the role of exchange bias in domain dynamics control

21
The Role of Exchange Bias in Domain Dynamics Control A. Benassi 1 , M.A. Marioni 1 , D. Passerone 1 and H.J. Hug 1,2 1 EMPA Swiss Federal InsHtute for Materials Science and Technology, Dübendorf (Switzerland) 2 Department of Physics, Universität Basel, Basel (Switzerland)

Transcript of the role of exchange bias in domain dynamics control

Page 1: the role of exchange bias in domain dynamics control

The  Role  of  Exchange  Bias  in    Domain  Dynamics  Control  

A.  Benassi1,  M.A.  Marioni1,  D.  Passerone1  and  H.J.  Hug1,2          

   

1-­‐  EMPA  Swiss  Federal  InsHtute  for  Materials  Science  and  Technology,  Dübendorf  (Switzerland)  2-­‐    Department  of  Physics,  Universität  Basel,  Basel  (Switzerland)  

Page 2: the role of exchange bias in domain dynamics control

Sample  and  measurements  A   perpendicular   anisotropy   ferromagneHc   film   (FF)   is   anH-­‐coupled  with   a   thinner  anHferromagneHc  film  (AF)  grown  on  top    

Upon  cooling  below  the  Neel  temperature,  the  AF  becomes  ordered  except  for  few  atomic   layer   at   the   interface,   here   the   defects   at   the   interface   give   rise   to   a  distribuHon  of  uncompensated  spins  (UCS)  

Being  the  Neel  temperature  of  the  AF  smaller  than  the  Curie  temperature  of  the  FF,  the  presence  of  the  ferromagneHc  domains   can   orient   the   uncompensated   spin   at   the   interface   during   the   cooling.   This   allow   us   to   fix   stably   the   FF  domain  structure  on  the  cooled  AF.      

the  domains  image  is  taken  through  the  AF  layer  because  the  FF  field  is  orders  of  

magnitude  stronger  

the  UCS  image  is  taken  saturaHng  the  FF  domains  with  an  external  field  

Schmid  e  al.  PRL  105  197201  (2010)            Joshi  et  al.  Appl.Phys.LeY.  98  082502  (2011)  

Hext

uncompe

nsated

 

frustrated

 

AF  

FF  

Page 3: the role of exchange bias in domain dynamics control

The  model  system  The  Landau-­‐Lifshitz-­‐Gilbert  (LLG)  equaHon  rules  the  precession  of  a  magneHc  dipole  in  an  external  field:  

∂m

∂t= − γ

1 + ξ2m×

�B+ ξ

�m×B

��

B = − 1

Ms

δH[m]

δm+Q(R, t)

�Q(R, t)� = 0

�Q(R, t)Q(R�, t�)� = δ(t− t�)δ(R−R�)2KBT ξ/Msγ

Bm

precession  term  

Bm

damping  term  dissipaHon  by  

microscopic  degrees  of  freedom    

Bm

stochasHc  term  thermal  fluctuaHons  

Under  the  following  approximaHons:      

•  scalar  magneHzaHon  uniform  along  the  FF  thickness  d

•  domain  walls  smaller  than  the  domain  size  

•  small  FF  thickness  d

the  magneHzaHon  in  the  FF  can  be  described  by  the  following  hamiltonian  power  expansion:    

Gilbert  IEEE  Trans.  On  MagneHcs  40  3434  (2004)            Brown  Phys.Rev.  130  1677  (1963)            Usadel  PRB  73  212405  (2006)  

m(r, t) = m(x, y, t)z

H =

�d3R

�−Ku(R)

m2

2+

A

2(∇Rm)2 +

µ0M2s d

�d2R�m(R�)m(R)

|R−R�|3− µ0Msm(Hext −HUCS(R))

Page 4: the role of exchange bias in domain dynamics control

The  model  system  

∂m

∂t= − γ

1 + ξ2m×

�B+ ξ

�m×B

��

B = − 1

Ms

δH[m]

δm+Q(R, t)

�Q(R, t)� = 0

�Q(R, t)Q(R�, t�)� = δ(t− t�)δ(R−R�)2KBT ξ/Msγ

Bm Bm

Under  the  following  approximaHons:      

•  scalar  magneHzaHon  uniform  along  the  FF  thickness  d

•  domain  walls  smaller  than  the  domain  size  

•  small  FF  thickness  d

the  magneHzaHon  in  the  FF  can  be  described  by  the  following  hamiltonian  power  expansion:    

Jagla  PRB  72  094406  (2005)            Jagla  PRB  70  046204  (2004)    

m(r, t) = m(x, y, t)z

The  Landau-­‐Lifshitz-­‐Gilbert  (LLG)  equaHon  rules  the  precession  of  a  magneHc  dipole  in  an  external  field:  

precession  term  

Bm

damping  term  dissipaHon  by  

microscopic  degrees  of  freedom    

stochasHc  term  thermal  fluctuaHons  

H =

�d3R

�−Ku(R)

m2

2+

A

2(∇Rm)2 +

µ0M2s d

�d2R�m(R�)m(R)

|R−R�|3− µ0Msm(Hext −HUCS(R))

Page 5: the role of exchange bias in domain dynamics control

The  model  system  

Under  the  following  approximaHons:      

•  scalar  magneHzaHon  uniform  along  the  FF  thickness  d

•  domain  walls  smaller  than  the  domain  size  

•  small  FF  thickness  d

the  magneHzaHon  in  the  FF  can  be  described  by  the  following  hamiltonian  power  expansion:    

m(r, t) = m(x, y, t)z

Jagla  PRB  72  094406  (2005)            Jagla  PRB  70  046204  (2004)    

anisotropy  term:  its   fluctuaHons   around   an  average   value   provides   strong  pinning   points   for   the   domain  walls  

Ku(R) = �Ku�(1− P (x, y))

�P (R)� = 0

�P (R)P (R�)� = θδ(R−R�)

H =

�d3R

�−Ku(R)

m2

2+

A

2(∇Rm)2 +

µ0M2s d

�d2R�m(R�)m(R)

|R−R�|3− µ0Msm(Hext −HUCS(R))

Page 6: the role of exchange bias in domain dynamics control

H =

�d3R

�−Ku(R)

m2

2+

A

2(∇Rm)2 +

µ0M2s d

�d2R�m(R�)m(R)

|R−R�|3− µ0Msm(Hext −HUCS(R))

The  model  system  anisotropy  term:  its   fluctuaHons   around   an  average   value   provides   strong  pinning   points   for   the   domain  walls  

Under  the  following  approximaHons:      

•  scalar  magneHzaHon  uniform  along  the  FF  thickness  d

•  domain  walls  smaller  than  the  domain  size  

•  small  FF  thickness  d

the  magneHzaHon  in  the  FF  can  be  described  by  the  following  hamiltonian  power  expansion:    

m(r, t) = m(x, y, t)z

Ku(R) = �Ku�(1− P (x, y))

�P (R)� = 0

�P (R)P (R�)� = θδ(R−R�)

Jagla  PRB  72  094406  (2005)            Jagla  PRB  70  046204  (2004)    

anisotropy  term:  it  represents  the  energy  cost  for  the  domain  walls.          we   do   not   have   real   Block   or  Neel  walls,   just   their  projecHon  along  z      

Page 7: the role of exchange bias in domain dynamics control

H =

�d3R

�−Ku(R)

m2

2+

A

2(∇Rm)2 +

µ0M2s d

�d2R�m(R�)m(R)

|R−R�|3− µ0Msm(Hext −HUCS(R))

The  model  system  anisotropy  term:  its   fluctuaHons   around   an  average   value   provides   strong  pinning   points   for   the   domain  walls  

Under  the  following  approximaHons:      

•  scalar  magneHzaHon  uniform  along  the  FF  thickness  d

•  domain  walls  smaller  than  the  domain  size  

•  small  FF  thickness  d

the  magneHzaHon  in  the  FF  can  be  described  by  the  following  hamiltonian  power  expansion:    

m(r, t) = m(x, y, t)z

Ku(R) = �Ku�(1− P (x, y))

�P (R)� = 0

�P (R)P (R�)� = θδ(R−R�)

Jagla  PRB  72  094406  (2005)            Jagla  PRB  70  046204  (2004)    

anisotropy  term:  it  represents  the  energy  cost  for  the  domain  walls.          we   do   not   have   real   Block   or  Neel  walls,   just   their  projecHon  along  z      

stray  field  energy:  responsible   for   the   domain  formaHon              non   local   term  to  be   treated   in  reciprocal  space  

Page 8: the role of exchange bias in domain dynamics control

H =

�d3R

�−Ku(R)

m2

2+

A

2(∇Rm)2 +

µ0M2s d

�d2R�m(R�)m(R)

|R−R�|3− µ0Msm(Hext −HUCS(R))

The  model  system  anisotropy  term:  its   fluctuaHons   around   an  average   value   provides   strong  pinning   points   for   the   domain  walls  

Under  the  following  approximaHons:      

•  scalar  magneHzaHon  uniform  along  the  FF  thickness  d

•  domain  walls  smaller  than  the  domain  size  

•  small  FF  thickness  d

the  magneHzaHon  in  the  FF  can  be  described  by  the  following  hamiltonian  power  expansion:    

m(r, t) = m(x, y, t)z

Ku(R) = �Ku�(1− P (x, y))

�P (R)� = 0

�P (R)P (R�)� = θδ(R−R�)

Jagla  PRB  72  094406  (2005)            Jagla  PRB  70  046204  (2004)    

anisotropy  term:  it  represents  the  energy  cost  for  the  domain  walls.          we   do   not   have   real   Block   or  Neel  walls,   just   their  projecHon  along  z      

stray  field  energy:  responsible   for   the   domain  formaHon              non   local   term  to  be   treated   in  reciprocal  space  

UCS   field:   as  measured  in  the  experiment  

Page 9: the role of exchange bias in domain dynamics control

H =

�d3R

�−Ku(R)

m2

2+

A

2(∇Rm)2 +

µ0M2s d

�d2R�m(R�)m(R)

|R−R�|3− µ0Msm(Hext −HUCS(R))

The  model  system  anisotropy  term:  its   fluctuaHons   around   an  average   value   provides   strong  pinning   points   for   the   domain  walls  

Under  the  following  approximaHons:      

•  scalar  magneHzaHon  uniform  along  the  FF  thickness  d

•  domain  walls  smaller  than  the  domain  size  

•  small  FF  thickness  d

the  magneHzaHon  in  the  FF  can  be  described  by  the  following  hamiltonian  power  expansion:    

m(r, t) = m(x, y, t)z

Ku(R) = �Ku�(1− P (x, y))

�P (R)� = 0

�P (R)P (R�)� = θδ(R−R�)

Jagla  PRB  72  094406  (2005)            Jagla  PRB  70  046204  (2004)    

anisotropy  term:  it  represents  the  energy  cost  for  the  domain  walls.          we   do   not   have   real   Block   or  Neel  walls,   just   their  projecHon  along  z      

stray  field  energy:  responsible   for   the   domain  formaHon              non   local   term  to  be   treated   in  reciprocal  space  

UCS   field:   as  measured  in  the  experiment  

External  field:  uniform

 bu

t  Hme  de

pend

ent    

Page 10: the role of exchange bias in domain dynamics control

Pueng  this  approximate  hamiltonian  inside  the  LLG  equaHon  we  obtain  the  following  equaHon  of  moHon:          where  everything  is  now  in  dimensionless  units:                

the  model  contains  only  three  non-­‐independent  dimensionless  parameters  related  to  the  material  proper.es:    

∂m

∂τ= (1−m2)

�α(1− p(r))m− 1

�d2r�

m(r�)

|r− r�|3 + hext(t)− hUCS(r) + q(r, τ)

�+ β∇2

rm

The  model  system  

Jagla  PRB  72  094406  (2005)            Jagla  PRB  70  046204  (2004)            Zhirnov  Zh.Eksp.Teor.Fiz.  35  1175  (1958)    

r = R/d

τ = tγµ0Ms/ξ

hext = Hext/Ms

hUCS = HUCS/Ms

q(r, τ) = Q(R, t)/µ0Ms

�KBT = KBT/µ0M2s d

3dimensionless  posiHon    dimensionless  Hme  

dimensionless  fields  

η = θ/d3β = A/µ0M2s d

2α = �Ku�/µ0M2s

dimensionless  temperature  dimensionless  thermal  noise  

�p(r)p(r�)� = ηδ(r− r�) dimensionless  anisotropy  noise  

uniaxial  anisotropy     exchange  sHffness   anisotropy  fluctuaHons  (strength  on  the  pinning  disorder)  

Page 11: the role of exchange bias in domain dynamics control

Pueng  this  approximate  hamiltonian  inside  the  LLG  equaHon  we  obtain  the  following  equaHon  of  moHon:          where  everything  is  now  in  dimensionless  units:                

the  model  contains  only  three  non-­‐independent  dimensionless  parameters  related  to  the  material  proper.es:    

The  model  system  

Jagla  PRB  72  094406  (2005)            Jagla  PRB  70  046204  (2004)            Zhirnov  Zh.Eksp.Teor.Fiz.  35  1175  (1958)    

r = R/d

τ = tγµ0Ms/ξ

hext = Hext/Ms

hUCS = HUCS/Ms

q(r, τ) = Q(R, t)/µ0Ms

�KBT = KBT/µ0M2s d

3dimensionless  posiHon    dimensionless  Hme  

dimensionless  fields  

η = θ/d3β = A/µ0M2s d

2α = �Ku�/µ0M2s

dimensionless  temperature  dimensionless  thermal  noise  

�p(r)p(r�)� = ηδ(r− r�) dimensionless  anisotropy  noise  

uniaxial  anisotropy     exchange  sHffness   anisotropy  fluctuaHons  (strength  on  the  pinning  disorder)  

∂m

∂τ= (1−m2)

�α(1− p(r))m− 1

�d2r�

m(r�)

|r− r�|3 + hext(t)− hUCS(r) + q(r, τ)

�+ β∇2

rm

Page 12: the role of exchange bias in domain dynamics control

Pueng  this  approximate  hamiltonian  inside  the  LLG  equaHon  we  obtain  the  following  equaHon  of  moHon:          where  everything  is  now  in  dimensionless  units:                

the  model  contains  only  three  non-­‐independent  dimensionless  parameters  related  to  the  material  proper.es:    

The  model  system  

Jagla  PRB  72  094406  (2005)            Jagla  PRB  70  046204  (2004)            Zhirnov  Zh.Eksp.Teor.Fiz.  35  1175  (1958)    

r = R/d

τ = tγµ0Ms/ξ

hext = Hext/Ms

hUCS = HUCS/Ms

q(r, τ) = Q(R, t)/µ0Ms

�KBT = KBT/µ0M2s d

3dimensionless  posiHon    dimensionless  Hme  

dimensionless  fields  

η = θ/d3β = A/µ0M2s d

2α = �Ku�/µ0M2s

dimensionless  temperature  dimensionless  thermal  noise  

�p(r)p(r�)� = ηδ(r− r�) dimensionless  anisotropy  noise  

uniaxial  anisotropy     exchange  sHffness   anisotropy  fluctuaHons  (strength  on  the  pinning  disorder)  

domain  wall  energy   domain  wall  width    domain  size  

domain  morphology  ∝�

β/α∝

�αβ

∂m

∂τ= (1−m2)

�α(1− p(r))m− 1

�d2r�

m(r�)

|r− r�|3 + hext(t)− hUCS(r) + q(r, τ)

�+ β∇2

rm

Page 13: the role of exchange bias in domain dynamics control

Model  validaHon:  from  micro  to  macro  The  procedure  for  the  determinaHon  of  the  three  material  parameters  is  made  in  such  a  way  to  fit  both  macroscopic  and  microscopic  properHes  of  the  sample:  

A   good   iniHal   guess   for  α,β   and  η   makes   the  measured  domain  image  at  0  mT  a  steady  state  of  our  equaHon  of  moHon.  A  good  choice  of  α,β  and   η   is   such   that,   if   we   use   the   measured  image  as  the  iniHal  condiHon  of  our  equaHon  of  moHon  and  we   let   it  evolve   in  Hme,   it  will  not  change.    

Benassi  et  al.  (waiHng  for  PRL  rejecHon)  

Page 14: the role of exchange bias in domain dynamics control

Model  validaHon:  from  micro  to  macro  The  procedure  for  the  determinaHon  of  the  three  material  parameters  is  made  in  such  a  way  to  fit  both  macroscopic  and  microscopic  properHes  of  the  sample:  

A   good   iniHal   guess   for  α,β   and  η   makes   the  measured  domain  image  at  0  mT  a  steady  state  of  our  equaHon  of  moHon.  A  good  choice  of  α,β  and   η   is   such   that,   if   we   use   the   measured  image  as  the  iniHal  condiHon  of  our  equaHon  of  moHon  and  we   let   it  evolve   in  Hme,   it  will  not  change.    The  measured  field  from  the  UCS  distribuHon  is    also   included   in   the  equaHon  of  moHon  and   it  helps  in  stabilizing  the  domain  configuraHon.      

Benassi  et  al.  (waiHng  for  PRL  rejecHon)  

Page 15: the role of exchange bias in domain dynamics control

Model  validaHon:  from  micro  to  macro  The  procedure  for  the  determinaHon  of  the  three  material  parameters  is  made  in  such  a  way  to  fit  both  macroscopic  and  microscopic  properHes  of  the  sample:  

Than  we  ramp  up  the  external  uniform  field  and  we   trim   the   parameters   in   such   a   way   to  reproduce   the   correct   path   to   saturaHon  looking  also  at   the  MFM  taken  at  100  mT,  200  mT  and  300  mT.  

Benassi  et  al.  (waiHng  for  PRL  rejecHon)  

Page 16: the role of exchange bias in domain dynamics control

Model  validaHon:  from  micro  to  macro  The  procedure  for  the  determinaHon  of  the  three  material  parameters  is  made  in  such  a  way  to  fit  both  macroscopic  and  microscopic  properHes  of  the  sample:  

Once   that   the  microscopic   properHes   are   well  reproduced  we  can  check  the  macroscopic  ones  (hysteresis  loops).  We  can  sHll  trim  a  bit  the  model  parameters  to  adjust  the  fine  detail.  Eventually   we   have   to   go   back   and   control  again  the  microscopic  behavior.        The  loops  were  measured  with  a  sample  cooled  in   a   saturaHng   field   so   the   printed   UCS  distribuHon  is  different  from  the  previous  one.  

Benassi  et  al.  (waiHng  for  PRL  rejecHon)  

Page 17: the role of exchange bias in domain dynamics control

Model  validaHon:  from  micro  to  macro  The  procedure  for  the  determinaHon  of  the  three  material  parameters  is  made  in  such  a  way  to  fit  both  macroscopic  and  microscopic  properHes  of  the  sample:  

Benassi  et  al.  (waiHng  for  PRL  rejecHon)  

αinit = 6.25

βinit = 0.85

ηinit = 1.5× 10−4

α = �Ku�/µ0M2s = 6.6 → �Ku� = 3.1× 106J/m3

β = A/µ0M2s d

2 = 0.88 → A = 2.2× 10−10J/m

η = 1.88× 10−4

perfectly  in    the  expected  range  

too  big  but  our  1D  walls  are  less  expensive  than  a  real  block  wall  and  A  must  compensate!  

Page 18: the role of exchange bias in domain dynamics control

Path  to  saturaHon:  the  full  dynamics    Now  that  the  parameters  are  fixed,  the  theoreHcal  model  allows  us  to  access  the  full  dynamics  in  Hme,  we  can  thus  invesHgate  the  domain  behavior  with  more  than  few  MFM  images      

Benassi  et  al.  (waiHng  for  PRL  rejecHon)  

As   demonstrated   by   the   experiments   the   domains,  retracHng  with  increasing  external  field,  will  try  to  avoid  the  frustrated  F/AF  coupling  regions.      

Page 19: the role of exchange bias in domain dynamics control

UnmounHng  the  machinery  

Benassi  et  al.  (waiHng  for  PRL  rejecHon)  

Now  we  can  switch  off  separately  the  different  hamiltonian  terms  and  try  to  understand  and  quanHfy  their  contribuHon:    

The  UCS  distribuHon  alone  has  not  enough  strength  to  pin  the   domains   and   even   at   0   mT   the   shape   of   the   steady  configuraHon  is  quite  different  from  the  original  one.    The  saturaHon  occurs  too  early!      The  anisotropy  fluctuaHons  have  enough  strength  to  keep  the  iniHal  configuraHon  pinned,  however,  without  the  help  of  the  UCS  local  field,  the  pinning  sHll  occurs    too  early.  

Page 20: the role of exchange bias in domain dynamics control

Now  we  can  also  try  to  predict  which  is  the  effect  of  a  different  UCS  distribuHon  on  the  exchange-­‐bias  (EB)  effect  and  on  the  coercivity  of  the  FF.                                    •  As  expected,  switching  off  the  UCS  field  the  EB  effect  goes  to  zero    

•  Doubling  the  average  value  of  the  UCS  field,  without  changing  the  fluctuaHons  strength,  increases  the  EB  effect  without  affecHng  the  coercivity  

•  Doubling  the  fluctuaHons  of  the  UCS  field,  without  shiking  its  average,  increase  strongly  the  coercivity  with  minor  changings  in  the  EB  loop  shik.  

Some  predicHons  on  the  macroscopic  properHes  

calc. 10 K

-0.5

0.0

0.5

1.0

-1.0-0.4 -0.2 -0.1 0.0 0.1 0.2 0.4-0.3 0.3

no UCS

double UCS averagedouble UCS fluctuations

Benassi  et  al.  (waiHng  for  PRL  rejecHon)  

Page 21: the role of exchange bias in domain dynamics control

Up  to  now  the  fluctuaHons  of  the  uniaxial  anisotropy  have  been  considered  to  be  uncorrelated  (white  noise),  however  they  have  something  to  do  with  the  granularity  of  the  sample.  Something  more  about  the  strength  and  the  correlaHon  of  these  fluctuaHon  can  be  inferred  from  Barkhausen  noise  measurements.                                  The  presence  of  Chromium  atoms  in  the  AF  decouples  the  magneHc  moment  of  neighboring  grains,  increasing  the  UCS  field  by  the  40%.  The  model  will  be  used  to  study  this  new  sample  in  which  the  role  of  the  UCS  map  as  been  enhanced.  ParHcular  aYenHon  will  be  given  to  the  return  point  memory  effects.    The  code  is  easily  parallelizable  allowing  for  the  descripHon  of  lager  system  or  for  the  coupling  of  two  interacHng  magneHzed  films        

Further  developments  

Benassi  et  al.  PRB  84  214441  (2011)