SEMI C28 grade claims on hydrofluoric acid rest on test methods, not assay

Two quotes for 49% hydrofluoric acid land on the same desk. One says industrial grade and prints an assay. The other says semiconductor grade, cites SEMI C28 and lists every trace metal in fractions of a part per billion. It looks stricter, and may be. But a limit that low is only as good as the method that can detect it, and SEMI’s public pages print no grade limits to check it against.

Before you compare the two, find the document, the edition and the methods behind each number.

How does a SEMI C28 grade differ from industrial grade hydrofluoric acid?

A SEMI C28 grade sets semiconductor-industry requirements with supporting test procedures, and SEMI C10 says each trace contaminant’s method detection limit should sit at or below its specification. Industrial grade hydrofluoric acid is bought mainly on assay: O-H-795 (1953), a federal technical-grade specification, set a 60% HF floor checked by titration, and 49% grades carry their own assay range.

The government purchase specifications show what a technical grade measured. O-H-795 covered technical hydrofluoric acid for removing sand from metal castings and for etching glass. Its tests were wet chemistry in platinum ware: a two-endpoint sodium hydroxide titration for total acidity and hydrofluosilicic acid, an evaporation residue for sulfuric acid and an iodine titration for sulfurous acid. MIL-A-24641A (1986) set the same 60% floor and the same titration, warming to 176 °F (80 °C) for the second end point.

Commercial sheets follow the same pattern. A producer’s 49% specification sheet names assay by titration, non-volatile acidity, sulfur dioxide by iodimetry and arsenic by ICP-AES, and a producer’s safety data sheet for 49% acid lists metal pickling, glass etching, chemical derivatives and semiconductor etching as uses. One strength serves four markets, so the grade, not the 49%, separates a pickling line from a wafer fab.

For metal cleaning, pickling and glass finishing, the bath recipe is built around strength, assay and a few acid impurities; the posts on uses of hydrofluoric acid by industry and glass etching and frosting cover those baths. On a wafer, trace metals are the point. A 1993 study of metal deposition from HF solutions reports that metals deposited during processing can increase junction leakage and degrade oxide breakdown strength, and a 2001 Journal of the Electrochemical Society paper notes that copper deposits easily on silicon from HF-based etching solutions.

An inquiry for “anhydrous 49%” acid names two different products: anhydrous hydrogen fluoride (AHF) and 49% aqueous acid. EPA’s 1993 report on hydrogen fluoride noted that bulk rail and truck shipments were mainly anhydrous HF and 70% aqueous acid. SEMI’s provisional standard for anhydrous HF gas, C3.43, was published in 1990 and withdrawn in 2001. The comparison of 49% and 70% hydrofluoric acid grades covers the two aqueous strengths.

What does SEMI C28 cover, and what does it leave to other documents?

SEMI C28, the Specification and Guide for Hydrofluoric Acid, standardizes requirements for hydrofluoric acid used in the semiconductor industry, with supporting test procedures, and its scope is the grades of HF used there. The current revision, C28-0618, carries Inactive status on the SEMI store but remains valid for use. A quote that says “SEMI grade” without naming the document and edition leaves both open, so ask for C28-0618 by name.

The revision history explains old references. C28 was first published as C28-0699, replacing the separate documents SEMI C1.8, C7.3, C8.3 and C11.3, with technical revisions coded 0301, 0704, 0705, 0306, 0611 and 0618 (month and year). SEMI M33, a wafer-surface test method withdrawn in 2007, still lists C7.3 as “Standard for Hydrofluoric Acid, Grade 2”. If a legacy internal specification cites C7.3, it points at a document C28 replaced; update the reference before you hold quotes against it.

C28 is not the only SEMI document for fluoride etch chemistry. Its neighbors are specified separately:

So a SEMI line on a quote for 4.9% (10:1) HF, solar-cell HF or a buffered etch points at C29, PV11 or C23, not C28, and none of these public abstract pages prints grade limits. A SEMI number is not always a product grade. SEMI F40, a practice for preparing distribution components for chemical testing, names 49% hydrofluoric acid as one of its test chemicals, and SEMI M51, a gate-oxide-integrity test method, lists C28 among its referenced standards. Neither is a purchase specification for the acid on your quote.

Which test methods stand behind a SEMI grade claim?

SEMI C1, the guide for the analysis of liquid chemicals, states that each liquid chemical standard consists of guidelines, specifications and validated analytical procedures, with procedures defined for each grade so that their sensitivity meets that grade’s requirements. In C1’s terms, a grade of liquid chemical (numbered 1 to 5) validated with the appropriate procedures can be described as “meeting SEMI Specifications”. The grade and its methods travel together, and a limit without its method is half a claim.

SEMI C10, in the edition approved August 25, 2014, supplies the sensitivity test: every relevant trace contaminant specified in a SEMI process-chemical standard should have a method detection limit, determined by regression of a calibration curve, equal to or less than its specification. Read any parts-per-billion table with that rule. A limit set below the detection limit of the method used cannot be confirmed by that method, and “not detected” means little unless the detection limit sits beside it on the certificate of analysis (CoA).

Validation evidence has a recognizable shape. In a 2016 manufacturer technical bulletin, cold-plasma ICP-MS on undiluted 48% HF recovered 90% to 107% of a 20 ng/L spike across the elements reported. Ask whoever signs the CoA for that kind of evidence, measured in the acid and not in water, with the spike at or below your own limit and on the same mass basis. A 20 ng/L spike is mass per volume; by mass it still sits above the sheet grade 3 metal limit of 0.01 parts per billion.

Particle limits need the same reading. ISO 21501-2:2019 covers calibration and verification of light-scattering liquid-borne particle counters, typically 0.1 to 10 micron, and notes that the measured size is equivalent to calibration particles in pure water because it depends on the refractive index of particle and medium. A particle count in 49% acid is therefore a calibration-equivalent size. For counters that reach smaller, SEMI C77 is the test method for counting efficiency where the minimum detectable size is between 30 nm and 100 nm.

NIST’s 2022 CHIPS strategy (NIST CHIPS 1000) states the buyer’s problem directly. Materials suppliers may differ in measurement accuracy or use their own conformance methods, so fabs may be uncertain about the accuracy of a supplier’s methods, and some form of certification or accreditation would help assure provenance and consistent quality.

How do you read the limits on a semiconductor-grade data sheet?

A concrete sheet shows how far the published numbers go. The technical data sheet (PDF) for semiconductor-grade 49% HF, version 1.0, revised 2026-09-30, lists three grades, each at 48.8 to 49.2 wt% HF. It labels them “Semiconductor grade 1, 2, and 3” and names no SEMI C28 grade, so the numbering is the sheet’s own and no SEMI grade should be read into it. The table calls them sheet grades.

Three things are not on the sheet: typical values, test methods and a SEMI grade. It prints specification limits only, and it states that binding specifications are those on the lot-specific Certificate of Analysis and in the sales contract. Its packaging line reads 235 kg net in new poly drums, with a shelf life of 1 year.

Borrow the C10 rule as a yardstick for the right-hand column, though the sheet is not a SEMI document. Sheet grade 3 holds each listed metal to 0.01 parts per billion, a tenfold step below sheet grade 2, so every method behind a metal result on the CoA needs a detection limit at or below that figure, measured in 49% acid. The sheet grade 3 particle limits start at 0.05 micron, below the typical range ISO 21501-2 describes, so the counter and its calibration belong on the certificate too.

Where is a purity claim measured?

A purity figure also has a place. NIST notes that materials pass touchpoints in production, storage, transport and delivery where contaminants can be generated or added. The International Roadmap for Devices and Systems (IRDS) says metal-ion control starts at the supplier, but that contact with shipping containers and delivery equipment may warrant additional purification, depending on how critical the process step is. The 1993 deposition study made the same point: chemicals that are pure to grade can pick up metal ions through poor handling or storage.

Whatever the container, a low-impurity claim describes the acid where its sample was drawn. Copper shows why the trail matters: in a 1996 Journal of the Electrochemical Society study of silicon in 50:1 HF spiked with 0 to 100 ppb copper, the copper deposited on the wafer rose with the copper in the acid.

On the fab side, SEMI F41 says bulk chemical distribution systems are typically qualified by testing particle levels and trace metals against levels agreed in advance. A clean certificate from the plant starts the record; it does not finish it.

What should the RFQ for semiconductor-grade 49% HF ask for?

1. The document and edition. SEMI C28-0618 for 49% acid; C29 for 4.9% (10:1) acid, PV11 for photovoltaic acid and C23 for buffered oxide etch. Where a sheet names no SEMI grade, write your own limits into the order.

2. Method and detection limit per analyte. For each metal and anion, the method and its detection limit, at or below your limit (the rule SEMI C10 sets for SEMI standards), determined in the acid and not in water, with spike recovery at or below your limit on the same mass basis.

3. The particle basis. The counter, its calibration under ISO 21501-2:2019 and the size channels reported, since grades on one sheet can be specified on different size bases.

4. The sampling point. Where the CoA sample was drawn, and the lot number on every certificate.

5. Lot-to-lot variation. The IRDS states that purity variation can matter to yield as much as absolute purity and that most semiconductor manufacturers apply statistical process control, such as 3-sigma limits, to critical process fluids, yet no industry-wide accepted standard defines that process. SEMI C64’s ship-to-control guidelines are listed as Inactive, so agree the terms in the contract.

6. Binding terms. The lot-specific CoA and the sales contract govern; a data sheet’s limits are where that negotiation starts.

Methodology: standard scopes, editions and statuses come from the SEMI store’s public abstract pages; grade limits from the RawSource technical data sheet, version 1.0. Government and public references: NIST CHIPS 1000 (2022), O-H-795 (1953), MIL-A-24641A (1986), EPA’s 1993 report on hydrogen fluoride, EPA’s AP-42 (1993), the EU reference document for large-volume inorganic chemicals (2007) and PubChem’s hydrogen fluoride record (CID 14917), including HSDB.

Industry and literature sources: the IRDS, ISO 21501-2:2019, peer-reviewed studies of metal deposition from HF solutions on silicon (MRS, 1993; Journal of the Electrochemical Society, 1996 and 2001), and unnamed producer documents: a 49% specification sheet, a 49% safety data sheet and a 2016 ICP-MS technical bulletin.