The structure behind the property
Two bars of the same steel, cut from the same melt, can differ by a factor of three in strength and by a factor of ten in toughness. Nothing in their chemical analysis tells them apart. What differs is their microstructure: the size of their grains, which phases are present, how the carbide is spread out, how many dislocations each holds. The earlier levels of this course explained why those features control properties, from the Hall–Petch effect in elasticity, plasticity and strengthening to the tempered martensite of processing and heat treatment. This lesson is about how materials scientists actually see them.
The features span an enormous range of sizes: grains of structural steel tens of micrometres across, pearlite plates a fraction of a micrometre apart, hardening precipitates in an aircraft aluminium alloy a few nanometres, atoms a quarter of a nanometre apart. No single instrument covers all of that, so materials scientists move between light, scanning electron, transmission electron and probe microscopes, choosing each for the question in hand. X-ray diffraction tells you which crystals are present on average; microscopy tells you where they are, how big, and what is wrong with them.
Resolution, contrast and why magnification is not enough
A microscope has two jobs. It must resolve, keeping two nearby features distinct instead of blurring them into one, and it must give contrast, making them look different from their surroundings. Magnification comes a distant third: once the image is big enough for the eye to see everything the instrument resolved, enlarging it further shows bigger blurs, not more detail. Microscopists call that empty magnification.
Resolution is limited by the waves that carry the image. A wave passing a fine feature spreads out by diffraction, and a lens collects only the part that comes towards it, so two points closer than roughly half a wavelength, divided by how wide a cone the lens accepts, cannot be separated by any lens, however perfect. Every step down the scale ladder needed a shorter wavelength or, in the probe microscopes, a way to do without lenses altogether.
Contrast comes from physics too, and different instruments get it from different physics: reflected light, electrons knocked out, diffraction by crystals, the force on a sharp tip. Knowing where the contrast comes from is what lets you read an image correctly.
Light microscopes and the Abbe limit
Definition— numerical aperture
The numerical aperture of an objective lens is , where is the half-angle of the cone of light the lens accepts from a point on the specimen and is the refractive index of the medium between the specimen and the lens (1.00 for air, about 1.52 for immersion oil).
Definition— resolution limit
The Abbe limit for the smallest resolvable spacing is . The closely related Rayleigh criterion, for two point sources whose diffraction discs just overlap, is .
Example— What a metallurgical microscope can resolve
Take green light, nm, near the peak of the eye's sensitivity. A 10× objective with NA 0.25 has an Abbe limit of nm, so it shows grains but not pearlite. A 40× objective with NA 0.65 gives 423 nm; a dry 100× objective with NA 0.90 gives 306 nm; a 100× oil-immersion objective with NA 1.40 gives 196 nm. By the Rayleigh criterion the same four objectives give 1342, 516, 373 and 240 nm. Either way, about 0.2 µm is the end of the road for ordinary light microscopy: the plates of fine pearlite, about 0.1 to 0.2 µm apart, are at the limit, and a precipitate 10 nm across is invisible however hard you look.
The eye resolves about 0.2 mm at a comfortable reading distance, so the useful total magnification is roughly times. Microscopists' rule of thumb is a useful magnification of 500 to 1000 times the NA; anything beyond is empty.
The same physics fixes the depth of field, the thickness of specimen that is in focus at once, which shrinks roughly as : about 0.7 µm for the NA 0.90 objective. That is why specimens for high-magnification light microscopy must be flat to well under a micrometre, and why a rough fracture surface is examined in the scanning electron microscope instead.
Materials are mostly opaque, so a metallurgical microscope works in reflected light, sent down through the objective. Polarised light shows the grains of non-cubic metals such as titanium and zinc without etching, because their reflectivity depends on orientation; differential interference contrast turns height steps of a few nanometres into colour; and confocal and interferometric microscopes map heights to nanometres. The Abbe limit is a limit on lateral resolution only.
Metallography: preparing a surface to look at
A fracture surface or a casting skin hides its microstructure. Metallography, founded by Henry Clifton Sorby in Sheffield, who in the 1860s was the first to polish and etch steel and examine it under a microscope, makes a flat section that reveals it. The steps are always the same, in a laboratory with proper equipment:
- Sectioning. A sample is cut from the part with an abrasive cut-off wheel cooled by liquid, so that the heat of cutting does not change the structure being studied.
- Mounting. The sample is set in a puck of polymer resin so it can be held flat and its edges protected.
- Grinding. Successively finer abrasive papers remove the cutting damage, each step removing the scratches of the one before.
- Polishing. Diamond suspensions down to a micrometre or less, then often colloidal silica, leave a mirror finish with a deformed layer far thinner than the features to be seen.
- Etching. A chemical etchant chosen for the alloy (ASTM E407 lists hundreds) attacks the surface unevenly. Grain boundaries, being disordered, dissolve faster and become fine grooves that scatter light and look dark; grains of different orientation, or different phases, etch at different rates and reflect differently.
Etchants are corrosive and some are toxic, so etching is done by trained staff in fume cupboards with eye and skin protection, never at home.
The result is the classic micrograph of ferrite grains, pearlite colonies, martensite needles or cast dendrites, still the first look in almost every failure investigation and steelworks quality laboratory, because it is fast, cheap and shows millimetres of sample at once.
Scanning electron microscopy
The scanning electron microscope (SEM) does not form an image with lenses at all. Its lenses focus an electron beam, accelerated through 1 to 30 kV, into a probe a few nanometres across, which is swept over the specimen line by line while detectors count what comes out; each count becomes one pixel. Because the beam converges at only a few milliradians, the depth of field is hundreds of times larger than a light microscope's, which is why fracture surfaces, with their dimples, cleavage facets and fatigue striations (fracture and toughness), are examined in the SEM.
The beam electrons do not stop at the surface. They scatter many times inside the specimen and spread into a pear-shaped interaction volume, and each signal comes from a different part of it.
- Secondary electrons (SE) are knocked out of the specimen's atoms with energies below 50 eV, so they escape only from the top few nanometres. Their yield rises where the beam meets the surface obliquely, so edges look bright and the image reads like a lit landscape: topographic contrast.
- Backscattered electrons (BSE) are beam electrons bounced back out. Heavier nuclei bounce more of them, so phases rich in heavy elements look bright: atomic-number contrast, ideal on flat polished sections. The yield also depends slightly on how the beam lines up with atomic planes, giving channelling contrast between grains.
- Characteristic X-rays identify the elements present (energy-dispersive spectroscopy, in spectroscopy and thermal analysis).
- Electron backscatter diffraction (EBSD): with the specimen tilted at 70°, backscattered electrons diffract into a pattern of bands that software indexes to the crystal orientation, to about half a degree, at thousands of points per second, giving maps of grains, boundaries, texture and phases.
How deep the beam goes can be estimated with the Kanaya–Okayama range, an empirical formula that grows with beam energy as and falls with density. The block computes it, with an empirical fit to measured backscatter coefficients , for several elements.
# Electron range (Kanaya-Okayama) and backscatter coefficient (an empirical fit) for some elements.
elements = { # Z, molar mass (g/mol), density (g/cm^3)
"C": (6, 12.011, 2.26),
"Al": (13, 26.982, 2.70),
"Si": (14, 28.086, 2.33),
"Fe": (26, 55.845, 7.87),
"Cu": (29, 63.546, 8.96),
"W": (74, 183.84, 19.3),
"Au": (79, 196.97, 19.3),
}
def ko_range_um(Z, A, rho, E_keV):
"""Kanaya-Okayama electron range in micrometres."""
return 0.0276 * A * E_keV**1.67 / (Z**0.89 * rho)
def backscatter(Z):
"""Fraction of beam electrons backscattered at normal incidence (empirical fit)."""
return -0.0254 + 0.016 * Z - 1.86e-4 * Z**2 + 8.3e-7 * Z**3
print("element Z range at 5 kV at 15 kV at 30 kV (um) backscatter eta")
for sym, (Z, A, rho) in elements.items():
r = [ko_range_um(Z, A, rho, E) for E in (5, 15, 30)]
print(f"{sym:>7} {Z:3d} {r[0]:13.2f} {r[1]:8.2f} {r[2]:13.2f} {backscatter(Z):15.3f}")
# Atomic-number contrast between two neighbouring phases in a backscattered-electron image
print()
for a, b in [("Al", "Si"), ("Fe", "Cu"), ("C", "Fe"), ("W", "Au")]:
ea, eb = backscatter(elements[a][0]), backscatter(elements[b][0])
print(f"BSE contrast between {a} and {b}: (eta_{b} - eta_{a}) / eta_{b} = {(eb - ea) / eb:.3f}")At 15 kV the beam penetrates 2.6 µm into aluminium but only 0.99 µm into iron and 0.53 µm into gold; at 30 kV, 8.2 µm of aluminium. The backscatter coefficient climbs from 0.064 for carbon to 0.279 for iron and 0.487 for gold. Neighbouring elements give weak contrast, 7 to 8 per cent between aluminium and silicon or iron and copper and only 2 per cent between tungsten and gold, while a carbon-rich inclusion in steel stands out at 77 per cent.
The range numbers explain SEM users' habits. Low voltages, 1 to 5 kV, keep the signal near the surface, sharpen fine surface detail and reduce charging, the build-up of trapped electrons on ceramics and polymers that deflects the beam; otherwise insulators are coated with a few nanometres of carbon or metal, or examined with a little gas in the chamber. A focused ion beam (FIB) of gallium or xenon ions in the same chamber mills material away with nanometre precision, cutting cross-sections at a chosen spot, slicing volumes for three-dimensional reconstruction and lifting out thin foils for the TEM.
Transmission electron microscopy and atomic resolution
In 1931 Max Knoll and Ernst Ruska in Berlin built the first electron microscope; by 1933 Ruska's instrument beat the light microscope's resolution. The transmission electron microscope (TEM) works like a projector: electrons accelerated through 80 to 300 kV pass through a specimen thin enough to transmit them, typically under 100 nm, and magnetic lenses form an image, or a diffraction pattern, of what comes out.
Electrons at these energies travel at well over half the speed of light, so their de Broglie wavelength must use the relativistic momentum. The block computes it and compares it with what the lenses actually achieve.
import math
h, m0, e, c = 6.62607015e-34, 9.1093837e-31, 1.602176634e-19, 2.99792458e8
def electron_wavelength(volts):
"""de Broglie wavelength of an electron accelerated through `volts`, with the relativistic correction."""
p = math.sqrt(2 * m0 * e * volts * (1 + e * volts / (2 * m0 * c**2)))
return h / p
print("For comparison, green light: 550 000 pm\n")
print("accelerating voltage wavelength (pm) without relativity speed v/c")
for kV in (1, 5, 30, 100, 200, 300):
V = kV * 1e3
lam = electron_wavelength(V)
lam_classical = h / math.sqrt(2 * m0 * e * V)
gamma = 1 + e * V / (m0 * c**2)
print(f"{kV:15d} kV {lam * 1e12:15.2f} {lam_classical * 1e12:18.2f} {math.sqrt(1 - 1 / gamma**2):9.3f}")
# A round magnetic lens always has spherical aberration (Scherzer, 1936). Balancing it against
# defocus gives the "point resolution" d = 0.66 Cs^(1/4) lambda^(3/4), at an aperture angle lambda/d.
lam = electron_wavelength(200e3)
print("\n200 kV TEM, spherical aberration coefficient Cs:")
for Cs_mm in (1.2, 0.5, 0.01):
d = 0.66 * (Cs_mm * 1e-3) ** 0.25 * lam**0.75
print(f" Cs = {Cs_mm:5.2f} mm point resolution {d * 1e12:4.0f} pm "
f"(= {d / lam:3.0f} wavelengths), aperture angle {lam / d * 1e3:4.1f} mrad")At 200 kV the wavelength is 2.51 pm (the non-relativistic formula would give 2.74 pm, 9 per cent too long), about 220 000 times shorter than green light. Yet a TEM does not resolve far below an atom, because magnetic lenses are poor lenses. Otto Scherzer proved in 1936 that any round, static electron lens has positive spherical aberration, focusing rays far from the axis too strongly. The usable aperture is then only about 10 milliradians, and a typical uncorrected 200 kV instrument with mm has a point resolution of about 245 pm, nearly a hundred wavelengths. Scherzer also showed the way out, with lenses that are not round; in 1998 Maximilian Haider, Harald Rose and Knut Urban's group demonstrated a working aberration corrector, and Ondrej Krivanek's group built correctors for the scanning version soon after. With near zero the formula would promise 74 pm; in practice chromatic aberration, higher-order aberrations and instrument stability set the limit, now around 50 pm in the best microscopes.
The TEM gives several kinds of contrast from the same specimen:
- Diffraction contrast. An aperture passes only the undiffracted beam (bright field) or one diffracted beam (dark field). Regions bent towards a Bragg angle diffract strongly and look dark in bright field; the distorted lattice around a dislocation does exactly that, which is how Hirsch, Horne and Whelan first watched dislocations move in 1956. Precipitates, stacking faults and boundaries show up the same way.
- Selected-area diffraction shows the diffraction pattern of one chosen grain or precipitate, identifying its structure and orientation.
- High-resolution (phase-contrast) imaging lets several diffracted beams interfere, giving fringes with the periodicity of the lattice, so rows of atoms can be seen end-on.
- Scanning TEM (STEM) scans a probe, well under 0.1 nm wide in a corrected instrument, and an annular detector collects electrons scattered to high angles. This high-angle annular dark field (HAADF) signal grows roughly as to , so heavy atom columns are bright and single heavy dopant atoms can be seen in a light matrix; the same probe excites X-rays and energy losses for spectroscopy column by column.
Specimen preparation is the hard part of TEM: metals are electropolished, ceramics ion-milled, and FIB lift-out cuts a foil from an exact site such as one transistor or one crack tip. Newer methods use the beam in new ways. In 4D-STEM a fast camera records a diffraction pattern at every probe position, mapping strain, orientation and fields, and electron ptychography computes the specimen from those patterns: in 2021 a group at Cornell University reported resolution of a few tens of picometres in a thick crystal, limited by the atoms' own thermal vibration.
Scanning probe microscopy
In 1981 Gerd Binnig and Heinrich Rohrer at IBM's laboratory near Zurich found a way to see atoms with no lens and no wave: feel them. The scanning tunnelling microscope (STM) brings a sharp metal tip within about a nanometre of a conducting surface and applies a small voltage. Classically no current can flow across the vacuum gap, but electrons tunnel through it quantum-mechanically, and the tunnelling current falls off exponentially with the gap :
where is the height of the barrier, close to the work function of the metals. A feedback loop moves the tip up and down with a piezoelectric scanner to keep the current constant as it scans, and the tip's height traces the surface. The atomic force microscope (AFM), invented by Binnig, Calvin Quate and Christoph Gerber in 1986, replaces the current by a force: the tip sits on a tiny cantilever spring whose deflection, read by a laser bounced off its back, measures the force between tip and surface. The AFM needs no conducting sample and no vacuum, so it maps polymers, ceramics, biological samples and surfaces in liquid.
import math
hbar, m_e, eV = 1.054571817e-34, 9.1093837e-31, 1.602176634e-19
# STM: the tunnelling current falls as exp(-2 kappa d), with kappa = sqrt(2 m phi) / hbar
for phi in (4.0, 4.5, 5.0): # barrier height, close to the work function, eV
kappa = math.sqrt(2 * m_e * phi * eV) / hbar
print(f"phi = {phi} eV: kappa = {kappa * 1e-9:5.2f} per nm; "
f"current falls x{math.exp(2 * kappa * 0.1e-9):4.1f} for each extra 0.1 nm of gap")
kappa = math.sqrt(2 * m_e * 4.5 * eV) / hbar
print("\nextra gap (pm) current, if it was 1 nA")
for dz in (0, 10, 20, 50, 100, 200):
print(f"{dz:14d} {math.exp(-2 * kappa * dz * 1e-12):8.4f} nA")
# AFM: Hooke's law for the cantilever, and its thermal jiggle sqrt(kT / k)
kT = 1.380649e-23 * 295
print("\nspring constant force for 0.1 nm deflection thermal rms motion")
for k in (0.1, 2.0, 40.0, 1800.0): # N/m: soft contact lever ... stiff quartz sensor
print(f"{k:10.1f} N/m {k * 0.1e-9 * 1e9:18.3f} nN {math.sqrt(kT / k) * 1e12:14.1f} pm")For a barrier of 4.5 eV, is 10.9 nm⁻¹, and the current falls by a factor of 8.8 for every extra 0.1 nm of gap; 10 pm, a twentieth of an atom's diameter, changes it by 20 per cent. This steepness is the whole secret of the STM. Nearly all the current flows through the single atom at the very end of the tip, so a tip that is blunt on the scale of the eye is atomically sharp where it matters, and height changes of a few picometres are measured easily.
The STM maps electron states near the Fermi level, not atoms as balls, so its images depend on the voltage, and atomic resolution normally needs ultra-high vacuum and clean surfaces. In 2009 Leo Gross and colleagues at IBM Zurich used a non-contact AFM with a carbon monoxide molecule on its tip to resolve the bonds inside a single pentacene molecule. In materials laboratories the AFM is an everyday tool for roughness, the thickness of two-dimensional materials such as graphene (two-dimensional and quantum materials), polymer morphology and, with conducting or magnetic tips, local currents and domains.
Measuring grain size from an image
A micrograph becomes data when you measure it. The most common measurement is grain size, because it controls strength through the Hall–Petch relation and enters every steel specification. The standard intercept method (ASTM E112, ISO 643) lays test lines across the image and counts the number of grain boundaries they cross. The mean linear intercept is the total length of test line divided by the number of crossings, and ASTM converts it into a grain size number
defined so that each increase of by one doubles the number of grains per unit area; a mean intercept of 20 µm is . The intercept is not the grain "diameter": a random line cuts most grains off-centre, so is shorter than the typical width of a grain.
The block below builds a synthetic microstructure whose answer is known, makes from it what an etched micrograph would look like, and measures it the way an automated image-analysis system would. The grains are a Voronoi tessellation: 60 random seed points in a 200 µm square, each pixel belonging to the nearest seed, as if all grains had nucleated at once and grown at the same speed until they met. For random seeds, the theory of random tessellations predicts , where is the mean grain area.
import math
import numpy as np
rng = np.random.default_rng(1863)
# 1. A synthetic polycrystal: 60 grains grown from random seeds (a Voronoi tessellation),
# made periodic so that a grain leaving one edge comes back at the opposite one.
n_px, px_um, n_grains = 400, 0.5, 60 # 400 x 400 pixels of 0.5 um: a 200 um field
L = n_px * px_um
seeds = rng.uniform(0, L, size=(n_grains, 2))
y, x = (np.mgrid[0:n_px, 0:n_px] + 0.5) * px_um
dx = np.abs(x[..., None] - seeds[:, 0])
dy = np.abs(y[..., None] - seeds[:, 1])
dx, dy = np.minimum(dx, L - dx), np.minimum(dy, L - dy)
labels = np.argmin(dx**2 + dy**2, axis=-1) # the grain each pixel belongs to
# 2. What an etched micrograph would show: grains in different greys (the etch attacks each
# orientation differently), dark grooves two pixels wide at the boundaries, and noise.
groove = np.zeros(labels.shape, dtype=bool)
for axis in (0, 1):
for shift in (1, -1):
groove |= labels != np.roll(labels, shift, axis=axis)
grey = rng.uniform(160, 220, size=n_grains)
image = np.where(groove, 60.0, grey[labels]) + rng.normal(0, 20, size=labels.shape)
# 3. The intercept method: lay test lines (every row and every column) across the image and
# count how often they cross a boundary. Mean intercept = total line length / crossings.
def box_blur(a):
return sum(np.roll(np.roll(a, i, 0), j, 1) for i in (-1, 0, 1) for j in (-1, 0, 1)) / 9
def crossings(mask):
"""Number of dark runs met along every row and every column (periodic image)."""
return int(np.sum(mask & ~np.roll(mask, 1, axis=1)) + np.sum(mask & ~np.roll(mask, 1, axis=0)))
line_um = 2 * n_px * L
truth = int(np.sum(labels != np.roll(labels, 1, axis=1)) + np.sum(labels != np.roll(labels, 1, axis=0)))
def report(name, hits):
l_bar = line_um / hits # mean linear intercept, um
G = -6.6439 * math.log10(l_bar / 1000) - 3.288 # ASTM E112 grain size number (l_bar in mm)
print(f"{name:32s} {hits:6d} {l_bar:6.2f} um G = {G:4.1f}")
print("segmentation crossings intercept")
report("ground truth (grain labels)", truth)
for t in (100, 120, 130):
report(f"raw image, darker than {t}", crossings(image < t))
for t in (110, 130, 150):
report(f"smoothed, darker than {t}", crossings(box_blur(image) < t))
print(f"theory for random seeds: (pi/4) sqrt(mean area) = {math.pi / 4 * math.sqrt(L**2 / n_grains):.2f} um")
# 4. A corner of the segmented image, 120 x 80 um, one character per 2 x 4 um.
seg = box_blur(image) < 130
print()
for r in range(0, 160, 8):
print("".join("#" if seg[r:r + 8, c:c + 4].mean() > 0.2 else " " for c in range(0, 240, 4)))The ground truth, counted from the grain labels, is a mean intercept of 20.01 µm, , close to the 20.28 µm the theory predicts. The "micrograph" measurements scatter around it. Thresholding the noisy image at 100 gets 20.23 µm, but at 130 noise speckles inside grains count as boundaries and the intercept drops to 14.78 µm (). Smoothing first removes the speckle, but too low a threshold then loses the fainter boundaries, those running diagonally across the pixel grid, giving 26.38 µm (). Smoothing and thresholding at 130 lands at 20.90 µm, 4 per cent long. The boundary map looks convincing whichever settings were used, which is exactly the danger.
The same tools measure phase fractions (by a theorem of stereology, the area fraction of a phase on a random section equals its volume fraction), particle sizes and pearlite spacings. In 2026 much segmentation is done by trained neural networks, which handle difficult contrast far better than a threshold, and the lesson of the block applies with more force: a network trained on one etch, magnification and microscope can fail silently on another.
Microscopy in research and industry
Correlative and automated work. Investigations combine instruments: a light micrograph to find a crack, an SEM to see its surface, EBSD to see which grains it followed, a FIB foil from its tip for the TEM. Automated SEMs with X-ray detectors scan thousands of particles unattended, to measure the minerals in a crushed ore, classify the inclusions in a steel or search for gunshot residue.
Three dimensions and time. Serial sectioning with a FIB, and X-ray tomography, reconstruct microstructures in three dimensions. In situ holders heat, strain, bias or flood the specimen with gas or liquid inside the TEM, so that reactions, dislocation motion and battery cycling can be watched directly. Cryogenic electron microscopy, developed for biology (Nobel Prize in Chemistry 2017), has since 2017 been applied to reactive battery materials such as lithium metal, which degrade under a room-temperature beam.
Data and access. A single 4D-STEM data set can run to hundreds of gigabytes, so analysis has moved into Python and open software: HyperSpy for spectrum images, MTEX for orientation maps, Fiji and scikit-image for images. Shared national facilities give researchers access to instruments no single group could afford, and training such as Microscopy Australia's MyScope is free to all.
The instrument makers are few: electron microscopes are built mainly in Japan, Germany, the Czech Republic, the Netherlands and the United States, and increasingly in China. Advanced instruments cost millions of euros and need quiet rooms shielded from vibration, temperature drift and stray magnetic fields.
Common misconceptions
Misconception
"More magnification shows more detail." Detail is limited by resolution, set by wavelength and aperture (or, in electron microscopes, by aberrations). Magnifying past about a thousand times the numerical aperture in a light microscope only enlarges the blur.
Misconception
"An SEM image is a photograph." It is a map of how many electrons came out at each point. Secondary-electron images look lit from one side because of the geometry of the detector and the edge effect, and colour in an SEM image is always added afterwards. Bright does not mean "higher" in a backscattered image; it means heavier.
Misconception
"Electron microscopes are limited by the electron wavelength." At 200 kV the wavelength is 2.5 pm, a hundred times smaller than the best resolution. The limits are lens aberrations, instrument stability, the specimen's thickness and its tolerance of the beam.
Misconception
"The bright dots in an atomic-resolution image are atoms." In a high-resolution TEM image they may be atoms or the spaces between them, depending on focus and thickness; matching to image simulations is needed. In HAADF-STEM they are columns of atoms seen end-on, and in an STM image they are electronic states at a particular voltage.
Exercises
Use J s, kg, C and m/s where needed.
1. Microscopy
Warm-upHumanAn oil-immersion objective has a numerical aperture of 1.25 and is used with light of wavelength 500 nm. What is its Abbe resolution limit, in nanometres?
2. Microscopy
CoreHumanA flat, polished, unetched cross-section of a lead-free solder joint contains a tin-rich matrix, copper–tin intermetallic particles and small silver–tin particles. In the SEM, which signal shows these three phases apart most directly in a single image?
3. Microscopy
CoreComputationalWrite electron_wavelength_pm(kilovolts) returning the de Broglie wavelength, in picometres, of an electron accelerated through the given voltage in kilovolts. Include the relativistic correction: with kinetic energy ,
The tests check 1, 100 and 300 kV.
4. Microscopy
CoreHumanIn a scanning tunnelling microscope the tunnelling current varies as , with for this tip and surface. The feedback loop is switched off and the tip is pulled back by 0.15 nm. By what factor does the current fall? Give the nearest whole number.
5. Microscopy
CoreHumanA micrograph is taken at 100× magnification. Five test lines, each 100 mm long on the micrograph, cross a total of 250 grain boundaries. Using the ASTM relation , with the mean linear intercept in millimetres at the specimen, what is the grain size number , to the nearest whole number?
6. Microscopy
CoreHumanA thin foil of a precipitation-hardened aluminium alloy is examined in a conventional (uncorrected) 200 kV TEM in bright field. Which of these can it show that a light microscope looking at a polished and etched section of the same alloy cannot? Select all that apply.
7. Microscopy
Warm-upHumanIn scanning transmission electron microscopy, an annular detector that collects electrons scattered to large angles gives images whose brightness rises roughly as to . What is this imaging mode called? (The usual abbreviation is enough.)
8. Microscopy
StretchComputationalAutomate the intercept method. A segmented micrograph is a 2-D NumPy boolean array mask, True where a pixel is grain boundary. Write:
count_crossings(mask): the number of runs ofTrue(maximal stretches of consecutiveTruepixels) met along every row plus along every column. The image is not periodic: a run touching the edge counts once.mean_intercept(mask, pixel_size): the total length of all rows and all columns,pixel_size, divided bycount_crossings(mask).
For example, the row [0, 1, 1, 0, 1] contains two runs.
Further reading and resources
- Beiträge zur Theorie des Mikroskops und der mikroskopischen Wahrnehmung · Ernst Abbe · 1873Archiv für mikroskopische Anatomie. Abbe's theory of image formation, with the resolution limit set by wavelength and aperture.
- Surface studies by scanning tunneling microscopy · G. Binnig, H. Rohrer, Ch. Gerber and E. Weibel · 1982Physical Review Letters. The first account of the scanning tunnelling microscope, from IBM's Zurich laboratory.
- Atomic force microscope · G. Binnig, C. F. Quate and Ch. Gerber · 1986Physical Review Letters. The invention of the atomic force microscope.
- Nobel Prize in Physics 1986: Ernst Ruska, Gerd Binnig and Heinrich Rohrer · Nobel Prize OutreachFor the design of the first electron microscope, and for the scanning tunnelling microscope.
- Electron microscopy image enhanced · M. Haider, S. Uhlemann, E. Schwan, H. Rose, B. Kabius and K. Urban · 1998Nature. The first working spherical-aberration corrector for a transmission electron microscope.
- Electron ptychography achieves atomic-resolution limits set by lattice vibrations · Z. Chen, Y. Jiang, Y.-T. Shao, M. E. Holtz, M. Odstrčil, M. Guizar-Sicairos, I. Hanke, S. Ganschow, D. G. Schlom and D. A. Muller · 2021Science. Resolution of a few tens of picometres in a thick crystal, limited by the atoms' own thermal motion.
- MyScope · Microscopy AustraliaFree online training in light, electron and probe microscopy and microanalysis, with simulators of the instruments.
- DoITPoMS · University of CambridgeFree teaching packages, including micrograph libraries and packages on electron microscopy and specimen preparation.
- ASTM InternationalPublishes E112, the standard test methods for determining average grain size, and E407, the standard practice for microetching metals and alloys.
- Fiji (ImageJ)Open-source image analysis widely used for micrographs.
- scikit-imageOpen-source image processing in Python, used for segmentation and measurement of microstructures.