Identifying which crystal form of an API is present by matching its powder X-ray diffractogram against reference patterns — the primary identity test in CMC release specifications
X-ray powder diffraction (XRPD/PXRD) is the single most important solid-state technique for confirming which crystalline form of a drug substance is present in a given lot of material. Because different polymorphs pack their molecules into different unit cells, each form produces a diffraction pattern as distinctive as a fingerprint — and getting a clean, artifact-free scan starts with careful sample preparation and correct instrument geometry, since preferred orientation or poor packing can distort relative peak intensities enough to cause a false form call.
Bragg-Brentano reflection geometry: • X-ray source (sealed Cu tube, Kα1/Kα2 doublet, or monochromated Kα1 only via Ge(111) or Johansson monochromator) irradiates a flat powder sample • Detector scans through angle 2θ while the sample rotates through θ, maintaining the focusing condition that keeps diffracted rays converging at the detector slit • Cu Kα is standard for organic/pharmaceutical materials (λ=1.5406 Å gives good peak separation across the 3-40° 2θ range where most API reflections fall); Mo or Co sources used for special cases (fluorescence avoidance with Fe/Co-containing samples)
Sample preparation — the most common source of error: • Back-loading into a zero-background (single-crystal Si, cut off-axis) holder minimizes preferred orientation vs. front-loading or pressing • Preferred orientation: needle- or plate-shaped crystallites tend to pack with a preferred crystallographic axis normal to the sample surface, artificially amplifying some peaks and suppressing others — a major cause of misleading relative intensities (though peak POSITIONS remain correct, so form identification by position is more robust than by intensity ratio) • Gentle spatula loading (not pressing or grinding, which can induce a polymorphic transition or amorphization via mechanical stress) preserves the as-received crystal habit • Sample must be representative — micronized/milled material for a dissolution study may show peak broadening (crystallite size effect via the Scherrer equation) unrelated to polymorphic form
Instrument calibration and validation: • NIST-traceable standards (Si 640d, LaB6 660c, or corundum) run periodically to verify 2θ zero-point accuracy and instrument resolution function • Zero-shift and sample-height displacement errors both shift all peaks systematically — corrected via internal or external standard calibration before quantitative comparison to reference patterns • Divergence slit, receiving slit, and soller slit configuration affect peak shape/resolution and must match between sample and any reference pattern being compared
As the goniometer steps through 2θ, the detector records diffracted X-ray intensity at each angle, building the raw diffractogram — intensity versus 2θ. Each peak represents constructive interference from a specific family of crystallographic lattice planes, and Bragg's Law provides the direct mathematical bridge between the angle where a peak appears and the physical interplanar spacing (d) of the crystal responsible for it.
Bragg's Law derivation and application: • nλ = 2d·sinθ, where n is the diffraction order (usually taken as 1), λ is the X-ray wavelength (1.5406 Å for Cu Kα), d is the interplanar spacing of the diffracting lattice planes, and θ is half the measured 2θ diffraction angle • Rearranged for direct use: d = λ / (2·sinθ) — every peak position in the raw scan converts directly to a d-spacing in Ångströms, independent of the specific compound (a purely geometric relationship) • Low 2θ angles (large d-spacings, e.g., 5-10° 2θ → d≈9-18 Å) probe large-scale lattice repeat distances, often corresponding to layer spacings or long unit cell axes • High 2θ angles (small d-spacings) probe finer structural detail within the unit cell
Anatomy of the raw pattern: • Sharp, well-resolved peaks indicate a well-crystallized material with long-range order • Peak position (2θ or d-spacing) is determined by unit cell geometry — this is the primary polymorph fingerprint • Peak intensity depends on structure factor (atomic positions and scattering power within the unit cell) and is secondarily affected by preferred orientation, so intensity alone is a less reliable diagnostic than position • Peak width (FWHM) relates to crystallite size (Scherrer equation: size ≈ Kλ/(β·cosθ)) and microstrain — informative for particle engineering but not primary for form identification • A broad, low, hump-like "amorphous halo" (no sharp peaks) instead of discrete reflections indicates absence of long-range crystalline order — the sample is amorphous or highly disordered
Background and artifact features: • Broad background hump beneath sharp peaks: air scatter, sample holder fluorescence, or amorphous content coexisting with the crystalline phase • Kα2 shoulder peaks (from the Kα1/Kα2 doublet, unless stripped or monochromated away) appear as small satellite peaks slightly above each main Kα1 peak, especially visible at higher 2θ • Instrumental knowledge of these artifacts prevents them from being mistaken for genuine additional polymorphic peaks
Indexing is the process of assigning each observed diffraction peak to a specific (hkl) Miller index — the family of crystallographic planes responsible for it — and, from the complete set of indexed peaks, solving for the unit cell parameters (a, b, c, α, β, γ) that define the repeating box of the crystal lattice. Successful indexing is itself powerful evidence that a pattern represents a single, phase-pure crystalline form, since a mixture of two polymorphs typically cannot be indexed by any single unit cell.
Why indexing from powder data is hard: • Single-crystal diffraction directly measures 3D reciprocal space; powder diffraction collapses this onto one dimension (2θ), so reflections from different (hkl) planes with similar d-spacings can overlap — the central challenge of "ab initio" powder indexing • Accurate, well-resolved peak positions from the low-angle region (which has sparser peak density and cleaner resolution) are weighted most heavily by indexing algorithms
Common auto-indexing algorithms: • DICVOL (dichotomy method): systematically searches cell parameter space, testing trial cells against the observed peak list • TREOR (trial-and-error): successive dichotomy on zones, historically the most widely used routine indexing program • Singular value decomposition / whole-pattern approaches (e.g., McMaille, using simulated annealing): more robust for lower-symmetry, harder-to-index patterns • Output: candidate unit cells ranked by a figure of merit (de Wolff M(20), or de Wolff/Smith F(N)) — M(20) > 20 with 20+ correctly indexed lines is generally considered a reliable, trustworthy solution
From unit cell to space group and crystal system: • The relative values and angles of a, b, c, α, β, γ classify the cell into one of the 7 crystal systems (triclinic, monoclinic, orthorhombic, tetragonal, trigonal, hexagonal, cubic) — organic pharmaceutical crystals are most commonly monoclinic or orthorhombic • Systematic absences (certain hkl classes missing from the pattern) narrow the possible space groups — e.g., a P2₁/c space group shows systematic absence of 0k0 reflections with odd k • Full space group and atomic coordinate solution typically requires additional structure-solution software (e.g., DASH, TOPAS, FOX) using simulated annealing or charge-flipping against the full powder pattern — effectively solving a crystal structure from powder data alone when single crystals are unavailable
Practical value of successful indexing for polymorph screening: • A pattern that indexes cleanly to ONE unit cell with all peaks accounted for confirms phase purity (single polymorph, no detectable second form) • A pattern that fails to index, or requires two independent unit cells to explain all peaks, signals either a polymorphic mixture or the presence of an unindexed impurity phase — both require follow-up before batch release
For a routine quality-control or release test, full ab initio indexing is rarely necessary — instead, the sample's diffractogram is compared directly against a pre-established internal reference library (patterns collected from authenticated, single-phase reference standards of each known polymorph) or an external database like the ICDD Powder Diffraction File (PDF). When quantifying the relative amount of each polymorph in a mixture — critical when a manufacturing process risks partial conversion — Rietveld whole-pattern refinement provides the gold-standard quantitative answer.
Search-match against reference libraries: • Internal polymorph library: patterns collected under identical, validated instrument conditions from well-characterized (single-crystal-confirmed) reference standards of each known form — the most reliable comparison for routine QC • Peak position list comparison: the sample's major peak 2θ positions (typically the top 3-5 most intense, most diagnostic peaks) are compared against each library form; a match within instrument peak-position tolerance (typically ±0.1-0.2° 2θ) across all diagnostic peaks confirms form identity • ICDD PDF (Powder Diffraction File): external reference database exceeding 1,000,000 patterns across all material classes; used when no internal reference exists yet, or to cross-check against literature-reported forms • Classical search-match algorithms (Hanawalt method: strongest 3 lines as a d-I triplet; Fink method: 8 strongest lines) pre-date modern computerized full-pattern matching but remain conceptually foundational
Rietveld refinement for quantitative phase analysis (QPA): • Whole-pattern fitting: a calculated diffraction pattern, generated from the known/refined crystal structure(s) of each phase present, is iteratively refined (least-squares) against every data point of the observed pattern — not just peak positions, but the full profile shape • Refined parameters: phase scale factors (directly proportional to weight fraction), unit cell parameters, peak shape/width parameters, preferred orientation correction, background • Goodness-of-fit metrics: Rwp (weighted profile R-factor, <10% considered a good fit for organic pharmaceuticals), Rexp (expected R-factor from counting statistics), and χ² = (Rwp/Rexp)² • Multi-phase refinement directly outputs the weight fraction of each polymorph present in a mixture — essential for setting and testing against a specification limit (e.g., "Form II ≤ 2.0% w/w" in a drug substance CoA) • Detection/quantification limits for a minor polymorph depend heavily on how well-resolved its diagnostic peaks are from the major form's pattern — well-separated diagnostic peaks can push quantification below 1% w/w; heavily overlapping patterns may only achieve 3-5% LOQ
Polymorphism is not an academic curiosity — an undetected or uncontrolled polymorphic transition in a marketed drug product can alter bioavailability, dissolution rate, or manufacturability enough to trigger a product recall. Regulatory guidance (ICH Q6A) explicitly recommends PXRD as an appropriate identity and, where relevant, quantitative control test for polymorphic drug substances, and the pharmaceutical industry's most infamous case study — ritonavir — remains the canonical justification for rigorous, ongoing polymorph screening throughout a product's lifecycle.
The ritonavir Form II crisis (1998) — the field-defining case study: • Ritonavir (Norvir®, Abbott Laboratories), an HIV protease inhibitor, was launched in 1996 as a semi-solid capsule formulation using what was believed to be the only known crystalline form (subsequently called Form I) • In mid-1998, batches began failing dissolution testing; investigation revealed a previously unknown, more thermodynamically stable polymorph — Form II — had spontaneously appeared, first in manufacturing and then even in some marketed capsules • Form II was significantly less soluble than Form I, sufficient to compromise bioavailability; once Form II nucleated in a manufacturing facility, it proved extremely difficult to eliminate (cross-contamination of the more stable form via airborne seed crystals and shared equipment) • Abbott was forced to pull product from the market and reformulate (ultimately to a liquid-filled capsule then later a tablet using a stabilized amorphous solid dispersion) — an estimated cost exceeding $250 million and a manufacturing gap of several months • Root cause, retrospectively: insufficiently exhaustive polymorph screening before launch failed to discover the thermodynamically more stable Form II, which existed but had never crystallized under the conditions initially used
ICH Q6A and pharmacopeial requirements: • ICH Q6A ("Specifications: Test Procedures and Acceptance Criteria for New Drug Substances and New Drug Products") provides a decision-tree framework (Decision Tree #4) for when polymorph testing and control should be built into a specification — triggered when multiple forms exist AND bioavailability, stability, or processability differ meaningfully between them • USP General Chapter <941> (X-Ray Diffraction) and Ph. Eur. 2.9.33 describe PXRD methodology standards for pharmaceutical characterization • Once a polymorph-sensitive specification is established, PXRD is typically run as an identity test at both drug substance release and, where relevant, on the final drug product to confirm no form conversion occurred during formulation/processing (e.g., wet granulation, which introduces water and can trigger a hydrate transition)
Other well-documented polymorph case studies used in industry training: • Carbamazepine: four anhydrous polymorphs (I-IV) plus a dihydrate — the most extensively studied model compound in academic crystal engineering literature, used to validate new PXRD indexing and screening methodologies • Chloramphenicol palmitate: Form A (active) vs. Form B (inactive) — an early, pre-regulatory-era cautionary tale from the 1960s illustrating that polymorph-driven bioavailability failure predates modern regulatory frameworks • Rotigotine (Neupro patch): a 2008 field recall after unexpected Form II crystallization reduced drug release from the transdermal patch matrix — a modern reminder that polymorph risk persists across all dosage forms, not just oral solids
The ritonavir case reshaped pharmaceutical development practice: virtually every modern small-molecule NCE program now runs a systematic polymorph screen (often 100+ crystallization conditions across solvents, cooling rates, and antisolvent combinations) before Phase III, specifically to find the thermodynamically most stable form before a manufacturing plant finds it by accident. PXRD is the primary analytical technique that makes this screening — and the ongoing lot-release identity testing that follows — possible.