Introduction to Automated Particle Analysis

8/2/2019 Introduction to Automated Particle Analysis

1/8

Introduction to Automated Particle AnalysisbyFocused Electron Beam

Copyright ASPEX Corporation ASPEX Corporation 175 Sheffield Drive Delmont, PA 15626-1723

P 800.573.7736 724.468.5400 F 724.468.0225 W www.aspexcorp.com

Frederick Schamber, Ph.D.Chief Technology OfficerASPEX Corporation


2/8

Copyright ASPEX Corporation Page 1

Introduction to Automated Particle Analysis by Focused Electron Beam

Frederick Schamber, PhDChief Technology Officer

ASPEX Corp

INTRODUCTION

This little document is intended to provide a brief entry-levelintroduction to the concepts and technology employed in theautomated particle analysis systems manufactured by ASPEXCorp. My objective is to convey a big-picture sense of whatour instruments do, and roughly how they do it. Along the way,Ill introduce some of the buzz-words that are used in our areaof technology. My ultimate objective is quite limited Im not

going to try to explore all of the interesting variations andapplications of electron beam imaging technology and weregoing to pretty much ignore the fascinating physics involved inthis instrumentation instead well be trying to home in on themost basic things one needs to know in order to understandhow/why our specific products work.

That being said, however, one can hardly talk about automatedelectron beam technology without discussing the ScanningElectron Microscope (or SEM as its usually known). As youcan gather from its name, the SEM is a microscope that is itmakes little things appear big so that we can study them. Theway it does this, however, isnt at all like the way a familiar lightmicroscope or magnifying glass works.

SEM FUNCTIONAL OVERVIEW

Figure 1 illustrates the way a SEM creates an image. Themost important thing in this figure is the beam, which is afocused stream of electrons. The function of the column is togenerate the beam and focus it down so it is a really tiny spotat the place where it hits the specimen, (How tiny is the spot?You could easily fit a million of them into the period at the endof this sentence.) The column can also laterally deflect thebeam so that the location where the beam spot strikes thespecimen can be precisely controlled. The electrons in thatfocused beam are traveling very fast, so complicated thingshappen when they hit the specimen. One thing that happens isthat some of those energetic electrons just bounce off theatoms of the specimen: we call those Back-Scattered Electrons(BSE). Another thing that happens is that some of the incidentelectrons get absorbed in the specimen, at the same timeknocking other electrons loose from the atoms of thespecimen, and we call those newly liberated ones SecondaryElectrons (SE). Also, the absorbed electrons give some oftheir energy to the atoms they strike and we say that thoseatoms are excited by the beam. One way that an excitedatom can get rid of this extra energy is by emitting a photon ifits a low-energy photon, we see it as visible light, but the mostimportant photons are the high energy ones we call X-rays. In

other words, the impact of the beam gives rise to a wholebunch of different kinds of emissions and it turns out that eachcarries different information about the local properties of thespecimen. For each kind of emission we can use a specializeddetector that sees those emissions and produces aproportional electrical signal. Each of those different kinds ofsignal depends on some property of the specimen so as wemove that focused beam spot across the specimen, the signalgets brighter or darker depending on how that property

varies across the specimen. We call this variation in signalbrightness contrast and we make use of it to generate apicture. We do this by moving the electron beam rapidly overthe surface of the specimen, and as we do this wesimultaneously change the brightness of the correspondingpixel of the image screen to correspond to the strength of thesignal. So where the signal from the specimen is strong, wesee bright pixels on the screen, where it is weak, we see darkpixels. The overall pattern of varying pixel brightness forms apicture of the specimen.

Now its important to note that weve drawn Figure 1 with reallydistorted scales. Specifically, the electron beam is about thediameter of a human hair at its widest point (where it comesout of the column) and converges down to a much smaller spot

where it hits the specimen. Also, the angle of deflection weredepicting in this figure is much larger than is usually the case.Instead, we typically sweep that really tiny beam of focusedelectrons over a comparably tiny region of the specimen saya square 1/100 th of an inch on each side. When we thendisplay the resulting pattern on our display screen, thedimensions of the screen might be 10 x 10 inches in otherwords, we are producing a picture that is magnified 1000 times(10 inches divided by 0.01 inch). By choosing how big an areawe want to scan, we can thus vary the magnification of ourpicture by scanning a bigger area, we reduce themagnification, by scanning a smaller area we increase it. Ofcourse, at the same time were doing that, we also need tomake sure that the size of the focused beam spot isappropriate. Its kind of like painting: you cant paint a clear

miniature picture with a big brush, nor can you efficiently painta large picture with a really tiny brush. Just as you want tomatch the size of the brush to the job, we want to make thediameter of the electron beam spot corresponds to the arearepresented by a single pixel on the display screen.

The last thing we need to point out in Figure 1 is the stage. Inone sense, the stage is nothing more than the platform we laythe specimen on. In practice, however, its a lot moreimportant than that. First of all, lets note that the object weve


3/8


depicted as a specimen is supposed to represent a piece offilter paper. When were looking at particles with our electronbeam, its quite common to deposit them on a filter paper orsomething similar. We need to realize that the particles we arelooking at are so small that they probably arent visible to theunaided eye (and some of them will still look like specks whenviewed under a light microscope). At the same time, the filterpaper we have collected them on might be an inch or two indiameter, so as one person put it, its like looking for BBs on a

football field (and thats actually a bit understated!) In orderto cover the entire area in sufficient detail, we need to break itdown into a series of smaller fields . We will detect all theparticles in a field and then move the specimen to the next oneand repeat the process. So the ability to move the stage insmall and precise increments is really important to our process(we dont want gaps or overlaps between the fields weanalyze).

EDXspectrometer Vacuum

Chamber

SE

detector

BSEdetector

Another way that Figure 1 is oversimplified is that it looks likewere examining our specimen out in the open. Thats not thecase. You need a pretty good vacuum in order to maintain afocused beam of electrons because the electrons in the beamwould be scattered all over the place if they encountered airmolecules. So an important aspect of the instrument is thevacuum chamber .

Figure 2: More Detailed System Diagram

Column ImageScree

Beam

Specimen

Detector

Stage

Figure 1: System Diagram

Figure 2 is a little more detailed than the first one. Once again,we have the focused beam being produced by the columnhitting the specimen mounted on the stage. This time we haveshown the specimen stage mounted inside of a vacuumchamber. Not shown in this figure, but obviously necessary,are the pumps and other equipment needed to evacuate thevacuum enclosure, but we dont need to go into those detailshere.

Another thing weve added to this figure is that were showing BSED IMAGING

Figure 3: BSED Diagram

the three principal kinds of detectors that are used for SEM:the Secondary Electron Detector (also called the SED), theBack-Scattered Electron Detector (also called the BSED), andthe x-ray detector, also called the EDX spectrometer. Werenot going to worry very much about the SED, since althoughits the most common kind of detector used for making prettypictures, we dont use it for automatic particle analysis. Onthe other hand, the BSE detector (BSED) is really important tous, so lets see why. You might have a hard time seeing theBSED in the above figure, because its a pretty small circulardevice tucked right under the pointed end ( polepiece ) of thecolumn. Figure 3 shows an exaggerated picture of what itactually looks like and does.

That round washer-like device represents the BSED itsabout 1 inch in diameter. It has a hole in the middle where theelectron beam passes through on the way to the specimen.The figure shows one of the incident electrons hitting an atomof the specimen and bouncing backwards so that it will hit thedetector. I trust that it is obvious that the scale of this picture isall out of whack electrons and atoms arent remotely that bigin comparison to the detector! But this picture still helps us toappreciate why the backscatter detector is so useful. As thefigure suggests, any solid material is mostly empty space theatoms are actually spaced quite far apart and so the incidentelectron may penetrate quite a ways before it hits anything. Asyou might also expect, the probability that the incoming


4/8


electron gets bounced back to the detector is going to dependon the size of the atoms in the specimen big atoms are easyto hit and will reflect a lot of electrons, whereas small atomsarent going to get hit so often. Thats exactly what happensand thats what Figure 4 shows.

The vertical scale of the graph is the back-scatter coefficient ,which is just another way of saying that its the probability thatan incident electron is back-scattered. The horizontal scale is

atomic number, which is proportional to the size/weight of theatom. Thus, materials with low atomic number (at the top ofthe periodic table) produce relatively few backscatteredelectrons and heavy elements (at the bottom of the table)produce a lot more. Consequently, materials that are mostlymade up of carbon, hydrogen, nitrogen, and oxygen ( organics )produce a w eak backscatter signal and look dark in an image.Those things we call minerals are somewhat heavier andappear a bit brighter, alloy metals (steel and brass) are stillheavier and appear brighter still, and so forth. This is a veryuseful effect since it allows us to easily discriminate classes ofmaterials just by looking at the brightness of the backscatteredelectron signal.

Figure 4: Backscatter Coefficient Vs. Atomic Number

To illustrate how useful the Backscatter Detector is, look at thepair of pictures in Figure 5. The picture on the left was takenwith an ordinary light microscope at about 50X magnification.Its a picture of metal particles removed from an automotivemechanism and mounted on a carbon-based adhesive. Itsreally pretty hard to discern how many particles are there andwhat their shapes are. Now look at the image on the right.This is exactly the same sample imaged via the BSED signal.Here the carbon-based adhesive looks essentially black(because it is low atomic number) and the metal particles standout clearly. Which image do you find easier to interpret? Andif the BSED image is a lot easier for you, imagine how mucheasier the problem is for computer software, which is still avery long ways from matching the sophistication of the humanvisual system!

Figure 5: Optical Image and BSED Image

ADVANTAGES OF SPECTROMETERS

But though the BSED signal is an important tool, the x-raysignal may be even more powerful. The reason for this isbecause the energy (or wavelength) of the x-ray emitted by anexcited atom is unique to the element that emits it. What thismeans is that if we measure the energies of the emitted x-rays,we can tell exactly which elements are present in thespecimen, and in what proportion. We call the x-ray detector aspectrometer, because it allows us to view the spectrum of xray energies that are emitted. Were not going to spend a lot oftime discussing this fascinating piece of technology, but Figure6 shows what an actual x-ray spectrum looks like.

Each of the peaks in this spectrum represents a particularemission, and labels have been applied to indicate theelements they arise from. This particular spectrum containssilver and iron emissions (the specimen was flecks of silver-coated steel). Now, if youre not used to looking at spectralike this, the pattern of peaks may seem mysterious, but x-rayemissions actually follow precise and relatively simple rulesthat make it fairly easy to identify what elements we are lookingat. And these rules are really robust it doesnt matter whatwe do to the specimen, whether we heat it, form chemicalcompounds, apply pressure, roughen the surface, or whatever

the characteristic spectral lines for iron (for example) willalways be found at the energies shown above.

So lets stop and summarize for a moment. What wevedescribed so far is a technology that allows us to look at reallytiny fragments of material and not only easily discriminateimportant classes of materials (i.e., minerals and metals) fromthe organic substrate on which they are presented, but oncewe have located these particles, we can tell precisely whattheir elemental composition is. And this works for a wide rangeof particle sizes from micron scale on the small end tomillimeter scale on the large end. O h yes, and we can alsotake very detailed pictures of these tiny objects, so we know

just about everything there is to know about them. And onelast thing to note: we dont have to work very hard to prepare

these samples for analysis. In fact, often its nothing more thancollecting the particles on some sticky tape and putting theminto the SEM. Are you impressed? You should be! Thepower and versatility of the SEM/EDX combination is such thatany materials laboratory worthy of the designation probablyowns one (or several) and uses it as one of its principal tools.


5/8


Figure 6: X-ray Emission Spectrum

SEM/EDX isnt a perfect tool, of course, and so well brieflyitemize some of the limitations. Probably the biggest limitationis that pesky vacuum requirement the simple fact is thatthere is no such thing as a material that is transparent to thepassage of electrons. Not only does this mean that we have toget rid of the air in our specimen chamber, it also means wecant look at things inside transparent containers like we dowith a light microscope. Further, since most liquids and manysolids evaporate at low pressures, there are lots of materialsthat we cant look at in a completely natural state (likebiological tissue, for example, which dries out and shrivels upwhen we put it in the vacuum). Then there is the fact thatelectrons are charged particles, and there can be a problemwith charge build-up when we try to view insulating materials.Finally, we need to recognize that although x-rays are emittedby all but the three lowest elements in the periodic table, x-rayanalysis isnt a very good tool for quantifying organiccompounds (which are mostly hydrogen, oxygen, and carbon).But, in the grand scheme of things, those are relatively modestlimitations to an extremely powerful technology, and by using

specialized instruments and techniques many of theselimitations can in fact be worked around.

ASPEX AND AUTOMATED ANALYSIS

What weve described so far is a relatively conventionalSEM/EDX system. In fact, we could be describing pretty muchany one of the tens of thousands of SEM/EDX installations thathave been in use around the world for the past 30+ years.

With a skilled operator sitting at the controls, any one of thosetens of thousands of units could do a credible job of examiningselected particles of interest. So if thats the case, whats sospecial about the ASPEX system as a particle analyzer? Theanswer lies in that word selected. The vast majority of those

tens of thousands of conventional SEM/EDX systems areuseful only for manual operation by a human operator. Theyare powerful machines that augment the human sensorysystem, but are completely non-functional unless human handsare present to manipula te their controls and human brains tointerpret their results. Its a powerful combination when the taskrequires insight and imagination, but if its a routine repetitivetask, its very inefficient.

Its now time to mention the most important parts of the systemdiagram in Figure 2: the supporting blocks of interfaces andcontrols and the way they are all interfaced to a controlling

computer. Thats the key to the power of the ASPEX systemsfor automated particle analysis. Rather than requiring a humanoperator to make decisions and operate controls, theseautomated systems are pre-programmed for their task, andthen accomplish it unattended. The net result is that they cananalyze collections of particles orders of magnitude faster thanany human can and do it round the clock without a break,and make fewer errors as well! Well now spend the rest ofour time seeing specifically how this is done.

One way we could automate the process is what is calledframe-based analysis. In this mode of implementation, aconventional SEM is used very much like a camera. Figure 7shows how frame-based particle analysis is done with acamera.

The frame-based analysis begins when the stage is positionedso that a field of view is presented to the camera (left figure).The camera then snaps a picture of that field and transfers itto the computer (middle). Software algorithms in the computer

then process this frame by locating the individual features and,by tracing around them, measuring them (right). As a finalstep, the computer may then direct the microscope optics toplace the beam at the coordinates of a particle so that an x-rayspectrum can be measured. Once all the particles in the fieldhave been analyzed, the stage is stepped to the next field, andthe process is repeated. If were doing this with a SEM and weare being smart, weve used our BSE detector to collect ourimages so that weve gotten a nice clear discriminationbetween our particles and the substrate theyre lying on.Because it is all automated, it can perform a lot faster than anyhuman operator. This all sounds pretty good, but it can bedone a lot better.

PARTICLE DETECTION

Heres how ASPEX systems perform the same task. Onceagain, the stage is positioned so that a field of view is centeredunder the optics. Instead of just snapping a picture of the field,however, ASPEX systems may subdivide this larger stagefield down into smaller mag fields that can be individuallydefined by deflecting the beam. (Note how the larger stagefield is divided down into 16 individual mag fields in Figure 8.)This is done so that we can work with lots of smallermanageable-sized fields rather than a few big ones. And sincewere moving between fields electronically, this is a lot fasterand more accurate than mechanically moving the stage.


6/8


Rather than capturing a high-resolution image of the field,however, the ASPEX system instead moves the beam acrossthis field in an array of fairly coarse steps. At each point, thebrightness of the BSED signal is noted. If the signal is brightenough to indicate that a particle is present at this position,then the software initiates a particle-sizing sequence as will beillustrated shortly.

Note the coarse grid of sampling points in Figure 9. Thespacing of the points is chosen to be no larger than thesmallest particle of interest, thus at least one hit is assuredfor every particle of minimum size. However, particles smallerthan the grid spacing will frequently be missed. In other words,the effect is like using a seive where particles too small to beof interest are allowed to fall through. That saves a lot oftime.

PARTICLE DATA ACQUISITION

When a particle is detected, a sizing sequence is initiated.There are several algorithms that can be used for this purpose,but for simple particle shapes, the rotating chord alogorithm isboth accurate and exceptionally fast. Figure 10 shows thesteps of the method.

The first step is to locate the center of the particle. This isdone by a bisected chord method as follows: (Left): Thecentering sequence begins when one of the sampling pointshits a particle (signal above detection threshold). The beam isthen moved horizontally in small steps until the signal againfalls below threshold. This defines the first horizontal chord.(Middle): The beam is then moved to the center of thishorizontal chord and then stepped first upwards until the upperextent of the particle is located, and then downwards to thebottom edge. When this is accomplished, this vertical chord isbisected and a new horizontal chord is established. (Right):By repeating the process of drawing a vertical chord, bisectingit horizontally, then bisecting the horizontal chord vertically, theprocedure converges quickly to a point where the vertical andhorizontal chords cross at the their respective centers this isthe geometric center of the particle.

Once the particle center is found, the last step is the rotatingchord process illustrated in Figure 11. A series of chords aredrawn through the particle center at equal angular spacing asshown. These chords allow us to provide some very usefulmeasures of the particle size and shape. For example, thelongest and shortest chords are a measure of the aspect ratio,or we can average the chord lengths for an average diameter.By connecting the tips of the chords, the perimeter and area ofthe particle are also determined.

Finally, the beam is placed again in the center of the particle,and its x-ray spectrum acquired.

Though this may sound like a great deal of activity, it allhappens very fast. In fact, for the vast majority of cases, thisprocedure locates and sizes particles several times faster thanthe frame-capture method (which is way faster than a humancould do it). The major reason for this improvement in speed isbecause it only spends time collecting detailed data whereparticles are known to be present, rather than wasting timecapturing and transferring vast numbers of empty pixels.Since there is almost always much more empty space on the

specimen then space occupied by particles, the result is a bigspeed advantage.

This method is also intrinsically more accurate since thechords can be drawn with arbitrarily fine spacing, whereasframe-based methods are restricted to the pixel spacing of the

captured frame image. The advantage of the ASPEX methodof dynamic sizing is particularly apparent when both large and

small particles are present in the same field. For a frame-based method to deal with this situation either the pixels mustbe made rather coarse (at the expense of precision inmeasuring small particles) or a huge array of small pixels mustbe captured, at the expense of speed. By contrast, the ASPEXmethod of dynamic sizing has no difficulty dealing with bothlarge and small particles in the same field it adjusts to theprecision needed.

Theres a great deal more that could be said about thismethodology, but this isnt the place for such detail. However,it probably should be mentioned that the rotating chord methodwill obviously fail for certain kinds of complex particle shapesas illustrated by the two shown in Figure 12. ASPEX has alsodeveloped a complex feature dynamic sizing algorithm thataccurately handles shapes of arbitrary complexity, albeit atsome cost in speed.

PARTICLE CLASSIFICATION

Once each particle is fully characterized (size, shape, andelemental composition) user-defined rules are then used toassign the particles to meaningful classes. For example,aerosol particles might be classified by their size, aspect ratio,and composition to correspond to various inhalation riskcategories. Again, this is automatically done by the software.

Once all of the analysis and characterization is performed forone mag field, it jumps to the next mag field and does thesame thing until all of the mag fields have been analyzed. Atthat point, the stage is moved to the next stage field and thewhole process repeated until the predefined area of the samplehas been covered.

How fast does this all go? It depends a lot on the specimenand what we are trying to accomplish. In favorable cases, ifwe dont need to collect x-ray spectra, we can processparticles at rates of something like 500/minute. If we need tocollect x-ray spectra for each particle, that can add a fewseconds per particle. By the standards of some technologiesused for particle analysis (such as flow counters, where theparticles are in a fluid passing rapidly in front of a sensor) thisis a relatively slow analysis. But given the quantity and qualityof information obtained (not just the number of particles, but

accurate assessment of dimensions and composition) electronbeam particle analysis is a marvelous technology that givesanswers that others cant.

DATA REPORTING

At the end of the complete specimen analysis, the results ofthe analysis are available as a large table containing thespatial coordinates, and size, shape, and compositionparameters for every particle analyzed. In a typical run, thismight range from a few hundred to many thousand particles.Accompanying each particle is a little thumbprint image so wecan see exactly what it looked like. And if we need to, we can


7/8


Stage moves tofield

Camera capturesframe

Figure 7: Frame-based Particle Analysis With A Camera

Software tracesparticle

Figure 8: Mag Fields in a Stage Field Figure 9: Sampling Point Grid

Figure 10: Locating the Center of a Particle


8/8


Figure 11: Rotating Chords Figure 12: Failure of Rotating Chord Method

tell the system to relocate any particle on the specimen thena human operator can analyze it in even greater detail. Thatsreally important in Gun Shot Residue detection where aforensic scientist must testify to having personally viewed theGSR particles.

The last step is to generate the final report. Software tools areprovided that allow customized reports to be readily configuredto suit the needs of individual customers. Some customerswant a lot of detail in their reports; other applications might onlywant a Go/NoGo indication.

There is actually one further step possible the results of allthe analyses may be exported to a database . This allows theuser to efficiently monitor long-term trends for purposes ofimproving the product or process.

CONCLUSION

So thats the big picture. It should be apparent that ASPEXsystems utilize relatively conventional mechanical andelectronic system components functionally similar to thosefound in conventional SEM/EDX instruments. The highlyrefined software provided with our systems provides a greatdeal of our unique value, but the ultimate value comes from theway all components of the system are designed to worktogether in a highly efficient manner for automatedapplications.

Introduction to Automated Particle Analysis

Documents