Phosphites as precursors in atomic layer deposition thin film synthesis

Phosphate materials have the ability to form redox inactive polymeric structures of polyanions with lithium as a charge compensating cation. Polyanions consist of phosphorus tetrahedrally coordinated by oxygen, sulfur, or nitrogen. They may form monomers or polymeric structures with varying lengths and readily form amorphous materials. When combined with aluminum in a near equal ratio, they can form complex porous structures resembling zeolites, also known as AlPO materials.¹ Similar structures may also be formed from polyanions such as silicates, sulfates, borates, and even transition element polyanions such as vanadates, molybdates, niobates, and many more.² The interactions between the polyanion and the charge compensating cation, Li in LiPON or Fe in FePO₄, will be affected by the covalent character of the polyanion itself. This can be controlled by the choice of central atom, here phosphorus, and its coordinates. When oxygen is substituted for nitrogen as a coordinate atom, the overall covalent character of the polyanion is increased. This effect is utilized in the formation of the LiPON structure. An alternative to substituting the coordinate atoms is altering the oxidation state of the central atom. A reduction of the oxidation state of phosphorous from +V to +III will result in an increased covalent character of the polyanion toward the cation, with the resulting shift in electrochemical potential due to inductive effects.³ Various phosphates have been thoroughly studied, but phosphites are rarely investigated. Phosphites also form polyanionic tetrahedral structures that tend to polymerize into asymmetric polyhedral superstructures. Three of the vertices in the phosphite bond to electronegative elements such as oxygen, while the remaining may bond directly to hydrogen, such as in phosphite acid, or occupy nonbonding electrons, such as for the phosphite precursors used in this study.

Phosphates are reported to be biocompatible, such as for apatitesm,⁴ potential proton conductors (Ca:LaPO₄),⁵ gate dielectrics (certain stoichiometries of aluminum phosphates),⁶ and as hosts for luminescent materials (titanium phosphates).⁷ We have used atomic layer deposition (ALD) to produce the films presented in this work.⁸ The ALD technique is advantageous in that uniform thin films can be deposited on flat as well as structured surfaces with atomic level thickness control. The films are grown using sequential self-limiting reactions between a reactant in the gas phase and active sites on the substrate surface. The process is relatively mild in that deposition temperatures are low enough to avoid thermal decomposition of the reactants and its products. This opens for the application of exploratory chemistry, such as the phosphites used here.

POCl₃, P₂O₅, and Me₃PO₄ have previously been used as phosphate precursors in ALD,^9,10 where P₂O₅ was the first reported precursor.¹⁰ In that work, the deposition was performed at temperatures above 400 °C where a high degree of decomposition of the precursor is likely to take place, hampering the self-limiting growth aspect of ALD. POCl₃ is a good phosphate precursor and does not require ozone as an oxygen source.¹¹ This is a major advantage when depositing on large structured substrates due to possible ozone depletion from catalytic decomposition with the substrate,¹² as well as the fact that ozone can oxidize other components in the film, thus hampering the deposition of reduced phases. The main drawback of POCl₃ is chlorine and its possible contaminations in the deposited film.¹³ For deposition of lithium containing compounds, the formation of stable LiCl is expected. Me₃PO₄ is the current primary choice of phosphate precursor in ALD, showing uniform growth and compatibility with several ALD systems.^{5,7,9,13–15} Furthermore, the phosphate content in the film can be tuned to above 60 at. %.¹⁵ Our approach is the deposition of phosphorous based films using precursors where phosphorous is in the +III oxidation state, namely, trimethyl phosphite (Me₃PO₃) and triethyl phosphite (Et₃PO₃). These precursors will replace trimethyl phosphate (Me₃PO₄) in the established deposition routes for aluminum phosphate (AlPO₄)¹³ and lithium phosphate (Li₃PO₄)^9,16 by ALD.

All deposition experiments were performed using an F-120 SAT hot wall ALD reactor (ASM Microchemistry Ltd.). The precursors used are listed in Table I. Li-O^tBu was kept at a source temperature of 130 °C, while all the other precursors were kept in external bubblers at room temperature. Me₃PO₃ has a flash point of 28 °C inferring extra safety precautions such as using external steel bubblers. The high vapor pressure of 23 mbar means that there is no need for an additional carrier gas or additional heating of the precursor. This is in contrast to the currently used Me₃PO₄, which has a notably lower vapor pressure (1.1 mbar) and needs the aforementioned measures. The pulse and purge times were kept constant for the precursors not under investigation. Table I also shows the experimental pulse and purge times. POCl₃ was used as the phosphate precursor for the XPS reference experiments and Me₃PO₄ was used for the FTIR reference experiments. The reference experiments were also deposited by ALD.

N₂ carrier gas was generated with an N₂ generator [Schmidlin-Sirroco 5 (99.999% Ar + N₂)]. The gas was further purified by a Mykrolis purifier before entering the reactor, keeping O₂ and H₂O levels below 1 ppm. Ozone was generated using an O₃-generator (InUSA-AC2025) supplied with O₂ (99.5% O₂, AGA) generating a flow rate around 500 cm³ min⁻¹ when pulsing. The reactor pressure was kept between 3.0 and 3.6 mbar during deposition with a total N₂ flow of 500 cm³ min⁻¹. Si(111) substrates were used. Thickness and refractive index (at 632.8 nm) was measured by spectroscopic ellipsometry using a wavelength range of 380–900 nm (Woollam alpha-SE, completeease software). An in-house designed crystal holder was used for quartz crystal microbalance (QCM) analysis, measured using a Maxtex TM400 unit and a PLO-10i oscillator. QCM was used to study the growth rates and saturation by relating the frequency response to mass change through the Sauerbrey equation calibrating the response with a known internal standard (TMA + H₂O).¹⁷ A pulse cycle is repeated 20 times and the average response and standard deviation of the 16 middle replicates are reported. XPS analysis was performed on a theta probe angle-resolved x-ray photoelectron spectrometer (Thermo Scientific). XPS analysis was performed on the surface layer, without sputtering. X-ray fluorescence (XRF) analysis was performed on a PANalytical Axios mAX mineral spectrometer and interpreted with the omnian and stratos analysis software. Time-of-flight elastic recoil detection analysis (TOF-ERDA) was done using 10.2 MeV ⁶³Cu⁵⁺ ions.¹⁸ 

The two phosphite precursors (Me₃PO₃ and Et₃PO₃) were investigated as possible candidates for ALD deposition by first attempting to deposit lithium phosphite (LiPO). The initial QCM analysis showed slight differences for the two precursors, particularly for the phosphite response, at 60 s in Fig. 1. Et₃PO₃ seems to reach saturation around the end of the pulse, while the Me₃PO₃ pulse has a high mass change throughout the pulse followed by a large mass loss during the following purge. This may be an indication of reaction with absorbed water into the bulk of the film from the water pulse, as similarly reported for LiOH deposited by ALD.¹⁹ In addition, the total mass change of the Me₃PO₃ subcycle is significantly larger than that of the Et₃PO₃ subcycle. We also discovered that an oxygen source was required to obtain films for both phosphite precursors. Deposition attempts using only alternating pulsing of the phosphite precursors and the cation precursor gave no film. This is in accordance with the Me₃PO₄ subcycle, which also gives no film without a cation subcycle.¹⁵ Water and ozone were tested as oxygen sources after the phosphite precursor pulse, both individually and in combination. We concluded that a water pulse after each precursor pulse resulted in a stable product. In addition, using only water was viewed as beneficial compared to ozone as it reduces the probability of oxidizing the phosphite groups to phosphate.

The Et₃PO₃ process is unstable and shows large thickness gradients on the deposited films in the temperature range 175–250 °C. Furthermore, Et₃PO₃ films with thicknesses of 50 nm or more were slightly opaque. This behavior is also observed in ALD films with high Li₂CO₃ and LiOH content,²⁰ which is the probable product if no phosphorous is incorporated into the film.²⁰ This assumption was confirmed by XPS, where the phosphorous content when using the Et₃PO₃ precursor never exceeded 2%. From here on, we will focus on the results using Me₃PO₃ as the phosphite precursor, as this was the only precursor that gave satisfactory products. From the QCM analysis, a 2/3 s pulse/purge on the Me₃PO₃ subcycle was found sufficient for saturative growth on Si substrates. The remaining pulse and purge parameters are given in Table I. The temperature window for the LiPO system was found by depositing in the temperature range 150–250 °C, as shown in Fig. 2. The error bars in Fig. 2 represent the standard deviation of ellipsometric analysis of seven points throughout the reaction chamber. At 225 °C, the product is consistent throughout the reaction chamber, as shown with the narrow error bars for both refractive index and growth rate at that temperature for Me₃PO₃. The refractive index has a plateau in the range of 150–225 °C, indicating a consistent product in that temperature range. The growth rate goes through a maximum at 175 °C and decreasing as the temperature is increased. The Me₃PO₃ process gives uniform films in the thickness range tested, and the growth rate stabilizes after around 600 ALD cycles, as indicated by Fig. 3. The variation in refractive index with film thickness for these films, with a drop in refractive index at 300 supercycles (Fig. 3) may be related to a growth mechanism where islets form and grow into structured shapes with apparent lower density before these grow together and reach a stable density.²¹ The surface roughness is further studied using AFM later in the article.

Aluminum phosphite (AlPO) was deposited using Me₃PO₃ as the phosphite precursor based on the performance of the precursors in the LiPO process. An initial QCM reactivity analysis was performed, and the phosphite response is very similar to what was previously seen for LiPO using the same precursor (Fig. 4). The pulse parameters for the phosphite precursor were set to the same value as for LiPO, due to the converging results. The TMA pulse and purge were set based on prior experience with this precursor in the same ALD reactor.

The stability window was determined by depositing film as a function of temperature in the range of 175–275 °C (Fig. 5). A slight decrease in the growth rate is observed with the increasing temperature. This is expected due to a decrease of the density of reactive sites with the increasing temperature.²² The refractive index remains almost constant and gives an indication of a stable process throughout the temperature range. Slight gradients were seen at the substrate edges at high temperatures, which could indicate that the Me₃PO₃ precursor starts to decompose. The growth as function of deposition cycles is given in Fig. 3 and shows a relatively constant growth rate with thickness at 0.1 nm/cycle.

ALPO elemental composition was analyzed using TOF-ERDA (Table II). The AlPO system was further used as a model to determine whether the composition of the product could be controlled by changing the subcycle composition, as the aluminum to phosphorous ratio of the AlPO films could be measured using a simpler elemental characterization procedure, XRF (Fig. 6). This was done by changing the relative number of TMA and Me₃PO₃ subcycles pulsed from 10% to 90% Me₃PO₃ and determining the composition by XRF (Fig. 7). In addition to determining at what level the composition of the product can be controlled, reaction stability analysis was performed using QCM and spectroscopic ellipsometry. Both methods show complimentary trends with respect to the growth rate with changing subcycle composition. The growth rate is the highest for pure TMA + H₂O cycles. When the subcycle ratio is around 1:1, there is a plateau of relatively constant growth rates before the growth rate again decreases for excess Me₃PO₃ subcycles. Uniform films were produced at all the compositions tested apart from when only Me₃PO₃ + H₂O was pulsed, where no film was formed, as expected. The phosphorous concentration in the films increases linearly with the phosphite subcycle content in the range of 10%–90% Me₃PO₃ pulsed, with a maxima of 25% phosphorous at 90% Me₃PO₃ pulsed [Fig. 7(a)]. The phosphorous content increases with increasing deposition temperature, until the concentration saturates at 25% phosphorous at 225 °C [Fig. 7(b)].

X-ray photoelectron spectroscopy (XPS) was used to quantify the composition and oxidation states of the components of selected AlPO and LiPO films. The results from the quantification are shown in Table II. The carbon content of the measured films was surprisingly high when measured by XPS, requiring further investigations. Selection of the films was analyzed with TOF-ERDA to get an accurate composition of the entire film. TOF-ERDA gives quantitative composition throughout the film, whereas XPS is only surface sensitive when sputtering is not used.²³ It should also be noted that sputtering in XPS can alter the lithium composition of the investigated film. TOF-ERDA analysis confirmed that the carbon detected by XPS was mainly located at the surface and we consider it to be contamination from handling the samples in air after deposition. XPS analysis shows a P2p 3/2 binding energy in LiPO at 133.0 eV. This is 0.6 eV lower than what reported for P(V) Li₃PO₄ by Kozen et al.²⁴ The P(V) LiPO₃ P2p peak has previously been reported at 134.5 eV by Chowdari et al.²⁵ P(V) AlPO₄ P2p has been reported at 134.4 eV.²⁶ This is the same as the measured binding energy for AlPO₄ at 134.3 eV, and 0.3 eV lower than the AlPO, at 133.0 eV. To our knowledge, there have been no reports of the P2p peak of P(III) phosphite using XPS. This downshift is an indication that our phosphorous might have retained the +3 oxidation state during deposition when using Me₃PO₃ as the phosphorous precursor and water as the reactant. More information can be obtained by comparing the P-peak shape of an AlPO film deposited using Me₃PO₃ precursor with the same peak on films deposited using POCl₃. In POCl₃, phosphorous is already in the +5 state. There are no significant differences in shape, only a shift of 0.5 eV (Fig. 8). This led us to believe that phosphorous in the films of this work retained the +3 oxidation state for LiPO, and AlPO uncertain. To confirm this, we performed FTIR analysis to search for the PO₄-signal at around 1200 cm⁻¹ (Fig. 9). There is a significant difference in the lithium systems. Li₃PO₄ has a sharp PO₄ absorption at 1132 cm⁻¹, and this absorption is not present for Li₃PO₃, thus confirming that we do not have phosphate in our films. The peak at 870 cm⁻¹ can be related to P-O stretch for P(III) phosphorous,²⁷ thus confirming that the oxidation state and the local structure of phosphorous are conserved during synthesis. The broad asymmetric peak at 1400–1600 cm⁻¹ is still unaccounted. It overlaps with the carbonate signal range, but due to the low carbon content in the material, this is an unlikely conclusion. The AlPO system is not as clear-cut: The peak at 1220 cm⁻¹ is shifted to 1200 cm⁻¹ when going from AlPO₄ to AlPO₃. There is an additional peak at 610 cm⁻¹ for AlPO₃, and the P-O peak at 890 cm⁻¹ is more pronounced and a lot broader. Broad peaks from amorphous Al-O stretching can appear in this region; the AlPO₃ peak around 850 cm⁻¹ can be assigned to isolated Al-O-Al sites, and downshift to around 870 cm⁻¹ in AlPO₄ for similar clusters.²⁸ This hints at less ordering when going from phosphite to phosphate.

The surface topography of AlPO and LiPO films deposited at 225 °C was investigated by AFM and shows a highly textured surface for the LiPO films while the AlPO films are less rough, as seen in Fig. 10(a). The measured LiPO film is 25 nm thick with an RMS roughness of 4.7 nm. In addition, narrow pillars that are 30 nm higher than the area immediately surrounding the peaks are visible. Thin, crystalline pillars or spikes are the likely shape of the surface when particles form through reservoirs of hygroscopic Li(OH) formed during growth. Such particles may react further when exposed with air to form Li₂CO₃. There are visible drift effects in the LIPO image; however, the image still demonstrates the large surface roughness. The surfaces of the AlPO films were found to be compliant to the assumption of uniform films, with an RMS roughness of 0.1 nm and some smaller peaks that are up to 4 nm higher than the surrounding area [Fig. 10(b)]. The topography for both films is clearly different to those deposited by Hämäläinen et al.^9,13 for the same systems, using fully oxidized phosphorous precursors. Those films show larger features that indicate microcrystallinity. The crystallinity of their film is proven by XRD, unlike the films presented in this work, where all are x-ray amorphous; this is surprising when considering the LIPO AFM measurements. Some care must be taken with this comparison since Hämäläinen et al. used SEM for topographical analysis.

Our investigations show that there are differences between phosphite and phosphate precursors. We observe increased reactivity for Me₃PO₃ as compared to using Me₃PO₄, where the latter requires a highly reactive coreactant for efficient deposition. FTIR analysis performed on the samples after deposition clearly indicates that phosphate is not present in Li₃PO₃ films. The phosphorous oxidation state is kept at +3 during the deposition. The AlPO analysis is not as clear, and the phosphate peak is visible in the FTIR spectrum but shifted by about 20 cm⁻¹. We see some peak shifts in both the FTIR and XPS data but not enough to give a clear indication of the phosphorous oxidation state for ALPO. If the product is indeed AlPO₄, something unknown must have been reduced in addition. Phosphates can form structures with varying degrees of polymerization, ranging from orthophosphates with Li₃PO₄ composition to chain structures with Li₃PO₃ composition. Extreme polymerization into P₂O₅ chains was not observed in this work.²⁹ By applying the possibilities of polymerization to the information in Table II, we can gain some insights into the film structure. The most interesting sample is the one characterized by TOF-ERDA, being less perturbed by surface carbon. The high oxygen content of the sample indicates an orthophosphate structure even when assuming that all carbon is in the form of carbonate. The AlPO system has even less phosphorous incorporated into the films than the LiPO system, too little to give AlPO₄ as the only product. For the AlPO film with the highest phosphorous content (AlPO 1:9), the average formula is $Al2.1PO5.5$⁠. By assuming AlPO₄ as the main product, the film must in addition contain 0.5 Al₂O₃. The average composition is close to the Al₂PO_5.5 previously reported by Meyers et al.,⁶ based on Al-O₆ octahedra as reported by Saraswati et al.³⁰ This explanation leaves 0.2 Al unexplained. The other AlPO films synthesized in this work have even less phosphorous incorporated. The films synthesized using a phosphite precursor is otherwise similar to films synthesized with phosphate precursors, albeit with slightly different stoichiometries. This is an indication that Me₃PO₃ is a good phosphorous precursor in ALD.

We have shown the successful deposition of LiPO and AlPO using trimethyl phosphite as the phosphorous precursor. The trivalent phosphorous remained as P(III) during the deposition for the LiPO system, and probably also for the AlPO system, based on similar shifts in the XPS data. Trimethyl phosphite has properties that are preferable in ALD synthesis such as high vapor pressure and good reactivity. This precursor can enable further exploration of phosphorous based ALD synthesis. Triethyl phosphite did not produce any film and was discarded as an ALD precursor in these synthesis routes.

The authors gratefully acknowledge the Research Council of Norway (Grant Agreement Nos. Nano-MILIB, 143732 and Solib, 622304) for financial support.

The data that support the findings of this study are available from the corresponding author upon reasonable request.