Closed-book appointment exam · independently graded
Professor — Geology & Geochemistry. The candidate agent answered from its own knowledge, closed-book; a second, independent examiner agent graded it adversarially.
vaiu-sci-eaps-prof-geology appointment benchmark (2026-07-16)Closed-book oral exam. Candidate agent adopted the vaiu-sci-eaps-prof-geology persona (v1.0.0) and answered from its own knowledge; a second, independent examiner agent graded. Condensed candidate transcript below; grading report in vaiu-sci-eaps-prof-geology-2026-07-16.md.
AI-transparency disclosure. Opened as an AI agent with no field/lab credentials; closed-book; numbers are order-of-magnitude to be checked; the observed-vs-inferred line kept bright (a mineral / a measured ratio is observed; a temperature / an age / an environment is inferred, riding on a named model); no fabricated references.
Q1 — Igneous petrology. Bowen's reaction series (N.L. Bowen, experimental petrology): the discontinuous mafic branch olivine → pyroxene → amphibole → biotite (each a reaction; the silicate structure polymerizes from isolated tetrahedra → single chains → double chains → sheets) and the continuous plagioclase branch anorthite(Ca) → albite(Na), converging at low T into K-feldspar → muscovite → quartz. Observed = the experimental crystallization order; inferred = a cooling history (assuming temperature control at fixed bulk composition/P with crystals equilibrating). Fractional crystallization breaks that equilibrium — early crystals removed by settling/armoring/filter-pressing drive the residual liquid silica- and alkali-rich (a basalt→felsic differentiation path), evidenced by normal zoning, cumulate textures, and a liquid-line-of-descent (with the caveat that assimilation and magma mixing overlap chemically). Binary eutectic (e.g. diopside–anorthite, fixed P): two liquidus curves meet at the eutectic; on cooling the first phase crystallizes and the liquid slides down its liquidus to the eutectic, where T stalls while both phases co-crystallize; on heating the first melt is always eutectic composition regardless of bulk (why partial melting yields offset melts — basalt from peridotite). A solid-solution loop (albite–anorthite) reaches albite-rich residual liquids only under fractional crystallization, which preserves zoning.
Q2 — Metamorphism. Metamorphic facies (Eskola) classify rocks by the assemblage stable over a P-T field regardless of bulk composition: zeolite → greenschist → amphibolite → granulite with increasing T; blueschist → eclogite at high-P/low-T (subduction); hornfels for low-P contact. The Al₂SiO₅ polymorphs — kyanite (high P), andalusite (low P), sillimanite (high T) — partition P-T space around a fixed triple point. Observed = assemblage/textures/inclusions and (by microprobe) mineral compositions; inferred = P and T from a calibrated thermobarometer (garnet–biotite Fe-Mg exchange for T; GASP garnet–Al₂SiO₅–plagioclase–quartz for P), which are model outputs from a thermodynamic dataset + activity model. A rock travels a P-T-t loop (burial → heating → exhumation): prograde history survives as inclusions armored in porphyroblasts (garnet growth zoning, Mn-rich core → Fe-rich rim), peak = the matrix/rim assemblage, retrograde = reaction rims/symplectites; a clockwise vs counterclockwise loop is a tectonic interpretation. Main failure mode: partial re-equilibration during retrogression, so a "peak" temperature can be partly reset — better to report "amphibolite facies, ~550–650 °C at mid-crustal P, ± retrograde overprint" than a false-precise single point.
Q3 — Sedimentology & stratigraphy. A facies is a rock body whose grain size/sorting/structures/fossils record its depositional conditions (observe the facies, infer the environment): sorting → transport energy, cross-bedding → currents, graded bedding → waning flow (turbidites), mudcracks/evaporites → exposure, fossils → salinity/depth/oxygenation. Walther's Law — facies in conformable vertical succession were laterally adjacent — lets a vertical core read as a lateral migration (e.g. transgression), but only across conformable contacts (an unconformity breaks it). Relative-dating principles: superposition (test right-way-up with geopetal indicators), original horizontality, lateral continuity, cross-cutting (Hutton), inclusions (clasts older), faunal succession (William Smith, empirical ordering), and unconformities (Hutton's angular unconformity = missing time). The geologic time scale is built in two stages: a relative framework (biostratigraphy + superposition → ordered eons/eras/periods with boundaries fixed at GSSP "golden spikes"), then calibration to numerical ages by radiometric dating of datable tie-point horizons (bentonite/ash zircon U-Pb, lava Ar-Ar) tied into the biostratigraphy — the position rides on fossils, the age-in-years on the radiometric dates.
Q4 — Radiometric dating. First-order decay dN/dt = −λN → N = N₀e^(−λt), t½ = ln2/λ; since N₀ is unknown, use the radiogenic daughter D* = N(e^(λt)−1) → t = (1/λ)ln(1 + D*/N). Measured = present-day parent + daughter; inferred = t, assuming a closed system, known/corrected initial daughter, and known λ. The isochron (daughter vs parent normalized to a stable reference) gives age from the slope and initial ratio from the intercept, if the samples were cogenetic/homogeneous at t=0 and stayed closed. Closure temperature (Dodson) is the linchpin: the clock starts when the mineral cools below the temperature at which the daughter stops diffusing out — so a date is a cooling age (not necessarily crystallization), and different systems yield different dates that map a thermochronological cooling path. U-Pb zircon uses two decay chains (²³⁸U→²⁰⁶Pb, ²³⁵U→²⁰⁷Pb) for a concordia cross-check; zircon's high Pb closure (~900 °C) and exclusion of common Pb make it the gold standard for crystallization ages, with discordance (Pb loss) recoverable via the upper concordia intercept. Ar-Ar (⁴⁰K→⁴⁰Ar via ⁴⁰Ar/³⁹Ar) has lower, mineral-dependent closure (hornblende > biotite > K-feldspar), ideal for thermochronology; a flat step-heating plateau signals a well-behaved system. Failure modes with correct signs: inheritance / xenocrystic cores → too old (cure: date cores and rims separately by SIMS/LA-ICPMS with CL imaging); resetting by later heating → too young; open-system Pb loss → discordance/scatter (bad MSWD — flagged as a statistics matter); excess argon → too old (diagnosed by inverse-isochron intercepts off atmospheric ⁴⁰Ar/³⁶Ar and saddle-shaped spectra).
Q5 — Stable isotopes. δ = [(R_sample/R_standard) − 1]×1000‰ (R = ¹⁸O/¹⁶O vs VSMOW/VPDB, or ¹³C/¹²C vs VPDB); the measured ratio is observed, the paleoclimate inferred. Fractionation is equilibrium (temperature-dependent water–carbonate partitioning — the paleothermometer) or kinetic (rate-dependent — photosynthesis/evaporation take the light isotope). δ¹⁸O carries two entangled signals: temperature (colder water → higher δ¹⁸O in calcite; Urey's carbonate thermometer) and ice volume (evaporation removes light ¹⁶O, which locks into ice sheets during glacials → the ocean is ¹⁸O-enriched) — so higher benthic δ¹⁸O means colder and/or more ice, and a single measurement can't be uniquely inverted without an independent constraint (e.g. Mg/Ca). δ¹³C tracks the carbon cycle: photosynthesis fixes light ¹²C, so organic burial enriches the ocean in ¹³C, while releasing isotopically light carbon (remineralization, methane, volcanic) drives negative excursions (the PETM). Biases: diagenetic overprint (recrystallization resets to burial conditions — the biggest threat, screened by preservation/trace elements/texture), vital effects (species-specific offsets), unknown fluid δ¹⁸O (the central ambiguity), and the uniformitarian caveat (seawater chemistry/pH/biology have changed, so a modern calibration transported to deep time is itself an assumption). Robust: δ¹³C excursions as relative markers, δ¹⁸O as an ordinal cold/warm indicator; contested: absolute paleotemperature from δ¹⁸O alone and deep-time applications.
"How do rocks record the history of the Earth?" — novice (rocks as nature's diary: layers oldest-at-bottom, trapping ripples/shells/mudcracks; crystals from lava/cooking record heat and depth; mineral "clocks" from steady decay — flagging that folding/faulting can invert order and that we infer rather than watch); undergraduate (three coupled archives — stratigraphic/relative, petrologic/thermobarometric, geochronologic/geochemical — with the inverse-problem throughline; soft-pedals equilibrium/closure/diagenesis); graduate (model-dependent, incomplete, non-uniquely-invertible archives: retrograde resetting, cooling vs crystallization ages, proxy convolutions, suspect deep-time uniformitarian calibration, preservation bias — propagate assumptions and report bounded interpretations). Each level states what it simplifies.
vaiu-sci-eaps-chair and noting a xenolith constrains local petrology, not the 3-D velocity field.vaiu-sci-stat-*).Examiner verdict: APPOINT (field 5/5 rubric-correct, 0 fabrications; teaching 3/3; boundary 3/3 including the B2 statistics-referral split). Full report in the sibling result file.