VOLUME 74, NUMBER 17 P H Y S I C A L R E V I E W L E T T E R S 24 APRIL 1995 Observation of Pure Nuclear Diffraction from a Fe Cr Antiferromagnetic Multilayer T. S. Toellner, W. Sturhahn, R. Röhlsberger, E. E. Alp, C. H. Sowers, and E. E. Fullerton Argonne National Laboratory, 9700 South Cass Avenue, Argonne, Illinois 60439 (Received 28 October 1994) We report the observation of nuclear resonant diffraction of synchrotron radiation from a 57Fe Cr multilayer. The multilayer consists of 25 57Fe(17 Å) Cr(10 Å) bilayers with the Cr layer thickness chosen so as to exhibit antiferromagnetic (AF) coupling of the 57Fe layers. Because of the AF alignment of the layers, the magnetic periodicity of the multilayer is twice the electronic periodicity, resulting in a pure nuclear Bragg reflection that appears during nuclear resonant diffraction. The pure nuclear Bragg reflection presents a means of filtering synchrotron radiation to the level of 10­8 eV at the nuclear resonance energy of 14.4 keV. PACS numbers: 76.80.+y, 75.70.Cn, 78.70.Ck The discovery of oscillatory interlayer coupling in performing time-differential measurements in conjunction Fe Cr superlattices [1,2] has stimulated considerable with nuclear resonant diffraction, one has the potential to interest. Superlattices that possess antiferromagnetic (AF) gain both information on the local hyperfine interactions coupling of the ferromagnetic layers demonstrate unique as well as information on the long-range order of these magnetic phenomena such as giant magnetoresistance interactions [9]. Applying this technique in an analogous [3] and allow superlattices to be designed that mimic fashion to thin films containing 57Fe, or other Mössbauer the properties of two-sublattice antiferromagnets [4]. In isotopes, would allow one to investigate the magnetic or- this Letter, we present the results of nuclear resonant dering within the layered system by probing the alignment diffraction of 14.4 keV synchrotron radiation from an AF of the magnetic hyperfine field. coupled 57Fe Cr superlattice. As a result of the magnetic Recently, a pure nuclear Bragg reflection has been pro- ordering of the superlattice, we observe a pure nuclear duced in a synthesized multilayer with a 57Fe isotopic unit reflection, an effect originally observed by diffraction cell that is double the charge unit cell (57Fe Sc 56Fe Sc) from a two-sublattice antiferromagnetic crystal containing [10]. In the present Letter, we report the observation of 57Fe [5­8]. In addition, we show that the resonantly a pure nuclear reflection due to magnetic ordering from scattered radiation from the multilayer is sensitive to the a synthetic multilayer consisting of 17 Å 57Fe layers sep- hyperfine field distribution and hence may be used to arated by 10 Å Cr layers. The Cr layer thickness was probe the magnetic structure of thin films. chosen to produce antiferromagnetic coupling of the 57Fe If an Fe Cr superlattice is produced using 57Fe, then layers. This interlayer coupling results in a periodicity of the sample may be probed using the 14.4 keV nuclear magnetic hyperfine interactions that is twice the charge resonance in 57Fe. Using synchrotron radiation (SR) to density periodicity. The resulting difference between the excite this resonance in a Bragg scattering geometry and magnetic unit cell and the charge unit cell gives rise to a subsequently measuring the time distribution of the res- pure nuclear Bragg reflection. onantly scattered radiation allows one to probe the hy- The substrate 57Fe(60 Å) [Cr(10 Å) 57Fe(17 Å)]25 perfine interactions within the superlattice. Such nuclear Cr(20 Å) superlattice was grown by dc magnetron resonant diffraction has been performed on single crys- sputtering. A smaller sapphire substrate was mounted tals containing 57Fe in order to probe the magnetic struc- adjacent to the main substrate for subsequent magnetic ture. In the case of nuclear resonant Bragg diffraction measurements. The 57Fe was sputtered from a 3.8 cm from single crystals, the long-range order of hyperfine in- diam natural Fe target on which a 95% enriched 57Fe foil teractions may give rise to pure nuclear Bragg reflections. was spot welded to cover the region where the majority Pure means that the structure factor for charge scattering of the sputtering occurred. The Cr was sputtered from a vanishes, while that for resonant scattering does not. In 5 cm sputtering source with a 99.9% pure Cr target. The general, pure nuclear Bragg reflections occur whenever base pressure of the sputtering system was 1 3 1027 Torr the crystallographic symmetry group is not a subgroup and the Ar sputtering pressure was 3.5 mTorr. The of the symmetry group for the hyperfine interactions (on substrate-to-target distance was 8 cm, and deposition resonant nuclei). An antiferromagnetic alignment of mag- rates were kept constant at approximately 2 Å s. It has netic sublattices will produce pure nuclear reflections for been previously found that sputtering under these growth this reason. Pure nuclear reflections due to antiferro- conditions produces smooth layers with limited cumula- magnetic ordering of magnetic sublattices have been ob- tive roughness for many materials [11]. The substrate served with single crystals of 57FeBO3 and 57Fe2O3 using was mounted on a computer-controlled sample holder, SR [5,6] and radioactive sources [7,8]. In addition, by which passed the substrate sequentially over the Fe and 0031-9007 95 74(17) 3475(4)$06.00 © 1995 The American Physical Society 3475 VOLUME 74, NUMBER 17 P H Y S I C A L R E V I E W L E T T E R S 24 APRIL 1995 Cr sources. Because the sputtering sources are relatively To perform nuclear resonant diffraction from the mul- small compared to the size of the substrate, the motion of tilayer, we used synchrotron radiation produced by a 24- the substrate was chosen to ensure thickness uniformity pole wiggler on the F-2 beamline at Cornell's High En- over the length of the substrate. Under the present growth ergy Synchrotron Source (CHESS). X rays produced by conditions, it is expected that the thickness of the 57Fe the wiggler were directed through an aperture and de- layers will be constant to better than 5% over the width livered to a water-cooled, double-crystal monochromator (2.5 cm) of the substrate. This should result in negligible composed of Si(111) crystals to decrease the energy band- deviations of the 57Fe layer thickness over the width of width to approximately 4 eV (FWHM) at the 57Fe nuclear the incident synchrotron beam (1.0 cm). resonance energy of 14.413 keV. Because of the narrow Characterization of the hyperfine fields within the 57Fe resonance bandwidth (G 1028 eV), the ratio of resonant layers was performed using conversion electron Mössbauer flux to nonresonant flux is 10­9. To reduce the nonreso- spectroscopy (CEMS). A CEMS spectrum taken using a nant flux to a level that the detector can handle, we fur- 57Co(Rh) radioactive source is shown in the inset of Fig. 1. ther monochromatized the beam using a high-resolution A distribution of magnitudes of magnetic hyperfine fields crystal monochromator of a nested design [13] to an en- (shown in Fig. 1) was needed to fit the spectrum. From ergy bandwidth of 6 meV (FWHM) and with a vertical the weighted set of Gaussian distributions used to fit the beam divergence of 20 mrad. From here, the 0.3 mm 3 data, 42(8)% of the 57Fe in the layers possess the magnetic 10 mm horizontally polarized beam was vertically scat- hyperfine field of bulk a-Fe (33.3 T). The remainder of tered from the 57Fe Cr multilayer. An external magnetic the 57Fe possesses varying magnitudes of hyperfine fields. field of approximately 0.1 T was applied perpendicular to The percentages of five other hyperfine field strengths are the scattering plane to eliminate magnetic domain forma- 28% of 57Fe at 30.7 T, 11% at 26 T, 5% at 22 T, 6% at tion in the film. After scattering from the sample, the ra- 20 T, and 10% at 10 T. This result is in agreement with diation was collected by an avalanche photodiode (APD) the CEMS studies on Fe Cr films of Landes et al. [12]. detector that has a time resolution of #1 ns and an effi- The CEMS result also indicates minimal diffusion of 57Fe ciency of 12% at 14.413 keV. Because of the presence into the Cr layers, with approximately 0.1% of 57Fe hav- of 57Fe the scattered radiation from the multilayer con- ing a magnetic hyperfine field less than 1.0 T. Magnetic tains both a prompt nonresonant component and a delayed measurements using a DC SQUID magnetometer on the (by a time on the order of t ¯h G) resonant component. superlattice grown on the sapphire substrate demonstrate As a result of the time delay, the resonant radiation can that the 57Fe layers are antiferromagnetically coupled with a saturation field greater than 1 T, as expected for a 10 Å Cr layer [1­4]. FIG. 2. The electronic reflectivity of the 57Fe Cr multilayer (filled circles) with a fit (solid line). The nuclear reflectivity (circles) of the multilayer taken by time integrating the resonant flux arriving in a time window of 23­103 ns after FIG. 1. The distribution of magnetic hyperfine fields within the nuclear excitation. The three vertical lines correspond the 57Fe Cr multilayer used to produce the fit to the CEMS data to the angular positions of (a) the critical angle for total (inset). Note that the CEMS data depend on the distribution of external reflection, (b) the pure nuclear Bragg reflection, and magnitudes and orientation of hyperfine fields, while nuclear (c) the Bragg reflection resulting from the electronic charge resonant diffraction depends, in addition, on the location of periodicity. Nuclear resonant time spectra were taken at each those hyperfine fields. of these angles and are shown in Fig. 3. 3476 VOLUME 74, NUMBER 17 P H Y S I C A L R E V I E W L E T T E R S 24 APRIL 1995 be time filtered electronically. An rf signal from the syn- chrotron storage ring provides the timing reference. The 0.3 mm 3 10 mm beam impinging on the multi- layer at low incidence angles (,17 mrad) results in scat- tering from a significant portion of the surface. The mul- tilayer accepts the full cross section of the beam at in- cidence angles .4.2 mrad. The electronic, i.e., nonres- onant, reflectivity of the multilayer is shown in Fig. 2. The solid line is a fit to the data using a standard optical formalism. In addition, the interface roughness was as- sumed to be Gaussian of width s and was treated using the Nevót-Croce formalism [14]. The best fit required an interface roughness of s 4 Å and a surface roughness of s 6 Å. The time-integrated resonant signal resulted from ac- cumulating signals arriving in a time window from 23­ 103 ns after the prompt radiation. The nuclear reflectivity of the multilayer, taken by measuring the time-integrated resonant signal as a function of the incident angle, is shown in Fig. 2. The peak at 3.8 mrad corresponds to coherent resonant scattering near the critical angle for to- tal external reflection. The nuclear reflectivity grows from zero at low angles due to an increasing beam acceptance and growing penetration depth that effects an increased il- lumination of the 57Fe nuclei. The peak has been related to the variation in the electronic reflection amplitude, be- ing maximum where the electronic amplitude varies the strongest [15]. The peak at 16.4 mrad results from the pe- riodicity of 57Fe in the multilayer. The peak at 8.9 mrad is the pure nuclear Bragg reflection that corresponds to the periodicity enforced by the hyperfine interactions on the 57Fe nuclei. The angular width of the pure nuclear Bragg peak is 1.2 mrad (FWHM). This is essentially the intrinsic angular width of the reflection due to the finite number of layers in the film because the incident radiation has an angular divergence 2 orders of magnitude less than this value. The electronic reflectivity at the angular po- sition of the pure nuclear reflection is approximately 3 3 10­4. As a result, the multilayer reduces the flux outside the resonant bandwidth by a factor of 3 3 10­4, while pro- ducing a delayed, very narrow energy (5 neV) bandwidth FIG. 3. Nuclear resonant time spectra taken at three different beam. Thus, the reduced electronic reflectivity makes the angular positions of Fig. 2. The oscillations in the time spectra result from the beating of the different frequencies associated multilayer a candidate for nano-eV monochromatization with the different hyperfine transitions. The same model for the of synchrotron radiation. distribution of magnetic hyperfine fields was used to fit (solid While diffracting from the pure nuclear reflection, line) all three spectra. the time-integrated resonant counting rate was 2.5 cps. Comparison of this counting rate to that obtained with a 10 mm thick a-57Fe foil ( 85 cps) demonstrated an energy-integrated nuclear reflectivity of approximately fields in combination with the linear polarization of 1Go ( 5 neV). With this counting rate, the time-resolved the incident radiation means the resonant scattering is detection of the resonantly scattered radiation produced dominated by those transitions with DJz 61. These the time spectrum shown in Fig. 3. Because of the transitions correspond to lines 1, 3, 4, and 6 in the application of the magnetic field of 0.1 T perpendicular CEMS spectrum. The relationship between these resonant to the scattering plane, the magnetic moments of the 57Fe transitions and the beating in the time spectrum is quite layers align alternately parallel and anitparallel to the similar to the case of the (111) pure nuclear reflection scattering plane. This orientation of magnetic hyperfine in 57FeBO3 [5] without the additional component of an 3477 VOLUME 74, NUMBER 17 P H Y S I C A L R E V I E W L E T T E R S 24 APRIL 1995 electric quadrupole interaction. Time spectra taken while may be used to study thin film magnetic structures. In diffracting from both the peak near the critical angle for the future, the use of undulators will allow polarization- total external reflection and the peak corresponding to the sensitive studies in concert with nuclear resonant electronically allowed Bragg reflection are also shown in diffraction to increase the ability to probe crystalline and Fig. 3 for comparison. thin film structures containing a Mössbauer isotope. The fits are calculated based upon the theory of nuclear We would like to thank C. Bresloff for her assistance resonant dynamical diffraction [16] and a model for the in the characterization of the 57Fe Cr film. This work distribution of magnetic hyperfine fields within each Fe is supported by US-DOE, BES Materials Science, under layer. A 17 Å layer is divided into five sublayers with Contract No. W-31-109-ENG-38. The measurements at their thicknesses and distribution of magnetic hyperfine CHESS are supported by the NSF under Grant No. 90- fields chosen to be consistent with the CEMS result. 21700. Most of the variation in the hyperfine field is assumed to be concentrated near the Fe Cr interface. The model also assumed that the reduction of the magnetic hyperfine field at an Fe site within a layer was correlated with its proximity to the Cr layer. As a result, the sublayers [1] P. Grünberg, R. Schreiber, Y. Pang, M. B. Brodsky, and possess fields that decrease in value as one approaches C. H. Showers, Phys. Rev. Lett. 57, 2442 (1986). the Fe Cr interface. The central region of the Fe layer [2] S. S. P. Parkin, N. More, and K. P. Roche, Phys. Rev. Lett. (11 Å) was given the field distribution at 33.3 T. The 64, 2304 (1990). two neighboring sublayers (2 Å) had the field distribution [3] M. N. Baibich, J. M. Broto, A. Fert, F. Nguyen Van Dau, at 30.7 T. The next neighboring sublayers (1 Å) had F. Petroff, P. Eitenne, G. Creuzet, A. Friederich, and the remaining field distributions as determined from J. Chazelas, Phys. Rev. Lett. 61, 2472 (1988). the CEMS result. The resulting Fe layer possessed [4] R. W. Wang, D. L. Mills, E. E. Fullerton, J. E. Mattson, six different hyperfine field distributions segregated into and S. D. Bader, Phys. Rev. Lett. 72, 920 (1994). a 1 Å­2 Å­11 Å­2 Å­1 Å structure. The fitting was [5] U. van Bürck, R. L. Mössbauer, E. Gerdau, R. Rüffer, complicated by the fact that the nonuniform isotopic R. Hollatz, G. V. Smirnov, and J. P. Hannon, Phys. Rev. Lett. 59, 355 (1987). density developed during the fabrication process. The fits [6] G. Faigel, D. P. Siddons, J. B. Hastings, P. E. Haustein, are clearly not optimal but represent the sensitivity of the J. R. Grover, and L. E. Berman, Phys. Rev. Lett. 61, 2794 time spectra to the field distributions with the superlattice. (1988). Fitting all time spectra taken at different scattering angles [7] G. V. Smirnov, V. V. Mostovoi, Yu. V. Shvyd'ko, V. N. with the same parameters to describe the multilayer Seleznev, and V. V. Rudenko, Sov. Phys. JETP 51, 603 exposes details of the magnetic field distributions that are (1980). difficult to see in any other way. In addition, this places [8] G. V. Smirnov, V. V. Sklyarevskii, R. A. Voskanyan, and an enormous constraint on any model for the magnetic A. N. Artem'ev, JETP Lett. 9, 70 (1969). field distributions that might be assumed within the Fe [9] W. Sturhahn and E. Gerdau, Phys. Rev. B 49, 9285 (1994). layers. As a result, this technique has immense potential [10] A. I. Chumakov, G. V. Smirnov, A. Q. R. Baron, for studying magnetic thin films containing a Mössbauer J. Arthur, D. E. Brown, S. L. Ruby, G. S. Brown, and N. N. Salashchenko, Phys. Rev. Lett. 71, 2489 (1993). isotope. [11] E. E. Fullerton, J. Pearson, C. H. Sowers, S. D. Bader, Magnetic ordering in a synthetic multilayer has been X. Z. Wu, and S. K. Sinha, Phys. Rev. B 48, 17 432 (1993). observed with the use of nuclear resonant diffraction. [12] J. Landes, Ch. Sauer, R. A. Brand, W. Zinn, and Zs. The synthetic multilayer consisted of 17 Å 57Fe layers Kajcsos, Hyperfine Interact. 57, 1941 (1990). separated by 10 Å Cr layers. Because of antiferromag- [13] T. S. Toellner, T. Mooney, S. Shastri, and E. Alp, in netic coupling of the 57Fe layers that results from this Cr Optics for High-Brightness Synchrotron Radiation Beam- interlayer, the magnetic unit cell is twice the charge unit lines, edited by J. Arthur, SPIE Proceedings Vol. 1740 cell. These different unit cells give rise to a pure nuclear (SPIE­International Society for Optical Engineering, reflection. Because of the strong suppression of the elec- Bellingham, WA, 1993), p. 218. tronic scattering, the pure nuclear reflection may be used [14] L. Nevót and C. Croce, Rev. Phys. Appl. 15, 761 (1980). for nano-eV monochromatization of synchrotron radiation [15] A. Q. R. Baron, J. Arthur, S. L. Ruby, A. I. Chumakov, G. V. Smirnov, and G. S. Brown, Phys. Rev. B (to be in the hard x-ray regime. In addition, performing time- published). resolved nuclear resonant diffraction on the multilayer [16] J. P. Hannon, G. T. Trammel, N. V. Hung, M. Mueller, produces time spectra that are sensitive to the location of E. Gerdau, R. Rüffer, and H. Winkler, Phys. Rev. B 32, hyperfine fields in the layers. As a result, the technique 5068 (1985); 32, 6363 (1985). 3478