Spherical aberration refined to an unphysical value

Dear colleagues,

I have an issue with the proper refinement of Spherical aberration of my dataset.

Shall I just disable in the settings prio to run it or there is a way to make it run properly?

Can the small box size be the reason for that?

Thank you in advance.


What is your overall resolution? And what is the box size? Refinement of Cs is only stable at fairly high resolution.

1 Like

Hi Oli,

Raw pixel size (A):

Box size 360x360

Current resolution 4.5 A (yes, quite at some distance from 3A and below).


4.5 Å is much too low resolution to refine Cs - it will always be unstable at this resolution


I typically don’t even think about refining beam tilt and other electro-optical parameters until <3.5A (tilt), <3A (trefoil) or 2-2.5A (tetrafoil/Cs). Mag. aniso is a little more forgiving; I’ve done it and got believable numbers at low as 7A (which can have a dramatic impact on larger objects - 200nm+).


Totally agree for tetrafoil/Cs - for beam tilt if it is the thing limiting resolution (i.e. if it is extremely severe) it can sometimes be worth refining at lower resolution, but usually in such cases it will be obvious there is a problem (FSC will look very odd)


Agreed, problem is that the pathological wavy-navy/valley FSC of extreme beam tilt can also mean pixel size mis-calibration, so solving that can be iterative.


Dear colleagues @olibclarke, @rbs_sci,

Thank you for your input.
Indeed, my test without CS looks ok in terms of FSC.

Chasing the high-res results I was trying to understand the case specific reason why over 300k particles with pixel size 0.9 A/pix do not converge for something better than 4.5 A.

The thickness looks fine. The dataset may still stay heterogeneous.

I will proceed with my investigation.

Kind regards,