Porosimetry determines how much void space a material contains, which pores are accessible, how filling and emptying proceed, and what pore-size or connectivity model is consistent with those observations. In semiconductor thin films, the small material volume and rigid substrate make conventional bulk adsorption difficult, so ellipsometric porosimetry is especially useful for porous low-k dielectrics, membranes, sensor films, and nanoporous coatings. It exposes the film to a controlled probe vapor and follows the optical response as relative pressure rises and falls. The instrument measures polarization change; porosity and pore size emerge only through an adsorption, dielectric-mixture, and pore-geometry model.

Porosity, accessible porosity, pore size, and connectivity are different measurands. Total porosity is the void-volume fraction relative to the film volume. Ellipsometric porosimetry primarily senses pores reached and filled by the chosen adsorptive under the measurement conditions; sealed pores may remain invisible. A constricted network can delay access to larger cavities, while a surface sealing layer can make an internally porous film appear nonporous to vapor. One porosity number therefore cannot describe pore topology or process damage by itself.

The dry film is first modeled by spectroscopic ellipsometry to establish thickness and effective dielectric response. During a vapor-pressure program, adsorption on internal surfaces and capillary filling replace pore vapor with condensed adsorbate, increasing optical polarizability. Repeating the fit at each pressure yields an adsorbate-volume trajectory or an equivalent optical-density trajectory. Desorption reveals how the network empties and may produce hysteresis.

The optical conversion requires a physically defensible effective-medium model. A common approach treats the porous film as a mixture of solid skeleton, pore vapor, and condensed adsorbate. For an isotropic Bruggeman mixture,

The volume fractions (f_i) are inferred from the measured effective dielectric function using assigned constituent dielectric functions. Bruggeman symmetry is an approximation, not a universal pore law. Anisotropic pores, connected channels, interfacial layers, density gradients, chemical interaction, and confinement-dependent adsorbate polarizability can violate it. Test alternative mixing rules and use multiple wavelengths or angles to expose thickness–index correlation.

The saturated uptake can estimate accessible pore volume when the pores are filled with a liquid-like adsorbate and the skeleton remains unchanged. If the film swells, densifies, dissolves, reacts, or changes surface chemistry during exposure, optical change is not equivalent to pore filling alone. Fit thickness and dielectric response together, examine whether the dry state returns after desorption, and report irreversible change separately.

Relative pressure connects a sorption event to a model-dependent pore radius. For capillary condensation in an idealized cylindrical pore, the Kelvin relation can be written

Here (p/p_0) is relative pressure, (\gamma) and (V_m) are liquid surface tension and molar volume, (\theta) is contact angle, (r_K) is Kelvin radius, and (t_{\rm ads}) represents the adsorbed layer. The derived pore radius depends on geometry, wetting, adsorbate properties, temperature, and thickness correction. At micropore dimensions, classical Kelvin assumptions become unreliable and density-functional or calibrated adsorption models may be more appropriate. A pore-size distribution is therefore conditional on the stated model.

Adsorptive selection determines which network the experiment can see. Choose a molecule compatible with the pore scale, surface energy, matrix chemistry, and intended process question. A polar probe may interact strongly with hydroxylated damage sites; a nonpolar probe may better represent hydrophobic pores but fail to wet another surface. Molecular size can exclude narrow necks. Vapor pressure must be accurately controlled at the sample temperature, and the saturation pressure must correspond to the actual adsorptive and temperature.

Degassing removes ambient water and residual solvents but can also alter fragile organics or collapse a weak network. Define evacuation temperature, duration, base pressure, and acceptance criterion. Establish equilibrium at each pressure step by an optical-rate or time criterion rather than a fixed dwell chosen without testing. Insufficient equilibration shifts filling pressure and broadens the apparent distribution. Record pressure at the sample, temperature stability, flow configuration, leak rate, and the complete pressure trajectory.

Adsorption and desorption branches contain different information. Hysteresis may reflect pore blocking, network effects, metastability, cavitation, or geometry—not simply two independent pore sizes. In an ink-bottle network, adsorption can be influenced by cavity filling while desorption can be controlled by narrower necks or cavitation. Report both raw branches and the model applied to each. Do not average them into one distribution without physical justification.

Open and closed porosity require complementary measurements. Ellipsometric porosimetry measures accessible uptake; X-ray reflectivity or density measurements can estimate total void fraction if skeleton density is known; positron annihilation methods can respond to closed nanovoids and connectivity; scattering reveals correlation lengths and ordered structures. The difference between total and accessible porosity can support a closed-pore or sealed-surface interpretation, but only after uncertainties and probe sensitivities are aligned.

For porous low-k dielectrics, process damage can change more than pore volume. Plasma exposure may remove hydrophobic groups, densify a surface layer, open previously closed pathways, enlarge connected damage regions, or increase water affinity. A larger uptake may indicate new accessibility or changed surface chemistry rather than newly created geometric void volume. Combine EP with FTIR, XPS, dielectric measurements, or depth-sensitive methods to separate chemical modification from topology.

Pore sealing is a particularly important ambiguity. A conformal or surface-localized coating may narrow pore necks, reduce accessible volume, or block vapor while leaving internal closed volume. Comparing multiple adsorptives of different size and polarity, varying exposure time, and using PALS or X-ray methods can distinguish reduced pore size from lost accessibility. In-situ EP during ALD can track this evolution, but the adsorptive test itself should not be assumed to reproduce precursor penetration.

Define total porosity, accessible volume, pore size, connectivity, or damage question
  -> Select EP and complementary methods based on film volume and closed-pore sensitivity
  -> Choose an adsorptive using molecular size, polarity, wetting, and matrix compatibility
  -> Establish dry-film thickness, dielectric model, and substrate response
  -> Degas with a validated temperature and verify a stable reversible baseline
  -> Step relative pressure through adsorption and desorption with equilibrium criteria
  -> Fit Ψ and Δ at each step while testing thickness change and effective-medium alternatives
  -> Convert uptake to accessible volume and apply a declared pore-filling model
  -> Inspect hysteresis, irreversibility, covariance, and pressure-temperature uncertainty
  -> Cross-check total porosity, chemistry, pore closure, and mechanical stability
  -> Archive raw spectra, pressure history, model, probe properties, and uncertainty

Mechanical response can be measured during pore filling but is not automatic. Capillary pressure can strain a supported porous film, and ellipsometry can detect thickness change while adsorption evolves. Converting strain into elastic modulus requires a pore-shape and boundary-condition model, known surface stress or capillary pressure, and separation of optical-density change from physical expansion. Substrate constraint, anisotropy, cracking, and irreversible swelling can invalidate a simple modulus calculation. Validate with nanoindentation, surface acoustic waves, wafer curvature, or another mechanical method.

Patterned structures complicate blanket-film assumptions. Trenches and lines generate diffraction and may have sidewall damage different from the field region. Scatterometric porosimetry combines a periodic-geometry optical model with vapor uptake to infer changes in patterned material, but critical dimensions, sidewall profiles, tensor response, and adsorbate filling can be correlated. Use independently measured geometry and compare blanket and patterned witnesses without assuming they experience identical plasma exposure.

Uncertainty must propagate through pressure, optics, mixing, and pore models. Pressure-transducer calibration, temperature gradients, saturation-pressure data, adsorptive purity, equilibrium tolerance, optical noise, film thickness, skeleton dielectric function, liquid dielectric function, mixing rule, contact angle, adsorbed-layer correction, and pore geometry all contribute. Repeat full cycles to quantify reproducibility and detect conditioning. Parameter covariance from the ellipsometric fit covers only part of this chain.

Inspect residual spectra at every pressure, not only the fitted uptake curve. A spectral residual that grows with pressure can reveal an invalid fixed skeleton response or swelling layer. Compare fits where thickness is fixed, free, or constrained by a mechanical model. Run blank-substrate and dense-film controls to measure vapor refractive-index effects, window adsorption, and chamber drift. Confirm that the dry optical state returns within uncertainty before declaring reversible physisorption.

Store raw Ψ and Δ spectra, wavelength and angle, pressure and temperature time histories, adsorptive identity and purity, saturation-pressure source, flow and equilibration criteria, degas recipe, chamber blank, film thickness, substrate and backside condition, effective-medium equation, constituent optical constants, pore model, contact-angle and adsorbed-layer assumptions, adsorption and desorption branches, residuals, covariance, exclusions, and software version. Preserve the isotherm before pore-size transformation so future models can be applied.

A defensible porosimetry result states exactly which void population was observed. Accessible porosity is not total porosity; filling pressure is not pore radius without a model; hysteresis is not a unique geometry label; and optical uptake is not necessarily pure condensation when the matrix swells or reacts. The strongest interpretation joins reversible sorption, a validated optical mixture, a suitable adsorption model, and a complementary method sensitive to the missing pore population.

The durable way to interpret porosimetry is through an accessible-versus-total-void-probe-chemistry-sorption-isotherm-effective-medium-capillary-model-hysteresis-connectivity-and-cross-validation lens.