While there are existing methods including the gravimetric method specified in JIS M 8214 for analyzing silicon, which is a substance contained in steel and other materials at a certain level or higher, these methods have problems regarding the complex analysis procedures and the accuracy of analysis. In addition, it is difficult to measure the silicon content in the low concentration range with ICP-OES or ICP-MS due to the elution of silicon from the device or mass interference. It is therefore accepted in general to adopt the method based on “silicon tetrafluoride vaporization separation-molybdenum blue spectrophotometry,” which is used in JIS methods in Japan, where the silicon contained in the sample is separated as a gas and collected in the absorbing solution to be measured. However, this method also has issues in the separation time (30 minutes) and measurement sensitivity (5 μg/g). We therefore developed a new method using continuous flow analysis equipment (CFA method).

In the equipment we developed, hydrofluoric acid is added to the sample that is made into a solution, which is then set into the auto-sampler. Sulfuric acid solution is mixed with the sample, which is segmented by several segment gases inside the flow channel, and makes the trace amount of silicon contained in the sample react with hydrofluoric acid to form silicon tetrafluoride gas, which then moves from the liquid phase to the gas phase along the reaction coil. The gas that has moved is absorbed into boric acid solution to conduct coloring quantification with molybdenum blue. The sample and the cleaning water are alternately introduced into the tube so that each sample undergoes reaction independently. Separation is possible in 90 seconds per specimen.

Since all the reactions complete within the sealed tube, it is possible to conduct high-sensitivity and high-accuracy analysis of silicon without being affected by external factors. Furthermore, analysis can be conducted without the effects of matrix as the silicon is separated from the sample by vaporization before measurement, and thus it is possible to take measurements in high matrix samples and concentrated samples.

The quantification lower limit value for the analysis method that was calculated using standard deviation based on repeated blank measurements was 0.003 mg/L. Addition and collection tests using an identical calibration curve, which were conducted by adding silicon to solutions containing Cu, Mg, Fe, Pb, and Ti at high concentrations, resulted in a favorable collection rate. When we used this method to quantify silicon in tantalum oxide (Ta₂O₅), we were able to quantify the silicon content at 0.29 μg/g level with relative standard deviation of 2.9%. In addition, favorable results were again attained when the method was applied to organic solvents and samples of certified substances such as brass and iron.

While substances used in semiconductors, gallium nitride and other gallium compounds that are expected to be used in the next generation semiconductors, are affected by silicon and other metallic impurities, there used to be no way of measuring silicon in low concentrations in these materials. Our results indicate the possibility of developing a simple method for measuring low-concentration silicon in semiconductors and semiconductor components by applying this method.