You're getting into the limits of the MIM/MOM range there. I don't think there are going to be commercial sources for that size capacitor (Digi-Key, Newark, etc...). You'll likely need to contact a company like Smoltek.

When you are trying to get low capacitance values with any accuracy, you need to have good control over your parasitic parameters and the best way to do that is to start small and stay small. Parasitic capacitance at the chip scale are in the range you're looking for and you can design a circuit at that scale with that in mind. As soon as you throw the circuit in a lead frame, your parasitics are jumping up an order of magnitude or more and then if you put it on a PCB your looking at another order of magnitude (or two, or three). Most of the commercial suppliers are targeting PCB scale manufacturing and their stock will reflect that. You'll probably need to find a supplier that is more focused on custom or research customers. Someone like muRata might be able to provide a laser trimmed silicone MIM/MOM. AVX also makes custom thin film capacitors but even their Ultr-Miniature Accu-P series only goes down to 50 fF stock.

Making your own might be an option but balancing dielectric constants with dielectric breakdown and thermal characteristics gets somewhat complicated pretty fast, not to mention the fact that you likely need micron scale accuracy for your plate dimensions to get decent tolerances in values. Of course you can always just put some copper tape on a piece of window glass and keep trimming the tape until you get the desired value but I somehow think that would not likely fit your application.

If you provide more information about the application, maybe someone else can chime in with a good supplier.

As far as I know, such capacitors are only available as reference or standard capacitors for calibration of capacitance measuring equipment. These are always 3-terminal "guarded" capacitors, as the residual "parasitic" capacitance between leads through the surrounding space would tend to dominate any intentional capacitance internal to a component.

I am most familiar with such capacitors made by General Radio in the past:

Type 1403-K, 1.0 pF (1,000 aF), +/- 0.1%

Type 1403-N, 0.1 pF (100 aF), +/- 0.1%

Type 1403-R, 0.01 pF (10 aF), +/- 0.3%

Type 1403-V, 0.001 pF (1 aF), +/- 1.0%

As a sometimes quite geeky metrology engineer, I've obtained these capacitors on eBay (with others up to 2.0 nF), but they are quite large (roughly 7 cm dia and 10 cm long, not including the GR input and output coaxial connections) and can be expensive.

An atto Farad is 1,000 times smaller that a femto Farrad, and a million times smaller that a pico Farad; a 1 aF capacitor is getting incredibly small!

According to my ancient Handbook of Chemistry and Physics (and assuming my midnight math is not in error) that would be the capacitance of an isolated sphere 18 nm in diameter.

To make a capacitor with circular plates, e.g., I would say the diameter should be about 10 times the spacing to reduce the effect of fringe fields. For an air (or vacuum) dielectric, the same reference shows that a 1 aF capacitor would have plates 14 nm in diameter with a 1.4 nm spacing.

For the same geometry the capacitance increases as the linear dimensions; a 0.1 fF (100 aF) capacitor would have plate diameters of 1400 nm, just sufficient to be detectable with a visible light microscope. I think this explains why it is hard to find them.

Bruce Carsten Thank you very much. You prompt response is really helpful

I should note a systematic error in my first response yesterday regarding GR Type 1403 capacitors: the equivalent of the capacitance in pF should have been in fF (femto Farads) instead of aF (atto Farads).

For example, a 0.001 pF capacitor is equivalent to 1 fF, not 1 aF.

Labib Labib I am working n air capacitors, I wish to measure the capacitance at 0.1 fF. I saw your question about smaller capacitors, Can you suggest some cost-effective method for measurement of capacitance with the least count of 0.1 fF?