Impurity profiling is the measurement of dopant or contaminant concentration as a function of depth beneath a wafer surface, distinguishing it from bulk or surface-only analysis. A fab needs to know not just how much boron, phosphorus, or arsenic sits in a layer, but exactly where the atoms sit relative to a junction, a gate stack, or an epitaxial interface. Chemical concentration measured by secondary ion mass spectrometry, electrically active carrier concentration measured by capacitance-voltage or spreading resistance, and sheet-level averages from a four-point probe each answer a different question, and confusing them produces bad process decisions. The working set covered here spans SIMS depth profiling, spreading-resistance profiling, capacitance-voltage and electrochemical CV profiling, Hall-effect and van der Pauw carrier measurement, and four-point-probe sheet resistance, tied together by junction-depth extraction, dose recovery, and the gap between chemical dose and activated carrier dose.

Calibrate the depth scale before trusting any single concentration number.

SIMS depth profiling erodes the surface with a primary ion beam while a mass spectrometer records secondary ion counts as a function of sputter time, and those counts are meaningless until sputter time is converted to a physical depth scale. Crater depth is measured afterward with a stylus profilometer or an interferometric tool, typically resolving steps down to 2 nm, and the assumed sputter rate is checked against that final crater floor rather than trusted blindly. A rate that drifts from 0.42 nm/s at the surface to a different steady-state value through a heterostructure will smear a sharp interface if a single linear scale is applied across the whole run. Near-surface transients and knock-on mixing degrade depth resolution from roughly 2 nm at the surface to 15 nm to 20 nm deeper, setting a hard floor on how sharp a reported junction can look.

Quantification depends on relative sensitivity factors that convert raw secondary ion intensity into atomic concentration, and those factors are matrix specific rather than universal. A boron RSF calibrated in silicon does not transfer cleanly to a silicon-germanium alloy or a heavily oxidized surface layer, because ion yield depends on the local chemical environment as much as on the dopant itself. Implant standards with a known dose, ideally traceable through NIST reference methods, anchor the RSF to roughly ±20% accuracy under good conditions, and that uncertainty band should travel with every reported concentration.

Spreading resistance profiling trades sample destruction for a continuous carrier-concentration record.

SRP bevels the sample at a shallow angle, often below 1°, to expose a long shallow ramp through the doped layer, then steps two closely spaced metal probes along that ramp while recording local resistance at each point. Converting spreading resistance to resistivity and then to carrier concentration relies on correction factors and calibrated reference curves, and probe spacing on the order of 500 µm to 1 mm along with contact force repeatability directly limits depth resolution, which typically runs from 5 nm near the surface to several tens of nm deeper into the bevel. SRP responds to free carriers rather than total dopant atoms, so it reports an electrically active profile that already reflects incomplete activation or deactivation near a heavily implanted surface.

Capacitance-voltage profiling extracts a carrier concentration profile from a MOS or Schottky diode by measuring capacitance while stepping a DC bias, superimposed with a small AC signal near 1 MHz, converting depletion-width change with bias into a depth-dependent carrier density under the depletion approximation. A mercury probe or deposited Schottky contact avoids full device processing, and a Keithley source-measure instrument or an LCR bridge supplies the bias and reads small-signal capacitance with a resolution around 150 mV of usable bias step. Electrochemical CV, ECV, extends the same physics beyond a single depletion region by alternating a brief electrolytic etch, controlled to remove single-digit nm increments, with a capacitance measurement at the freshly exposed surface, stitching together a much deeper profile than a static CV diode allows. Semilab and other vendors package ECV as a turnkey tool, and a corona-Kelvin noncontact charge step is sometimes used ahead of a CV measurement to probe surface potential without a deposited contact.

Hall-effect and van der Pauw measurements report activated carriers, not total dopant atoms.

A van der Pauw structure with four contacts near the sample perimeter, measured under an applied magnetic field, yields sheet carrier density and Hall mobility from a single small die, and repeating that measurement after successive thin-layer removal by anodic oxidation or a controlled etch builds a carrier profile analogous to SRP but with a direct mobility readout at every step. The technique measures only carriers free to conduct, so a profile with a nominal chemical concentration of several times 1e20 cm-3 near the surface can show a much lower Hall-derived carrier concentration if a large fraction of the implanted dose sits on interstitial rather than substitutional lattice sites after an incomplete anneal.

Four-point-probe sheet resistance measurement forces a known current through the two outer probes of a linear array and reads voltage across the two inner probes, with a standard 1 mm probe spacing and a geometric correction factor applied for finite sample size. A Keysight or Keithley source-measure unit supplies sub-mV voltage resolution at low currents, and the resulting sheet resistance, often expressed in ohm per square with typical implant layers running from under 50 ohm per square to several kilo-ohm per square, folds carrier concentration and mobility into a single number that cannot separate the two without an independent Hall measurement.

Depth resolution and dynamic range pull profiling techniques in opposite directions.

SIMS delivers the sharpest depth resolution near the surface but its dynamic range collapses where the dopant signal approaches instrumental background, typically in the 1e16 cm-3 to 1e17 cm-3 range depending on species and matrix, and knock-on mixing broadens any buried interface regardless of primary beam energy. SRP and CV-based methods trade some of that near-surface sharpness for a much wider usable dynamic range because they measure a bulk electrical response rather than counting individual sputtered ions, so a profile needing both a sharp near-surface gradient and an accurate deep tail usually needs two techniques stitched together. XPS and AFM contribute complementary near-surface information, XPS through chemical-state sensitivity and AFM through direct topography of a bevel or crater floor, while ellipsometry tracks oxide or cap-layer thickness that would otherwise bias a depth-scale calibration, and DLTS separately reveals deep-level traps that can mimic a deactivated dopant signature.

Define what the profile must answer: chemical dose, activation fraction, or junction depth
  -> select SIMS for chemical concentration with nm-scale near-surface resolution
  -> select SRP, CV, or ECV for electrically active carrier concentration
  -> select Hall / van der Pauw plus layer removal for mobility-resolved carrier depth data
  -> select four-point probe for a fast sheet-resistance sanity check across the wafer
  -> calibrate depth scale against measured crater or bevel geometry, not assumed rate alone
  -> calibrate RSF or resistivity-to-concentration curves against traceable reference standards
  -> compare chemical and electrical profiles at the same nominal depth
  -> profiles agree within calibrated uncertainty?
  -> no: suspect incomplete activation, matrix effect, or depth-scale mismatch and re-measure
  -> yes: extract junction depth, dose, and activation fraction with quantified confidence

Dose and junction depth extracted from a profile are only as trustworthy as the calibration chain behind them.

Integrating a SIMS concentration profile over depth recovers an implant dose that should be checked against the dose set on the implanter, and a mismatch beyond roughly a few percent is a signal to question RSF calibration, sputter-rate drift, or channeling rather than the implant tool. Junction depth can be read two different ways from the same wafer: a chemical junction depth where the SIMS profile crosses a reference background concentration near 1e17 cm-3 to 1e18 cm-3, and an electrical junction depth where SRP-derived carrier type crosses zero net doping, and these two depths coincide only when activation is complete and uniform. A junction reported near 180 nm from SIMS with an electrical junction depth measured 20 nm to 30 nm shallower by SRP is direct evidence of a deactivated or amorphized region that a chemical count alone would never reveal.

Reference standards anchor every number in this workflow. NIST-traceable implant standards calibrate SIMS RSFs, and certified resistivity standards calibrate SRP and four-point-probe conversions. A single reading at 25 °C with a fresh standard is not sufficient; bracket the measurement campaign with repeat verification and flag any run where the standard reads outside roughly ±5% of its certified value.

Chemical concentration and activated carrier concentration answer different questions, and dopant-engineering decisions need both.

A post-implant anneal at temperatures commonly reaching 1000 °C to 1050 °C activates dopants by moving them onto substitutional lattice sites and repairing lattice damage, and the activation fraction, the ratio of electrically active carrier concentration to total chemical concentration, is rarely 100% even after a well-tuned anneal. A SIMS profile alone cannot report that fraction because it counts every dopant atom regardless of lattice site; a Hall or CV profile alone cannot report the total implanted dose because it only sees carriers that conduct. Running both on companion samples from the same wafer or lot under matched conditions is the only way to separate an implant-dose problem from an activation-anneal problem when a device's electrical results miss target.

Cross-validate every profile against an independent measurement before it drives a process change.

A SIMS profile that looks sharp and a CV profile that looks smooth are not in conflict once the physical reasons for depletion-region rounding are understood, and a four-point-probe sheet-resistance reading that falls outside its historical band is a fast, cheap trigger to pull a full depth profile rather than a standalone verdict. Combining a chemical profile, an electrical carrier profile, and a sheet-resistance check gives three independent views of the same doped layer, and agreement across all three, within stated calibration uncertainty, is what should authorize a process change rather than any single number in isolation. Viewed through a dopant-engineering-control lens, impurity profiling is less about any one instrument's precision and more about reconciling chemical dose, activated carrier concentration, and junction geometry into one internally consistent picture before that picture is allowed to steer an implant, anneal, or epitaxial recipe.