PHYSICAL REVIEW B VOLUME 57, NUMBER 9 1 MARCH 1998-I Influence of magnetism on superconductivity in epitaxial Fe/Nb bilayer systems Th. Mu¨hge, K. Theis-Bro¨hl, K. Westerholt, and H. Zabel Institut fu¨r Experimentalphysik/Festko¨rperphysik, Ruhr-Universita¨t Bochum, D-44780 Bochum, Germany N. N. Garif'yanov, Yu. V. Goryunov, and I. A. Garifullin Kazan Physicotechnical Institute of Russian Academy of Sciences, 420029 Kazan, Russian Federation G. G. Khaliullin* Max-Planck Institut fu¨r Physik komplexer Systeme, D-01187 Dresden, Germany Received 27 May 1997; revised manuscript received 22 August 1997 We studied the superconductivity of molecular-beam-epitaxy-grown epitaxial Nb 110 thin films covered by Fe 110 monocrystalline layers with thicknesses dFe between 4 and 30 Å. We find a nonmonotonous depen- dence of the superconducting transition temperature Tc on dFe , however, without a maximum in Tc(dFe) as observed earlier in Nb/Fe layers prepared by sputtering techniques. We argue that the change in the Tc(dFe) curve is caused by a smaller thickness of the nonmagnetic Fe-rich interlayer at the Nb/Fe interface and can be explained within the same model developed previously for the explanation of the maximum in Tc(dFe) in sputtered Nb/Fe layers. S0163-1829 98 04209-X The study of superconductivity in multilayer systems highest quality interface with less alloying than in the case of combined of ferromagnetic FM and superconducting SC sputtered systems. This might lead to a change of the Tc(dFe) layers is a topic of great current interest see, e.g., Refs. 1­3 dependence. and references therein . The renewed interest in this classical The Nb/Fe bilayers were prepared by a conventional 3-in. proximity effect at a FM/SC interface4 partly originates from RIBER EVA 32 metal molecular beam epitaxy system for the observation of a maximum in the SC transition tempera- details about this equipment see, e.g., Ref. 8 . The base pres- ture T sure in the sample chamber is 5 10 9 Pa and the working c in several systems,1,2 when plotting Tc versus the thickness of the FM-layer d pressure 2 10 8 Pa. Nb was evaporated by electron beam F . This has given rise to the speculation that the FM/SC multilayer systems might repre- evaporation from a 14-cm3 crucible. An evaporation rate of sent the first experimental example of a -type Josephson 0.5 Å/s was found to be optimal for the growth of high coupling between successive SC layers, which has been pre- quality single crystal Nb 110 films with SC transition tem- dicted theoretically.5 perature Tc 8.5 K. Fe was deposited from an effusion cell In our recent study on Fe/Nb/Fe-trilayer systems,6,7 for providing a flux of high stability. A rate of 0.1 Å/s was used which a -type coupling can be excluded for geometrical and the final thickness was determined by the evaporation time. The placement of the effusion cells is inclined at a reasons, we found, however, a similar Tc(dFe) curve with a certain angle to the 3-in. sample holder. This gave us the clearly resolvable maximum at dFe 12 Å. We have devel- opportunity for the preparation of an Fe wedge on top of the oped a new model for the interpretation of the peculiar Nb layers. As substrate material a single crystal Al T 2O3 c(dFe) dependence in Fe/Nb, taking the presence of a non- (112¯0) was used, which was outgassed and finally annealed magnetic magnetically ``dead'' Fe-rich interlayer at the at 1000 °C for more than 1 h before starting the evaporation Nb/Fe interface into consideration. This interlayer acts in a process. The substrate temperature during the evaporation of twofold manner. On the one hand it reduces Tc due to the the first Nb layer was 900 °C. Subsequently the layer was introduction of a strongly repulsive interaction for the Coo- annealed at 950 °C for 15 min. The following Fe wedge type per pairs. On the other hand, it screens the SC layer from the layer was evaporated at 100 °C. Finally the Fe wedge was strongly pair-breaking exchange field of the FM layer. As covered by a thin Nb and Pd film of 30 Å thickness in order explained in detail in Ref. 7 with an appropriate thickness of to prevent Fe from oxidation. The quality of each layer was the nonmagnetic interlayer the onset of ferromagnetism in controlled by in situ reflection high-energy electron diffrac- the Fe layer can cause an enhancement of Tc first, instead of tion characterization. We have prepared three samples with a rapid suppression of Tc , which would be expected consid- wedge-shaped Fe layers: m855 with dNb 220 Å and with ering proximity effects between the FM layer and the SC dFe 10***26 Å; m856 with dNb 250 Å and with layer alone. dFe 7***23 Å; m879 with dNb 225 Å and with dFe 4***9 Since the interlayer at the Fe/Nb interface plays an essen- Å. The width of each sample was 50 mm. For the magneti- tial role in our model, the aim of the present paper is a zation, resistivity, and ac-susceptibility measurements the comparative study of Tc(dFe) for heterostructures prepared samples were cut into stripes with a width of 3 mm. in single crystalline form by the molecular beam epitaxy X-ray reflectivity measurements were performed ex situ at MBE technique. The optimized growth of Fe on Nb by three positions of each sample, using an additional vertical state of the art MBE techniques is known to provide the slit of 0.2 mm in order to exclude effects from the Fe gradi- 0163-1829/98/57 9 /5071 4 /$15.00 57 5071 © 1998 The American Physical Society 5072 BRIEF REPORTS 57 FIG. 1. X-ray reflectivity scans taken on three films cut from the FIG. 3. Transition curves as obtained by ac susceptibility for m855 sample. The solid line is a fit by the Parratt formalism Ref. five samples with different d 9 . The Fe thickness as determined from the fit was a 10 Å, b 15 Fe cut from the m856 sample. Å, and c 23 Å. The inset shows the 110 Bragg reflections with tion as a function of 1/dFe is presented for a series of films well resolved Laue oscillations. cut from m855 and m879 samples. The saturation magneti- ent on the interpretation of the spectra. In Fig. 1 three typical zation values are determined using the Fe thickness dFe as reflectivity measurements are presented for films cut from obtained by the x-ray reflectivity measurements. the sample m855 with d For the residual resistivity ratio RRR R Fe 10, 15, and 23 Å, respectively, 300 K /R10 K we showing well resolved thin film thickness oscillations Kies- obtained a factor of about 10­15 for all of our samples. sig fringes . The broadening of the total reflectivity edge is Taking the phonon resistivity of bulk Nb as ph 14.5 due to the top Pd layer with a rather high electron density cm, we estimate a residual resistivity of less than 1.6 and a correspondingly high critical angle. Fits with the Par- cm for our samples. Using the experimental values for ratt formalism9 shown as solid lines in Fig. 1 give interface the density of states at F and the Fermi velocity for Nb, we roughnesses of less than 5 Å, indicating the high structural obtain l 3.75 10 6 cm2,10 corresponding to a quality of our samples. The film thicknesses obtained from mean free path of 250 Å. According to Ref. 5 the value of these fits were used for the final calibration of d the SC coherence length can be estimated by Fe . The inset of Fig. 1 shows a typical Bragg scan of a sample with dNb 350 Å and dFe 100 Å. The presence of Laue oscillations from the Nb as well as the Fe layers is an indi- S BCSl 3.4 cation of the structural coherence of the films. All samples show 110 growth of Nb and Fe with a structural coherence where BCS is the BCS coherence length of pure Nb. We then length comprising the total film thickness. In-plane x-ray obtain S 180 Å for our thin films, which is three times Bragg reflection measurments by grazing incidence diffrac- larger than the S estimated for our sputtered films.7 tion show that Fe and Nb grow within a single domain with The SC transition temperature Tc was measured by ac the Fe 001 axis parallel to the Nb 001 axis. susceptibility. Typical transitions for one series of samples The magnetic state of the Fe layers was studied by a su- are presented in Fig. 3. Except for the sample with dFe 14 perconducting quantum interference device magnetometer at Å, all transitions are very sharp. The origin for the slightly a temperature of 10 K with the surface of the samples paral- broader transition of the sample with dFe 14 Å can be un- lel to the magnetic field. In Fig. 2 the saturation magnetiza- derstood by taking into account the shape of the Tc(dFe) curve see Fig. 4 , with the strong thickness dependence around dFe 14 Å. Hence a small gradient of dFe caused by the wedge shape of the Fe layer can lead to a definite broad- ening of the SC transition. The SC transition temperature Tc , defined as the mid- point of the ac-susceptibility transition, is plotted in Fig. 4 for three series of samples with variable dFe and fixed dNb . As mentioned above, Tc for a pure SC Nb film is about 8.5 K. Thus there is a strong initial Tc supression at lower dFe followed by a flat Tc(dFe) curve for 5 Å dFe 13 Å and final dropping to a constant lower Tc value between 13 Å dFe 16 Å. Thus in the present Fe/Nb bilayers Tc depends nonmo- notonously on dFe , similar to our previously studied Fe/ Nb/Fe trilayers,6,7 however, here without a maximum in FIG. 2. Saturation magnetization versus reciprocal thickness of Tc(dFe). It is neccessary to note that a strong resemblance the Fe-layer 1/d exists between our data and the data reported on the Nb/Gd Fe for a series of films cut from the m855 closed circles and m879 opened triangles samples. The solid line is a system,2,3 when comparing sputtered and MBE samples. In linear fit. both cases the sputtered samples show an increase in Tc 57 BRIEF REPORTS 5073 b The exchange field of the FM layer is strongly pair breaking and reduces Tc . If one assumes ideally transparent interfaces and neglects any influence of spin-orbit interac- tion, one can easily estimate the expected Tc suppression. The exchange splitting of the conduction band in Fe IFe is about 1 eV. In the Cooper limit the effective exchange field acting on the Cooper pairs in the Fe layer is given by Ieff IFe dFe /dNb . 1 Considering the Clogston limit, a field of H 2 /g B will completely quench superconductivity. With the gap param- eter for Nb and the exchange field H Ieff / B a complete quenching of superconductivity is expected for dFe 0.5 Å. This corresponds to less than one monolayer. In the sputtered Fe/Nb/Fe trilayers the FM layer is well FIG. 4. Superconducting transition temperature Tc versus dFe screened by the nonmagnetic interlayer and effect a domi- for three series of samples closed triangles: m856; closed circles: nates. This give rise to the Tc maximum just at the onset of m855; opened triangles: m879 . The broken lines are guides for the ferromagnetism. eye. In the MBE-grown samples, the nonmagnetic interlayer is thinner, thus the FM layer is less well screened and has a stronger direct negative influence on T when the magnetic layer becomes FM, while the MBE c . Both effects a and b just compensate and give rise to the first plateau in samples show a shoulderlike feature with a downward jump T around 13***20 Å. We believe that for both systems the c(dFe) in Fig. 4. At larger dFe the negative direct effect of the exchange field dominates in any case and a downward change of the Tc dependence on the ``magnetic'' layer thick- jump in T ness is due to a smaller value of the non FM interlayer at the c(dFe) with a large slope is expected. Tc ap- proaches a second plateau at an FM layer thickness interface in the MBE samples compared to the sputtered d d samples. Fe dNM comparable to the penetration depth of Cooper pairs into the Fe layer The linear dependence of M M 12 Å.13 It is necessary to note S on 1/dFe Fig. 2 is caused that the absence of any anomaly in the magnetization at the by the presence of a nonmagnetic Fe/Nb at the interface.6,7 A Fe thickness range between 5 and 25 Å Fig. 2 indicates that linear fit gives a thickness of the nonmagnetic layer of the jump in the T d c(dFe) curve at 13***16 Å is caused by the NM 5 Å. This is smaller than dNM for Fe/Nb/Fe trilayers exchange field only. prepared by rf sputtering techniques,7 where the magneti- In summary, we have studied the influence of magnetism cally ``dead'' layer thickness was larger than 7 Å. We have on the superconductivity in high quality epitaxial Fe/Nb bi- assumed7 that the magnetically ``dead'' layer arises due to an layers. We obtained a nonmonotonous T intermixing of Nb and Fe at the interface. Therefore the dif- c(dFe) curve, how- ever, without the maximum observed in our sputtered ference in the nonmagnetic layer thickness may be due to the samples. We explained this finding in the model developed smaller particle energies during thermal evaporation as com- previously for sputterd Fe/Nb/Fe trilayers.7 This model was pared to sputtering. It is well known that Fe atoms diluted in based on a combined action of the FM Fe layer and the Nb lose their magnetic moment. This has been experimen- nonmagnetic Fe/Nb interface on the superconductivity. We tally established11 and theoretically explained.12 concluded that the actual shape of the T Let us discuss the observed T c(dFe) curve in c(dFe) dependence. From Fe/Nb multilayers is mainly determined by the thickness of our magnetization data we concluded that for dFe dNM fer- the nonmagnetic interface. Therefore FM/SC proximity sys- romagnetism is absent. As explained in detail in Ref. 7, the tems offer an additional tool for the study of interface re- effective electron-electron interaction in this Fe-rich non- gions and its complex interplay with the physical properties magnetic interlayer is strongly repulsive and stems from the in multilayer system. coupling of the conduction electrons to the local paramagnon fluctuations. This repulsive interaction suppresses Tc by the We would like to thank J. Podschwadeck and C. Leschke proximity effect thus giving rise to the initial Tc suppression for technical support. This work was supported by the in Fig. 4. A special situation is given when a FM Fe layer Deutsche Forschungsgemeinschaft Grant No. DFG-ZA161/ occurs first for dFe dNM , since then there are two compet- 6-2 and by the Russian Fund for Fundamental Research ing effects: Project No. 96-02-16332a , which are gratefully a The Zeeman splitting of the virtual Fe-derived d state acknowledged. One of us I.A.G. acknowledges NATO in the nonmagnetic layer weakens the repulsive interaction support under the collaborative research Grant No. and this enhances Tc .7 HTECH .CRG961371. *Permanent address: Kazan Physicotechnical Institute of Russian 2 J. S. Jiang, D. Davidovic´, Daniel H. Reich, and C. L. Chien, Phys. Academy of Sciences, 420029 Kazan, Russian Federation. Rev. Lett. 74, 314 1995 . 1 H. K. Wong, B. Y. Jin, H. Q. Yang, J. B. Ketterson, and J. E. 3 C. Strunk, C. Su¨rgers, U. Paschen, and H. v. Lo¨hneysen, Phys. Hillard, J. Low Temp. Phys. 63, 307 1986 . Rev. B 49, 4053 1994 . 5074 BRIEF REPORTS 57 4 N. R. Werthamer, Phys. Rev. 132, 2440 1963 . Th. Mathieu, C. Mathieu, and B. Hillebrands, Phys. Rev. B 53, 5 Z. Radovic´, M. Ledvij, L. Dobrosavljevic´-Grujic´, A. I. Buzdin, 11 613 1996 . 9 and J. R. Clem, Phys. Rev. B 44, 759 1991 , and references L. G. Parratt, Phys. Rev. 95, 354 1954 . 10 therein. H. Weber, E. Seidl, C. Laa, E. Schachinger, M. Prohammer, A. 6 Th. Mu¨hge, N. N. Garif'yanov, Yu. V. Goryunov, G. G. Khaliul- Junod, and D. Eckert, Phys. Rev. B 44, 7585 1991 . 11 B. T. Matthias, W. Peter, H. J. Williams, A. M. Clogston, E. lin, L. R. Tagirov, K. Westerholt, I. A. Garifullin, and H. Zabel, Corenzwit, and R. C. Sherwood, Phys. Rev. Lett. 5, 542 1960 . Phys. Rev. Lett. 77, 1857 1996 . 12 P. Lang, B. Dittler, R. Zeller, and P. H. Dederichs, J. Phys.: 7 Th. Mu¨hge, K. Westerholt, H. Zabel, N. N. Garif'yanov, Yu. V. Condens. Matter 4, 911 1992 . Goryunov, I. A. Garifullin, and G. G. Khaliullin, Phys. Rev. B 13 G. Verbanck, C. D. Potter, R. Schad, P. Belien, V. V. 55, 8945 1997 . Moschalkov, and Y. Bruynseraede, Physica C 235-240, 3295 8 K. Theis-Bro¨hl, R. Scheidt, Th. Zeidler, F. Schreiber, H. Zabel, 1994 .