Jump to content

Requests for technical support from the VASP team should be posted in the VASP Forum.

Spin spirals: Difference between revisions

From VASP Wiki
m Huebsch moved page Spin spirals to Construction:Spin spirals without leaving a redirect
m Huebsch moved page Construction:Spin spirals to Spin spirals without leaving a redirect
 
(2 intermediate revisions by the same user not shown)
Line 1: Line 1:
Spin spirals are magnetic structures in which the direction of the magnetization rotates continuously from one unit cell to the next. In VASP they are modeled with a generalization of the Bloch condition, which captures incommensurate magnetic order in the primitive cell and avoids the need for large supercells.{{cite|marsman:prb:02}}
This page describes the underlying formalism. For step-by-step instructions on setting up, running, and analyzing a spin-spiral calculation, see [[Spin-spiral calculations]].
== Generalized Bloch condition ==
== Generalized Bloch condition ==
Spin spirals may be conveniently modeled using a generalization of the Bloch condition (set {{TAG|LNONCOLLINEAR}}<tt>=.TRUE.</tt> and {{TAG|LSPIRAL}}<tt>=.TRUE.</tt>):
Spin spirals may be conveniently modeled using a generalization of the Bloch condition{{cite|sandratskii:pssb:86}}{{cite|marsman:prb:02}} (set {{TAG|LNONCOLLINEAR|.TRUE.}} and {{TAG|LSPIRAL|.TRUE.}}):


<span id="GeneralizedBlochTheorem">
<span id="GeneralizedBlochTheorem">
Line 14: Line 18:
</span>
</span>


''i.e.'', from one unit cell to the next the up- and down-spinors pick up an additional phase factor of <math>\exp(-i{\bf q}\cdot {\bf R}/2)</math> and <math>\exp(+i{\bf q}\cdot {\bf R}/2)</math>, respectively,
''i.e.'', from one unit cell to the next the up- and down-spinors pick up an additional phase factor of <math>\exp(-i{\bf q}\cdot {\bf R}/2)</math> and <math>\exp(+i{\bf q}\cdot {\bf R}/2)</math>, respectively, where '''R''' is a lattice vector of the crystalline lattice, and '''q''' is the so-called spin-spiral propagation vector.
where '''R''' is a lattice vector of the crystalline lattice, and '''q''' is the so-called spin-spiral propagation vector.


The spin-spiral propagation vector is commonly chosen to lie within the first Brillouin zone of the reciprocal space lattice, and has to be specified by means of the {{TAG|QSPIRAL}}-tag.
The spin-spiral propagation vector is commonly chosen to lie within the first Brillouin zone of the reciprocal lattice, and is specified by means of the {{TAG|QSPIRAL}} tag.


=== Magnetization density ===
The generalized Bloch condition above gives rise to the following behavior of the magnetization density:
The generalized Bloch condition above gives rise to the following behavior of the magnetization density:


:<math>
:<math>
{\bf m} ({\bf r} + {\bf R})= \left(  
{\bf m} ({\bf r} + {\bf R})= \left(
\begin{array}{c}
\begin{array}{c}
m_x({\bf r}) \cos({\bf q} \cdot {\bf R}) - m_y({\bf r}) \sin({\bf q} \cdot {\bf R}) \\
m_x({\bf r}) \cos({\bf q} \cdot {\bf R}) - m_y({\bf r}) \sin({\bf q} \cdot {\bf R}) \\
m_x({\bf r}) \sin({\bf q} \cdot {\bf R}) + m_y({\bf r}) \cos({\bf q} \cdot {\bf R}) \\
m_x({\bf r}) \sin({\bf q} \cdot {\bf R}) + m_y({\bf r}) \cos({\bf q} \cdot {\bf R}) \\
m_z({\bf r})  
m_z({\bf r})
\end{array}  
\end{array}
\right)
\right)
</math>
</math>


This is schematically depicted below:
The components of the magnetization in the ''xy''-plane rotate about the spin-spiral propagation vector '''q''', while the out-of-plane component <math>m_z</math> retains the usual cell periodicity. This is depicted schematically below:
the components of the magnization in the ''xy''-plane rotate about the spin-spiral propagation vector '''q'''.
 
 
:[[File:Spinspiral.png|400px]]
 
 
'''N.B.''': This does not mean that the magnetisation density may not have contributions along the ''z''-direction.
These, however, will not be affected by the generalized Bloch condition, ''i.e.'', <math>m_z ({\bf r})</math> will simply show the usual cell periodicity.


If one explicitly wants to keep the magnetisation density from developing components along the ''z''-direction set:
[[File:Spinspiral.png|center|400px]]


{{TAGBL|LZEROZ}} = .TRUE.
{{NB|mind|The generalized Bloch condition does not mean that the magnetization density may not have contributions along the ''z''-direction; these are simply unaffected by it. To keep the magnetization density from developing a component along ''z'', set {{TAG|LZEROZ|.TRUE.}}, which sets <math>m_z({\bf r}) = 0</math> at each step of the electronic minimization.|:}}


This will set <math>m_z ({\bf r}) = 0</math> at each iteration step of the electronic minimisation.
== Modified Hamiltonian ==
 
== Basis set considerations ==
The generalized Bloch condition redefines the Bloch functions as follows:
The generalized Bloch condition redefines the Bloch functions as follows:


Line 73: Line 67:
</math>
</math>


where in <math>H^{\uparrow\uparrow}</math> and <math>H^{\downarrow\downarrow}</math> the kinetic energy of a plane wave component changes to:
where in <math>H^{\uparrow\uparrow}</math> and <math>H^{\downarrow\downarrow}</math> the kinetic energy of a plane-wave component changes to:


:<math>
:<math>
Line 83: Line 77:
</math>
</math>


In the case of spin-spiral calculations the energy cutoff of the basis set of the individual spinor components is specified by means of the {{TAG|ENINI}}-tag.
Because the only change to the off-diagonal exchange-correlation potential is a phase modulation, a spin-spiral calculation has approximately the same computational cost as a standard noncollinear calculation{{cite|hobbs:prb:00}} of the primitive cell.
 
Additionally one needs to set {{TAG|ENMAX}} appropriately:
{{TAG|ENMAX}} needs to be chosen larger than {{TAG|ENINI}}, and large enough so that the plane wave components of both the up-spinors as well as the components of the down-spinor all have a kinetic energy smaller than {{TAG|ENMAX}}.
This is the case when:
 
:<math>
\mathtt{ENMAX} \geq \frac{\hbar^2}{2m}\left( G_{\rm ini} + |q| \right)^2
</math>


where
== Basis-set considerations ==
Because the two spinor components are shifted in reciprocal space by <math>\pm{\bf q}/2</math>, the plane-wave cutoff must be chosen carefully. The cutoff of the basis set of the individual spinor components is specified by means of the {{TAG|ENINI}} tag, while {{TAG|ENMAX}} must be chosen large enough that the plane-wave components of both spinors have a kinetic energy below {{TAG|ENMAX}}. This is the case when


:<math>
:<math>
G_{\rm ini}=\sqrt{\frac{2m}{\hbar^2}\mathtt{ENINI}}
\mathtt{ENMAX} \geq \frac{\hbar^2}{2m}\left( G_{\rm ini} + |q| \right)^2,
\qquad
G_{\rm ini}=\sqrt{\frac{2m}{\hbar^2}\mathtt{ENINI}}.
</math>
</math>


In most cases it is more than sufficient to set {{TAG|ENMAX}}={{TAG|ENINI}}<tt>+100</tt>.
In practice it is more than sufficient to set {{TAG|ENMAX}} to {{TAG|ENINI}} + 100 eV. The practical settings, and the runtime warning VASP prints when {{TAG|ENMAX}} is too small, are described in [[Spin-spiral calculations]].
 
To judge whether {{TAG|ENMAX}} was chosen large enough one will always get a warning at runtime, ''e.g.''
 
  -----------------------------------------------------------------------------
|                                                                            |
|          W    W    AA    RRRRR  N    N  II  N    N  GGGG  !!!          |
|          W    W  A  A  R    R  NN  N  II  NN  N  G    G  !!!          |
|          W    W  A    A  R    R  N N  N  II  N N  N  G      !!!          |
|          W WW W  AAAAAA  RRRRR  N  N N  II  N  N N  G  GGG  !            |
|          WW  WW  A    A  R  R  N  NN  II  N  NN  G    G                |
|          W    W  A    A  R    R  N    N  II  N    N  GGGG  !!!          |
|                                                                            |
|      To represent the spin spiral you requested, with a kinetic            |
|      energy cutoff of ENINI=  300.00 eV, choose ENMAX >  331.21 eV          |
|      Currently ENMAX=  400.00 eV                                            |
|                                                                            |
  -----------------------------------------------------------------------------


== Symmetry ==
== Symmetry ==
Generally the introduction of a spin-spiral will lower the symmetry of the system.
The introduction of a spin spiral generally lowers the symmetry of the system, and VASP cannot currently account for the presence of a spin spiral in its symmetry analysis. For this reason the use of symmetry has to be switched off completely ({{TAG|ISYM|-1}}) in spin-spiral calculations.
At present VASP can not correctly account for the presence of a spin-spiral in its symmetry analysis.


Therefore the use of symmetry has to be switched of completely:
== Related tags and articles ==
[[Spin-spiral calculations]]


{{TAGBL|ISYM}} = -1
{{TAG|LNONCOLLINEAR}},
 
== Initialization of the magnetic subsystem ==
As for all calculations on magnetic systems it is highly advisable to initialise the magnetic subsystem by means of the {{TAG|MAGMOM}} tag.
 
Note that, in case one does '''not''' specify initial magnetic moments, the initial order of the magnetic subsystem will be ferromagnetic (see the {{TAG|MAGMOM}} default values),
and even if the groundstate of your system is not ferromagnetic, the initial ferromagnetic state might very well be a local minimum in which the system can remain stuck during the electronic minimisation.
 
Of course, in case of spin-spiral calculations the magnetic subsystem is not completely characterized by the magnetic configuration within the unit cell,
but will additionally depend on how the magnetization density changes from one unit cell to the next as determined by the spiral propagation vector.
 
Consider the following two examples:
 
*Two magnetic atoms per unit cell, both with initial magnetic moments ''M'' along the ''y''-axis and <math>q=(0,0,\frac{1}{2})</math>:
 
{{TAGBL|MAGMOM}} = 0 M 0  0 M 0
{{TAGBL|QSPIRAL}}= 0.0 0.0 0.5
 
: ''i.e.'', a double layer antiferromagnet:
:[[File:ss1.png|500px]]
 
 
*Two magnetic atoms per unit cell, both with initial magnetic moments ''M'' along the ''y''- and ''x''-axis, respectively, and <math>q=(0,0,\frac{1}{2})</math>:
 
{{TAGBL|MAGMOM}} = 0 M 0  M 0 0
{{TAGBL|QSPIRAL}}= 0.0 0.0 0.5
 
: ''i.e.'', a spin spiral:
 
:[[File:ss2.png|500px]]
 
 
'''N.B.''': both of the aforementioned magnetic arrangements obey the same generalized Bloch condition, <math>q=(0,0,0.5)</math>, and during the electronic minimisation one may transform into the other if that lowers the total energy.
In other words the generalized Bloch condition dictates the change in the magnetization density from one unit cell to the next, but it does not explicitly constrain the magnetic order ''within'' the unit cells themselves.
 
== Local magnetic moments ==
Analysing the magnetisation density from spin-spiral calculations in terms of site resolved local magnetic moments is a bit more involved than usual.
Problems arise from the fact that in most cases the spin-spiral period will not be commensurate with the unit cell (otherwise there would be no reason to use the generalised Bloch theorem).
This means that ''implicitly'' the magnetisation density is not cell periodic, as illustrated in the figure below:
 
:[[File:ss3.png|300px]]
 
The usual ways to analyse the site resolved local charge density and magnetisation (''e.g.'' setting the {{TAG|LORBIT}}-tag: output in the {{FILE|PROCAR}} file and at the end of the  {{FILE|OUTCAR}} file) do not account for this.
As a workaround one may (ab)use the infrastructure created for the ''constrained magnetic moment'' approach, as follows:
 
{{TAGBL|I_CONSTRAINED_M}} = 1
{{TAGBL|LAMBDA}} = 0.0
 
''i.e.'', one switches on the [[I_CONSTRAINED_M|constrained magnetic moment approach]] ({{TAG|I_CONSTRAINED_M}}<tt>=1</tt>) but sets the penalty potential to zero ({{TAG|LAMBDA}}<tt>=0.0</tt>).
 
'''N.B.''': do not forget to set {{TAG|RWIGS}} appropriately.
 
The magnetisation density will correctly integrated inside site centered sphere of radius {{TAG|RWIGS}},
and the resulting local magnetic moments are written under <tt>M_int</tt>, to the {{FILE|OSZICAR}} file.
For instance:
 
  E_p =  0.00000E+00  lambda =  0.000E+00
<lVp>=  0.00000E+00
  DBL =  0.00000E+00
  ion        MW_int                M_int
  1  1.178  0.000  0.000    1.573  0.000  0.000
RMM:  8    -0.819213822792E+01    0.53417E-07  -0.43965E-08  2542  0.310E-03
 
In the above, the local magnetic moment on ion 1 (after iteration 8) is <math>M=1.573 \hat{x} \;\mu_{\rm B}</math>.
 
== Related Tags and Sections ==
{{TAG|LSPIRAL}},
{{TAG|LSPIRAL}},
{{TAG|QSPIRAL}},
{{TAG|QSPIRAL}},
{{TAG|LZEROZ}},
{{TAG|LZEROZ}},
{{TAG|LNONCOLLINEAR}},
{{TAG|MAGMOM}},
{{TAG|ENINI}},
{{TAG|ENINI}},
{{TAG|ENMAX}},
{{TAG|ENMAX}},
{{TAG|ISYM}},
{{TAG|ISYM}},
{{TAG|I_CONSTRAINED_M}},
{{TAG|MAGMOM}}
{{TAG|LAMBDA}},
{{TAG|M_CONSTR}},
{{TAG|RWIGS}}


----
== References ==
<references/>


[[Category:Magnetism]][[Category:Spin spirals]][[Category:Theory]][[Category:Howto]]
[[Category:Magnetism]][[Category:Spin spirals]][[Category:Theory]]

Latest revision as of 08:18, 22 June 2026

Spin spirals are magnetic structures in which the direction of the magnetization rotates continuously from one unit cell to the next. In VASP they are modeled with a generalization of the Bloch condition, which captures incommensurate magnetic order in the primitive cell and avoids the need for large supercells.[1]

This page describes the underlying formalism. For step-by-step instructions on setting up, running, and analyzing a spin-spiral calculation, see Spin-spiral calculations.

Generalized Bloch condition

Spin spirals may be conveniently modeled using a generalization of the Bloch condition[2][1] (set LNONCOLLINEAR = .TRUE. and LSPIRAL = .TRUE.):

[math]\displaystyle{ \left[ \begin{array}{c} \Psi^{\uparrow}_{\bf k}(\bf r) \\ \Psi^{\downarrow}_{\bf k}(\bf r) \end{array} \right] = \left( \begin{array}{cc} e^{-i\bf q \cdot \bf R / 2} & 0\\ 0 & e^{+i\bf q \cdot \bf R / 2} \end{array}\right) \left[ \begin{array}{c} \Psi^{\uparrow}_{\bf k}(\bf r-R) \\ \Psi^{\downarrow}_{\bf k}(\bf r-R) \end{array} \right], }[/math]

i.e., from one unit cell to the next the up- and down-spinors pick up an additional phase factor of [math]\displaystyle{ \exp(-i{\bf q}\cdot {\bf R}/2) }[/math] and [math]\displaystyle{ \exp(+i{\bf q}\cdot {\bf R}/2) }[/math], respectively, where R is a lattice vector of the crystalline lattice, and q is the so-called spin-spiral propagation vector.

The spin-spiral propagation vector is commonly chosen to lie within the first Brillouin zone of the reciprocal lattice, and is specified by means of the QSPIRAL tag.

Magnetization density

The generalized Bloch condition above gives rise to the following behavior of the magnetization density:

[math]\displaystyle{ {\bf m} ({\bf r} + {\bf R})= \left( \begin{array}{c} m_x({\bf r}) \cos({\bf q} \cdot {\bf R}) - m_y({\bf r}) \sin({\bf q} \cdot {\bf R}) \\ m_x({\bf r}) \sin({\bf q} \cdot {\bf R}) + m_y({\bf r}) \cos({\bf q} \cdot {\bf R}) \\ m_z({\bf r}) \end{array} \right) }[/math]

The components of the magnetization in the xy-plane rotate about the spin-spiral propagation vector q, while the out-of-plane component [math]\displaystyle{ m_z }[/math] retains the usual cell periodicity. This is depicted schematically below:


Modified Hamiltonian

The generalized Bloch condition redefines the Bloch functions as follows:

[math]\displaystyle{ \Psi^{\uparrow}_{\bf k}(\bf r) = \sum _{\bf G} \rm C^{\uparrow}_{\bf k \bf G} e^{i(\bf k + \bf G -\frac{\bf q}{2})\cdot \bf r} }[/math]
[math]\displaystyle{ \Psi^{\downarrow}_{\bf k}(\bf r) = \sum _{\bf G} \rm C^{\downarrow}_{\bf k \bf G} e^{i(\bf k + \bf G +\frac{\bf q}{2})\cdot \bf r} }[/math]

This changes the Hamiltonian only minimally:

[math]\displaystyle{ \left( \begin{array}{cc} H^{\uparrow\uparrow} & V^{\uparrow\downarrow}_{\rm xc} \\ V^{\downarrow\uparrow}_{\rm xc} & H^{\downarrow\downarrow} \end{array}\right) \rightarrow \left( \begin{array}{cc} H^{\uparrow\uparrow} & V^{\uparrow\downarrow}_{\rm xc} e^{-i\bf q \cdot \bf r} \\ V^{\downarrow\uparrow}_{\rm xc}e^{+i\bf q \cdot \bf r} & H^{\downarrow\downarrow} \end{array}\right), }[/math]

where in [math]\displaystyle{ H^{\uparrow\uparrow} }[/math] and [math]\displaystyle{ H^{\downarrow\downarrow} }[/math] the kinetic energy of a plane-wave component changes to:

[math]\displaystyle{ H^{\uparrow\uparrow}:\qquad |{\bf k} + {\bf G}|^2 \rightarrow |{\bf k} + {\bf G} - {\bf q} /2|^2 }[/math]
[math]\displaystyle{ H^{\downarrow\downarrow}:\qquad |{\bf k} + {\bf G}|^2 \rightarrow |{\bf k} + {\bf G} + {\bf q} /2|^2 }[/math]

Because the only change to the off-diagonal exchange-correlation potential is a phase modulation, a spin-spiral calculation has approximately the same computational cost as a standard noncollinear calculation[3] of the primitive cell.

Basis-set considerations

Because the two spinor components are shifted in reciprocal space by [math]\displaystyle{ \pm{\bf q}/2 }[/math], the plane-wave cutoff must be chosen carefully. The cutoff of the basis set of the individual spinor components is specified by means of the ENINI tag, while ENMAX must be chosen large enough that the plane-wave components of both spinors have a kinetic energy below ENMAX. This is the case when

[math]\displaystyle{ \mathtt{ENMAX} \geq \frac{\hbar^2}{2m}\left( G_{\rm ini} + |q| \right)^2, \qquad G_{\rm ini}=\sqrt{\frac{2m}{\hbar^2}\mathtt{ENINI}}. }[/math]

In practice it is more than sufficient to set ENMAX to ENINI + 100 eV. The practical settings, and the runtime warning VASP prints when ENMAX is too small, are described in Spin-spiral calculations.

Symmetry

The introduction of a spin spiral generally lowers the symmetry of the system, and VASP cannot currently account for the presence of a spin spiral in its symmetry analysis. For this reason the use of symmetry has to be switched off completely (ISYM = -1) in spin-spiral calculations.

Related tags and articles

Spin-spiral calculations

LNONCOLLINEAR, LSPIRAL, QSPIRAL, LZEROZ, ENINI, ENMAX, ISYM, MAGMOM

References