Concentration and Refractive Index Profiles of Titanium- and Iron-Diffused Planar LiNbO, Waveguides

Transcription

1 D. KP et al.: Conentration and Refrative ndex Profiles of LiNbO, Waveglides 141 phys. stat. sol. (a) 139, 341 (1993) Subjet lassifiation: 78.65; 66.3; S1l.l Fuhhereih Physik (a) and Znstitut fur Clzmie (hi der Clniwrsirdr Osnahriik ') Conentration and Refrative ndex Profiles of Titanium and rondiffused Planar LiNbO, Waveguides BY D. KP (a), B. GATHER (b), H. BENDC (b), and E. KRATZG (a) Depth profiles of titanium, iron, and niobium onentrations are investigated with an eletron miroprobe in planar optial LiNbO, :Ti waveguides additionally doped with iron. The profiles are well desribed by Gaussian funtions allowing the determination of diffusion onstants. The extraordinary refrative index of the wavguides an be desribed by a linear superposition of iron and titanium profiles. 1. ntrodution LiNbO, is a promising material for appliations in holographi volume storage and integrated optis [ 11. Optial waveguides fabriated in single domain rystals of this material by titanium indiffusion or proton exhange are used as integrated swithes, modulators, or spetrometers [2,3]. Photorefrative iron and opperdoped LiNbO, waveguides enable effiient two and fourwave mixing [4]. n partiular, phaseonjugate waves an be generated by anisotropi fourwave mixing in a waveguide with very large effiienies [5]. n this paper we report on the fundamental properties of LiNbO, : Ti waveguides additionally doped with iron (LiNbO, : Ti : Fe). Photorefrative properties of the waveguides an be improved by hightemperature diffusion of transition metals, e.g. iron, from the surfae. This proess, however, leads to a strong hange of the refrative index profiles of the waveguide, too. We analyse the dependene of the refrative index hange on the onentration of indiffused iron and determine the diffusion onstants of titanium and iron. The knowledge of the profiles of both titanium and iron permits the investigation of depthdependent photorefrative properties of the waveguides, e.g. ondutivity or photovoltai onstants [4]. 2. Experimental Methods Our waveguides are prepared in two steps using yut LiNbO, wafers of ongruently melting omposition. At first a thin titanium layer with a thikness of some tens of pm is vauumdeposited on the LiNbO, substrate and then annealed at a temperature of 1 "C in an atmosphere of wet argon gas. The addition of water prevents the outdiffusion of lithium from the surfae of the substrate. n a seond step, a thin iron layer is indiffused in the same manner as the titanium layer. Here we use an atmosphere of wet oxygen, and thus the iron ions are oxidized to Fe3+ during annealing. ') Barbarastr. 7, D4969 Osnabruk, Federal Republi of Germany. 16 physid (a) 139/1

2 242 D. KP, B. GATHER, H. BENDG, and E. KRATZG The effetive refrative indies of the TE and TM modes of the waveguides are measured by the prism oupling method (dark line spetrosopy). From the effetive refrative indies the profiles of ordinary and extraordinary refrative index are reonstruted by the use of an inverse WKB method [6]. For the measurement of the onentration profiles the rear fae of the planar LiNbO, waveguides is polished. This fae and the surfae of the waveguide form an angle of (9 & 1)'. The eletron beam of a miroprobe (aeleration voltage 15 kev, beam width.2 pm, step width.2pm) sans the polished rear fae starting from the surfae. n two suessive measurements the niobium and titanium and the niobium and iron onentrations are determined, respetively. Beause of the low atomi number of lithium (Z = 3), this element is not detetable by this method. The niobium signal is used as referene whih allows to defet the surfae edge reproduibly. Absolute values of the titanium and iron onentrations at the surfae of the waveguide are determined by Xray photoeletron spetrosopy (XPS). This method also provides information about the valene state (FeZf/Fe3+) of the indiffused iron. The penetration depths of titanium and iron in the LiNbO, substrate are large ompared to the thiknesses of the vauumdeposited metal layers. Thus we expet [7] that the onentration profiles of both dopants an be desribed by Gaussian funtions (diffusion from a finite soure), The penetration depth Q depends on the temperature dependent diffusion onstant D and the diffusion time r. We assume a loal, linear relationship between the hange of the extraordinary refrative index An,(y) and the onentrations,(y) [7, 81, and superimpose linearly the influenes of titanium and iron ions, Then the parameters <,,i, tf, should be proportional to the absolute onentrations of titanium and iron, respetively. 3. Experimental Results Using the miroprobe the onentration profiles of titanium, iron, and niobium are measured in the waveguiding layer. The results for the relative onentrations enable a omparison to the refrative index profiles obtained by mode spetrosopy. Furthermore, it is possible to evaluate the diffusion onstants of titanium and iron in LiNbO,. n Fig. 1 the titanium onentration Ti is shown for the set of waveguides Fe3 to Fel2. The thikness of the deposited titanium layer (8 nm) is equal for all waveguides, while the thikness of the iron layer inreases from 3 nm (Fe3) to 12 nm (Fel2O). The diffusion times of titanium ( h) and iron (18 h) and the annealing temperature (1 "C) are the same for all investigated waveguides. n addition, the reonstruted extraordinary

3 Conentration and Refrative ndex Profiles of Planar LiNbO, Waveguides 243 refrative index profile An,(y) of the waveguide FeO (8 nm titanium, annealed h) is inluded in the figure, too. This waveguide has the same onentration profile of titanium as the irondoped samples indiating that the influene of the iron ions on the diffusion of titanium an be negleted. n the region near to the surfae the values of the titanium onentration deviate slightly for different waveguides, but no systemati dependene on the amount of indiffused iron an be derived. The figure onfirms a good orrelation between the onentration profiles of titanium and the reonstruted refrative index profile. By fitting Gaussian funtions to the measured titanium profiles aording to (1) we obtain an averaged penetration depth of Qri = (3.6 &.2) pm. With the diffusion time t = 41 h and (2) we evaluate the diffusion onstant of titanium in LiNbO, at 1 C as DTi = (4.4?.5) x m2 s'. At the surfae the niobium onentration is redued in all investigated waveguides by about 4% ompared to the bulk value. The region of this dereased niobium onentration orresponds rather well to the penetration depth of the titanium ions. Furthermore, by Xray photoeletron spetrosopy we measure a titanium onentration of 3.3 mol% (Fe8) at the surfae. The experimental values of the iron onentration Fe in the waveguiding layer are ompared in Fig. 2. The iron onentration inreases with inreasing thikness of the deposited iron layer. The solid lines show the desription of the dopant profiles by Gaussian funtions. With the exeption of the waveguide with the highest amount of indiffused iron (Fe12) we obtain good agreement of measured values and fitted Gaussian urves. An explanation of this exeption is proposed later. The given onentration values are (as well as the measurements of titanium onentrations) no absolute quantities, but desribe the onentration relations of the waveguides Fe3 to Fe12. n the following the surfae value.5 Ln. 7.4 e p.3 V.2.1 Q.9 : P depth (pm) Fig. 1. Titanium onentration Ti in different titanium and iron indiffused (LiNbO, :Ti: Fe) waveguides ompared with the reonstruted refrative index profile Ane of the titanium indiffused (LiNbO, :Ti) waveguide. The desription of the waveguides orresponds to the thikness of the deposited iron layer. FeO is the titanium indiffused waveguid without iron doping 6*

4 244 D. KP, B. GATHER, H. BENDG, and E. KRATZG i '.O 2.8. Y n &.6 LL Q V.4.2 a %A. a'3 Fe3 Fe4 Fe6 Fe8 A FelOO a Fe depth (pm) Fig. 2. ron onentration Fe in different titanium and iron indiffused (LiNbO, :Ti : Fe) waveguides. The desription of the waveguides orresponds to the thikness of the deposited iron layer. The solid lines are fits by Gaussian funtions (s = ) of the Gaussian funtions fitted to the onentration profiles is alled the relative iron onentration Fe (x = ) of the waveguide. The averaged penetration depth of the iron ions is QF = (1.7 &.6) pm, onsiderably exeeding the value of titanium, Additionally, one has to take into aount the shorter diffusion time of iron of 18 h (for omparison: titanium 42 h). For the diffusion oilstant of iron in LiNbO, at 1 'C we obtain the value of D,, = (1.8 f.2) x lo" m2 s'. An absolute measurement of the iron onentration at the surfae by Xray photoeletron spetrosopy yields a value of 3.7 mol% for the waveguide Fe8. Only Fe3+ is deteted, while the onentration of Fez+ is below the measuring auray. ' 1 8 " 9 6 Y / 4 Fig. 3. Correlation of iron onentration Fe (relative values at the surfae for 2 the waveguides Fe3 to F12) with th thikness of the deposited iron layer. The solid line shows a linear regression of the two quantities

5 Conentration and Refrative ndex Profiles of Planar LiNbO, Waveguides aj o N depth (,urn) Fig. 4. Reonstruted refrative index profiles An, (extraordinary polarisation) of the irondoped (LiNbO, :Ti: Fe) waveguides (Fe3 to Fe12) and the titanium indiffused (LiNbO, :Ti) waveguide (FeOl n Fig. 3 the relative iron onentration Fe (x = ) is orrelated to the thikness of the deposited iron layers. The solid line eluidates the linear dependene of the two quantities. Obviously the indiffused amount of iron does not reah the limit of its solubility in LiNbO, even for the heavily doped samples. The reonstruted extraordinary refrative index profiles of the waveguides FeO and Fe3 to Fe12 are shown in Fig. 4. The refrative index inreases ontinuously with inreasing iron onentration. A similar dependene is obtained for the ordinary refrative index profiles a, N ( Fe Fig. 5. Parameter tpe obtained by fitting of two superimposed Gaussian funtions to the refrative index profiles Ane aording to (3) as a funtion of the relative iron onentration Fe of the LiNbO, :Ti : Fe waveguides (Fe3 to Fel?Ol

6 246 D. KP, B. GATHER, H. BENDG, and E. KRATZG t.8 v). =..6 e,,".4.. V.2 An, Fe1... fit with two Gaussian funtions depth (,urn) t Q, Q u ' Fig. 6. Fit of the onentration profiles T, Fe, to the reonstruted extraordinary refrative index profile Ane for the waveguide Fe1. The dopant profiles rti, Fer are desribed by Gaussian funtions (dashed lines. partly onealed). The parameters of the fit aording to (3) are Sri =.14 and SFe =.86 With the knowledge of the onentration profiles of titanium and iron in the waveguiding layer and the assumption of a loal, linear dependene of the refrative index hange An on the dopant onentrations i, it is possible to distinguish between the influene of titanium and iron ions on the extraordinary refrative index. The ordinary refrative index depends nonlinearly on the titanium onentration [6], and for this reason statements about the additionally irondiffused samples are diffiult. The extraordinary refrative index profiles, however, are orrelated to the onentration profiles of titanium and iron aording to (3). For this purpose we use the fitted Gaussian funtions of Fig. 1 and 2. The parameter tri =.14 is obtained from the refrative index profile of the titanium indiffused waveguide FeO ( tfe = ). With the assumption of equal titanium distributions in the waveguides (FeO, Fe3 to Fe12) this value is kept onstant for the following fits to the extraordinary refrative index profiles of the irondoped samples. Fig. 5 illustrates the obtained values of the parameter tf as a funtion of the relative iron onentration. The measured dependene of the refrative index hange on the iron onentration is pratially linear. As an example Fig. 6 shows the result of the omparison of the onentration profiles with tri =.14 and tfe =.86 and the reonstruted extraordinary refrative index profile An, of the waveguide Fe1. The measured onentration profiles Ti and Fe are fitted by Gaussian funtions with a width of eti = 3.3 pm and efe = 11. pm, respetively. The reonstruted refrative index profile is desribed very well by the fitted urve (superposition of the two Gaussian funtions). For the other waveguides we obtain a similar good agreement.

7 Conentration and Refrative ndex Profiles of Planar LiNbO, Waveguides Disussion The maximum values of the titanium onentration at the surfae of the waveguides Fe3 to Fe12 deviate by about 2% (Fig. l), but a systemati dependene on the total onentration of indiffused iron annot be derived. Some pm away from the sufae differenes between the profiles of the waveguides are omparatively small. One possible reason for the deviations of the measured values at the surfae is the nonperfet preparation of the polished endfae for the miroprobe analysis. Relative to this unertainty, errors in the reproduability of the thikness of the deposited titanium layers, or slightly different annealing onditions of the waveguides are negligible. The titanium profiles of the investigated irondoped waveguides oinide with the refrative index profile of the waveguide FeO (only titanium indiffused). The redued niobium onentration near the surfae of the waveguiding layer agrees well with the penetration depth of the titanium ions and is onsiderably different from the width of the iron profile. This indiates the inorporation of titanium ions on niobium sites. Quantitatively the value of niobium redution at the surfae of 4%, measured with the miroprobe, orresponds well to the titanium onentration of 3.3 mol%, obtained by Xray photoeletron spetrosopy. The measured onentration profiles of iron an be desribed by Gaussian funtions (Fig. 2). Only the waveguide Fe12 is an exeption. The onentration profile of this waveguide learly shows deviations from a Gaussian lineshape near the surfae and is more similar to an inverse errorfuntion (erf (x) = exp ( t2/2) dt). This funtion is a solution of the diffusion equation with an infinite soure; thus it is expeted for the profile shape of the indiffused iron for thik deposited layers. The measured iron onentrations inrease nearly linearly with inreasing thikness of the iron layer (Fig. 3). A saturation of iron in LiNbO, is not reahed even for the highest onentrations of more than 5 mol% (Fe12) iron relative to niobium (Fe8: 3.7 mol%). n homogeneously doped rystals, where iron oxide is already added to the melt, suh high onentrations lead to problems: The rystals exhibit tensions, and they often rak when ooled to room temperature. After annealing in an atmosphere of oxygen the waveguides are strongly oxidized; iron is mostly in the Fe3+ state. This is onfirmed by Xray photoeletron spetrosopy, measuring the Fe3' onentration, with no detetable Fe2'. An estimation of the Fe2+ onentration yields a value whih is at least two orders of magnitude lower than the onentration of Fe3+. From the diffusion profiles of titanium and iron we dedue the orresponding diffusion onstants. The obtained value of DTi = (4.4 &.5) x m2 sc1 is in good agreement with earlier results [7, 91. The estimated error of this value results from the maximum deviation of the penetration depth eti around the averaged value QTi. t should be noted that the diffusion onstant of titanium depends deisively on the Li/Nb ratio of the substrate and dereases from lithiumdefiient (48.1 mol% Li) to lithiumrih (5. mol% Li) samples by one order of magnitude [lo]. The alulated diffusion onstant of iron in LiNbO, is D,, = (1.8 t_.2) x lo" m2 s'. This value is 4 times larger than the one for titanium, while the radii of the two kinds of ions are nearly equal (Pauling radius, oordination number 6, Ti4+ = 61 pm, Fe3+ = 65 pm). An explanation of the large differenes may be different inorporation sites (iron on lithium site, titanium on niobium site) and the bonding energies of the two ions, respetively. X

8 248 D. KP et al.: Conentration and Refrative ndex Profiles of LiNbO, Waveguides The height of the ordinary and extraordinary refrative index profiles Ano, of the waveguides Fe3 to Fe12 inreases with inreasing amount of indiffused iron (Fig. 4). We an separate the influene of titanium and iron on the extraordinary refrative index: the lineshape of the refrative index profile An,(y) an be desribed by the onentration profiles of titanium and iron (Fig. 6). The iron ions lead to a linear inrease of the extraordinary refrative index. 5. Summary and Conlusions n this paper we examine the diffusion profiles of doublediffused LiNbO, : Ti : Fe waveguides. These waveguides are interesting objets for the study of two and fourwave mixing and effiient generation of phaseonjugate waves, respetively. The onentration profiles of titanium, iron, and niobium are measured by an eletron miroprobe. The profiles of both titanium and iron an be desribed very well by Gaussian funtions. At the surfae the onentration of niobium is dereased, and the region of this redution agrees well with the diffusion profile of titanium. Absolute onentration values of the two dopants at the surfae are measured by Xray photoeletron spetrosopy. Up to onentrations of more than 5 mol% iron the limit of solubility of iron in LiNbO, is not reahed. The extraordinary refrative index An,(y) an be desribed by a linear superposition of the iron and titanium onentrations F(y) and Ti(y) aording to (3). This result allows the fabriation of waveguides with tailored properties. Aknowledgements Finanial support of the Deutshe Forshungsgemeinshaft, SFB 225, is gratefully aknowledged. We thank T. Albers and M. Neumann for XPS measurements and P. Hertel for helpful disussions. Referenes [] R. S. WES and T. K. GAYLORD, Appl. Phys. A 37, 191 (1985). [2] A. DONALDSON, J. Phys. D 24, 785 (1991). [3] V. E. WOOD, P. J. CRESSMAN, R. L. HOLMAN, and C. M. VERBER, Topis appl. Phys. 61, 45 ( 1989). [4] D. KP, R. FNK, T. BARTHOLOMAUS, and E. KRATZG, Optis Commun. 95, 33 (1993). [5] D. KP and E. KRATZG, Optis Letters 17, 1563 (1992). [6] P. HERTEL and H. P. MENZLER, Appl. Phys. B 44, 75 (1987). [7] J. VOLLMER, J. P. NSUS, P. HERTEL, and E. KRATZG, Appl. Phys. A 32, 125 (1983). [8] T. BRFMFR, D. KOLFWE, H. KOSCMEDER, and W. HLAND, Fresenius J. anal. Chem. 333, 485 (1989). [9] M. FUKUMA, J. NODA, and H. WASAK, J. appl. Phys. 49, 3693 (1978). [lo] R. J. HOLMES and D. M. SMYTH, J. appl. Phys. 55, 3531 (1984). (Rewired June 2, 1993)