From a $50,000 GFAAS to the Lead Lottery.

The reason I bought it was the same reason most people read Lead Safe Mama: I wanted to know what was in the food, specifically. XRF is the wrong tool for that. ICP-MS and GFAAS are the right instruments, but the prep method matters more than the instrument.

The standard prep is wet-acid digestion. You take a food sample, transfer it to a trace-metal-grade vessel, add concentrated nitric acid (around 1,000 dollars per liter for the pharmaceutical-grade version), heat it to break the sample down to a clear digest, then analyze the digest. Sounds clean. It is not. Every step has a contamination vector.

The vessels have to be acid-washed to a level where a fingerprint is a problem. The acid itself, even pharmaceutical grade, has trace lead. The lab air has trace dust, and lead dust is everywhere if your house is older than 1978. A single speck of dust on a 28-gram digest can shift the result from nondetect to a "positive" reading at sub-ppb levels.

And the sample is not homogeneous. A single slice of bread does not have one lead concentration. The lead is concentrated in the crumb edges and the crust, not evenly distributed. To get a defensible number you have to homogenize the whole loaf, weigh out a representative sub-sample, and digest that. Most home-bench operators are not doing this. I was not, at first. The within-bag variation between two pinches from the same loaf was bigger than I expected.

The conclusion you reach is uncomfortable: brand consistency is a marketing claim, not a chemistry one. The number you publish for one bag tells you about that bag. The next bag from the same shelf can be very different. The lab method is fine. The single-sample sampling plan is the structural problem.

Burn the sample. Capture the effluent gas. Analyze the captured stream. The lead does not go anywhere; it ends up in the trapped condensate or the absorption-train solution, where you can quantify it cleanly. No nitric acid. No acid-washed vessels. No fingerprint problem. You still need to homogenize the sample before you burn it, and you still need a clean burn chamber, but the contamination vectors are far fewer than the wet-prep version.

This is a known method (Wickbold-style combustion, oxygen flask combustion, catalytic combustion variants). It is not what the standard contract-lab pipeline uses for food samples, partly because contract labs are set up for ICP-MS-with-acid-digestion and partly because the sample throughput on combustion rigs is lower. But for consumer-product screening at the price point we want to hit (lab fee under 100 dollars per sample, or under 20 dollars in-house when you have the rig running), combustion is the more defensible approach.

Plan: build the next-generation Lead Lottery test on combustion-prep, not acid-prep. Same instruments downstream (ICP-MS or GFAAS for the quantitation), different front end. Less contamination, less acid cost, less time per sample, and more confidence in the resulting number.