An experimental study of copper self-diVusion in CuO, Y2Cu2O5 and YBa2Cu3O7-x by secondary neutral mass spectrometry Jan A. Rebane,a Nikolay V. Yakovlev,a Dmitry S. Chicherin,a Yuri.D. Tretyakov,a Lidia I. Leonyukb and Valery G. Yakuninc aChemistry Department,Moscow State University, 119899 Moscow, Russia bGeology Department,Moscow State University, 119899 Moscow, Russia cPhysics Department, Moscow State University, 119899Moscow, Russia Copper self-diVusion has been studied in CuO, Y2Cu2O5 and YBa2Cu3O7-x ceramics by the SNMS (secondary neutral mass spectrometry) technique in the temperature range 700–900 °C with the stable isotope 63Cu used as a tracer.The lowest diVusion rate and the highest value of activation energy were found for the Y2Cu2O5 phase. Copper self-diVusion was also studied along the c axis of YBa2Cu3O7-x single crystals.It was shown that the anisotropy of Cu diVusion in YBa2Cu3O7-x is rather high. The diVusion rate in the c direction of single crystals is more than two orders of magnitude lower than that in ceramic samples. Since the discovery of high-temperature superconductivity a surface of polished ceramic substrates.The roughness of the number of experimental studies have been concerned with substrates before the deposition was between 0.1 and 0.2 mm mass transport in superconducting oxides. The most common as a determined by the profilometer Talystep (Rank–Taylor– method applied involved a study of the diVusion of radioactive Hobson). The films were deposited by RF sputtering using tracers and the use of diVerent sectioning methods or SIMS thick-film 63CuO targets.The ratio of copper isotopes (secondary ion mass spectrometry) for the determination of 63Cu/65Cu in the films was 95/5. The ratio of copper isotopes depth profiles. In this work we tried to apply a relatively new 63Cu/65Cu in the substrates was 69/31 (the natural ratio). method of depth profile analysis, plasma-SNMS, that has The ‘film’ copper with an isotope ratio 63Cu/65Cu=95/5 was emerged as a result of the development of SIMS.The former used as a tracer in this work. has two advantages in comparison to SIMS. First, a very low To study the anisotropy of copper diVusion in YBa2Cu3O7-x energy of primary ions (several hundred eV) allows the per- the same films were deposited on the (001) surface of formance of depth profiling with very high depth resolution, YBa2Cu3O7-x single crystals (the size of the crystals was ca.which is especially valuable for the analysis of mass transport 3×3×0.1 mm). For YBa2Cu3O7-x all diVusion annealings in thin films and heterostructures. Secondly, relatively easy were performed in an oxygen flow, and for Y2Cu2O5 and CuO quantification of the data is possible, as SNMS is less matrix- in air.dependent than SIMS. Depth profile analyses of the diVusion couples were performed by the SNMS technique in direct bombardment mode (DBM) using an INA-3 set-up (Leybold AG). In this exper- Experimental imental mode the low pressure Kr (Ar) plasma in the analysis chamber of the spectrometer is supported by electron-cyclotron Polycrystalline Y2Cu2O5 was synthesized by the conventional resonance.The sample is situated inside the plasma and ceramic process starting from Y2O3 and CuO. Y2Cu2O5 and separated from it by a grounded aperture. The primary ions CuO ‘low density’ ceramic substrates were prepared by convenare extracted from the plasma volume by a negative voltage tional sintering at 1050 and 950 °C, respectively.The final applied to the sample surface (600 V in our case). These ions density of these substrates was around 90% of the theoretical are used for the sputtering of the sample. Neutral particles values for Y2Cu2O5 and CuO. The average grain size was ca. generated in the course of the sputtering have to travel 10 mm for the both specimens. approximately 5 cm through the plasma to the entrance of the High-density CuO substrates were prepared by hot pressing ion optics.This distance is long enough for eVective post- (pressure 4 GPa, temperature 700 °C, duration 60 min). The ionization by plasma electrons. If plasma parameters, sample density reached in this case was 97%. It was not possible to holder assembly and the extraction voltage applied to the estimate the grain size from SEM photographs. The density of sample are properly adjusted, very high depth resolution can substrates was determined by picnometric weighing in CBr3H.be achieved (up to 2–5 nm).1 The uncertainty of this procedure was determined as ±1% Some problems related to sample charging arise in the (absolute) by using Ge single crystals as a reference material.course of the analysis of insulating materials (Y2Cu2O5 in our CuO substrates were pre-annealed in an oxygen flow prior case). The depth profiling of 63CuO/Y2Cu2O5 diVusion couples to polishing to avoid possible reduction of copper. was performed with a Ni grid (grid period=50 mm) pressed The single crystals of the YBa2Cu3O7-x phase were grown onto the sample surface by a mask.The profilometric measure- from a melt of composition 3525572% (YO1.5–BaCO3–CuO). ments confirmed the rectangular shape of the small craters The mixture was heated in an alumina crucible to 1000 °C and restricted by conductive stripes of the grid and the uniform cooled to room temperature at the rate of 1–5 °C h-1. Single depth of these craters over the ion bombarded area (typically crystals were separated by breaking apart the crucible. 1–2 mm in diameter). EPMA (electron probe microanalysis) was performed with The following working parameters of SNMS analysis were an accuracy better than 1%, using a CAMECA analyzer and used: RF power=150 W; Kr (Ar) pressure=3.0×10-3 mbar; showed that the crystals had the stoichiometric composition. Helmholtz coil current=4.9 A and accelerating potential for The diVusion couples were prepared by deposition of thin films of CuO enriched with the stable isotope 63Cu onto the the primary ions=600 V.J. Mater. Chem., 1997, 7(10), 2085–2089 2085Discussion The experimental procedures employed in this work are quite diVerent from the traditional and very well developed methods of diVusion experiments such as techniques using radioactive tracers and serial sectioning for depth profiling.2,3 The first problem related to the application of thin film diVusion couples is that the thin film, which actually is the source of the tracer, constitutes approximately a quarter of the whole diVusion profile.Moreover it would be advisable to perform some correction for diVerences in sputtering rates between substrates and films.However, our preliminary studies showed that the sputtering rates for CuO, Y2Cu2O5 and YBa2Cu3O7-x do not diVer by more than 25–30%. So the average sputtering rate was applied for the transformation of sputtering time into depth. Fig. 1 Experimental points and simulated curve for ‘film’ copper The other problem that arises in the case of thin film diVusion profile in a CuO/CuO diVusion couple (‘low density’ sub- diVusion couples is related to the grain growth.The temperastrate) annealed for 20 min at 800 °C ture of CuO deposition was relatively low (400 °C) and insuYcient to allow perfectly crystalline material. Intensive Table 1 Copper diVusion coeYcients in polycrystal CuO samples grain growth during the first few minutes of a diVusion annealing under elevated temperature changes the surface T/°C; t /min D/cm2 s-1 D/cm2 s-1 morphology and the resultant broadening of transitions of diVusion experiment high density (97%) low density (90%) between layers in an experimental SNMS depth profile is 850; 10 (8.4±0.2)×10-14 caused not only by mass transport but also by recrystallization. 800; 10 (2.3±0.7)×10-14 (4.8±0.4)×10-14 It is almost impossible to separate these two processes. 800; 20 (4.0±0.2)×10-14 Taking this eVect into account all diVusion couples were 750; 15 (1.5±0.1)×10-14 pre-annealed at 800 °C for 1–3 min. The depth profiles of 750; 30 (1.7±0.08)×10-14 copper isotopes in these pre-annealed samples were used as 700; 15 (3.6±0.8)×10-15 the starting point for further calculations. 700; 30 (3.6±0.8)×10-15 (2.8±0.2)×10-15 700; 60 (5.8±0.2)×10-15 The depth profiles of ‘film’ copper in the starting samples 650; 30 (1.6±0.1)×10-15 before diVusion annealings were not step-like. To correct this 650; 60 (2.3±0.1)×10-15 the ‘additional time’ was calculated in the following way. In 600; 30 (4.0±0.4)×10-16 the first stage the copper diVusion coeYcient was calculated 600; 60 (4.4±0.4)×10-16 assuming that the initial profile was step-like.Then the calculated diVusion coeYcient was used to estimate the time that would be necessary to allow the development of the real initial The second series of diVusion experiments was performed profile starting from an ideal one. In the next stage the sum with ‘high-density’ substrates (97% of the theoretical value) of ‘additional time’ and the real time of a diVusion annealing prepared by hot-pressing.In this case it was not possible to was applied for the calculation of the copper diVusion simulate the whole diVusion profile by a simple error function. coeYcient. As a rule after four to seven iterations both A combined fitting function with an additional term corre- ‘additional time’ and the diVusion coeYcient stop changing.sponding to grain-boundary diVusion was applied in this case The final diVusion coeYcients were on average two to three [eqn. (2)] times lower than the ones calculated under the assumption of C(x,t)=A{1-erf [(x-x¾)/(4Dt)1/2]}+C exp[-B(x-x¾)6/5] an ideal shape of the initial diVusion profile. (2) In this experimental series the thickness of 63CuO films was ca. 0.3 mm and as a consequence the surface concentration of Results the tracer decreased in the course of the diVusion annealings. An experimental study of copper self-diVusion in CuO Experimental diVusion profiles at 700 °C of ‘film‘ copper are shown in Fig. 2. It is clearly seen that the concentration in the The first series of experiments was performed with ‘low density’ films is uniform but decreases with increasing diVusion time.CuO polycrystal substrates. The average grain size for the This means that diVusion in the film is much quicker than in substrate material was ca. 10mm and the thickness of the the substrate, probably due to higher defect concentration. 63CuO film was ca. 1 mm. The diVusion profiles were simulated This uniform decrease of the tracer concentration in the film using the error function solution of the Fick’s second law means that the Gaussian term for volume diVusion is more equation for the case of contact of two semi-infinite bodies appropriate for the description of volume diVusion than the where C(x,t)=concentration of the tracer, D=copper diVusion error function term.However, application of eqn.(3) does not coeYcient and x¾=the thickness of the CuO film. change the calculated values of the copper diVusion coeYcients by more than a factor of two. C(x,t)=A{1-erf [(x-x¾)/(4Dt)1/2]} (1) C(x,t)=A exp[-(x-x¾)2/4Dt]+C exp[-B(x-x¾)6/5] An experimental profile in this diVusion couple at 800 °C (3) and annealing time of 20 min is shown in Fig. 1. The parameters A, x¾ and D were chosen to minimize the value of x2, the The temperature dependences of copper self-diVusion measure of the goodness of the fit.x2 values were in the range coeYcients for both series of experiments are compared in 0.5–2. The position of the film/substrate boundary was deter- Fig. 3 and Table 1 and the data are fitted to an Arrenius mined by the position of the 50% point of the ‘film’ copper equation [eqn.(4)] diVusion profile. Calculated values of copper self-diVusion D=D0 exp(Ea/RT) (4) coeYcients are given in Table 1 together with the errors determined in the course of fitting. where Ea is the activation energy and D0 is the intercept. 2086 J. Mater. Chem., 1997, 7(10), 2085–2089Fig. 4 The temperature dependence of the copper self-diVusion coeYcient in Y2Cu2O5 (Ea=415±20 kJ mol-1) Fig. 2 Experimental SNMS profiles of ‘film’ copper in a CuO/CuO diVusion couple (‘high density’ substrate). Filled circles, before annealing; open circles, after 15 min annealing at 700 °C; triangles, Table 2 Copper diVusion coeYcients in a polycrystal Y2Cu2O5 sample after 30 min annealing at 700 °C. temp./°C; time of diVusion experiment D/cm2 s-1 850; 10 min (1.4±0.1)×10-13 850; 30 min (7.7±0.6)×10-14 800; 190 min (1.5±0.2)×10-14 750; 21 h (1.0±0.2)×10-15 700; 76 h (1.4±0.2)×10-16 presented in Fig. 4 and Table 2. The activation energy of copper diVusion in this phase is almost twice as large as that in YBa2Cu3O7-x and even more than twice the value for CuO. At the same time the absolute values of diVusion coeYcients are the lowest for the temperature range studied.The structure of Y2Cu2O5 can be considered as a stack of CuMO layers parallel to the ab-plane, separated by layers of YO6 octahedra. Copper atoms have distorted squarepyramidal coordination (CuMO distance around 2 A° ) with a fifth oxygen atom at a distance of 2.8 A ° .7,8 The copper diVusion mechanism in Y2Cu2O5 is probably rather complex.First, there are no data in the literature on any cation or oxygen non-stoichiometry of this phase. As a Fig. 3 The temperature dependence of the copper self-diVusion consequence the concentration of vacancies might be much coeYcient in CuO determined for ‘low’ ($; Ea=162±7 kJ mol-1) and ‘high density’ (#; Ea=150±15 kJ mol-1) substrates lower than that in CuO or YBa2Cu3O7-x.Secondly, the value of intercept (Table 3), which is proportional to the entropy of diVusion, is too high even for an interstitial mechanism. The Both absolute values of diVusion coeYcients and activation value of intercept for interstitial diVusion in metals is in the energies are in good agreement. range 10-3–1 cm2 s-1.9 The literature data shows that CuO is non-stoichiometric, most probably due to cation vacancies4 as in other divalent Experimental study of anisotropy of copper diVusion in oxides of the elements of the first transition series such as NiO, YBa2Cu3O7-x CoO and FeO.Hence the very low value of the intercept (5×10-6–5×10-7 cm2 s-1) for the temperature dependence The last set of experiments was performed with YBa2Cu3O7-x of the copper self-diVusion coeYcient can be considered as a single crystals.consequence of a vacancy diVusion mechanism in CuO. This mechanism of cation diVusion has also been established for Table 3 Activation energies and pre-exponentials of the temperature NiO and CoO.5 However the values of the intercept for NiO dependences of copper self-diVusion coeYcients in CuO, YBa2Cu3O7-x and CoO (5×10-2 and 5×10-3 cm2 s-1, respectively)6 are and Y2Cu2O5 much higher than that found for CuO.Such a diVerence can sample Ea/kJ mol-1 log10 (D0/cm2 s-1) be caused by diVerent types of crystal structures: the rocksalt structure of NiO and CoO diVers from the structure of CuO. CuO (97%) 150±15 -6.1±0.9 CuO (90%) 162±7 -5.5±0.4 Copper self-diVusion in Y2Cu2O5 polycrystals YBa2Cu3O7-x (polycrystal) 240±9 -0.65±0.4 YBa2Cu3O7-x (crystal c axis) 280±30 -1.3±1.5 The same experimental procedures were applied to study of Y2Cu2O5 415±20 6.3±0.9 copper self-diVusion in Y2Cu2O5 and the data obtained are J.Mater. Chem., 1997, 7(10), 2085–2089 2087Fig. 5 The temperature dependence of the copper self-diVusion coeYcient in YBa2Cu3O7-x. Filled circles, literature data for YBa2Cu3O7-x polycrystals (Ea=256±4 kJ mol-1); open circles, our data for ceramic YBa2Cu3O7-x samples (Ea=240±9 kJ mol-1); triangles, copper self-diVusion coeYcients along the c axis of Fig. 6 The structure of YBa2Cu3O7-x YBa2Cu3O7-x single crystals (Ea=280±35 kJ mol-1). The data on copper diVusion in YBa2Cu3O7-x found in the Table 5 The anisotropy of cation diVusion in YBa2Cu3O7-x literature10 and obtained in the course of our studies11 are cation Dpoly/Dc ref.shown in Fig. 5. The good agreement between our data and literature data for polycrystalline samples gives us faith in our Ba 1000 14 experimental procedures. The new data on copper diVusion Cu 200–400 this work along the c axis are also presented in Table 4. Ni 50–60 10,13 The anisotropy of copper self-diVusion in this phase is Co 1000 10 surprisingly large.The ratio Dpoly/Dc is estimated to be 200–400 (Dc=cation diVusion coeYcient along the c axis). This result can be explained in the following way. The layers of square diVusion.14 For Ba, Dpoly/Dc is at least 3000–4000, i.e. one pyramids CuO5, separated by yttrium ions, are presumed to order of magnitude higher. The authors of ref. 14 emphasise be rigid. No additional ordering or disordering at elevated that this value represents a lower limit for the anisotropy of temperatures was observed in these layers. On the other hand, Ba diVusion as the measurements were performed on melt- O(1) atoms are very mobile and statistically distributed textured ceramic samples rather than single crystals. The actual between O(1) and O(5) positions (Fig. 6) under the conditions value of the anisotropy could be even larger as there is no of the diVusion experiments. This is confirmed by very high continuous chain of Ba positions in YBa2Cu3O7-x along the anisotropy of oxygen diVusion in YBa2Cu3O7-x (four to six c axis and diVusing in this direction Ba has to substitute for Y. orders of magnitude).12 The anisotropy of Co diVusion measured by Routbort et al.10 Gupta et al.13 suggested that the most probable mechanism is of the same order of magnitude as that for Ba.He suggested of copper diVusion in YBa2Cu3O7-x is via vacancies. The that the reason for this is that Co preferentially substitutes negatively charged copper vacancies that inevitably exist in only for copper in the position Cu(1), as confirmed by neutron the lattice will attract highly mobile and positively charged diVraction.As a consequence the jump distance along c axis oxygen vacancies in the Cu(1)MO(1) layer. As a consequence is at least two times longer than the ab plane. the energy barrier for a copper atom to jump onto an adjacent vacant Cu(1) position will be significantly lower than that for Conclusions a jump onto a Cu(2) vacancy.The available data on the anisotropy of cation diVusion in It is shown that the combination of stable isotope tracers, thin YBa2Cu3O7-x are summarised in Table 5. film diVusion couples and secondary neutral mass spectrometry In spite of large absolute value, the anisotropy of copper can be successfully used for an experimental study of cation diVusion is lower than the anisotropy measured for Ba diVusion in complex oxide materials.Copper self-diVusion has been studied for CuO, Y2Cu2O5 Table 4 Copper diVusion coeYcients along the c axis of YBa2Cu3O7-x and YBa2Cu3O7-x ceramic samples. The lowest diVusion rate single crystals and highest values of the intercept and activation energy were found for the Y2Cu2O5 phase.temp./°C; t /h Self-diVusion in CuO most probably involves a vacancy of diVusion experiment Da/cm2 s-1 mechanism, because the non-stoichiometry in this compound 925; 4 (2.3±0.3)×10-14 is accommodated by vacancies in the cation sublattice exactly 900; 4 (1.8±0.2)×10-14 as found in NiO, CoO and FeO. This assumption was con- 850; 5 (7.7±0.5)×10-15 firmed by low values of the activation energy and the intercept. 800; 10 (6.5±0.7)×10-16 The anisotropy of copper diVusion in YBa2Cu3O7-x was 750; 17 (2.1±0.3)×10-16 found to be in the range 200–400: approximately one order of 750; 36 (3.5±0.3)×10-16 magnitude smaller than that for Ba and Co.Such a high value for anisotropy of copper self-diVusion may be related to the aTwo exponential functions. 2088 J.Mater. Chem., 1997, 7(10), 2085–20897 J. L. Garcia-Munoz and J. Rodriguez-Carvayal, J. Solid State very high mobility of oxygen vacancies in the Cu(1)MO(1) Chem., 1995, 115, 324. layer of YBa2Cu3O7-x structure. 8 R. D. Adams, Ja. A. Estrada and T. Datta, J. Supercond, 1992, 5, 33–38. This research was partly supported by RFBR grant N 9 W. K.Warburton and D. Turnbull, in DiVusion in solids—recent 96–03–33027. developments, ed. A. S. Nowick and J. J. Burton, Academic, New York, 1975, p.180. 10 J. L.Routbort, S. J.Rothman, Nan Chen and J. N. Mundy, Phys. References Rev. B., 1991, 43, 5489. 11 A. Strelkov, J. Rebane and Y. Metlin, J. Mater. Chem., 1993, 3, 1 K.-H.Muller and H. Oechsner,Mikrochim. Acta, 1983, 10, 51. 735. 2 K. HauVe, Reaktionen in und an Festen StoVen, Springer-Verlag, 12 Yupu Li, J. A. Kilner et al., Phys. Rev. B., 1995, 51, 8498. Berlin, 1955. 13 D. Gupta, R. B.Laibowitz and J. A.Lacey, Phys. Rev. L ett., 1990, 3 I. Kaur and W. Gust, Fundamentals of Grain and Interphase 64, 2675. Boundary DiVusion, Ziegler Press, Stuttgart, 1989. 14 N. Chen, S. J.Rothman and J. L. Routbort, J. Mater. Res., 1992, 4 S. Asbink and A. Waskowska, J. Phys.: Condens. Matter, 1991, 7, 2308. 3, 8173. 5 N. L. Peterson, in DiVusion in solids—recent developments, ed. Paper 7/02798B; Received 24th April, 1997 A. S. Nowick and J. J. Burton, Academic, New York, 1975, p.157. J. Mater. Chem., 1997, 7(10), 2085–2089 2089