How to make IIP3-NF larger?

change ur process to 0.13um, 90nm...

new process always deteriorate linearity
and I did see why it can improve NF.

For a better NF, burns more power always help for NF, an increase the over-drive of the transconductance stage is good for IIP3

thankyou!
I guess there maybe some optimum value for IIP3-NF under certain current .
Is it possible to achieve iip3 around 10dBm NF<15dB current tail<1mA 2G downconverting to 10MHz using .35 process?

I've never used 0.35
Thus I do not know exactly
You'd better calculate it yourself.
Bosco Leung's book on microwave VLSI circuit design would be helpful.

Noise depends heavily on current through electronic device, so keep it lower for better noise performance. IP3 is proportional to the product of V and I. So if cannot increase I due to noise requirements you may increase the voltage. I remember many years ago I designed RF front end with cascode LNA and Vcc was about 120 V, but current kept small. Linearity and noise performance met the spec.

"Noise depends heavily on current through electronic device, so keep it lower for better noise performance"

I don't agree with this statement. Noise in CMOS transistors for example are usually thermal and reducing the current means increasing the channel resistance and hence increasing the NF. This could be true if your noise is mainly shot noise.

May be. I am not IC designer and do not work with CMOS. I just remember the noise theory and equations that clearly show how noise depends on current.