Electronic Grade Ipa Semi G1 To G5 Metal Limits And What Each Grade Actually Cleans
Electronic-grade 2-propanol (CAS 67-63-0) is a five-tier contamination hierarchy rather than a single solvent. The designation electronic grade IPA: SEMI G1 to G5, metal limits and what each grade actually cleans is governed by residual metal concentration, water content, nonvolatile residue, and particle load. The solvent matrix itself remains 2-propanol with a density of 0.786 g/cm³ at 20 °C, a boiling point of 82.6 °C, a vapor pressure of 4.4 kPa at 20 °C, a surface tension of 21.7 mN/m at 20 °C, and a flash point of 12 °C. The public semiconductor reference is SEMI C7, Specification for 2-Propanol, supported by ICP-MS metal analysis against NIST-traceable multi-element standards.
Metal limits and what each grade actually cleans are linked through an evaporation residue budget. If a 300 mm wafer retains 10 mL of IPA after spin-off, a 1 µg/kg Fe ceiling contributes approximately 7.9 ng of Fe if the entire retained volume dries. The calculation changes only in scale from grade to grade: a 100 µg/kg metal ceiling in a G1 product would leave approximately 0.79 µg of a single element under the same retention assumption. This is why G1 may be appropriate for a stainless-steel bench degreaser but not for patterned wafer contact. The solvency does not change; the residue left after evaporation does.
How Do Metal Ceilings Separate One 2-Propanol Molecule into Five Usable Grades?
The official framework must be read with supplier specification sheets because the public edition of SEMI C7 and several advanced tiers are often implemented through bilateral quality agreements. The table below lists representative procurement windows for electronic-grade IPA labelled G1 through G5. Metal ceilings are expressed as µg/kg, equivalent to ppb, and are measured per element by ICP-MS after direct injection or evaporation, depending on detection-limit requirements.
| Grade | Metal per element by ICP-MS | Water | Particle criterion | What this grade is typically allowed to clean |
|---|---|---|---|---|
| SEMI G1 | ≤ 100 µg/kg each | ≤ 0.15 vol% | ≥ 0.5 µm: ≤ 25 particles/mL | Non-wafer equipment: stainless-steel chamber parts, quartzware exteriors, ultrasonic bath tanks, screws, load-lock covers |
| SEMI G2 | ≤ 20 µg/kg each | ≤ 0.1 vol% | ≥ 0.5 µm: ≤ 25 particles/mL | Wafer cassettes, carrier plates, spinner bowls, transfer blades, and non-critical fluid lines |
| SEMI G3 | ≤ 1 µg/kg each | ≤ 0.05 vol% | ≥ 0.5 µm: ≤ 10 particles/mL | Legacy wafer surfaces as a direct rinse, pre-diffusion and pre-oxidation cleans, and moderate-voltage power device wafers |
| SEMI G4 | ≤ 0.1 µg/kg each | ≤ 0.03 vol% | ≥ 0.3 µm: ≤ 5 particles/mL | Post-CMP wafers, copper damascene interconnect surfaces, pre-epitaxial wafers, and advanced BEOL structures |
| SEMI G5 | ≤ 0.01 µg/kg for Na, K, Ca, Fe, Cu, Zn, Cr, Ni, Al; ≤ 0.05 µg/kg for other transition metals | ≤ 0.02 vol% | ≥ 0.2 µm: ≤ 2 particles/mL | Gate-oxide preclean, high-k/metal gate, EUV mask substrates, and high-aspect-ratio memory stack contact surfaces |
In a production-scale immersion cleaner, G1 IPA is typically dispensed through point-of-use filters into closed stainless-steel vessels at 40 kHz ultrasonic energy and 55 °C bath temperature. The solvent swells and removes light machine oils, fingerprints, and rosin-based particulate films from disassembled parts. It does not remove oxide scales, inorganic salts, or metal stains; those require acid or chelating treatments before IPA rinsing. G2 is allowed on wafer cassettes and transfer equipment because its 20 µg/kg per-element ceiling leaves a lower evaporation residue. Direct wafer contact begins at G3 in many older fabrication flows, where the device defect budget tolerates a 1 µg/kg individual metal ceiling. G4 is the standard direct-wafer rinse in logic and memory fabs operating below 28 nm because the post-CMP copper surface is particularly sensitive to Fe, Cu, and Ni re-deposition. G5 is reserved for surfaces in which a single metal spike above 0.01 µg/kg for sodium or iron can shift threshold voltage or degrade gate oxide integrity.
Direct injection ICP-MS for organic solvents requires aerosol dilution, solvent-load matching, and cool plasma or collision-cell conditions to avoid carbon-based polyatomic interferences. For G5, the method detection limit for Na, K, Ca, and Fe in IPA should be below 0.005 µg/kg; otherwise the 0.01 µg/kg specification cannot be verified. The use of NIST-traceable standards and cleanroom sampling procedures is mandatory. If sampling is conducted in a standard laboratory glass bottle, sodium and boron leaching can exceed the G5 limit within hours. The grade certificate is therefore valid only at the point of package; contamination is not self-limiting after dispensing.
Metal Limits Are Only Half the Specification for Electronic-Grade IPA
Water content is a separate rejection parameter because it changes the drying mechanism. A G4 product with 0.03 vol% water may pass the metal table, but excessive water can raise the surface tension of the final film and slow Marangoni drying. Karl Fischer titration is the reference method for water in electronic-grade IPA under SEMI C7; gravimetric residue after evaporation is used for nonvolatile residue. Particle counts are measured by laser particle counters in size bins of 0.5 µm, 0.3 µm, or 0.2 µm, depending on grade. A solvent with compliant metals but 25 particles/mL of 0.5 µm particles can still leave hard defects on a wafer.
Acidity, chloride, sulfate, phosphate, and acetone content are also controlled. Chloride is often tested by ASTM D512 or ion chromatography; acetone is monitored because it can influence residue solubility and flash point. For a G5 product, an anion spike of chloride or sulfate at low µg/kg levels is sufficient to create ionic contamination after evaporation. The point of use must be equipped with high-purity pressure dispensing and filtered nitrogen blanketing because IPA is hygroscopic and can absorb water from cleanroom air. At 23 °C and 50% RH, the equilibrium water content of a partially filled IPA container can exceed 0.1 vol%. A high-purity certificate is therefore not self-preserving in an open container or an unblanketed dispense vessel.
When a Final Rinse Shared by a Bench Degreaser Meets a 5 nm Gate Stack
The cleaning mechanism of 2-propanol is solvation of low-polarity organic residues, not complexation or oxidation of metals. In a wet-bench sequence such as SC1 (NH₄OH/H₂O₂/H₂O at 1:1:5) or SC2 (HCl/H₂O₂/H₂O at 1:1:6), the aqueous chemistry removes metallic and particulate contamination; IPA displaces the water film and carries away traces of the wash solution. If the IPA itself contains Fe at 0.1 µg/kg, the rinse does not remove that Fe; it deposits it. Therefore, the grade does not determine what the solvent can dissolve—it determines what the solvent may touch without recontaminating the surface.
On a copper damascene wafer after CMP, IPA is used in brush scrubbers and spin rinses to displace slurry-laden water from low-k trenches. A direct G3 rinse may be acceptable for a 130 nm product, but the same liquid in a 5 nm flow leaves too much cation residue in the critical dimension. The relevant surface is not the only factor: rinse time, spin speed, and exhaust rate determine how little liquid remains. Single-wafer spin processors typically run at 1,200–1,800 rpm during IPA rinse, with dispense rates of 0.5–1.5 L/min. Production-scale tools add point-of-use purifiers and metal-stable polymer dispense lines because the container itself can contribute sodium, calcium, and zinc at ppt levels if borosilicate glass or unpassivated stainless steel is used.
For a production-scale ion-source housing, G1 cleaning is performed after removal of photoresist carbon and sputtered metal films. The IPA removes residual organics and wipes away particulate residues. The same G1 product cannot be used on wafers because a single evaporated droplet can exceed the local defect budget. G2 is used in cassette and transfer-tool cleaning where the surface will later contact wafers, such as FOUP shelves and robot paddles. G3 is allowed as a direct wafer rinse only where metal-contamination tolerance is wide enough to absorb the evaporation residue. G4 is commonly specified for post-CMP cleans because the wafer surface contains exposed Cu and low-k dielectric; the metal residues from a G1 rinse would be in the same concentration range as the contaminants being removed. G5 is specified where trace Na, K, Ca, and Fe cannot be tolerated in the final drying film.
Particle Counts, Drying Performance and Flammability Boundaries
Sub-0.3 µm particle counts matter when the IPA is used as a Marangoni drying agent. The low surface tension of 21.7 mN/m at 20 °C allows the vapor to create a surface-tension gradient that pulls water from the wafer. Residual nonvolatile particles in the IPA remain at the drying front. This is why G5 particle criteria are often set at ≥0.2 µm: ≤2 particles/mL. A liquid particle counter with a 0.2 µm detection threshold is required; a 0.5 µm counter can miss sub-0.3 µm particles that are large enough to bridge a narrow trench.
Flammability is an operational boundary independent of metal grade. 2-Propanol has a flash point of 12 °C, a flammability range of approximately 2.0–12.7 vol% in air, and an autoignition temperature near 425 °C. Wet benches and spin coaters must be exhausted to keep vapor concentration below the lower flammability limit. Point-of-use containers should be grounded and kept closed; the solvent is incompatible with strong oxidizers, including concentrated nitric and sulfuric acids, and should not be mixed with such oxidizers in waste drains.
Electronic-grade IPA is not a replacement for aqueous metal removal, photoresist stripping, or post-etch residue removal. A G5 product can clean a surface only after the preceding aqueous sequence has already removed the contaminant. If the residue is an inorganic salt, the IPA rinse will not dissolve it; if the residue is an organic film, the IPA will solvate it but the spent solvent must be removed before drying. Published data for the exact G5 per-element limits must be obtained from the supplier certificate and the applicable SEMI C7 revision because public summaries do not contain all elements at all nodes.