Experimental separability of channeling giant magnetoresistance in Co/ Cu/ Co W. E. Bailey* Materials Science Program, Department of Applied Physics, 500 West 120th Street, Columbia University, New York, New York 10027, USA S. E. Russek National Institute of Standards and Technology, Boulder, Colorado 80303, USA X.-G. Zhang and W. H. Butler † Metals and Ceramics Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831-6114, USA Received 7 January 2005; revised manuscript received 24 March 2005; published 13 July 2005 The magnitude of the electronic channeling contribution is a significant open issue in the understanding of giant magnetoresistance GMR. We show that for the technologically important system Co/ Cu/ Co, channel- ing GMR can be isolated and quantified experimentally through measurement in the limit of rapid surface diffuse scattering. First-principles based Boltzmann transport calculations are compared with experimental in situ magnetoconductance data, which support the possibility of a significant contribution from channeling. Cyclic control of atomic-scale surface roughness, applied during in situ measurement, will enable a quantita- tive estimate of the channeling contribution. DOI: 10.1103/PhysRevB.72.012409 PACS numbers: 75.47.De, 72.10.-d, 73.50.-h, 75.47.-m I. INTRODUCTION Spin-valve giant magnetoresistance, particularly in the technologically important Co/ Cu/ Co trilayer system, has eluded a complete theoretical description. The relative weights of two separate proposed sources for the current-in- plane giant magnetoresistance CIP-GMR, spin-dependent diffuse scattering and electronic channeling, could not yet be estimated. In this article, we describe how channeling and spin-dependent scattering contributions to GMR may be separated experimentally in a single Co/ Cu/ Co “spin-valve” trilayer, allowing quantification as a function of interface mi- cro structures, to which channeling is thought to be espe- cially sensitive. The most conventional explanation for GMR is given in terms of spin-dependent scattering SDS. GMR from bulk and interfacial SDS sum, in the context of the Valet-Fert theory, 1 to yield a total R = R AP - R P . This additive property, combined with the ability to manipulate bulk SDS alone through selective doping, has enabled experimental separa- tion of bulk and interfacial SDS contributions to CPP-GMR in the inversion experiments of Hsu et al. 2 and Vouille et al. 3 Satisfactory agreement between experiments and first- principles-based models of CPP-GMR has been found subsequently. 4,5 An additional source of GMR arises in the CIP-geometry due to intrinsic electronic structure and grazing incidence of currents at interfaces. This alternate source, the electronic channeling or “waveguide” effect, has been identified in ab initio calculations of conductivity in the Co/Cu/Co111 system. 6 The channeling contribution to GMR is special since it involves electrons which cannot, at constant k  , travel from one Co layer to the other. In the parallel magne- tization state, spin-up majority electrons with large k  are confined to the Cu layer since they lack a k  -conserving state in majority Co. This reduces the total resistance of the par- allel state and is a source of GMR in the absence of spin- dependent diffuse scattering. The magnitude of the channeling contribution is a major open question in the understanding of GMR. Careful experi- ments which compare CPP and CIP values for similar struc- tures have simply indicated a “sizable” channeling contribu- tion: Vouelle et al. 2 attribute the absence of inversion of CIP- GMR to a major role of channeling in CIP. In this article, we show how the channeling contribution to GMR may be measured and isolated from SDS-related contributions. Using first-principles based Boltzmann trans- port calculations, we show that the channeling contribution G chan is independent of surface scattering, confined as it is to the interior Cu. G SDS , on the other hand, can be nearly suppressed for high rates of scattering at the exterior Co surface. The two contributions are additive in G, in analogy to the additive property of bulk and interfacial SDS contri- butions to R measured in CPP. We discuss how in situ magnetoconductance measurement, combined with control over diffuse scattering at the top surface, can be used to separate both contributions in a single sample. II. MODEL CALCULATIONS The purpose of the model is to compare any separate size effects of channeling and spin-dependent diffuse scattering GMR in Co/Cu/Co trilayers. The size effect is an important problem in spin valve GMR, as it reduces the GMR of trilay- ers by nearly an order of magnitude to 20% at room tem- parature compared with multilayer values to 110%. 7,8 Sur- face diffuse scattering, parameterized by a probability of specular reflection p, is thought to be the origin of this effect. Separate trends in the size effect will motivate an experimen- tal technique to isolate the channeling contribution. We focus therefore on the Co-layer size effect of GMR Gt Co  in trilayers and its dependence on the surface specularity pa- rameter p. A hybrid ab initio/Boltzmann semiclassical transport PHYSICAL REVIEW B 72, 012409 2005 1098-0121/2005/721/0124094/$23.00 ©2005 The American Physical Society 012409-1