Physica B 283 (2000) 162}166


                 Polarized neutron re#ectivity characterization of
                                    weakly coupled Co/Cu multilayers
              J.A. Borchers  *, J.A. Dura , C.F. Majkrzak , S.Y. Hsu , R. Lolee ,
                                                       W.P. Pratt , J. Bass 
             NIST Center for Neutron Research, National Institute of Standards and Technology, 100 Bureau Drive STOP 8562,
                                                       Gaithersburg, MD 20899-8562, USA
                        Department of Physics and Astronomy, Michigan State University, East Lansing, MI 48824, USA


Abstract

   Room temperature resistivity studies on (1 1 1) Co(6 nm)/Cu(6 nm) multilayers with weak interlayer coupling yield
a giant magnetoresistance (GMR) of several percent for the as-prepared state relative to the saturated (i.e., large magnetic
"eld) state. After application of a magnetic "eld, the magnetoresistance for the coercive state is only half to two-thirds as
large. Using specular and o!-specular polarized neutron re#ectivity, we have determined the magnetic structure of these
multilayers in the as-prepared and coercive states. Measurements of as-prepared samples show a strong antiparallel
correlation of in-plane ferromagnetic Co domains across the Cu interlayers. At the coercive "eld, the Co domains are
uncorrelated along the growth direction. Thus, the larger GMR for the as-prepared state arises from long-range
antiparallel magnetic order along the growth-axis direction that is destroyed upon application of a magnetic "eld. For
both the as-prepared and coercive states, the size of the in-plane ferromagnetic domains is approximately 0.5}1.5  m.
These domains give rise to pronounced di!use scattering in rocking curves through the antiferromagnetic peak
position.   2000 Elsevier Science B.V. All rights reserved.

Keywords: Giant magnetoresistance; Magnetic multilayer; O!-specular scattering; Magnetic domains; Polarized neutron re#ectivity



   Giant magnetoresistance (GMR) has been observed in                     whether the magnetic order associated with MR(0) or
magnetic multilayers [1}3] with alternating ferromag-                     MR(H!) is closer to the antiparallel state.
netic and non-magnetic metallic layers. In these mater-                      Here we describe specular and di!use polarized neu-
ials, a maximum in the magnetoresistance (MR), is                         tron re#ectivity (PNR) studies of Co/Cu multilayers with
associated with an antiparallel alignment of adjacent                     t! "6 nm. In the as-prepared state, we observe a dis-
ferromagnetic layers. The resistance decreases when an                    tinct magnetic re#ection at the half-order position.
external magnetic "eld aligns the magnetization of the                    MR(0) thus originates from antiparallel correlations
ferromagnetic layers parallel to each other. In multilayers               among the ferromagnetic Co layers. Application of
with thick non-magnetic layers (t '4}5 nm), the inter-                    a small magnetic "eld irreversibly destroys this re#ection.
layer magnetic coupling is weak. In this limit, the mag-                  We observe that the peak in the magnetoresistance at the
netoresistance, (MR(0)) for the as-prepared multilayer is                 coercive "eld MR(H!) arises from a randomization of
often larger than the maximum value obtained at the                       the Co layer domains along the growth direction. For
coercive "eld MR(H!) after saturation [4,5]. The initial                  both states, the presence of di!use magnetic scattering
MR(0) is not recovered after "eld cycling [6,7]. For                      reveals that the Co moments, which preferentially lie in
proper analysis of MR data, it is important to know                       the growth plane, are ordered in small domains with
                                                                          an average in-plane size of 0.5}1.5  m. (This domain
                                                                          size has been con"rmed by scanning electron microscopy
  * Corresponding author. Tel.: #1-301-97-56-597.                         with polarization analysis [SEMPA] measurements
  E-mail address: julie.borchers@nist.gov (J.A. Borchers)                 [5]). These results demonstrate that the combination of

0921-4526/00/$- see front matter   2000 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 1 - 4 5 2 6 ( 9 9 ) 0 1 9 2 0 - 1



                                       J.A. Borchers et al. / Physica B 283 (2000) 162}166                                     163

specular and di!use PNR provides a full pro"le of the
magnetization along the growth direction and within the
sample plane.
  For this study, we examined several multilayers
of nominal composition [Co(6 nm)"Cu(6 nm)], with
N"10 or 20 bilayers. The samples were sputtered onto
Si substrates as described elsewhere [4]. The "eld de-
pendence of the magnetization and magnetoresistance
was measured at room temperature for comparable sam-
ples grown at the same time. The current-in-plane MR(0)
was typically 4}7%, and MR(H!) was half to two-thirds
as large [4,5].
  Fits to specular X-ray re#ectivity data con"rm that the          Fig. 1. Specular PNR as a function of QX"(4 / ) sin   for
Co and Cu interlayers are well modulated. Typical full-            [Co(6 nm)"Cu(6 nm)]   in the as-prepared state in a 1.5 Oe "eld.
widths of the interfacial roughness (averaged over the             The di!use scattering has been subtracted, and the data have
sample plane) were +1.5 nm for our original samples,               been corrected for the polarization e$ciencies. The open and
compared to 4.0 nm for samples that were prepared later.           shaded circles correspond to the (!!) and (##) data respec-
To characterize the structural disorder within the sample          tively. The open and shaded squares designate the (#!) and
                                                                   (!#) data. The arrows point to the appropriate vertical axis
plane, we probed the di!use scattering by performing               for the NSF and SF data.
rocking curves at various QX positions [8,9]. For samples
with interfacial-roughness widths of 1.5 nm, the di!use
scattering is insigni"cant relative to the specular. Thus,         PNR scans  along the QX direction for the
the interfaces are relatively smooth, and the limited inter-       [Co(6 nm)"Cu(6 nm)]   sample. As expected, the NSF
facial  mixinga is due to interlayer interdi!usion or              cross sections show a "rst-order structural superlattice
atomic-scale roughness. In contrast, the samples with              peak at QX"0.057 As\ +2 /d where d"11.4 nm is the
interfacial widths of 4.0 nm exhibit pronounced di!use             bilayer repeat distance. A pronounced peak is also evi-
scattering. The measured interfacial widths probably               dent at the half-order position (QX"0.031 As\ +2 /2d).
originate from larger structures such as islands, corruga-         A "t to the data indicates that +1.5% of the Co mo-
tion or steps within the growth plane. In the rest of this         ments are ordered in large ferromagnetic domains (in-
paper we focus on a [Co(6 nm)"Cu(6 nm)]   multilayer               plane size '100  m) that are aligned antiparallel across
that shows some di!use scattering.                                 the Cu. The orientation of these domains in the sample
  We performed the specular and di!use PNR studies on              plane is likely random since the magnetic intensity is
the NG-1 re#ectometer at the NIST Center for Neutron               evenly distributed in all four cross sections. The magnetic
Research at room temperature. Fits to the specular data            SF scattering at the "rst-order position indicates that at
yield a pro"le of the magnetization as a function of depth         least an additional 2.5% of the Co moments are aligned
for regions of the multilayer with ferromagnetic domains           parallel across the Cu. Along the growth axis, sets of
having in-plane dimensions larger than the coherence               distinct domains aligned parallel and antiparallel may
length of the neutrons, +100  m [10]. Di!use measure-              coexist or there may be a collection of non-collinear
ments are sensitive to the in-plane length scale of the            domains [12]. The latter possibility is unlikely because
structural and magnetic roughness [8,9,11] and of the              the "rst-order SF magnetic peak is broader than the
magnetic domains [11]. They also provide information               half-order peak, suggesting a di!erence in the growth-
about the interlayer correlations among ferromagnetic              axis coherence length of the parallel and antiparallel
domains having in-plane dimensions smaller than                    magnetic structures. (We note that the coherence of the
+100  m. For specular and di!use experiments, we                   parallel magnetic structure is also shorter than the struc-
measured all four cross sections, (!!), (##), (#!)                 tural coherence length as indicated by the di!erence
and (!#). (The # and ! signs indicate polarizations                between the full-widths of the SF and NSF "rst-order
of the incident and scattered neutrons parallel or anti-           re#ections).
parallel to the guide "eld.) The (!!) and (##) non-                   This analysis of the specular PNR data accounts for
spin-#ip (NSF) data depend on the chemical structure, as           only about 4% of the total Co magnetization. Because
well as the projection of the in-plane magnetization par-          specular re#ectivity averages over features with in-plane
allel to the guide "eld. The (#!) and (!#) spin-#ip
(SF) cross sections originate from the projection of the
in-plane magnetization perpendicular to the "eld [10].                  Di!use data have been subtracted from the total re#ectivity
The instrumental polarization e$ciency is '95%.                    to give the specular re#ectivity. This is exact only if the
  We "rst characterized the magnetic structure of                  sample has regions with large and small domains that scatter
each sample in the as-prepared state. Fig. 1 shows specular        respectively in the specular and di!use.



164                                      J.A. Borchers et al. / Physica B 283 (2000) 162}166
















Fig. 2. Total PNR (shaded symbols) relative to the di!use
scattering (open symbols) for [Co(6 nm)"Cu(6 nm)]   and
[Co(6 nm)"Cu(6 nm)]   (C2) in the as-prepared state. The dif-
fuse scattering was measured by o!setting the angle   by a small
amount and then scanning QX. The circles and squares corres-
pond to (!!) and (##) data, respectively. The up and down
triangles mark (#!) and (!#). No data corrections have
been made. The arrows designate the vertical axis for each cross
section. The inset shows the magnetic structure.

                                                                     Fig. 3. Transverse QV scans at the half-order position
dimensions less than +100  m, the remaining moments                  (QX"0.314 As\ ) for [Co(6 nm)"Cu(6 nm)]   in the as-prepared
appear to be ordered in small, discrete domains spread               (shaded symbols) and saturated (H"400 Oe) states (open sym-
over the growth plane.  For the as-prepared state, a com-            bols). The (##) and (!!) NSF cross sections (circles and
parison of the di!use scattering (i.e., Q                            diamonds respectively) are shown in (a) and the (!#) and
                                          X scan with   o!set
by 0.13) to the total re#ectivity [Fig. 2(a)] shows that the         (#!) SF cross sections (squares and triangles) are shown in (b).
total scattering is dominated by di!use at the half-order
position. These results indicate that most of the small              all of the samples seem thus to order in domains with
domains in these samples are oriented antiparallel [inset            small in-plane dimensions ((100  m) that are randomly
Fig. 2(a)] along the growth direction. The narrow                    oriented across the Cu layers [5]. As a consequence, the
QX width of the di!use re#ection indicates that the anti-            GMR is not as large as that measured for the as-prepared
parallel order is correlated through the entire multilayer           antiparallel state. This irreversible loss of antiparallel
thickness. For a second [Co(6 nm)"Cu (6 nm)]   sample,               magnetic order is analogous to the magnetic structure
the half-order re#ection shown in Fig. 2(b) is entirely of           reported for weakly coupled Fe/Nb multilayers [13].
di!use character, and the QX width is greater. We note                  In the as-prepared and coercive magnetic states,
that the data for most of our samples resembled those in             additional details of the in-plane magnetic structure can
Fig. 2(b). In these multilayers the growth-axis coherence            be gained from transverse QV scans centered at the
is limited to a few bilayers. The as-prepared state respon-          half-order position (QX "0.0314 As\ ). Fig. 3(a) and (b)
sible for the maximum GMR is thus dominated by small,                show NSF and SF data for [Co(6 nm)"Cu(6 nm)]   in the
ferromagnetic Co domains that are aligned antiparallel               as-prepared and saturated states. Pronounced dips are
across the Cu layers.                                                centered at the sample angles  "0 and  "2  where
  Consistent with the magnetoresistance data, this anti-             either the incident or scattered beam is parallel to the
parallel order is irreversibly destroyed by the application          sample face and is thus refracted away from the detector
of a "eld. After saturation in a large "eld, the half-order          [14]. The as-prepared data in Fig. 3 are otherwise com-
re#ection in the PNR data for all of the samples con-                posed of a sharp specular re#ection at QV"0 As\  on top
sidered is absent at the coercive "eld (H!+50 Oe) in both            of a broad, di!use peak. The coercive state data look
the total re#ectivity and di!use data. No other magnetic             similar to the as-prepared data [5]. We conclude that the
feature is evident. In the coercive state, the Co moments in         di!use scattering in the as-prepared and coercive states
                                                                     originates entirely from magnetic features within the
                                                                     sample plane because it disappears when the Co mo-
    Specular PNR is also insensitive to spins perpendicular to       ments are aligned in a saturating "eld of 400 Oe (Fig. 3).
the growth plane, but magnetization measurements indicate that       In addition, much of the di!use scattering is SF, which
the Co moments lie in-plane.                                         has no chemical contributions.



                                        J.A. Borchers et al. / Physica B 283 (2000) 162}166                                          165

                                                                    position does increase in a 400 Oe "eld, as shown in
                                                                    Fig. 4. At least part of the di!use intensity for this sample
                                                                    can thus be attributed to magnetic interfacial roughness.
                                                                    This result agrees with the observation of structural dif-
                                                                    fuse scattering in transverse X-ray scans. However, other
                                                                    samples do not show substantial chemical roughness
                                                                    within the sample plane, and the di!use magnetic scatter-
                                                                    ing at the "rst-order position subsequently disappears in
                                                                    high "elds. Most of the magnetic di!use scattering from
                                                                    our samples is apparently characteristic of the micron-
                                                                    size in-plane magnetic domains directly observed by
                                                                    SEMPA [5].
                                                                       Our PNR measurements have shown that the larger
Fig. 4. Transverse QV scans at the "rst-order position              MR(0) for as-prepared [Co(6 nm)"Cu(6 nm)]
(Q                                                                                                                        , multi-
  X"0.575 As\ ) for [Co(6 nm)"Cu(6 nm)]   in the as-prepared        layers arises from antiparallel ordering between fer-
state (shaded symbols) relative to the saturated state (open
symbols) in H"400 Oe. Only the NSF cross sections are               romagnetic Co domains across the Cu layers. This
shown. The 400 Oe (!!) data (circles) and (##) data                 antiparallel state occurs only after growth and cannot be
(squares) are split because the Co moments are aligned parallel     restored upon application of a "eld. The smaller MR(H!)
to the "eld. Insets (a) and (b) show multilayers with in-plane      at the coercive "eld is associated with random Co do-
domains and magnetic roughness, respectively.                       mains. In both the as-prepared and coercive states, the
                                                                    typical in-plane domain size is +0.5}1.5  m as deter-
                                                                    mined from the full-width of the magnetic di!use scatter-
      An estimate of the in-plane magnetic correlation              ing. Our measurements of the magnetic di!use scattering
length can be obtained from the transverse full-width of            demonstrate the power of PNR as a probe of domains in
the half-order di!use peak. This value corresponds either           buried magnetic layers and are a complement to other
to the size of the ferromagnetic domains and/or to the              neutron and X-ray resonant scattering results for related
correlation length of the magnetic roughness [11].                  thin "lm systems [15}18].
Lorentz "ts of as-prepared and coercive data similar to
those shown in Fig. 3(a) and (b) give correlation lengths
of 0.5}1.5  m. (A direct analysis of the transverse data            Acknowledgements
using a kinematic or dynamical formalism [8,9,11] would
yield a more accurate estimate, but this work is still in              Research supported by NSF DMR 98-20135, MRSEC
progress.) SEMPA imaging of the "rst and second Co                  Program DMR 98-09688, MSU-CFMR and Ford Re-
layers in the [Co(6 nm)"Cu(6 nm)]                                   search Laboratory.
                                          show clearly that
the Co moments are ordered in irregularly shaped mi-
cron-sized domains for the as-prepared state [5]. The
observation of strong magnetic di!use scattering in the             References
PNR data (Fig. 3) indicates that these domains are dis-
tributed throughout the entire sample.                                [1] M.N. Baibich, J.M. Broto, A. Fert, F. Nguyen Van Dau,
      Transverse scans through the "rst-order Q                           F. Petro!, P. Etienne, G. Creuzet, A. Friederich, J.
                                                   X position
provide information about other in-plane magnetic                         Chazelas, Phys. Rev. Lett. 61 (1988) 2472.
structures that may contribute to the di!use magnetic                 [2] S.S.P. Parkin, N. More, K.P. Roche, Phys. Rev. Lett. 64
                                                                          (1990) 2304.
scattering. By comparing the "eld dependence of the                   [3] S.S.P. Parkin, R. Bhadra, K.P. Roche, Phys. Rev. Lett. 66
di!use peak at "rst- and half-order, we can di!erentiate                  (1991) 2152.
between in-plane domain formation and magnetic                        [4] W.P. Pratt Jr., S.-F. Lee, J.M. Slaughter, R. Loloee,
roughness (i.e., structural roughness between the Cu and                  P.A. Schroeder, J. Bass, Phys. Rev. Lett. 66 (1991) 3060.
ferromagnetic Co layers). If the Co moments order com-                [5] J.A. Borchers, J.A. Dura, J. Unguris, D. Tulchinsky,
pletely in small domains at low "elds [Fig. 4 inset (a)],                 M.H. Kelley, C.F. Majkrzak, S.Y. Hsu, R. Loloee,
any di!use intensity at the half- and "rst-order positions                W.P. Pratt Jr., J. Bass, Phys. Rev. Lett. 82 (1999) 2796.
should disappear when the Co domains are aligned par-                 [6] P.S. Schroeder, S.-F. Lee, P. Holody, R. Loloee, Q. Yang,
allel in a saturating "eld. If instead the interfaces are                 W.P. Pratt Jr., J. Bass, J. Appl. Phys. 76 (1994) 6610.
rough across the growth plane [Fig. 4 inset (b)], any                 [7] Ch. Rehm, F. Klose, D. Nagengast, B. Pietzak, H. Maletta,
                                                                          A. Weidinger, Physica B 221 (1996) 377.
low-"eld di!use scattering at half-order should transfer              [8] S.K. Sinha, E.B. Sirota, S. Garo!, H.B. Stanley, Phys. Rev.
to the "rst-order position as the Co spins align in high-                 B 38 (1988) 2297.
"elds [11]. In fact, the NSF (!!) di!use scattering for               [9] V. HolyH, T. Baumbach, Phys. Rev. B 49 (1994) 10 668.
the [Co(6 nm)"Cu(6 nm)]   sample at the "rst-order                  [10] C.F. Majkrzak, Physica B 221 (1996) 342.



166                                          J.A. Borchers et al. / Physica B 283 (2000) 162}166

[11] S.K. Sinha, in: D.A. Neumann, T.P. Russell, B.J. Wuensch            [14] C. MuKnster, T. Salditt, M. Vogel, R. Siebrecht, J. Peisl,
       (Eds.), Neutron Scattering in Materials Science II, Vol.                Europhys. Lett. 46 (1999) 486.
       376, Mat. Res. Soc. Symp. Proc. Materials Research So-            [15] G.P. Felcher, Physica B 192 (1993) 137.
       ciety, Pittsburgh, 1995, p. 175.                                  [16] M. Takeda, Y. Endoh, H. Yasuda, K. Yamada, A. Kamijo,
[12] A. Schreyer, J.F. Ankner, Th. Zeidler, H. Zabel, C.F.                     J. Mizuki, J. Phys. Soc. Japan 62 (1993) 3015.
       Majkrzak, M. SchaKfer, P. GruKnberg, Europhys. Lett. 32           [17] J.F. MacKay, C. Teichert, D.E. Savage, M.G. Lagally,
       (1995) 595.                                                             Phys. Rev. Lett. 77 (1996) 3925.
[13] Ch. Rehm, D. Nagengast, F. Klose, H. Maletta, A. Weidin-            [18] Y.U. Idzerda, V. Chakarian, J.W. Freeland, Phys. Rev.
       ger, Europhys. Lett. 38 (1997) 61.                                      Lett. 82 (1999) 1562.