How to supress LO second harmonic in Gilbert cell mixer?

I am designing a low voltage(1.8V) Gilbert cell mixer, which includes(from bottom to top) tail current mos, RF input mos, LO switching mos, and resistive load. Since LO has large amplitude, there would be a large common-mode second harmonic of LO at the drain of RF input mos, which may cause the RF input mos to go into triode region, or even feed through to the RF port. Is there any trick to supress the common-mode 2nd order LO signal? Adding a cascode is not really a good solution for low voltage application.

The Gilbert cell naturally generates even harmonics. If you cannot balance the mixer by DC bias, you simply must use a good low-pass filter to get rid of unwanted second harmonic (and higher).

I know someone add a low voltage at the two IF1/IF2 to surpress the spurs or harmonics.

If you drive LO signal as square wave, it wil have no even order harmonics..
It's coming from Fourier expansion of square wave..