Alum and bisulfite: specify the substance, not the trade word

“Alum” is not one substance

21 CFR 182.90 lists the substances migrating to food from paper and paperboard products that are generally recognized as safe for their intended use. The first entry reads: “Alum (double sulfate of aluminum and ammonium potassium, or sodium).” That is not aluminum sulfate. It is the ammonium, potassium or sodium double salt—a different compound, with a different formula and a different CAS number from the Al₂(SO₄)₃ that a mill orders as papermaker’s alum. The list goes on to name sodium aluminate, aluminum hydroxide, sodium hydrosulfite, sodium silicate and talc separately, each by its own name, because the rule is precise where the trade vocabulary is not.

This is the ordinary failure mode of the word. Write the CAS number and the hydrate on the purchase document, not the nickname, and check that the substance named in a specification is the substance being quoted against it. Note also what appearing on this list does and does not mean: § 182.90 addresses substances migrating into food from packaging. It is not an approval of any supplier’s material, and it is not a purity specification—as the next section shows, the rule that governs the list says so itself.

The percentage has a basis and the basis has a hydrate

The aggregated public record for aluminum sulfate carries “Papermaker’s Alum,” “Filter Alum” and “Pickle Alum” among its synonyms for CAS 10043-01-3, and records that the substance “is also obtained as an 18-hydrate Al₂(SO₄)₃·18H₂O,” for which it gives a molecular weight of 666.43 against 342.15 for the anhydrous salt. It then adds the sentence that ends most arguments about a number: the “article of commerce usually contains 5 to 10% less water than theory.” So a percentage can be expressed on an anhydrous basis, on a hydrated basis, or as an as-delivered solution strength, and those are three different figures describing the same drum.

Two more entries in the same record are worth carrying into a specification conversation, with their limits stated. “On long boiling of aqueous solution, insoluble basic salt precipitates”—basicity and insolubles are real, measurable attributes, not merchant folklore. And “one part of technical aluminum sulfate reduces the alkalinity as calcium carbonate by 0.55 parts,” a figure quoted with no temperature, no concentration and no method attached, which is exactly the sort of number that should be traced to its originating record before anyone designs against it. The same record also contains a consumption breakdown labelled 1981. A dated figure is not a current one.

Iron comes from the raw material, not from a polish

EPA’s supply-chain profile describes the manufacturing route: aluminum sulfate is produced by reacting aluminum hydroxide with sulfuric acid, and “the process may begin with ground bauxite in place of aluminum hydroxide.” Liquid alum, the more commonly sold form, is prepared by diluting that solution; evaporating and crystallising it gives the dry product. The aggregated chemical record completes the picture with the reason a low-iron product exists at all: the iron-free grade, quoted there at less than 0.005% ferric oxide maximum, is made “using pure alumina trihydrate in place of bauxite or clay.”

The consequence for a purchase is straightforward. Iron content is set upstream, by the feedstock, and a buyer asking for it has to ask for a limit with a method and units rather than for a word. The consensus commodity standard here is AWWA B403, Aluminum Sulfate—Liquid, Ground, or Lump, which EPA’s profile cites; it is sold by AWWA and was not read for this guide, so nothing here paraphrases its requirements. Where a specification invokes it, obtain the edition invoked and cite the clauses by number. Citing a standard by designation without having read it commits both parties to test methods and sampling frequencies nobody has looked at.

A reducing agent loses the thing you bought it for

The aggregated record for sodium bisulfite states the stability problem in one line: “on exposure to air it loses some sulfur dioxide and is gradually oxidized to sulfate.” It has no boiling point—it decomposes—and it is recorded as “corrosive to aluminum.” Its listed uses read like a mill inventory: bleaching groundwood, wood pulp digestion, and use as an antichlor in paper making. Each of those is a different job with a different endpoint.

Two purchasing consequences follow, and only two. First, reducing capacity is a measurement with a date on it, in the same way a hypochlorite strength is, so a delivery record has to carry a fill date, a storage arrangement and an expected turnover rate against tank volume, not merely a grade. Second, a compatibility note like “corrosive to aluminum” belongs in the transfer-interface field of the purchase record, where the wetted materials are listed—not in an application recommendation. What this guide will not do is set a dose, a ratio or an endpoint. Residual control as a specification problem is treated in this library’s water and wastewater treatment collection, under “Dechlorination is two problems wearing one word,” and the decay of an oxidant strength under “A hypochlorite percentage is a measurement with a date on it.” Neither is repeated here, and neither supplies a dose either.

Purity with no number attached is not a specification

21 CFR 186.1 governs the list quoted at the top of this guide, and its wording on purity is the most useful sentence in it. Indirect food ingredients are GRAS “providing they comply with the purity specifications listed in this part or, in the absence of purity specifications, are of a purity suitable for their intended use in accordance with § 170.30(h)(1) of this chapter,” used “at a level no higher than reasonably required to achieve its intended technical effect in the food-contact article.” The rule also states outright that part 186 does “not authorize direct addition of any food ingredient to a food.”

Where a mill makes a food-contact grade, “suitable for its intended use” is a determination the mill and its own specialist make, and the only way to buy against it is to write the attributes down: each one with a limit, a method, units and a reporting limit. Alliance sends a Certificate of Analysis when a customer asks, at no charge; name in the purchase order which document is meant, because a typical or sample certificate describes what the product has generally looked like while a lot certificate reports results for the material being shipped. A sample COA is not a lot certificate. Nothing here states that a particular Alliance product is suitable, approved or recommended for any application, and independent specialist review of this collection is pending.

• Which compound is actually being quoted—CAS number, hydrate, and form as delivered?
• Is the stated percentage on an anhydrous basis, a hydrated basis, or as-delivered solution strength?
• Which attributes does your specification control—iron, insolubles, basicity, free acid—each with a limit, a method and units?
• For a reducing agent: what is the fill date, the storage arrangement, and the turnover rate against tank volume?

Requirement → evidence → decision boundary

An editorial checklist for your review—not a table of product specifications.

Revision record

1.0 · September 12, 2026 — Source-based editorial edition. No original testing or product qualification is claimed.

For pulp, paper and paperboard mills and their purchasing teams specifying bulk caustic, bleaching chemistry, coagulants and reducing agents. This collection does not select a bleaching sequence, set a dose or an application rate, design a stock or water circuit, or make a discharge, permit or waste determination.