SciBar at The Vat and Fiddle: A Naïve Person's View of Chemical (and Other) Measurement

What does an analytical chemist actually do? Prior to coming to Nottingham, Quentin had a varied career – nuclear reactor research, oceanography, oil refinery quality control, making artificial air pollution for environmental health and working on imaging and spectroscopy systems. Basically, it all comes down to measuring stuff. Sometimes this is stuff that other people would rather not think about. For example, more than 33kg of cocaine and its metabolites go through the Nottingham sewage network every year and end up in the River Trent.

How do they measure stuff? There is the atomic absorption system. There's the liquid chromatography system. Then there is an infrared spectrometer. There are things that use plasma to look at lots of trace elements simultaneously. There are even things that can tell the difference between an omnivore and a herbivore. There is also the visual tuberator – a box with a magnifying glass stuck on the outside. While it looks incredibly simple, it covers a lot of complicated chemistry. In short, there are many, many techniques.

So what? Well, we've learnt a number of things from all this measurement:

• Calcium levels are higher in skimmed milk compared to full fat milk (this is because milk is an emulsion and the calcium is in the "watery" bits)
• Potassium levels are higher in premium beer (this is because cheaper brands contain more water and the potassium comes from the malt and the hops)
• Selling water is a good business plan
• There are drug precursors in nutmeg (a good chemist can make drugs from nutmeg oil)
• There is a mineral water subculture that is all kinds of crazy (water with stabilised oxygen supplements?!)
• Branded, over-the-counter drugs are pretty much a waste of money (own-label Ibuprofen, for example, is just as good)

But does it really matter? Yes, in 2007, 100 people died in Panama from taking a cough medicine. The government thought that they had put "99.5% glycerine" into 260,000 bottles of medicine. It turned out that they had put diethylene glycol into it instead. This kills you by renal failure – not a nice way to go. This same mistake has happened twelve times globally and, in 1985, it lead to the destruction of 30 million litres of wine.

But what about that naivety in the title of the talk? Well, there's bias, certainty and uncertainty, the boundary between something and nothing, and the degree of confidence in the estimate of certainty. All of this came to a head in a murder trial from 1999. A man was executed for the murder of a nine-year-old girl. The murder weapon was never found so the key piece of evidence was analysis of the bullet. It was analysed using ICP-OES to get a trace element profile. An expert witness with 21 years of experience said that this analysis could determine if a bullet came from a particular box of ammunition. Mathematically, the likelihood that two .22 bullets come from the same batch in a year is 0.0000025%.

So, the bullet was analysed and was matched with bullets found in the suspect's home. But the science was never there – it's not possible to match a bullet to bullets from the same box. Lead smelting experts subsequently showed that the science was flawed. The National Academy of Science now says that this evidence is no longer admissible in court. So, despite having 21 years of experience and uncertainty to two orders of magnitude, the wrong outcome was reached. The expert had been taught that something was true and then propagated this belief. There was a false prognosis and assumptions.

We need to try and look for the unquestioned assumptions. For example, the modern balance. These things look pretty straightforward, essentially a weighing scale and from the digital display on the front, you'd think that they were accurate to around plus or minus 0.0002g. However, during a break in the talk for the audience to grab a beer, Quentin invites us all up to come and do some science and test the balance. It turns out that uncertainty scales with the mass – as the mass increases, so does the error.

Why is this interesting? Exponential dispersion models are used for inter-laboratory comparisons for certifications of labs. William Horwitz looked at thousands of these and they appeared to be linked on the Horwitz curve – concentration against standard deviation is a straight line. But this isn't restricted to physical science. It applies to other things too, such as crime.

This can all be traced back to a French mathematician Poisson and his eponymous distribution. He began by contemplating juries in post-revolutionary France to determine whether or not they made a good decision. His work and that of Gauss introduced fluctuation scaling – constant uncertainty. This was built upon by Maurice Tweedie and Bert Jorgenson to produce exponential scaling models.

All this has shown us that everything has fundamental variability. Understanding this allows us to know things such as how many photons hit a device, how to calibrate the response of a police force and how to understand chemical measurement at a very abstract level. However, this does all mean that every time we make a conclusion, we have to ask ourselves, "Do you know how certain you are?"

SciBar returns to The Vat and Fiddle on Wednesday 30 November, where Dr Gareth Cave from University of Nottingham will talk about Nanoparticles: From Therapeutic Vectors to Chicken in a Basket

SciBar MeetUp Page