Judge a second electronic-grade hydrofluoric acid supplier on variation, not limits alone

Your etch bay has run one supplier’s electronic-grade 49% hydrofluoric acid for years, and procurement now wants a second source on the approved list. The candidate’s data sheet matches the incumbent’s limits line for line, which proves less than it looks. Two lots can sit inside the same limits and still differ from each other, and two laboratories can report against the same limit with methods of different sensitivity. A qualification that stops at the sheet sees neither difference.

What should you check before adding a second hydrofluoric acid supplier?

Before approving a second electronic-grade hydrofluoric acid supplier, check three things: that every limit sits inside your own specification, that each test method can detect the impurity at or below its limit, and how far results move from lot to lot. The International Roadmap for Devices and Systems (IRDS, 2024) says purity variation can matter to wafer yield as much as absolute purity.

Confirm you are comparing the same product. EPA’s 1993 Hydrogen Fluoride Study describes electronic-grade HF as usually 49 percent aqueous acid, typically bought at that strength and used at 1 to 10 percent. It also says aqueous HF is produced commercially mainly as a 70 percent solution, with electronic and reagent grades at 5 to 52 percent, so an industrial hydrofluoric acid supplier quoting the stronger acid is quoting a different product. The comparison of 49% and 70% hydrofluoric acid grades covers the two strengths.

What SEMI C28 and SEMI C1 give a qualification

SEMI C28, the Specification and Guide for Hydrofluoric Acid, standardizes requirements and supporting test procedures for the grades of hydrofluoric acid used in the semiconductor industry. It was first published as C28-0699, replacing four separate documents (SEMI C1.8, C7.3, C8.3 and C11.3). The current revision, C28-0618, carries Inactive status on the SEMI C28 standard page but remains valid for use, so a qualification file should name the revision.

SEMI C1, the Guide for the Analysis of Liquid Chemicals, carries the logic underneath. Its public abstract says liquid-chemical purity has been shown to affect device yield, and that each grade (1 to 5) validated with the appropriate procedures can be described as meeting SEMI Specifications. Those procedures are defined per grade, so their sensitivity meets that grade’s requirements. SEMI C10 adds the arithmetic: every relevant trace contaminant specified in a SEMI process-chemical standard should have a method detection limit equal to or less than its specification.

A SEMI claim is therefore two claims at once, a limit and a validated method able to detect at or below it. Ask any candidate that cites SEMI C28 for the grade, the revision and the validation behind each method. A sheet that names no SEMI grade makes no SEMI claim, and its grade numbers are the supplier’s own, not SEMI grade numbers.

Where do the published standards stop?

The IRDS moves the question from the limit to the spread. It states that variation in process-fluid purity can matter as much to wafer yield as absolute purity, and that statistical process control (SPC) on incoming materials reduces variation at the point of supply. It reports that most semiconductor manufacturers run SPC, such as 3-sigma limits, on critical process fluids. These statements cover high-purity process chemicals in general, not HF alone.

In its gases section, the same roadmap reports that no industry-wide accepted standard defines the SPC process, and that a survey of several large manufacturers found that the criteria for “in control” in their bulk and specialty gas purchases vary substantially. SEMI C64 gives statistical guidelines for deriving ship-to-control limits from process data, but the SEMI store lists its edition as Inactive. The IRDS also reports manufacturers adopting ship-to-control standards individually with their suppliers, in a passage on deposition precursors, not acids.

A limit sheet tells you where a lot must fall, not where the lots do fall. The spread you will accept is a figure you agree with each supplier, and no published standard sets it for you, so write it into the qualification plan and the contract.

How was each number measured?

NIST’s measurement strategy for semiconductor manufacturing, NIST CHIPS 1000 (2022), states the method problem plainly: materials suppliers may differ in measurement accuracy or use their own conformance methods, so fabs may be uncertain how accurate a supplier’s methods are, and certification or accreditation would help. Its example is dielectric properties, not acid purity, but the logic carries over. A “less than or equal to” figure is only as good as the method behind it.

For trace metals, ask for validation data as well as the limit. One manufacturer’s 2016 technical bulletin reports spike recoveries of 90% to 107%, across the elements reported, for a 20 ng/L addition to undiluted 48% HF measured by cold-plasma ICP-MS: evidence that a method recovers what is added at trace level. Ask each candidate for the equivalent on its own method, with the detection limit for every metal on its sheet and spike recoveries at or below the limit you buy to.

Particle counts carry the same trap. The scope of ISO 21501-2:2019, which covers calibration of liquid-borne particle counters for pure water and chemicals, notes that measured size is equivalent to calibration particles in pure water, because the reading depends on the refractive index of particle and medium. SEMI C77 tests the counting efficiency of counters whose minimum detectable size is between 30 nm and 100 nm. Before comparing two sheets, ask which counter, which calibration and which size thresholds produced each figure.

What the data sheet prints, and what it leaves out

This grade’s own technical data sheet, for the semiconductor-grade 49% acid (V1.0, revision 2026-09-30), prints specification limits only, with no typical values and no test methods, and a buyer qualifying a source asks for both. It names its own grades 1, 2 and 3 and no SEMI C28 grade, so none should be read into it. Binding values, the sheet says, are those on the lot-specific Certificate of Analysis (CoA) and in the sales contract.

Size thresholds differ even within one sheet. Grades 1 and 2 set particle limits at 0.2, 0.3, 0.5 and 1.0 micron, while grade 3 sets them at 0.05, 0.1, 0.2 and 0.5 micron, and the sheet notes that grade 3 is specified on a different particle size basis. The sheet lists one package, 235 kg net new poly drums, and a shelf life of 1 year.

Qualify the package and the route, not only the acid

The IRDS defines the point of supply for process chemicals as the drum, tote or bulk supply, and the point of delivery as the outlet of final filtration of the chemical distribution unit. Metal-ion control starts at the supplier, it says, but contact with shipping containers and delivery equipment may warrant more purification, depending on how critical the step is.

Materials of construction are part of that route. A DOE national-laboratory assessment of fluoropolymer use states that few materials withstand semiconductor processing without releasing leachable or extractable constituents into process solutions, and a 1993 study of metal deposition from HF solutions adds that process chemicals pure according to grade can easily pick up metal ions through improper handling or storage.

So qualify the second source as delivered, in the package and on the route its production lots will use. Treat a later change of package or filling site as a change of the point of supply, and requalify.

What to request from each candidate

1. Any SEMI C28 grade and revision claimed, with its method validation and detection limits.

2. Typical values and test methods for every parameter.

3. Consecutive production-lot results for assay and your critical metals, enough to set the 3-sigma limits the IRDS describes.

4. Counter, calibration and size thresholds behind every particle figure.

5. Qualification lots in the package and from the filling point production will use.

6. Notification terms when a lot falls outside agreed ship-to-control limits (SEMI C64 gives guidelines for deriving them).

7. A lot-specific CoA with every shipment, checked on receipt.

The limits for each grade are in the spec table on the Hydrofluoric Acid 49% Semiconductor Grade page and in the technical data sheet (PDF); the lot CoA and the sales contract hold the binding values.

Methodology: standards content is from SEMI’s public abstract and scope pages for C1, C10, C28, C64 and C77 and the ISO 21501-2:2019 scope; roadmap statements from the IRDS; method statements from NIST CHIPS 1000; leachables from DOE’s fluoropolymer assessment (OSTI 2370520); strength data from EPA’s 1993 Hydrogen Fluoride Study; grade limits from the RawSource technical data sheet V1.0 (revision 2026-09-30).