If you want a color selective photodiode, using LEDs perhaps isn't the best approach.
Spectrum, as discussed in the comments, is different, added to that a wavelength transmission of the chip encapsulant, as well as the primary optic. On top of that, the primary optic doesn't have AR coating, which can start to mess things up.
Then you have temperature dependencies, silicon impurities, and all sorts of effects that will introduce unknown characteristics besides the basic physics of the junction. You may spend time to find an LED that works for your need, but since they are not designed to work in this mode, the manufacturer can at any time modify the characteristics, while still operating as the datasheet, it may change the behavior as a photodiode.
On top of that, the LED chip-making process is more like a kitchen recipe than a set process. They will be differences between wafers, batches, on the silicon level, but also encapsulant, optics, and so forth.
Using a device in a way it wasn't meant to be used is always a bit risky. Then also it depends if you intend to make 1 device or 5'000.
The correct approach rather would be to use a photodiode and to place optical passband filters on top. Those are quite cheap, can be tuned to specific wavelengths, have known cut-off and thermal behavior and you will have a predictable, and stable result.
More detail on filters:
Optical Filters can be made by many different means:
- Using material properties like different types of glass or polymers that all have specific spectral transmission properties.
- Chemically colored glass or polymers.
- Interferential by deposition of thin-film layers.
It is possible to create a filter by just stacking a few different polymer or glass materials to achieve the desired spectral sensitivity on the photodiode.
Example of colored glass form this Japanese company
More detail on low-cost filters:
When one searches for optical filters, filters from Thorlabs and such tends to be the only results google provides. Thorlabs and the likes make high-grade optics that are used for research, military, and special scientific devices, and the cost is high.
However, there are many companies providing much lower cost optics like Schott in Switzerland, or many many companies in China. The optics may not be as high grade as Thorlabs, but they are way-way sufficient for this type of application.
For instance, we had a filter custom made which is RG1000 with additional interferential layers and we pay a few $ a piece for a 20x20mm filter.
The cost is directly dependent on the size, for a photodiode a 4x4mm filter is probably enough.
One may not forget that interferential filters are everywhere, at a low cost. Most camera lenses have one. Another example is the reflector cups of the old Halogen Lamps, which are glass coated with an interference filter called Dichroic, letting the IR through to reduce heat.
Also, a simple colored piece of plastic, like a paper holder, can work.
I worked in a company that was stacking pieces of different types of glass to achieve similar things.
An optics trade show is a great place to go to find a lot of suppliers.
Depending on the optical needs, it is also possible to use the reflectance, rather than the transmission of the material by placing the photo-diode facing a reflector or in an integration chamber. Then you can paint the surface or use metals and other materials' reflectance properties. You need to be careful of oxidation though.
You can also use phosphor up or down-conversion as well, this is particularly useful if you are trying to detect outside the photodiode sensitivity region.