Food-Grade Silica COA: What the 99 Percent Figure Hides

Created on 11.20
A purchasing manager in Izmir sent me a COA in April. One page, clean layout, silicon dioxide 99.2 percent. He wanted to know why his premix line had started blocking at the hopper every second shift.
I asked for the rest of the sheet. He sent the same page again. That was the whole document. Assay, moisture, a lot number, a signature.
This happens often enough that I stopped being surprised by it. Assay is the easiest number to test and the easiest number to print. It is also the number that tells you almost nothing about whether the powder will flow, disperse, or sit quietly in a bag for two years.

What an assay result actually proves

For an amorphous synthetic silica sold as E551, silicon dioxide content is measured on the dried sample, and some methods add an ignition step. Our W38A food grade is specified at 98.0 percent minimum on a dry basis. That limit is real and we hold it lot after lot. But it is a ceiling test, not a fingerprint.
Two lots can both read 99.1 percent and behave nothing alike. One disperses in ten seconds in a ribbon blender. The other leaves fish eyes in the finished sauce and the customer blames the recipe. The assay cannot see the difference, because the difference sits in particle size, pore volume, surface chemistry and how much water the powder is still carrying when it leaves the silo.
Food-grade silica COA sheet beside two samples of precipitated silica powder with identical assay but different particle properties

Loss on drying and loss on ignition are two different tests

Loss on drying is run at 105 degrees C. It measures free moisture, the water that comes off in a normal oven. Loss on ignition is run much hotter, usually around 1,000 degrees C, and it burns off chemically bound water and residual organics as well.
For W38A the two limits are 7.0 percent and 8.5 percent. If a supplier only quotes one of them, you do not know which question was answered. A powder at 6.5 percent moisture passes the drying test and can still be a sticky mess in a humid filling room, because the number was taken in a lab at 20 degrees C and 40 percent RH, not on your mezzanine in August.
Technician measuring loss on drying of food-grade silicon dioxide in a laboratory oven and muffle furnace

Soluble salts: the limit nobody reads until there is a taste problem

Precipitated silica is made by reacting sodium silicate with acid. The reaction leaves sodium sulfate in the slurry, and the washing step removes it. Wash too little, and the salt stays in the dried product. Our ceiling is 2.0 percent, measured as soluble salts.
In a savory seasoning this shows up as an odd salty back note that the flavour house cannot explain. In a beverage powder it shifts the taste profile. In the plant it shows up as scale on the drier walls and corrosion on mild steel where the powder settles. None of this appears on an assay-only COA.

The 325-mesh residue is what your customer sees

We hold 325-mesh residue at 0.1 percent maximum. That is roughly 45 microns. Above that, particles stop behaving like a powder in a liquid and start behaving like grit.
Where it matters: a white sauce, a tablet face, a clear beverage, a cosmetic cream. One oversize particle in a white matrix is visible. This is also why the milling step is not a commodity operation. We run air-jet milling on the food grades and hold patents on the process, mainly because a mechanical mill heats the product and broadens the tail of the distribution, and a broad tail is exactly what shows up as specks.
325-mesh sieve test of air-jet milled precipitated silica powder showing residue and particle size control

Why the pH window is narrow

W38A is specified at pH 6.5 to 7.5 in a water slurry. That window is not decoration. Silica outside it can react with flavour compounds, shift the colour of a natural pigment, or change how a preservative behaves in a dry mix that later gets wet.
Compare that with fumed silica, which sits at pH 3.7 to 4.5. Fumed is the right material for a silicone sealant or a coating, and the wrong material for a food premix, even though both are called silicon dioxide on the label. If your supplier offers to substitute one for the other on price, the conversation is over.

Lead, arsenic and the metals list

Our food grade holds lead at 1 ppm maximum, arsenic at 2 ppm, heavy metals as lead at 10 ppm. These are tight numbers and they depend on the raw material, not just the process. Sand with a different geological origin brings a different metals profile, and no amount of washing fixes a bad source.
Ask which method was used. ICP-MS gives you a clean detection limit. Older colorimetric methods can pass a lot that a modern instrument would fail. The number on the sheet is only as good as the method behind it.

GB 25576-2020 and ISO 3262-19 are not the same document

GB 25576-2020 is the Chinese national standard for food additive silicon dioxide. ISO 3262-19 is the international specification for precipitated silica. They overlap on the basics and differ in test details, sample preparation and a few limits.
If you sell into the Chinese domestic market, GB is the one your customer will check. If you export to Europe or quote to a multinational, the ISO numbers are what the QA team will compare against. We print both columns on the same SPEC sheet, which sounds like a small thing until you sit in an audit and have to explain why two documents give slightly different ignition losses for the same lot.

Different jobs need different silicas, not different prices

The anti-caking family is not one product. W58A and W58B are built for low-dose work, typically 0.5 to 1.0 percent in a powder mix, where you want free flow without changing bulk density or mouthfeel. A325 and A200 are coarser and denser, and they suit spice blends and instant powders where the blend has to stay uniform through a packaging line.
WPS200 and WPS300 are spray-dried carriers. They hold liquid flavours and oils on a porous structure, and the loading rate depends on pore volume, not on assay. ZOILSIL-G300 goes further in the other direction, with BET surface area of 500 to 750 square metres per gram and a D50 of 15 to 25 microns, which is why it works in edible oil bleaching and beer stabilisation rather than in a dry cake mix.
Put the wrong one in and the line tells you within a shift.

What belongs in your purchase specification

Start with the standard you are actually selling against, then add the items that break your process:
  • Silicon dioxide content, with the basis and the method stated.
  • Loss on drying at 105 degrees C, and loss on ignition at 1,000 degrees C, as separate lines.
  • Soluble salts, with a ceiling.
  • 325-mesh residue, with a ceiling, not the word "fine".
  • pH of the water slurry, with the concentration stated.
  • Lead, arsenic and heavy metals, with the test method.
  • Bulk density and D50, if the material goes through a hopper or a dosing screw.
  • Packaging, including the liner, because a 25 kg bag and a 500 kg jumbo behave differently in a humid warehouse.

A trial that costs one shift

Take 25 kg of the candidate and 25 kg of your current material. Run both at 0.5 percent and 1.0 percent in the same base. Fill twenty bags from each and leave ten of them open in a room at 80 percent relative humidity for a week.
Then check four things: whether the powder still flows through a funnel, whether the bags have set into lumps, whether the finished mix shows specks on a white plate, and whether the taste panel can pick the difference. That is half a day of work and it answers more than three pages of brochure text.

What ships with our lots

Every lot leaves with a COA that lists the test method next to the result, a TDS, an MSDS, and an elemental impurities statement. Food and feed grades carry the production licence and the registration documents for the destination market. Sealed food and pharma grades carry a 24-month shelf life, and we ask you to re-test moisture after that rather than assume the number still holds.
If your incoming inspection is one assay titration, the COA will match it every time and you will still get the occasional blocked hopper. The sheet is there to stop that.

FAQ

Q1. What does silicon dioxide 99% mean on a food-grade silica COA?
A: It is the assay result on the dried or ignited sample. It confirms the main component. It says nothing about moisture, salts, particle size or metals, which are usually the numbers that decide whether the powder works in your line.
Q2. Why do I need both loss on drying and loss on ignition?
A: They measure different water. Loss on drying at 105 degrees C captures free moisture. Loss on ignition at around 1,000 degrees C also removes chemically bound water and residual organics. A lot can pass one and fail the other.
Q3. What is a normal soluble salt limit for food-grade silica?
A: Our W38A holds 2.0 percent maximum. The number comes from the washing step after precipitation. High salts show up as taste defects, scale on equipment and corrosion, so it belongs on the spec sheet.
Q4. Why does 325-mesh residue matter if the powder looks fine?
A: Because the eye sees the tail, not the average. A 0.1 percent residue limit at 325 mesh keeps oversize particles out of white sauces, clear drinks and tablet faces where a single speck is visible.
Q5. Can I use fumed silica instead of precipitated silica in food?
A: Not as a drop-in. Fumed silica sits at pH 3.7 to 4.5, has a different surface and a much lower bulk density. It is the right material for sealants and coatings. Food and feed grades are precipitated and specified in the 6.5 to 7.5 range.
Q6. What is the difference between GB 25576-2020 and ISO 3262-19?
A: GB 25576-2020 is the Chinese food additive standard. ISO 3262-19 is the international precipitated silica specification. The limits overlap and the test details differ. Export customers usually want the ISO numbers, domestic customers want GB.
Q7. How much silica should I dose as an anti-caking agent?
A: Most powder mixes run between 0.5 and 1.0 percent. Above 2 percent you start changing bulk density and mouthfeel without gaining much flow. The right figure depends on the base powder, the humidity and the storage time.
Q8. Why did my silica work last year and fail this year?
A: Usually a raw material or drying change, not a formula change. Ask for the loss on drying, the soluble salts and the D50 for both lots side by side. The assay will look the same and the other three will not.
Q9. Does particle size affect how much silica I need?
A: Yes. A finer, higher surface area material covers more particle surface per gram, so you can often dose less. A coarser material flows better in the bag and doses more evenly through a screw. The two effects pull in opposite directions.
Q10. What is the shelf life of food-grade silica?
A: 24 months in the sealed original bag, stored dry and cool. After that, re-test moisture before use. Silica does not spoil, but the bag can pick up water and the powder can cake.
Q11. Is your silica FDA approved?
A: We hold an FDA attestation and ship from an FDA-registered facility. That is the accurate wording. Amorphous silica is generally recognised as safe for its intended food use, and the specific limits sit in the relevant regulation and standard.
Q12. Which documents should arrive with each lot?
A: A lot-specific COA with test methods, a TDS, an MSDS and an elemental impurities statement. Food and feed shipments also need the production licence and the additive registration for the destination market.
Contact
Leave your information and we will contact you.
Phone
WeChat
WhatsApp