Atomic layer deposition of PbTiO₃ and PbZr_xTi_1-xO₃ films using metal alkyl and alkylamide precursors

Atomic layer deposition (ALD) processes have been extensively developed for simple oxides (compounds of one metal cation and oxygen), as described in several recent reviews.^1–3 Far less development effort has been done for ALD of complex oxides (compounds of two or more metal cations and oxygen). Simple oxide films are typically utilized for their dielectric or barrier properties. Complex oxide films such as PbTiO₃ and lead-zirconium-titanate PbZr_xTi_1-xO₃ (PZT) are technologically important for their ferroelectric properties, in applications such as memories;^4–6 and for their piezoelectric properties, in applications such as microelectromechanical systems (MEMS).^7,8 PZT films are presently prepared for these applications by chemical solution deposition,^4,6 sputtering,^9,10 or by chemical vapor deposition.^11,12 A reliable ALD process for deposition of ferroelectric films into vertical trenches would be beneficial to enable high density memory devices.¹³ A reliable ALD process for the deposition of piezoelectric films onto vertical sidewalls would enable three-dimensional actuation for MEMS.¹⁴ 

The published literature contains several reports of ALD of lead based Perovskite films, including PbTiO₃,^15–19 PbZrO₃,²⁰ and PZT,^21–24 which is the solid solution of PbTiO₃ and PbZrO₃. Many of these works achieved degrees of success, including stoichiometric or near stoichiometric film compositions for PbTiO₃,^15,17,18 PbZrO₃,²⁰ and PZT.²⁴ In each case, the films were amorphous in the as-deposited condition. The formation of the perovskite phase on subsequent annealing, as evidenced by x-ray diffraction (XRD), was demonstrated for PbTiO₃ (Refs. 16 and 17) and for PZT (Refs. 21–24) films. Ferroelectricity was demonstrated by polarization versus electric field (P-E) scans for selected ALD deposited PbTiO₃ (Ref. 17) and PZT (Refs. 21, 23, and 24) films. Piezoelectricity was demonstrated by scanning piezoforce microscopy for an ALD deposited PZT film.²⁴ Deposition of uniform films of PZT by ALD was demonstrated into patterned trench structures with depth/width aspect ratios of 1.5/1.0 (Ref. 22) and 1.7/1.0.²¹ In spite of these reported successes, many of the above referenced studies also reported difficulties with composition control, particularly with regard to the Pb concentration, including “unusual growth behavior”¹⁸ and “complex interrelated incorporation behavior.”²² As yet, no significant industrial implementation of ALD processes for PbTiO₃ or PZT films has been reported, even with the strong commercial incentive to do so.

In the referenced studies on ALD of Pb based perovskite films, the intended approach was to produce the complex oxide films by combining ALD processes for the component simple oxides. This is a logical approach and would be straight forward to implement in practice. ALD processes for simple oxides commonly use precursors based on metal alkyl or metal alkylamide compounds.^1,3 The relatively weaker metal-ligand or metal-amine bonds react more readily with common oxidizers at low temperatures, compared to the metal–oxygen bonds in metal alkoxide or β-diketonate precursors. Table I lists the precursors used in the referenced previous studies. Many of these studies used the β-diketonates Pb(thd)₂ or Pb(tod)₂ as the Pb precursor, reacted with either water or ozone. No evidence was provided as to the reactivity of these precursor/oxidizer combinations at the ALD temperatures (typically≤300 °C) or to their ability to deposit the component oxide for Pb. Other studies used the mixed ligand alkylamide compound Pb(NEt₂MeOⁱPr)₂ reacted with H₂O,¹⁸ or the metal alkyl compound Pb(Ph)₄ reacted with O₃.^16,20,25 These later combinations are expected to be more reactive at low temperatures, and coincidentally were demonstrated to result in ALD of Pb oxide films.^18,25 Table I also shows that a frequently used precursor for Ti was Ti(thd)₂(OⁱPr)₂ reacted with H₂O.^16,18,21,22 ALD of the TiO₂ simple oxide is more commonly accomplished using either Ti(OⁱPr)₄ (Refs. 26–31) or the metal alkylamide compound tetrakis dimethylamino titanium (TDMAT).^30,32–37 Both of these compounds are highly reactive with water or ozone in the temperature range for common ALD processes. Table I also shows that the previous studies frequently used Zr(thd)₄ reacted with H₂O (Refs. 22 and 24) for the Zr precursor. ALD of the ZrO₂ simple oxide is more commonly accomplished using metal alkylamide compounds such as tetrakis ethylmethylamino zirconium (TEMAZ).^38–43 Both TEMAZ and TDMAT are highly reactive with water or ozone at low temperatures.

The primary objective for this study was to develop reliable ALD process technology for PbTiO₃ and PZT films based on precursors which have shown good compatibility with ALD processes for the component oxides of Pb, Zr, and Ti. The Pb precursor selected in this study was the metal alkyl tetraethyl lead Pb(Et)₄. Similar to Pb(Ph)₄, tetraethyl lead is not reactive with water under neutral conditions. Both Pb(Ph)₄ and Pb(Et)₄ are highly flammable, and therefore should be highly reactive with ozone. Pb(Et)₄ has the practical advantage of higher vapor pressure than Pb(Ph)₄. The selected precursors for Ti and Zr were the metal alkylamide compounds TDMAT and TEMAZ. Both of these compounds are highly reactive with either ozone^{32,33,38–41} or water^{34–37,42,43} in the temperature range of interest for ALD processes (≤300 °C).

The first objective for this study was to demonstrate the control of film composition in the as-deposited condition. Ideally, the film composition should be controlled solely by the relative number of Pb, Zr, and Ti precursor doses, for predictable implementation in device fabrication. A second objective for this study was to demonstrate uniform deposition of PbTiO₃ and PZT films on vertical sidewalls in patterned substrates, to demonstrate the potential for implementation in advanced MEMS and high density memory devices.

The ALD process development work was done in a tube furnace reactor with a quartz tube of 0.75 in. inside diameter. The metal organic precursors were Pb(Et)₄, TEMAZ and TDMAT. Reaction of the Pb(Et)₄ precursor was done using ozone. Reaction of the TEMAZ and TDMAT precursors was done using either water or ozone. H₂O (Aldrich reagent grade) was supplied from a glass bubbler at room temperature. Ozone was supplied from a US Filter ozone generator using ultrahigh purity oxygen. According to the manufacturer, the O₃/O₂ conversion rate is 1%–2%. Initially, we investigated a range of values for the ALD parameters, and their effects on film composition and growth rate. We later settled on the following process parameters, which were used for all results reported here. The deposition temperature was 250 °C for all samples. Precursor dose time was 1.0 s for all three precursors. Purge time after each Pb(Et)₄ precursor dose was 8 s. Purge time after each TDMAT or TEMAZ precursor dose was 20 s. The H₂O dose time was 1.0 s. Purge time after H₂O exposure was 20 s. No purge gas flow was used. The O₃ exposure time was 10 s, at a flow rate of 150 standard cm³/min as measured by the input O₂ flow rate to the ozone generator. Purge time after ozone exposure was 10 s.

The overall ALD process sequence for PbTiO₃ and PZT films consisted of combinations of the separate subprocesses for the component oxides of Pb, Zr, and Ti. Each subprocess consisted of (1) precursor dose, (2) purge, (3) reaction, and (4) purge steps. We initially demonstrated the subprocesses to produce component oxide films for PbO, ZrO₂, and TiO₂. We then combined the subprocesses to produce complex oxide PbTiO₃ and PZT films. The combined process always started with a TiO₂ subprocess, followed by a series of PbO subprocesses repeated a specific number of loops. For ALD of PZT films, we replaced a fraction of the TiO₂ subprocesses with ZrO₂ subprocesses to control the Zr/Ti composition of the film. For all process runs, each TiO₂ or ZrO₂ subprocess was followed by a series of PbO subprocesses. The entire ALD process was computer automated, which allowed control of the relative number of Pb, Zr, and Ti precursor doses over a wide range.

The substrates for investigations of film composition were Insotek platinum coated silicon (Pt/Si) wafer sections, with no patterning. The specified structure of the Pt/Si substrates was 150 nm (111)Pt/20 nm TiO₂/300 nm SiO₂/(100)Si. The concentrations of Pb, Zr, and Ti for all films were determined by x-ray fluorescence (XRF), using an Amptek system with a Ag x-ray source operating at 20 kV and 10 μA. The XRF system uses a broad x-ray beam, resulting in an average composition for the entire sample. The XRF technique requires calibration using samples of known composition. Calibration of the XRF analysis for Pb and Ti was done using results obtained by energy dispersive x-ray spectroscopy (EDS) of a subset of the samples. The EDS analysis was done in an FEI Quanto 200 scanning electron microscope (SEM) with an Oxford Instruments X-Act detector. The SEM operating voltage was 10 kV. This EDS analysis could not determine the Zr concentration in films on Pt coated substrates, due to overlap of the Zr Lα (2.044 eV) and Pt Mα (2.050 eV) peaks. Therefore, a separate subset of samples was analyzed by x-ray photoelectron spectroscopy (XPS). The XPS analysis was performed using a Physical Electronics VersaProbe II instrument equipped with a monochromatic Al Kα x-ray source. The XPS results were used to calibrate the XRF analysis for Zr, using the Zr Kα (15.775 eV) peak, which does not overlap any substrate Pt peaks. All reported compositions are in atomic percent. Oxygen concentration was not measured for any of the films. The chemical formulas reported here assume the most common oxidation state for the stoichiometric compounds. The target composition for the PbTiO₃ and PZT films was stoichiometric Pb_1.0Ti_1.0O₃ and Pb_1.0(Zr_0.5Ti_0.5)O₃, respectively.

Thickness measurements for the component oxide films on Pt/Si substrates were done by optical interferometry, using a Filmetrics system with a white light source. This technique worked well for the component oxides. The interferometry technique is less accurate to determine thickness of the complex oxide films, since it requires knowledge of the film optical constants, which can vary significantly with composition and crystallinity. XRD was done using a Rigaku Geigerflex system, with a Cu Kα x-ray tube.

Following the work to demonstrate ALD of the component oxides, and to investigate film composition for the complex oxide films, we combined the ALD processes to deposit films of PbTiO₃ and PZT onto patterned substrates for investigation of step coverage. The substrates used for step coverage investigations were silicon on insulator (SOI) wafer sections, with 20 μm thick (100) silicon above the buried oxide layer. The substrates were photolithographically patterned and deep reactive ion etched to produce vertical trenches. The etch process used the buried oxide layer as the etch stop. After patterning, the entire substrate was coated with platinum by sputtering. Thickness measurements for the complex oxide films deposited on the patterned substrates were done by SEM observation of fractured cross sections of the samples.

Initially, ALD processes were demonstrated for single component TiO₂ and ZrO₂ films. The TDMAT and TEMAZ precursors could be easily reacted with either water or ozone, in agreement with previous studies on ALD of TiO₂ (Refs. 30 and 32) and ZrO₂,^38–43 respectively. Initially, component oxide films were deposited using increasing precursor source temperatures for each material, while keeping the precursor dose time constant at 1.0 s. For each material, the deposition rate increased as the precursor vapor pressure increased, then leveled off, indicating that the precursor dose process was saturated. Source temperatures for the TDMAT and TEMAZ precursors were selected as 70 and 80 °C, respectively, to be well above this saturation point. As an additional test that the precursor dose step was saturated, films of each material were deposited using a precursor dose time of 2.0 s, with no observed increase in deposition rate. The oxidizer dose times were constant in all runs, at 1.0 s for H₂O and 10 s for O₃. ALD trials were also conducted for each material, as a function of purge time after the precursor dose and after the oxidizer (H₂O or O₃) dose, using the selected precursor source temperatures and the 1.0 s precursor dose time. The selected purge times of 20 s after each H₂O dose and 10 s after each O₃ dose were well into the range for which the deposition rate was independent of purge times, confirming that the reaction step is saturated and the overall process kinetics are ALD limited. Using these selected process parameters, the measured deposition rates were 0.108 ± 0.012 nm/cycle for ZrO₂ and 0.106 ± 0.021 nm/cycle for TiO₂. No significant difference was observed in deposition rates using either H₂O or O₃ for the oxidizer for ALD of TiO₂ or ZrO₂ films. All films were smooth, reflective, and pinhole free as observed by optical microscopy.

For ALD of the Pb component oxide, reacting the Pb(Et)₄ precursor with water resulted in no film growth. Reacting the Pb(Et)₄ precursor with ozone resulted in smooth, reflective, and pinhole free films. ALD trials for PbO were also done as a function of the Pb(Et)₄ source temperature. The selected temperature of 58 °C was well above the point at which the deposition rate became independent of the precursor vapor pressure. The PbO film deposition rate was also found to be independent of purge time after ozone exposure, for purge times ranging from 4 to 32 s. The 8 s purge time was selected to be well within this saturated reaction region, without resulting in excessively long times for the overall ALD process. Under the selected conditions, the PbO film deposition rate was also found to be independent of the O₃ exposure time in the range of 5–30 s, and independent of the ozone flow rate, in the range of 150–300 standard cm³/min, as measured by the input O₂ flow to the ozone generator. The measured deposition rate for the ALD PbO films was 0.112 ± 0.022 nm/cycle. X-ray diffraction showed the Pb oxide films to be amorphous in the as-deposited condition. The oxygen content in these films was not measured in this study. We list the chemical formula here as PbO, since this is the most common oxidation state. However, a similar ALD study using Pb(Ph)₄ and ozone found the as-deposited films to be a mixture of tetragonal and orthorhombic PbO₂, as determined by x-ray diffraction.²⁵ 

Next, we combined the Pb and Ti subprocesses to investigate ALD of PbTiO₃. ALD using a Pb/Ti precursor dose ratio of 1/1 consistently resulted in approximately 30% Pb incorporation in the films, as shown in Fig. 1. This initial 30% Pb incorporation level was found to be independent of the oxidizer used for the Pb precursor, as well as independent of many other ALD process parameters, including substrate temperature, oxidizer dose conditions, and purge times. The oxidizer used for the Ti precursor also had no effect on the results. Data for reaction of the Ti precursor with either H₂O or O₃ are included in Fig. 1. Increasing the Pb/Ti precursor dose ratio up to 12/1 had no effect to increase the film Pb content above the initial 30% concentration, when the Pb precursor was reacted with water. When the Pb precursor was reacted with ozone, the Pb content in the films increased linearly with the Pb/Ti precursor dose ratio, as shown in Fig. 1. The trend line in Fig. 1 represents best case behavior. There were several samples for which the measured Pb content was unusually low, as indicated by the data points which fall below the trend line in Fig. 1, but still at or above 30% Pb. There are apparently two mechanisms affecting Pb content in these films. The first mechanism is independent of the oxidizer used, and results in 30% Pb incorporation when the Pb/Ti precursor dose ratio is 1/1. The second mechanism operates to increase the Pb content above 30%, only when ozone is used as the oxidizer for Pb. Therefore, the unusually low Pb content samples most likely result from conditions associated with the oxidizer during these deposition runs.

The results of Fig. 1 show that a wide range of compositions can be achieved for Pb-Ti oxide films by adjusting the Pb/Ti precursor dose ratio, when the Pb precursor is sufficiently reacted. Stoichiometric Pb_1.0Ti_1.0O₃ was deposited at a Pb/Ti precursor dose ratio of 15/1. The deposition rate for stoichiometric PbTiO₃ films was approximately 1.0 nm/cycle, as measured by optical interferometry.

Once knowledge of the Pb/Ti composition behavior was established, we began investigating ALD of PZT, by replacing some of the TiO₂ subprocesses with ZrO₂ subprocesses. For the PZT ALD trials, all three precursors were reacted with ozone. Table II summarizes results for four PZT ALD process trials on Pt/Si substrates. Zr appears to incorporate more efficiently than Ti in these films. A 1/3 dose ratio for the Zr/Ti precursors resulted in film composition very near the 0.5/0.5 target Zr/Ti composition for films on Pt/Si substrates, as shown in Fig. 2. The Pb content in these films decreased as we substituted Zr for Ti, as shown in Fig. 3. For ALD of PZT films there is a clear trend of decreasing Pb content with increasing Zr/Ti precursor dose ratio. The trend line of Fig. 3 extrapolates to the expected value of Pb/Ti = 1/1 for a Zr/Ti precursor dose ratio of zero. Deposition rates for the Pb deficient PZT films ranged from 2.0 to 2.9 nm/cycle, as measured by optical interferometry. As discussed earlier, the accuracy of the interferometry technique to determine thickness of the complex oxide films is poor, since it requires knowledge of the film optical constants. The optical thickness measurements can vary significantly from the actual SEM thickness measurements.

Using the processes established, we deposited films of each material on patterned SOI substrates, to investigate step coverage for the ALD PbTiO₃ and PZT films. The PbTiO₃ film was deposited for 120 ALD supercycles using a Pb/Ti precursor dose ratio of 15/1. The resulting overall film composition, as measured by XRF, was Pb_0.84Ti_1.16O₃. The PZT film was deposited for 60 ALD supercycles using a Pb/Zr/Ti precursor dose ratio of 15/0.5/0.5. The resulting overall film composition, as measured by XRF, was Pb_0.78Zr_0.40Ti_0.60O₃.

Results of SEM analysis of the ALD PbTiO₃ and PZT films are shown in Figs. 4 and 5, respectively, including cross section images of the film deposited into 5 μm wide by 20 μm deep trenches, enlarged cross section images of the films at the corners of the trenches, and plane view images of the films on the sample top surface. Measurements of film thickness from the enlarged cross section images are summarized in Fig. 6 for both materials. The results show the film thickness on the trench sidewall and trench bottom to be similar, but less than the film thickness on the sample top surface. The top surface film is thicker by a factor of about 1.6 for the PbTiO₃ film and a factor of about 1.4 for the PZT film. The plane view images show the films to be dense and pinhole free.

This work demonstrated atomic layer deposition processes for PbTiO₃ and PZT films, as well as for single component PbO, ZrO₂, and TiO₂ films. The overall ALD process for PbTiO₃ or PZT films is not a simple combination of the component oxide processes. Interdependencies were observed for the Pb, Zr, and Ti compositions in both PbTiO₃ and PZT films. Although the PbO and TiO₂ component films deposit at about the same rate, when these subprocesses were combined 15 Pb precursor doses are required for each Ti precursor dose to produce stoichiometric PbTiO₃. Figure 1 shows that a Pb/Ti precursor dose ratio of 1/1 results in a Pb concentration of 30%. A significantly lower Pb incorporation rate is observed for subsequent Pb precursor doses. Similar (though less severe) behavior was observed for ALD of the complex oxide LaAlO₃.⁴⁴ A La/Al precursor dose ratio of 1/1 resulted in a film composition of La_0.6Al_1.4O₃. Subsequent La precursor doses resulted in a lower La incorporation rate, with a La/Al precursor dose ratio of 4/1 being required to achieve stoichiometric LaAlO₃.

When ZrO₂ subprocesses are substituted for TiO₂ subprocesses, interdependencies were again observed for the Zr/Ti and Zr/Pb composition of the ALD films. Zr was found to incorporate more efficiently than Ti during ALD of PZT films, even though ZrO₂ and TiO₂ component films had similar growth rates when deposited as simple oxides. A Zr/Ti precursor dose ratio of 1/3 achieved the 0.5/0.5 target for the Zr/Ti ratio of the PZT films. However, substituting Zr for Ti during ALD of PZT caused the Pb incorporation rate to decrease. For this reason, we did not achieve the target Pb/(Pb + Zr + Ti) concentration of 0.50 for the PZT films, even though the Pb/(Zr + Ti) precursor dose ratio was the same that achieved stoichiometric PbTiO₃ without Zr.

Once these composition interdependencies were identified, it was possible to achieve a wide range of compositions for Pb-Ti oxide films, including stoichiometric PbTiO₃ as well as Pb-deficient and Pb-rich compositions. This afforded an opportunity to study the annealing and crystallization behavior of these films as a function of composition, which was done in a separate study.⁴⁵ That investigation found all films were amorphous in the as-deposited condition. Only films with an A/B ratio ≥ 1.0 (where A = Pb and B = Zr + Ti) crystallized to the expected perovskite phase on annealing. For the films represented in Fig. 1, only the PbTiO₃ films ≥50% Pb crystallized to the perovskite phase on annealing. This was true even for highly Pb rich films up to 73% Pb. None of the PZT films produced in this study crystallized to the perovskite phase on annealing, since they all had A/B ratios ≤1.0. The development of ferroelectric properties in the films was highly dependent on film composition and annealing conditions. PbTiO₃ films close to the stoichiometric composition showed the highest degree of ferroelectric properties, as evidenced by a nonzero irreversible Rayleigh coefficient and hysteresis in the P-E scans. Figure 7 shows P-E results for a stoichiometric PbTiO₃ film deposited by ALD on Pt/Si. A remanent polarization (P_r) of 13.5 μC/cm² is observed.

The use of metal alkyl and metal alkylamide precursors is shown to be a viable route for ALD of PbTiO₃ and PZT films. ALD of PbTiO₃ and PZT complex oxide films is not a linear combination of the component oxide processes, exhibiting interdependencies for the film Pb, Zr, and Ti composition. Once these interdependencies were quantified, the film composition could be controlled by the relative number of Pb, Zr and Ti precursor doses, and a wide range of film compositions were produced. The resulting PbTiO₃ and PZT films are dense and pinhole free. ALD of the films over patterned trenches with vertical sidewalls showed good thickness uniformity through the trench sidewalls and bottom. However, the film thickness within the trenches was reduced relative to the film thickness on the surface outside of the trench by a factor of about 2/3.

Subsequent work showed that PbTiO₃ films deposited on Pt/Si substrates crystallized on annealing to the expected perovskite structure for films with stoichiometric to Pb rich compositions. Stoichiometric PbTiO₃ films showed ferroelectric properties after annealing, as evidenced by a nonzero irreversible Rayleigh coefficient and hysteresis in the P-E scans.

This work was funded by the Army Research Office under STTR Phase II Contract No. W911NF-14-C-0163.