A new study has shown that the bright and saturated colours produced by some dye molecules can fall outside of the 8-bit sRGB colour space used by digital cameras, potentially leading to misinterpretations in automated analyses.1
To capture the wealth of information available by watching a reaction, scientists are increasingly using smartphones and digital cameras in their labs. That includes Marc Reid’s group at the University of Strathclyde, UK, which is developing video-analysis software to convert everyday visual observations into quantitative kinetic data. ‘Our research team is trying to get away from the frustration of seeing [chemical phenomena] but not being able to turn them into numbers,’ explains Reid. ‘Instead of using eyes, we use the next thing – cameras. You can take these “digital eyes” and turn curiosities into quantifiable measurements.’
Previously, Reid and his colleagues had found that the sRGB (standard red green blue) gamut used by many digital cameras can be inadequate for reaction monitoring.2 ‘One of our video-based measurements led to a plot with a funny kink in it that we’d never seen before,’ says Reid. ‘The kink was identifiable because we knew the chemistry followed simple first-order kinetics, but the camera wasn’t telling that story.’
‘A camera does not record colour the way your eye sees it’, continues Reid. ‘It stores colour as three numbers (red, green and blue), each running from 0 to 255.’ These three numbers allow digital cameras to record 16,777,216 colours, however, even this number of colours can be insufficient when studying reactions. ‘Very vivid, very saturated colours fall outside the ringfenced set of colours a camera can encode. When that happens, the camera does not warn you. It simply substitutes the closest colour it is able to store.’
Now, Reid and his team have studied the well-known hydroxylation reaction of crystal violet to better understand how gamut clipping can affect scientific analysis. Using a commercial camcorder, they found that the colour of the reaction mixture remained outside the sRGB colour gamut for the first 80 seconds of the 1000-second experiment and was therefore recorded incorrectly, even at a low starting concentration of 0.05mM (20.4mg/l). Fitting an exponential decay curve to the full data set, including the clipped data, then systematically underestimated the rate constant for the reaction.
A camera does not record colour the way your eye sees it
‘This [finding] is extremely important nowadays with the rise of self-driving labs, where cameras are abundantly used to check reactions,’ comments Vittorio Saggiomo, an expert in applying modern technologies to research challenges from Wageningen University and Research in the Netherlands. ‘The magnitude of the error is huge; fitting exponential decay models to camera data can underestimate reaction rates by nearly an order of magnitude.’
Reid and his co-workers have developed a flowchart to help others diagnose and mitigate for colour gamut clipping. ‘The diagnostic is trivially easy, and I would like every reader to take it away’, adds Reid. ’Look at your sample, then look at it on the camera screen. If it looks more vivid in the flask than it does on the display, you may be clipping. Capture [your data] in RAW, or use a camera supporting a wider gamut such as DCI-P3 or Adobe RGB.’
Steve Christie, a chemist at Loughborough University in the UK whose group uses modern techniques and technologies to advance synthesis, says the flowchart ‘offers a logical and pragmatic way forward, whilst being cautious of the clipping issue’. ‘The main takeaway is that researchers need to be careful using colour monitoring, and make sure the colours they may be looking at fall within the gamut,’ he adds.
