Sie sind auf Seite 1von 9

ARTICLE IN PRESS

Journal of Magnetism and Magnetic Materials 321 (2009) 29022910

Contents lists available at ScienceDirect

Journal of Magnetism and Magnetic Materials


journal homepage: www.elsevier.com/locate/jmmm

Effect of material selection and background impurity on interface property


and resulted CIP-GMR performance
Xilin Peng , Augusto Morrone, Konstantin Nikolaev, Mark Kief, Mark Ostrowski
RHO, Seagate Technology, 7801 Computer Ave. South, Bloomington, MN 55435, USA

a r t i c l e in f o a b s t r a c t

Article history: In this paper, we investigated the effect of background base pressure, wafer-transferring time between
Received 21 February 2009 process modules, and stack layer material selection on the current-in-plane giant magneto-resistive
Available online 3 May 2009 (CIP-GMR) interface properties and the resulted CIP-GMR performance. Experimental results showed
Keywords: that seed layer/AFM interface, AFM/pinned layer (PL) interface, pinned layer/Ru interface, and reference
CIP-GMR layer (RL)/Cu spacer interface are among the most critical ones for a CIP-GMR device. By reducing the
Interface background impurity level (water moisture and oxygen), optimizing the wafer process ow sequence,
Lattice mismatch and careful stack-layer material selection, such critical interfaces in a CIP-GMR device can be preserved.
Seed layer Consequently, a much robust GMR performance control can be achieved.
AFM
& 2009 Elsevier B.V. All rights reserved.
SAF
TEM

1. Introduction From new material development and interface engineering/


controlling point of view, CIP-GMR and CPP-GMR devices
To meet the ever-increasing demand for higher areal density (including metallic CPP-GMR and MgO-TMR) bear a lot of
storage requirement, an increased output voltage from the storage similarity. For example, both CPP-GMR and CIP-GMR devices have
device is required. Various efforts have been undertaken to very close bottom-half structure (from seed layer, through
improve the traditional current-in-plane giant magneto-resistive synthetic anti-ferromagnetic (SAF) to the barrier/spacer),
(CIP-GMR) device amplitude since its invention [1] and introduc- although the seed layer requirements for CIP-GMR and CPP-
tion to the storage industry by IBM [2] a decade ago. Insertion of GMR are not the same. Seed layer has to meet the requirement for
an oxide layer on top of the free layer (FL) and/or within the minimized current shunting for CIP-GMR [5], in addition to
reference layer [3,4] have been found effective in boosting such providing good texture development for other layers grown onto
CIP-GMR device output, due to the specular reection of electrons it. In CPP-GMR design, on the other hand, the seed layer with
at nano-oxide layer (NOL) interface, which preserves their spin reduced parasitic resistance (metallic CPP-GMR in particular)
polarization. Using the rst Cu anti-ferromagnetic (AFM) coupling should be the primary choice. Moreover, the CIP-GMR device
peak (at about 12 A) as spacer layer in a trilayer design [5] can performance can be easily measured on sheet lm level without
further improve the signal output for CIP-GMR by reducing the stringent nano-patterning requirement, as compared to CPP-GMR
electrical shunting via the Cu spacer layer. However, the current- devices, and thus, fast device performance feedback is possible.
perpendicular-to-plane (CPP) tunneling magneto-resistive (TMR) This provides substantial benet for new material selection,
devices have replaced the current-in-plane giant magneto- process development, and fundamental study.
resistive devices and become the main stream for the storage Since the material layers in typical GMR devices are approach-
application in the last few years due to the signicant progress in ing atomic level (ranged from angstroms to nanometers), interface
MgO-TMR devices, which achieved over 200% of MR ratio [6,7]. engineering/control is as important as bulk material selection, if
For the new generations of storage heads beyond 1 Tb/in2, either not more crucial. It is therefore benecial if we can leverage the
very lower resistance  area product (RA) MgO-TMR, or CPP-GMR CIP-GMR interface property information for CPP-GMR device.
systems with [8] or without current connement path are Those critical interfaces for both CIP- and CPP-GMR systems
expected. This requires new materials and stack layer interface include, but not limited to: (1) seed layer/anti-ferromagnetic layer
design innovation. interface, (2) AFM/pinned layer (PL) interface, (3) synthetic anti-
ferromagnetic interfaces (ferromagnetic layer/Ru/ferromagnetic
layer), and (4) reference layer/spacer interface. Those interfaces
 Corresponding author. can signicantly affect the AFM texture development (and the
E-mail address: xilin.peng@seagate.com (X. Peng). corresponding AFM pinning eld and blocking temperature), SAF

0304-8853/$ - see front matter & 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmmm.2009.04.047
ARTICLE IN PRESS
X. Peng et al. / Journal of Magnetism and Magnetic Materials 321 (2009) 29022910 2903

Fig. 1. Critical interfaces for specular, simple CIP-GMR, single NOL-GMR, and DNOL-GMR stacks employed in this study and the important interfaces, which may be affected
by moisture and O2 exposure.

coupling strength, interface spin-dependent electron scattering, bi-layer stack, simple SAF, specular stack, simple bottom spin
and thus the achievable overall GMR performance. For example, it valve (BSV), to BSV with single and dual nano-oxide layers (DNOL)
has been found that the seed layer material selection [913], as detailed below
critical seed layer thickness [10,11,14], oxygen impurity content
[15] had strong impact on the AFM texture growth, its grain size,  bi-layer stack for AFM pinning: seed/AFM (PtMn or IrMn)/
and the corresponding GMR ratio of the lm stack. Background CoFex/Cap layer
impurity effect can be more profound when high oxygen-afnity  isolated SAF structures: seed/CoxFeyBz/Ru/CoxFeyBz/Cap layer
seed layer (like Ta) is used. Mao et al. [3] have reported that IrMn  specular stack with layer structure: Ta 30 A/NiFe 30 A/CoFe
grown onto oxide substrates (TaO, CoFeO, etc.) had much weaker 15 A/(NOL)/Ta 30 A
(111) texture than that deposited onto metallic Ta and Ta/CoFe  simple bottom spin valve with layer structure: seed layer/
seed layers. PtMn/CoFe/Ru/CoFe/Cu/CoFeNiFe/Cap layer
In this article, we have systematically investigated such critical  single nano-oxide layer BSV (NOL-BSV) with layer structure:
interfaces for the CIP-GMR devices and their dependence on seed/PtMn/CoFe/Ru/CoFe/Cu/CoFe/NOL1/Cap layer
the material choices, background impurity, and process ow  dual nano-oxide layer BSV (DNOL-BSV) with layer structure:
sequence. This learning is also applicable to the CPP-GMR seed/PtMn/CoFe/Ru/CoFeNOL1CoFe/Cu/CoFe/NOL2/Cap
(including TMR) device stack design. Due to the length limitation, layer
we were not able to cover other critical interfaces, such as the
spacer (barrier)/free layer interface, FL/Cap layer interface, and the All the GMR samples and bi-layer stacks were annealed at
patterned device side wall/permanent magnet (PM) (for CIP-GMR) 300 1C, 2 T eld for 4 h under protective N2 environment.
interface or device side wall/isolation layer (for CPP-TMR) inter- Magneto-optical Kerr effect (MOKE) measurement was employed
face in this study, though they are as crucial as other mentioned for AFM pinning eld and SAF coupling characterization.
interfaces. Two point probing was used to measure the GMR stack
sheet resistance and magneto-resistance change under external
eld sweep from 2000 to +2000 Oe. Further high-resolution
2. Experimental details transmission electron microscopy (TEM) was used to characterize
the interface and lattice growth for selected layer and layer
All lms were prepared via direct current (DC) magnetron interfaces.
sputtering using ultra-high vacuum (UHV) system with a base Although an ultra low pressure (107 Torr) and very small
pressure of 109 Torr, except for those experiments where a dose (a few to a few tens Langmuir (L). Note: 1 L 106 Torr s) of
controlled background pressure/impurity study was specied in O2 exposure at interfaces between Cu/CoFe or within the RL CoFe
detail in the context. Silicon with 3000 A of thermal oxide was has been found being benecial to minimize the interface
used as substrate. Crystal structures for selected single lms and intermixing. This commonly referred oxygen surfactant effect
lm stacks were characterized by X-ray diffraction (XRD) with Cu has been successfully employed to alter the interface roughness,
target at 45 kV and 40 mA current on Philips XRD machine. thus altering the interlayer exchange coupling and GMR ratio.
To fully quantify the impact of material selection (seed, AFM, However, further increase in the O2 exposure is found to be
CoFe, SAF, etc.) and the background impurity on the critical detrimental.
interfaces for various CIP-GMR stack congurations, we have Here, we have intentionally altered some critical interfaces by
evaluated different CIP-GMR congurations ranged from simple either controlling O2 exposure at 106 Torr chamber pressure or
ARTICLE IN PRESS
2904 X. Peng et al. / Journal of Magnetism and Magnetic Materials 321 (2009) 29022910

Table 1
Crystal structure, lattice constants [16,17], and the corresponding densely packed crystal plane d-spaces for various metals and their alloys commonly used for CIP-GMR
devices.

Material Structure Lattice constant d (111) Lattice mismatch

a (A) c (A) Relative to IrMn (%) Relative to PtMn (%) Relative to CoFe (%)

IrMn fcc 3.78 2.67 0 1.4 6.6


PtMn fct 3.97 3.70 2.71 1.5 0 8.1
CoFe10 fcc 3.55 2.51 6.2 7.5 0
NiFe20 fcc 3.54 2.50 6.4 7.8 0.3
Cu fcc 3.61 2.56 4.4 5.7 1.9
Ir fcc 3.84 2.71 1.6 0.1 8.2
Pd fcc 3.89 2.75 2.9 1.4 9.7
Pt fcc 3.92 2.77 3.8 2.3 10.6
NiCr40 fcc 3.56 2.52 5.8 7.2 0.4
NiFe12Cr40 fcc 3.56 2.52 5.8 7.2 0.4
Al fcc 4.05 2.86 7.1 5.6 14
d (0 0 0 1)

Ti hcp 2.92 4.67 2.92 9.2 7.7 16.4


Ru hcp 2.70 4.28 2.70 1.2 0.3 7.8
Co hcp 2.51 4.07 2.51 6.1 7.4 0.1
Os hcp 2.74 4.32 2.74 2.3 0.9 9.1
d (11 0)

Ta bcc 3.31 2.87 7.2 5.7 14.3


Cr bcc 2.88 2.50 6.6 7.9 0.4
Fe bcc 2.87 2.49 7.0 8.4 0.9

by controlling vacuum base pressure through baking procedures. 9000


The resulted magnetic and electrical performance (coupling eld, Ta100A
8000
GMR, sheet resistance, and delta R) change of such partial or full Ru100A
GMR stacks were explained in the light of critical interface 7000 Ta50ARu100A
Intneisty, a. u.

variation. By understanding the importance of such interface 6000


engineering/control for CIP-GMR devices, CIP-GMR devices with 5000
high MR ratio and better magnetic stability (pinning eld, SAF
coupling for instance) have be enabled via correct material 4000
selection, impurity contamination level control and process ow 3000
sequence optimization. 2000
1000
0
3. Results and discussion
-1000
35 40 45 50
3.1. Important interfaces for GMR system Diffraction angle, degrees
As we have discussed previously that the interface engineering Fig. 2. XRD patterns for 100 A Ta, 100 A Ru, and 50 A Ta+100 A Ru single lms
and control for CIP-GMR is crucial since the layer thickness is grown on thermal oxide silicon wafers. (Cu Ka 45 kV, 40 mA.) The sharp peak at
42.21 correspond to Ru.
approaching atomic level in these lm stacks. Fig. 1 highlighted all
the critical interfaces for various lm stack congurations from
specular stack to DNOL-BSVs. These interfaces include: seed/AFM
interface, AFM/pinned layer interface, SAF interfaces, reference The lattice mismatch can cause the lm stack distortion,
layer/Cu spacer interface, and Cu/free layer interfaces. Solid unstability and interface irregularity/defects [16]. A good seed
arrows in Fig. 1 highlighted these interfaces. Interfaces above layer usually enhances the nucleation and provides least lattice
the free layer will not be discussed here. mismatching for the growing layer. Table 1 lists some selected
seed layer materials and their lattice mismatching percentage
relative to IrMn, PtMn, and CoFe is calculated based on the inter-
3.2. Seed layer requirements atomic distance within the most densely packed crystal planes
((111) for fcc, (0 0 0 1) plane for hcp and (11 0) plane for bcc
3.2.1. Lattice matching vs. texture development crystals, respectively). Judging from lattice mismatching data
The primary role of seed layer is to promote the designed only, Ir, Pd, Pt, Ru, and Cu appear to be good seed layer candidates
texture development of the top layers grown onto it. Simple seed for AFM (PtMn and IrMn), followed by NiCr, NiFeCr, CoFe,
layer (like NiCr, NiFeCr, etc.) and bi-layer seed layer (such as Ta/Ru, and NiFe.
Ta/NiCr, Ta/NiFeCr, NiCr/CoFe, NiCr/NiFe, NiFeCr/CoFe, and NiFeCr/ Ta is an interesting seed layer material, it usually exists in
NiFe) have been widely employed for different designs. For bottom amorphous or nano-crystal status from XRD pattern as shown in
spin valve-type CIP-GMR devices, both IrMn and PtMn AFMs have Fig. 2. It provides a very good wetting base for Ru growth. Zheng
also been used. et al. [9] reported that Ta/Ru provided much better texture growth
ARTICLE IN PRESS
X. Peng et al. / Journal of Magnetism and Magnetic Materials 321 (2009) 29022910 2905

5500 Seed layer = NiFeCr x+/CoFe10A


NiC r10/IrMn100 30

GMR (%) or Rs (ohm/sq)


NiC r30/IrMn100
4500 NiC r50A 25
NiC r50/IrMn100
Ta50/IrMn100 20
Intensity, a. u.

GMR ratio, %
3500
15 Rs, Ohms/sq

2500 10
5
1500
0
25 27 29 31 33 35
500 NiFeCr thickness (A)

Fig. 4. Seed layer thickness effect on the NOL-GMR and sheet resistance.
-500
35 40 45 50
Diffraction angle, degrees
Fig. 3. XRD patterns for 1050 A NiCr+100 A IrMn, 50 A NiCr only, and 50 A
Ta+100 A IrMn lm stacks grown on thermal oxide Si wafers. (Cu Ka 45 kV, 40 mA.) 3.2.2. Critical thickness
The broad peak at 441 corresponds to NiCr, while the sharp peak at 41.61 is
Fig. 4 shows the sheet resistance (Rs) and GMR ratio
attributed to IrMn.
dependence for a single NOL-BSV as a function of NiFeCr seed
layer thickness. It is evident that critical NiFeCr thickness needs to
be 433 A (when used in combination with 10 A of CoFe here). This
for Co (0 0 0 2) for perpendicular recording medium, compared thickness is 450 A when single NiCr or NiFeCr seed layer is used,
with Ru seed layer only. This is primarily due to the much easy which is in agreement with what Lee et al. [10,11,14] reported
nucleation of Ru on amorphous Ta seed layer in comparison with previously.
on thermal oxide silicon. The missing of underneath Ta layer is The large sheet resistance and low GMR at o31 A of NiFeCr
thought to be the main cause in Aley et al.s [13] results that the seed-layer thickness is thought to be caused by the discontinuous
IrMn (111) texture is not favored by Ru and Cu seed layers, even and defective stack layer formation when the seed layer is not
though their lattices match fairly well with those of IrMn, since thick enough, such defective layers may scatter more electrons
the Ru and Cu seed layers in Aleys report are directly deposited and causes a higher Rs [10,11]. The Rs dropping with further
onto thermal oxide SiO2. In contrast, NiCr seed layer without Ta increase in NiFeCr thickness due to shunting being not signicant
underlayer does favor the strong IrMn (111) texture formation here, though, probably due to the relatively high resistivity of
once it reaches a critical thickness. This thickness is about 4050 A NiFeCr.
in our study from XRD pattern in Fig. 3. NiCr diffraction peak was
evident only after the NiCr thickness reached 50 A. Accordingly,
3.2.3. Resistivity consideration
the IrMn (111) developed well on 50 A NiCr, but not on 10 and
For CIP-GMR device, higher resistivity seed layer and thinner
30 A NiCr, which did not show XRD diffraction peak.
spacer layer are benecial to minimize current shunting [5]. For
Although we observed that amorphous Ta-only seed layer can
CPP-GMR device, on the other hand, materials with lower
also promote the IrMn (111) texture growth (see Fig. 3), this may
resistivity are required as seed layer and other non-active layers
not be a surprise. On amorphous substrate, it is believed that IrMn
(such as AFM, Cap, etc.) only add parasitic resistance to the device
(111) texture can be formed to minimize its surface energy (the
and do not contribute to the magneto-resistance change (MR
closely packed plane of (111) has to lie on the lm growing plane
ratio). Therefore, low-resitivitiy seed layers, such as Ta/Ru and
in the absence of a large strain or distortions from the underlying
Ta/Cu may be better.
layer [18]). However, the IrMn grain size may be relatively smaller
Here, we have experimentally measured both CIP resistance
compared with other crystal seed layers, and thus may affect the
and CPP resistivity for a variety of materials (including seed layer)
blocking temperature.
and shown in Fig. 5a and b. CPP RA for various materials was
As a result, other seed layer combinations, such as Ta/CoFe,
determined by patterning the devices into 0.35 mm circular shape
Ta/NiFe, NiFeCr/CoFe or NiCr/CoFe or NiFe, Ta/NiFeCr, Ta/NiCr,
and using Cu as both top and bottom electrodes to sandwich the
Ta/Ru, Ta/Cu, etc. have been widely studied for both CIP-GMR
materials to be studied (i.e. Cu/materials to be studied/Cu) rst,
[3,10,11], CPP-TMR [8] and perpendicular medium [19,20].
then isolating these devices with 40 nm Al2O3 and nally measure
Bi-layer seed layer with Ta is apparently advantageous to
the CPP resistance using a four-point probe. We did not correct the
remove the substrate surface impact, since amorphous Ta can act
Cu contact resistance here and assume that it is negligible. The
as a very good wetting layer for the second seed layer growing
data from Fig. 5 clearly indicated that NiFeCr and NiCr will not be
on top of it. For instance, Ru(0 0 0 1) and NiFeCr(111) texture is
a good seed layer for CPP-GMR due to the large CPP RA. For the
much improved with 420 A of Ta under layer deposited onto
same reason, CPP-GMR AFM (IrMn and PtMn) thickness needs to
thermal oxide SiO2 substrates. As we will discuss more later that
be reduced as well to minimize the parasitic resistance.
Ta surface is very suspective to impurity attack and careful
process sequence design is recommended to avoid the Ta surface
contamination. For instance, Mao et al. [3] found that IrMn (111) 3.2.4. Material selection and background impurity impact on trilayer
diffraction peak became broader and weak once the Ta is oxidized. CIP-GMR performance
It is therefore recommended to deposit the Ta underlayer and its Due to the fact that stack layers are in the range of a few to a
overlying seed in the same vacuum chamber if possible to few tens of angstroms, the impact of seed layer material selection
minimize the background impurity impact on the interface during and impurity impingement to the CIP-GMR performance can be
wafer transferring. signicant. To simplify the question on hand, we selected here a
ARTICLE IN PRESS
2906 X. Peng et al. / Journal of Magnetism and Magnetic Materials 321 (2009) 29022910

Normalized CIP resistivity, mohms-


Normalized CIP resistivity for various materials
1.2
1.0 Normalized CIP
resistivity
0.8
cm 0.6
0.4
0.2
0.0 IrMn

Ta

PtMn

Co32Fe48B20

NiFe12Cr40

NiCr40

Ru

NiFe20

CoFe10

Fe80Co20

Cu
Material ID

Normalized RA for various materials


1.2
Normalized CPP RA

1.0
Normalized CPP RA
0.8

0.6

0.4

0.2

0.0
PtMn

NiFe12Cr40

FeCoB6

NiFe20

FePt

Fe80Co20

CoFe10

CoPt

Ru x 10
NiCr40

IrMn

Ta

Material ID

Fig. 5. (a) CIP resistivity for various selected materials commonly used for both CPP- and CIP-GMR stacks and (b) CPP RA.

Table 2
Seed layer effect on the trilayer NOL-GMR ratio.

Sample ID Stack details Rs (O/sq) GMR (%)

X1 no seed/CoFe12.530/Cu12.5/CoFe12.530/Cu4/NOL/TaN50 12.3 1.1


X2 IrMn40/CoFe12.530/Cu12.5/CoFe12.5 30/Cu4/NOL/TaN50 11.9 0.1
X3 Ta30/Cu10/CoFe12.530/Cu12.5/CoFe12.5 30/Cu4/NOL/TaN50 8.9 0.1
X4 Ta30/Ru10/CoFe12.530/Cu12.5/CoFe12.5 30/Cu4/NOL/TaN50 10.25 0.1
X5 NiCr40/CoFe12.530/Cu12.5/CoFe12.5 30/Cu4/NOL/TaN50 6.8 19.4
X6 NiCr40/Ru10/CoFe12.530/Cu12.5/CF12.5 30/Cu4/NOL/TaN50 6.6 18.2
X7 NiCr40/Cu10/CoFe12.530/Cu12.5/CoFe12.5 30/Cu4/NOL/TaN50 6.1 19.2
X8 NiCr40/Al10/CoFe12.530/Cu12.5/CF12.5 30/Cu4/NOL/TaN50 8.8 0.8
X9 NiCr40/Ti10/CoFe12.530/Cu12.5/CoFe12.5 30/Cu4/NOL/TaN50 12.6 0.05
X10 NiCr40/Ta10/CoFe12.530/Cu12.5/CoFe12.5 30/Cu4/NOL/TaN50 8.5 0.2
X11 NiCr40/Ru10/CF12.530/Cu12.5/CF12.5 30/Cu4/NOL/TaN50 6.6 18.2
X12 NiCr40/Ru10-O2/CoFe12.530/Cu12.5/CoFe12.5 30/Cu4/NOL/TaN50 7.5 11.4

trilayer NOL-GMR stack to avoid the AFM and SAF inuence and Cu and Ru is not thick enough to fully develop good texture, which
summarized the data in Table 2. Samples X1X4 from Table 4 in turn, affect the CoFe magnetic layer growth. To further prove
clearly show very low GMR when bottom FL is directly grown this suspetion, a 10 A of Cu, Ru, Al, Ti, and Ta was added on top of
on thermal oxide SiO2 (X1), IrMn (X2), Ta30/Cu10 (X3), and 40 A NiCr to form a bi-layer seed layersee to samples X6X10).
Ta30/Ru10 (X4), compared with X5 where 40 A of NiCr seed layer Indeed, NiCr/Cu and NiCr/Ru showed normal GMR and Rs
is used. It is suspected that on amorphous Ta underlayer, 10 A of (though samples X6 showed reduced Rs compared with X5 due to
ARTICLE IN PRESS
X. Peng et al. / Journal of Magnetism and Magnetic Materials 321 (2009) 29022910 2907

the shunting caused by low-resistance Ru layer), but not for the The impact of background impurity on the seed layer proper-
samples with NiCr/Al, Ti, and Ta seed layers. Due to the large ties was further studied using a specular stack (Fig. 6). Stack sheet
lattice mismatching between NiCr and Ti/Al/Ta, as well as resistance oscillation between odd and even wafers was obvious
between Ti/Al/Ta and CoFe (see Table 1), texture of Ti/Al/Ta when wafers were run in a cascading mode (multiple wafers in
grown on NiCr and the overlying CoFe texture grown onto Ti/Al/Ta process tool during the run and some wafers had to wait in the
is unlikely to develop well, this will ultimately affect the CoFe transfer mode before a process mode was available for that
magnetic layer properties and degrade the trilayer stack CIP-GMR particular wafer). Such Rs oscillation is due to the transferring
performance. Another possibility of the low GMR ratio and time difference of the odd and even wafers from one chamber (C)
large sheet resistance for samples X8X10 is that the interface to another (D) when we run in cascading mode (see Fig. 7a). Since
between Al/CoFe, Ti/CoFe, and Ta/CoFe may be less stable and Ta readily reacts with residual moisture and background oxygen,
could be intermixed, compared with Cu/CoFe and Ru/CoFe longer transferring time means more impurity impingement on
interfaces. the Ta surface and results in higher Rs. It is easy to understand that
Samples X1112 from Table 2 highlighted the impact of the O2 poor vacuum increases such resistance oscillation amplitude due
exposure on the GMR ratio. A dose of 60 L of O2 on the NiCr/Ru to higher impurity concentration. Once a serial process mode
seed layer surface reduced the GMR from 18.2% to 11.4%, while was employed (one wafer in the process tool only at all times),
increased the sheet resistance. such oscillation pattern was gone (see Fig. 7b).

3.3. AFM/pinned layer interface and impurity impact

Table 3 shows the impact of O2 exposure (60 L) impact on the


seed layer/AFM and AFM/pinned interfaces. Previous XRD has
conrmed that Ta/Ru served as a good seed layer for IrMn (111)
AFM texture development in view of the good lattice matching
between Ru (0 0 0 1) and IrMn (111). When the Ta/Ru interface is
contaminated; however, the pinning eld reduced from 960 to
800 Oe; the IrMn pinning eld only exhibited minor reduction
Fig. 6. The most critical interface in a specular stack, affected by wafer transferring
from 960 to 920 Oe when Ru/IrMn interface was subjected to O2
process.
exposure. In contrast, the pinning eld completely disappeared
when IrMn/CoFe interface is attacked by O2. This highlighted that
the sensitivities of the interfaces towards impurity is in this order:
AFM/pinned layer interface4Ta/Ru interface4Ru/IrMn interface.
Specular stack
100 As a result, it is strongly recommended that AFM/PL processed
within the same chamber without wafer transferring to enable
C>D transfer time, sec

80 more robust process control.

60 3.4. SAF (PL/Ru/RL) interface dependence on material selection and


impurity
40
Tables 4 and 5 summarized the resulted SAF coupling strength
20 after subject to O2 exposure (60 L). It is apparent from Table 4
that O2 exposure (60 L) on pinned layer had a large impact on Ru
0 growth and completely destroyed the SAF coupling. This is
0 1 2 3 4 5 6 7 presumable because the formation of CoFeOx after O2 exposure
Wafer number and Ru (9 A here) does not grow well onto CoFeOx. However,
once the Ru has been deposited onto the pinned layer CoFe, the O2
exposure on Ru surface has very little impact onto the SAF
Rs-serial -good vac
coupling, probably due to Rutheniums low oxygen afnity.
Specular stack Rs-cascading-good vac
86
On the other hand, the SAF coupling was also found to be very
Rs_cascading-poor vac
sensitive towards the magnetic layer microstructure. Tsunoda
85
et al. [21] reported that the pinning eld for CoFex using IrMn AFM
Rs (ohms/sq)

84
showed a peak at about 30% of Fe content. We know that there is a
83 fcc to bcc transition for CoFex when Fe content is above
82 2530 at% and the spin polarization of CoFe (for CPP-GMR in
81 particular) increases with the Fe content [8]. So, CoFe30 was also
80 examined in this study as both PL and RL.
79 From Table 5, it seems that amorphous FeCoB and bcc Fe-rich
78 FeCo PL failed to provide good SAF coupling, while CoFe30 was
0 1 2 3 4 5 6 7 among the magnetic layers that provided highest SAF coupling,
Wafer number though its high temperature thermal stability (not shown here) is
inferior to that of CoFe10. Additional high-resolution TEM images
Fig. 7. (a) Chamber C to chamber D transferring time of odd/even wafers during from Fig. 8 revealed that Ru showed much sharper interface when
the specular stack process when run in cascading mode (multiple wafers in the grown on crystal ferromagnetic layer CoFe30, compared with that
process tool at one time) and (b) the resulted sheet resistance variation due to the
grown on amorphous ferromagnetic layer CoFeB. For a better TEM
difference of such transferring time and the impact of background base vacuum.
Note: good vacuum 109 Torr (with 12 h chamber baking) transfer mode; poor study of the FM/Ru interface, relatively thick (30 A) was used here,
vacuum 108 Torr base pressure (without baking). i.e. Ta20 A/FM1 50 A/Ru 30 A/FM2 50 A/Cap strucutre was used for
ARTICLE IN PRESS
2908 X. Peng et al. / Journal of Magnetism and Magnetic Materials 321 (2009) 29022910

Table 3
O2 exposure effect on seed/IrMn and the IrMn/pinned layer interfaces.

Sample ID Stack Mokem results


Hc (Oe) Hex (Oe) Kerr intensity

SEED1 Ta30/Ru30/IrMn70/CoFe30 50/Ta50 440 960 25.3


SEED2 Ta30O2 exposure/Ru30/IrMn70/CoFe30 50/Ta50 600 800 24.8
SEED3 Ta30/Ru30O2 exposure/IrMn70/CoFe30 50/Ta50 440 920 25.2
SEED4 Ta30/Ru30/IrMn70O2 exposure/CoFe30 50/Ta50 260 10 27.4

Table 4
O2 exposure effect at SAF interfaces on the SAF coupling strength.

Sample Stack Normalized SAF


ID coupling

SAF1 Ta30/Ru30/CoFe30 30/Ru9/CoFe30 30/Ta50 1


SAF2 Ta30/Ru30/CoFe30 30O2 exposure/Ru9/CoFe30 0
30/Ta50
SAF3 Ta30/Ru30/CoFe30 30/Ru9O2 exposure/CoFe30 1
30/Ta50

Table 5
Material and its structure effect on the SAF coupling strength.

Sample ID SAF material combinations Normalized SAF coupling

SAF-M1 Fe64Co16B20/Ru/Fe64Co16B20 0.00


SAF-M2 Fe80Co20/Ru/Fe80Co20 0.39
SAF-M3 Fe64Co16B20CoFe10/Ru/Fe64Co16B20 0.61
SAF-M4 CoFe30/Ru/Fe64Co16B20 0.75
SAF-M5 CoFe10/Ru/Fe64Co16B20 0.83
SAF-M6 CoFe10/Ru/CoFe10 0.89
SAF-M7 CoFe30/Ru/CoFe30 1.00

Note: Magnetic ux was kept the same for all the SAF material combination by
adjusting their thickness.
Fig. 8. TEM images showing the ferromagnetic layer (FM)/Ru interfaces with
structure: Ta 20 A/FM1 50 A/Ru 30 A/FM2 50 A/Cap, for (a) FM1 crystal CoFe30,
FM2 amorphous Co60Fe20B20, and (b) FM1 FM2 amorphous Co60Fe20B20.
Note: right side of images are the enlarged ones from left side.

TEM imaging to better evaluating the FM/Ru interface. Actual


Ru thickness used in real SAF structure is much thinner (9 A)
in Table 4 and the Ru growth defects on amorphous FM1 can between chambers does not occur at such critical interfaces.
cause less sharp interface and effectively change the SAF coupling Detailed residual gas analyzer (RGA) data has shown (not
(see SAF-M1 and SAF-M2). Interestingly, inserting 10 A of CoFe10 presented here) that residual moisture from the etching chamber
between amorphous FeCoB and Ru, restored the SAF coupling up and load locks (release from the wafer surface) for extended run
to 60% of maximum level for SAF-M3 sample. The relatively poor gradually released into transfer mode when slit valves were open
SAF coupling for Fe-rich PL may be also related to the interface in between transfer mode and etching mode or load lock. This has
stability between Fe and Ru layer. been identied as the primary reason to cause the GMR ratio to
The impact of Ru/RL interface was found less signicant than gradual drop for single NOL-GMR in Fig. 9. For example, in the
PL/Ru interface, accounting for only 1030% of SAF coupling wafer movement conguration #1, NiFeCr+NiFe seed deposition
variation when amorphous RL and crystal RL were used for the requires wafer transferring from chamber F to chamber D and as
same crystal PL (compare samples SAF-M4/M7 and SAF-M5/M6). the moisture level in transfer mode got higher and higher, wafer-
Therefore, crystal and Co-rich PL is recommended for better SAF to-wafer (WTW) GMR ratio is gradually reduced from wafer #1 to
coupling. wafer #11 and the GMR ratio range over mean (R/M) reached
about 7%. While for conguration #2, such R/M was reduced to
o3%, by eliminating the impurity impact on two critical
3.5. Reference layer/Cu spacer interface dependence on impurity interfaces: NiFeCr/NiFe and NiFe/PtMn interfaces (i.e. seed/AFM/
SAF were processed with the same chamber F and were not
Reference layer/Cu spacer is another critical interface affecting exposed to the transfer mode). However, spacer Cu deposition
CIP-GMR performance. Due to the necessary wafer transferring still required the wafer transferring from F to D, which normally
between various process modules during lm stack deposition, it took a few tens of seconds. By eliminating the interface between
is important to arrange targets and optimize the process ow reference layer CoFe and Cu spacer (CoFe/Ru) in conguration #3,
sequence in such a way that the required wafer transferring the WTW GMR range over mean was successfully reduced
ARTICLE IN PRESS
X. Peng et al. / Journal of Magnetism and Magnetic Materials 321 (2009) 29022910 2909

to 0.5%, while the average GMR ratio did not show obvious process sequence ow are critical to preserve the key interfaces
changes. (not to transfer wafers at critical interfaces). This learning is
To reduce the interlayer exchange coupling between RL and applicable to CPP-GMR as well.
free layer via Cu spacer, a smoothening treatment in CIP-GMR was
usually employed. This smoothening operation, unfortunately,
required transferring wafers from deposition module D to etching
module E (see Fig. 10). Dual NOL
200

D to E transfer time, sec


When wafers were run in a cascading mode (for fast
throughput) and the background impurity level was high, different
150
impurity absorption level on CoFe RL for odd/even wafers was
resulted due to their transferring time variation (see Fig. 11a). This
100
impurity absorption at CoFe(RL)/Cu interface resulted in the GMR
ratio oscillation as shown in Fig. 11b. By removing the chamber E
smoothening treatment and keeping CoFe(RL)/Cu spacer deposited 50
within the same chamber D (at the expense of slightly higher
interlayer exchange coupling) or reducing the moisture level of the 0
0 1 2 3 4 5 6 7
chamber E, we have successfully removed the GMR oscillation
Wafer number
pattern (not shown here).
Dual NOL
20
4. Conclusions
18

GMR ratio, %
The seed layer, SAF material selection and the background 16
impurity effect on the CIP-GMR SV performance have been
studied. It was found that the Ta/NiCr, Ta/NiFeCr, and Ta/Ru are 14
among the best seed layers to promote AFM (111) texture
12
development. For CPP-GMR, Ta/Ru seed is better due to its lower
parasitic resistance than Ta/NiCr counter part. AFM/pinned layer, 10
pinned/Ru and reference layer/Cu spacer interfaces are three of 0 1 2 3 4 5 6 7
the most critical locations to control the CIP-GMR AFM pinning, Wafer number
SAF coupling and the GMR ratio, respectively. Background oxygen
Fig. 11. (a) Chamber D to chamber E transferring time of odd/even wafers during
and water moisture can signicantly alter the above critical
the DNOL-BSV stack process at the interface highlighted in Fig. 10, and (b), the
interfaces. Employing of UHV system (with better than 109 Torr resulted sheet GMR ratio variation due to the difference of such transferring time
base pressure) and optimized target arrangement and wafer and the impact of background base vacuum.

Fig. 9. Congurations for simple bottom CIP-GMR wafer process and the resulted WTW GMR range over mean.

Fig. 10. Most critical interface in a dual NOL-GMR affected by transferring time.
ARTICLE IN PRESS
2910 X. Peng et al. / Journal of Magnetism and Magnetic Materials 321 (2009) 29022910

Acknowledgements [9] M. Zheng, G. Choe, K.E. Johnson, L. Gao, S.H. Liou, IEEE Trans. Magn. 38 (2002)
1979.
[10] C.L. Lee, A. Devasahayam, M. Mao, J. Kools, P. Cox, K. Masaryk, D. Mahenthiran,
The authors are grateful to Dr. Christoph Mathiu for assistance J. Munson, J. Appl. Phys. 93 (2003) 8406.
in MOKE measurement, to Ky Tran for testing support. Managerial [11] C.L. Lee, A.J. Devasahayam, C.C. Hu, Y. Zhang, M. Mao, J.C.S. Kools, K. Rook, IEEE
support from Dr. Cal Hardie is also acknowledged. Trans. Magn. 40 (2004) 2209.
[12] M. Fecioru-Morariu, G. Guntherodt, M. Ruhrig, A. Lamperti, B. Tanner, J. Appl.
Phys. 102 (2007) 053911.
References [13] N.P. Aley, G. Vallejo-Fernandez, R. Kroeger, B. Lafferty, J. Agnew, Y. Lu,
K. OGrady, IEEE Trans. Magn. 44 (2008) 2820.
[14] W.Y. Lee, M.F. Toney, D. Mauri, IEEE Trans. Magn. 36 (2000) 381.
[1] B. Diney, V.S. Speriosu, S. Metin, S.S.P. Parkin, B.A. Gurney, P. Baumgart, [15] H.N. Fuke, K. Saito, M. Yoshikawa, H. Iwasaki, M. Sahashi, Appl. Phys. Lett. 75
D.R. Wilhoit, J. Appl. Phys. 69 (1991) 4774. (1999) 3680.
[2] NIKKEI Electronics, 704 (1998) 2324. [16] H. Fuke, H. Fukuzawa, H. Yuasa, S. Hashimoto, H. Iwasaki, USP7289305 (Oct.
[3] M. Mao, C. Cerjan, J. Kools, J. Appl. Phys. 91 (2002) 8560. 30, 2007).
[4] A.A. Jibouri, M. Hoban, Z. Lu, G. Pan, J. Appl. Phys. 91 (2002) 7098. [17] M.J. Mehl, D.A. Papaconstantopoulos, Phys. Rev. B 54 (1996) 4519.
[5] M.A. Seigler, IEEE Trans. Magn. 43 (2007) 651. [18] C.Y. You, H.S. Goripati, T. Furubayashi, Y.K. Takahashi, K. Hono, Appl. Phys. Lett.
[6] S.S.P. Parkin, C. Kaiser, A. Panchula, P.M. Rice, B. Hughes, M. Samant, 93 (2008) 012501.
S.-H. Yang, Nat. Mater. 3 (2004) 862867. [19] A. Hashimoto, S. Saito, N. Itagaki, M. Takahashi, Appl. Phys. Lett. 89 (2006)
[7] S. Yuasa, T. Nagahama, A. Fukshima, Y. Suzuki, K. Ando, Nat. Mater. 3 (2004) 262508.
868871. [20] M. Shibamoto, K. Yamanaka, D.D. Djayaprawira, N. Watanabe, Intermag. Asia
[8] X. Peng, P. Kolbo, K. Nikolaev, S. Chen, Z. Wang, T. Boonstra, P. Anderson, 2005 Digest (2005) 583 April 48.
S. Kalderon, P. Czoschke, A. Morrone, D. Dimtrov, S. Xue, Y. Chen, J. Magn. [21] M. Tsunoda, K. Nishikawa, T. Damm, T. Hashimoto, M. Takahashi, J. Magn.
Magn. Mater. 321 (2009) 1889. Magn. Mater. 239 (2002) 182.

Das könnte Ihnen auch gefallen