Nicolas
Godard
*,
Patrick
Grysan
,
Emmanuel
Defay
* and
Sebastjan
Glinšek
Materials Research and Technology Department, Luxembourg Institute of Science and Technology, Rue du Brill 41, L-4422 Belvaux, Luxembourg. E-mail: nicolasgodard@live.be; emmanuel.defay@list.lu; Tel: +352 275 888 1
First published on 8th December 2020
Chemical solution deposition (CSD) is a well-established process for the fabrication of metal oxide functional layers such as piezoelectric lead zirconate titanate (PZT) thin films. The latter exhibit an enhanced electromechanical response in the presence of {100} crystalline orientation with respect to the substrate surface. Lead titanate (PbTiO3) seed layers are commonly used to promote the growth of this orientation on platinized silicon, which otherwise usually affords the {111} orientation. In this work, we present a comparative study of two solvents used for the preparation of PZT and PbTiO3 solutions for CSD, i.e. the popular but highly carcinogenic and teratogenic solvent 2-methoxyethanol and the benign solvent 1-methoxy-2-propanol. In addition to tremendous health and safety benefits, we show that 1-methoxy-2-propanol-derived PbTiO3 seed layers promote the {100} orientation more efficiently than their 2-methoxyethanol-derived counterparts, owing to their nanocrystalline microstructure and strong orientation. This resulted in a noticeable enhancement of the electrical and piezoelectric properties, with Pr = 38 μC cm−2, Ec = 55 kV cm−1, εr = 1300, tanδ = 0.04 and e31,f ≈ 14 C m−2.
Platinized silicon (bulk Si/SiO2/TiOx/Pt) is frequently used as a substrate for piezoelectric stacks, owing to the excellent stability and electrical conductivity of platinum. However, the growth of CSD-derived PZT layers directly on platinized silicon usually affords the {111} orientation, as the film adopts the texture of the underlying Pt(111) electrode.5 It is nevertheless possible to induce the growth of the {100} orientation of the PZT film via the use of seed layers deposited on the substrate. Both lead(II) oxide (PbO) and lead titanate (PbTiO3) seed layers were shown to promote the development of {100} texture in PZT.6 The use of seed layers is also a popular strategy for lowering processing temperatures, thus resulting in ecological and economic benefits associated with the reduced energy consumption, as well as technological integration on temperature-sensitive substrates.7
The seed layer can be also deposited by CSD. In the case of PbTiO3, the chemistry of the precursor solution can be essentially very similar to the one used for the fabrication of the PZT layer. A typical solution preparation route for PbTiO3 solutions involves lead(II) acetate and a titanium(IV) alkoxide dissolved in a polar solvent such as 2-methoxyethanol. However, as the seed layer is intended to be very thin (<20 nm), PbTiO3 solutions are usually more dilute. 2-Methoxyethanol is a popular solvent for CSD processing of PZT or PbTiO3 layers as it has excellent chelating properties which can stabilize water-sensitive alkoxides.8
Dekleva et al. studied the behavior of the PbTiO3 sol–gel system in 2-methoxyethanol and evidenced the formation of complex heterometallic structures in solution.9 Sengupta et al. investigated the local coordination environment of the metals in dried PbTiO3 gels and found that homocondensation of the metal species had occurred preferentially during the sol–gel transition and heat treatment.10 Another study by Arčon et al. evidenced the presence of Pb–O–Ti linkages in PbTiO3 precursor solutions based on 2-methoxyethanol.11 The solvent is therefore a key component of these chemical systems as it provides a homogeneous dispersion of the metal cations in solution and mediates the formation of the amorphous metal oxide network. Although stable solutions with a long shelf life can be prepared using 2-methoxyethanol, this solvent is highly carcinogenic and teratogenic. Nevertheless, it is still featured in a number of recent contributions for the processing of PZT thin films.12–14
1-Methoxy-2-propanol has a structure similar to that of 2-methoxyethanol as it possesses an identical backbone and only differs by a methyl substituent in alpha of the alcohol moiety (Table 1). However, unlike 2-methoxyethanol, its degradation in the body does not produce harmful by-products. Indeed, 2-methoxyethanol is successively oxidized to methoxyacetaldehyde and methoxyacetic acid, then integrated into the Krebs cycle, yielding 2-methoxycitrate. The latter interferes with the natural metabolite citrate, impacting important metabolic pathways involved in cell development, hence the adverse carcinogenic effects.15 Thanks to the methyl moiety, the oxidation of 1-methoxy-2-propanol after O-demethylation succesively yields lactaldehyde, lactate and pyruvate, which are natural metabolites.16 1-Methoxy-2-propanol can therefore be considered as a safe solvent and a potential candidate for the replacement of 2-methoxyethanol.
Nevertheless, its use for both the deposition of PbTiO3 seed layers and subsequent growth PZT layers has not been investigated yet. Although it is known that the nature of the seed layer has an influence on orientation of the PZT film grown on top of it, it is not obvious whether the solvent used for processing also plays a role in this process. In the present work, we will present a comparative study between the hazardous solvent 2-methoxyethanol and the safe solvent 1-methoxy-2-propanol for the fabrication of CSD-derived PZT layers on platinized silicon.
Fig. 2 XRD patterns of micron-thick PZT thin films deposited on platinized silicon using (a) 2ME and (b) 1M2P-based solutions (PbTiO3 seed layer and PZT film). Note that the absolute scale is identical for both XRD patterns. A linear scale representation of these XRD patterns is shown in Fig. S2 (ESI†). |
The difference in the quality of the PZT orientation likely resides in the crystallization mechanism of the PZT layer, which is mediated by the PbTiO3 seed layer. To confirm that the seed layer induces the observed orientation effect, we performed a cross-check experiment by making a matrix of both layers deposited from different solvents, as illustrated in Fig. S1 in the ESI.† From the XRD patterns of 200 nm-thick PZT films spin-coated on PbTiO3 seed layers, it appears that the 1M2P-derived seed layer is indeed responsible for the strong {100} orientation in PZT, regardless of the solvent used for the deposition of the latter layer. We therefore proceeded with a closer examination of the microstructure of 2ME and 1M2P-derived seed layers in order to gain a deeper understanding of the solvent influence.
XRD patterns and surface SEM micrographs are shown in Fig. 3. The XRD peaks recorded in θ–2θ geometry and associated with the PbTiO3 phase have relatively low intensities as the seed layer is very thin. Substrate-related parasitic peaks are also present in the XRD patterns. However, it clearly appears that 1M2P-derived PbTiO3 seed layers exhibit a strong (100) orientation with a striking (100)/(001) peak splitting compared to the 2ME-derived ones. A marked difference also appears upon observation of the top surface by SEM. On the one hand, 2ME-derived seed layers consist of nanoislands, where individual PbTiO3 grains adopt irregular shapes. On the other hand, the use of 1M2P promotes the growth of PbTiO3 nanocrystal clusters with well-defined facets. Despite the apparent random orientation of these nanocrystals, it is clear from the XRD analysis that the (100)/(001) orientation prevails statistically. The 2:1 intensity ratio between the (100) and (001) peaks could be explained by stress relaxation of the stand-alone grains, in contrast with the 2ME-derived layers, where the (001) contribution is more pronounced. Indeed, PbTiO3 generally exhibits a negative coefficient of thermal expansion in the ferroelectric phase due to a strong decrease of the out-of-plane lattice parameter with increasing temperature.20,21 This in turn induces a compressive thermoelastic stress in PbTiO3 at room temperature, which favors the (001) orientation. The SEM top views together with the AFM measurements (respectively Fig. 3c and Fig. S5d, ESI†) show that the well-defined grains in 1M2P-derived PbTiO3 are well separated from each other (thus inducing stress relaxation) contrary to 2ME-derived PbTiO3. Out-of-plane piezoresponse force microscopy (PFM) analysis of the PbTiO3 seed layers was performed and results are shown in Fig. S5 (ESI†). In both cases PFM signal is present across the whole surface, proving that both layers are continuous.
Fig. 3 (a) XRD patterns of PbTiO3 seed layers deposited using 2ME and 1M2P-based solutions. The peak at ∼44.5° can be attributed to the substrate (see Fig. S3, ESI†), while the low-intensity peak at ∼28° could be associated to the (110) reflection of rutile (TiO2). A linear scale representation of these XRD patterns is shown in Fig. S4 (ESI†). SEM top views reveal the presence of (b) PbTiO3 nanoislands in the case of the 2ME-derived seed layer and (c) nanocrystals for the 1M2P-derived seed layer. |
Several explanations could be invoked to explain the differences in microstructures and orientations. First, it could be hypothesized that the amount of precursor material deposited on the substrate by spin coating could influence the growth of the crystalline phase. Due to the difference in rheological properties, we have determined in previous experiments that 1M2P-derived layers are approximately 5/4 times thicker than 2ME-derived films processed in the same conditions. We can estimate that the average thicknesses are ∼13 nm and ∼16 nm in the case of 2ME and 1M2P-derived seed layers, respectively. Assuming that the expected thickness of the liquid film is proportional to (where ω is the angular velocity of spinning), we investigated the influence of the spinning rate (therefore amount of deposited material) on the microstructure of the seed layers. The results are presented in Fig. S6 (ESI†). It appears that adjusting the spinning rate such to deposit a similar amount of material does not have a significant influence on microstructure development. Interestingly, when the 1M2P-based PbTiO3 solution is spun at 4700 rpm, a greater amount of well-defined isolated nanocrystals can be seen on the surface (Fig. S6e, ESI†). In these conditions, the development of such different microstructures could be ascribed to different wetting interactions between the precursor solution and the substrate.
Another possible explanation could reside in solution chemistry. The different structures of 2ME and 1M2P likely result in different reactivities towards metal precursors. As a consequence, the species formed in solution upon mixing and refluxing presumably have different structures and distributions of the metal cations. According to Muralt, in the case of sputter-deposited PbTiO3 seed layers, crystallization of the (100) orientation was ascribed to a high partial pressure of PbO in the system.22 In the present case, we could hypothesize that the metal species obtained upon reaction with 1M2P possess lead-rich clusters that could in turn provide a lead-rich environment during crystallization, thus favoring the growth of the (100) orientation.
We also performed an experiment to study the influence of acetylacetone, which is added as a chelating agent for the stabilization of titanium(IV) isopropoxide. As can be seen in Fig. S7 (ESI†), the nanocrystal microstructure of the seed layer is only obtained upon combined use of 1M2P and acetylacetone. We can therefore conclude that the highly oriented seed layers result from a synergistic effect between 1-methoxy-2-propanol and acetylacetone.
In order to gain additional insights, we performed thermogravimetric analysis (TGA) coupled with differential thermal analysis (DTA) of powders obtained by drying the acetylacetone-based PbTiO3 solutions. The results presented in Fig. S8 (ESI†) reveal that the crystallization of the 2ME-derived PbTiO3 powder occurs at higher temperatures than its 1M2P-derived counterpart and follows a different crystallization path, as shown by exothermic contributions to the DTA signal. The decomposition of 2ME-derived powders and conversion into perovskite phase could proceed via carbonate species, which are undesired intermediates as they persist up to high temperatures (∼600 °C).23 The nature of the solvent participating as a metal ligand could influence the formation of carbonates. Indeed, the primary alcohol in 2ME could be oxidized to a carboxylic acid and undergo decarboxylation, which in turn would generate carbonates. The presence of a secondary alcohol in 1M2P could prevent this phenomenon. Further investigations would be required to verify this hypothesis and provide a better understanding of the crystallization mechanism.
The effective transverse piezoelectric coefficient e31,f was evaluated via the four-point bending method.18Fig. 4c shows the evolution of the e31,f with a bias electric field. Films derived from 2ME and 1M2P-based solutions both exhibit a remanent piezoelectric coefficient close to 7 C m−2. However, the maximum values of the piezoelectric coefficient were noticeably higher for the 1M2P-derived sample, with ±13.8 C m−2, as compared to −12.6 C m−2 and 12.1 C m−2 for the 2ME-derived sample. A similar trend was observed when performing piezoelectric characterization via the converse method, where a 20 Hz electrical excitation was applied to the sample and the displacement was monitored as function of the electric field. The characteristic butterfly loops are shown in Fig. 4d. By performing a linear fitting in the −100 to 0 kV cm−1 range and using the formula from Mazzalai et al.19 we extracted values of −13.9 and −14.6 C m−2 for the 2ME and 1M2P-derived samples, respectively. An improvement of the electromechanical response up to 15% in the case of the sample fabricated through the 1M2P route could be ascribed to the improvement of the {100} orientation provided by the solvent. The e31,f values that we report are consistent with the ones obtained for typical CSD-derived {100}-textured PZT layers that feature a chemical gradient (see Fig. S9, ESI†).4,24 Note that the chemical gradient is quantitatively similar for both 2ME and 1M2P-derived layers, but more pronounced in the vicinity of the substrate for the 1M2P-derived PZT film, as shown in Fig. S10 (ESI†).
This last observation could account for the moderate improvement of the piezoelectric response, which results from the competition between orientation and chemical gradient. We speculate that a strong enhancement of the piezoelectric response could be achieved through multi-layered deposition of solutions with varying Zr/Ti ratios to counteract the development of the chemical gradient, as originally proposed by Calame and Muralt.24 However, this investigation is beyond the scope of the present study.
We can therefore conclude that 1-methoxy-2-propanol provides a suitable replacement to 2-methoxyethanol for the processing of PZT films, as the latter exhibit an enhanced electromechanical response that could be attributed to the strong {100} orientation induced by 1-methoxy-2-propanol-derived PbTiO3 seed layers. Moreover, 1M2P is a safe solvent, whereas 2ME is highly carcinogenic and teratogenic and therefore banned in industry.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0tc04066e |
This journal is © The Royal Society of Chemistry 2021 |