Secondary Ion Mass Spectroscopy (Tim Spila)
Sep 5, 2023
Tim Spila
06/03/15
AMC 2015
Today, I'm going to talk about secondary IMS spectrometry.
As Rick was good at pointing out, when we do surface analysis, we're usually talking
about different things coming in and different things going out.
For SIMS, we're looking at primary ions going in, and then we look at the ions that
come out.
We call those secondary ions, just like you would in a secondary scanning electron microscope,
you're looking at secondary electrons.
Those are the electrons that come out.
The nice thing about SIMS is, because we're measuring mass, the mass of the ions that
are coming out, allows us to examine things like isotopes and differences there, which
you can't do with other techniques.
The other thing I'd note is SIMS is complementary to OJ and XPS, and it's very nice because
we work in SIMS tend to be at the 1% and lower concentrations, where the XPS and OJ,
0.1%, 1% and higher, so you can start with one and then work your way to the other.
Have the nice technique comparison here that you guys can look at later.
Just a couple highlights, like I said, SIMS is good as best below 1% and can measure down
into the parts per million range.
We have, to be able to be quantitative though, this is where SIMS is a little bit difficult,
we need to have standards to be able to be quantitative, whereas with XPS and OJ, there's
a lot more known about physics, so therefore you don't need to necessarily have standards,
although they can help.
And one of the other things I'd like to talk about is the fact that for molecular imaging,
you can get some molecular information from XPS, as Rick's talking about, but you can
also get molecular imaging from the tough SIMS.
So that's nice.
This is a block diagram of the SIMS technique.
So we have an ion source that we start with.
Could have lots of different options for ion sources, cesium, oxygen, nitrogen has been
used, gallium.
More recently, people are using heavier things and molecules, so gold, bismuth, and then
you can have bismuth 2, bismuth 3, and then very recently, people used C60 buckyballs,
and now massive argon clusters of like 5,000 argon atoms in a cluster, and even water droplets.
We have our sample, and as Rick mentioned, we have to be vacuum compatible, so therefore
if you have samples that are like biological, we can look at them, but you have to be able
to mount them so that they can be put into a vacuum.
Finally, we go through an energy analyzer and the mass spectrometer, and then finally
we have many detectors that we can use.
The most common is an electron multiplier and Faraday cup, so that will give us counts,
but we can also do some imaging work with resistive anode encoders or channel plates
so that we can get out all these different things, a mass spectrum, and then if you look
at the masses with respect to depth, you can get a depth profile.
You can also image, and then you can combine it all into an image depth profile, and I'll
get to show you some of those examples of each.
Now, when we talk about SIMs, we generally work in one of two regimes.
We can either work in the dynamic regime or the static regime, and the idea, you can kind
of think about this as, if I'm in the dynamic regime, I'm removing material so I can go
down deep into my sample.
If I'm in the static regime, I'm only looking at the very surface.
We kind of define it as less than 10 to the 13 ions per square centimeter.
If you think of silicon having 10 to the 15th ions per square centimeter on its surface,
you're affecting 1% or less of the atoms on the surface.
The idea being that if you come back to the same spot and do the same measurement, you
should get the same result.
With the dynamic SIMs, of course, we're looking at elemental, but with the static SIMs, we
can look at elemental and molecular, so, and then, of course, like I said, we're doing
depth profiles generally with dynamic SIMs, you can get a mass spectrum, and the one instrument
we tend to use a lot for that would be the Kamika 5F, but we can also do dynamic measurements
with the TRIFT.
The static SIMs is almost exclusively done on the TOF SIMs, the TRIFT, and we can get
elemental and molecular information.
This is a picture of Judy Baker working the Kamika 5F.
As you can see, we have two sources that we can use on this, a cesium source and a duoplasmatron,
which we use with oxygen.
The secondary beam then goes through the static analyzer and the electromagnet to get us our
mass, and then we have three detectors, a Faraday cup, an electromultiplier, and also
we can do imaging with a phosphorous screen or the resistive anode encoder.
The mass is determined by bending the ions to the electromagnet, so therefore, if it
doesn't have the right mass, it doesn't get through, so we can look at one mass at a time,
but we can switch between them by just changing the power to the magnet, but it's only one
at a time.
With the Time-of-Flight SIMs, which is our other SIMs, it's a TRIFT 3 from Physical
Electronics, instead of using a continuous beam hitting our sample, we use a pulse beam
coming from a gold source.
We also have cesium and oxygen beams for doing material removal, but for analysis, we'll
use the gold beam.
The Tough Time-of-Flight detector works slightly differently, so we have a pulse beam coming
in and we remove all the ions, secondary ions, at time equals zero, and then we accelerate
them away from the sample surface, so we've given them a certain amount of energy.
Well, kinetic energy is one-half mass times velocity squared, so if we put a detector
really far away, all the light things will arrive first and the heavy things will arrive
later, so that we don't have to have such a big room and so that we can increase our,
enhance our sensitivity.
We actually bend it through some electrostatic analyzers before we reach our detector here,
and so what's nice about that is we have, for every spot on the sample surface, we get
a whole spectra, and so therefore, if you were thinking about doing, if you thought
about it, you have X, Y, and Z information, if you're doing depth profile, and then you
have a mass to charge ratio and the intensity, so it's five-dimensional data and the computer
stores all of that.
Now ion beam sputtering is a very technique that produces lots of things, so you come
in with a primary ion beam and you create a lot of damage to your sample surface, and
then all types of things come out.
Those things include positive and negative ions, neutral species, you can get light,
you can get electrons, you tend to have an amorphous crater here due to the bond breaking
due to the primary ion beam, and so the fraction of ions that you get out of it can be very
dependent on a lot of things, one of them being the primary ion beam that is used.
Now from Penn State, Professor Garrison did some work looking at hitting a sample with
C60 versus gallium and the type of damage that's done.
And I have a—she had done a molecular dynamic simulation to look at it, and so this is
the gallium atom that's coming in and hitting our surface, and you can see the type of damage
that comes out.
If you were very quick, you could see the secondary ion that came out, so this is the
amount of damage and the depth of the damage that occurs from a single ion impact.
And this is from a top-down view.
I'd like to go back and redo that just so you can see the—catch the ion that comes
out going this way.
Boom.
So all this atom motion, just from hitting the sample, the—so now the second thing
that we'll look at then is if we use C60, the idea here being that we have less energy
per atom that's hitting the surface, and you can see all the things that are coming out.
You're having molecules come out, atoms together and all that, and the damage layer is not
significantly deeper than the layers that you're going to be looking at material from.
The goal of this with C60, this is why people are using the larger cluster beams now, is
because the—you can go to the idea now where I'm going to look at molecules coming out
as I go deeper in.
Under—with traditional ion guns, you're breaking so many bonds that you can't look
at moleculars as you go deeper into a sample.
Now you have the ability that you might be able to do that.
Now we want to—generally with all the techniques, we want to try to be quantitative.
SIMS is very difficult to be quantitative, but if you look at it kind of from first principles,
you have your ion beams coming in and you're measuring things coming out.
So we can say our current that we're getting from our species that we're interested in,
perhaps our dopant in a semiconductor wafer, is proportional to the primary particle flux
coming in, the sputter yield, i.e. how many atoms come off the surface for each atom that
I'm impinging on it, and then what fraction of that comes off as an ion, then you also
have what's my dopant's coverage at that surface, and then the transmission of the
system.
Well, one of the things that we're talking about, sputter yield and ion sputter yield,
sputter yield generally ranges about 5 to 15 atoms come off per atom going in, but our
ion sputter yield can vary by orders of magnitude, like 10 to the minus 6 to 10 to the minus
2.
That's four orders of magnitude that it can vary over, and it can be affected by lots
of things—matrix effects, surface coverage of reactive elements, background pressure
in the sample environment, orientation, angle of emission.
It's a very, very complicated process, so generally we don't use first principles to
figure out what our ion sputter yields are, but if we use the standard samples so we have
something to compare against—it has the same matrix, has the same surface coverage
when we're doing the measurement—we can actually use some of these things, like reactive
elements on the surface or background pressure in the chamber, to enhance our signals.
That allows us to measure even lower concentrations, and so we do that.
In general, we can enhance the sputter yields by using ion sources that help us out.
So, for instance, if you look at the yellow atoms here, they want to be positive ions
when they come off, so we're going to use an oxygen gun to enhance that.
The green ions here, they're very electronegative, so they want to grab an electron, and so we'll
use a cesium ion to give that to them.
The physics behind it's a little complicated, but you can kind of just use the chart and
say, okay, what do I want to look for, and so what type of beam am I going to use?
This is another way of plotting that same idea.
The relative secondary ion yields cesium with negative secondary ions, oxygen with positive
secondary ions.
Highlight different things, so over here you can see calcium, zinc, they're pretty low,
but if you go to oxygen you get good calcium.
Zinc isn't really good.
Zinc is one of those strange elements that tends not to be good with either positive
or secondary ions, but what they found was that if you actually look for zinc plus cesium
or cadmium plus cesium with positive secondary ions, you can actually get good signal.
So we do have ways around it, even with those difficult elements.
Zinc tends to be a very popular dopant for gallium arsenide.
So you'll say, okay, how do I be quantitative?
And so, of course, we have the equation that we looked at before, so let's define something
called an RSF, all right, so you can kind of think of it as, well, we can say the concentration
is going to be the counts that we get from the dopant divided by the counts of the matrix
times the RSF, or you can reverse it around and say our RSF is proportional to the atomic
density divided by the ratio of the counts.
And then, of course, a lot of times to make a standard you'll do ion implantation because
then you know exactly how much you put into your sample, and so we have this, and so you
can make the equation a little bit more complicated to deal with the Gaussian profile.
But then from that we can have accuracies within about 10 percent.
And so I did this with a phosphorus standard that I have, and so, of course, it's kind
of a circular argument, I'm telling you that we're using a phosphorus standard to calculate
what our phosphorus standard is, but I did it under two different conditions, with an
oxygen beam with a cesium beam, kind of show you how I was saying that phosphorus generally
would want to be done with a cesium beam, and this will kind of be an example why.
So if I use a cesium beam, which is the light blue measurement for the phosphorus signal,
you can see we have more counts than we do for the blue with it, where we have used the
oxygen beam.
Of course, when we do our analysis to generally generate, we know that the area under the
curve is a certain number of atoms of phosphorus, so of course we can get the same curve shape,
but you'll note because with the cesium beam we have more counts for our phosphorus, we're
going to have a nicer looking signal.
And so phosphorus and boron are very popular dopants for silicon, and so we're constantly
switching back between oxygen and cesium so that we can have the best sensitivity for
that.
One last definition for SIMS is mass resolution.
So with elements, you tend to have, you can have moleculars that are very close to atomics,
and so having good mass resolution, even though the same nominal mass, having good mass resolution,
you're able to resolve one from the other.
And so we define mass resolution as m divided by delta m, so mass divided by the change
in mass that you're interested in, or the width of the peak.
So looking at sulfur at mass 32, you have a peak, this is plaid linear, but you can
also have two oxygens at mass 16, and so this is kind of how close they are together.
If you look at the, here, here's 32.
And so for this, you have to have a mass resolution defined of 1600 to be able to get it.
That's pretty high.
The main difficulty one that we have is looking for phosphorus, because silicon plus one hydrogen
can be very close to phosphorus.
But those are the most difficult ones.
Otherwise, the instruments we have have the ability to resolve the peaks.
We can get that high of mass resolution if we're willing to throw away some signal.
So some examples, this is a depth profile that was done on the Kamenka Sims.
So gallium arsenide, carbon doped and silicon doped, and they did some annealing in hydrogen.
They did some annealing in a nitrogen atmosphere, and what they found was that the hydrogen
actually can go away.
So what's neat about these?
Well, first off, you get very nice results at good depths, microns into your sample.
Second, we can see hydrogen, which is not seeable by many of the analytical techniques.
And third, you can see, this is a log scale, so that's one, two, three, four, five, six
orders of magnitude in concentration that we're looking over.
So very large dynamic range.
Another experiment that was done, you may recognize that name, was looking at diffusion
of oxygen into a nickel three aluminum alloy.
And so what they did was, they did four hours of 18 oxygen, isotope of oxygen, followed
by 16 hours of normal 16 mass, 16 oxygen.
And because we're looking at ions, at masses, you can look at the isotopic analysis.
So the results, and so basically what they were trying to determine is what's moving
when you're doing oxidation.
Is the aluminum or the nickel moving, or is the oxygen just diffusing in?
And what they were able to find out is that both the nickel and the oxygen diffuses in
the sample, but the aluminum does not.
Some work that we were doing when I was a graduate student looking at highly doped boron
in silicon.
And so this is a SimSTEPPS profile through a modulated dope sample where we varied the
boron flux, and it shows the different dopant levels that we have.
So it goes from 8.48 times the 13th to 1.2 times 10 to the 16th.
And what's interesting about this is, we have nice profile up and down for the lower levels,
but as you go higher in concentration, you turn off the boron flux, but you're still
having some.
So there's actually some surface diffusion.
There's so much boron that stays at the surface and then goes into the upper silicon layer
that doesn't have any in it.
And then we're able to see that very nicely with the Sims.
This is some boron delta-doped layers.
I apologize for the quality of the work, but so you can see that we can dope with very
sharp spikes, and we can see those sharp spikes in the Sims.
So you know, you're talking orders of magnitude, so 50% would be right there.
And so that's very, very nice.
This is more recent work.
So the difficulty, the problem that they had is they wanted a non-lithographic patterning
of wafer-scale nanostructure arrays.
So they've got all these little guys sticking up, and they want to have those well-doped.
The problem is if you dope them before etching the arrays, you end up with very poor structures
and it doesn't work well.
So what they attempted to do here was to make the arrays and then dope it after the fact
using a spin-on dopant.
Previous results we did with these guys showed that we can dope the nanowires using the spin-on
dopants, but now the question was, can we do something about the tops?
And so what they did was they etched away a little area and was able to dope it.
And so we did some Sims on it, and depending on how long they doped it, we can get different
concentrations at the various tops, and that helps them with making contacts to them.
Also did some work on lithium-ion batteries.
And so lithium-ion batteries consist of a negative electrode generally made out of graphite,
and a lithium-containing positive electrode, which will have other metals in it—nickel,
cobalt, manganese—and an electrolyte, which is generally LiPF6.
One problem, though, that they've been having with making these batteries work longer and
better and move charge back and forth is that there's an impedance rise at the negative
electrode up here, the graphite part, and so what was causing that?
So we did some Sims work initially, and we followed the manganese, carbon, nickel, and
cobalt to see what was going on, and what we found was that there's a very large rise,
as you can see in the manganese signal, for what's known as an aged sample.
So we had a fresh-formed and aged sample, so that means that there's a significant
amount of manganese that's moving from this electrode up into the graphite electrode,
from the lithium electrode.
And so we were able to determine that, and so then the question was, what can we do to
limit that?
So we did some additional experiments where we put an alumina coating on top of this,
and from that, then, we did the same type of experiment.
And so now you do see that there is some aluminum that moves, and there's still some manganese
that moves, but the effect from the cycling and the measurements of that is that even
though there is some rise in impedance, it's significantly inhibited, but we do still see
some motion of these elements, and so therefore, there is still some loss due to that.
But this helped to increase the lifetime of these batteries significantly.
Now, I had mentioned before about dynamic SIMs and static SIMs.
All the examples I've shown you so far have been with dynamic SIMs, now I'm going to talk
about some static SIMs.
So just to remind you of the differences, with dynamic SIMs, we're doing some significant
material removal and looking at elements in our depth profiles, but with static SIMs,
we can look at surface analysis, and so we can also see molecular information, and then
we, as I said, the analysis is complete before a significant fraction of the molecules are
destroyed.
The TOF SIMs tend to be very capable in this static SIMs regime.
One of the other nice things about TOF SIMs is we can have an extreme dynamic range, since
the amount of time you wait is proportional to the mass that you're looking at, we can
have an extreme mass range.
So this is actually just a simple sample of gold on top of silicon wafer.
And so 197 is the mass of a gold atom, and so 1, 2, 3, 4, all the way up to about 20
in a cluster, we can see that.
So this mass range of 10,000 amu, and that's about 50 atoms in size, and you can see it
was still going.
Under normal circumstances, we don't see much past 1,000 amu with this system.
We can do trace analysis work, so this was some looking at things on the surface, some
contaminants on the surface of silicon wafer, gallium arsenide wafer, can have extreme mass
resolution on the TOF SIMs.
We can also do line scans, so this is some quantum dots of indium arsenide on gallium
arsenide, and you can kind of see lateral resolution.
Lateral resolution on the TOF SIMs is on the order of a micron, although in this case we're
seeing things about half a micron in size.
We can do depth profiling with the TRF SIMs, so in this case they used a gallium beam for
analysis and then used an oxygen gun to do the depth profiling, and you can see the gallium
signal going up and down opposite that of the aluminum signal.
And approximately you have one period for every two nanometers in depth, so that type
of depth resolution is possible.
Because of that, and because we can use a low energy ion beam for material removal in
the TOF SIMs, one of the projects I was working with, with Professor Sebauer, they're looking
at charge point defects and how they interact with the space charge region in the near surface.
So one of the things they did was they would anneal the samples in natural abundance oxygen,
and then they would come back and anneal it with isotopic oxygen, and then they could
follow the difference between the two as a function of depth.
And so they found out that the oxygen 18 will pile up in the first 10 to 30 nanometers for
both zinc oxide and titanium oxide, and then they were able to use their models to look
at the effect of the drift versus the diffusion that would naturally occur to determine that
there is an effect, and that the diffusion from a high concentration at the surface goes
to the lower in the bulk.
And then if you have drift in the opposite direction due to the space charge region,
you'll have a piled up surface.
If they're in the same direction, then you'll have the depletion of the mobile defects at
the surface.
Now, when doing depth profiling on the TOFSIMs, you have to be very careful, though, because
we're actually going to align both the analysis beam with the material removal beam.
And so ideally you want something that looks like this.
You have a nice big crater that's flat on the bottom.
Everything's at the same depth.
But a more realistic crater shape is something that looks like this for the TOFSIMs.
So you have some bending.
So that means that the only area that's good for measurement would be here in the middle
that I show there.
So if you do it nice and correctly, this is some work on neodymium and uranium oxide,
and so you have this nice profile where they grew a film and annealed it, and so it has
a nice Gaussian shape, and everything's wonderful.
But one of the problems that this researcher had was he had very, very small samples, so
he's trying to find shortcuts to the method, and so he's trying to use a different sample
to do the beam alignments.
And if you do that poorly, you could end up having your analysis area be here and spread
out over many different depths, and then you get results that look not the way you expect
them, but very different and very difficult and impossible to analyze.
And so the takeaway here is to have samples that are big enough that you can do your alignment
and your profiling on the same sample.
Some work that was done with Josh Ritchie in the Nozo and Moore groups, he had some
samples where he had a molecule that if he puts it under light, will break off.
And so by doing patterning, he could then make areas that were with the large molecule
and without the molecule.
And so the key here is we noticed that bromine was a very important element here.
And so we imaged the surface for bromine, and we were able to see areas where the molecule
was still intact versus areas where the molecule wasn't.
And even though sulfur is in both of them, it's less shielded in the second case, so
we could actually use that as well.
And so this is an example of kind of the top resolution that we're looking at in the TOFSIM.
So this is about 50 microns, and so you have about 20 pixels per micron in that case.
Originally they tried to work this in the XPS and was not able to see it, and that's
why we ended up in the SIMS.
In general, I like the idea of trying things in the XPS and the OJ before coming to the
SIMS, because of the fact you have some other advantages.
But this ended up being one of those situations where the SIMS was very useful.
Another work that has been done by Eric Munro was looking at images of an isolated neuron.
So this is from a snail.
Very nice, because the neurons tend to be like a millimeter in size.
And so what they had was, we have the mass spectrum that gives us peaks, and so you're
able to generate images from the different peaks in the mass spectra.
And so one of the nice things they were able to do is, if the images looked the same, they
could probably figure out that they came from the same peak.
It's just that the molecule broke in the process of going to the detector.
And so that is what they call a cracking pattern, so you can identify it.
So this mass here is with vitamin E, and then this one here is a vitamin E fragment.
And then the lateral imaging was helped to determine the cracking patterns.
Vitamin E, they actually found out, tended to be, if you starve the cells of vitamin
E, it tended to localize here at the part where the cell contracts with what they call
the process, which is where it reaches out to talk to other cells.
And so they found out that the vitamin E was actually very important for communication
between the cells.
So that was some work that was done with the Tuft SIMS.
Another researcher was here, Kenzie Amaya, actually was looking at images from a songbird
brain.
And so this is actually, you can kind of see the little squares, actually many, many images
that he stitched together to form, so I guess it was 194 individual images.
And so this is the whole image of the brain.
One of the things they were trying to look at is, if a songbird, a male songbird isn't
taught to sing by other songbirds, by a certain age it will never learn to sing.
And so they were looking at differences in the brain chemistry related to that.
And so they found that there's localization of certain fatty acids in the RA and HVC sections
of the brain when it is able to learn to sing that don't exist in the birds that weren't
learning to sing.
We can also do something completely different, some friction testing of samples using diamond
like carbon films.
And so you have a ball and you have a coated disk.
And so this is some work that was done on the Tuft SIMS.
And so we did imaging and we made the oxygen red and the carbon green, and so you can actually
overlay this to see, this is a wear track, and so in the middle of the wear track we
don't have any oxygen, but we do have some pile up on the edges.
And so that was related to lifetime ability for the wear of the diamond like carbon films.
And we can make 3D images out of that, and so on the software you can actually rotate
this around.
You can see the wear track, you can see the oxygen and the OH signals as well.
You can actually take a series of depth profiles and make a movie out of it.
And so this first column is for hydrogen and then we have CH2, C2H2 and oxygen with the
wear track doing a depth profile.
So here we'll just give a quick summary of the SIMS.
So we have ions that come in and we measure ions that come out.
And we can get information on surface mass spectra, 2D surface ion images, elemental
depth profiling and 3D image depth profiling.
We can detect elements from hydrogen and up.
We have sensitivity in the parts per billion range up to the atomic percent.
And our analysis area tends to be about on the order of one micron up to several millimeters.
And finally, I'd just like to thank again our sponsors.
Specifically for me, I'd just like to point out Cameca and Physical Electronics who provided
the instruments that we're using.
But also specifically our Platinum sponsors.

