Vacuum 60 (2001) 401}405 E!ect of interface structure correlation on magnetoresistance of Fe/Cr multilayers Amitesh Paul*, Ajay Gupta, S.M. Chaudhari, D.M. Phase Inter-University Consortium for DAE Facilities, University Campus, Khandwa Road, Indore-452017, India Received 12 January 2000; received in revised form 29 March 2000 Abstract E!ect of interfacial roughness on giant magnetoresistance (GMR) in Fe/Cr multilayers has been studied. A set of samples is prepared by simultaneously depositing on a set of #oat-glass (FG) substrates with varying rms surface roughness. This causes the correlated part of the rms roughness to vary from sample to sample. Another set of specimen is irradiated with 200 MeV Ag ions in order to induce uncorrelated roughness at the interfaces. In both the cases morphological and other microstructural features of di!erent multilayers remained similar, thus allowing one to separate the e!ect of interface roughness from that of morphological changes. GMR measurements on these multilayers show that increasing interfacial roughness causes GMR to decrease nonlinearly. It is found that the e!ect of uncorrelated part of the roughness is much stronger than that of the correlated part. 2001 Elsevier Science Ltd. All rights reserved. Keywords: Magnetic "lms and multilayers; Giant magnetoresistance; Interface structure and roughness 1. Introduction pressure, sputtering power [1,3}6], substrate tem- perature [7] or by post-deposition treatments like Giant magnetoresistance (GMR) in metallic ion irradiation [8] and thermal annealing [9]. But multilayers [1] continues to be a topic of great the neglect of the e!ects of associated changes in the interest. A fairly good understanding of the basic morphological and other microstructural features phenomena, including the origin of the interlayer of the "lms on GMR are responsible for the contra- coupling and the spin-dependent electron scatter- dictory results in the above studies [1,3}12]. It may ing has been reached [2]. However, one aspect of be noted that variations in the deposition condi- the GMR phenomenon which is still not under- tions or the post-deposition treatments, besides af- stood properly is the role of the interface roughness fecting the interface quality, are also expected to in determining the GMR. Experimentally the e!ect a!ect other "lm properties like grain size and mor- of interface roughness on GMR has been studied phology, grain texture, internal stresses and defect by varying the deposition conditions like sputtering concentration in the bulk of the layers, etc. which in turn a!ect the GMR in the multilayers [10}16]. Therefore, in the present work we have tried to * Corresponding author. Tel.: #91-0731-463913; fax: #91- separate the e!ect of interface roughness on GMR 0731-462294. from that of morphological changes. The interface E-mail address: prasanna@iucindore.ernet.in (A. Paul). roughness can in general be written as the sum of 0042-207X/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 2 - 2 0 7 X ( 0 0 ) 0 0 2 2 4 - 4 402 A. Paul et al. / Vacuum 60 (2001) 401}405 two contributions " # , where is the irradiated with 200 MeV Ag ions up to #uences of correlated part and is the uncorrelated part of 2.0;10 , 5.0;10 , 1.0;10 , 5.0;10 , the interface roughness [13]. Therefore, two sets of 1.5;10 and 3.0;10 ions/cm using 15 UD multilayers are prepared in which either the corre- Pelletron at Nuclear Science Centre, New Delhi. lated part or the uncorrelated part of the interface XRR and AFM were used to characterize the roughness is varied in a controlled manner. Char- substrate as well as the multilayers. XRD was used acterization of the multilayers using X-ray re#ectiv- to determine the grain texture and size of the multi- ity (XRR), X-ray di!raction (XRD), conversion layers. CEMS was used to gather information electron Mossbauer spectroscopy (CEMS) and about the intermixing at the interface. For XRR Atomic Force Microscopy (AFM) shows that ex- a powder X-ray di!ractometer model D5000 of cept for the interface roughnesses, other micro- Siemens with CuK radiation was used. CEMS structural features of the multilayer (ML) like grain measurements were done using a gas #ow propor- size, coherence length, grain texture, intermixing at tional counter and a 50 mCi Co source in Rh the interface, internal stresses etc. remain un- matrix. Magnetoresistance at room temperatures changed, thus allowing one to selectively study the was measured using the standard four-probe tech- e!ect of interface structure on the GMR. nique with a constant current source and a nanovoltmeter in an external "eld upto 1 T. The "eld was applied parallel to the plane of the "lm 2. Experimental and perpendicular to the electrical current which was also in the same plane. Float glass was used as substrate for depositing multilayers. Substrates with varying surface rough- ness were prepared by etching the #oat-glass sub- 3. Results strates in dilute HF for varying periods of time. Six substrares with etching times of 0, 15, 60, 300, 600 Table 1 summarizes the results of various and 1200 s, designated as specimens 1}6, respec- measurements of the substrate as well as of the "lms tively, were taken (set I). XRR measurements in the case of set I. Column 3 of Table 1 gives the showed that the rms surface roughness varied non- rms roughness of the substrates obtained by com- monotonically with etching time. Multilayers were puter "tting of the X-ray re#ectivity data on sub- deposited on these substrates simultaneously in strates etched for di!erent periods of time. One may a UHV chamber using two e-beam guns (TELEM- note that the surface roughness exhibits a non- ARK Model No. 528) at a rate of 0.01 nm/s. The monotonic variation with etching time: after reach- base pressure in the chamber was 8.0;10\ mbar. ing a maximum value of 1.25 nm for etching time of The source to substrate distance was kept at 60 cm, 300 s, it again decreases with further etching. This in order to ensure uniformity of layer thickness re#ects some sort of layer by layer removal of the (within 0.5%) on di!erent substrates. Thicknesses surface during etching. of individual layers were controlled during depos- XRD measurements showed that the "lms have ition using a standard quartz-crystal oscillator. a texture along the (1 1 0) direction. However the Multilayers consisted of the following deposition texture does not vary from sample to sample. The sequence: substrate/Cr (10.0)/[Fe (3.0)/Cr width of the (1 1 0) re#ection was used to determine (1.2)];20/Fe (5.0), where the numbers in brackets the structural coherence length of grains along give the layer thickness in nm. Cr spacer layer the momentum transfer vector q using the Scherrer thickness of 1.2 nm corresponds to the "rst peak method, and is reported in Table 1. It may be noted in the antiferromagnetic coupling between Fe that has several times the thickness of individual layers [1]. layers indicating a high-degree of coherency be- Another set of multilayers (set II) of composition tween adjacent Fe and Cr layers. Further, one "nds [Fe (3.0 nm)/Cr (1.2 nm)];20/Fe (5.0 nm) samples that does not vary with substrate roughness. The deposited on FG and Si (1 1 0) substrates were d-spacing of (1 1 0) planes as calculated from the A. Paul et al. / Vacuum 60 (2001) 401}405 403 Table 1 Microstructural parameters of Fe/Cr multilayers on #oat glass (FG) substrates with di!erent etching times ¹. is the rms roughness of the glass substrates after etching for di!erent periods of time. The values of the lattice spacing d of (1 1 0) planes, structural coherence length , and average grain size in x}y plane t is also reported. The relative area A under the broad hyper"ne "eld component in the CEMS gives the fraction of total iron atoms located at the interfaces and is a measure of the thickness of the intermixed layer. The last column gives the saturation magnetoresistance of the multilayer No. ¹ (s) (nm) d (nm) (nm) t (nm) A (%) GMR (%) 1 0 0.67$0.05 0.2025$0.0005 16.3$1.0 248 26 3.52$0.01 2 15 0.77 0.2026 15.3 290 * 3.08 3 60 0.92 0.2025 15.6 295 * 2.94 4 300 1.25 0.2025 15.5 291 26 2.83 5 600 0.95 0.2028 15.9 266 25 3.19 6 1200 0.85 0.2029 16.1 261 * 3.22 di$cult to obtain a good theoretical "t to the experimental data. However, the following in- formation could be obtained: (i) from the position of the Bragg peak one can "nd that the bilayer periodicity is 4.4 nm instead of the designed value of 4.2 nm. This di!erence may be due to some error in the tooling factor of the thickness monitor, (ii) the height of the Bragg peak which is related to the average interface roughness, decreases with increasing substrate roughness, con"rming that the roughness of the substrate is at least partly transfer- red to the successive layers. The polycrystalline nature of the "lms is clearly visible from AFM pictures. For each sample 10 di!erent frames of 1 m;1 m were taken and the average grain size in the "lm plane was calculated. The results are reported in Table 1. CEMS Fig. 1. X-ray re#ectivity scans of Fe/Cr multilayers on #oat measurements were done in specimens 1,4 and 5. glass (FG) substrates with di!erent etching times and on micro- The spectra were "tted with two distributions of scopic glass slide (SG). For clarity, various curves are shifted hyper"ne magnetic "elds. The distribution in the relative to each other along the y-axis. range 28 T(B (36 T corresponds to the bulk of the iron layers while the broad distribution in the position of (1 1 0) re#ection is also reported in range 0 T(B Table 1. The d-value and hence the internal stresses (30 T corresponds to the iron atoms at the interfaces [17]. The fraction of total in the "lm also do not vary from sample to sample. iron atoms located at the interfaces, which is pro- Fig. 1 shows the re#ectivity pattern of the multi- portional to the relative area under the broad sex- layers deposited on di!erent substrates. The "rst tet, is a measure of the thickness of the intermixed Bragg peak due to multilayer periodicity is clearly layer at the interface and is reported in Table 1. The visible. However beyond the "rst Bragg peak the GMR de"ned as R re#ectivity pattern becomes obscure due to strong !R /R ;100 with R and R di!use scattering. The presence of the Cr seed layer, being the resistance values at zero and saturat- ing "elds, respectively, is also reported in Table 1. an electron density gradient in the substrate and Fig. 2 gives the variation of the GMR with the a possible oxidation of the Fe capping layer made it roughness . 404 A. Paul et al. / Vacuum 60 (2001) 401}405 Fig. 4 gives the variation of GMR as a function of irradiation #uence. 4. Discussions Table 1 shows that grain size, grain texture, structural coherence length , internal stresses and the thicknesses of the interface layers are similar for all the multilayers of set I grown on di!erent substrates. Furthermore, since all the "lms were deposited simultaneously, the deposition condi- tions like deposition rate and substrate temper- Fig. 2. Variation of percentage GMR with surface roughness of ature are identical for all the specimens. Therefore, the substrates. the individual layer thicknesses as well as the den- sity of defects in the bulk of the layers is expected to be similar. Thus, the only di!erence between vari- ous multilayers deposited on di!erent substrates is in their interface roughness, and the observed vari- ation in GMR can be attributed to variation in the interface roughness only. The di!erence in the interfacial roughness in dif- ferent multilayers is essentially due to the di!erence in the roughness of their substrate which is trans- mitted to the successive layers. Therefore, the di!er- ence among various multilayers is expected to be in their correlated part of the interfacial roughness. It is interesting to note that with increase in etching time, as the substrate roughness decreases for etch- ing time beyond 300 s, the GMR of the correspond- ing multilayers also shows an increase. Thus, the Fig. 3. XRR scans of [Fe(3.0 nm)/Cr(1.2 nm)];20 multilayers observed variation in GMR in this set is due to the showing irradiated spectra with di!erent irradiation #uences of correlated part of the interface roughness. (a) 2;10 (b) 5;10 (c) 1;10 (d) 5;10 (e) 1.5;10 and From Fig. 4 one "nds that the e!ect of 200 MeV (f) 3;10 ions/cm irradiated with 200 MeV Ag ion. Ag ion is again to cause a decrease in GMR. Since the modi"cations at various interfaces induced by irradiation are not expected to be correlated, the XRD measurements in the specimens of set II observed decrease in GMR is because of an in- show that irradiation does not a!ect the mor- creased uncorrelated part of the roughness. A com- phological parameters like grain size, texture etc. parison of Figs. 2 and 4 shows that while the e!ect Irradiation a!ects the XRR pattern indicating cha- of an increase in the correlated part of the rough- nges in the interface structure (Fig. 3). Because of ness by almost 100% is to cause a decrease in GMR the small contrast between Fe and Cr in their by only 20%, the e!ect of swift heavy ion irradia- refractive indices it is not possible to "t the re#ec- tion is to decrease GMR by more than 60%. Thus tivity data to get reliable information. However in the present study shows the e!ect of the uncor- some earlier studies it has been shown that in related part of the interface roughness is much Fe/Tb multilayers, the interface roughness varies stronger compared to the correlated part of the almost linearly with irradiation #uence [17]. roughness. A. Paul et al. / Vacuum 60 (2001) 401}405 405 References [1] Fullerton EE, Kelly DM, Guimpel J, Schuller IK, Bruyn- serade Y. Phys Rev Lett 1992;68:859. [2] Barnas J, Bruynserade Y. Phys Rev B 1996;53:5449. [3] Nakashini H, Okiji A, Kasai H. J Magn Magn Mater 1993;126:451. [4] Petro! F, Barthelemy A, Hamzic A, Fert A, Etienne P, Lequien S, Cruzet G. J Magn Magn Mater 1991;93:95. [5] Takanashi K, Obi Y, Mitani Y, Fujimore H. J Phys Soc Japan 1992;61:1169. [6] Rensing NM, Payne AP, Clemens BM. J Magn Magn Mater 1993;121:436. [7] BelieKn P, Schad R, Potter CD, Verbanck G, Moshchalkov VV, Bruynserade Y. Phys Rev B 1994;50:9957. [8] Kelly DM, Schuller I, Koreniviski V, Rao KV, Larsen KK, Fig. 4. GMR of the as-deposited and irradiated specimens of Bottiger J, Gyorgy EM, Van Dover RB. Phys Rev [Fe(3.0 nm)/Cr(1.2 nm)];20 multilayers deposited on #oat B 1994;50:3481. glass (- - - - - -) and Si substrate (0) as a function of #uence [9] Laider LH, Hickey BJ, Hickey TRA, Hase TRA, Tanner irradiated with 200 MeV Ag ions. BK, Schad R, Bruynserade Y. J Magn Magn Mater 1996;156:332. [10] Schad R, BelieKn P, Verbanck G, Potter CD, Fisher H, Lefebvre S, Bessiere M, Moshchalkov VV, Bruynseraede Y. Phys Rev B 1998;57:13692. 5. Conclusions [11] Velez M, Schuller I. J Magn Magn Mater 1998;184:275. [12] Slonczewiski JC. Phys Rev Lett 1991;67:3172. In conclusion, the e!ects of correlated as well [13] Schreyer A, Ankner JF, Zeidler Th, Zabel H, SchaKfer M, Wolf JA, GruKnberg P, Majkrzak CF. Phys Rev B 1995;52:16066. as uncorrelated interface roughness on GMR [14] Parkin SSP, York BR. Appl Phys Lett 1993;62:1842. in Fe/Cr multilayers have been studied in [15] Fullerton EE, Conover MJ, Mattson JE, Sowers CH, specimens with similar morphological structure. Bader SD. Phys Rev B 1993;48:15755. The increasing interface roughness causes GMR [16] Colino JM, Schuller IK, Schad R, Potter CD, BelieKn P, to decrease. The e!ect of uncorrelated roughness Verbanck G, Moshchalokov VV, Bruynserade Y. Phys Rev B 1996;53:766. is much stronger than the e!ect of correlated [17] Gupta A, Paul A, Gupta R, Avasthi DK, Principi G. roughness. J Phys Condens Mater 1998;10:9669.