Hi Ralf, Thanks for replying to me. I believe I have the 5 Gaussian approximation for the values of a1-a4, b1-b4, and c for electron scattering factors. I think this is the same format as in the International Table volume 4. I haven't been able to locate a library file in phenix directory so I presume that the information is incorporated into the codes or in cctbx. George Wisedchaisri On Tue, 20 Jul 2010, Ralf W. Grosse-Kunstleve wrote:

...

I have a file containing electron scattering factors for 300kV EM that has been used successfully for refinement with CNS before for electron diffraction data. How can I tell Phenix to use electron scattering factors during refinement? I think currently Phenix has 4 choices for X-ray and neutron scattering factors but not for electron scattering. Is there a good way to implement electron scattering factors into phenix library?

No, sorry. The underlying libraries support custom coefficients for the scattering factors (must be a Gaussian approximation), but there is no practical way for users to define them. I'll make this a priority for the next official phenix release. Ralf _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Hi Ralf, Is it possible yet to incorporate user-defined custom X-ray scattering factors? Thanks, -Andy On Tue, 20 Jul 2010, Ralf W. Grosse-Kunstleve wrote:

...

I have a file containing electron scattering factors for 300kV EM that has been used successfully for refinement with CNS before for electron diffraction data. How can I tell Phenix to use electron scattering factors during refinement? I think currently Phenix has 4 choices for X-ray and neutron scattering factors but not for electron scattering. Is there a good way to implement electron scattering factors into phenix library?

No, sorry. The underlying libraries support custom coefficients for the scattering factors (must be a Gaussian approximation), but there is no practical way for users to define them. I'll make this a priority for the next official phenix release. Ralf _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Hi Ralf, This is a good news for electron crystallography! It looks like you have provided one of the most recent Gaussian fitting of electron scattering factors from Peng et al. Acta Crys A52, 1996 which is supposed to be an improvement to the widely used Doyle and Turner 1968 by using 5 gaussian fittings instead of the 4 + a constant fitting (also used for X-ray scattering factors). I have a question though. Because electron scattering depends on the velocity of the electron which in turn depends on the acceleration voltage of the microscope. The gaussian parameters need to be corrected using the relativistic electron velocity by multiplying the ai parameters with m/m0 = (1 – v2/c2) ^-1/2 where v is the relativistic electron velocity and c is the velocity of light (for example m/m0 = 1.58707 for a 300keV electron microscope). These days it is probably common for many electron crystallography labs to use 200-300keV microscopes for data collection. Is there a way to implement this correction into Phenix? or alternatively allow users to supply their own parameters? I normally supply my own files for electron scattering factors of appropriate acceleration voltage in Refmac and CNS (but can only do so in the 4 gaussian + a constant format). I think the 5-Gaussian fit is more accurate and it would be nice if other refinement programs can support such parameters in the future. George Wisedchaisri On Fri, 17 Jun 2011, Ralf Grosse-Kunstleve wrote:

Hi Andy,

...

Is it possible yet to incorporate user-defined custom X-ray scattering factors?

No, sorry, we never got to it. But we added a table of electron scattering factors:

http://cci.lbl.gov/cctbx_sources/cctbx/eltbx/e_scattering.py

Is this the right table for your purposes?

Your question reminded me to add "electron" as a possible choice for the

refinement.main.scattering_table=electron

parameter, which enables use of the table above. This will be in the next development build (phenix dev-791 or higher).

Ralf _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Hi Ralf, No, the factor m/m0 is independent of atom types for the same acceleration voltage (m0 is electron rest mass, m is relativistic mass of electron which depends on acceleration voltage. The Guassian ai and bi coefficients already take care of fitting scattering factor f(e) for each atom type. In another word, if you can implement a function that calculates and multiplies m/m0 to all the ai coefficients for all atom types in the table, users will only need to input the acceleration voltage. Thanks, George Wisedchaisri On Mon, 20 Jun 2011, Ralf Grosse-Kunstleve wrote:

Hi George,

...

I have a question though. Because electron scattering depends on the velocity of the electron which in turn depends on the acceleration voltage of the microscope. The gaussian parameters need to be corrected using the relativistic electron velocity by multiplying the ai parameters with m/m0 = (1 – v2/c2) ^-1/2 where v is the relativistic electron velocity and c is the velocity of light (for example m/m0 = 1.58707 for a 300keV electron microscope).

Is the factor m/m0 different for each atom type?

Ralf _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Can anomalous (f') and dispersive (f") contributions be ignored?

I know this was not at all your point, but are these the right terms? I thought both f' and f" were called anomalous--or better, resonant--contributions, whereas dispersive contributions are due to different amounts of resonant scattering at different wavelengths, say "peak versus inflection." Jacob

Hi Ralf, Hmm... You raised a very good point. Multiplying a_i by a constant would be the same effect as scaling Fcalc, and that should be fine as scaling is normally done during the refinement anyway. I will be curious to test this with electron diffraction data. Hint: electron diffraction data to 1.9A resolution is available from pdb under accession code 2B6O. George On Mon, 20 Jun 2011, Ralf Grosse-Kunstleve wrote:

Hi George,

...

No, the factor m/m0 is independent of atom types

Can anomalous (f') and dispersive (f") contributions be ignored?

Ignoring these the structure factor equation is

f_calc += f0 * dw * occ * sym_weight * exp(2 * pi * i * x * h)

dw = isotropic or anisotropic Debye-Waller factor occ = occupancy factor (between 0 and 1) sym_weight = usually 1 except for special positions x = fractional coordinate of atom h = Miller index

f0 = Sum[a_i * exp(-b_i * stol**2)]

If you scale all the a_i by a constant factor, it is the same as scaling the final f_calc by the same constant factor. Therefore the m/m0 will just change the overall isotropic scale factor, which we determine dynamically inside phenix.refine; the value isn't usually interesting. I.e. I'd expect that you can use phenix.refine as-is.

But maybe I'm missing something important? I've never worked with electron diffraction data.

Ralf

On Mon, Jun 20, 2011 at 1:45 PM, Goragot Wisedchaisri wrote:

...

Hi Ralf,

No, the factor m/m0 is independent of atom types for the same acceleration voltage (m0 is electron rest mass, m is relativistic mass of electron which depends on acceleration voltage. The Guassian ai and bi coefficients already take care of fitting scattering factor f(e) for each atom type. In another word, if you can implement a function that calculates and multiplies m/m0 to all the ai coefficients for all atom types in the table, users will only need to input the acceleration voltage.

Thanks,

George Wisedchaisri

On Mon, 20 Jun 2011, Ralf Grosse-Kunstleve wrote:

...

Hi George,

...

I have a question though. Because electron scattering depends on the velocity of the electron which in turn depends on the acceleration voltage of the microscope. The gaussian parameters need to be corrected using the relativistic electron velocity by multiplying the ai parameters with m/m0 = (1 – v2/c2) ^-1/2 where v is the relativistic electron velocity and c is the velocity of light (for example m/m0 = 1.58707 for a 300keV electron microscope).

Is the factor m/m0 different for each atom type?

Ralf _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb