couple of months ago, we took a look at one way of using MEMS cantilevers to detect gases. Our focus was on the optics used to read the status of the cantilevers, but, however detected, the use of cantilevers to measure concentrations of substances is common – at least in research papers.
Since then, this topic of detecting… something (officially called an “analyte”) has come up several times, each with a different twist or approach. So this week we follow up with variations and alternatives to the cantilever theme. They’re all different: one of the take-aways is that this cat can be skinned lots of different ways.
Who needs a straight line??
We’ll start by riffing off of the cantilever theme. The “normal” way of using a cantilever is to “functionalize” the top surface – that is, put something on there that will attract your target analyte. When the analyte floats in to say hello, it sticks to the cantilever and changes the cantilever’s behavior.
That behavior might be static – the otherwise-straight “diving board” might curve up or down slightly. Or it might be dynamic – detected by getting the cantilever resonating and measuring changes in that resonance in the presence of the analyte.
But a common denominator in all of this is that the cantilever is operating in a linear regime. It’s acting like an ideal spring, exhibiting harmonic resonance. Using a dynamic approach, you look at changes in the resonant frequency to detect the presence of gasses.
But a recent paper asked the question, “What happens if we go outside the linear regime?” When you apply a force to a cantilever, you have to be gentle to keep things linear. The tip can’t travel too far; outside of that range, the geometry and materials properties create non-linear behavior. That’s normally considered bad, but… is it?
Of course, “going non-linear” can mean a lot of things. (It can even be the engineering equivalent of “going postal.”) In fact, it literally means doing something – anything – except acting linear. So, by itself, it’s not particularly descriptive. A research team from Delft and Cambridge Universities investigated what happens to the amplitude when you sweep a driving frequency from below the natural harmonic resonance (in the linear range) to above it, working with different driving voltages – ones that would drive the cantilever within and above linearity.
Within the linear range, as you might expect, the resonant frequency* stayed pretty much in place as drive voltage changed. But at a certain critical voltage, the behavior became bistable: amplitude would decrease with increasing frequency, but would then increase when the frequency came back down, finally snapping back at a frequency far lower than the frequency at which it went non-linear, as shown in the following figure. (I’ve intentionally focused only on the shape, leaving numerical details to the paper.)
What becomes interesting about this is the position of the frequencies at which the “bifurcation” and return happen. This showed some promise as a way of detecting an analyte – in their case, ethanol.
Such non-linear responses have value to the extent that they end up spreading out measurement points so that it’s easier to resolve a potentially noisy answer. They found that the movements of these two voltages were three times that of the resonant frequency in linear mode. This makes for less ambiguous detection.
It’s one paper, and it’s far from commercialization, but, at the very least, it might help non-linear behavior to escape from the penalty box.
From Dusk to Radon
Meanwhile, as we ensconce ourselves safely in our homes for a stress-free slumber far from the perils of the outside world, an unseen, unsmelled peril may, in fact, be pervading our space. Depending on where you live, naturally occurring radon is a real issue.
You can test for it, but this tends to involve expensive equipment with an ionization chamber run by professionals. There are solid-state alternatives, but, according to a new company, RSens, their lower price is matched by lower performance. RSens is proposing a smaller, cheaper, more accurate radon sensor that homeowners could keep in their homes for constant monitoring.
“But why do you need constant monitoring once you know how much radon there is?” you might ask. Or, I asked, anyway. The idea is that, depending on the “connection” between the house and the ground, as breezes blow and windows open or close, radon can build up or deplete. In the longer term, you could connect a sensor like this to your home (or business) ventilation to, for instance, turn on a blower if the radon concentration gets too high, venting it to the outside, perhaps.
How they propose to do the detection involves a very different approach from anything we’ve seen. Radon decay causes emission of alpha particles, and, rather than detecting radon outright, it’s the alpha particles that are detected at a signature energy of 5-7 MeV.
Striking silicon, they can cause an electron/hole pair to be formed. So the RSens detector, dubbed Aria, has a matrix of good-old-fashioned NPN transistors. The high-resistance collectors are all connected to VCC, and the emitters drive an A/D converter. The trick is that the bases are floating, and it’s the cumulative area of this floating base that is the real detector. When an alpha particle strike creates electron/hole pairs in the base, this tiny current is amplified (remember beta?) into a larger current through the emitter.
These guys are in the final stretch of their Indiegogo campaign, so it’s an open question as to whether or when this will be showing up in a model home near you. But, as a sensing approach, it certainly represents a departure from other techniques we’ve seen.
Next is a company that came to my attention at the recent MEMS Executive Congress. Called Cambridge CMOS Sensors, they’ve taken something of a detoured route to where they are now. They started out making hotplates. No, this isn’t yet another restaurant venture (and the nice folks up north shouldn’t confuse “hotplate” with “hot dish,” ya know?).
A hotplate is, fundamentally, an infrared emitter. It’s used in high-accuracy gas sensors (“Non-Dispersive Infrared Sensors,” or NDIRs). Picture a tube with the IR emitter at one end. An opening lets gas into the tube, and, at the other end, there are optics and filters and such that capture a particular wavelength and deliver it to a detector. Such a module is typically designed to detect a single specific gas.
But they also realized they could take advantage of another phenomenon: certain metal oxides (MOXes), when heated, catalyze reduction and oxidation (“redox”) reactions of specific analytes at specific temperatures. Those reactions will change the resistance of the MOX. The nature of the reaction and the optimal temperature will depend on the MOX and the analyte. So they can put a MOX layer over the hotplate and use it as a solid-state sensor.
The drawing above is very similar to one they created, but I added a couple of details. The thing is, they’ve got “CMOS” in their company name for a reason: they build their sensors using a standard CMOS process (although the economics are such that they still build their ASIC separately). But you might look at the figure and think hard to come up with a MOX layer that’s part of a normal CMOS flow. And that’s because it’s a departure – but an additive one that comes at the end.
The tungsten hotplate, gold detectors, and aluminum contacts are all standard CMOS stuff. What’s missing is the MOX. That’s added afterwards, in a separate room or building, by using an ink. In fact, because an ink can be applied selectively, they can build an array of hotplates with a variety of MOXes to create a multi-analyte sensor. And they can heat from 0 to 600 °C in around 25 ms, so they can run a temperature profile quickly to detect a variety of substances at the appropriate temperatures.
Of course, because they’re adding the ink elsewhere, they have to passivate first. That means they can’t measure the resistance of the MOX directly; they have to use an AC approach to measure the resistance through the passivation layer. It also means they can use this both in gaseous and liquid environments (assuming suitable packaging).
Why go through all this trouble when you could use cantilevers instead? This appears to be a more precise approach, competing not so much with other MEMS sensors, but with bigger, bulkier industrial sensors like the NDIR modules.
A Hyper Spectre
Next, a brief stop at a topic we’ve covered in the past: hyperspectral imaging – yet another way to detect materials. VTT, a Finnish research house, also presented at the MEMS Executive Congress. They have built a variety of spectral detectors based on Fabry-Perot interferometry. This is an approach that bounces light between two mirrors to measure its frequency.
In the MEMS context, they use Bragg mirrors – alternating layers (such as are found in masks intended for EUV use, which requires all reflective optics). A tensioned membrane forms one mirror; it is suspended above the surface, onto which the second mirror is deposited. The airgap in between determines the frequency to be detected (as shown in the highly abstracted figure below).
Because the airgap distance determines the filtered frequency, a simple version as shown would hit only a narrow range. But they’ve got a platform involving a piezoelectric actuator that lets them tune the airgap on a single device.
When bugs are a good thing
Finally, something completely different: using microbes (bacteria, yeast, algae, etc.) as your sensors. I ran across an interesting survey paper that was both an overview and more information than I could ever digest on recent biosensor developments.
In general, the idea is that you take the microbe and somehow “immobilize” it (hopefully without damaging it); you then measure something that reflects the microbe’s response to a target analyte. Yeah, that’s pretty vague. You might be looking for a protein or, preferably, an enzyme created by the microbe.
There are numerous detection approaches, each with pros and cons, many of which have been around for years. Some techniques are electrochemical (amperometry, potentiometry, conductometry, voltammetry, and microbial fuel cells) and some optical (fluorescence, bioluminescence, and colorimetry).
Apparently the downside of using microbes has been the fact that they haven’t been as specific or sensitive as might be desired, and the reactions have been slow to evolve – making it hard to do an “instant read.” But genetic engineering (which plays a role in a number of the measurement techniques) is also improving the response of microbes to specific analytes, either by creating a better response or by using arrays of microbes and measuring the “fingerprint.”
Nanotechnology is also helping – in particular with micro-electrodes and microfluidics. For some techniques, miniaturization has been difficult; these new manufacturing techniques have allowed that barrier to be breached.
Finally, there’s this immobilization concept. The idea is that you need to pin the microbes down – by trapping them in goo or encapsulating them or any of a number of other techniques – all of which can compromise the microbes or hamper measurement. But it was noted that a newly-recognized microbe, Caulobacter, naturally has adhesive abilities that let it build a monolayer with no external substances. Its genome is known, so it can be modified for various purposes.
This is a whole separate world, of course, but as the biosensors improve and become more readily available, there may be less of a divide between them and the more familiar (to us) sensors.
So there you go. Five completely different approaches to complement the one we’ve already seen:
- Functionalized cantilevers, linear (from before)
- Functionalized cantilevers, non-linear (research)
- What’s effectively a big NPN transistor with a giant floating base (Aria)
- Hotplates and MOXes for redox reactions (Cambridge CMOS Sensors)
- Bragg mirrors for hyperspectral imaging (VTT; Imec, from before, has their own approach)
- Microbes
Some are still in research; others are available now. I’ll be curious to see how well they all coexist in different applications… or whether they’ll ultimately duke it out in a dramatic Götterdämmerung.
*Fifth flexural mode, for anyone keeping track…
More info:
Nanomechanical gas sensing with nonlinear resonant cantilevers (behind paywall)
RSens Aria radon detector (the only site appears to be the Indiegogo campaign)
Microbial biosensors: A review (behind paywall)
Will all these ways of detecting substances continue to coexist? Or do you see some winning or losing?