Influence of the oxidation state of SrTiO 3 plasmas for stoichiometric growth of pulsed laser deposition films identified by laser induced fluorescence

Influence of the oxidation state of SrTiO3 plasmas for stoichiometric growth of pulsed laser deposition films identified by laser induced fluorescence Kasper Orsel,1,a Rik Groenen,2 Bert Bastiaens,1 Gertjan Koster,2 Guus Rijnders,2 and Klaus-J. Boller1 1Laser Physics and Nonlinear Optics, Department of Science and Technology, MESA+ Institute for Nanotechnology, University of Twente, Enschede, The Netherlands 2Inorganic Materials Science, Department of Science and Technology, MESA+ Institute for Nanotechnology, University of Twente, Enschede, The Netherlands

Here, we investigate the stoichiometry of the film growth while systematically varying the chemical background gas composition, changing the mixture of O 2 and Ar, while maintaining the absolute total pressure constant.For an increasing fraction of O 2 , we observe a transition from non-stoichiometric to stoichiometric film growth.Specifically, the film quality shows a distinct leap to perfectly stoichiometric STO, which occurs at a certain O 2 fraction (≈60%) of a total pressure of 0.1 mbar.A total background pressure of 0.1 mbar is chosen as the plasma expansion is no longer in the ballistic regime and has a strong interaction with the background gas, moderating the kinetic energy of the ablated species. 7Argon and oxygen have a similar atomic weight and PLD plasmas exhibit very similar expansion dynamics in either Ar or O 2 . 9By mixing Ar and O 2 , we maintain a constant total pressure in all measurements with comparable gas dynamic processes, such as expansion, propagation speed, and collision rates, for all gas mixtures.As the ablation laser fluence is kept constant for all measurements, changes in plasma composition will be primarily caused by a change in chemical processes and not by changing dynamics.
By comparison with optically recorded TiO and SrO spatiotemporal distributions in the plasma, we find that the oxidation state of Ti and Sr upon arrival at the substrate is correlated with the stoichiometry of growth.The optical measurements show that all Ti needs to be oxidized at the substrate location for achieving stoichiometric growth.Based on the SrO measurements, we conclude that this is also the case for Sr.These results provide a direct link and thus predictability for controlling the chemical plasma composition and subsequent growth of complex oxide thin films.
The spatiotemporal mapping of the plasma constituents is carried out in a custom built PLD chamber, a detailed description of which can be found in a previous article of ours. 10 We use laser induced fluorescence (LIF) instead of the more commonly used optical emission spectroscopy (OES), 7,11,12 because OES is limited to the detection of excited particles that happen to spontaneously fluoresce.Another disadvantage of OES is that the excited state populations are orders of magnitude lower than the ground state populations when the plume has expanded close to the substrate. 13In contrast, LIF enables us to excite and detect ground state species at freely selectable locations and times and in a chemically specific manner.][16] Target ablation is done with laser pulses generated by a KrF excimer laser (248 nm, 30 ns duration FWHM, operating at 2 Hz).A mask, placed in the KrF laser beam to select a spatially uniform beam, is imaged onto the target, resulting in a laser spot of 0.91 × 2.42 mm 2 .Through control of the laser output energy, the laser fluence is kept at 1.3 ± 10% J/cm 2 during all measurements.The UV excitation wavelengths for LIF in the range from 250 to 350 nm are generated by frequency doubling the output of a dye laser pumped with the second harmonic (532 nm) output of a Q-switched Nd:YAG laser (7 ns FWHM).The UV output, which has a pulse duration of 4 ns FWHM, a bandwidth of 8.1 pm and 75 µJ per pulse, is transformed into a thin sheet in the plane of the forward propagation of the plasma plume from the ablation spot on the target to the center of the substrate.The sheet has an in-plane focus of approximately 0.4 mm thickness.
To image the neutral titanium distribution in the plasma plume, we use a transition from the atomic ground state (3d 2 4s 2 a 3 F) to the 3d 2 ( 1 G)4s4p( 3 P • )v 3 F • state at 294.1995 nm.Relaxation occurs to the 3d 3 ( 4 F)4sb 3 F state via fluorescence at 445.3313 nm. 17A narrowband interference filter with a bandpass of 437-447 nm transmits the LIF of Ti while suppressing the thermally induced spontaneous emission (SE) of the plasma.The excitation of titanium oxide is done using an X 3 ∆ → D transition at 301.47 nm and detecting the red-shifted LIF between 301 and 310 nm. 18trontium oxide is excited using an X 1 Σ → C 1 Σ transition at 350.38 nm and detected using the red-shifted LIF from relaxation to various different rotational and vibrational states of X 1 Σ. 18 For both TiO and SrO, the SE of the plasma is suppressed by a colored glass band-pass filter transmitting 275-375 nm.A background subtraction using measurements with the LIF beam blocked is applied to all LIF measurements to remove any residual SE from the plasma that is transmitted by the band-pass filters.To reduce the influence of shot-to-shot fluctuations, all measurements are averaged over 30 shots.To improve the quality of the spatiotemporal mapping measurements, we decided to carry these out with the substrate removed, because in our setup with an oblique angle of the LIF beam, the substrate caused unwanted optical reflections.This modification avoids artifacts 106103-3 for the relevant observation times, until the plume arrives at the location of the substrate.The films deposited for the XRD measurements are grown on a substrate heated to 710 • C. The influence of substrate heating on the propagation dynamics of the plasma plume, 19 present during deposition but absent during LIF measurements, is largely avoided by using laser substrate heating instead of the more commonly used resistive substrate heating.We performed comparative measurements of plume propagation with either a laser heated target at 710 • C or at room temperature, 20 which showed little difference, demonstrating the validity of the LIF measurements without heated substrate.In contrast, plume propagation with a resistive heated substrate at 710 • C showed, compared to room temperature, large changes in propagation similar as described by Sambri et al. 19 To investigate the chemical composition of the plasma plume and, more specifically, the effects of the background gas composition on the oxidation of the plasma constituents, we attempted to map the spatio-temporal distribution of all relevant and expectantly dominant plasma constituents which are Ti, TiO, TiO 2 , Sr, and SrO.The location, density, and arrival time of these species is mapped in a background gas of which the total pressure is held constant at 0.1 mbar.The partial pressure of O 2 is step-wise increased from 0% to 100% in order to step-wise increase the chemical reactivity.
The LIF signal of strontium atoms excited at either 293.18 and 689.25 nm could be detected, unfortunately, only close to the noise level and only at small delay times (<10 µs) in pure Ar.Excitation at 460.73 nm might yield a higher fluorescence signal.However, generating this wavelength is not possible with our current setup.Similarly, measuring TiO 2 densities proved difficult with excitation lines at 562.56 nm and 556.67 nm as described by Zhuang et al., 21 due to strong background noise from scattering of the LIF laser, when using the LIF laser in the visible range.
To be able to relate the LIF signal to the material density, all excitation transitions used are driven into saturation.In saturation, a variation in laser power, for example, pulse-to-pulse fluctuations or spatial inhomogeneities, will not result in variations of the LIF signal.Thereby, the LIF signal becomes proportional to the density of the excited species.However, since the measurements are not calibrated on an absolute scale, the density of one species cannot be absolutely compared to that of another species.
Figure 1 shows typical results of distribution measurements for a cross-section of the plasma plume in the propagation plane from the target (z = x = 0 mm) to where the substrate is positioned during growth experiments (x = 0 mm, z ≈ 50 mm).Two images of half of the plasma plume are shown next to each other for easier comparison.Figure 1 region (around z = 50 mm).Figure 1(b) shows the distribution of SrO (left) next to TiO (right) under the same conditions.These measurements display a strong spatial separation of the different plasma species.It can be seen that SrO is located primarily in the outer most edges of the plume, whereas TiO is largely absent from the plasma edges and is concentrated in the center of the plume.Neutral titanium is found to be confined even more to the center, as well as close to the target.In Fig. 2 are shown the ground state population densities of Ti, TiO, and SrO along the propagation axis (at x = 0 mm in Fig. 1) as function of the distance from the target, z, at different partial pressures of O 2 .The densities and oxidation state of plasma species along the propagation axis are of particular interest, especially close to the substrate location, as this displays the material that arrives at the center of the substrate, the location where growth is typically being optimized.Species further away from the propagation axis will either deposit on the outer ranges or edges of the substrate or even disappear into the PLD chamber.These measurements are taken for a delay of 35 µs between ablation and detection, at which time the plasma plume has expanded to the substrate location, which is typically located 50 mm from the target.To emphasize the strong differences in spatial changes of densities, all densities are normalized to the maximum of the Ti, which itself is normalized to unity.Since our approach monitors the relative spatiotemporal distribution for the various species, the ratio in absolute data between different species reflects their ratio in fluorescence yield.The SrO density determined as described above is close to our detection limit (10 −3 in Fig. 2), because the LIF signal is only about a factor of ten above the background noise when no O 2 is supplied (100% Ar).The noise is caused by residual spontaneous emission of the plasma that is not removed by the background subtraction.In the Ti distribution, a remarkable drop by almost two orders of magnitude can be seen for all distances from the target when the oxygen fraction is increased but is less than 80%.At 80% and above (which is pure O 2 ), the Ti density does not decrease and the distribution does not change anymore.In other words, the Ti distribution along the z-axis retracts away from the substrate position (z = 50 mm) with increasing O 2 fraction.The on-axis TiO density distribution is low when no O 2 is supplied (lowest trace).The highest TiO concentration, at all z-values, is seen with 20% O 2 , and decreases moderately with increasing O 2 fraction, while the shape of the TiO distribution does not change much except for an overall factor.Another observation is that the density distribution in pure O 2 is actually very similar to the case when no O 2 is supplied at all (pure Ar).The SrO distribution shows an increase by almost an order of magnitude in density in the front of the expanding plume with increasing O 2 fractions, while the density closer to the target slightly decreases.Similar to Ti, the distributions at 80% and 100% O 2 are nearly identical.
In Fig. 3 we show the total amount of Ti, TiO, and SrO present in the plume as function of delay time from target ablation for different partial pressures of O 2 .This plot allows to identify the chemical processes taking place in the entire plasma plume, and specifically, to separate the influence of the oxygen that is inherently present in the plasma plume, supplied by the target, and the oxygen that is supplied by the background gas.The total number of particles shown in Figure 3 is acquired by spatially integrating the LIF signal for Ti, TiO, and SrO over the entire plasma plume, taking the cylindrical symmetry around the z-axis (x = y = 0) into account.Similar to the density distribution in Figure 2, all plots of the particle numbers are normalized to the maximum number of particles of Ti, which itself is normalized to unity.The first 20 µs are not shown, as part of the plasma plume close to the target is obscured by the target holder and therefore not accessible for LIF measurements.Similar to the density distributions, Ti shows the most prominent change with increasing oxygen fraction.Without any O 2 supplied, the total amount of Ti atoms is nearly constant in time, whereas the Ti amount decreases strongly, by almost two orders of magnitude, when the oxygen fraction is increased.With O 2 present, the amount of Ti decreases vs. time, with a steepening slope the more O 2 is present.In comparison, a large increase in total amount of TiO is visible when a small amount of oxygen is added, but the amount of TiO decreases again with higher O 2 fractions, returning to values in pure O 2 that are almost identical to those without O 2 .As function of delay time, the low amount of TiO particles is almost constant in a pure argon background, turning into a weak decrease vs. time with increasing oxygen fraction.The total number of SrO molecules as function of the oxygen fraction in the background gas shows only a small increase.Also, as function of delay time, the number of SrO particles remains almost constant for all oxygen fractions.The strong fluctuations in amount over time are caused by the large amount of noise present close to the detection limit.
Our physical interpretation of the spatiotemporal data in Figs. 2 and 3 is that a strong mixing of the background gas into the plasma plume takes place, facilitating the oxidation of Ti to TiO and of TiO to TiO 2 .Both plots show that both the density and amount of titanium steadily decrease with an increasing oxygen fraction in the background gas.Strong mixing of the background gas and plasma plume has been previously concluded by Sasaki and Watarai. 22t higher levels of oxygen (80% and 100% O 2 ), it can be seen that no significant amount of neutral Ti reaches the substrate location.Density distributions at longer delay times even show the front of Ti retracting towards the target.In comparison, the total amount of Ti in pure Ar remains almost the same over time, indicating that little to no oxidation takes place in the plasma plume with oxygen released from the target.This shows that it is the O 2 from the background gas that is responsible for fully oxidizing the plasma species.Interestingly, very little difference is visible between the Ti densities at 80% and 100% oxygen, indicating that above a certain threshold, somewhere between 60% and 80%, the amount of available oxygen no longer is the bottleneck for oxidation of Ti.The described reaction dynamics are confirmed when inspecting the next species in the oxidation of Ti, TiO, and TiO 2 , which is the TiO concentration.Our measurements show that the density of TiO strongly increases when a small amount of oxygen (20%) is added to the background gas, but steadily decreases again when the percentage of oxygen is increased further.We conclude that the latter decrease is caused by further oxidation of TiO to TiO 2 , because TiO is an intermediate between Ti and TiO 2 and the Ti density continuously decreases with an increasing O 2 fraction, although the TiO 2 molecular fluorescence signal turned out to be too weak for detection with our setup.
No SrO is detected in the plasma when ablating in pure argon, or at least only present in small quantities.The density measurement in pure Ar yielded signals of the same strength as the noise, as shown in Figure 2, making these measurements less reliable.Nevertheless, the lack of SrO in pure Ar and the relatively small increase of the amount of SrO with O 2 increasing from 20% to 100% suggests that Sr oxidizes strongly and only requires little O 2 to fully oxidize.This is supported by the observation that adding a small amount of oxygen in the background gas lets SrO appear in the center of the plasma plume.When increasing the oxygen fraction, an increasing amount of SrO is formed on the front edge of the plasma plume (Figure 2), where interaction with the background gas is strongest.A small decrease in SrO amount is visible near the center of the plume.A possible explanation for this behavior is that the reaction Sr + O 2 SrO + O is strongly dependent on temperature and relative concentration of the four products. 23As the O concentration inside the plume will increase due to reactions of Ti and TiO with O 2 , the reaction SrO + O → Sr + O 2 could become more common, leading to a decrease in SrO concentration in the plume center.
The spatiotemporal mapping measurements of Ti, TiO, and SrO show a significant change in plasma composition of the plume, especially in the front edge of the plume, when it reaches the substrate.To be able to link these plasma composition measurements directly to the thin film growth, we have grown STO films on STO under the same conditions (varying the Ar/O 2 mixture at constant pressure), with the substrate at z = 50 mm, for a subsequent analysis with X-ray diffraction (XRD).The LIF and XRD measurements are done at slightly different absolute pressures, 0.1 and 0.08 mbar, respectively.However, we verified that this small difference does not have a noticeable effect on our results, based on XRD and LIF comparisons of 100% O 2 at 0.08 mbar vs. 80% O 2 at 0.1 mbar.Figure 4 shows symmetrical XRD scans around the (002) Bragg reflection of SrTiO 3 for these thin films.The film grown in 0% oxygen (pure Ar) shows a clear film peak, as indicated by the arrow, aside from the strong substrate peak at 46.5 • .This indicates an increased lattice parameter with respect to the substrate. 24This typically arises from cation defects in the lattice, increasing the volume of the unit cell.For homoepitaxially grown films, the increased volume results in an expansion of the c-axis, since these films are in-plane strained with regard to the SrTiO 3 substrate.With increasing O 2 fraction, the film peak moves closer to the substrate peak, indicating an improvement in stoichiometry.
A remarkable transition can be seen with further increased oxygen pressure (75% and above), in that the film peak vanishes.This proves that a significant increase in film quality takes place at a certain oxygen pressure, here at about 75%, where a clear transition takes place from nonstoichiometric to stoichiometric film growth.
At this point, we note that a sharp transition to high-quality growth within a narrow range of certain parameter settings is a well-known heuristic observation in PLD.A possible mechanism contributing to the change in stoichiometry might be the effect of preferential scattering, as described by Wicklein et al. 8 This effect implies that lighter plasma species scatter stronger than heavier species.In the case of STO, where Ti is significantly lighter than Sr, the plasma plume would be Sr rich when it reaches the substrate.However, oxidation of the plasma species increases their mass.Whereas the weight ratio for Sr/Ti is 1.8, for the oxidized species (SrO/TiO 2 ) it is only 1.3, reducing the effects of preferential scattering.In this paper, we present independent spatiotemporal mapping data of the species in the plasma plume that are responsible for the growth.In fact, when comparing the spectroscopic data with the structural data of the grown material, we find an important correlation: the transition from non-stoichiometric to stoichiometric growth in the XRD measurements coincides with an oxygen pressure where neutral atomic titanium ceases to reach the substrate, while both Ti and Sr reach the substrate in their fully oxidized state, which are, respectively, TiO 2 and SrO.Furthermore, the decrease of non-stoichiometry aligns with the decrease of Ti and with the increase of oxidized species.This strongly suggests that full oxidation of plasma constituents is clearly a necessary requirement for stoichiometric growth of STO.
To summarize, we have mapped dynamic oxidation processes of plasma species in the PLD process and have shown the strong influence of the composition of the background gas on the plasma plume composition.Comparing these spectroscopic data with X-ray diffraction data of the grown material quality, we find a clear correlation between the chemical composition of the leading edge of the plasma plume and the stoichiometry of the grown film.Above a specific O 2 fraction of the background gas, approximately 75%, when full oxidation of Ti and Sr in the leading edge of the plasma plume is reached, the film growth transides from non-stoichiometric to stoichiometric growth, as determined with X-Ray diffraction.This suggests that the oxidation of species of the plasma is a crucial mechanism in the stoichiometric reconstruction of the synthesized oxide thin films.
To illustrate how our findings may be applied in scaling stoichiometric PLD growth, one may consider the following example.If the target-substrate distance is decreased from 50 mm to 40 mm, while maintaining all other parameter settings, this would prohibit stoichiometric film growth, even in pure O 2 , as Figure 2 shows that the oxidation of Ti would not be complete.Similarly, an increase in target-substrate distance would allow for stoichiometric growth in lower oxygen fractions than 75%, as the oxidation is allowed to progress further.Other examples and according spatiotemporal mapping can be devised for transversely upscaling PLD, for increasing the area of stoichiometric growth.This research is supported by the Dutch Technology Foundation STW, which is part of the Netherlands Organization for Scientific Research (NWO) and partly funded by the Ministry of Economic Affairs (Project No. 10760).
FIG. 1. Normalized density distributions of several components of a STO plasma plume cross-section in the propagation plane from the target (z = x = 0 mm) to the center of the substrate position (x = 0 mm, z = 50 mm).The lhs of (a) displays the Ti ground state density and the rhs displays TiO.In (b), respectively, SrO and TiO are displayed.The densities shown are measured at 20 µs delay after ablation and obtained with 0.1 mbar pure O 2 as background gas.The lack of signal close to the center of the target (x = 0, z = 0) is caused by the target holder obscuring one edge of the LIF excitation beam.

FIG. 4 .
FIG. 4. XRD scans around the (002) Bragg reflection of STO.It shows scans of samples grown at partial oxygen pressures and a total pressure of 8 • 10 −2 mbar.
FIG.3.Total amount of Ti, TiO, and SrO in the plasma plume, acquired by spatially integrating the LIF signal over the entire plume as function of the delay with respect to the moment of target ablation, for different partial pressures of O 2 .All measurements are at 0.1 mbar total background pressure.In the first 20 µs, a significant part of the plasma plume is obscured by the target holder and therefore not accessible for LIF measurements.