Journal of Non-Crystalline Solids 349 (2004) 291–298 www.elsevier.com/locate/jnoncrysol
Section 6. Properties of glasses with nano-particles
Structural evolution of LaBGeO5 transparent ferroelectric nano-composites P. Gupta a, H. Jain a
a,*
, D.B. Williams a, O. Kanert b, R. Kuechler
b
Department of Materials Science and Engineering, Center for Optical Technologies, Lehigh University, PA 18015, USA b Institute of Physics, University of Dortmund, 44221 Dortmund, Germany Available online 30 October 2004
Abstract LaBGeO5 is a model material for understanding the formation of transparent ferroelectric nanocomposites (TFNs) that are being developed for several photonic applications. It is one of the few oxides that forms glass easily, devitrifies congruently, and then becomes ferroelectric upon crystallization. Through a two-step heat treatment we have successfully prepared this TFN with stillwellite-type nano-crystallites that are homogeneously formed in the bulk of the sample. The microstructures of the samples have been investigated with optical, scanning electron, and transmission electron microscopy; the crystallites undergo a ferroelectric transition at the Curie temperature that is less than the glass transition temperature. The changes in structure at the molecular level have been monitored by 11B high-field Magic Angle Spinning Nuclear Magnetic Resonance (MAS NMR). The 11B spectra exhibit two well-resolved lines, which can be attributed to planar BO3 and tetrahedral BO4 units, respectively. The relative fraction of the two units depends strongly on the degree of devitrification. Ó 2004 Elsevier B.V. All rights reserved. PACS: 77.84.Lf; 42.70.Nq; 76.60.cq; 81.07.b
1. Introduction Optically transparent ferroelectric glass nano-composites (TFNs), which are glass-ceramics containing ferroelectric nanosize crystallites, have recently received much attention because of their excellent nonlinear optical properties and easy formability [1]. TFNs containing optically nonlinear crystalline phases such as b-BaB2O4, Ba2TiGe2O8, LiNbO3 or BaTiO3 have been successfully prepared and tested to show second harmonic generation [2–8], where the ferroelectric crystalline phase is considered as the main source of second order nonlinearity. In spite of considerable interest in the microscopic
*
Corresponding author. Tel.: +1 610 758 4217; fax: +1 610 758 4244. E-mail address:
[email protected] (H. Jain). 0022-3093/$ - see front matter Ó 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.jnoncrysol.2004.08.212
origin of overall nonlinear response, these systems are too complicated for fundamental studies due to the presence of multiple crystalline phases, preferential crystallization of surface, and variable nanostructure. In this respect, LaBGeO5 is a model material because it crystallizes congruently i.e. there is no redistribution of different elements upon divitrification. LaBGeO5 single crystal with stillwellite structure is ferroelectric and exhibits nonlinear optical properties [9–14]. The corresponding LaBGeO5 TFN has been prepared and shows second harmonic generation [15–19]. Second-order optical nonlinearity, d33, of surface crystallized LaBGeO5, LiBGeO4 and Ba2TiGe2O8 TFNs has been reported to be 0.13 pm/V, 0.7 pm/V, 7 pm/V, respectively [20]. However, preparation of homogeneously bulk crystallized LaBGeO5 has not yet been reported. Also a systematic understanding of the development of the macroscopic nonlinear response of TFNs
292
P. Gupta et al. / Journal of Non-Crystalline Solids 349 (2004) 291–298
with the growth of nanocrystallites inside glass matrix is lacking. The Curie temperature (Tc = 530 °C) of LaBGeO5 single crystal is below the glass transition temperature (Tg = 670 °C), and therefore, the ferroelectric transformation causes stresses in the TFN upon cooling to room temperature. So, we have begun investigating the effect of a confining glass matrix on the development of ferroelectric crystallites in the bulk crystallized LaBGeO5 TFN with the help of optical and electron microscopy. To complement these nano-structural studies, molecular level structural changes resulting from devitrification of the glass have been monitored by Nuclear Magnetic Resonance and X-ray Photoelectron Spectroscopy (XPS). We hope that the information on the structural evolution of bulk LaBGeO5 TFN will provide an insight into the macroscopic nonlinear optical response and clues to design TFN nanostructure with optimum optical response.
2. Experimental procedure LaBGeO5 glass was prepared using a conventional melt-quenching technique. High-purity powders of La2O3 (99.99%), H3BO3 (99.99%) and GeO2 (99.95%) were mixed in the stoichiometric composition La2O3 Æ 2H3BO3 Æ 2GeO2 to provide a 35-g batch. The batch was mixed for 24 h and then melted in a platinum crucible for 1 h at 1250 °C. The melt was poured into a stainless steel mould kept at 400 °C to form glass plates of 3 mm thickness. The glass transition, crystallization onset and crystallization peak temperatures were determined by Differential Thermal Analysis (DTA) at a heating rate of 10 K/min. Previous crystallization studies on the LaBGeO5 glass by Takahashi et al. and Sigaev et al. [15,19] were used as a guide for nucleation and growth temperatures, especially to produce glass nano-composites by a two-step heat treatment consisting of separate nucleation and growth steps. Partially crystallized samples were mechanically polished using CeO2 powder to remove the thin opaque crystalline surface layer. Polarized-light optical microscopy was performed using an Olympus BH2-UMA microscope. The fractured surfaces of selected samples were examined by JEOL 6300 scanning electron microscope. Transmission electron microscope (TEM) specimens were prepared by crushing samples to very fine powder and then suspending the fine powder particles ultrasonically in acetone. Suspended particles were carefully collected using a pipette and placed on a holey carbon film supported by a copper grid. These specimens were then studied using a JEOL 2000 FX TEM operated at 200 kV.
The NMR experiments were performed at room temperature with a high-field (14.1 T) Varian Infinity Plus pulse spectrometer. 11B magic angle spinning (MAS) spectra were acquired at 192.4 MHz using a 9 kHz spinning speed of the 5 mm sample rotor. The free induction decay (FID) was measured by applying short (3 ls) nonselective radio frequency pulses with a recycle delay of about 3 s (the 11B spin-lattice relaxation time some seconds). 32 FIDs were averaged to obtain each spectrum. The shape of the two-line spectra is determined by 11B central (1/2 ! 1/2) transitions broadened by second-order quadrupole perturbation and shifted by the isotropic part of the respective chemical shift [21]. The XPS spectra were acquired using a Scienta spectrometer (ESCA-300) with monochromatic Al Ka X-rays (1486.6 eV) on a freshly in situ fractured sample surface. The instrument was operated in a mode that yielded a Fermi-level width of 0.4 eV for Ag metal. At this level of resolution, the instrument contribution to the line width was very small (<10%). The XPS data consisted of a survey scan over the entire binding energy range and selected scans over the valence-band or core-level photoelectrons peaks of interest at 90° of X-ray incidence. An energy increment of 1 eV was used to record the survey scans and 0.05 eV for the valence-band and core-level spectra. Data analysis was conducted with the ESCA-300 software package using a Voigt function and Shirley background subtraction.
3. Results 3.1. Optical microscopy Depending on the number and size of crystallites, samples appeared from fully transparent to fully opaque e.g. the fully crystallized sample was opaque. The resulting microstructures observed by polarized transmission microscopy are shown in Figs. 1 and 2 for selected conditions where the birefrengent crystallites appear bright and colored against the dark and isotropic glass background. Also micrographs show unfocused, blurry, and spot-like crystallites in the background of the clearly focused crystallites throughout the material due to the uniform crystallization of glass. Crystallites show spherulitic growth [22], which is evident from the optical micrograph of a partially crystallized sample (Fig. 3) and the scanning electron micrograph of a fully crystallized sample (Fig. 4). The effect of nucleation time on the number density of crystallites is evident when Fig. 1(a)– (c) are compared. Similarly, the effect of growth temperature on particle size is noted by comparing Fig. 2(a) and (b); the effect of the growth time on the microstructure is illustrated by Fig. 2(b) and (c).
P. Gupta et al. / Journal of Non-Crystalline Solids 349 (2004) 291–298
293
Fig. 1. Optical micrographs of partially devitrified LaBGeO5 samples, showing the effect of nucleation time on the number density of crystallites (growth at 745 °C for 3 h). Nucleation at 670 °C for: (a) 3 h, (b) 12 h and (c) 18 h.
Fig. 2. Optical micrographs of partially devitrified LaBGeO5 samples, showing the effect of growth time and temperature on particle size (nucleation at 670 °C for 3 h). Growth: (a) at 745 °C for 3 h, (b) at 755 °C for 3 h, and (c) at 755 °C for 6 h.
294
P. Gupta et al. / Journal of Non-Crystalline Solids 349 (2004) 291–298
3.3. NMR
Fig. 3. Optical micrograph of a partially crystallized sample (nucleation at 670 °C for 3 h and growth at 765 °C for 6 h) for volume fraction analysis.
Fig. 6 shows 11B MAS NMR spectra of fully crystallized (opaque) (Fig. 4), partially crystallized (translucent) (nucleation: 670 °C for 3 h, crystallization: 765 °C for 6 h), and 100% glass samples (transparent). The 11 B exists in three and four coordination, which produces peaks at different frequencies [21]. The crystalline sample consists of only tetrahedrally coordinated boron atoms, but the glass sample consists of three-fold as well as four-fold coordinated boron atoms. The concentration of the two types of boron in the glass is nearly equal. The partially crystallized sample shows the fraction of three-coordinated boron at a value between 0 and 0.5. 3.4. XPS X-ray photoelectron spectra were obtained from 100% glass and fully crystallized samples. The oxygen 1s peaks from both samples were decomposed in accordance with the information obtained from NMR data on similar samples and the crystal structure of LaBGeO5 single crystal [10]. The procedure for deconvoluting the oxygen 1s peak is given in Section 4.1 and the results of analysis for the fully crystallized and glass samples are shown in Fig. 7(a) and (b), respectively.
4. Discussion 4.1. Effect of devitrification on the local structure 11
Fig. 4. SEM micrograph of fractured surface of a fully crystallized (opaque) LaBGeO5 sample.
3.2. Transmission electron microscopy (TEM) Based on results obtained from optical microscopy studies, the heat treatment conditions were extrapolated so that the sample would have only nanoscale crystallites. Such an optically transparent, colorless sample was prepared for the TEM studies. Fig. 5(a)–(c) respectively show a bright-field image, a dark-field image and an electron diffraction pattern of a crystallite found in this sample. The bright-field image consists of a bright crystalline part which is a single crystal. This single crystal is one of the biggest crystallites found in the nano-crystalline sample and has been chosen for easier analysis. The dark part in the bright-field image is the glass matrix because the diffraction pattern from this part shows only diffuse rings without any distinct spots. The dark-field image was taken using the 3 0 0 reflection circled in the diffraction pattern (Fig. 5(c)).
B NMR data in Fig. 6 show that the LaBGeO5 glass has three- as well as four-coordinated boron atoms in about equal concentration. When glass changes to the crystalline form all the three-coordinated boron atoms transform into four-coordinated boron atoms. Thus the NMR data for the fully crystallized sample is in agreement with the crystal structure of LaBGeO5, which consists of only four-coordinated boron atoms [10]. In the crystal structure of LaBGeO5 (Fig. 8) boron atoms exist only in four-coordinated B[ 4 tetrahedral units, which make a helical chain around the 31 axis, where [ represents bridging oxygen (BO). Two out of the four oxygen atoms in a B[ 4 unit are shared with the Ge[2 O2 tetrahedral units, while 2 the remaining two oxygen atoms are shared with other B[ 4 tetrahedral units. If we consider covalently shared oxygen atoms or BO as 1/2 oxygen unit and ionically bonded oxygen or nonbridging oxygen atoms attached to Ge (NBO (–Ge)) as 1 oxygen unit then one molecular unit of La2O3 Æ B2O3 Æ 2GeO2, has a total of four out of 10 oxygen atoms attached to two boron atoms and 6 oxygen atoms out of ten attached to two Ge atoms.
P. Gupta et al. / Journal of Non-Crystalline Solids 349 (2004) 291–298
295
Fig. 5. Transmission electron microscopy of a single crystal found in the TFN sample: (a) bright-field image, (b) dark-field image taken using the marked diffraction spot in 5(c), (c) electron diffraction pattern.
Thus, for each formula unit consisting of 10 oxygen atoms we have:
Signal Strength [a.u.]
BO4
4 BOs attached to two B atoms 2 BOs attached to two Ge atoms 4 NBOs attached to two Ge atoms
BO 3
glass partly crystallized fully crystallized
20
10
0
-10
-20 ppm
Fig. 6. 11B MAS NMR of LaBGeO5 glassy, partially crystallized (nucleation at 670 °C for 3 h and growth at 765 °C for 6 hrs) and fully crystallized samples.
That is, the fraction of BO and NBO in the crystal is fBO = 0.6 and fNBO = 0.4, respectively. In the case of the 100% glass sample, the NMR results show that half of the B is three coordinated. Since the concentration of La3+ is more than enough to change all the bridging oxygen connected with boron to nonbridging oxygen, it is assumed that three-coordinated boron exists in the BØ2O form rather than in the neutral BØ3 form. Thus, half of the oxygen atoms
296
P. Gupta et al. / Journal of Non-Crystalline Solids 349 (2004) 291–298
Fig. 7. Deconvolution of oxygen 1s peak in the XPS data for LaBGeO5 : (a) fully crystallized sample, (b) 100% glass sample.
Fig. 8. Unit cell of LaBGeO5 crystal at room temperature. Big and small circles are La and oxygen atoms, respectively. The B[ 4 tetrahedron makes helical chain around the 31 axis and it has its all 2 oxygen atoms covalently bonded to other tetrahedrons. The Ge[2 O2 tetrahedron has its two oxygen atoms covalently bonded to two B[ 4 tetrahedron and remaining two oxygen atoms ionically bonded to La atom.
associated with boron atoms in the crystal transform according to the following reaction: B[ 4 (crystal) ! BØ2O (glass) Further, for one formula unit of glass consisting of a total of 1 B[ 4 þ 1B[2 O units, there are 3 BOs attached to 2 B atoms (2 in B[4 units and 1 in BØ2O units). There is one NBO for each B2O3 molecule (in the BØ2O
unit), 2 BOs attached to 2 Ge atoms, and 4 NBOs ionically connected to the 2 Ge atoms. The oxygen 1s electron in NBO (–B) is assumed to be more tightly bonded than the one in NBO (–Ge) bonds due to the smaller radius and higher electronegativity value of boron atoms. Thus, the peak due to NBO (–B) in the XPS core-level spectrum will be at a higher binding energy than that due to the NBO (–Ge) bond. So the three peaks in the oxygen 1s spectrum of glass are assigned with decreasing binding energy to BO, NBO (–B), and NBO (–Ge) units. Since the area under an oxygen 1s peak provides total number of the particular type of oxygen atoms present. The area ratio of the three peaks should be 5, 1 and 4 respectively from the calculation above. To complement NMR results, oxygen 1s peaks in XPS spectra of the fully crystalline and the 100% glass samples were deconvoluted using ESCA-300 software. The oxygen 1s peak of the crystalline sample was expected to contain fewer peaks based on NMR results and, therefore, fully crystalline spectrum was analyzed first. Then the resulting parameters were used for analyzing the XPS spectrum of the glass. Fig. 7(a) shows a deconvoluted oxygen 1s XPS spectrum for the crystalline sample, which consists of all types of BOs as one peak and the NBOs belonging to the Ge part of the network as the other peak. The ratio of BO and NBO (–Ge) peak areas has been fixed to the calculated value of 6/4 (NMR results), giving a very good fit with reasonable full-width-at-half maximum (FWHM) values. In the case of glass, we fixed the binding energy and the FWHM of the BO and NBO (–Ge) peaks to the value obtained from the deconvolution of the XPS spectrum of fully crystalline sample. It then became necessary to add a third peak between these two peaks, which we have identified as due to the new type of NBOs only found in the glass, viz. the ones in the BØ2O units. As expected, this Oxygen 1s peak is at a higher binding energy than the NBO (–Ge) peak. Its position is nearer to the NBO (–Ge) than to the BO peak, which one would also expect for the two types of NBO peaks. The ratio of the areas of the three peaks has been fixed to 5:1:4 for BO, NBO (–B) and NBO (–Ge), respectively. Their characteristics are given in Table 1. Fig. 7(b) shows that the experimental spectrum is in excellent agreement with the assumed model of the glass structure and independent NMR data. 4.2. Effect of devitrification on the microstructure and nanostructure Optical microscopy and TEM were used as tools to monitor the growth of the crystallites. Fig. 1 shows that when the nucleation time is increased, the number of crystallites increases. Fig. 2(a) and (b) show that when growth temperature is increased, the size of the crystallites increases. The same effect of increase in size is ob-
P. Gupta et al. / Journal of Non-Crystalline Solids 349 (2004) 291–298 Table 1 Relative area fraction, peak position and FWHM values of bridging and nonbridging oxygen peaks as obtained from the deconvolution of the fully crystallized and 100% glass XPS spectra
Crystallized sample BO NBO–Ge Glass sample BO NBO–B NBO–Ge
Relative area (%)
Position (eV)
FWHM (eV)
60 40
531.32 530.00
1.78 1.64
50 10 40
531.32 530.49 530.00
1.78 1.27 1.64
as glass matrix. On the other hand BO3 line has a contribution only from the glass matrix. If we separate the glass and crystalline parts for the experimentally observed BO4 line and BO3 line for a general partially crystallized sample, ½BO4 expt ¼ ½BO4 cryst þ ½BO4 glass
served when growth time is increased (Fig. 2(b) and (c)). Samples shown in Figs. 1 and 2 contain crystallites in the micrometer-size range, which affect the transparency. The aim of this investigation was to fabricate glass containing nanosize crystallites. So these initial optical microscopy observations were used as a guide to fabricate glass-ceramic samples containing nano-crystallites. We used the same nucleation temperature and much longer nucleation time (18 h) while growth temperature and growth time were reduced to 730 °C and 1/2 h respectively. The result was a completely clear glassceramic. Since the crystallites were sub-micrometer size, they were not observable through optical microscopes. The selected area diffraction (SAD) pattern in Fig. 5(c) shows that the nano-size crystallite is a single crystal of stillwellite-type LaBGeO5, which belongs to the trigonal crystal system and P31 space group [10] at room temperature. It has the hexagonal reciprocal lattice as seen in the Fig. 5(c), which shows the h0001izone axis from the hexagonal lattice. Thus, the crystal system of LaBGeO5 is consistent with prior observations from the literature [10]. The single crystal in the dark-field image (Fig. 5(b)) shows bright and dark lines, which are an indication of the presence of the ferroelectric domains as it has been reported for LaBGeO5 [10,12]. 4.3. Determination of volume fraction of crystallites from NMR results We have analyzed the NMR results to determine the volume fraction of the crystallites from the relative intensities of the two 11B lines related to the BO3 and BO4 units (Fig. 6). The notation [] has been used to represent the respective intensities of peaks. From the spectrum of the 100% glass and fully crystallized samples, we find that for the 100% glass: [BO4]expt=0.5 and [BO3]expt=0.5 for the fully crystallized sample: [BO4]expt=1 and [BO3]expt=0. In a partially crystallized sample, we will have the BO4 line having contribution from crystallites as well
297
ð1Þ
With [BO4]cryst=0 for purely glass sample and [BO4]cryst=1 for the fully crystalline sample, ½BO3 expt ¼ ½BO3 glass
ð2Þ
The ratio [BO3]expt/[BO4]glass = a, where a is a constant independent of the amount of the crystallites in the glassy matrix. In the present experiment, a = 1. Then the fraction of the crystallites c in the partially crystallized samples is given by, c¼
½BO4 cryst ½BO4 cryst þ ½BO4 glass
ð3Þ
Since [BO4]cryst=[BO4]expt– [BO4]glass (Eq. (1)) and [BO4]glass = [BO3]expt./a (Eq. (2)), we get c¼1
½BO3 exp t a ½BO4 expt
ð4Þ
For the case of a partially crystallized sample (Fig. 6), [BO4]expt. = 0.6 and [BO3]expt. = 0.4. With a = 1, we predict c = 1/3 or 33%. This result based on the NMR calculation has been verified by the optical microscopy. The optical micrograph of the same sample in Fig. 3 has crystalline volume fraction of 26% from volume fraction analysis. This value is in reasonable agreement with the value predicted by the NMR studies, especially considering the uncertainty in estimates from the twodimensional information available form the micrograph.
5. Conclusion TFNs in the LaBGeO5 system have been successfully prepared through a two-step heat treatment process with trigonal nano-crystallites homogeneously formed in the bulk of the sample. Nucleation time, growth time and growth temperature are control parameters for producing the desired size and number density of the crystallites in the glass. TEM micrographs indicate that the micro-crystallites have trigonal crystal structure of bulk LaBGeO5. At the molecular level, the glass to crystalline form change of LaBGeO5 brings about no change in the composition while there is a striking change in the coordination of boron. The boron coordination changes from a 2D three-coordination in glass to a 3D tetrahedral four-coordination in crystal. LaBGeO5 glass contains three coordinated boron as BØ2O and four coordinated boron as B[ 4 in equal amounts while crystallized LaBGeO5 contains boron in only the B[ 4 form.
298
P. Gupta et al. / Journal of Non-Crystalline Solids 349 (2004) 291–298
Acknowledgments We would like to thank Mr David Ackland and Mr Arlan O. Benscoter for the help in microscopy work, Dr Alfred C. Miller for the help with XPS work and Dr D. M. Smyth for helpful discussion on stillwellite crystal structure.
References [1] H. Jain, Ferroelectrics 306 (2004) 111. [2] Y. Takahashi, Y. Benino, T. Fujiwara, T. Komatsu, J. NonCryst. Solids 316 (2003) 320. [3] Yu-Hua Kao, Y. Hu, H. Zheng, J.D. Mackenzie, K. Perry, G. Bourhill, J.W. Perry, J. Non-Cryst. Solids 167 (1994) 247. [4] T. Komatsu, T. Tawaramaya, K. Matusita, J. Ceram. Soc. Jpn. 101 (1993) 48. [5] H.G. Kim, T. Komatsu, R. Sato, K. Matusita, J. Non-Cryst. Solids 162 (1993) 201. [6] M.V. Shankar, K.B.R. Varma, J. Non-Cryst. Solids 243 (1999) 192. [7] K. Shioya, T. Komatsu, H.G. Kim, R. Sato, K. Matusita, J. Non-Cryst. Solids 189 (1995) 16. [8] N.F. Borrelli, A. Herezog, R.D. Maurer, Appl. Phys. Lett. 7 (1965) 117.
[9] N. Horiuchi, E. Osakabe, Y. Uesu, B.A. Strukov, Ferroelectrics 169 (1995) 273. [10] E.L. Belokoneva, W.I.F. David, J.B. Forsyth, K.S. Knight, J. Phys-Condens. Matter 9 (1997) 3503. [11] Y. Uesu, N. Horiuchi, E. Osakabe, S. Omori, B.A. Strukov, J. Phys. Soc. Jpn. 62 (1993) 2522. [12] B.A. Strukov, B.V. Mill, E.L. Belokoneva, S. Yu. Stefanovich, V.N. Sigaev, Y. Uesu, SPIE 2967 166-170. [13] A. Rulmont, P. Tarte, J. Solid, State Chem. 75 (1988) 244. [14] A.A. Kaminskii, H.J. Eichler, D. Grebe, R. Macdonald, A.V. Butashin, S.N. Bagaev, A.A. Pavlyuk, Phys. Stat. Sol. 198 (1996) K9. [15] Y. Takahashi, Y. Benino, V. Dimitrov, T. Komatsu, J. NonCryst. Solids 260 (1999) 155. [16] Y. Takahashi, Y. Benino, T. Fujiwara, T. Komatsu, J. Appl. Phys. 89 (2001) 5282. [17] V.N. Sigaev, S.Yu. Stefanovich, P.D. Sarkisov, E.V. Lopatina, Mat. Sci. Eng. B 32 (1995) 17. [18] V.N. Sigaev, E.V. Lopatina, P.D. Sarkisov, S.Yu. Stefanovich, V.I. Molev, Mat. Sci. Eng. B 48 (1997) 254. [19] V.N. Sigaev, E.V. Lopatina, P.D. Sarkisov, A. Marotte, P. Pernice, Thermochim. Acta. 286 (1996) 25. [20] Y. Takahashi, Y. Benino, T. Fujiwara, T. Komatsu, Jpn. J. Appl. Phys. 41 (2002) L1455. [21] Lin-Shu Du, J.F. Stebbins, J. Non-Cryst. Solids 315 (2003) 239. [22] C. Fredericci, E.D. Zanotto, E.C. Ziemath, J. Non-Cryst. Solids 273 (2000) 64.
