I am going to build a device to detect the physical state of water. I came up with an idea, that the easiest solution is to measure the intensity of the reflected wave, which length is 4150nm. According to https://www.nature.com/nature/journal/v439/n7073/fig_tab/nature04415_F1.html there is a significant difference in absorption of liquid and solid water. The device is created for ordinary applications, so it doesn’t need to be as sensitive as ones for laboratory use. I found that there is a possibility to buy a thermopile with 4.3 μm band-pass filter, but there is a lack of 3050nm detectors. The device should also have the second reference detector, but I don’t know which wavelength will be the most desirable. Another point is selection of IR emitter that will match the detectors, I am not sure which one should I buy.
Or maybe there is a completely different solution that is easier/better than presented above?
IR bandpass filters with various center wavelengths are either available from stock or can be made according to customer specifications. Specialized detector manufacturers have a number of band pass filters for popular gas measurement applications on stock, some might be useful for your application, see e.g. a 3.09 µm filter from Infratec https://www.infratec-infrared.com/sensor-division/products/ir-filter.html
Not particularly. But both measurements are implementing IR absorption photometers, using the same IR sources, detectors, filters, lenses, windows etc.