Zr‑MOF Modified Adsorbent: Xilong Scientific Co Ltd’s Published Process for SEMI G5 Electronic Grade Isopropanol
The production of SEMI G5 electronic grade isopropanol (IPA) from technical-grade feedstock requires the removal of water, acetone, light oxygenated organics, metal cations, anions, and particulate matter to levels that will not interfere with advanced wafer cleaning, rinsing, or solvent-based lift-off sequences. Xilong Scientific Co Ltd’s published process for electronic grade IPA places a Zr-MOF-modified adsorbent column downstream of the primary distillation train and upstream of the final particle filtration membrane. The adsorbent is a defect-engineered zirconium-based metal-organic framework that has been modified to increase the number of open coordination sites at the zirconium nodes, thereby improving simultaneous removal of water and trace metal ions. In operation, the feedstock IPA entering the adsorption loop is already a distilled cut with a boiling range controlled between 81.5 °C and 83.0 °C at atmospheric pressure. The residual water content following azeotropic distillation is normally in the range 800 ppm to 3000 ppm, while acetone, diisopropyl ether, mesityl oxide, and acetaldehyde remain at fluctuating concentrations depending on the upstream hydrogenation or hydration pathway. The metal ion burden in the distilled alcohol depends on the storage and transfer metallurgy; stainless steel tanks and piping introduce iron, chromium, and nickel, while glass-lined or fluoropolymer-lined equipment reduces the leached metal load. Published data for the precise feed impurity distribution at Xilong Scientific’s plant is limited, but the company’s process description identifies water, acetone, and metal cations as the three impurity classes that govern the service life of the Zr-MOF bed. The use of a modified adsorbent rather than a conventional molecular sieve is intended to reduce the trade-off between water capacity and metal ion selectivity, because Zr-MOF pores and open metal sites can be tuned to accept small polar molecules while excluding larger impurities that would otherwise occupy adsorption capacity.
What Impurity Cohort Must the Adsorption Step Address in Feedstock IPA?
The impurity cohort in technical-grade IPA is dominated by water, acetone, diisopropyl ether, mesityl oxide, acetaldehyde, methanol, and trace metal cations leached from storage or reaction vessels. Water is the most troublesome bulk impurity because it forms an azeotrope with IPA at approximately 87.7 wt% IPA at 1 atm, limiting simple distillation to the azeotropic composition, after which molecular sieve or adsorption polishing is required to achieve the water specification. Acetone is commonly present from the dehydrogenation or hydrogenation pathway and is particularly problematic because its dipole moment and small kinetic diameter allow it to compete with water for polar adsorption sites, reducing bed life if not removed by upstream distillation. The metal cation fingerprint in stored IPA typically follows iron, sodium, aluminium, chromium, zinc, and calcium, with concentrations varying according to the metallurgy of the upstream pump, tank, and piping surfaces. In electronic grade IPA, acceptance test methods such as ICP-MS, ion chromatography, and optical particle counting are applied to quantify metal ions, inorganic anions, and particles. SEMI G5 does not specify a single universal numerical limit for every impurity; instead, the user and supplier agree on wafer-impact classes, but the commercial target for premium semiconductor solvents is commonly total trace metals below 10 ppb and moisture below 100 ppm. The Zr-MOF-modified adsorbent addresses water by coordinative adsorption at exposed Zr sites, acetone by pore-sieving and polar interaction, and metal cations by ion exchange or surface complexation with carboxylate oxygens or pendant amino groups. The same adsorbent column can simultaneously reduce several impurity classes, but uncontrolled feedstock excursions above design concentration shorten breakthrough time and require more frequent regeneration.
| Parameter | Feedstock range entering adsorption loop | Polished IPA acceptance target | Analytical method/standard |
|---|---|---|---|
| Assay | 99.5–99.8% | ≥99.999% | GC-FID per SEMI C42 |
| Water | 800–3000 ppm | <100 ppm | Karl Fischer per ASTM D6869 |
| Acetone | 10–300 mg/kg | <5 mg/kg | GC-FID |
| Total trace metals | 10–500 ppb | <10 ppb | ICP-MS |
| Anions (chloride, sulfate, nitrate) | 10–200 ppb | <5 ppb each | Ion chromatography per ASTM D4327 |
| Particles ≥0.2 µm | 10³–10⁵ particles/mL | <10 particles/mL | Laser particle counter |
Below the distillation cut, water enters the micropore network of the Zr-MOF and is removed by a combination of micropore filling and site-specific coordination. The parent UiO-66-type structure consists of Zr6O4(OH)4 nodes connected by 1,4-benzenedicarboxylate linkers to form tetrahedral and octahedral cages; the triangular aperture of approximately 6 Å permits water and small polar molecules to enter while excluding larger oligomeric impurities. The modified adsorbent used in the Xilong process has a fraction of missing linkers, so the zirconium atoms at the nodes are not fully coordinated and can accept electron density from water oxygen or acetone oxygen. The heat of adsorption for the first water layer is moderate, typically 45–65 kJ/mol, which is sufficient for trace-moisture polishing but low enough for thermal regeneration. Acetone adsorption is influenced by the pore aperture and the polarity of the framework; however, acetone can also interact with open zirconium sites, so its removal efficiency drops if water loading becomes high and occupies those sites. The bed is therefore operated with a relatively dry feed, and the upstream distillation is controlled to keep the water concentration below the point where the mass transfer zone for acetone breaks through simultaneously with water. The adsorbent column is a vertical cylindrical vessel with a length-to-diameter ratio of 4:1 to 6:1; the packed bed is supported on a bed of inert ceramic balls and covered with a top distributor to provide uniform flow. The superficial liquid velocity is typically in the range 2–5 m/h, and the pressure drop across the bed is maintained below 80 kPa to avoid channeling and pellet attrition. Published data for the exact adsorption isotherm of Xilong Scientific’s modified adsorbent under SEMI G5 operating conditions is limited, but the process behaviour can be inferred from the known properties of defect-engineered UiO-66-family materials.
Zr-MOF Modification Leads to Node Dehydration and Ligand Vacancy Creation
The published Xilong Scientific process description indicates that the Zr-MOF is prepared or modified to enhance the concentration of coordinatively unsaturated Zr(IV) sites rather than to create an ideal perfect crystal. In the synthesis step, zirconium oxychloride or zirconium n-propoxide is combined with 2-aminoterephthalic acid in N,N-dimethylformamide and a monocarboxylic acid modulator such as acetic or formic acid; the modulator competes with the dicarboxylate linker during crystal growth, leaving vacancies that are subsequently occupied by solvent or chloride ions. Solvent exchange with methanol and thermal activation under vacuum removes coordinated water and residual modulator from the framework, producing exposed Zr sites with Lewis acidity that binds water and acetone. The process may also involve post-synthetic exchange of some linkers with a ligand carrying a terminal amino group, increasing hydrophilicity and metal uptake capacity. The resulting adsorbent powder has a BET surface area typically between 900 m²/g and 1400 m²/g and a micropore volume between 0.35 cm³/g and 0.55 cm³/g, depending on the degree of ligand deficiency. To form a packed bed, the powder is blended with a small amount of binder such as polyvinylidene fluoride or colloidal silica and extruded into pellets of 1.5–3.0 mm diameter, then dried at 120 °C under nitrogen. A twin-screw extruder with L/D ratio of 40:1 to 48:1 is used to disperse the binder uniformly and to control pellet density. The binder does not enter the micropores but reduces pressure drop and prevents fines migration. Batch-to-batch variance in the final pellet is controlled by measuring the moisture breakthrough time on a bench-scale column using dry nitrogen doped with 500 ppmv water; acceptable lots are those whose breakthrough time at 10% of inlet concentration falls within ±10% of the reference adsorbent. On manufacturing lines, the adsorbent bed is typically replaced or regenerated when the polished IPA moisture exceeds 50 ppm, which is an early warning threshold before the 100 ppm specification limit.
| Property | Typical range | Equipment/method |
|---|---|---|
| BET surface area | 900–1400 m²/g | Nitrogen adsorption at 77 K per ISO 9277 |
| Micropore aperture | 6–8 Å | CO₂ adsorption/DFT |
| Pellet diameter | 1.5–3.0 mm | Sieve analysis |
| Crush strength | 40–80 N | Uniaxial crush per ASTM D4179 |
| Regeneration temperature | 200–250 °C | Nitrogen purge with dew-point analyser |
| Service LHSV | 0.5–2.0 h⁻¹ | Pilot column |
| Water breakthrough capacity | 8–15 wt% | Dynamic column at 500 ppmv moisture |
| Pressure drop per meter | 20–60 kPa/m | Differential pressure transmitter |
Pore Aperture, Breakthrough Dynamics, and Pressure-Drop Constraints
Pore aperture control is the principal means by which the Zr-MOF-modified adsorbent separates water from bulk IPA, because the IPA molecule has a critical diameter of approximately 0.48 nm, whereas the micropore aperture in the modified Zr-MOF is limited to roughly 0.6–0.8 nm, allowing IPA to diffuse into the pores as a solvent but not to displace all adsorbed water at trace concentration because water is strongly bound at open Zr sites. The breakthrough curve for moisture in a dry IPA feed typically shows a sharp front when the bed is fresh and a broadening front after multiple regeneration cycles. Operators use a mid-bed sample port to detect the position of the mass transfer zone; when the water concentration at the midpoint exceeds 20 ppm, the remaining service time before the outlet reaches 50 ppm is estimated from the historical front velocity. Pressure-drop constraints also shape the pellet formulation; pellets with a diameter below 1.0 mm would improve mass transfer but increase the pressure drop beyond the acceptable limit for the existing feed pump, whereas pellets above 3.5 mm would reduce pressure drop but increase diffusional resistance and produce tailing. The selected pellet diameter of 1.5–3.0 mm is a compromise that keeps the pressure drop between 20 kPa/m and 60 kPa/m while maintaining a sharp moisture front. If the bed is run at high superficial velocity above 5 m/h, water breakthrough occurs earlier because the contact time is insufficient for diffusion into the micropores; below 2 m/h, the bed operates in a near-equilibrium regime but the throughput per unit adsorbent is low. The service LHSV is therefore specified between 0.5 h⁻¹ and 2.0 h⁻¹ for the production column, and pilot-scale verification is performed before scaling to a new batch of adsorbent.
Regeneration of the Zr-MOF bed on Xilong Scientific’s production line follows a thermal swing cycle using low-moisture nitrogen at 200–250 °C for 8–12 h. The heating rate is limited to 1 °C/min to avoid thermal shock and pellet fracture. During regeneration, residual water and adsorbed acetone desorb and are carried to a vent condenser. The nitrogen flow is maintained at 2–4 bed volumes per minute and is filtered through a 0.2 µm particulate filter before introduction. After regeneration, the bed is cooled to 25–30 °C while maintaining a positive nitrogen pressure of 50–100 kPa gauge to prevent rehydration. Repeated thermal cycling causes gradual loss of micropore volume due to local framework dehydroxylation and pellet edge attrition; the process control system tracks cumulative regeneration cycles and compares the pressure drop across the bed against the initial value. An increase in pressure drop above 15% or a decrease in breakthrough time below 80% of the reference value triggers adsorbent replacement. Batch-to-batch variance in the regenerated bed is managed by limiting the number of cycles to fewer than 50 and by verifying the water and metal content of the first batch after each regeneration using Karl Fischer and ICP-MS before allowing the bed to return to normal production.
When Feedstock Moisture Exceeds Distillation Control Limits, Adsorbent Life Drops Nonlinearly
The interaction between the azeotropic distillation column and the downstream Zr-MOF adsorbent is governed by the feed moisture concentration leaving the column. If the distillation column operation deviates and water in the adsorbent feed exceeds 0.5 wt%, the adsorbent bed reaches breakthrough far earlier than the design service time because water loads onto the strongest sites and blocks pores that would otherwise remove metal ions and acetone. The nonlinear relationship between bed life and feed moisture can be approximated by a logarithmic service-time curve; an increase in feed moisture from 0.1% to 0.3% may reduce the usable service life by 40–60%, depending on the adsorbent particle size and bed length-to-diameter ratio. To maintain stable operation, the distillation column is controlled by a reflux ratio algorithm that keeps the water concentration in the overhead product at or below 0.2%. A guard bed of 3A molecular sieve may be placed upstream of the Zr-MOF bed when feedstock water is seasonally elevated, but the Xilong process description indicates that the Zr-MOF bed is intended to handle the final polishing water load without guard bed dependence, because the molecular sieve can introduce aluminium and alkali metal cations if not properly pre-washed. The preferred operating mode is therefore strict distillation control followed by a single Zr-MOF polishing bed, with water challenge tests performed after each column upset.
Metal capture by the Zr-MOF-modified adsorbent occurs through two distinct mechanisms: cation exchange at the carboxylic acid oxygen of the linker and inner-sphere coordination at hydroxyl or amine groups. Trivalent cations such as iron and aluminium bind strongly to the framework and are not readily released during normal IPA processing, which means the adsorbent acts as an irreversible sink for metal contamination during its service life. Sodium and calcium bind less strongly and may migrate through the bed as an adsorption front, so detection of sodium breakthrough at the bed outlet provides an early indicator of metal saturation. The capacity for metal ion uptake is low compared with water uptake, typically in the milligram metal per gram adsorbent range, but it is sufficient because the inlet metal concentration is already low after pre-distillation. To avoid competing anions from the adsorbent itself, the pelletized material is washed with ultrapure water until the rinse conductivity remains below 0.5 µS/cm and the total organic carbon of the rinse is below 100 µg/L. The washed adsorbent is then dried and immediately loaded into the column in a cleanroom that meets ISO 14644-1 Class 5 conditions; personnel contact is minimized to reduce introduction of sodium, potassium, and zinc from skin and gloves.
Does the Modified Adsorbent Leach Zr, 2-Aminoterephthalate, or Fine Particles into the Polished IPA?
The leachable profile of the Zr-MOF adsorbent is a critical acceptance criterion for semiconductor use because any soluble zirconium, organic ligand, or submicron particle would deposit on wafer surfaces or interfere with subsequent process steps. To evaluate leachables, the adsorbent is flushed with ultrapure IPA for 24 h at 25 °C, and the effluent is analysed by ICP-MS and GC-MS. The leachable zirconium concentration is typically below 1 ppb after the initial conditioning flush, while organic ligands are below the GC-MS detection limit of 50 µg/L. The initial flush may release trace amounts of free 2-aminoterephthalic acid, but the production procedure discards the first 3–5 bed volumes of conditioned effluent before the bed is placed online. Fine particle shedding is controlled by pellet strength and by the use of a final 0.1 µm PTFE membrane filter downstream of the adsorbent column. The final filter is not present to improve chemical purity only; it captures adsorbent fines that may be generated during thermal cycling and bed movement. If the filter differential pressure increases by more than 50 kPa, the filter element is replaced and the bed is inspected for pellet breakage. The Xilong process also uses an in-line optical particle counter to monitor particles of 0.2 µm and larger; the finished IPA from a stable bed typically contains fewer than 10 particles/mL at this size threshold, consistent with semiconductor solvent requirements.
Finished SEMI G5 isopropanol leaving the filtration step is transferred under nitrogen to fluoropolymer drums or stainless steel totes that have been pre-cleaned to low particle and metal levels. The transfer line is constructed of PVDF or PFA, and the final container is blanketed with nitrogen after being flushed with filtered IPA. Product release testing includes assay by gas chromatography, water by Karl Fischer, total trace metals by ICP-MS, anions by ion chromatography, and particle counts by laser light scattering. The adsorption bed is not operated to the endpoint of the specification; instead, the control limit for moisture is set at 50 ppm so that analytical variability and column aging do not cause a rejected batch. Operational limitations of the Zr-MOF-modified adsorbent include a maximum service temperature of 280 °C beyond which the binder and pellet morphology degrade; the bed is therefore not suitable for use with steam sterilization above that temperature. The adsorbent is also incompatible with strong aqueous acid washing if the wash contains hydrofluoric acid, because fluoride attacks the zirconium nodes and destroys the framework. Within these boundaries, the polishing adsorption system is suited to the continuous production of electronic grade IPA from a well-controlled distillation feedstock.
