n PHYSICAL REVIEW B 1 NOVEMBER 2000-IIVOLUME 62, NUMBER 18 Epitaxial mismatch strain in YBa2Cu3O7Àd ÕPrBa2Cu3O7 superlattices M. Varela,1,2 D. Arias,1,* Z. Sefrioui,1 C. León,1 C. Ballesteros,2 and J. Santamaria1 1GFMC, Departamento de Fisica Aplicada III, Facultad de Fisica, Universidad Complutense de Madrid, Madrid 28040, Spai 2Departamento de Fisica, Universidad Carlos III de Madrid, Legane´s 28911-Madrid, Spain ~Received 18 April 2000! The effects of epitaxial strain in ultrathin YBa2Cu3O72d layers have been investigated by x-ray diffraction and transmission electron microscopy. The samples used were high quality@YBa2Cu3O72d (YBCO)N / PrBa2Cu3O7 (PBCO)M#1000 Å superlattices, grown by high oxygen pressure sputtering, withN ranging be- tween 1 and 12 unit cells andM55 unit cells. Superlattice structure is refined by fitting x-ray spectra to a structural model containing disorder related parameters. Epitaxial mismatch strain and the absence of step disorder are found for YBCO layer thickness below 4 unit cells. A surprising reorganization of interatomic distances results, which seems to correlate with the decrease in the critical temperature. For larger YBCO layer thickness, stress relaxes and step disorder builds up. Transmission electron microscopy observations show the presence ofa-axis orientated microdomains, which seem to be correlated to the release of epitaxial strain, provided in plane mismatch is smaller in this orientation. b tu te at iti ra du e n n fe o bl xi ha al it in o th e e s re wt w h fit uf il c his ut ne a ss t the eti- al tice ses. d- tand ure, le, of t- pa- ults h the re truc- ap- an ice tive re- ns- - ion e to at truc- I. INTRODUCTION There is wide evidence on the important role played structure on the superconducting properties of high-Tc cu- prates. A large number of studies have been devoted to s the structural behavior of these complex oxides under ex nal applied forces, such as uniaxial strains or hydrost pressure,1,2 or cation substitution,3 looking for correlations between structural rearrangements and changes in the cr temperature. The main point of interest is to find what int cell distances are relevant to the mechanism of supercon tivity. Structural changes may seriously influence the sup conducting properties of ultrathin epitaxial films. Whe reducing the layer thickness down to a small number of u cells, lattice mismatch to the substrate may seriously af structure, in the form of epitaxial strain. The net effect epitaxial stress will result in a strain pattern not attaina under hydrostatic pressures or the superposition of unia strains.4 On the other hand, step disorder may cause t although the average thickness may stay at the desired v layer thickness fluctuations may locally break the continu of the layer, affecting the transport properties of ultrath layers. Strain induced structural modifications should be c sidered when trying to address this issue. Anyway, from starting point, Locquetet al. have been able to double th critical temperature in the La1.9Sr0.1CuO4 high-Tc supercon- ductor using mismatch strain.5 They show that compressiv in plane epitaxial strain can generate much larger increa of Tc than those obtained by comparable hydrostatic p sures. In presence of lattice mismatch there are two main gro modes depending whether thickness is above or belo critical value, tc . For layer thickness belowtc there is a pseudomorphic growth of the film on the substrate, throug deformation of the crystal lattice. In this way, if the mis between a substrate and the growing epitaxial layer is s ciently small, the first atomic layers which are deposited w be elastically strained to match the substrate through a PRB 620163-1829/2000/62~18!/12509~7!/$15.00 y dy r- ic cal - c- r- it ct f e al t, ue, y n- is es s- h a a fi- l o- herent interface, which minimizes the interface energy. T strain may not only modify in plane lattice parameters, b also may result in a lattice distortion along the out of pla direction. According to the Poisson effect, film growth on substrate with slightly smaller~larger! in plane lattice param- eters may lead to a compression~expansion! in theab plane that can result in an expansion~contraction! in the out of plane direction.6 On the other hand, as the layer thickne increases, the stored strain energy becomes so large a critical thickness that misfit dislocations become energ cally favorable to relax the crystal strain. When this critic thickness is exceeded, the lattice relaxes to its own lat constants, and a new distribution of structural defects ari It is clear that quantitative structural investigations inclu ing disorder related parameters are essential to unders the properties of ultrathin films. X-ray diffraction~XRD! is a nondestructive widely used technique to analyze struct which supplies structural information on the atomic sca averaged over a length scale~structural coherence length! which may be around hundred angstroms. The extraction quantitative information requires the fit of the diffraction pa tern to a structure model containing a large number of rameters in these complex materials and therefore, res may not be very reliable for single epitaxial films, whic usually show a reduced number of diffraction peaks. In case of ultrathin films the x-ray diffraction pattern gets mo and more featureless when thickness is reduced, and s ture refinement becomes meaningless. An alternative proach towards the study of this problem consists of analysis of ultrathin superconducting films in superlatt structures. Superlattice configurations provide a sensi source of quantitative structural information through the finement of x-ray diffraction spectra. Complementary, tra mission electron microscopy~TEM! supplies real space mea surements of the microstructure with a spatial resolut about a few angstroms. Thus we have a structural prob study the atomic structure of defects and their distribution short lateral scales which complements the averaged s tural data provided by XRD. 12 509 ©2000 The American Physical Society lly 1% he is n s- ne bo a it u de ls p re h - u ou ca D re ri i o O ve O at C he ss ro a s de te si in ge re 40 on b y he ate ed ral ses tion ch le. a n- he ize first f he r- ces nd int lat- nd re- fit ted o lu- s g ns, ith - very ate lat- 12 510 PRB 62M. VARELA et al. In a recent Letter, we have reported on epitaxia strained YBa2Cu3O72d ~YBCO! layers in YBa2Cu3O72d /PrBa2Cu3O7 ~YBCO/PBCO! superlattices.7 Since in plane YBCO lattice parameters are about smaller than those of PBCO, then YBCO layers sandwic between PBCO show significant epitaxial strain. In th work, apart of presenting additional XRD data compleme ing the previous report,7 we report complementary transmi sion electron microscopy observations, we have exami the appearence of step disorder for sample thickness a the critical value, and discussed the effect of epitaxial str on doping. We analyze the structure of ultrathin layers w thickness decreasing from a hundred angstroms to one cell. Although the thickness dependence ofTc for very thin layers has been widely discussed in recent past,8 the issue of the superconductivity of a single unit cell has been long bated and remains an open question.9,10 Our study indicates that epitaxial strain for YBCO thickness below 4 unit cel causes important changes in interatomic distances. We pose that these structural modifications may be partly sponsible for the changes in the critical temperature. T possibility of overdoping of the CuO2 planes as a conse quence of structural rearrangements is examined, altho deoxygenation of epitaxially strained superlattices rules this chance. While samples with thickness below the criti value ~4 unit cells! show no step disorder, pointing to a 2 growth mechanism, structure relaxation for thicker layers sults in the appearance of step disorder. This probably t gers a change of the growth mechanism into 3D. Transm sion electron microscopy~TEM! observations are als reported to look at the local nature of defects. II. EXPERIMENTAL The samples for this study were epitaxial YBCO/PBC superlattices grown on~100! SrTiO3 ~STO! substrates using a high pressure~3.4 mbar pure oxygen! multitarget sputter- ing system. High pressure oxygen atmosphere yields a thermalized growth at a very slow rate, 0.13 Å/s for YBC which allows an accurate control of film thickness. Substr temperature was held at 900 °C. The thickness of the PB layer was fixed at 5 unit cells~;60 Å! while the thickness of YBCO layers was changed from 1 to 12 unit cells. T YBCON /PBCO5 motive was repeated up to a total thickne of 1000 Å. The growth began always with PBCO layer, p vided PBCO shows a smaller lattice mismatch to STO~about 0.5% in a and 1% in b! than YBCO. High angle XRD spectr were analyzed using the SUPREX 9.0 refinement program11 which allows obtaining not only fractional unit cell position of the different elements in the c direction, but also disor related parameters such as step disorder, interdiffusion, in face strain, etc. Cross section specimens for transmis electron microscopy were prepared by mechanical grind dimpling and argon ion milling with an acceleration volta of 5 kV and an incidence angle of 8°. HREM studies we carried out in a JEOL 4000 EX microscope, operated at kV and in a Philips CM200 field emission analytical electr microscope operated at 200 kV and equipped with a dou tilt beryllium holder. d t- d ve in h nit - , ro- - e gh t l - g- s- ry , e O - , r r- on g, 0 le III. RESULTS AND DISCUSSION While a structural analysis of single epitaxial films b XRD is limited to obtaining lattice parameters along t growth direction, artificial superlattices are very adequ systems to study the structural implications of strain growth for two central reasons. First, a modulated structu strain profile can be obtained varying the relative thicknes of the components. And second, the structural modula introduced by the artificial periodicity results in a feature ri XRD pattern, which makes a spectrum refinement reliab Figure 1 shows a low angle diffraction spectrum for @YBCO1/PBCO5#1000 Å sample. Low angle spectra are se sitive to the chemical modulation of the layers through t change of the refraction index from layer to layer. Finite s oscillations can be observed below and around the Bragg peak~inset! corresponding to a total thickness o 1000 Å. This is an indication of a surface planitude of t order of one unit cell.12 Sharp satellite peaks from the supe lattice modulation are observed, denoting very flat interfa between YBCO and PBCO layers. Rocking curves arou the ~005! peak, showing FWMH as small as 0.1–0.2°, po to a very small mosaic spread in single films and super tices.F scans around the~102! reflection show a perfect in plane matching, not allowing to resolve between YBCO a PBCO lattice parameters. High angle spectra are shown in Fig. 2 with the cor sponding fits for@YBCON /PBCO5#1000 Å samples withN 51 @Fig. 2~a!# and N58 u.c. @Fig. 2~b!#. The refinement curve, displaced vertically for clarity, shows an excellent to data. The confidence factor of the fit,x2, was highly sen- sitive to the values of the fitting parameters, and was tes to sit at a meaningful mimimum for their final value. T exclude local minima in the multidimensional space of so tions the sensitivity ofx2 to different intracell distances wa checked~see Fig. 3!. This was done by manually displacin single parameters from the final value in both directio monitoring the increase inx2. Negligible small interdiffu- sion was found at the interfaces (,5% in the first layer!. Random step disorder was negligible for samples w FIG. 1. Low angle x-ray diffraction pattern of a @YBCO1 /PBCO5#1000 Å superlattice. Finite size oscillations corre sponding to the total thickness can be observed, pointing to a flat surface. The well defined low angle superlattice peaks indic a high structural perfection of the superlattice. Low angle super tice peaks are indexed asm, and the satellite peaks around the~001! Bragg peaks are indexed asn. to ai ils e gl , C o e n ic e ck O Å tiv n s - ity der red nit ra- as 4 k- y nit ects r d in on - the % - uO th n re r ate PRB 62 12 511EPITAXIAL MISMATCH STRAIN IN . . . YBCO layer thickness below 3 unit cells, and was found show up for thicker layers. For the thinnest YBCO layers interface mismatch str may play a major role in determining the structural deta We found a systematic and monotonous increase of thc parameter when YBCO layer thickness increases from 1 12 unit cells, approaching the value reported for sin 1000 Å thick films ~see Fig. 4!. PBCO lattice parameter however, remained practically unchanged. In plane PB lattice parameters are larger than those of YBCO by ab 1%. Therefore, thin YBCO layers sandwiched in betwe may show in plane expansion and eventually out of pla compression. In plane expansion of YBCO due to latt mismatch for layer thickness below four unit cells has be recently established in YBa2Cu3O72d.13 Increasing YBCO thickness would relax epitaxial stress when the critical thi ness, 3 or 4 unit cells from Fig. 4, is exceeded and YBC would tend to recover single film lattice parameters~dotted line in the figure are lattice parameters of typical 1000 thick films!. Figure 5 shows resistance curves for representa @YBCOn /PBCO5#1000 Å superlattices withn51, 2, 4 and 8 unit cells. Sharp superconducting transitions can be see all cases. The critical temperatureTc was determined using the zero resistance criterion. It was also measured by ac FIG. 2. XRD spectra for~a! @YBCO1 /PBCO5#1000 Å and ~b! @YBCO8 /PBCO5#1000 Å superlattices together with their fits, dis placed two decades vertically for clarity. Extra peaks due to C precipitates present in the sample are marked for clarity, toge with the ~100! substrate peak. FIG. 3. x2 sensitivity to main intracell YBCO distances alongc direction@squares: Y-Cu~2! distance; circles: Cu~2!-Ba distance; tri- angles: Ba-Cu~1! distance#. A common minimum is obtained whe varying manually one of them, keeping the rest fixed in their spective final values. n . to e O ut n e e n - e in us- ceptibility and values agreed within 2 K. It is widely ac cepted that coupling through PBCO layers, and proxim effects, enhance superconductivity in superlattices. In or to check the effect of coupling through PBCO, we measu the critical temperatures of a set of@YBCO1 /PBCOn#1000 Å samples, with PBCO thickness increasing from 1 to 20 u cells. Samples were superconducting with a critical tempe ture depending on the thickness of the PBCO spacer shownin the insetof Fig. 5. It can be observed that above PBCO unit cellsTc becomes independent of the spacer thic ness, as reported previously.14 For the samples of this stud we held PBCO layer thickness in 60 Å, approximately 5 u cells, so, as discussed later, we can discard coupling eff as a source ofTc variations when changing YBCO laye thickness. TheTc of @YBCON /PBCO5#1000 Å superlattices decreases with the number of YBCO unit cells as depicte Fig. 4. Tc values agree with previously reported data similar superlattices.15,16 Additionally, interdiffusion could also be invoked to ac count for the depression of the critical tmperature when YBCO thickness is reduced. It is well known that that a 45 ~atomic! Pr interdiffusion into the first YBCO layer, would result in aTc of 30 K for the resulting alloy.15 Intriguingly, er - FIG. 4. Dependence ofTc and c parameters on YBCO laye thickness, being PBCO thickness fixed in 60 Å~5 unit cells!. Dot- ted lines represent the typical 1000 Å thin film values. FIG. 5. Resistance curves of@YBCOn /PBCO5#1000 Å superlat- tices withn51,2,4, and 8~from bottom to top!. Inset:Tc values for @YBCO1 /PBCOm#1000 Å samples withm ~PBCO spacer! ranging from 1 to 20 unit cells. Coupling through PBCO is found to satur for PBCO layer thickness over 3 unit cells. he - th u- th t th e e m ex a t ie e in e x- io er ce d. n ev i lt d - d ell e Å s ns ice een unt f per- d d- ould of he ess ers, nd ich the s in ins the tly . nd- t is the to s ap- s ta ou the e 12 512 PRB 62M. VARELA et al. this value is very similar to the one obtained for t @YBCO1 /PBCO5#1000 Å superlattices. Our samples, how ever, showed negligible interdiffusion, less than 5% in first layer. The SUPREX software is sensitive to interdiff sion, which is accounted for taking a weighted average of scattering powers of the constituent materials.11,15 Simula- tions of the effect of increasing such chemical disorder up 25% in the first layer are depicted in Fig. 6, showing how intensity of high order satellites is substantially depress Thus, interdiffusion should be discarded as a source forTc reduction in our ultrathin superconducting layers. From the observation of Fig. 4, the changes inTc seem to correlate with those of thec lattice parameter, and it may b then tempting to explain the effect of epitaxial stress in ter of the results of hydrostatic pressure or uniaxial strain periments. It is well known that YBCO shows an anom lously low Tc dependence under pressure, probably owing the opposite dependencies on uniaxial stress ina and b di- rections in plane, dTc /dea51212 K and dTc /deb 52244 K.17 We can use these uniaxial strain dependenc of Tc to estimate roughly theTc changes expected from th observed strains in YBCO/PBCO superlattices. Assum that the structural changes for the one unit cell YBCO sup lattice would occur as a result of an in plane expansionea 50.007 andeb50.011 due to matching with PBCO, the e pected decrease ofTc would be of 1.2 K. Concerning the 1.42% decrease in thec lattice parameter, sincedTc /dec 528 K, a very small increase is expected on the base uniaxial strains. It is clear, therefore, that the superposit of the equivalent uniaxial strains of the lattice paramet due to epitaxial stress can be ruled out as a possible sour the changes inTc when the YBCO thickness is reduce However, it is worth stressing that in plane and out of pla strains have different signs, a situation impossible to achi under hydrostatic pressure. Another point of interest whether the Poisson effect holds in this system as a resu epitaxial stress. Since there are no external forces applie the superlattice, the perpendicular components of stressscc must vanish. The equilibrium condition should then beec 52(Ccaea1Ccbeb)/Ccc . If both ea and eb are tensile, a compressive effect is expected inec , according to the Pois son effect. An estimate can be done using the elastic mo determined by Leiet al.18 (Cca589 GPa,Ccb593 GPa, and FIG. 6. Simulated XRD patterns of@YBCO1 /PBCO5#1000 Å su- perlattices with 0, 10, and 20~from bottom to top! atomic percent interdiffusion of PBCO into the YBCO layer. Dots are experimen data. The depression of the intensity of high order satellites is lined ~dotted ellipses!. e e o e d. s - - o s g r- of n s of e e s of to uli Ccc5138 GPa). If we again assume that for the 1 unit c YBCO sampleea andeb are those expected from the lattic mismatch to the PBCO, a value for thec lattice parameter of 11.51 Å is obtained, which is close enough to the 11.49 obtained from the x ray fits, if one keeps in mind that film show shorterc lattice parameters than bulk samples. It tur out that although the Poisson effect may hold for the latt parameters as a result of in plane lattice mismatch betw YBCO and PBCO, its effect is not strong enough to acco for the drastic decrease observed inTc when the thickness o YBCO layers is reduced. We want to stress that theTc re- duction when YBCO thickness is decreased is not a su lattice effect. The sameTc , and structural results obtaine for 5 PBCO cells were obtained for thicker~up to 20 unit cells! PBCO layers. Moreover, single YBCO layers san wiched between 20 PBCO cells showed the sameTc as the superlattices, although for these samples the structure c not be refined. This means that whatever the influence epitaxial strain within the superconducting layer is, it is t same for superlattices and single films for PBCO thickn over 5 unit cells. However, aside from changes in the lattice paramet the overall stress pattern gives rise to very significant a inhomogenous changes in the intracell distances, wh might be responsible in their own for the changes in superconducting properties. Figure 7 gathers the change some significant intracell distances in thec direction, like separation between consecutive CuO2 planes, distance from the planes to the barium ion, from the barium to the cha and distance from planes to chains. Error bars describe range over which the goodness of the fit did not significan change, i.e., the width of thex2 minimum at a 3% increase Dotted lines in the figure mark the values for the correspo ing distances for bulk samples. The important new resul that epitaxial stress causes very nonhomogenous strain in YBCO cell when the thickness of this layer is reduced up 1 unit cell: the distance between CuO2 planes decrease ~3.8%! when the thickness is reduced and the barium proaches the chains~4%! and moves away from the plane ~3%!. Meanwhile, the change in thec lattice parameter is l t- FIG. 7. Changes in the main YBCO intracell distances along c axis when varying YBCO layer thickness.~a! Distance from planes to chains.~b! Distance between neighboring CuO2 planes. ~c! Plane to barium.~d! Barium to the chain. Dotted lines are th corresponding bulk values after Ref. 19. is va ni nt u a o in e an ec uO e h a um ng in to e tra he A ul t n a on p e ve w s r w pl ce ge g fo e d 0 ili ca ly g n pi- nd for tic ual esult tep g a alue of d in gth. on in in the of . 2. der ked s- ess es O hen x- se PRB 62 12 513EPITAXIAL MISMATCH STRAIN IN . . . only of 1.42%. When the thickness of the YBCO layer increased intracell distances get close to bulk material ues. In fact the fractional atomic positions in the u cell in the thicker YBCO @YBCO8 /PBCO5#1000 Å and @YBCO12/PBCO5#1000 Å superlattices, were in agreeme within 0.5%, with those reported for bulk samples from ne tron data.2 While separation between CuO2 planes and the distance between the barium and the chains become sm under epitaxial stress, as qualitatively~but not quantitatively! expected from the Poisson effect, the opposite happens t distance from the barium to the planes. It is worth remark that, although quite large, structural changes are not unr istic. According to the values of the elastic moduli,18 the 1.42% change in thec lattice parameter is the one expected for uniaxial stress in the order of only 2 GPa along this dir tion. Concerning the distance between consecutive C2 planes which decreases by 3.8% when the YBCO thickn is reduced, it increases by a 4.9% when the rare eart substituted from Y to Nd, increasing the ionic radius by factor of 1.088.19 The distance between the CuO chains and the bari decreasing by 4%, is also known to show a large cha when oxygen content is reduced. In fact this distance creases by 5% in going from the fully oxygenated YBCO the antiferromagnetic insulator YBa2Cu3O6.20 Such changes in intracell distances may have drastic fects on the electronic structure affecting carrier concen tion ~doping or chain to plane charge transfer!. In particular, the changes in the position of the Ba ion may modify t position of the apical oxygen affecting charge transfer. simple reasoning in terms of electrostatic interaction wo suggest that Ba atom approaching the chains may poin overdoping arising from epitaxial strain. In fact in oxyge depleted underdoped samples Ba goes away from the ch when oxygen is removed: for example, for an oxygen c tent x56.6, the Ba atom changes its position by 2.7% a proaching the planes.2 This value is well in the range but in opposite direction to the changes observed in strained lay In order to elucidate the possibility of overdoping we ha obtained a series of deoxygenated@YBCO1 /PBCO5#1000 Å superlattices. Oxygen content was adjusted in as gro samples byex situheat treatments in controlled oxygen pre sure following a stability line of the pressure temperatu phase diagram. For other samples oxygen composition adjusted in situ during sample cool down after sam growth.21 No difference was observed between both pro dures.c lattice parameter was found to increase when oxy content was reduced, as usually observed in YBCO sin films. Intracell distances changed upon oxygen removal, lowing similar trends than those reported for oxygen d pleted bulk samples.2 Annealing of the oxygen deplete samples in pure oxygen at one atmosphere and at 55 completely recovered the initial structure andTc of the fully oxygenated strained superlattice, showing the reversib and control in the process. On removing oxygen, criti temperatures were found to decrease in all cases, similar oxygen defficient YBCO thin films, definitely discardin overdoping as a source ofTc changes in strained ultrathi layers. l- t - ller the g al- - ss is , e - f- - d to ins - - rs. n - e as e - n le l- - °C ty l to For YBCO layer thicknesses larger than 3 unit cells e taxial strain is found to relax, and lattice parameters a intracell distances approximate the reported values YBCO thicker samples (1000 Å). The release of elas strain above the critical thickness is followed by a grad increase in step disorder at the interfaces, probably as a r of a change in the growth mode from 2D to 3D islands. S disorder is quantified by the SUPREX software assumin Gaussian layer thickness fluctuation around a mean v with standard deviations. While s was found to be zero for the epitaxially strained YBCO layers (N,3), increasing YBCO thickness above 4 unit cells resulted in an increase the step disorder parameter,s, up to values of 0.7 for the N512 sample. Multiplyings by the value of thec lattice parameter gives the mean interface roughness average length scales of the order of the structural coherence len Figure 8 shows the interface roughness dependence YBCO layer thickness. Error bars denote the interval with the refinement does not seem to be sensitive to changess. The random layer thickness fluctuation is responsible for absence of high order satellite peaks in the XRD spectra samples with the thicker YBCO layers, as depicted in Fig In order to further analyze the nature of structural disor generated by strain relaxation at a local scale, we chec the films by high resolution electron microscopy. Cros sectional samples were studied with YBCO layer thickn changing in the desired range. Low magnification imag FIG. 8. Step disorder versus YBCO layer thickness. For YBC thickness under three unit cells, step disorder is negligible. W increasing YBCO thickness over the critical one for strain rela ation ~3–4 unit cells! the step disorder arises. FIG. 9. ~a! HREM image of a@YBCO1 /PBCO5#1000 Å sample, with the c-axis perpendicular to the electron beam. An antipha boundary with a displacement ofc/3 along the@001# direction emerging form a substrate step is marked with arrows.~b! High magnification image of the same sample. le d t m om r e a rv t - is hi e h C c r of en e - ial itu- - l e- in to e c 4 or r re- ch t he of in s to in 12 514 PRB 62M. VARELA et al. show the lateral uniformity of the layers over length sca larger than thousands of angstroms. Figures 9~a! and 9~b! show images from an epitaxially straine @YBCO1 /PBCO5#1000 Å sample. Figures 10~a! and 10~b! show images corresponding to the@YBCO8 /PBCO5#1000 Å unstrained superlattice. The latter images were taken with sample@100# direction tilted a few degrees out of the bea direction to enhance the contrast features. Although the c positional contrast between both materials is hard to obse due to the structural similarity of YBCO and PBCO, th fluctuation in YBCO layer thickness is noticeable, confirm ing the presence of random step disorder for epitaxial str released YBCO layers. Antiphase boundaries were obse in strained and relaxed samples as a result of substrate s one unit cell high (c/3). Figure 9~a! shows a typical conser vative antiphase boundary with a displacement ofc/3 along the @001# direction, emerging from a substrate step and ra ing up to the film surface.a-axis oriented microdomains were found in superlattices with a relaxed structure~see Fig. 10!, not showing up in the strained samples. Although t a-axis orientation is known to appear at substantially low growth temperatures (600 °C), we would never expect t growth mode at our high substrate temperatures (900 ° a-axis growth provides a path for a reduced lattice mismat and the wide distributions of these defects appearing in FIG. 10. ~a! HREM image of a@YBCO8 /PBCO5#1000 Å sample. ~b! a-axis oriented microdomains. . t s he - ve - in ed eps - s r is ). h, e- laxed superlattices, might be connected to the relaxation the epitaxial stress. In fact, the lattice mismatch betwe YBCO and PBCO forc oriented growth~0.7% and 1.1% along a and b axis, respectively!, is reduced to 0.5% along thec direction (c/3) for a-axis growth. It is also worth noting that in this orientation, the difference between STO lattic parameter and PBCOc/3 length is only of 0.05%, therefore, we speculate that the apparition ofa-axis oriented micro- twins might favor a reduction of the interface energy in re laxed YBCO layers. In summary we have shown that to some extent epitax stress may be complementary of pressure or cation subst tion in producing internal strains affecting supercon ductivity. Superlattice x ray fitting of high quality @YBCON /PBCO5#1000 Å samples allows obtaining precise information about epitaxial strain in YBCO layers. Intracel strain increases continuously when YBCO thickness is r duced down to one unit cell. Deep nonuniform changes some interatomic distances have been found, which seem correlate with the decrease ofTc when thickness is reduced. In this complicated scenario it is very difficult to ascribe th changes inTc to the changes in any particular interatomi distance. Epitaxial strain is released in layers thicker than unit cells, showing lattice parameters close to those found f thicker (1000 Å) films. While no step disorder is found fo the strained layers, suggesting a 2D growth mechanism, laxed samples show a random thickness fluctuation whi might be connected to a 3D growth mode. Although a present it is not possible to conclude on a direct effect on t superconductivity mechanism, epitaxial strain is a source quite inhomogeneus changes in intracell lattice distances ultrathin YBCO layers. Such nonuniform structural change must be taken into account as extrinsic factors when trying explain the decrease in the critical temperature of ultrath YBCO films. ACKNOWLEDGMENTS Financial support from CICYT Grants No. MAT94-0604 and No. MAT97-0675 is acknowledged. . l. i, s. r- , . *On leave from Universidad del Quindio, Armenia, Colombia. 1A.A.R. Fernandes, J. Santamaria, S.L. Bud’ko, O. Nakamura Guimpel, and I.K. Schuller, Phys. Rev. B44, 7601~1991!. 2J.D. Jorgensen, S. Pei, P. Lightfoot, D.G. Hinks, B.W. Veal, Dabrowski, A.P. Paulikas, and R. Kleb, Physica C171, 93 ~1990!. 3J.P. Attfield, A.L. Kharlanov, and J.A. McAllister, Nature~Lon- don! 394, 157 ~1998!. 4I.K. Schuller, Nature~London! 394, 419 ~1998!. 5J.-P. Locquet, A. Catana, E. Ma¨chler, C. Gerber, and J.P Bednorz, Nature~London! 394, 453 ~1998!. 6A. Fartash, M. Grimsditch, E. Fullerton, and Y.K. Schuller, Phy Rev. B47, 12 813~1993!. 7M. Varela, Z. Sefrioui, D. Arias, M.A. Navacerrada, M. Lucia M.A. Lopez de la Torre, G.D. Loos, C. Leon, F. Sanche Quesada, and J. Santamaria, Phys. Rev. Lett.83, 3936~1999!. 8I.N. Chan, D.C. Vier, O. Nakamura, J. Hasen, J. Guimpel, Schultz, and Y.K. Schuller, Phys. Lett. A175, 241 ~1993!. 9 Jakob, T. Hahn, C. Tome´-Rosa, and H. Adrian, Europhys. Let 19, 135 ~1992!. , J. B. s. , z- S. . 10J. Hasen, I. Lederman, and I.K. Schuller, Phys. Rev. Lett.70, 1731 ~1993!. 11E.E. Fullerton, I.K. Schuller, H. Vanderstraeten, and Y Bruynseraede, Phys. Rev. B45, 9292~1992!. 12O. Nakamura, E.E. Fullerton, J. Guimpel, and I.K. Schuller, App Phys. Lett.60, 120 ~1992!. 13J.P. Contour, M. Drouet, O. Durand, J.L. Maurice, and A. Gauzz Physica C282-287, 689 ~1997!. 14J.M. Triscone and O” . Fischer, Rep. Prog. Phys.60, 1673~1997!. 15E.E. Fullerton, J. Guimpel, O. Nakamura, and I.K. Schuller, Phy Rev. Lett.69, 2859~1992!. 16Q. Li, X.X. Xi, X.D. Wu, A. Inam, S. Vadlamannati, W.L. McLean, T. Venkatesan, R. Ramesh, D.M. Hwang, J.A. Ma tinez, and L. Nazar, Phys. Rev. Lett.64, 3086~1990!. 17W.E. Pickett, Phys. Rev. Lett.78, 1960 ~1997! and references therein. 18M. Lei, J.L. Sarrao, W.M. Visscher, T.M. Bell, J.D. Thompson A. Migliori, U.W. Welp, and B.W. Veal, Phys. Rev. B47, 6154 ~1993!. 19G.D. Chryssikos, E.L. Kamitsos, J.A. Kapoutsis, A.P. Patsis, V ri- o, Jr PRB 62 12 515EPITAXIAL MISMATCH STRAIN IN . . . Psycharis, A. Koufoudakis, C. Mitros, G. Kallias, E. Gama Seale, and D. Niarchos, Physica C254, 44 ~1995!. 20R.J. Cava, A.W. Hewat, E.A. Hewat, B. Batlogg, M. Marezi K.M. Rabe, J.J. Krajewski, W.F. Peck, Jr., and L.W. Rupp, ., Physica C165, 419 ~1990!. 21Z. Sefrioui, D. Arias, M. Varela, J.E. Villegas, M.A. Lo´pez de la Torre, C. Leo´n, G.D. Loos, and J. Santamaria, Phys. Rev. B60, 15 423~1999!.