Voltage-pulsed vs. laser-pulsed APT
Time-of-flight only works if you know when the clock started. The standing voltage is held just below the evaporation threshold, and a short pulse tips atoms over the edge at a known instant. The choice is what to pulse: the field, or the temperature. That one decision sets which materials you can run, and what your mass spectrum looks like.
A voltage pulse adds 15–25 % to the field for about a nanosecond. It is clean (evaporation happens at one field and one cold temperature), but the pulse has to travel down the specimen like a transmission line, so the specimen must conduct. It also flexes the tip mechanically on every pulse, a fatigue test at hundreds of kilohertz that brittle materials fail.
A laser pulse instead heats the last few tens of nanometres of the apex by a few hundred kelvin for under a nanosecond. No current needs to flow, so semiconductors, oxides, and even geological minerals evaporate happily: this is the single change, arriving with commercial ultrafast lasers in the mid-2000s, that turned APT from a metals technique into a general materials technique. The price: the tip stays warm for a moment after each pulse, so some atoms leave late, and every peak in the spectrum grows a thermal tail.
Voltage pulsing gives the sharpest mass spectra but only works on conductors that can survive being flexed ten billion times; laser pulsing runs almost anything, at the cost of thermal tails (Part 2 shows what those do to peak ranging) and a sensitivity to exactly where and how hard the laser strikes.
Go deeper: the physics of the trigger
Field evaporation is thermally activated escape over a field-lowered barrier. The rate per atom is:
ν = ν₀ · exp( −Q(F) / kBT ) , Q(F) ≈ Q₀ (1 − F/Fev) ν₀ ≈ 10¹¹–10¹³ s⁻¹ (an attempt frequency); Q₀ is of order an eV; Fev is the zero-barrier evaporation field, tens of V/nm. Near threshold the exponent is enormous, so small changes in F or T swing the rate by orders of magnitude, which is exactly what makes a nanosecond trigger possible.The two pulsing modes are the two knobs in that exponent. A voltage pulse raises F;
within a fraction of a percent of threshold, a 20 % field excursion turns evaporation from
negligible to certain. A laser pulse raises T from ~50 K to several hundred kelvin;
the barrier is unchanged but the Boltzmann factor explodes. Both drop the effective barrier crossing
into a ~1 ns window; the difference is what happens at the edges of that window.
Why the cold base temperature? Between pulses the standing field must produce essentially zero evaporation (that is the DC background in the spectrum), and surface atoms must not migrate: at higher base temperatures atoms roll along the surface before evaporating and lateral resolution degrades. 20–80 K is the compromise.
Why the thermal tail? The apex cools by conduction down the shank, with a time constant of nanoseconds. Atoms that evaporate while the tip is still warm start their flight late; since m/z is computed from the total flight time, they land on the high-mass side of their peak. The tail length grows with laser energy and with anything that slows cooling: long thin tips, poor thermal conductors, oxides. Reflectron instruments partially compensate by making late, slower ions take a shorter path.
Why laser opened semiconductors and oxides. A voltage pulse needs the specimen to carry a fast current pulse to the apex: resistive specimens smear the pulse (the clock start becomes fuzzy → mass resolution collapses) and piezoelectric/brittle ones crack under the cyclic Maxwell stress, which scales as F². Photons need no conductivity: absorption at the apex does the work. Kellogg & Tsong demonstrated pulsed-laser atom probe in 1980; commercial UV (355 nm) picosecond systems made it routine on LEAP-class instruments from ~2005–2013 onward.
FIB lift-out & needle sharpening
The specimen is the instrument. Everything in Parts 1 and 2 assumed a smooth needle of ~50 nm end radius with the feature you care about in the first hundred nanometres of the apex. The focused ion beam is how you put a specific grain boundary, not a random one, exactly there.
Electropolishing, the classical route, turns a wire of a conductive alloy into a beautiful needle but gives you no say in what ends up at the apex. The FIB lift-out route inverts that: choose the feature first in the SEM image, then carve the specimen around it. The price is gallium damage and a full shift at the microscope; the reward is site-specific APT on grain boundaries, interfaces, devices, and anything else you can see.
The annular mill: where the needle happens
Mounted on its post, the blank is still a micron-scale block. The atom probe needs a sub-100 nm apex. The bridge between those scales is annular milling: the ion beam scans a ring, everything inside the inner radius is shadowed and survives, everything outside is removed. Shrink the ring, mill again, repeat: the block turns into a cone, and the cone into a needle.
Go deeper: the details that decide whether the run survives
- Protect before you look. The first thing the ion beam touches is destroyed, so the first step is always a protective cap (electron-beam Pt or Ni first, then ion-beam Pt) over the region of interest. In the finished needle that cap is your depth marker: when Pt stops arriving in the mass spectrum, the real specimen is starting.
- Put the feature 20–100 nm below the apex, not at it. The first tens of nanometres of any FIB tip are Ga-implanted and milling-damaged; runs also frequently fracture in the first seconds. Leave sacrificial material above the feature.
- Finish gentle. 30 kV Ga implants to ~20 nm in most metals. A final polish at 2–5 kV shrinks the damaged skin to 2–5 nm, and most of that evaporates before the region of interest. Skipping the low-kV step is the classic source of mystery Ga peaks and amorphised apexes.
- Watch the taper. A full shank angle of ~10–20° is the sweet spot: too blunt and the voltage needed exceeds the supply before the feature is reached; too sharp and the tip is fragile and the field of view runs off the sides. The reconstruction also needs the shank angle as an input: measure it in the SEM before the run (Part 1).
- One lamella, many tips. A single lift-out bar yields 4–8 mounted blanks. Atom probe runs end early and often: redundancy is not optional. A realistic overall yield for a difficult interface specimen is well under 50 %, and everyone's first dozen needles are donations to the vacuum system.
- Cryo and Xe variants. Ga embrittles some materials (Al grain boundaries, notoriously); Xe plasma FIB removes the Ga problem entirely and mills faster. Cryo-FIB extends lift-out to frozen liquids, batteries mid-cycle, and soft matter.
Grain-boundary segregation
A grain boundary is two atomic layers of disorder that decide whether a polycrystal is tough or glass-brittle. Solute atoms trade strain energy for a comfortable boundary site and pile up there: a monolayer-scale chemical event, invisible to almost every probe. Measuring it atom-by-atom is the job APT was born for.
The thermodynamics is a competition. Each solute atom lowers the system energy by the segregation
energy ΔGseg when it sits in a boundary site, but boundary sites are few, and
entropy prefers atoms spread through the bulk. McLean wrote the balance down in 1957, and it is worth
seeing it play out live.
Boundary coverage against annealing time
Play with the two ends of the temperature slider. Cold, strong segregation: the boundary saturates toward a full monolayer and the enrichment factor reaches the hundreds: this is phosphorus in steel, the classic temper-embrittlement villain, or boron in Ni₃Al doing the opposite and holding the boundary together. Hot: entropy wins, the boundary empties, and quenching from too high a temperature freezes in a clean boundary that segregates later during service. The physics is one exponential.
Auger spectroscopy sees boundary chemistry only after fracturing the specimen along it: so only boundaries weak enough to break, with monolayer depth ambiguity. STEM-EDS/EELS sees a projected 1–2 nm column and struggles below a percent. APT gives the absolute count: atoms per square nanometre, element by element, boundary by boundary, with the boundary's position and both grains in the same dataset (measured as Γ, Part 2). Numbers like "0.4 monolayers of B" come from APT and essentially nowhere else.
Go deeper: the McLean isotherm and its fine print
The exponential is the whole story. ΔGseg = 25 kJ/mol at 700 K gives a factor of ~74; at 500 K the same energy gives ~420. Small energies, huge enrichments, which is why trace impurities at parts-per-million bulk levels can still coat boundaries.
The fine print matters in real alloys. Sites are not identical: ΔG varies with boundary character, so a Σ3 twin may stay clean while a random high-angle boundary loads up; APT on multiple boundaries in one specimen shows exactly this spread. Solutes compete for the same sites (site competition: B displacing S in steels). And kinetics can be the real limiter: reaching equilibrium requires diffusion over the collection distance, so short anneals undershoot the isotherm: watch the curve in the sandbox take longer to arrive when the temperature drops.
Reporting convention: quote Γ (atoms/nm², from the ladder diagram or proxigram integral) rather than a peak concentration, for all the reasons in Part 2: Γ survives the blurring that peak values do not.
Correlative APT + TEM
TEM tells you where you are and what phase you are in; APT tells you which atoms are there, peak overlaps permitting. Neither statement is reliable from the other instrument. Running both on the same needle is not a luxury; it is how the two biggest error sources in each technique get cancelled.
| Property | TEM / STEM | APT |
|---|---|---|
| Geometry | 2-D projection of a foil | 3-D atom-by-atom reconstruction (model-dependent) |
| Field of view | nm → tens of µm, zoom at will | ~100 nm × 100 nm × 100s nm, once |
| Crystallography | Full: diffraction, orientation, defects | Weak: faint lattice planes near poles |
| Composition | ~0.5–1 at.% floor (EDS/EELS), through-thickness average | Ion-by-ion counting; ppm-class trace sensitivity in favourable cases |
| Light elements | Hard (H invisible, B/C/N marginal) | Strong: B, C, N routine; H detectable but attribution needs care |
| Specimen fate | Survives: image it again tomorrow | Destroyed: the specimen is the data |
| Systematic weakness | Projection overlap; thickness unknown | Reconstruction distortions (Part 1) |
The workflow that uses this properly runs the TEM first, on the finished needle, before it goes anywhere near the atom probe:
- Image the needle. True apex radius and shank angle, the two numbers the reconstruction otherwise has to guess (Part 1), measured directly, per specimen.
- Locate the feature. Confirm the boundary or interface really is 40 nm below the apex, at what inclination, before committing the tip. No more running blind and hoping.
- Identify the phases. Diffraction names the phase; APT then quantifies it. The combination (structure from one instrument, absolute chemistry from the other) is what neither can claim alone.
- Run the atom probe. The needle is consumed; the TEM images are its only surviving portrait, and the reconstruction is calibrated against them rather than against defaults.
The TEM step is not free: the electron beam deposits carbon, can amorphise a delicate apex, and every transfer is a chance to break the tip. Low dose, short sessions, and a vacuum transfer holder for reactive materials: hydride formation in Ti and Zr needles during air transfer is a documented artefact that correlative workflows exist to avoid, not to cause.
Sources & further reading
- G. L. Kellogg, T. T. Tsong, “Pulsed-laser atom-probe field-ion microscopy,” J. Appl. Phys. 51, 1184 (1980): abstract.
- K. Thompson et al., “In situ site-specific specimen preparation for atom probe tomography,” Ultramicroscopy 107, 131 (2007): the standard FIB lift-out protocol.
- T. J. Prosa, D. J. Larson, “Modern focused-ion-beam-based site-specific specimen preparation for atom probe tomography,” Microsc. Microanal. 23, 194 (2017): the current review of lift-out and sharpening practice, including correlative workflows.
- D. J. Larson, T. J. Prosa, R. M. Ulfig, B. P. Geiser, T. F. Kelly, Local Electrode Atom Probe Tomography: A User’s Guide, Springer (2013): instrument geometry, pulsing modes and run conditions.
- D. McLean, Grain Boundaries in Metals, Oxford (1957): the segregation isotherm.
- P. Lejček, Grain Boundary Segregation in Metals, Springer (2010).
- M. Herbig et al., “Atomic-scale quantification of grain boundary segregation in nanocrystalline material,” Phys. Rev. Lett. 112, 126103 (2014): correlative TEM + APT on individual boundaries.
- B. Gault, M. P. Moody, J. M. Cairney, S. P. Ringer, Atom Probe Microscopy, Springer (2012).