It depends what your application is whether you need to avoid Ka2 in the pattern, by using a primary monochromator, or remove Ka2 from the pattern, using software. If you are trying to do structural determination it is best to use a primary monochromator. I do not have a primary monochromator so I always see Ka2 at higher angles. I do a lot of quant work using Rietveld and I never strip Ka2 because the contribution of this, and other lines, can be modelled and accounted for. For qualitative work I also leave Ka2 alone as stripping it can leave artefacts in the background near the peak which just look wrong. Also if a Ka2 of one phase overlaps with a Ka1 of another in a multiphase assemblage, then stripping Ka2 blindly will reduce your chances of identifying the minor components if you are using an intensity based search match procedure. I think that most data reductions like smoothing and stripping should be avoided. Present the raw data, every peak tells a story, don't throw it away.
You can remove the K-alpha 2 from your data by two ways, firstly you can use the specific filter for different X-ray radiation source which block the k-alpha 2 component from the analysis during measurement, secondly you can also remove the contribution of k-alpha 2 component from your obtained data.
It depends what your application is whether you need to avoid Ka2 in the pattern, by using a primary monochromator, or remove Ka2 from the pattern, using software. If you are trying to do structural determination it is best to use a primary monochromator. I do not have a primary monochromator so I always see Ka2 at higher angles. I do a lot of quant work using Rietveld and I never strip Ka2 because the contribution of this, and other lines, can be modelled and accounted for. For qualitative work I also leave Ka2 alone as stripping it can leave artefacts in the background near the peak which just look wrong. Also if a Ka2 of one phase overlaps with a Ka1 of another in a multiphase assemblage, then stripping Ka2 blindly will reduce your chances of identifying the minor components if you are using an intensity based search match procedure. I think that most data reductions like smoothing and stripping should be avoided. Present the raw data, every peak tells a story, don't throw it away.