Heterogeneous Integration of III-V Semiconductor Nanowires with Si and Graphene for Photovoltaics
Sep 5, 2023
"Heterogeneous Integration of III-V Semiconductor Nanowires with Si and Graphene for Photovoltaics"
Parsian Mohseni, Electrical and Computer Engineering
CNST Nanotechnology Workshop 2013
All right, well it's certainly my pleasure to be here today to share with you some of
our work on heterogeneous integration of 3-5 crystals with silicon as well as with graphene
through the use of nanowire structures, and our main device application here is photovoltaic
solar cells.
So I should point out before starting that this is a largely collaborative effort between
the research groups of Professor Shulian Li, John Rogers, Joe Lighting, Eric Popp, as well
as some others both here at U of I and other universities throughout the states and internationally.
So a good place to start is with the question of what motivates us to study these material
systems and these devices.
We know currently in the current state of solar research, our goal is not to simply
continually just increase the conversion efficiency of our devices, but it's really the fundamental
factor is the dollars to watt ratio.
We want to move towards third generation solutions.
So whereas first generation photovoltaics is predominantly based on bulk single crystalline
silicon solar cells, these allow for moderate conversion efficiencies but are also associated
with really high costs.
So a better solution would be to move towards second generation applications which are based
on thin films, cadmium telluride, organic, SIGs, chalcogenides, and this allows for significant
cost reduction, but it's also associated with a decrease in conversion efficiency.
But that's okay because the overall dollars to watt ratio is actually reduced.
Now what we aim to do through the use of nanowires is to really get into this third generation
regime that through the use of nanomaterials and novel hybrid structures, we can take advantage
of the novel physics that they provide to not only increase our conversion efficiencies
beyond the single band gap limit, but also to keep our costs in par with second generation
solutions.
So then the question is still why nanowires?
There's three main reasons why we want to use nanowires for solar cells.
First off, they allow for absorption enhancement.
If you take a look at this photograph here, I'm showing you a bare gallium arsenide substrate.
It's perfectly reflective.
You can see yourself and it acts as a mirror.
Now if we take the same substrate and grow a nanowire array on it, you can see that it's
almost allowing for unity absorbance.
Now this is based on two principles.
First of all, the nanowires allow for multiple scattering interactions and they also act
as a graded refractive index medium, allowing for ease of coupling between air and the substrate
itself, both of which allow for superior light trapping beyond the level of what can be achieved
even with the best multilayer anti-reflective coatings.
Second of all, nanowires allow us to sort of relax the lattice matching criteria for
heterepitaxy.
In heterepitaxy, when we grow one material type with a particular lattice constant on
another material with a widely different lattice constant, then there's strain associated
that interface and that strain can lead to dislocations which present adverse effects
for our devices.
Now with nanowires, because of the small footprint that they provide as well as the large surface
area or volume ratio, part of that strain can be relaxed along the sidewalls.
This allows for a wide degree of bandgap engineering, a combination of dissimilar materials.
So we can grow high quality crystalline materials, in this case gallium arsenide on silicon for
dramatic cost reduction and we can also add functionality by growing these nanowires,
these high quality crystals on flexible substrates for added functionality.
In this case, these gallium arsenide wires are grown on a flexible mat of carbon nanotubes.
Also the last point I should make about nanowires is that they allow for a wide degree of freedom
in the structures that we can make.
With planar designs, we're really only free to modify our doping profiles and our material
changes in one direction, which is the lateral growth direction of our crystals.
But nanowires offer the opportunity to really modify our geometry in three dimensions.
And part of that is growing these so-called core shell structures where we have one material
or doping type in the core and we encapsulate it radially with a different doping type or
a different material.
So what this allows us to do is really decouple or orthogonalize the direction of light absorption
and carrier collection.
More specifically, if we tune the thickness of our shell to the order of the minority
carrier diffusion length, then we can collect our photogenerated carriers before they thermalize
to the bandage or before they recombine with other carriers.
So the use of core shell architectures allows us to more efficiently collect our carriers.
So as a quick outline for this talk, I'd like to focus on two material combinations.
First off, 3-5 crystals with silicon substrates.
I'll go over some growth and some characterization of this heterogeneous integration.
And then we'll consider some device applications using nanowires both as passive components
as well as active elements in hybrid devices.
Then we're going to switch gears slightly and consider the similar 3-5 nanowires grown
on graphene substrates moving towards flexible devices.
Again, I'll consider some growth and characterization results.
I'll discuss a new phase segregation effect that we've seen.
And then I'll show some recent results from our prototype devices.
So here I've shown you a series of indium arsenide and indium gallium arsenide nanowires
grown vertically on silicon substrates.
Again, these are highly lattice mismatched systems, but the crystalline quality in the
nanowires is actually very well preserved.
We can tune the composition of these ternary materials by just modifying the molar flow
ratios, the incidence of our indium to gallium that we use during our growth.
And in so doing, we can tune the band gap of this material to a particular range of
the solar spectrum that we aim to absorb light from.
These nanowires are grown through a seed-free and a pattern-free approach.
These are not vapor-liquid-solid seed-based growth, and this is not a pattern-based selective
area epitaxy.
It's direct epitaxy on the substrate.
I should point out that in tuning this material composition, we see slight compositional variances
from one material type to another, namely that with the introduction of gallium, these
wires tend to be slightly thicker and slightly shorter.
So there is a material dependence on the morphology.
We know from our material characterization, here I'm showing you x-ray diffraction rocking
curves that we can tune the lattice constant of these materials based on the composition
that we grow in MOCVD, and we can tailor the band gap of these materials, again, based
on the composition.
We can see bulk-type optical behavior from these nanowires, which is very promising for
device applications, and I should point out, based on our PL results, that since these
are nanowires and they have large surface area-to-volume ratios, there's lots of recombination
that can occur at the surface, so we really need to passivate the surface, and one way
to do that is through in-situ passivation, basically encapsulate our core with a material
that quenches all those recombination sites at the surface, and in so doing, we can increase
our luminescence intensity by roughly a factor of five.
So moving on to consider some device applications, first off, I'd like to show you a passive
way of using these nanowires.
What we've done in this approach is we've taken an existing silicon solar cell, the
p-n junction is built into the silicon substrate itself, and we're growing these nanowires
on the backside of the silicon cells.
So what they allow is for two-fold improvement.
First of all, they can re-scatter the otherwise transmitted light from the backside of these
solar cells and allow for reabsorption back into the core, as shown here, and second of
all, they act as a low-bandgap rear-point contact for more efficient collection of our
photogenerated holes.
So in terms of the device performance, we can see that over the entire spectral response
range of the silicon solar cell, that we can increase our external quantum efficiency by
roughly a factor of 40%, and this is, again, because we're reabsorbing some of those transmitted
photons, and then we're more effectively collecting them, and this translates into an increase
in our short-circuit current density, as well as our open-circuit voltage, leading to a
conversion efficiency increase of roughly 3%, just by putting nanowires on the backside.
So that's the passive approach.
We can also use these nanowires as active components to our devices.
We can grow these nanowires on the top side of a p-type silicon substrate, and we can
dope the nanowires, institute, make them n-type, so therein lies the p-n junction that allows
for carrier separation, and we can encapsulate the nanowire array in a polymer matrix, planarize
it so that we can deposit a top contact to the wires, and then the backside contact is
standard for contacting silicon, as you normally would.
So we do see a clear photovoltaic effect from this material system.
We characterize it by a conversion efficiency of roughly 2.5%.
Now based on the dark IV characteristics, we know that this has an ideality factor of
roughly 1.9, meaning there's a high degree of recombination, and again, that comes out
of the surface states on the nanowires.
So I'll talk about how surface passivation can lead to improvement of the conversion
efficiency in a few slides.
For now, we can also, so in this system, I showed you indium gallium arsenide nanowires,
where we're tuning the ternary composition based on the ratio of indium to gallium, the
ratio of the group 3s.
We can also modify, instead of the ratio of the group 3 species, the ratio of the group
5 species, so as to grow indium arsenide phosphide nanowires.
And as you can see here, again, this is based on a seed-free, pattern-free growth mechanism
in MOCVD.
We can grow a dense array of high aspect ratio nanowires, very uniform in terms of their
structure, and we get nice wafer-scale uniformity of coverage of these wires.
We know from our, again, our XRD experiments that the composition of these materials can
be tuned from all the way from indium arsenide to roughly 75% indium arsenide phosphide,
and we know from our TM analysis that the composition of these wires is fairly uniform
throughout their length on the single nanowire scale, and we know from our XRD and our PL
results, given the very narrow line widths, that the ensemble of this system also has
a very narrow distribution in terms of composition.
So composition uniformity is actually fairly important for us, and we can demonstrate that
we've achieved that.
Again, in the case of, as in the case of indium gallium arsenide wires, we can use these wires
in an active manner.
We can dope them N-type and grow them on a P-type substrate, again, submerge them in
a polymer matrix, backfill and deposit our transparent conductive oxide on the top, and
here I'm really showing you the difference between a passivated system and a non-passivated
system.
If we don't use a shell, if we don't passivate the surface, then we have a device that has
a performance of roughly 1.9%, has a really low fill factor, again, hinting towards recombination
on the surface, but once we grow a passivation layer, a material of a higher band gap, and
again, this is done in situ in a single growth, we can quench those surface states and not
allow recombination of the surface.
So those are all these extra carriers that we're now collecting, which can be seen as
evidenced by the higher short circuit current density and the increase in the open circuit
voltage and the increase of our fill factor going from roughly 45 to closer to 60%.
And we've roughly doubled our conversion efficiency just by using this passivation layer.
So I have about five minutes left, and I'm going to switch gears now and talk about growth
of similar nanowire structures on graphene moving towards flexible devices.
Using the same pattern-free, seed-free growth mechanism, we can grow dense nanowire arrays.
Over here I'm showing you a tilted view SCM image, low magnification of the sample as
it's grown.
These nanowires are vertically oriented, very high density, roughly 700 million of these
nanowires per centimeter squared, and they grow vertical to the substrate, which points
towards epitaxial guidance from the substrate itself, from the graphene layer itself.
We've also seen that these nanowires exhibit an aerial selectivity.
They tend to only grow on regions that have graphene coverage, meaning that we can pattern
a graphene layer amongst the background of oxide to have regular arrays of nanowires.
We can also start to introduce gallium to grow ternary materials such as indium gallium
arsenide.
As before, you see the nanowires are very nice, uniform, large area density.
They are vertically oriented, but as before, with increasing gallium content, this xi value
points towards a molar flow ratio that incorporates more gallium.
As we increase the gallium content here, we see a slight morphological variance.
We see the nanowires start to take on, about 40% of them start to take on this sickle-shaped
morphology, which hints towards compositional inhomogeneity based on strain in the system.
Now an interesting observation that we made through the analysis of our indium gallium
arsenide systems is that this is a naturally and spontaneously phase-segregated material.
During the entire growth, we're growing one segment.
We aim to grow indium gallium arsenide, meaning that we supply indium, gallium, and arsenic
atoms during our growth.
So this should be a single-phase system, but our XRD results for all the indium gallium
arsenide signals demonstrate two peaks.
They demonstrate an indium gallium arsenide signal that tends towards the gallium arsenide
111 location with increasing gallium content.
So that's the indium gallium arsenide signal, but there's also this persistent indium arsenide
signal obtained from every single one of these in-gas samples.
So to really understand what's happening, how this two-phase system comes to be, we
need to consider this on the single nanowire scale.
So what's shown here is one such indium gallium arsenide nanowire that's grown on graphene,
and through high-angle annular dark field imaging and scanning transmission electron
microscopy, we can see a clear contrast variance in this material.
We see a region of high contrast in the center and a region of dark contrast towards the
outside, and based on our EDX line scan analysis, we know that the region of high contrast is
actually an indium arsenide phase, and the region of dark contrast is an indium gallium
arsenide phase.
So why does this segregation come to be?
It turns out that indium arsenide is very closely lattice registered to graphene itself.
The unit cell, the lattice constant of indium arsenide along the 110 direction is roughly
a six-fold multiple of the atom-to-atom spacing in graphene.
So because of this atomic configuration, indium arsenide wants to settle in a location that's
most natural for itself, and it segregates gallium towards the outside, forcing it to
settle really only where it can form covalent bonds with the sidewalls of the nanowire itself.
We can get around the space segregation mechanism by tricking the crystals during growth, and
how we trick it is to provide an indium arsenide base.
So we can grow the nanowires for two minutes just flowing indium arsenide, and then we
can slowly start to introduce gallium, and in that case the gallium can actually stick
to the indium arsenide crystals, and we can grow a single phase uniform composition indium
gallium arsenide segment, and now this allows us to move towards device applications, because
we can tune the indium gallium arsenide composition to that region of the solar spectrum that
we aim to absorb, and we can introduce doping in this material just through in-situ introduction
of silicon and zinc in different segments, so we can dope part of this nanowire N-type
and dope the other part P-type.
We can again backfill with a polymer, planarize the top surface and deposit indium tin oxide
on the top.
Here I'm showing you an SEM image of the sample after the planarization process where the
tips of the nanowires have been exposed for contacting.
In terms of device performance, the good news is that we do see a clear photovoltaic effect.
First, under dark testing, we see a clear sign of asymmetric rectification in the IAV
behavior.
This diodic behavior is not observed when we only grow N-type wire, so this points towards
the idea that our diodic behavior is not based on Schottky barriers and that our dopants
have been incorporated and activated as desired.
Under AM 1.5G illumination, we see a clear shift in the current.
We see a short-circuit current of roughly 12, close to 13 milliamps per centimeter squared.
This leads to a conversion efficiency of roughly 0.6%.
This is a low conversion efficiency, but I should point out that this is already a record-breaking
efficiency for substrate-free nanowire solar cells, and there's lots of room to improve
this system.
To summarize quickly, as I'm just at the cusp of 20 minutes, we've demonstrated heterogeneous
integration of 3-5 crystals with silicon and with graphene, moving towards flexible devices.
We can use these nanowires both as passive and active elements in hybrid systems, and
we've demonstrated this photovoltaic effect from the nanowires just on a graphene substrate
that simultaneously acts as an epitaxial guiding layer as well as the back contact of our devices.
So with that, I'd like to acknowledge the support of my advisor, Professor Shu-Ling
Li, as well as some of our colleagues here at UIUC, Professor Wasserman and Professor
Minju Larry Lee at Yale and Professor Ning at ASU, and I'd like to acknowledge the support
of the following financial institutions.
Thank you for your attention.
Good.
Now we have time for a few questions.
What determines...
Yeah, this is a good question.
Again, like I said, this is not a vapor-liquid solid growth mechanism, so the density is
not determined by seeds, nor is it determined by patterned approaches you would have with
selective area epitaxy.
The real answer is that it is a strain-based nucleation mechanism.
The start of this growth process is similar to a former Weber island-based growth.
So it is a strain-based system, and we know that the relative difference in the strain
changes the density.
With more gallium incorporation, we have a lower density system.
So you've observed ways to modulate the density?
We have, simply through compositional control.
Yes.
Thank you.
Yes.
Yes, we have.
And is there a dependence there in terms of...
It's interesting.
Actually, the device that I showed you is based on a multilayer graphing consisting
of four layers.
So we use a four-layer system instead of a single-layer system to improve the conduction
in our backside contact.
The real change with multilayer graphing is when you stack multiple layers, you introduce
kinks or folds, and then the nanowires no longer tend to take off perfectly vertical.
Because they can grow on a kink or a fold, they can offshoot from the vertical direction.
But otherwise, the density is actually improved with increasing number of graphing layers
because defects in the graphing can be passivated from one layer to the next.
The edges of the graphing is where you have all your dangling bonds.
So that's where you don't get nanowire growth, you get polycrystalline sort of parasitic
growth.

