After reviewing research papers on the topic, I retract the statement that the particle counts are synthesized. A light-scattering type nephelometer is capable of discriminating particle sizes, typically into classes or bins of a certain width. The conversion from a count (particles/dL) to a gravimetric weight (μg/m^3) is impacted by (assumed) density.
I would advise against taking the particle counts too literally, as it isn’t actually counting individual particles.
As with all sensors, there are limits. For the Plantower PMS5003 used in the PA-II Classic, the lower size limit is 0.3μm, accuracy below 0.5μm drops precipitously and there is noise from particles below those limits. (See the 3rd paper below for a study on sub-1μm particle measurement with this sensor.)
The sensor outputs summations of diameters at the level and higher (i.e., >0.3μm, >0.5μm, >1.0μm, etc.). A less-than reading is computed by subtracting the higher bound from the lower and should be expressed as ‘0.3μm >= x >= Yμm’ to accurately portray the available data.
To answer your last question – The hardware is capable of size discrimination and adjustments can be developed & applied to obtain meaningful data, but it can’t tell us density. This is the limitation of nephelometer technology. They work best when the optical properties of the material under test are known. With air pollution, the components are different depending on what your pollution sources are, and its very difficult to come up with general rules that apply well everywhere. And yes, if the adjustment factors are off, you get useless data; GIGO.
If you want more details on how the hardware works and how the adjustment factors are developed, I’ve linked some research papers below that may be of interest.
This paper discusses EPA’s early efforts to develop a nationwide correction for PM2.5 from PurpleAir sensors. The paper is a good primer on the difficulties of calibrating air quality sensors, and dealing with the limitations of the Plantower sensors compared to Federal spec air quality measuring equipment (note that the current adjustment is vastly different than the one presented here). EPA evaluated particle counts as a basis for calculating PM2.5 but discarded it due to complex interactions with humidity:
Lance Wallace, who frequents this board, published a study in 2022 that uses the particle counts to derive a PM2.5 measurement with better limits-of-detection than the Plantower-derived PM2.5 output:
There’s certainly opportunities for research on PurpleAir performance with ultrafine particles. This paper, while a bit more technical, discusses sensor performance (and its quirks) with small particle sizes:
https://doi.org/10.1080/02786826.2019.1696015
For further reading, the board maintains a list of research papers on PurpleAir sensors which can be found here:
Hope this helps!