如题，

   为什么VASP中OUTCAR里分轨道的磁矩和不等于总磁矩？

同样的问题也有，

  为什么Bader电荷number of electrons 与各原子charge总和不同？

同样的问题还有，

   为什么分原子的LDOS或者分轨道的PDOS之和不等于总DOS？

这几个问题都涉及到分轨道或分原子的贡献加起来不等于总的贡献。

那么，这里就涉及到分解到每原子的过程是如何分解的问题。

在VASP中PDOS和分原子轨道磁矩输出一般是LORBIT >= 10的设置。

另外还有一个起作用的参数是RWGIS，即每个原子的 Wigner-Seitz半径。

默认值是-1，

VASP手册中建议：

• RWIGS has to be supplied for each atom type if LORBIT is set to <10. In this case, the spd- and site projected wave function character of each band is evaluated, and the local partial DOS (see sections PROCAR and DOSCAR) and local magnetic moments are calculated.
• If LORBIT >= 10, RWIGS is ignored
• RWIGS must be set in calculations with constraining the local magnetic moments (see I_CONSTRAINED_M)

以上LORBIT <10以及限制磁矩大小的时候必须要给每个原子的RWGIS。

并且只有一种原子的体系和多种原子的体系给出了如下说明：

For mono-atomic systems RWIGS can be defined unambiguously. The sum of the volume of the spheres around each atom should be the same as the total volume of the cell (assuming that you do not have a vacuum region within your cell). This is in the spirit of atomic sphere calculations. VASP writes a line

Volume of Typ   1:     98.5 %
to the OUTCAR file. You should use a RWIGS value which yields a volume of  approximately 100%.
For systems consisting of more than one atom type, there is no unambiguous way to define RWIGS and several choices are possible. In all cases, the sum of the volume of the spheres should be close to the total volume of the cell (i.e the sum of the values given by VASP should be around 100%).
One possible choice is to set RWIGS such that the overlap between the spheres is minimized.
a more elaborate way is to apply a Bader-type charge analysis to estimate the radius of an atom in a specific compound.
in most cases, it is simplest to choose the radius of each sphere such that they are close to the covalent radius as tabulated in most periodic tables. This simple criterion can be used in most cases, and it relies at least on some ``physical intuition.
 
Please keep in mind that results are qualitative i.e. there is no unambiguous way to determine the location of an electron (and hence the local magnetic moments). With the current implementation, it is for instance hardly possible to determine charge transfer. What can be derived from the partial DOS is the typical character of a peak in a DOS. Quantitative results can be obtained only by careful comparison with a reference system (e.g. bulk versus surface).
 
以上手册内容表明，通过RWGIS分解得到的每个原子的成分只能用于定性分析，而不能定量来讨论具体的数值加和关系。因为分解过程中涉及到RWGIS的设置，我们只能尽可能的让晶体中所有原子的RWGIS所计算的体积之和接近100%。而不能完美的将空间分解开。同样的问题在计算bader电荷的时候也会遇到。
 
因此，上面三个问题就可以理解了。
 
关于VASP中PDOS相加应不应该总DOS的问题小木虫上luzihen给出了详细的讨论。如下链接：
http://muchong.com/html/201611/10847313.html