Assume you have a LM324. This OPamp contains bipolar transistors at inputs. To work properly an input bias current of about 45nA is flowing out of each input, remember that input transistors are PNP. This current is the base current of PNP, which must flow to make the OPamp to work properly.
As the LM324 has (almost) identical transistors at each input (non-inverting and inverting), like most of all OPamps available, out of both inputs a current of about 45nA is flowing. The two currents are nearly identical, but can differ by about +-5nA. The difference of both currents is called input offset current.

If you have now high ohmic resistors at inputs, the input bias current will cause relevant voltage drops, but their effect at output will cancel, if you keep them identical. This can be done by chosing the resistors in such a way, that each input bias current sees the same effective ohmic resistance from according input to ground or virtual ground. Then only the difference of input bias currents, the input offset current leads to an error. This error has the same effect like the original input offset voltage and so, it must be kept low.

An example:
LM324 has at inverting input a resistance of 100kOhm to ground, non-inverting input is grounded directly (zero impedance). Then, input bias current causes a voltage drop of 45nA*100kOhm = 4.5mV. This voltage drop has the same effect as the original input offset voltage. Remarkably is, that original input offset voltage of LM324 is only 2mV typically. So, the too high resistance at inverting input heavily increases offset voltage of OPamp.
But if you have also at non-inverting input a 100kOhm resistance to ground, then only the input offset current of 5nA must be taken into consideration: The additional offset voltage is less than 5nA*100kOhm = 0.5mV!

It's clear, that this cancelling methode is only efficient, if input bias current of an OPamp is rather high. If it's low, like with JFET-OPamps or CMOS-OPamps, then the impedance matching is not required.

An example:
Input bias current of LF356, which is an JFET-OPamp, is only 30pA typically. Voltage drop across above 100kOhm resistance is only 3µV, which is 1000 times smaller than the original input offset voltage of LF356. So, there's no advantage at all to do this impedance matching.



If nothing special is told in datasheet, then it's a bipolar chip, all internal transistors are bipolar, like the LM741, LM324, NE5534, etc.
If it's not a plain bipolar chip then it's told in datasheet.

Let's have some examples:

1. LF356, TL084, TL052, etc: OPamp with JFET inputs. The input transistors are FET, concretely spoken JFET (junction FET, gate and source form a reverse polarized pn-junction). All other transistors are bipolar. Sometimes also some other internal circuitry is made of JFETs.

2. CA3140: BiMOS-OPamp with CMOS-transistor inputs and bipolar transistor output. Input transistors and some other internal circuitry are made of CMOS transistors (Complementary Metal Oxid Semiconductor, gate and source are isolated by a highly insulating silicon-oxid layer). All other transistors are bipolar. The OPamp is called a BiMOS-OPamp, because of the mix of bipolar and CMOS transistors.

3. CA3130: BiMOS-OPamp with CMOS-transistor inputs and CMOS-transistor output. Some other additional circuitry is also made of CMOS-transistors. All other transistors are bipolar.

4. TLC274, TLC2274, etc: LinCMOS-OPamp, all internal transistors are CMOS.

And then there are also DiFET, BiFET, CBFET, bipolar/JFET, etc., etc.

Kai