"Loop: 50m x 50m with 2000 Hz"

At 2kHz (!) that's not a loop antenna; it's a slightly bent wire.

Just one turn?

Extreme SWR will affect the efficiency, not to mention trying to feed an extremely low Z. This will reduce the radiated energy.

So as far as not seeing anything in the 'far field' (?), perhaps the mathematical results are so low in amplitude that they've fallen below the formatting of your output.

Perhaps you can adjust the frequency upwards (in steps, to at least 1 MHz) to see if the far field results reappear, and then you can contemplate why they're so low at 2 kHz.

Yes, it's just one turn.

And yes, it's not a "antenna".

I am interested in the nearfields (H-field) from 0-50 meter. If I increase the frequency, the far field pattern grows and the efficiency grows to 80-90%. At 2kHz efficiency is 0% .

My question concerns the relation between antenna matching and nearfields! If the efficiency is 0 % the nearfields is still high. Do you think the output from 4nec2 is correct?

It seems like your version of NEC, or more-likely the formatting of your output, is running out of bits. The efficiency will be extremely small, but is should never reach exactly zero until the frequency reaches exactly zero. But at some point the math will run out of bits. But my gut tells me that the output formatting is the more likely place to check first.

My point about exploring the frequency range was to discover where these zeros change to the smallest possible unit. And then you can see what this least significant bit is worth. If it suddenly ticks up from 0% to 1%, then you'll see that the problem is the formatting of the output.

Your approach (2kHz loop) will be soaking in the micro-percent end of the efficiency range. So if the output is not showing you the complete version of the numbers, then you need to change something with the software.

And, predicting the next stage, once you see the numbers then you'll be left dealing with the insanely low radiation resistance and might-as-well-be-zero efficiency.

You may to research the US Navy's ELF network. Their antennas (used to communicate with subs) fill entire valleys.

You think 4nec2 is not the right tool to model such low frequencies?

Is there a complete description how 4nec2 calculates all output parameters? For example, it's not clear which voltage is calculated? There is an input voltage source and the output?

Thanks so far!

Although I'm not an expert with 4nec2, I expect that your data resolution problem is in the GUI, not the NEC engine itself.

You could answer the question for yourself by simply doing a quick search for the frequency where the results just start to be non-zero. Using a binary search method (splitting the difference until you find the transition), it shouldn't take too long. If the result (far field, efficiency, whatever...) suddenly switches from "0%" to "1%" (for example) then the software (the GUI) is useless for exploring situations where the numbers will be below the resolution of the output.

There are others on this group that might have more recent hands-on experience with 4NEC2.