FEI provides various standard configurations that provided integrated solutions for various industrial and research applications of Automated Mineralogy based on the Quanta series of Scanning Electron Microscopes.

These standard configurations are fitted with multiple Backscattered Electron (BSE) Detectors and Energy Dispersive Spectrometers (EDS).

Quanta Microscopes also provide, as standard, the capability to use variable pressure ESEM operation.

Quanta Microscopes use an Everhart-Thornley Secondary Electron (ETSE) detector. ETSE detectors cannot be used in ESEM mode and a special high pressure ESEM detector is required.
All systems can be used stand-alone in ESEM mode without modification. However ESEM mode can only be used independently of the Automated Mineralogy software.

• BSE detectors can be used in ESEM mode.
• Conventional secondary electron imaging is not used in Automated Mineralogy operation and cannot operate in ESEM.
• EDX analysis gives poor results in high pressure mode.
• The Automated Mineralogy software does not integrate ESEM information.

If an ESEM detector is installed, it is rather large and requires a special port and the detector may interfere with AM operation.
Non-standard configuration like this would need to be specified in detail in consultation with FEI technical staff.

FEI provides various standard configurations that provided integrated solutions for various industrial and research applications of Automated Mineralogy based on the Quanta series of Scanning Electron Microscopes.

These standard configurations are fitted with multiple Backscattered Electron (BSE) Detectors and Energy Dispersive Spectrometers (EDS).
Wavelength Dispersive Spectrometers (WDS) systems can be fitted to the Quanta SEM chambers. However, 

• WDS is a rather large piece of equipment inside the chamber and limits the number of EDS detectors that can be fitted.
• WDS can be operated independently of the Automated Mineralogy software.
• WDS can be integrated with other x-ray analysis software that combines EDS and WDS.
• WDS cannot be integrated automatically with the Automated Mineralogy software.

Non-standard configuration like this would need to be specified in detail in consultation with FEI technical staff.

iDiscover calculates the average density per sample or fraction using the area % of each mineral and the density of each mineral phase.
This example table shows the steps.
Vol% Mineral Density Mass% Mass Unit

1. Measure: Volume %: Counting points, lines or areas.
2. Estimate Mineral Densities: Look up text books or Mineral Data Bases
3. Calculate: Mass Units: Volume x Density for each mineral phase
4. Calculate Total Mass Units: Sum the Mass Units
5. Mass %: Divide Mass Units for each Mineral by Total Mass Units x 100
6. Average Density: Divide Total mass / Total Volume

Species Identification Protocol (SIP) development is the creation of a library of target chemical compounds that are expected in a given sample and which are later interpreted as minerals in a primary mineral list.

SIP development is a science, a skill and an art. It requires a combination of knowledge about mineral occurrences, mineral composition, their metallurgical importance and knowledge about Energy Dispersive X-ray (EDX) spectra especially elemental peaks positions and their possible overlaps. It also requires knowledge about x-ray generation volumes around the analysis point and the effect that density has on spatial resolution. It therefore requires the analyst to have relevant domain knowledge in order to make decisions and interpretations. The degree of detail required depends on the application and the time available for development.

The SIP definition can be as broad or as specific as required. Broad categories are good for getting going. Specific important broad categories can be progressively refined as further detail is obtained.
Generally for SIP development, an example of the mineral or a standard makes the job a lot easier. The major challenges for rare but important phases such as PGMs are the large number of host phases and their rarity.

It is impossible to prepare standards of all the required phases from scratch and SIP development relies on the creation of a basic set from “theoretical” considerations followed by a continuous improvement of investigative work as each new occurrence is located and added to the SIP list. It is possible but very difficult to do this remotely, i.e. off-line, using the SIP Editor functionality alone. It is best done on-line with actual samples.

As part of this process, measurement made will detect compounds that require further on-line analysis. These usually need interactive investigation using the detailed quantitative EDX analysis capability.
Many minerals are non stoichiometric and have wide ranges of composition even for a named mineral. For example the (Pt,Pd,Ni) Sulphides cooperite and braggite both have a wide range of composition. Many PGMs show significant and detectable elemental substitution which makes the naming convention awkward and often not practicable as a target list in a SIP.

One of the many great things about QEMSCAN is the ability of the software to record the x-ray spectrum analysis from every single analysis point regardless of its classification and replay the measurements off-line against a new SIP under development.
This ability to replay the original measurements is used to great advantage in training the system to recognise minerals and in continually updating the SIP as new phases are found.

Misidentification
During this development process, there will inevitably be misidentification problems.

Many of these issues can be addressed by a careful review of the SIP entries, better grouping at the primary mineral level and better definitions. This could take some time. There are usually many parts to a solution. The name associated with a specific category is assigned by the SIP developer and usually relates to the phase that was used in the original definition. A check on the naming of categories may solve the problem. A better definition of the mineral will improve the situation and a better definition of the phase with which it is confused will also have a major impact.

Elemental Overlaps
These problems occur because nature can be unkind to us. Some elements have similar peak energies and are harder to resolve.
As far as overlaps go, there are some overlaps that are impossible to resolve using any Energy Dispersive x-ray system. In these cases, if they are important, mineralogy knowledge can be used to discriminate them or they can be grouped together or the identification can be based on existence of other “marker” elements. In general, a higher atomic element overlaps with a lower atomic number element at a specific energy window. The contribution of the high atomic number element to the average atomic number of the mineral can also used in further discrimination – possible using BSE.

EDS mineral misidentifications generally fall into several categories.

• Poor definitions that can usually be fixed by some detailed investigative work.
• Elemental overlaps which are impossible to resolve using EDS systems in general no matter how detailed the spectra.
• Elemental overlaps which are difficult to resolve using EDS detectors. (e.g barium and titanium)
• Polymorph compositions which are not unique but depend on crystal structure (e.g. graphite and diamond)
• Mineral definitions on boundary phases that produce mixed spectra and which are not recognised as boundaries but misidentified as another mineral (e.g copper sulphide, pyrite boundaries).
• Higher energy lines have either too high an energy to be excited at all or not efficiently enough to be detected at standard operating conditions.
• Lower energy lines (e.g carbon, oxygen, fluorine & sodium) that are absorbed by the detector window and are difficult to quantify and are essentially qualitative.

Nevertheless, it is usually possible to get very reliable mineral discrimination if you keep in mind that SIP development is a science, a skill and an art.

1. QEMSCAN acquires a rapid EDX spectrum with a low photon count (typically 1000 photons) at each analysis point in a few milliseconds.
2. The EDX spectrum is analysed by a proprietary Spectral Analysis Engine (SAE) especially designed for analysing low-count spectra to determine the elements present and their relative proportions (the mineral composition).
3. The composition derived from a low photon count in the spectrum is accurate but the precision is sufficient to identify most rock forming minerals.
4. For more precise composition, more photons are used in the spectrum.
5. The composition is then compared to a library of saved mineral composition definitions.
6. The mineral definitions can be very broad such as “Iron Sulphide” which must have Iron and Sulphur but the proportions don’t matter that much or very narrow such as Pyrite which has a specific composition Fe 47% S 53%.

Magnification is a comparison of the size of an image of an object to the size of the original object, as viewed by an observer.

For any work related to images, such as those used in satellite imagery, astronomy, microscopy, photography etc, the scale of the image is the fundamental property, not the magnification.

With a telescope or optical microscope, it is possible to view an object with one eye through the instrument and with the other eye directly. With SEM it is possible to view the image on the Monitor of the computer. The magnification is then obtained by comparing the size of the image with the size of the original.

Once the image is recorded and reproduced, the size of the image, and hence, the magnification depends on the scale at which the image is reproduced.

A stored image can be reproduced at any scale whatsoever and the "magnification" value used when it was recorded is irrelevant and usually very misleading.

If the image is printed on a piece of paper it can have any magnification value, depending on the scale at which it is printed. If an image is project onto a screen, it can have any magnification value depending on, the lens system in the projector and the proximity of the projector to the screen.

It is very important to include on displayed images a scale bar that changes size in proportion as the image is enlarged, reduced or cropped rather than the “magnification”. To achieve this it is essential to record, with each image, the field width of the original image or the pixel size of the original image in physical units.

Grain size variation of the main value bearing phases and composites is an important parameter in ore characterisation assessment. This parameter has been successfully employed, in conjunction with other QemSCAN indicators, to predict metallurgical responses from exploration and mine samples.

An estimate of the average grain size of each mineral can be obtained from the distribution of their intercept lengths. As you know, size is a subjective measurement that depends on shape and other factors. We have commonly used two methods of estimating size: Estimated Mean Sieve Size (EMSS) and Phase-Specific Surface Area (PSSA).

EMSS is an empirically calibrated measurement of the geometric mean intercept length. The calibration is performed by comparing the actual mean screen size for a given fraction with the geometric mean of the intercepts on the particles measured from that fraction. The same calibration factor is then applied to the geometric mean intercept length of the individual phases.

The PSSA grain size is derived from the Phase-Specific Surface Area (PSSA). The PSSA is a stereologically correct, and therefore, unbiased estimate of the surface-area per unit-volume of the phase. A large surface area per unit volume is indicative of a fine-grained mineral and vice versa. Rather than using the PSSA, which has dimension 1/L and is therefore larger for smaller particles, we derive a grain-size from the PSSA by using the diameter of an equivalent sphere with the same surface-area per unit-volume. The PSSA derived grain size is thus D=6/PSSA.

Generally the two values are very similar, but you will find in our reports that we now only use PSSA derived grain size data in preference to the EMSS data because the former is a stereologically-correct, and we have found it to be more robust and better correlated with grind-size estimates in metallurgical applications.

The Surface Area per Unit Volume of the particles themselves is used in an identical way to derive an equivalent particle size. The “All Material” category in the first row of our tables gives the mean size for the particles themselves derived by this method. This size determination is very robust and is correlated very well with particle size determined from screening.

For metallurgical and geological studies, we combine results for different samples, different size fractions, calculate grade and recovery, compare chemical assays and mineralogical results, derive mineral and elemental distributions and perform various other calculations.

These calculations involve Modal Analyses, Chemical Assays, Particle Size Distributions (PSD) and Mineral Composition.
Modal analysis measurements are volumetric measurements. Chemical assays and mineral composition measurements are mass measurements. To combine results, we need to work in either mass or volume throughout.
We could:

• Convert the mass values used in chemical assays, PSD and mineral compositions to volume and work entirely in volume, or
• Convert the modal analysis volume values to mass and work entirely in mass.

We have chosen to convert modal volume percent values to mass percent.
iExplorer does all this for you. It converts the volumetric measurements into mass values by using the mineral density. All that is required is that the mineral densities be represented correctly. The density can be used automatically from the minerals data base, or entered manually into the mineral properties.

Calculated mineral mass results obtained depend on user-defined mineral densities in the primary mineral list.

Similarly chemical elemental results obtained depend on user-defined mineral compositions in the primary mineral list.

To ensure accurate results, the user must ensure that accurate estimates of both density and composition are available for all phases.

You cannot average density by weighting the values by weight %.
To average density across fractions, you need to calculate the mass and the volume of the each fraction, calculate the total mass and total volume and then divide the total mass by the total volume.

Given the weight % in a fraction,

1. Calculate the volume represented by each fraction of the sample. This is done by dividing the mass % by the density.
2. Sum the mass and sum the volume.
3. The combined (or average) density is then the total mass divided by the total volume.
   

An example:

If you multiply each mineral density by the percentage of mass, add them up and divide by 100 you will get the incorrect answer 2.775.

Chemical assays are expressed as mass percent of an element. Measurements in section are estimates of volume percent of a mineral.How do we reconcile chemical assays with mineral percentages?

To average density across fractions, you need to calculate the mass and the volume of the each fraction calculate the total mass and total volume and then divide the total mass by the total volume.

Given the weight % in a fraction,

1. Calculate the volume represented by each fraction of the sample. This is done by dividing the mass % by the density.
2. Sum the mass and sum the volume.
3. The combined (or average) density is then the total mass divided by the total volume.

An example:

If you multiply each mineral density by the percentage of mass, add them up and divide by 100 you will get the incorrect answer 2.775.

The temperature reached at a given spot of depends on:

• Energy of the each electron 15 or 25 KeV (kiloelectronvolts) - KeV is a (very small) unit of energy. 1 kiloelectronvolt = 1.60E-16 Joule
• The number of electrons/second – the beam current (Amperes)
• How long you leave the beam there Joules
• Leaving the beam for a few seconds on epoxy will burn a hole in the epoxy.
• Electron dispersion in the sample
• A function of both electron beam energy and the sample properties
• Thermal characteristics of the sample
• Conductivity - is the heat conducted away
• Reactivity - does heat convert it to another phase
• Moisture - does it evaporate

The run-length distribution for each mineral (and the entire particle) is used in calculating this property. This distribution is very skewed (has a long tail to the right).

One method of analysing such distributions is to transform the data by some method or other to produce a "normal" looking distribution so that the statistics of normal distributions can be used. In this case the transformation is made by taking the log of the intercept length. This reduces the skewness because the log of long intercepts is a small number.

If you then take the arithmetic mean of this log distribution, it turns out that it is equivalent to the geometric mean of the intercept distribution. Geometric mean = nth root of the product of the intercept lengths. Log(a.b.c.d) = log(a) + log(b) + log(c) +log(d).
This is the geometric mean size. This method is rather empirical as it deemphasises long intercepts which carry a lot of a given mineral and has no basis whatsoever in stereology. It also involves a calibration fudge factor for screen size fractions which is then applied to the mineral value (1.8 from memory).

We don't use it much any more. We generally use the 3-D Surface Area per Unit volume (S/V) which is estimated in a non-biased way from the "# of intercepts ends" per "intercept length" in 2D.  This is the PSSA. The PSSA amounts to the arithmetic mean of the intercept distribution.

We then convert PSSA to "size" by assuming an equivalent sphere diameter. D=6/PSSA. We have found this “size” to be a very robust number that correlates very well with grind size when used as a predictor. It is used widely now to estimate grind size from ore characterization samples.

Wavelength Dispersive X-ray Spectroscopy (WDS or WDX) and Energy Dispersive X-ray Spectroscopy (EDS or EDX) are complementary X-ray microanalysis techniques. Both methods measure x-ray energies to identify individual elemental components in a sample, but with different approaches. In either case, x-ray photons are emitted by the bombardment of the sample by an electron beam in an electron microscope.

In EDS, data are collected for all energies at once, and displayed as a histogram of counts versus x-ray energy.

In WDS, x-rays are separated using diffraction, and individual wavelengths are detected at different spectrometer positions one at a time.

The chief advantages of WDS are better peak resolution and decreased noise from electronics or stray radiation. WDS provides better peak separation and increased peak-to-background ratios. WDS is most useful when:

• Compared with EDS, resolution is typically several orders of magnitude better; peak overlaps are virtually eliminated.
• The high energy resolution leads to order-of-magnitude better sensitivity than EDS. The sensitivity improvement is often most dramatic in the light element region.
• The better peak-to-background ratios and elimination of overlaps yield more accurate quantitative analysis for those elements in low concentration or involved in the overlap.
• Because the WD spectrometer can be positioned at the center of the peak, where peak-to-background is maximized, WDS provides superior x-ray maps, especially for low concentrations.
• The serial nature of spectrum acquisition means that analysis is slow compared to EDS.
• Because of the relatively inefficient diffraction process and the long detectore distances involved, considerably higher specimen currents (up to 100 nA, typically) are needed for WDS compared to EDS.
• WDS is often operated by setting the crystal directly on a specific peak. The acquisition time is devoted then only to the points of real interest. For minor elements, WDS then becomes faster relative to EDS, which spends most of its time counting unimportant x-rays from the major elements.

The chief advantages of EDS are that the detector is much cheaper, it has no moving parts, data are collected from all energies simultaneously and acquisition is many orders of magnitude faster.
The chief disadvantage of EDS are that:

• The elemental detection limits are around 0.5-1 %.
• Light elements are not well discriminated.

The Full functionality of the SEM and X-ray Analysis is retained

FEI’s Automated Mineralogy capability is an add-on to an FEI SEM with EDS X-ray detectors.

• All the functionality of the electron microscope and the X-ray analysis system are retained, and
• Many additional features are gained through the addition of FEI’s Automated Mineralogy Solutions.

Mineral analysis system

FEI’s Automated Mineralogy Systems identify and measure minerals. Mineral classification is done on-line on the basis of chemical composition using X-ray analysis. A mineral map of the sample – in particulate form or as a rock section – is generated with a pixel spacing as low as 1 micron. Mineral identification is done at all points on the image.

With the SEM alone images are BSE, SEI or elemental maps only

Fully automatic operation

FEI’s Automated Mineralogy systems are fully automatic and can measure multiple samples without manual intervention. Typically they operate un-attended 24 hours per day giving very high productivity.

With the SEM alone operation is generally interactive

Application to a wide range of mineral systems

FEI’s Automated Mineralogy Systems bring more than 25 years of experience of measuring a wide range of minerals and ore types. Mineral identification programs exist for all well-known mineral types. For new applications there is flexible software to enable the user to add new minerals.

With the SEM alone you must accumulate all this experience yourself

Ready to go!

FEI’s Automated Mineralogy Systems are “ready to go!” There is no need to spend a long time developing your own programs to carry out your particular analyses. You can be measuring samples within a week of installation.

With the SEM alone you must develop your own systems

Special purpose image analysis software for metallurgical interpretation

A wide range of software is available for analysis of the images so that results can be presented in a way that highlights the interesting and important features. Outputs have been designed with metallurgical and geological use in view. The results are generally used for ore characterization, ore testing, plant optimization and plant auditing, e.g. mineral modal analysis, mineral grain sizes, mineral associations, mineral surface areas, liberation, stereological correction, grade recovery curves, etc.

With the SEM alone, you must develop your own interpretation programs.

Special Purpose Software For Sample Investigation

Features of the operation are available to enhance manual SEM investigations. Particles and minerals of particular interest can be readily located, investigated using the mineral identification system, spectra collected and analyzed. This converts the SEM to a point mineral analyzer.

With the SEM alone, only the basic SEM operations are available

Sample Management System

FEI’s Automated Mineralogy systems have special-purpose software that keeps track of the large number of samples measured and sets up the measurement parameters for the analyses.

With the SEM alone, the operator must do sample logging and tracking.

Off-Line Image Analysis Software

Once the samples are measured the results are stored as images, which can be investigated and analyzed by stand-alone software installed on any PC. This again increases the productivity of the system.

With the SEM alone, the software all resides on the SEM.

Windows Based System

The Automated Mineralogy software is Windows based making it very easy to use and efficient to operate. All measurement and data analysis programs are accessed through customized screens.

FEI’s Automated Mineralogy systems are state-of-the-art products, which make maximum use of the SEM when it is used for mineralogical work. It expands the capability of the SEM and greatly increases its productivity through automation.