The Transistor-Injected Quantum Transition Laser: A Novel Mid-IR Source
Sep 5, 2023
"The Transistor-Injected Quantum Transition Laser: A Novel Mid-IR Source"
John Dallesasse, Professor of Electrical and Computer Engineering, University of Illinois
CNST Nanotechnology Workshop 2013
Unlike the previous speaker who's been 14 years into his topic, we're actually on the
other end of that, just at the concept phase.
So today what I'm going to be talking about are some new device concepts that we have
for a device that we're calling the Transistor Injected Quantum Transition Laser, which combines
features of the transistor laser developed here at Illinois by Professors Holonyak and
Feng with devices from Bell Labs, particularly faced in Capasso's group, the Quantum Cascade
Laser, which has been very important in the mid-IR, and I'll talk a little bit about that
as we go forward.
I'd like to acknowledge the grad student that I have involved in this particular project,
Kano Chen, she had a paper, a poster out there if anybody is interested in some further
details on some of the things that we're looking at with regard to this and some of the modeling.
But obviously I'd like to thank my group, entire group as well, which is still fairly
small.
With regard to the topics I'd like to cover today, I want to briefly talk about the motivation
for this.
Obviously in a conference like this, people understand the importance of the mid-IR, but
highlighting a few of the reasons why we're interested in this wavelength range is important.
I want to talk a little bit about, again, really the inspiration for this device, which
was the transistor laser and QCL.
And then we'll introduce the basic concept of the Transistor Injected Quantum Transition
Laser and talk about what we think some of the advantages of this device are going to
be.
Finally, we'll talk a little bit about some of the efficiency improvement estimates that
we've been working on and then summarize.
So with regard to the motivation for this, again, the mid-IR region of the spectra is
an incredibly rich region for chemical transitions, and as a consequence, any mid-IR source is
very useful for a number of applications, biomedical imaging, chemical sensing, chemical
process gas detection, greenhouse gas monitoring.
All of those applications become very important, especially for a device where effectively
you can move on to a given or around a given chemical transition and dither a device wavelength
around that in order to effectively look for amplitude changes in either specific wavelength.
And that's one of the things that we feel that we can do with this device, and rather
than relying on device heating, which is a slow process and typical more for the quantum
cascade laser, we can actually do it by a fairly rapid modulation of what ultimately
is our base collector junction in our transistor.
Obviously, there's military applications as well.
We like to look at dual-use technologies.
Mid-IR is also very important for things like defense against manpads, surface-to-air missiles.
It's potentially useful for producing atmospheric links for ground-to-air and air-to-air communication
in the, again, atmospheric passband windows.
And of course, chemical detection in the IR is also important for things like explosive
detection in homeland security applications.
So with regard to the background of this, there's really two devices that ultimately
were the inspiration for looking at this particular structure.
The first device was the transistor laser.
And for those of you not familiar with that, back in around the 2004 timeframe, Professor
Feng here at Illinois was looking at very high-speed HBT devices and also was working
very closely with Professor Holonyak.
What was recognized was that recombination, which is a natural part of the operation process
of any sort of bipolar transistor in a direct-gap material, produces light.
And those photons are useful and can be used.
So what they did was made a modification to a traditional HBT structure.
They added a quantum well into the base region, which allowed somewhat more efficient collection
of the fast-recombining electron-hole pairs.
The slow-recombining electron-hole pairs, or slow-recombining electrons, effectively
get swept out in the reverse-biased base-collector junction of the device.
As a consequence, you have a device that can be modulated to very high speeds, and they've
demonstrated for transistor lasers up to about 40 gibbets per second, relaxation, oscillation,
free oscillation, with no need for external modulation.
So it's a very interesting device, a device that is still early in its, I suppose, development
or understanding of what can be done with it.
But in and of itself, it's an interesting device and something that my group is also
doing some work on as well, because it could actually be an interesting functional unit
for systems that require both electronic and photonic interactions.
The quantum cascade laser, of course, is another very interesting device.
It kind of represents the cutting edge of epitaxy in the sense that you're producing
very complex quantum structures where you're actually getting into the mechanics of the
recombination process.
We're engineering the electron wave functions in this structure such that we can produce
the optimal interaction between the quantum state in one region and the quantum state
in another by effectively varying the thicknesses of our layers.
We're able to create the energy separation of those quantum transitions, and we also
are working with fundamental properties such as state recombination lifetimes.
By varying the quantum overlap between one state and the next, you effectively are able
to obviously vary or control the time that that recombination event is going to occur.
You're really, unlike a semiconductor laser or a band-to-band process where we're sort
of stuck with what we have, in the sense we can vary our quantum well thicknesses and
things like this, here we're actually engineering the physics area.
We're using physics to engineer the recombination process itself.
It's a very interesting device, but it has some problems.
The biggest problem with the device is that it's a two-terminal device.
We have this structure that we're producing where we're relying on quantum transitions
that we've engineered into this device, but when we change the bias voltage, we're changing
the field across those regions.
Changing the field changes the overlap between effectively one state and the next, so we're
effectively changing the fundamental characteristics of the device, the state lifetimes and the
transition energies as we change the bias on this device itself.
It's an N plus IN device, so from a semiconductor laser person's perspective, and some people
cringe when I say this, it's a resistor laser, but still a very interesting resistor laser.
What are some of the challenges with the QCL?
Again, I mentioned one of these a moment ago.
Effectively, if you look at what's happening as you climb your LI curve of the device,
you're effectively changing the band bending again, which is changing both your state recombination
lifetimes as well as the recombination energies.
The net result of that is you have a very strong field dependence of the population
version of the device.
Effectively, there's a sweet spot where these things operate, and then outside of that sweet
spot, there's a rapid deterioration of the performance of the device.
Another huge problem, which was actually talked about earlier today by Katerina, is the problem
of free carrier absorption.
In the mid-IR portion of the spectra, you're in a range where basically using even just
the simple Druda model, which begins to ... You actually have more complicated processes
going on in P-type material than N-type material, because obviously you've got multiple sub-bands
present.
You have heavy holes, light holes, your split-off band.
In the simplistic sense, if you go back to the simple models and data from people like
Spitzer and Whalen, the dependence of free carrier absorption on wavelength goes as effectively
the power of the wavelength.
You have a very strong increase in free carrier absorption that occurs in the mid-IR and longer
wavelength portion of the spectra.
As a consequence of that, you'd like to be able to come up with some way, if possible,
to defeat that.
In thinking about those challenges, and again, having the advantage of simultaneously thinking
about QCLs for some other projects, and also looking closely at the transistor laser work
that was going on here, the idea basically was generated to utilize the transistor effect
to inject the carriers into our cascade region, and to effectively stabilize the field in
our cascade region by effectively putting it in the middle of a reverse bias-based collector
junction.
Effectively, what we do is we have a transistor or an HVT.
We have a standard base junction.
We have a collector junction.
We have an intrinsic region that contains our quantum cascade structure.
We apply a reverse bias to that structure, which basically applies the field.
We engineer the doping in both the base region and the collector region to create a depletion
region around that quantum cascade region, which basically strips out the carriers that
are contributing to free carrier absorption.
By engineering our doping levels and optimizing or doing a trade-off between our base collector
voltage and the desired amount of free carrier absorption that we want to remove, we can
basically generate a device that has much better power efficiencies in the end, or much
lower optical losses as a consequence of power efficiencies than the quantum cascade laser
has.
These are basically some of the diagrams showing what the structure looks like for an edge
emitting device, where if you look at the figure here, we basically show the overall
structure.
This is an edge-on view, again, of an edge emitting device.
We have an emitter structure, which is here.
We have our base region, which is here.
Obviously, as we inject a small amount of base current based upon the beta of the transistor,
we then pull in a large number of electrons from the emitter contact.
They flow across the base.
They're collected in the reverse bias base collector junction, undergo the quantum transitions,
which provide gain, and then effectively are pulled out.
In this structure, although this isn't key to the concept itself, we're using some other
techniques, processing techniques, to help do mode shaping, lateral oxidation of materials,
which is commonly used in vertical cavity surface emitting lasers, allows us to effectively
put fingers of high resistive, low refractive index material into the structure in places
that we want to put it.
Those fingers basically allow us both to do optical mode shaping as well as to control
where we're injecting the current in the device.
Again, the basic operating principle is the operating principle of a transistor, an NPN
transistor.
Again, this is showing the band diagram and what that looks like under your typical bias
condition.
What are some of the advantages of this device?
There's advantages that depend upon how we choose now to bias that three terminal structure
that we now have as opposed to effectively the two terminal structure of a standard QCL.
The first one is wavelength tuning.
In this particular instantiation or application, what we would do is we would have a common
emitter configuration where effectively we use one circuit to control the base current
and as a consequence, the laser power.
We have a fixed laser power or a fixed current going in.
Then we would have a small dithering circuit placed on the reverse biased base collector
junction to cause very small changes right around, again, in regions where we still have
a good overlap between the different wave functions that are contributing to the state
lifetimes.
But, again, by having a small dither, we're effectively able to produce a little minor
sweep in the wavelength right around a given part of the spectra.
That can be done very rapidly.
We can also modulate this device at high speed.
The N plus IN QCL, which basically has a very high operating voltage, while the fundamentals
will allow that to operate very rapidly, there are some practical things which cause that
to be problematic.
Here we can basically use a common base configuration where, once again, we either fix or dither
the laser wavelength and then have a separate control which basically RF modulates or digitally
modulates the emitter base junction to effectively provide amplitude modulation.
Effectively, this sort of configuration allows us to do amplitude modulation with one circuit
or slight wavelength modulation with a second circuit.
Again, it allows perhaps some interesting modulation schemes to be employed similar
to phase modulation of things that are being used in long-haul telecommunication systems.
This application would largely be more like free-space optical communication.
The other advantage of the device is in the area of free carrier absorption.
How do we minimize free carrier absorption?
I mentioned this earlier.
This graph really helps to illustrate how that occurs.
We effectively have our cascade region here, which we have under reverse bias.
Under reverse bias, we create a depletion region around that cascade region.
We have a forward bias circuit which injects our electrons through our base region and
then through our cascade region.
This can be by having a slightly more lightly doped P-type material.
Again, we're able to reduce the doping significantly over the N plus layer that's required in the
QCL.
We can also play with our doping density here, but what's most important is that we create
this depletion region around the structure.
If we create a depletion region around the structure, there are no free carriers to contribute
to the free carrier absorption process.
We're actually able to improve the efficiency as well.
This is an additional advantage beyond the rapid wavelength modulation and ability to
high speed modulate.
That efficiency impact can be measured by looking at parameters like high power operation
and wall plug efficiencies.
There's also, again, a gain stabilization advantage related to basically, again, fixing
that base collector voltage and applied field.
Again, we're still in the preliminary stages and further modeling is being done, but in
some simple 1D modeling that we did, effectively looking at the mode shape overlap with the
regions of the crystal that would have free carrier absorption and also applying that
across the various metrics that are important for the quantum cascade laser, threshold current
densities, DQEs, total power improvement.
We've come up with some estimates with regard to what we'd see just from a pure efficiency
improvement of the device, ranging from something like a 25% reduction in threshold current,
again, per the simplistic model that we have, all the way up to, again, with also doing
some mode shaping to effectively reducing the threshold current by two thirds, improving
slope efficiency by as much as a factor of two and wall plug efficiency by as much as
a factor of two.
Again, we're still taking these simple one dimensional models and extending them and
doing full two dimensional models and looking at the overlap of those modes with the doped
material in the structure and in particular, the P-type material is problematic because
there's not a lot of data in the literature right now on mid IR absorption for lightly
doped P-type material, so we're having to do some estimates based upon the data that
is available, but again, we think that these are reasonable and what could be expected
for this sort of device.
With that, we've basically introduced the concept for this device, reviewed some of
the projected advantages, and then showed an estimated improvement performance for that
device as well.
Thank you very much.
Questions?
Really interesting.
I have two questions.
You mentioned the possible, you know, infrared spectroscopy, and so I wondered what do you
think the tuning range would be in the relevant infrared range?
Sure.
The second question is, I didn't understand when you have the common base configuration
how the base emitter voltage or current actually amplitude modulates.
Okay, let me answer the second question first because it's easier than the first one.
So basically, in that configuration, it's effectively the base emitter, you know, the
voltage you're applying to the base emitter junction that's controlling the amount of
current that's going through, you know, your transistor, so it's a pure transistor effect.
So that one is, you know, straightforward.
With regard to the other, you know, that there's sort of an element of sort of it depends
to that.
I mean, you know, so we can achieve limited wavelength tuning by effectively focusing
on just doing a small voltage modulation in and around, you know, the particular wavelength
that we're interested in at, you know, at sort of a fixed temperature.
So that's going to be relatively small, probably enough to span either side of an absorption
peak, you know, but not much more than that.
We would have the ability to temperature tune as well.
So if we wanted to look at separate peaks, we in principle could move, you know, the
overall structure to a different wavelength or to a different temperature and then, you
know, again, do the same dithering there.
You know, or if we began to look, you know, and there's a lot of work in the literature
now on, you know, on multi-wavelength QCL structures, you know, where effectively you're
putting different quantum transition regions in to broaden your gain spectra.
If we do something like that, which is down the road, we effectively could build any structure
into the reverse bias base collector junction and potentially use something, you know, again,
this is now way far out, you know, integrate this gain structure perhaps into, you know,
a silicon tuning block effectively, you know, where effectively we do something like a Y-branch
tunable laser that uses the additive Vernier effect as an example to pick out different
wavelength transitions with a broadly tunable laser.
I mean, there's a company that's done that in the telecommunications wavelength range
where they've integrated a standard 3-5 gain media with tuning arms that have been integrated
into effectively a silicon photonic structure in order to, you know, realize a broader tuning
range.
So, I know that's a long-winded answer to your question, you know, but I think that
down the road, there's more functionality that we could have.
In the short term, it would just be, you know, enough to maybe sense, you know, a single
peak.
Let's thank our speaker.
We'll all be happy.
Thank you.

