72
W. C. Law and S. De W. Wong
although a thin coupling layer has to be included between the two CoFeB as a strain
relief from crystallization from both sides of the MgO.
5.6 Free Layer
The free layer is also another key research focus as it is responsible for the storage of
the bit information. There are various design concepts pertaining to the free layer to
deal with specific issues with minor trade-offs. In general, CoFeB has been commonly
used as the free layer in pMTJ due to its high degree of flexibility [95, 149], low
Gilbert damping factor ranging from 0.0035 to 0.032 [149, 150, 151, 152, 153, 154],
and amorphousness during deposition [50, 155]. CoFeB may crystallize in bcc(100)
orientation upon annealing beyond 300 °C, resulting in a good lattice matching with
MgO and subsequently high TMR [30, 156]. A significant amount of research has
been focused on the annealing dependence of CoFeB and how its crystallization,
boron and oxygen absorption by the adjacent layers (e.g. Ta, Mo, W and MgO), and
magnetic dead layer effect due to rough interfaces and/or interlayer diffusion [29,
86, 126, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165]. The magnetic dead layer
is another concern as it may reduce the effective thickness of the free layer, and
therefore the thermal stability.
However, there is a limit in the maximum thermal stability achievable using a
single CoFeB layer since its origin of PMA comes from interfacial effect. To increase
the thermal stability, free layer with two CoFeB layers coupled by a thin non-magnetic
layer (such as Ta, Mo or W) has been reported to achieve almost double the thermal
stability while maintaining comparable TMR and J c [86, 166, 167, 168]. Both sides of
the CoFeB layers can be sandwiched by MgO tunnel barriers for maximum interfacial
anisotropy effect as well as improving symmetrical switching during write operations
from P → AP state. The standard seems to be leaning towards double MgO-based
pMTJ stack as discussed above, but some novel concepts are also introduced below.
Liu et al. suggested to substitution one of the CoFeB layer within the free layer
section with Co/Ni multilayers instead as they bear similarities in some of the key
magnetic properties [169]. Co/Ni has a slightly higher gilbert damping value than
CoFeB, but has a much higher M s and PMA generated from magnetocrystalline
anisotropy (MCA, see Sect. 3.3) [169, 170, 171, 172, 173, 174]. Since Co/Ni multilayers do not rely on interfacial PMA, the magnetic volume can now scale according
to the demands. That having said, strong MCA is observed when Co/Ni multilayers
are in the (111) crystalline direction, which means that the coupling layer between
Co/Ni and CoFeB must be sufficiently thick or amorphous to prevent the transfer of
fcc crystalline texture to the CoFeB free layer. Furthermore, the MCA of Co/Ni is
heavily reliant on the contiguous layers to transfer the crystalline texture, which is
restricted to only one side of the interface since the other side would be the texture
breaking coupling layer. This means that double-MgO will not be deployed in MTJs
with such hybrid free layer.
W. C. Law and S. De W. Wong
although a thin coupling layer has to be included between the two CoFeB as a strain
relief from crystallization from both sides of the MgO.
5.6 Free Layer
The free layer is also another key research focus as it is responsible for the storage of
the bit information. There are various design concepts pertaining to the free layer to
deal with specific issues with minor trade-offs. In general, CoFeB has been commonly
used as the free layer in pMTJ due to its high degree of flexibility [95, 149], low
Gilbert damping factor ranging from 0.0035 to 0.032 [149, 150, 151, 152, 153, 154],
and amorphousness during deposition [50, 155]. CoFeB may crystallize in bcc(100)
orientation upon annealing beyond 300 °C, resulting in a good lattice matching with
MgO and subsequently high TMR [30, 156]. A significant amount of research has
been focused on the annealing dependence of CoFeB and how its crystallization,
boron and oxygen absorption by the adjacent layers (e.g. Ta, Mo, W and MgO), and
magnetic dead layer effect due to rough interfaces and/or interlayer diffusion [29,
86, 126, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165]. The magnetic dead layer
is another concern as it may reduce the effective thickness of the free layer, and
therefore the thermal stability.
However, there is a limit in the maximum thermal stability achievable using a
single CoFeB layer since its origin of PMA comes from interfacial effect. To increase
the thermal stability, free layer with two CoFeB layers coupled by a thin non-magnetic
layer (such as Ta, Mo or W) has been reported to achieve almost double the thermal
stability while maintaining comparable TMR and J c [86, 166, 167, 168]. Both sides of
the CoFeB layers can be sandwiched by MgO tunnel barriers for maximum interfacial
anisotropy effect as well as improving symmetrical switching during write operations
from P → AP state. The standard seems to be leaning towards double MgO-based
pMTJ stack as discussed above, but some novel concepts are also introduced below.
Liu et al. suggested to substitution one of the CoFeB layer within the free layer
section with Co/Ni multilayers instead as they bear similarities in some of the key
magnetic properties [169]. Co/Ni has a slightly higher gilbert damping value than
CoFeB, but has a much higher M s and PMA generated from magnetocrystalline
anisotropy (MCA, see Sect. 3.3) [169, 170, 171, 172, 173, 174]. Since Co/Ni multilayers do not rely on interfacial PMA, the magnetic volume can now scale according
to the demands. That having said, strong MCA is observed when Co/Ni multilayers
are in the (111) crystalline direction, which means that the coupling layer between
Co/Ni and CoFeB must be sufficiently thick or amorphous to prevent the transfer of
fcc crystalline texture to the CoFeB free layer. Furthermore, the MCA of Co/Ni is
heavily reliant on the contiguous layers to transfer the crystalline texture, which is
restricted to only one side of the interface since the other side would be the texture
breaking coupling layer. This means that double-MgO will not be deployed in MTJs
with such hybrid free layer.