Tip:
Highlight text to annotate it
X
So, moving on towards lighting.
we have some other aspects we need to consider then for
displays and some of the other more early applications that I mentioned.
The ifrst thing to keep in mind is that there's enormous opportunity
here for the beneifit of the world in terms of energy savings.
Which frankly is a better story than almost all
the ones you'll hear, in terms of energy efficiency.
because not only can LED's have a
dramatic impact on reducing our global carbon footprint but it can be
cost effective because the technology has so many advantages over the incumbent.
Everybody can do okay.
The suppliers can do okay, the user can do okay,
and the environment can benefit if we adopt LED technology.
so this projection is by the Department of Energy.
in 2010, roughly 700 terawatt hours of energy
was consumed towards lighting. and
[INAUDIBLE]
electricity usage for lighting.
And that can be reduced, according to their model, which
was a fairly conservative adoption curve, to about half by 2030.
That's a big deal.
There's not many technologies where you can literally cut consumption by half.
And this is really what we're doing by moving from the, you know, vacuum
tube like era of conventional lighting to
something much more controllable and much more efficient
inherently like LED technology.
[COUGH]
to give you an idea what that
means because nobody, you know, 300 terawatt hours
[LAUGH]
what's that mean?
It's kind of the equivalent, it's roughly
the equivalent of 50 1 gigawatt power plants.
50 coal-fired power plants that what you find
is that the incandescence, even though they aren't
the largest volume of lamps because commercial things
tend to, commercial applications tend to dominate the residential.
They're huge energy hogs. And so that even though they are not as
common, the fact that they're so inefficient is
that there are huge advantages to replacing them.
also, you remember the, the road map I showed before
had LED's going up to 150, 200 lumen per watt, right?
So even then, now you're talking
about replacing fluorescents to tremendous advantage.
Another thing that's not often talked about but for example, this is a great
[LAUGH]
example here.
this fluorescent tube which you can't see on camera
but is right above me here has a raw
[UNKNOWN]
efficacy of something like 80 lumens per watt.
It's extremely poor at putting its light where you need it.
So if you calculate the utilized efficacy, there's a luminary
utilization efficiency we, do we apply to a luminaire like that.
It's roughly on the order 50% or so.
So the, the overall system efficacy can be literally
half of what the raw lamp of this efficiency is.
And this is something that, where LED's have a huge advantage because the source
is so small, so directed, you can put the light exactly where you want it.
So, there's this other factor of two or so, that aren't
really captured in these plots that we're talking about, the LEDs.
[SOUND]
so energy consumption's opportunity is is a big deal, reduction's a big deal.
You know, 50 power power plants, $30 billion a year, 200
metric tons of carbon emissions per year.
And if you look at the world, the total savings are about four times higher.
The whole, the world gets on this.
So there's an enormous opportunity here to
convert over for the betterment of everyone.
we go now into technology a little bit.
LED's are made very similar to the way silicone IC's are made.
The same kind of silicon
[UNKNOWN]
processing different materials. Simply start with a substrate.
you deposit by there's, there's two common ways to deposit
[UNKNOWN]
metal organic chemical vapor deposition where you take precursors
like trimethylaluminum, trimethylgallium and mixed
with in this case ammonia to form gallium nitride along with gallium indium nitride.
molecular beam epitaxy is another way to deposit materials,
more reserved for the university or aca, academic research.
It doesn't quite have the
[INAUDIBLE]
capability of MOCVD.
So, most of your LED's today are, used using MOCVD
technology to deposit the layer structure to make the diode.
After we have the layer structure down,
we can go through standard wafer fab techniques.
again, very similar to what you might do with an IC.
Except there has to be
special consideration for optical issue.
These are optical devices, we have to make sure we get all the light out.
And that ends up being a big focus.
And I'll go into that some detail. the
[INAUDIBLE]
are then singulated by various techniques
and packed, packaged into LED specific packages.
These are a couple of those from some of the
top tier suppliers for the, at the component levels today.
and then I'm going to luminaire. for a retrofit luminaire or maybe some
[INAUDIBLE]
built-in new fixture depending on the product.
go to market strategy of the company in question.
typical InGaN-GaN LED structure looks something like this.
So here you have again, the conventional approach is like a sapphire substrate.
The nucleation layer has grown to form gallium nitride.
gallium nitride is not lattice matched to sapphire, so you can
imagine the dislocation density in the crystal is very high and
that's true.
There are tricks to reduce the dislocation density that
almost the, little bit all of the, the LED suppliers
use to try to reduce the dislocation density on
the order of 10 to the 8 per square centimeter.
So that's about one dislocation for every square micron.
that's kind of the good enough point, where you can make reasonably good LED's.
not as good as you could, and I'll talk about that a little bit more later.
Your n-type layer,
the quantum wells, which are the active emitters.
They tend to be 2 or 3 nanometers in thickness
and with the indium nitride composition shows you to choose
[UNKNOWN]
the color that's required.
typically there's an asymmetric injection situation.
The whole masses are much higher than electron masses and you need some
sort of balancing activity to keep
the electrons and hole of recombining together.
And so typically, you have what's so-called
electron, AlGaN electron blocker layers growing on
top of the active region here before you finish off with the P-GaN material.
Electrodes are applied and if it's an epi-up device like this one, you need some
sort of current spreading technology because P-GaN
magnesium-doped p-type GaN tends to be very resistive.
And so you need some help to spread the current there.
look at the energy bandgap profile.
Here's the valence band, a conduction band.
you typically have, you know, the n-type layers with holes inject
electrons injected this way and then holes injected the other way.
The object is to get them to overlap and make light.
we talked about internal efficiency.
internal quantum efficiency has really two components.
There's an injection component and a radiative component.
these are not easy to break down.
there's a lot of work in the industry, in academia trying to break out.
Leakage current versus non-radiative recombination mechanisms.
all kinds of opportunity for you guys studying characterization.
And how to design experiments to really pull these things apart.
It is not straightforward.
I would submit that well designed LED's have injection
efficiencies that are very high.
I believe the ones that we have at SORAA are very high but you can get it wrong.
If you misplace the junction or you mess up on
this AlGaN layer, you can have substantial leakage in current.
and that, that stuff was, you'll never see.
It will recombine in the p-layers and you're done.
once you get the injected current in the
active region, you need to convert that to light.
And that's also not for free.
it's driven by, you know, roughly three mechanisms.
Again one, this is a fairly simple way of looking at the situation, so we have
a so called A-B-C model where we consider
a non-radiative recombination like
Shockley-Read-Hall recombination and defects.
spontaneous mission which is what we like.
The bimolecular radiative recombination for electron-hole pairs generate
a photon but we also have Auger recombination.
And one of the interesting
things, at least in the GaN role has been over the last five years or
so, the notion that, in fact, Auger scattering
plays a role even these wide bandgap semiconductors.
we published the first paper calling attention to that
in 2007 at Philips under a lot of controversy.
But when you look at the data, I
think it's still supported that in these wide bandgap
semiconductors higher line electron bands and valence bands
can play roles that are, again, not so straightforward
to figure out.
And we do see evidence under high carrier
densities of consistent non-ready recombination due to Auger.
so this is the game that we fight every day.
and try to get the internal efficiencies as high as
possible under whatever conditions we want to operate the diodes in.
after internal efficiencies, so the photons are generated.
We got to get them out of the crystal and you
might not think this much of this probably haven't looked it.
But it ends up being a big deal because the refractive indices of these
materials either GaN or, in the case of AlGalP is even worse, are very high.
And so electromagnetic phase batch matching boundary conditions require
a refraction of of waves at the semiconductor ambient interface.
And once you go beyond the critical angle the,
the ray, the light is trapped inside the crystal.
And any kind of working device is going to have
metal contacts, electrodes, maybe wire bonds, some, something in it.
You're, you know, even dope ends or the
active region itself that's going to absorb photons.
You cannot have the photons rattling around inside this chip forever.
The light extraction will be extremely poor.
old LED technology with absorbing substrates and
just one surface for extraction, typical extraction efficiency 5 to 10%.
Horrible.
And there's been all kind of
to get that to be much, much better. that's kind of shown here.
This is a review paper, again, from several years ago.
back in the early 90s late 80s,.
There, if an LED was bright enough that you could see it, it was doing its job.
Because most, back then LED's were used as
indicators to tell you that your clock radio's
on or, you know, you can see this button is turned on or what have you.
or, you know, signage, low level signage.
But as we started to get performance better
and better, and this is really in the 90s.
This started to kick off, people realized
you could use the LED's for lighting applications.
All of a sudden focusing on efficacy was a much bigger deal.
And so you had all the LED companies at that time starting putting a
lot more investment into improving the performance
of the LED's to get efficiency up.
And so there's this dramatic improvement in light extraction efficiency.
I won't go into all these details of
read this paper I can, I can direct you to it.
but the efficiencies went from this 5% regime up to 80%.
So 80% of the photons generated in the active layer get out
of the chip and are measured and collected, which is pretty amazing.
and I'll talk about where we're at now
[LAUGH]
we're fighting for every 1 or 2%.
So source target right now is a 90% efficient light extraction efficiency.
these are some of the tricks that people play.
there's various things that you have to deal with, so as
I said before, you know, GaN is usually deposited on sapphire.
Sapphire has a fairly low
[UNKNOWN]
compared to GaN.
So you get trapping in the GaN layers against the sapphire.
So one common strategy is to pattern the sapphire substrates.
So light scatters at the growth interface with the sapphire and gets out.
Light extraction can be made higher that way.
It's also common in those cases to use if it's an
[UNKNOWN]
device, to use an indium tin oxide
[SOUND]
transparent electrode to get more light out of the LED.
another way is to remove the sapphire altogether.
So Osram developed what they call ThinGaN technology where the device is inverted
and bonded to a metalized carrier, like silicon or *** arsenide or germanium.
Sapphires are moved in a
[UNKNOWN]
to provide a vertical thin film device, and so we get light now
[UNKNOWN]
that way.
Lumileds has an approach where they use their flip chip technology in
combination with sapphire removal and roughness
to get light, high light extraction.
This is a kind of interesting device because it
looks very analogous to sun-powered diode, a high efficiency sun-powered
[UNKNOWN]
diode.
So if you are interested in a very efficient
solar device, you can look at LED's and get ideas about
[LAUGH]
the same kind of thought process and vice versa.
And also, there's
[UNKNOWN]
transparent substrate, you can do things like chip shaping.
So you know, a parallel pipe is not the best thing, if you want to break
up this light trapping characteristics so you
can do other things to improve light extraction.
what we can measure is external efficiency.
So this is photons and the detector versus electrons injected into the diode.
And this is how it looks as a function of
wavelengths of the InGaN material system is here, AlGaInP material system.
the AlGaInP material systems I said before.
bandgap potentials get really small when you add aluminum increase
the aluminum mole fraction you have this really rapid falloff.
And this is just because carriers are not confined and
they aren't in the direct valleys anymore so the efficiency, I guess, pretty low.
And the temperature sensitivity in the system is very high.
InGaN materials are very good to the blue because most people working in the blue.
this result here is our violet result,
which we, in fact, just published last week.
I guess an APL with our, and I'll go through that in a minute.
rapid fall off as we go into the green and this has some to,
something to do with polarization fields.
things like spontaneous and piezoelectric polarization fields get
exacerbated as we increase in the nitride mole fraction.
And the other problem we have is this strain
issue that I mentioned before which complicates the performance.
But, the, the, for white light, you're really focusing in
these areas, violet or blue emission down converting with phosphorous.
And this is how you do it with
phosphorous, in this case for a blue LED can
be mixed with a typical common phosphorus YAG, so-called YAG.
It's a garnet.
A
yellow phosphor can be mixed, for example, with a
nitride phosphorus as a family of 258 nitride phosphorus.
[UNKNOWN]
that are fairly common.
And once you have these three color points, you
can mix and match to make anything you'd like.
And again, this typical cartoon.
You can have a spectrum that looks like this.
And basically the idea is you are matching a black body radiator.
this is something we all view as white, and you can
see we have to overshoot a little bit in the blue.
There is a requisite gap but until we get to the phosphor
emission, and then we track the black body and then you turn off.
so that's how phosphor LED is made.
the, if, if we want to have the biggest impact near-term
on energy consumption, you have to go after the retrofit market.
There are 30 billion installed sockets in the world, and while it's nice
to think about, you know, brand new luminaires that all run on low voltage.
And are, you know, integrated with the building control systems for the future.
And that will happen. It's going to take a long, long time.
In the meantime, we have 30 billion sockets that are housing
energy hogging technology like halogen or incandescent
or fluorescent and they need to be replaced.
So this is really the first market of focus, at least for SORAA.
in general the feedback we have, when we were getting into looking more at
opportunities for LED's, is that typically, the
quality was insufficient, especially the brightness was insufficient.
It's very expensive.
And so we were thinking about ways to, to deal with that.
another barrier to option adoption is light quality.
And this is something that I would
submit the compact fluorescent guys completely got wrong.
there was a huge focus for fluorescence on cost at the expense of anything.
Was the idea, was it that the consumer
would never convert unless it wasn't cheap enough?
What they failed to realize is that people didn't covert because of the price.
They didn't convert because they hated the light quality.
And that's true in the US today 30 years later, the penetration of
CFL's into US market is like 15 or 20%. It's complete failure.
and it, when hit, when McKinsey went back and you know, surveyed everybody.
Residential office, all of these different sectors to look at these various aspects.
Lifetime, purchase price, efficiency, light quality is number one in every
one of these sectors, with the only exception being the residential consumer.
There's a real
[UNKNOWN]
sensitive, slightly preferring price
[UNKNOWN]
quality, but only a little bit. Every
[UNKNOWN]
every other application, quality is a big deal.
We don't want to get that long again.
And I think there's some risk of that.
if you go in the source today and just grab
some LED product off the shelves at Home Depot, or Costco.
You may not be getting something that you're going to be happy
[UNKNOWN]
when you get, when you get home.
It doesn't mean there's not good products out there.
But there certainly a risk.
that there are some products out there that are not so great.
as I said before we're involved with lighting for a long time and we expect
a certain things in terms of quality of
light, like color rendering, color uniformity, being quality
[UNKNOWN].
All of these things we need be provided and there's
no reason you can't provide these in LED lighting products.
the other barrier to adoption outside of quality is, is the cost thing, you know?
Fluorescence have done a good job with rebate programs,
apparently be, being fairly inexpensive at the end-user level.
You know, a few dollars or even less, a dollar a piece for
lumen compact fluorescent. LED is about 10x higher than that today.
and that, this is something that obviously has to be addressed.
[COUGH]
and I'll talk a little about our view, SORAA's view on
[INAUDIBLE]
address this problem. And it kind of goes back to Moore's Law.
If you look at the power of the computing industry, it was really
driven by the fact that transistor density just increased and increased and increase.
And so, you know, what required a mainframe several years ago
or maybe a decade ago is, is probably in your cellphone today.
It's on that order, right? It's a huge increase in computing power.
there is a similar trend
in LED's at a lumen per square millimeter level.
It's a little bit different, but really it's how
much light am I getting from my semiconductor wafer.
This is what drives the cost. It
drives the size of the fab that I need.
It also drives the size of my final product, the
light source, the optics, the heat sink I need to design.
And what we see, if we look at lumens per square millimeter over the last, you
know, a decade or so, you see a nice strong trend but is now rolling over.
Roland Haitz who was a components group manager at HP
was the first one to kind of identified this exponential
trend in light output but it hasn't been holding up, the last few years.
And, in fact, if anything, what you see, a
lot of the LED companies now doing, if I get
more light is not increasing the power density of
the LED's but it's putting more LED's into the package.
Say this is a fairly large packages that lead
to fairly large and arguably clunky light system designs.
So with SORAA's approach, we decided, you know, why do we have to go down this road?
Why can't we
do something better.
And so we developed what we call GaN on GaN technology, which will be
the focus for the rest of my talk to be able to do that.
And the idea is to focus on really high power density devices that enable us
to go to a different regime in terms of performance, but also in terms of cost.
[MUSIC]