Crystal Size Influence on Acetone Residuals in API Isolation

In pharmaceutical manufacturing, the isolation of an active pharmaceutical ingredient (API) from a final acetone crystallization step directly determines the residual solvent burden in the drug substance—a critical quality attribute governed by ICH Q3C(R9) guidelines. The crystal size distribution and habit profoundly influence the quantity of acetone retained within the dried product because solvent can be superficially adsorbed, capillary‑entrapped between agglomerates, or occluded within intraparticle voids that form during rapid crystallite aggregation under elevated supersaturation. Observations from multiple development campaigns using acetone as a Class 3 anti‑solvent or single cooling‑crystallization medium reveal that a mean crystal diameter (d50) below 50 µm frequently leads to residual acetone levels exceeding the 5000 ppm (0.5%) concentration limit per ICH Q3C(R9) Option 1 after standard agitated nutsche filter‑dryer (ANFD) processing cycles. The underlying mechanism involves high specific surface area combined with mechanical entrapment of mother liquor in fine capillary networks formed when needle‑like or dendritic morphologies with aspect ratios above 4:1 create tight filter cakes of low permeability and poor drainability. By contrast, equant crystals with d50 in the range of 150–300 µm—typically produced through controlled cooling at cooling rates not exceeding 0.2 °C min⁻¹ and seeded with attrition‑milled product of identical crystal form—yield filter cakes that allow efficient displacement washing with 2–3 bed volumes of chilled acetone/water mixtures (95:5 v/v) and faster evaporative solvent removal, achieving final acetone residuals below 1000 ppm without exceeding temperature thresholds that risk polymorphic transformation. The permitted daily exposure (PDE) for acetone of 50 mg day⁻¹ sets the absolute toxicological limit, but for high‑dose formulations where daily intake exceeds 10 g of drug substance, the residual specification must be derived from the PDE by the formula concentration (ppm) = (PDE × 1000) / daily dose, often driving requirements below the default 5000 ppm and placing even greater emphasis on crystal engineering to minimise solvent carry‑over.

Crystal Morphology Dictates Acetone Entrapment Mechanism

The retention of acetone in isolated API crystals is not simply a function of surface area but is commanded by the interplay between growth‑driven lattice imperfections and macroscopic habit. When a crystallisation is operated beyond the metastable zone width—achieved by high anti‑solvent addition rates or cooling ramps exceeding 1 °C min⁻¹—the resulting particles exhibit pronounced skeletal or hopper morphologies where internal solvent inclusions become trapped as the crystal faces advance around liquor‑filled cavities. Scanning electron microscopy of gold‑sputtered samples from such batches has repeatedly shown honeycomb cross‑sections containing acetone‑filled voids of 0.8–5 µm diameter, which are sealed off from the external surface and therefore inaccessible to conventional washing or vacuum stripping at temperatures safely below the API’s glass transition or desolvation‑induced lattice collapse point. Crystals grown under low supersaturation, however, develop faceted, dense exteriors as predicted by the BCF surface‑diffusion model; the near‑equilibrium growth conditions allow solvent molecules to diffuse away from the advancing step train, yielding an inclusion‑free core. Industrial batch records for a development‑stage BCS Class II compound crystallised from acetone confirmed that a shift from a mean aspect ratio of 5.2 (needles) to 2.1 (platelets) through 2 wt% seeding with milled product reduced the residual acetone after identical ANFD washing and double‑cone drying at 50 °C/30 mbar for 6 h from 6200 ppm to 850 ppm, a 7.3‑fold decrease. The mechanism was attributed to elimination of occluded solvent and a reduction in capillary‑held liquor within agglomerates, as the plate‑like particles packed with larger interstitial channels that permitted more complete de‑liquoring under the 0.46 bar pressure differential applied across the 5 µm retention‑rated PTFE filter cloth.

How Does Filter Cake Permeability Scale with Particle Size Distribution in Acetone‑Wetted Isolations?

The resistance to flow through a filter cake composed of crystalline particles obeys the Kozeny‑Carman relationship, wherein the permeability coefficient scales with the square of the Sauter mean diameter (D[3,2]) and the cube of the cake porosity. For a typical acetone‑soaked crystal mass isolated in a 0.6 m² ANFD equipped with a retreat‑curve impeller and a hydraulic pressing plate, a shift in D[3,2] from 35 µm to 120 µm lowers the specific cake resistance by more than an order of magnitude, from approximately 8 × 10¹¹ m kg⁻¹ to 5 × 10¹⁰ m kg⁻¹, enabling effective displacement washing at differential pressures as low as 0.3 bar. In production campaigns where needle‑like fines (d90 < 40 µm) were generated by a rapid anti‑solvent addition at the 5 L min⁻¹ scale, filter blinding occurred within the first 30 s of pressure application, and the resultant channeling gave a washing efficiency of merely 40% as measured by the reduction in chemical oxygen demand of the mother liquor expressed as TOC. When the same compound was crystallised with a linear cooling ramp of 0.15 °C min⁻¹ from 55 °C to 20 °C in the presence of 1.5 wt% seed crystals milled to a d50 of 80 µm, the D[4,3] of the final slurry reached 210 µm, cake porosity exceeded 0.42, and the pressure‑equalising step during nitrogen‑pressurised deliquoring achieved a residual moisture (acetone) of 12% w/w before thermal drying, compared to 38% for the fines‑laden variant. The relationship between particle size distribution span (span = (D90 − D10)/D50) and residual solvent is also significant; a span below 1.2 ensures uniform cake compressibility and avoids localised regions of high solvent retention that are undetectable by averaged loss‑on‑drying measurements but cause excursions when individual blend uniformity samples are tested by headspace GC according to USP <467>.

When Polymorphic Transitions Limit Drying Temperature for Desolvation

A narrow processing window with critical threshold risks emerges when the API exhibits a thermally induced polymorphic conversion within 5 °C of the temperature required to efficiently remove acetone from small‑diameter crystals. Thermogravimetric analysis coupled with online mass spectrometry (TGA‑MS) performed at 5 °C min⁻¹ on a sample with d50 of 42 µm consistently reveals a biphasic acetone loss profile: a rapid surface‑desorption peak centred at 56 °C accounting for approximately 65% of total solvent, and a broad, diffusion‑limited shoulder extending from 80 °C to 105 °C corresponding to the release of occluded solvent. For a metastable Form II of a proprietary API that converts irreversibly to the thermodynamically stable Form I at 63 °C (onset by differential scanning calorimetry at 10 °C min⁻¹), the drying temperature cannot exceed 60 °C without risking uncontrolled form change and the consequent particle bridging that reintroduces solvent into new interparticle capillaries. At this limiting temperature the intraparticle diffusion coefficient for acetone in the crystal lattice, estimated from desorption curve fitting with Fickian diffusion models for a sphere of radius equal to the Sauter mean diameter, is on the order of 1 × 10⁻¹⁴ m² s⁻¹, requiring drying times in excess of 48 h under vacuum (10 mbar) to reach 5000 ppm when the crystal size falls below 50 µm. Larger crystals of 180 µm mean diameter, although still affected by the temperature ceiling, release surface and shallow‑occluded acetone more rapidly because a greater proportion of the solvent is located within 10 µm of the external surface; full compliance with the 5000 ppm limit was achieved after 12 h of drying in the same double‑cone rotary dryer operating at 60 °C/10 mbar. Real‑time monitoring of polymorphic purity by in‑line Raman spectroscopy during drying has been deployed to terminate the cycle as soon as the Form I characteristic peak at 1682 cm⁻¹ appears above a threshold of 0.2% of the Form II peak area, safeguarding both residual solvent compliance and solid‑form integrity.

Micronisation of dried API frequently exposes formerly occluded solvent pockets and regenerates acetone levels that were below the limit of quantitation in the pre‑milled material. Jet‑milling of a batch originally containing 620 ppm acetone, performed at a grinding pressure of 6 bar and a classifier speed yielding a target d50 of 5 µm, raised the headspace GC reading to 3400 ppm because internal pores of 0.3–1.5 µm, previously sealed by crystal growth, were fractured open. This phenomenon necessitates that residual solvent specifications be set on the final micronised drug substance rather than on the unmilled intermediate, and that crystallisation‑growth parameters be adjusted to produce crystals with a mean diameter at the upper end of the permissible range for downstream processing—typically 60–90 µm for micronisation feed—to minimise the volume fraction of occluded solvent before size reduction. When such a size target cannot be met due to bioavailability constraints, a post‑crystallisation solvent‑exchange slurry wash with a lower‑boiling Class 3 solvent such as ethyl acetate (PDE 50 mg day⁻¹, limit 5000 ppm) has been employed to displace acetone from the accessible pore network; subsequent drying at 45 °C under 30 mbar achieves ethyl acetate residuals below 1000 ppm even in fines‑dominated populations. Published data for the specific configuration of solvent‑exchanged micronised APIs with varied initial crystal size distribution are limited, but mass‑balance calculations based on measured intraparticle porosity by mercury intrusion porosimetry indicate that a crystal porosity exceeding 3% v/v will carry sufficient solvent to exceed the ICH limit unless the trapped solvent is physically displaced before the final drying step.

Crystal Size Fraction (µm)	d50 (µm)	Span	Residual Acetone Post Drying (ppm)	Drying Cycle	Observations
20–50	34	1.9	5200	45 °C / 30 mbar / 8 h	Needle agglomerates; high occluded solvent; failed ICH limit
50–125	82	1.4	1750	50 °C / 30 mbar / 6 h	Plate‑like habit; partial inclusion; acceptable for low‑dose products
125–300	195	1.1	420	55 °C / 20 mbar / 4 h	Equant, dense crystals; surface acetone only; easily below limit

Data aggregated from three development campaigns employing acetone as primary solvent and identical ANFD washing protocol (2 × 2 L acetone rinses per 1 kg wet cake) with vacuum shelf drying. Analytical determination by headspace GC‑FID per USP <467> with a DB‑624 (6% cyanopropylphenyl‑94% dimethylpolysiloxane) column, 30 m × 0.32 mm × 1.8 µm, injection temperature 140 °C, equilibration 80 °C for 30 min. LOQ 10 ppm. Span defined as (D₉₀−D₁₀)/D₅₀ from laser diffraction (Malvern Mastersizer 3000, wet dispersion in acetone‑saturated heptane).

Solvent	Class (ICH Q3C)	PDE (mg/day)	Concentration Limit (ppm, Option 1)	Analytical Method Reference
Acetone	Class 3	50	5000	USP <467> Method IV
Ethyl acetate	Class 3	50	5000	USP <467> Method IV
Isopropyl alcohol	Class 3	50	5000	USP <467> Method IV
Methanol	Class 2	30	3000	USP <467> Method IV

Regulatory thresholds per ICH Q3C(R9) and pharmacopoeial testing methodology. For acetone‑based isolations, the Class 3 default limit of 5000 ppm applies only when daily intake of drug substance is ≤10 g; higher doses require calculation from the PDE of 50 mg/day, which may yield tighter limits down to 250 ppm for a 200 g daily dose. In such instances, even modest crystal‑size effects become the dominant risk control lever.

A case encountered during tech transfer of a final API step involved a twin‑screw continuous crystalliser with residence time less than 2 min that produced a crystal bed with d90 of 28 µm. Despite post‑crystallisation batch‑wise ripening under dynamic light scattering control at 40 °C for 4 h, the crystal size could not be increased beyond a d50 of 45 µm without dissolving the metastable form. Drying in a conical screw dryer at 40 °C/5 mbar required 72 h to reduce acetone to 4800 ppm, and intermittent sampling revealed an asymptotic approach to a residual floor of 2200 ppm attributed to lattice‑occluded solvent in vacuoles identified by focused ion beam sectioning. The deviation was resolved by substituting the final wash solvent with 2 bed volumes of anhydrous ethanol (ethyl alcohol) for acetone displacement, a step validated by demonstrating that ethanol, though a Class 3 solvent with the same PDE of 50 mg/day, has a higher diffusivity in the crystal matrix at sub‑50 °C temperatures; after the solvent switch, the residual ethanol was below 800 ppm after 24 h of drying. This example underscores the operational boundary that crystal size distributions with d50 < 50 µm and span >1.8 are unlikely to meet residual acetone specifications without either exhaustive drying that challenges polymorphic stability or a specifically designed solvent‑exchange protocol validated against the PDE and analytical sensitivity of the finished dosage form.