Precise calibration of optical fiber sensor for ammonia sensing using multivariate analysis

Detection in chemical sensing which needs to be carried out in a specific controlled environment, becomes complex in multivariate environment. This complication is caused by chemical interference, sensor degradation or drifting of the signals with time. A minute drifting or overlapping of the signals affects the calibration, especially in the detection of sub-ppm level of concentration of any chemical species. The presence of other compounds can well interfere providing false positive readings, deterring calibration of the system in precise quantification of any compound. This problem is known to also happen in our optical fiber sensor for the detection of ammonia. A clad-modified polymer optical fiber sensor for ammonia detection is explored in this work where oxazine 170 per chlorate dye is used as a recognition element to detect ammonia. The sensor was tested in water media and the sensitivity of the sensor we found was 0.0006 ppm-1cm-2. However, the lower sensitivity causes significant overlaps in between signals corresponding to different concentrations. To resolve this problem, multivariate analysis method, such as principal component analysis (PCA) was explored to interpret the datasets for precision of measurement and classification of each concentration. PCA generates unique regression curve which represents each concentration of ammonia considering principle components. The significance of this research lies in its versatility dealing with the existing challenge of calibration of sub-ppm level measurement of any volatile compound, such as ammonia.