Recent developments in molecular precursors for atomic layer deposition

In recent years the industrial, consumer driven demand for smaller and more sophisticated microelectronic devices has directed research into nanoscale thin film deposition processes such as chemical vapour deposition (CVD) and atomic layer deposition (ALD). Common CVD methods employ volatile compounds, which are vaporised and continuously drawn into a deposition chamber. These precursor compounds chemically react either in the vapour phase or directly on a heated substrate and build up a film of target material, while unwanted volatile components are removed by reduced pressure or an inert gas purge. In the case of CVD, single source precursors (SSPs), i.e. molecules which already contain the necessary atoms in a proper ratio, corresponding to the desired complex phase required, come from a single precursor molecule. Alternatively, multiple source precursors can be used to produce the desired materials, in which two or more precursors act as vectors for individual atoms, the ratio of which in the final desired complex is tuned by varying the amount of precursor introduced to the reaction chamber, incorporating the desired multiple components into the growing film. Because of the fundamental reaction chemistries involved in CVD processes, impurities such as carbon may be introduced into growing thin films from organic ligand combustion products.

Another way to promote the desired chemistry is to employ co-reagents, i.e. separate reactive gases, which are simultaneously introduced into the reaction chamber. Oxygen, water and ozone are commonly used for oxides, and ammonia for nitrides. CVD rapidly covers the substrate in a thin film of the target material, tuneable to nanometres in thickness. However, as device requirements push toward smaller and more spatially demanding structures, methods such as CVD and various physical vapour deposition (PVD) techniques are being superseded by Atomic Layer Deposition (ALD). ALD offers advantages over alternative deposition methods due to its simplicity, reproducibility, control over film composition and thickness, and high conformality of deposited films at the atomic level. These unique characteristics originate from the cyclic, self-saturating nature of ALD processes.^1–3 

Unlike CVD, ALD is based on sequential surface reactions of molecular precursors separated by an inert gas purge. In this manner, the reactants are kept separated until the adsorbed species react at the surface in a self-limiting process, without the influence of gas phase reactions. Inherent to the ALD process is the ability to accurately control the thickness of the deposited films at an atomic level by simply varying the number of deposition cycles. Another important feature of ALD resides in the self-limiting nature of the surface reactions taking place during deposition, which precludes multilayer growth (i.e. allows control of the film thickness) and can simplify the study of the mechanism of surface grafting and/or film growth. A growth cycle for ALD is shown schematically in Fig. 1. For a more comprehensive summary of ALD and its many applications, the reader is referred to existing reviews on the topic. Several excellent reviews provide an overview of ALD in all of its facets, while this article aims to summarise the most recent advances in new precursor development. However, in order to understand the need for new precursors for ALD, it is important to briefly outline some very general principles that determine precursor properties and aid precursor selection.

The key to ALD is the design and synthesis of new precursors specifically for the aforementioned purposes. For many years precursors for ALD have been sourced from successful precursors originally developed for CVD, and specifically the more reactive precursors used for metalorganic-CVD (MOCVD). In order to achieve the unique characteristics of ALD, and to be suitable as a practical vapour deposition process, precursors must display specific properties.

Ideally, compounds must in the first instance be sufficiently volatile (minimum values of 0.1 Torr equilibrium vapour pressure have been suggested).⁴  Precursors should also vaporise rapidly and at a reproducible rate, conditions that are most typically met by liquid precursors. In order to achieve a self-terminating surface reaction, precursors must be thermally robust, with no intramolecular reactions or thermal decompositions on the substrate or in the gas phase. However, the precursor must be sufficiently chemically labile to react with the substrate surface and subsequently with the second precursor pulse. The implications of this are that at the point where the precursor has reacted with all the available reactive sites, there is a plateauing of film growth for the cycle, and the following step can begin. This self-limiting growth manifests itself in a typical saturation/growth per cycle plot, as can be seen from the plot in Fig. 2. Assuming that the precursor is thermally stable at the deposition temperature, no further film growth reactions can occur once the maximum surface saturated dose has been delivered. Whilst this holds for an ideal situation, in reality the saturation of 100% of the surface sites is unlikely, primarily because of kinetic and steric factors which limit surface coverage. Concordantly, theoretical growth rate per cycle (GPC) for classical ALD will always be suboptimal. As a result of the properties outlined previously, most ALD growth processes exhibit an “ALD window”; a temperature window where the growth rate is independent of substrate temperature (Fig. 2B). This can vary in size and is highly dependent on precursor.⁵ 

Precursors must also be highly reactive towards surface adsorbed species, and towards subsequent co-reagents. High reactivity generally results in relatively fast kinetics, lower ALD temperatures, and faster surface saturation, resulting in shorter cycle times.⁶ 

A further precursor requirement is that reaction by-products must be volatile and thus easily purged in order to prepare for the subsequent half-cycle. Moreover, by-products should not be, in and of themselves, corrosive or react with the growing thin film, which would result in non-uniform films, film etching and corrosion of the ALD reactor. Precursors which react with both the surface and with co-reagents exothermically tend to produce high purity films, as ligands are completely removed. This large thermodynamic driving force (large negative ΔG values) also typically allows lower deposition temperatures, which in turn, can result in the formation of high quality films.

For many novel precursor applications which require scale up, thermal ALD processes are often too slow, primarily because the precursors in use find their true application in MOCVD and were never originally designed with ALD chemistries in mind. In these instances, the need to develop new ALD-only precursors has to-date been somewhat circumnavigated by the use of alternative ALD chemistries and the application of more reactive co-reagents (e.g. H₂O₂ in-place of H₂O), or highly reactive gaseous species such as plasmas, radicals, ions and electrons which can be formed by using energy enhanced ALD processes such as plasma enhanced-ALD. By using these energy enhanced processes, less reactive precursors, foraged from MOCVD applications have been able to be applied to ALD.⁷ 

The success of ALD is built on chemistry, and unfortunately the time is rapidly approaching where the engineering of ALD processes using “established” precursors is no longer be able to keep pace with demand for both well-known industrially relevant materials and new curiosity-driven materials research. With ALD maturing into the deposition method of choice for many applications, its advancement is restricted by the availability of precursors which can keep pace with the demands of industry, such as low temperature ALD windows and the compatibility of ALD with other upstream or downstream device fabrication processes. For this, and several other reasons, new precursors distributed across the periodic table are needed.

It is important to note that whilst many references are made to ALD properties such as “growth per cycle (GPC)”, direct comparisons between precursors and processes – particularly for different materials – are not absolute, with many other factors exerting influence on this film thickness, including material densities, relative size of atoms, crystallinity and interatomic planes (i.e. wide-spaced laminar materials). Further to this, the authors do not attempt to dwell on the GPC of multinary systems, where the overall growth per cycle is not necessarily equivalent to the sum of a number of independent processes. It must also be put forward that whilst there is a good understanding of the “textbook” ALD process for simple systems, the lines between “atomic layer deposition” and “chemical vapour deposition” become less distinct with processes that move beyond monolayer growth, processes in which physisorption of precursor plays a more important role than surface-based chemical reactions, and processes that do not have the ability to react in the traditional binary fashion. A degree of contention exists over the definition of “true” ALD, and exactly which characteristics of ALD growth are most important in defining it as a process.

This review is intended to give the reader to a brief overview of “new” metal and metalloid precursors focussing on selected examples of some the most recent advances in new precursor development. This review is not intended to be exhaustive, and focusses on what the authors believe to be significant works and contributions to the field within recent years. The authors apologise in advance that not all work can be covered and afforded the same depth of coverage. As such, for chemistries we have not covered, particularly in the field of precursors as yet untested for ALD, we would direct interested readers to reviews in the area.^8–11 

It is perhaps unsurprising that the many advantages of ALD have been turned towards the continued development of thin film battery materials. For some time now, considerable attention has been drawn to the benefits of various ALD coatings for electrodes, which for the most part rely on standard metal oxide materials such as Al₂O₃ and TiO₂, in addition to the development of ALD routes to the electrode materials themselves. Further to this, more recent research has turned towards the fabrication of solid-state electrolytes (SSE) deposited in a similar fashion, and intrinsic to these have been the development of precursors for group 1 metals.

Initial studies into the suitability of a range of lithium compounds as ALD precursors were made by Putkonen et al. in 2009.¹²  The study focussed on five lithium systems with differing ligand functionalities: Li(THD) (THD=2,2,6,6-tetramethyl-3,5-heptanedionato), LiCp, LiⁿBu, LiNCy₂ and LiO^tBu, with the lithium β-diketonate, Li(THD), and the lithium alkoxide, LiO^tBu, proving most effective at the deposition of Li-containing films. Whilst the volatility of tetrameric Li(THD) is somewhat limited¹³  with sublimation under operating pressures of ca. 3 mbar occurring in the region of 175–200 °C, films of LiCO₃ were successfully deposited with sequential pulsing of precursor and ozone. Depositions were carried out between 185 °C and 300 °C onto soda lime glass, with a consistent growth rate of ca. 0.30 Å per cycle between 185 °C and 225 °C, after which a significantly diminished growth rate is observed. LiNCy₂ was found to decompose at temperatures required to ensure volatility, whilst LiCp/H₂O and LiⁿBu/H₂O deposited films were found to be irreproducible and non-self-limiting respectively. Subsequent work on the efficacy of LiO^tBu/H₂O as a lithium oxide precursor by Aaltonen et al. saw growth rates of 2.8 Å per cycle when used along with AlMe₃/H₂O to deposit films of Li₂O–Al₂O₃ at 225 °C.¹⁴  It is worth noting that the H₂O process towards lithium containing films greatly reduces the amount of carbon in the film and does not lead so readily to the formation of LiCO₃, as is observed with the Li(THD)-O₃ process. Furthermore, reaction of LiO^tBu with H₂S and H₂S/Al(NMe₂)₃ has been found to effectively deposit Li₂S and Li_xAl_yS thin films respectively, for use as cathodic and SSE materials in LiS batteries.^15,16 

Alongside the widespread use of Li(THD) and LiO^tBu, lithium hexamethyldisilazide (LiHMDS, see Fig. 3) has also been shown to be effective in the deposition of lithium containing films. The reaction of LiHMDS with H₂O has become a standard research process, which when coupled with CO₂ affords films of lithium carbonate, and when coupled with NH₃ affords LiN.¹⁷  Further to this, LiHMDS was shown by Hämäläinen et al. to act as a single source lithium and silicon precursor when deposited with ozone pulses.¹⁸  Lithium trimethylsilanolate (LiTMSO, Fig. 3) has been investigated by Ruud et al.¹⁹  for the deposition of a range of lithium containing thin films, including its use as a single-source precursor for both lithium and silicon within Li_xSi_yO_z, in a study that discounted the efficacy of a range of other potential lithium precursors including lithium benzoate, lithium trifluoroacetate and lithium acetate.

Interesting precursor pairings by Nisula et al.²⁰  and Kozen et al.²¹  have utilised LiHMDS and LiO^tBu respectively, coupled with diethyl phosphoramidate, (H₂N(O)P(OEt)₂), to deposit high quality lithium phosphorus oxynitride (LiPON) films for use as a solid state electrolyte material, whilst other phosphorus precursors such as triethyl phosphate have been shown to yield films of Li₃PO₄.²²  Other more complex structures such as LiFePO₄,²³  Li_xAl_ySi_zO²⁴  and Li_xAl_yO¹⁴  rely on the integration of LiHMDS/H₂O or LiO^tBu/H₂O into multicomponent processes. A recent interest in LiF deposition has led to a range of processes being developed utilising Li(THD)^25,26  and LiO^tBu²⁷  with TiF₄.

The ubiquity of lithium within battery applications has resulted in limited research into the atomic layer deposition of other alkali metal containing films. Despite this, a number of precursor compounds based on the efficacy of their lithium analogues have been explored, and the relatively higher abundance of sodium has resulted in a greater expansion of research into the incorporation of sodium into next generation batteries. Sodium containing thin films were first successfully deposited by Østreng et al.²⁸  in a study that investigated a range of simple sodium and potassium systems including Na/K(HMDS), Na/K(TMSO), Na/K(O^tBu). Of the six compounds investigated, both HMDS derivatives and the potassium trimethylsilanolate proved unsuitable for deposition. Sodium and potassium aluminate films were obtained from Na/K(O^tBu) using both H₂O and O₃ as oxygen sources with variable super cycles of AlMe₃:H₂O pulses. Less success was achieved in an attempt to realise sodium silicate films through the use of NaTMSO, with film conformity proving unachievable. Further research by Stønsteby et al. has utilised the alkali tert-butoxides with H₂O in multicycle processes to deposit sodium and potassium perovskite materials via ALD.²⁹  Examples of Group (I) metal precursors used in generating thin film products is shown in Table 1.

Quite remarkably the application range for atomic layer deposition (ALD) has now been extended to include the deposition of rubidium-containing films. Rubidium t-butoxide has been utilised as a precursor for the deposition of rubidium containing materials. The deposition of rubidium containing films is reported as proof of concept through the incorporation of rubidium into two very different systems. Doping amounts of Rb have been added to TiO₂ to yield a Rb doped oxide, Rb:TiO_x with up to 20 at% Rb. Studies revealed that higher amounts of Rb lead to the uncontrolled formation of rubidium carbonate, as determined by XPS. In the rubidium niobate system, the level of rubidium incorporation can be controlled up to the stoichiometric 1 : 1 Rb:Nb ratio allowing the formation of epitaxial films of perovskite RbNbO₃.³⁶ 

In a significant development porous lithium aryloxide films, [Li₂{p-C₆H₄O₂}], have been produced by Karppinen and co-workers, using LiHMDS as the lithium source, along with p-hydroxyquinone (HQ) in an atomic/molecular layer deposition process. Remarkably these Li₂Q metal–organic frameworks are crystalline, and undergo the reversible coordination of H₂O without the loss of structural integrity.³⁴  In related studies the use of LiTHD, along with NaTHD and KTHD have been used to generate a novel molecular layer deposition process for the production of a family of group 1 metal-terephthate frameworks, i.e. [Li₂{p-O₂CC₆H₄CO₂}], [Na₂{p-O₂CC₆H₄CO₂}] and [K₂{p-O₂CC₆H₄CO₂}] respectively.³⁵ 

Beryllium oxide (BeO) has garnered much interest as a highly thermally conductive high-κ dielectric material, with a dense structure that permits its use as a gas-diffusion barrier.^37,38  Due to the primary perceived application of BeO within the microelectronics industry, routes towards the atomic layer deposition of BeO have been readily exploited. As with many other standard ALD processes for which metal-alkyl systems display unparalleled reactivity and volatility, the standard beryllium precursor for this purpose is [BeMe₂], and has found use in a vast number of literature studies.^39–47  Growth per cycle values of 0.14 nm at 150 °C have been achieved with H₂O as an oxidant, whilst at higher temperatures, where the H₂O process demonstrates limited growth, the use of O₃ facilitates growth rates of up to 0.11 nm per cycle.³⁹ 

The atomic layer deposition of MgO was first achieved through the use of diethylmagnesium and H₂O.⁴⁸  This process was followed by the widespread use of the sandwich complex bis-cyclopentadienyl magnesium [Mg(Cp)₂], or its alkylated derivative, [Mg(η⁵-C₅H₄Et)₂], with H₂O,^49–53  along with magnesium β-diketonate, [Mg(THD)₂] and H₂O₂ or O₃.^54,55  Due to the coordinatively unsaturated magnesium environment within [Mg(THD)₂], a tendency towards oligomerisation or the formation of adducts is observed and a wide range of melting points and other reported properties are attributed to this.^55–57  Work has been done in an attempt to elucidate these properties and their effect on the efficacy of [Mg(THD)₂] as a precursor by Hatenpää et al.,⁵⁵  where alcoholic adducts of [Mg(THD)₂] were found to be monomeric but unsuitable for ALD. Films of MgF₂ have subsequently been deposited through the use of [Mg(THD)₂] and TiF₄, which demonstrated the use of a significantly less hazardous fluoride precursor than more conventional HF, which also exhibits a tendency to etch silica substrates.^58,59  The most significant recent advance in magnesium precursor chemistry is the development of monomeric and volatile bis(N,N′-di-sec-butylacetamidinato)magnesium, Fig. 4, [Mg{(ⁱPrN)₂CMe}₂], by Lou et al. which was used to deposit films of Mg_xCa_1−xO onto GaN substrates.⁶⁰  The volatilities and potential utility of these amidinate systems was highlighted by Sadiq et al. in 2001, though no deposition was attempted.⁶¹  An interesting computational investigation by Kazadojev et al. into the theoretical efficacies of a variety of chelating ligands for MgO and CaO deposition has also been reported.⁶² 

Unsurprisingly, the familiar THD complex of calcium has long been utilised in the deposition of calcium containing thin films, with films of CaS, SrS and BaS being deposited in the late 1980s by Tammenmaa et al. via reaction of the group 2 chelates with H₂S.^63,64  In addition to the low reactivity towards H₂O,⁶⁵  and in line with the other alkaline earths, coordinative unsaturation within these complexes is problematic,⁶⁶  with [Ca{THD}₂] often existing in trimeric form, prompting investigation into monomeric adducts such as the tetraethylenepentamine complex [Ca{THD}₂(tetraen)], Fig. 5, which exhibits higher volatility, but slightly lower growth rates than [Ca{THD}₂] when used to deposit CaS.⁶⁷  The latter is likely due to the increased steric hindrance of the spectator amine ligand. A natural advancement in the precursor chemistry of calcium ALD, in an attempt to increase reactivity towards H₂O was the utilisation of calcium cyclopentadienyl complexes, with bis(1,2,4-triisopropylcyclopentadienyl)calcium being used in conjunction with H₂O by Kukli et al. in 2006 to deposit CaO.⁶⁸  Precursor couplings of ozone and calcium hexafluoroacetlyaceonate, {hfac}, have subsequently been used to deposit films of CaF,⁶⁹  whilst more recently, Kim et al. demonstrated the use of novel dimeric bis(N,N′-diisopropylformamidinato)calcium and bis(N,N′-diisopropylacetamidinato)calcium, (Fig. 5) to deposit CaS and CaO.^60,70  One of the more interesting developments within calcium ALD is the use of bis(tris(pyrozolyl)borate)calcium, [Ca{Tpb}₂], by Saly and co-workers in 2010, to afford CaB₂O₄ in a reaction with H₂O.⁷¹  This is a little reported example of a single-source ALD precursor with demonstrable ability to confer two metals to a film in precise stoichiometries. A similar reaction was used to deposit analogous films of SrB₂O₄ and BaB₂O₄ with related precursor systems by the same authors.^72,73 

The use of strontium β-diketonate, [Sr{THD}₂] is well established⁶³  within atomic layer deposition, facing many of the same challenges as the magnesium, calcium and barium variants, being oligomeric in the solid state and exhibiting low volatility, low reactivity towards H₂O, and undesirable decomposition to incorporate carbon impurities.^74–76  In situ [Sr{THD}₂] synthesis has in the past sought to eliminate the short shelf-life of the precursor with limited success,⁷⁷  and more reliable precursor pairings subsequently developed have eliminated many of the disadvantages of [Sr{THD}₂]. These processes include the reactions of cyclopentadienyl derivatives [Sr(Cp)₂], [Sr(η⁵-C₅Me₅)₂] and [Sr(η⁵-C₅H₃ⁱPr₃)₂] with both H₂O and H₂S, and [Sr(η⁵-C₅H₄^tBu)₂] and [Sr(η⁵-C₅Me₄Pr)₂] with dimethoxyethane and H₂O, with all processes successfully depositing the desired strontium chalcogenide films. A variety of adducts of these cyclopentadienyl complexes have also been investigated.^75,78–81  Some of this work forms part of an interesting experimental-computational precursor study by Holme and Prinz⁷⁵  in which a variety of well-documented precursor ligand systems are assessed on the basis of a number of factors including metal-ligand bond strength, electron withdrawing substituents and intra-ligand bond strengths with a view to preventing ligand decomposition. Focus was primarily on cyclopentadienyl complexes in addition to a range of alkylated and fluorinated acetylacetonates, or “acac” compounds, the parent family of the widely used β-diketonate systems. The study indicates the presence of weaker metal-ligand bonding in cyclopentadienyl complexes, in part explaining their higher reactivity towards H₂O and oxygen, whilst the presence of weak intra-ligand bonding within acac species supports the propensity of these complexes to undergo thermal decomposition over an extended temperature range.⁷⁵ 

In a bid to improve upon the precursor properties of [Sr{THD}₂], a novel precursor was synthesised and trialled by Kim et al. in 2007⁸²  that featured the addition of a chelating pendant arm that served to disrupt oligomerisation and increase the volatility of the system. The new precursor, [Sr{MTHD}₂] (Fig. 6), possesses a melting point 60 °C lower than [Sr{THD}₂] and was thermally stable to 400 °C, advantageous over the disputed thermal stability of its more established parent compound.^76,82  In a similar vein, the heteroleptic derivative of [Sr{THD}₂], [Sr{THD}{demamp}] (demampH={1-[2-(dimethylamino)ethyl]methylamino}-2-methylpropan-2-ol), was described in 2015 by Lee et al.⁸³  The new complex displays an intermediate level of volatility and reactivity towards oxygen, sitting roughly between [Sr(η⁵-C₃H₂ⁱPr₃)₂] and [Sr{THD}₂].⁸³  More recently, dimers of strontium and barium imidazolates (Imid) have been shown by Norman et al.⁸⁴  to deposit the respective group 2 oxides, discussed in greater depth within the barium precursor overview.

Many of the drawbacks associated with the β-diketonate {THD} complexes discussed previously are accentuated within barium deposition, with oligomerisation of the larger alkaline earth metal more pronounced, existing in tetrameric form amongst others, with most non-adducted species susceptible to degradation over time.^66,85  However, ALD of BaS has been achieved with [Ba{THD}₂] synthesised in situ, as with its strontium counterpart, in addition to more conventionally synthesised precursor.^63,86  Significantly more focus has been directed towards the deposition of barium containing films utilising cyclopentadienyl derivatives as ligand systems, such as [Ba(η⁵-C₅H₂^tBu₃)₂] · THF_x, [Ba(η⁵-C₅Me₅)₂] · THF_x, [Ba(η⁵-C₅H₂ⁱPr₃)₂], [Ba(η⁵-C₅Me₄Pr)₂] and [Ba(η⁵-C₅Me₄C₂H₄NMe₂)₂] with O₂ plasma or H₂O and [Ba(η⁵-C₅Me₅)₂] with H₂S^{78,80,87–89}  (Fig. 7). Systems such as these offer a range of parameters and deposition windows to the prospective user.

More recent investigation into novel barium containing precursor systems has seen highlighted a number of chelating systems based around β-ketiminato backbones in an attempt to increase both volatility and reactivity whilst maintaining the structural integrity of ligands at elevated temperatures.^74,90  These complexes have not been trialled for atomic layer deposition to date, but exhibit good thermal stability and volatility, in addition to controllable coordinative saturation. A recent publication by Acharya et al.⁹¹  described the successful trial of a novel pyrrole-based barium precursor [Ba(PyMe₄)₂] with H₂O as an oxidant, facilitating low temperature deposition of BaO. No structural detail was described, and with such coordinatively unsaturated systems it is unclear as to the extent of oligomerisation or possible adducts present within the system. Other notable advances in precursor development attest to the application of aforementioned dimers of strontium and barium, di-strontium and di-barium tetra(2-tert-butyl-4,5-di-tert-amylimidazolate), (Imid), Fig. 7, in the deposition of metal oxide films with O₃ as an oxidant.⁸⁴  Examples of Group (II) metal precursors used in generating thin film products are shown in Table 2.

As with many other oxides of metals, group 3 and their rare earth (lanthanide) counterparts have long attracted attention as versatile materials with a range of potential applications. Because of the range and size of their band gap (4–5 eV) and high relative permittivity or dielectric constant (κ=12–30) together with high band offsets to silicon they have been considered and studied as potential dielectric materials in microelectronic devices. These properties along with others unique to this range of metals, including high thermal conductivities, electroluminescence and mechanical strength, have made ALD processes in which these materials can be deposited long sought after. ALD processes for the deposition of these rare earth metal oxides (RE₂O₃, RE=Sc, Y La, Ce, Pr, Nd, Pm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb and Lu) have conventionally been dominated by systems in which the bulky ligands such as β-diketonate, substituted cyclopentadienyl or bis(trimethylsilyl)amide ligands have been employed in an attempt to satisfy the steric demands of the RE³⁺ ions. Whilst the volatility and stability of these compounds is well known and many are commercially available, from a precursor perspective these materials are not ideal. Thin films grown from materials such as [RE{THD}₃]_x require high growth temperatures, as well as activation with strong oxidising agents such as O₃ and O₂-plasma due to the low reactivity of these systems with H₂O. In contrast the tris-cyclopentadienyl complexes, [RE{CpR}₃] R=Me, Et, ⁱPr, are considerably more reactive towards H₂O, resulting in higher growth rates per cycle, and often display monomeric structures (depending on both the size of the RE³⁺ ion and the substituent groups on the Cp-ligand), which results in greater volatilities. However, the thermal stability of RE-Cp systems is, in general, lower than the RE β-diketonate complexes, resulting in concomitant MOCVD alongside the ALD, which leads to non-uniform growth rates and an uncontrolled deposition process. While bis(trimethylsilyl)amide complexes, [RE{N(SiMe₃)₂}₃], are generally easily prepared and the bulky {SiMe₃} substituents aid volatility, the reactivity of the N–Si bonds often results in the formation of metal silicate thin films and are often avoided when pure RE-oxide materials are desired. While these precursors continue to dominate as “go-to” precursors because of their availability, reproducibility, and known chemistry, there has been some progress in the development of new precursors for rare earth metals. Examples of innovative ALD precursors used for Group III and lanthanide thin films are presented in Table 3.

Homoleptic alkoxide complexes of rare earth complexes are one group of precursors which have been used in the ALD of RE oxide films: The sterically bulky alkoxide complexes [RE{OCMe₂ⁱPr}₃] (RE=Y, Gd, La Pr), [RE{OCⁱPr₃}₃] (RE=Y, Gd, La) and [RE{OCEt₂^tBu}₃] (RE=Y, Gd, La) shown in Fig. 8 have been utilised as ALD precursors, with water as the co-reagent over a range of temperatures (250–400 °C) with varying success. However, not many details regarding the self-limiting nature of the process were reported.^102,103  Other examples of Group (III) metal precursors used in generating thin film products are summarised in Table 3. While these systems show some degree of volatility, high TG residues indicated a degree of thermal instability, a feature which was replicated in the ALD experiments displaying inconsistent growth rates. Homoleptic alkoxide complexes (Fig. 8) containing the bidentate ligand 1-methoxy-2-methyl-2-propanolato (mmp) have also been the subject of investigation; PrO_x and Gd₂O₃ films have been both been deposited by liquid injection ALD using a toluene solution of [RE(mmp)₃]_x/tetraglyme and water as precursors. The growth rate per cycle was found to increase with increasing precursor dose for both PrO_x and Gd₂O₃, indicating the lack of a saturating, self-limiting ALD-type growth.¹⁰⁴ 

The homoleptic N,N′-dialkyl-2-alkyl amidinates of the general formula [RE{(R′N)₂CR}₃] are the most recent class of ligands to be utilised in the ALD of rare earth oxide thin films. In 2003 Lim et al. first reported the deposition of La₂O₃ thin films in an water-based ALD process using the tris-amidinates [La{(ⁱPrN)₂CMe}₃] and [La{(ⁱPrN)₂C^tBu}₃] as precursors.^106,107  Growth rates of 0.9 Å per cycle at a deposition temperature of 300 °C were reported, however limited details on the growth behaviour, film composition, or properties were given. Whilst Winter and co-workers¹⁰⁸  have reported the synthesis, characterisation and molecular structures of a range of rare earth amidinate complexes, [RE{(^tBuN)₂CMe}₃] (RE=Y, La, Ce, Nd, Eu, Er, Lu), all monomeric in nature, they were significantly less volatile than the related ⁱPr derivatised complexes reported by Gordon.^109,110 

In addition, the compounds were reported to be significantly less reactive towards H₂O and required the use of stronger oxidizing agents such as O₃. Of the systems described only the Erbium complex was investigated for ALD, yielding nearly stoichiometric Er₂O₃ films, with low impurity levels reported. Growth rates for the Er₂O₃ films were reported to increase from 0.37 Å per cycle at 225 °C to 0.55 Å per cycle at 300 °C.

[Sc{(ⁱPrN)₂CMe}₃]/H₂O and [Y{(ⁱPrN)₂CMe}₃]/H₂O precursor combinations have also been employed for the ALD growth of Sc₂O₃ and Y₂O₃ films respectively.^109,110  For the Y₂O₃ process a self-limiting ALD-type growth was observed, with a growth of 0.8 Å per cycle at optimized deposition temperatures, with low impurity concentrations and relatively good dielectric properties (κ=11–12). However, the films were reported to be oxygen rich (O/Y ratio=1.7–2.0). In the case of Sc₂O₃ only very few details were given on the growth behaviour, film composition or properties however a growth rate of 0.7 Å per cycle was reported.

In an attempt to blend the properties of the rare earth tris-cyclopentadienyl complexes, [RE{Cp}₃], with the reactivity observed in the tris-amidinate complexes, heteroleptic species of the general form [(η⁵-C₅H₄R′)₂RE{(ⁱPrN)₂CMe}] (R′=Me or ⁱPr) have been synthesised and investigated (Fig. 8). In 2014 Park et al. described the liquid precursor [(η⁵-C₅H₄ⁱPr)₂Y{(ⁱPrN)₂CMe}]. The heteroleptic precursor, which exhibits a vapour pressure of 1 Torr at 168 °C, displays self-limiting growth in an ALD window from 350 to 450 °C and a growth rate of 0.6 Å per cycle.¹¹¹  Subsequent studies focusing on the development of mixed Hf/Y oxide materials showed a decreased growth rate of 0.4 Å per cycle at a deposition temperature of 180 °C.¹¹⁵  The related heteroleptic complexes [(η⁵-C₅H₄ⁱPr)₂RE{(ⁱPrN)₂CMe}] (RE=Pr, Gd Dy) as well as the yttrium derivative have also been investigated with both H₂O and O₃ as the oxygen source. The thermal stability of the precursors was reported to decrease in the order Y>Dy>Gd>Pr (i.e. with increasing ionic radius). In general, growth rates were found to be higher with H₂O as the co-reagent compared to O₃, except in the case of PrO_x deposition.¹¹³  In related studies Er₂O₃, Dy₂O₃ and Y₂O₃ have been produced from [(η⁵-C₅H₄Me)₂Er{(ⁱPrN)₂CMe}] and [(η⁵-C₅H₄ⁱPr)₂RE{(ⁱPrN)₂CMe}] (RE=Y or Dy), with H₂O (thermal ALD) and O₂ plasma (PE-ALD) counter oxidants. Both systems are reported to show ALD growth characteristics.¹¹⁴ 

Work on the development of rare earth guanidinate complexes has also been initiated with Devi and co-workers, developing both the yttrium and erbium complexes [RE{(ⁱPrN)₂CNMe₂}₃] (RE=Y or Er) (Fig. 8). In the case of [Y{(ⁱPrN)₂CNMe₂}₃] with H₂O as the co-reagent, ALD in the temperature range of 175 °C to 250 °C displayed saturative behaviour consistent with an ALD process and a growth rate of 1.1 Å per cycle.¹¹⁶  The corresponding erbium derivative has been exploited in a molecular layer deposition (MLD) process with 3,5-pyridine dicarboxylic acid to generate a luminescent metal organic framework (MOF), [Er₂{(O₂C)₂C₅H₃N}₃]_∞, with an exceptional growth rate of 6.4 Å per cycle at 245–280 °C.¹¹⁷ 

Group 4 metal oxide materials, either as simple binary oxides, i.e. TiO₂, ZrO₂ and HfO_2, or as mixed metal oxides have been the focus of considerable interest for some time. In the case of TiO₂, its application in photo-catalysis, as a transparent conducting oxide and as a material for dielectric and microelectronic applications have made it one of the most actively researched materials for deposition by ALD. As a result the topic of TiO₂ and the fabrication of thin films by ALD has recently been the subject of an extensive review by Karppinen and co-workers.¹¹⁸  Independently, Devi⁸  and Blanquart et al.¹¹⁹  also have produced excellent, and extensive, reviews of the precursor chemistry of Group IV elements. Similarly ZrO₂ and HfO₂ based materials are also of considerable interest, and are currently exploited in high volume manufacturing as high permittivity dielectric materials.

The desire for thin films of TiO₂, ZrO₂ and HfO₂ has driven the development of precursors for some time. However, with the established utility of a number of highly efficient ALD processes based around precursors such as group 4 metal halides, e.g. TiCl₄, ZrI₄ and HfCl₄, metal alkoxides e.g. Ti(OMe)₄, Zr(OⁱPr)₄ and Hf(OCMeEt₂)₄, metal amides e.g. Ti(NMe₂)₄, Zr(NMeEt)₄ and Hf(NMeEt)₄, and metal cyclopentadienyl complexes e.g. [CpTi(NMe₂)₃], [(MeCp)₂ZrMe₂], [Cp₂ZrCl₂] and [(MeCp)₂Hf(OMe)₂], there has been relatively little development in new precursors for the ALD of TiO₂, ZrO₂ and HfO₂.

While Group IV metal β-diketonates have played an extensive role in the development of MOCVD applications there is only one homoleptic β-diketonate complex, namely Zr(THD)₄, with ozone, that has been reported as a prospective precursor for the ALD of any Group IV metal oxide. However it is important to note the heteroleptic species [Ti(OⁱPr)₂{THD}₂] and [Ti(O^tAmyl)₂{THD}₂] (Fig. 9) have both been shown to achieve self-limiting growth above 360 °C with O₃ as the oxygen source.¹²⁰  Otherwise, this general class of precursors is associated with the low GPC characteristic of β-diketonate complexes (Table 4).

Whilst a range of homo- and heteroleptic metal amides and metal alkoxide complexes of Group IV are well established, they have lent themselves as suitable departure points for the limited number of new Ti, Zr and Hf ALD precursors that have been developed recently. The mixed alkoxide-alkylamide compounds, [Ti(OⁱPr)₂(NMe₂)₂]¹²¹  and [Hf(O^tBu)(NEtMe)₃]¹²²  have both been used successfully for the ALD of TiO₂ and HfO₂, respectively. Whilst homoleptic alkoxide and alkylamide precursors usually have thermal stabilities well below 300 °C, these heteroleptic compounds exhibited self-limiting growth above 300 °C, with both processes displaying high growth rates and affording oxide films of high purity. TiO₂ grown from [Ti(OⁱPr)₂(NMe₂)₂] and either ozone or water, did not exhibit a constant GPC, however self-limiting growth was confirmed at 325 °C with a remarkably high GPC, for TiO₂, of 0.8 Å. HfO₂ was similarly grown using a [Hf(O^tBu)(NEtMe)₃]/O₃ ALD process at 300 °C, and a GPC of 1.6 Å.¹²² 

In a similar vein, donor-functionalised alkoxide precursors have also been developed. Although several alkoxide−donor-functionalized alkoxide precursors have been reported in the literature, only the studies on the complexes [Hf(O^tBu)₂(mmp)₂]¹²³  (Hmmp=HOCMe₂CH₂OMe, 1-methoxy-2-methylpropan-2-ol) or [Ti(OⁱPr)₂(dmae)₂]¹²⁴  (dmae=dimethylaminoethoxide) using water for the generation of HfO₂ and TiO₂ thin films reported self-limiting growth. Disappointingly high carbon and hydrogen contamination was found in the films grown from [Hf(O^tBu)₂(mmp)₂]. The self-limited growth of TiO₂ from [Ti(OⁱPr)₂(dmae)₂] was demonstrated at the low temperature of 100 °C, but no compositional data relating to the thin films were reported. TiO₂ thin films have also been produced using the novel precursors [Ti(NMe₂)₃{dmap}] (Hdmap=1-dimethylamino-2-propanol). Plasma enhanced ALD using [Ti(NMe₂)₃{dmap}]/O₂ plasma has allowed the deposition of high quality TiO₂ in an ALD window between 60 and 120 °C at a high growth rate of 0.9 Å per cycle.¹²⁵ 

Amidinate ligands have been shown to significantly improve the thermal stability of M–N containing compounds whilst maintaining the reactivity of related metal amide systems, resulting in amidinate complexes having good reactivity with ozone and water. Their high reactivity results in the efficient removal of the ligands, and low impurity levels in the deposited films. Whilst a number of amidinate compounds have been reported, very few have reported the application of these systems in ALD processes. To date only tetrakis(N,N′-dimethylacetamidinate) zirconium, [Zr{(MeN)₂CMe}₄]¹²⁶  and the alkoxide-amidinate precursor [Ti(OⁱPr)₃{(ⁱPrN)₂CMe}]¹²¹  have been reported for the ALD of Group 4 metal containing materials, i.e. ZrO₂ and TiO₂ respectively. In the case of [Zr{(MeN)₂CMe}₄]/H₂O self-limiting growth of ZrO₂ could be achieved at 300 °C, but with a low GPC (0.24 Å). Unfortunately, the study did not involve any compositional analysis.¹²⁶  In contrast [Ti(OⁱPr)₃{(ⁱPrN)₂CMe}], which is a liquid at room temperature with a low evaporation temperature (65 °C under 5–10 mbar), presented good reactivity toward both ozone and water. Using water as the oxygen source, the deposition exhibited a region of constant GPC between 300 and 350 °C. Self-limiting growth was confirmed at 325 °C with saturation of the GPC at 0.5 Å. Deposition within the temperature range 275–375 °C, resulted in what was described as “very pure” thin films with carbon and nitrogen contamination below 1 at.%. Notwithstanding the deposition temperature, the titanium dioxide films deposited adopted the anatase phase. Whilst the ozone process presented nearly identical data with respect to purity, crystallinity and growth behaviour, the ALD window was extended from 275 to 350 °C with ozone as the oxygen source. However, saturated GPC was slightly lower (0.4 Å).¹²¹  In situ mechanistic studies have also been performed on [Ti(OⁱPr)₃{(ⁱPrN)₂CMe}]/D₂O using quartz crystal microbalance and quadrupole mass spectroscopy at 275 °C.¹²⁷ 

Whilst amidinate complexes of Group IV metals have, in general, been little explored, the related guanidinate systems have seen more attention. The mono guanidinate complexes [Ti(NMe₂)₃{(ⁱPrN)CNMe₂}],¹²⁸  [Ti(NMeEt)₃{(ⁱPrN)CNMeEt}]¹²⁹  and [Zr(NMeEt)₃{(ⁱPrN)CNMeEt}]¹²⁹  have been investigated as ALD precursors. The investigation into the ALD of TiO₂ from [Ti(NMe₂)₃{(ⁱPrN)CNMe₂}] highlighted a precursor with a thermal stability and a process with a high GPC, and self-limiting growth confirmed at 330 °C.¹²⁸  The related monoguanidinate compounds [Ti(NMeEt)₃{(ⁱPrN)CNMeEt}] and [Zr(NMeEt)₃{(ⁱPrN)CNMeEt}] were also evaluated and exhibited similar properties. Evaluation of these complexes as ALD precursors was performed with both H₂O and ozone as the oxygen source. Using water, both precursors exhibited self-limiting behaviour at 275 °C. Both water and ozone processes presented an ALD window of between 225–325 °C, with a GPC of 0.4 Å for the growth of TiO₂ across this range, independent of the oxygen source, compared to a growth of 0.8 Å for ZrO₂ using water. For the ALD process with ozone as the oxygen source, the GPC increased steadily with the deposition temperature (1.0 Å per cycle).¹²⁹ 

In related studies both the zirconium and hafnium bis-guanidinate complexes [Zr(NMeEt)₂{(ⁱPrN)CNMeEt}₂]¹³⁰  and [Hf(NMe₂)₂{(ⁱPrN)CNMe₂}₂]¹³¹  precursors have been evaluated for the ALD of ZrO₂ and HfO₂ thin films. The zirconium precursor, when water was used as the oxygen source, displayed a steady GPC (0.85 Å) over a temperature range of 225 to 400 °C, whereas with ozone the GPC increased steadily with the deposition temperature from 0.8 Å at 250 °C to 1.3 Å at 400 °C. The characteristic self-limiting ALD growth was confirmed by the study of the GPC as a function of the precursor pulse length at 300 °C, with high GPCs of 0.90 and 1.15 Å per cycle with water and ozone, respectively.¹³⁰  For the hafnium complex self-limiting growth was achieved at 200 °C with a high GPC of 1.2 Å.¹³¹ 

As noted above, Group IV metallocene complexes have long been established as ALD precursors for the deposition of thin films by ALD.¹¹⁹  Recently this family has been extended beyond cyclopentadienyl complexes to introduce cycloheptatrienyl complexes, specifically [(Cp)Ti(CHT)] and [(MeCp)Zr(CHT)] (CHT=C₇H₇), for the ALD of TiO₂ and ZrO₂. These precursors do not react with water and require ozone as the oxygen source. The zirconium compound appeared to perform well with self-limiting growth occurring up to a temperature of 350 °C, with a GPC rate of 0.8 Å. Contrastingly the titanium compound was significantly less thermally stable (T_max=300 °C) and had only a modest GPC (0.35 Å). Interestingly, results show that the films deposited from [(Cp)Ti(CHT)] had a strong tendency to form the high permittivity rutile phase upon annealing at temperatures below 600 °C.¹³² 

It should be noted that although there are a plethora of reports which feature established materials, such as TiN, as well as new materials, e.g. TiS₂, which feature ALD as the primary fabrication process, these materials in and of themselves have not resulted in the design, synthesis and development of new precursors, relying on established metal precursors, such as TiCl₄, with co-reagents other than H₂O and O₃.

As with group 4 precursors, much focus in the development of group 5 precursors has focused on their development for the generation of oxide materials, and as with oxides of the Group 4 elements oxides of vanadium, niobium and tantalum, have been of primary interest because of their application in electronic devices. Nb₂O₅ and Ta₂O₅ are both high-κ materials, with large band gaps that have been investigated for numerous applications such as antireflective coatings, dielectric layers and diffusion barriers.¹¹⁹  In stark contrast to Nb₂O₅ and Ta₂O₅, the properties of vanadium oxide thin films depend significantly on their stoichiometry, structure, and morphology.¹³³  Vanadium exists in many stable oxidation states, and as such results in a number of oxide materials, i.e. VO, V₂O₃, VO₂ and V₂O₂, and in several polymorphic forms, therefore depositing thin films of vanadium oxide with a specific oxidation state and structure has been a challenge, with particular attention focused on +4 and +5 oxidation states of vanadium oxide. Vanadium dioxide (VO₂) displays a semiconductor-to-metal phase transition that occurs close to 67 °C which is accompanied by an abrupt change in its resistivity and near-infrared transmission. This property makes VO₂ of interest for a number of applications including smart windows, resistive memories, and switches in microelectronics.¹³⁴  Vanadium pentoxide (V₂O₅) meanwhile, has potential applications as an electrode in lithium ion batteries.¹³⁵ 

As with Group 4, ALD processes which use the Group 5 precursors are dominated by a handful of highly established ALD processes based around metal halides i.e. VCl₄, NbCl₅, NbF₅, TaCl₅, TaF₅ and TaI₅, metal alkoxides and oxo-alkoxides e.g. [OV(OⁱPr)₃], [V(O^tBu)₄], [Nb(OEt)₅] and [Ta(OEt)₅], and metal amide and metal amide-imido complexes e.g. [V(NEtMe)₄], [Ta(NMe₂)₅], [Ta(NEt₂)₅], [Nb(N^tBu)(NEtMe)₃] and [Ta(N^tBu)(NEt₂)₃], all of which can, and do, act as suitable departure points for precursor modification and development.¹¹⁹ 

A summary of Group V complexes used as precursors for thin films is shown in Table 5. Amongst the precursors reported for the ALD growth of vanadium based materials, only the compounds [OV(OⁱPr)₃], [V(NEtMe)₄] and [V{(ⁱPrN)₂CMe}₃] (Fig. 10) have been reported for vanadium oxide deposition. [OV(OⁱPr)₃] is the common precursor for the ALD of vanadium oxide, and the deposition process is well characterized, however OV(OⁱPr)₃ has limited thermal stability (below 170 °C) and only a low GPC.^136,137 

ALD of the amide system [V(NEtMe)₄], with water as the oxygen source displays a constant GPC of 0.5 Å in the temperature range of 125 to 200 °C. Self-limiting film growth was confirmed at 150 °C with both ozone and water. The water process had the lower GPC of 0.45 Å. While the films were reported to be smooth and possess low degrees of contamination (C impurities) at 200 °C, it was found that the vanadium oxide thin films contained a mixture of +4 and +5 oxidation states of vanadium.¹³⁸  Greater control over oxidation state of the final materials was found to be achievable when ozone was used a co-reagent in the ALD process alongside [V(NEtMe)₄]. Self-limited growth was established at 200 °C, with a GPC of 0.3 Å. Analysis of the thin films by XPS revealed the films to be VO₂.¹³⁹ 

Hock and co-workers have recently reported the development of the vanadium 3+ precursor [V{(ⁱPrN)₂CMe}₃]. As part of an investigation into the application of this precursor with various oxidising agents, the reaction of [V{(ⁱPrN)₂CMe}₃] with H₂O, H₂O₂ and O₃ was monitored in situ using quartz crystal microbalance (QCM) and quadrupole mass spectroscopy (QMS). With ozone, self-limiting growth of V₂O₅ was reported from 150 to 225 °C with QCM studies showing an average mass gain of 45.3±3.4 ng cm⁻² per cycle. In contrast, reaction with O₂ showed no mass gain during in situ monitoring, however when O₂ was dosed after H₂O ALD growth at 200 °C was observed (average mass gain of 17.7±0.9 ng cm⁻² per cycle). Whilst the as-deposited films were amorphous by Raman and XRD spectroscopy, after annealing in a N₂ atmosphere at 450 °C films were identified as V₂O₅. ALD (200 °C) with H₂O₂ followed by a post deposition anneal in H₂ showed the presence of V₂O₃ by Raman spectroscopy. In an attempt to encourage selective V⁴⁺ deposition, H₂ (4%) was added the ALD sequence after a H₂O₂ pulse. Self-limiting growth of VO₂ was reported at 200 °C with QCM studies showing an average mass gain of 17.5±1.3 ng cm⁻² per cycle.¹⁴⁰  In related studies the same group have reported the ALD of VS₄, (Parónite), in which [V{(ⁱPrN)₂CMe}₃] reacts with H₂S to oxidise V³⁺ to V⁴⁺ while reducing S²⁻ (from H₂S) to {S₂²⁻}. Self-limiting growth of VS₄ was established by QCM between 150 ° C and 200 °C, with a GPC of 0.33 Å.¹⁴¹  [V{(ⁱPrN)₂CMe}₃] has also been used as a co-reagent in the ALD of thin films of V_xIn_(2−x)S₃.¹⁴² 

As noted earlier, the imido-amide complexes [{^tBuN}Nb(NEt₂)₃]¹⁴³  [{^tBuN}Nb(NEt_Me)₃],¹⁴³  [{^tBuN}Ta(NEt₂)₃],^144,145  and [{^tBuN}Ta(NEtMe)₃]¹⁴⁵  have all been reported as precursors for the growth of Nb₂O₅ and Ta₂O₅ respectively, with water as the oxidant at 275 °C (Nb) and 325 or 350 °C (Ta), with GPC values of 0.4 Å 0.5 Å, 0.65 Å and 0.6 Å reported. With the exception of [{^tBuN}Ta(NEtMe)₃] these precursors have also been utilised alongside ozone as the oxygen source at identical temperatures.^143,145 

As a derivative of the imido-amide complexes, the highly thermally stable pyrazolate complex [(^tBuN)Ta{3,5-di-tert-butylpyrazolate}₃], has been utilised as an ALD precursor for the formation of Ta₂O₃. With a relatively low vapour pressure, a source temperature of 160 °C under a reduced reactor pressure of 2–3 mbar was necessary to sublime the precursor. Self-limiting growth was confirmed at 325 °C when ozone was used as the oxygen source. Analysis of the thin films indicated low amounts of impurities in the films, however the GPC was low (0.3 Å) compared to more volatile complexes in this class.¹⁴⁶ 

With the proliferation of amidinate ligands across the periodic table in a wide array of ALD precursors it is perhaps not surprising that Group 5 metals have also received attention. The heteroleptic amide-imido-amidinate complex compound, [(^tBuN)Ta{(ⁱPrN)₂CMe}₂(NMe₂)]¹⁴⁷  has been investigated for the ALD of Ta₂O₅, using water as the oxygen source. Self-limiting growth was confirmed at 325 °C with a rather low GPC of 0.28 Å.

Chromium, molybdenum and tungsten precursors for application in ALD are extensively focused around the development of precursors for the deposition of metal oxide thin films, such as CuCrO₂, Cr₂O₃, MoO₃ and WO₃, with the exception of tungsten, where the ALD of metallic tungsten thin films (W) as well as tungsten nitride (W_xN) dominates. Precursors for these materials have been reviewed elsewhere,¹⁴⁸  however, as with other elemental groups, established precursors for ALD are selected from a handful of reactive MOCVD precursors, specifically species such as CrO₂Cl₂, Cr(acac)₃, MoCl₅ and WF₆. For the metal nitrides, Mo_xN and W_xN, the bis-imido-amide complexes [(^tBuN)₂M(NMe₂)₂] (M= Mo or W) have been comprehensively explored. Similarly for Group VII elements where only manganese has been investigated for the production of MoO_x, MnS, MnTe and MnAs thin films by ALD, established precursors include MnCl₂, [Mn{THD}₃], [(η⁵-C₅H₄-Me)₂Mn] and [(η⁵-C₅H₄-Et)₂Mn].¹⁴⁸ 

Despite the relative inactivity with the arena of precursor development over the past two decades several new precursors have been reported recently. Kalutarage et al. reported the development of a family of transition metal (Cr, Mn, Fe, Co, Ni, Cu) complexes bearing α-imino alkoxide ligands (Fig. 11). These complexes can reportedly be sublimed, between 90 and 160 °C at 0.05 Torr, on ∼0.5 g scales in ∼3 h with <5% nonvolatile residues. In the case of the manganese derivative [Mn₂{^tBuNCHC(^tBu)(Me)O}₄] the sublimation temperature is higher than that of the other complexes in this range, because of its dimeric structure. ALD at 225 and 180 °C for the Mn and Cr complexes, [M{^tBuNCHC(^tBu)(Me)O}₂] (Fig. 11), with the BH₃(NHMe₂), showed self-limiting growth of metallic Cr, and in the case of the manganese precursor, MnO_x films respectively, as identified by XPS analysis: oxide generation is thought to be the result of ex situ oxidation. For the Cr precursor the GPC was reported to be 0.08 Å and for the Mn precursor 0.095 Å.¹⁴⁹  Subsequent investigation of the manganese precursor [Mn₂{^tBuNCHC(^tBu)(Me)O}₄] demonstrated its utility as a manganese vector in the deposition of Cu/Mn alloy thin films alongside the copper precursor [Cu{OCHMeCH₂NMe₂}₂] and the reducing agent BH₃(NHMe₂). Deposition rates of about 0.09 Å were reported on a variety of substrates, with Cu : Mn ratios of about 70 : 30 obtained by controlling the number of Cu and Mn cycles. X-ray photoelectron spectroscopy supported the presence of metallic Cu and Mn within the Cu/Mn alloys.¹⁵⁰ 

Chabal and co-workers reported a novel molybdenum precursor, [(η⁵-C₅H₄-Si(CH₃)₃)Mo(CO)₂{η³-2-CH₂CMeCH₂}], and its use with ozone as the co-reactant in an effort to grow MoO₃ films at low temperatures by ALD. The ALD process at 250 and 300 °C is characterized by a nucleation delay attributed to the reaction barrier for the initial reaction of the metal precursor and the formation of the surface products. The results suggest an initial delay in the ALD process, which can be negated if higher substrate temperatures (350 °C) during the initial Mo precursor pulses are used, after which a lower-temperature (250–300 °C) ALD process is possible with no incubation period, yielding what are reported as good quality MoO₃ films.¹⁵¹ 

Atomic layer deposition of crystalline molybdenum oxide thin films α-MoO₃, β-MoO_3, and an unidentified sub-oxide MoO_x phase (2.75≤x≤2.89) has been demonstrated using a new, commercially available precursor [MoO₂(THD)₂] with O₃ at 250 °C. The ALD process (approx. 1 Å per cycle) reportedly displays a well-controlled film growth with good uniformity and conformality with low levels of impurities. Studies showed that the relative amounts of the different phases could be controlled by changing the deposition conditions. Post-deposition annealing treatments studied by high-temperature X-ray diffraction enabled further control of film composition. In particular, single-phase α-MoO₃ and MoO₂ films were achieved by annealing in O₂ and ‘forming gas’, respectively.¹⁵²  Marks and co-workers have also reported the synthesis and design of new oxo-complexes for the fabrication MoO₃ and WO₃. Remarkably, these two isostructural precursors [MoO₂{(^tBuN)₂CMe}₂] and [WO₂{(^tBuN)₂CMe}₂] (Fig. 11) display very different reaction profiles. Using the molybdenum precursor and O₃ amorphous, ultrathin molybdenum oxynitride (approximate composition MoO_2.48N_0.18, as determined by XPS) films are grown at 200 °C. In contrast using [WO₂{(^tBuN)₂CMe}₂] and H₂O (200 °C) yields amorphous WO₃ films. For both processes GPC data is not reported. Despite attempts to produce effective ALD growth with [MoO₂{(^tBuN)₂CMe}₂]/H₂O and [WO₂{(^tBuN)₂CMe}₂]/O₃ combinations all attempts failed underscoring a very different reaction chemistry for the two isostructural complexes, the origin of which is not clearly understood.¹⁵³  Related studies by the same group have gone on to develop the cyclohexyl and isopropyl derivatives, [MoO₂{(CyN)₂CMe}₂] and [MoO₂{(ⁱPrN)₂CMe}₂]. Quartz-crystal microbalance and X-ray photoelectron spectroscopic studies confirm that the isopropyl derivative is an improved ALD precursor versus the R=t-butyl derivative for the deposition of nitrogen free MoO₃. A linear growth rate was observed at 150 °C using [MoO₂{(ⁱPrN)₂CMe}₂] and O₃, with a reported growth rate of ∼0.065 Å per cycle, as determined by ex situ by spectroscopic ellipsometry measurements, which was slightly smaller than the growth rate determined from QCM (∼0.085 Å per cycle).¹⁵⁴  A summary of ALD precursors used to generate thin films for Group VI and VII metals is shown in Table 6.

Thin films of the first row transition metal elements, and their derivatives (i.e. oxides and sulphides) have many important and current applications; metals such Fe Co and Ni have potential application in spintronic devices¹⁵⁵  whereas Cu is the conductor of choice for interconnects in electronic devices. In addition to the interest in metallic thin films, oxides and sulphide materials such as Fe₂O₃, NiO, Co₃O₄, Cu₂O, FeS, NiS, Co₉S₈ and CuS, which find application in a range of semiconducting energy harvesting conversion and storage applications. In 2013 Kinsley et al. published a review of precursors, and their chemistry, for the ALD of metallic first row transition metals.¹⁰  Precursors and processes for metal oxide¹⁴⁸  and metal sulphide¹⁵⁶  materials deposited by ALD have also been reviewed. For this reason, this section will cover examples of new precursors for Fe, Co, Ni and Cu which have, in general, been published subsequently to these reviews.

As noted in the previous section Kalutarage et al. in 2013 reported the development of a family of α-imino alkoxide complexes of the metals Cr, Mn, Fe, Co Ni and Cu of the general form [M{^tBuNCHC(^tBu)(Me)O}₂] (Fig. 12).¹⁴⁹  As with the Mn complex discussed earlier the Fe complex is dimeric and displays low volatilities, whereas the Co, Ni and Co analogues are monomeric and display appreciable volatilities (sublimation is achieved between 70–80 °C at 0.05 Torr). Self-limiting growth rates of 0.07 (Fe), 0.07 (Co) and 0.09 (Ni) Å per cycle were reported for an ALD process at 180 °C using BH₃(NHMe₂) as the co-reagent/reductant. XPS analysis was used to confirm the deposition of metallic Fe, Co and Ni metal thin films. In the case of the Cu derivative, ALD was not explored as the precursors showed volatilities below that of comparable Cu precursors, specifically [Cu{dmap}₂] and presented no obvious advantages as a prospective ALD precursor.

In 2001 Winter and co-workers reported the design and synthesis of a family of first row transition metal complexes of chromium(ii), manganese(ii), iron(ii), cobalt(ii), and nickel(ii), containing 1,4-di-tert-butyl-diaza-1,3-butadienyl (^tBu2DAD) ligands (Fig. 12).¹⁵⁷  Since then, the cobalt and nickel complexes have been utilised in a number of ALD processes. In a 3-step ALD process for the deposition of cobalt metal Winter et al. describe the utility of the cobalt complex [Co{^tBu2DAD}] with formic acid and 1,4-bis(trimethylsilyl)-1,4-dihydropyrazine.¹⁵⁸  Reaction of the Co(ii) formate intermediate with 1,4-bis(trimethylsilyl)-1,4-dihydropyrazine at 180 °C resulted in the formation of pure (>97.1% by XPS) Co films. A saturative growth rate of 0.95 Å per cycle, which is significantly higher than comparable reported processes for Co metal deposition, is reported. The as-deposited Co films are reported to be of high purity, in addition to possessing a sheet resistivity close to that reported for bulk Co metal.^159,160  In an extension to this work the Co and Ni complexes, [Co{^tBu2DAD}₂] and [Ni{^tBu2DAD}₂], have both been used in a 2 component ALD process with either ^tBuNH₂ or Et₂NH as the co-reactant to form Co and Ni metal films respectively.^161,162  For the Co system, growth on a range of substrates proved to be relatively uniform across the temperature range 160 to 220 °C, with a self-limiting ALD window between 170–200 °C and GPCs of 0.95 Å (Pt substrate) 0.98 Å (Cu substrate) and 0.98 Å (Ru substrate).¹⁶¹  In the case of the Ni precursor, saturation plots demonstrated a self-limited growth window between 180–195 °C with a GPC of 0.6 Å.¹⁶²  In both cases the metallic thin films were determined by XPS to be of high purity; >98% (Co) and >97% (Ni).

The same Co precursor, [Co{^tBu2DAD}₂], has also been reported in the ALD of Co₃O₄ thin films with O₂ as the co-reagent, across a temperature range of 125–300 °C. Interestingly, below 265 °C mixed phase CoO_x predominated (a mixture of CoO and Co₃O₄), whereas above 265 °C thin films appeared to be phase pure Co₃O₄. Growth rate data was not included as part of the report.¹⁶³ 

A family of N-heterocyclic carbene (NHC) stabilised Co complexes of the type [Co(CO)(NO)(NHC)(PR₃)] have been reported alongside an evaluation of their thermal properties and an evaluation of their application to ALD of Co based thin films. ALD experiments on the complexes [Co(CO)(NO)(ⁱPrIm)(PMe₃)], [Co(CO)(NO)(ⁱPrIm)₂], [Co(CO)(NO)(Me^tBuIm)₂] and [Co(CO)₂(NO)(ⁱPrIm)] (Fig. 12) were all run at 250 °C using a NH₃/H₂ mixed gas as co-reagent, however no details of ALD growth rates, or saturation plots were disclosed.¹⁶⁴ 

Perhaps not unsurprisingly given their proliferation across the other elements of the periodic table amidinate complexes of Co have also been reported. The amidinate complex bis(N-tert-butyl-N′-ethylpropionamidinato) cobalt(ii) [Co{(^tBuN)₂CMe}₂] has been used at 265 °C by atomic layer deposition from H₂ to produce Co metal thin films.¹⁶⁵  ALD half reactions between [Co{(^tBuN)₂CMe}₂] and H₂, on Cu substrates, have also been the focus of XPs investigations. Surface chemistries between these two reagents were evaluated by XPS. Adsorption of [Co{(^tBuN)₂CMe}₂] proved self-limiting and the precursor was shown to be reduced readily on Cu with and without H₂ co-reactant to form Co⁰.¹⁶⁶ 

ALD of the pyrite like metal disulfides FeS₂, CoS₂ and NiS₂ have also been reported. Deposition using the complexes [M{(^tBuN)₂CMe}₂] (M=Fe, Co or Ni) with H₂S plasma (3% in Ar) at 200 °C displayed GPCs of 1.06 (Fe), 1.29 (Co) and 1.24 (Ni) Å, with an apparently large ALD window (80–200 °C).¹⁶⁷  Contrastingly, application of the Co system [Co{(^tBuN)₂CMe}₂] with H₂S in a thermal ALD process displayed self-limited growth behaviour at 120 °C. Saturated growth rates were determined to be 0.27 Å per cycle producing the Co rich sulphide Co₉S₈, as determined by XPS.¹⁶⁸  Similar studies using [Ni{(^tBuN)₂CMe}₂] and H₂S again in a thermal ALD process, to produce NiS_x Films_, have also been reported.^169,170  In one instance the reaction was also monitored by QCM studies with a GPC of 9.3 ng cm⁻² reported.¹⁶⁹ 

The new cobalt complex [Co{DMOCHCOCF₃}₂] has been prepared by the reaction of cobalt(ii) acetate tetrahydrate with two equivalents of deprotonated ligand 1-(dimethyl-1,3-oxazol-2-yl)-3,3,3-trifluorprop-1-en-2-ol (DMOCHCOHCF₃). This new precursor has been reported to show a high degree of thermal stability. Optimised ALD processes and growth studies were performed on Si (100) and carbon/TiO₂ nanofibers substrates, at 145 °C with ozone (O₃) as the co reagent. Studies revealed a well-defined temperature range with constant growth per cycles of 0.2 Å per cycle between 150 °C and 200 °C of phase pure Co₃O₄.¹⁷¹ 

In a radical departure from complicated and often very air sensitive ALD precursors Väyrynen et al. have reported the simple diamine adduct [CoCl₂{κ²-Me₂NCH₂CH₂NMe₂] alongside a full ALD study of the precursor and its application to the fabrication of CoO thin films. Using water as a precursor the process was investigated over a 225–300 °C range, below the limit of the precursor's thermal stability. Saturation of the film growth with respect to both precursors was determined. At 275 °C high purity CoO was deposited at a growth rate of 0.2 Å per cycle regardless of Co precursor pulse length.¹⁷²  The implications of this new complex could be profound and is the first of what will be a much larger number of simple adduct systems to be assessed in the future.

As noted earlier in this section Cu has an assured place as the conductor of choice for interconnects in electronic devices.¹⁷³  As such, it has long been a material which has garnered attention as have precursors for the deposition of high purity Cu metal and the selective deposition of semiconducting oxide materials such as Cu₂O and CuO. Since the development of Cu(i) amidinates as precursors to copper metals by Barry and Gordon¹⁷⁴  only a small number of new precursors have been developed including the amino-alkoxide complex [Cu{dmap}₂].^175,176  The NHC-copper(i) amide system [{ⁱPr₂Im}Cu{N(SiMe₃)}] has been shown to deposit Cu metal, with growth rate per cycle of 0.2 Å, at 225 °C in a plasma enhanced ALD process using H₂ plasma as the co-reagent.¹⁷⁷ 

Atomic layer deposition has long been an attractive method by which thin films for advanced technological applications such as microelectronics and nanotechnology can be deposited. One material group in ALD that has matured in past 15 years and proven to be of importance to a wide range of technological applications are the noble metals, specifically the elements Ru, Os, Rh, Ir, Pt, Pd, Ag and Au. While these metals are known to be good conductors of both electricity and heat, they are perhaps even better known because of their application to catalysis and because of their resistance to corrosion and oxidation.¹⁷⁸ 

A summary of ALD precursors used to generate thin films for Group VIII, IX, X and XI metals is shown in Table 7. Whilst the noble metals have attracted attention, their oxides are not often considered. However, for materials such as RuO₂ and IrO₂, which are conductive and biocompatible and thus are possible candidate materials for biological applications and implantable devices, in addition to being structurally compatible electrode materials with high-κ dielectrics such as TiO₂ and SrTiO₃. In conjunction with the review by Miikkulainen et al. which provides an overview of precursors and general ALD trends across the periodic table,¹⁴⁸  a more recent review by Hämäläinen et al. provides comprehensive coverage of the topic across the noble metals.¹⁷⁸  The review by Hämäläinen et al. is comprehensive, exploring both noble metal containing precursors and their thermal ALD to produce noble metals and noble metal oxides. The review covers reaction mechanisms in various types of processes as well as specific issues regarding nucleation, in addition to tabulating and comparing the deposition temperatures, film growth rates, purities as well as evaporation temperatures for noble metal precursors. A review highlighting the specific chemistries of ruthenium β-diketonates and their application to ALD has also been published.¹⁸²  As a result we shall only describe here the development of new precursors which have appeared in the literature since the publication of these reviews.

Popovici et al. have described the ALD of Ru thin films starting from the new precursor (ethylbenzyl)(1-ethyl-1,4-cyclohexadienyl)Ru(0) (C₁₆H₂₂Ru) in combination with molecular oxygen (O₂) as co-reagent on both SiO₂ and TiN starting surfaces in the temperature range 225 and 325 °C. The study shows that the Ru growth behaviour is strongly dependant on the starting surface and explores how this can influence the film properties and growth kinetics.¹⁷⁹ 

Other developments in the area of precious metal precursor are centred on the development of a small number of new Au-precursors. In 2016 Mäkela and colleagues described their evaluation of a family of seven new gold precursors for application in the formation of gold containing thin films. From this shortlist of perspective precursors the liquid precursor [(Et₃P)Au{N(SiMe₃)₂}] was selected for investigation with potential reducing agents and hydrogen sulphide. The study describes the best in show results with respect to growth rate and film properties, obtained from [(Et₃P)Au{N(SiMe₃)₂}] and BH₃(NHMe₂). Unfortunately self-limiting growth profiles, characteristic of true ALD could not be confirmed; however the process did produce polycrystalline Au films of relatively high purity.¹⁸⁰  In studies by the same group of researchers another family of Au(iii)dithiocarbamate complexes were screened for potential ALD application. Of the systems investigated one complex, namely [Me₂Au{S₂CNEt₂}] showed optimal thermal behaviour, being volatile while displaying thermal stability. ALD growth experiments with O₃ at temperatures between 120–250 °C were investigated. Self-limited growth was confirmed at 180 °C with a rate of 0.9 Å per cycle growth of Au. The report describes the as deposited thin films of Au as uniform polycrystalline and highly conductive, with only trace amount of impurities (as determined by XPS).¹⁸¹ 

Perhaps even more so than any other group in the periodic table, the ALD of thin films containing Group XII elements, and in particular zinc containing materials is dominated by the alkyl metal reagents, such as diethyl zinc (DEZ) and to a lesser extent, dimethyl zinc (DMZ).^148,183  Whilst other precursors such as [Zn(OAc)₂], ZnCl₂, [ZnMe(OⁱPr)] and [Zn{THD}₂] have been explored in very specific ALD processes, in the past 20 years there is only one example of a new and novel ALD precursor for zinc based thin films.

In 2016 Devi and co-workers reported the design, synthesis, characterisation and viability as precursors for atomic layer deposition (ALD) of ZnO, of a family of zinc ketoiminates. Of the seven complexes explored one compound was outstanding with respect to its thermal properties, [Zn{OC(Me)CHC(Me)NCH₂CH₂OEt}₂] ([Zn(eeki)₂]) (Fig. 13). Deposition of ZnO thin films by ALD was studied using [Zn(eeki)₂] as the metal and water as the oxygen sources, respectively, over a temperature range of 150–300 °C. An ALD window was observed from 175–300 °C with an average growth rate of ∼1.3 Å per cycle. Growth rates were shown to be minimal at low temperatures highlighting that the activation energy (i.e. reaction with either the surface or H₂O) of the process was not sufficient to form a film as it was reactivity limited. The authors suggest that the broad ALD window provides a means of lengthening the window for stable high temperature ALD processes such as MLD processes as it widens the precursor cross-over window for organic molecules as well as dopants. The crystallographic orientations of the deposited thin film layers were found to be temperature dependent and contrary to the literature observation of c-axis dominant ZnO thin films grown using diethyl zinc. UV–Vis measurements revealed that the transparency is high in the visible range with a band gap of 3.29 eV.¹⁸⁴ 

For cadmium materials the precursor library is even more meagre, with dimethyl cadmium (DMCd) being the only reported ALD precursor. To the best of our knowledge there have been no developments in the arena of cadmium precursors in the past two decades. In the case of mercury, to the best of our knowledge no mercury based precursor has been reported to date.

The atomic layer deposition of group 13 materials is perhaps the most established and widely studied within the field, owing largely to the high reactivity of the simple boron-group alkyl compounds. The deposition of Al₂O₃ is the de facto “gold standard” of ALD processes, with a vast number of applications across multiple fields, whilst the subsequent use of (III–V) materials such as aluminium, gallium and indium nitrides, phosphides and arsenides in electronic applications has seen a wealth of research in the area over the last few decades. A description of the widespread use of simple alkyl, chloride, alkoxide and amide precursors is neither novel or within the scope of this overview, however examples of simple processes and precursors for many boron-group materials can be found within a 2013 review of inorganic ALD films by Miikkulainen et al.¹⁴⁸  Some examples of ALD precursors used to generate Group XIII materials is shown in Fig. 14 and summarised in Table 8.

Despite the unparalleled simplicity, reactivity and efficiency of established boron-group precursors, a number of avenues of research are being directed towards the development of less reactive, more exclusive and importantly less toxic reagents for group 13 materials. This is particularly prevalent in the deposition of boron-containing films, with a standard process revolving around the highly toxic and flammable diborane.¹⁸⁵  To this end, alkoxide compounds such as triisopropyl borate, [B(OⁱPr)₃], have been shown to be effective in the deposition of boron as a film dopant, and are seemingly less reactive and toxic than their alkyl counterparts.^185,186  Recently, B₂F₄ was shown for the first time to be used in the deposition of bismuth aluminate films with H₂O or reducing agent Si₂H₆ (for elemental B deposition), alongside trimethylalumimium (TMA), as co-reagents. Film growth was found to quickly decline after substrate saturation with processes consisting of solely B₂F₄/H₂O and B₂F₄/Si₂H₆. However, with alternate TMA/H₂O pulses, repeatable growth of bismuth aluminate was obtained.¹⁸⁷  Single-source boron containing precursors have also been used by Saly et al. to deposit calcium, strontium and barium borate with [M{Tpb}₂] (M=Ca Sr & Ba) and water, as highlighted in previous alkaline earth discussions.^71–73 

In similar attempts to temper the reactivity of aluminium precursors, a number of heteroleptic species such as non-pyrophoric dimethylaluminium isopropoxide [Me₂AlOⁱPr] have been developed and utilised in aluminium deposition, but are not recent enough developments to warrant more detailed discussion and can be found mentioned in a number of publications and the review described previously.^148,188  The alkoxy ether complex aluminium tris(1-methoxy-2-methyl-2-propoxide), or [Al{mmp}₃], (Fig. 13) was first reported to deposit Al₂O₃ films via solution injected ALD with H₂O at growth rates of ∼0.6–7 Å per cycle (200–350 °C) in 2005 by Min et al.,¹⁸⁹  and benefits from the enhanced stability of a chelating oxygen pendant moiety on each alkoxide ligand, though it is noted that the added bulk with respect to conventional monodentate ligand systems is the likely cause of the reduced growth per cycle cf. more established precursors.

A recent study by Mai et al. (2017),¹⁹⁰  synthesised three prospective non-pyrophoric aluminium precursors based on an aluminium centre with two varied, singly coordinated amines in addition to a stabilising donor functionalised alkyl ligand, 3-(dimethylamino)-propyl, or {dmp}. The three complexes, of the formula [Al(NR₂)₂{dmp}], where R=Me, Et and ⁱPr, exhibited good thermal stability and high volatility. In addition to the three amino precursors, a dimethyl aluminium derivative with {dmp}, [AlMe₂{dmp}] was also investigated. (Fig. 13). Of the synthesised precursors, [Al(NⁱPr)₂{dmp}] was shown to deposit Al₂O₃ films with H₂O (100–180 °C, 1.0 Å per cycle), whilst the fully alkylated compound [AlMe₂{dmp}] gave a GPC of 0.7 Å in a low-temperature (60–180 °C) oxygen plasma enhanced deposition.¹⁹⁰ 

Recent investigations into novel gallium precursor chemistries are more limited than their aluminium analogues, though a 2014 study by Ramachandran and co-workers¹⁹¹  trialled the application of [Ga{THD}₂] in the formation of Ga₂O₃ films. Although the precursor was shown to have enhanced volatility and thermal stability over its parent acetylacetonate, [Ga(acac)₃] – previously used in gallium ALD¹⁹²  – growth was only observed with an O₃ oxidant and exhibited a very low growth per cycle of 0.1 Å within a temperature window of 100–400 °C.¹⁹¹  In a bid to decrease the pyrophoric nature of gallium alkyl precursors, an investigation by Lee et al.¹⁹³  sought to apply the volatile, non-pyrophoric and reactive dimethylgallium isopropoxide (DMGIP), [GaMe₂(OⁱPr)], with an H₂O co-reagent to the ALD of gallium oxide. The process yielded a narrow temperature window (280–300 °C) with a low growth rate of ca. 0.3 Å per cycle. Interestingly, when compared to the analogous aluminium compound DMAIP, (Me₂Al(OⁱPr)), mentioned briefly previously, this growth rate is exceedingly low taking into consideration that fact that the growth rate for the aluminium species is nearly comparable to the standard pyrophoric trimethylaluminium process.^188,193 

There is relatively more abundance within novel indium precursor development than there can be found within that of the earlier members of the group, though a wide variety of existing processes are covered in the 2013 review by Miikkulainen et al.¹⁴⁸  Research of a similar vein to that taking place within aluminium and gallium ALD is being undertaken, with the varying of simple amido and alkyl substituents and their effect on deposition properties explored. Such a study utilised dimethylamino dimethyl indium [InMe₂(NMe₂)] and H₂O, with deposition of In₂O₃ occurring between 250 °C and 400 °C, with a GPC of 0.61–65 Å per cycle between 300–350 °C.¹⁹⁴  A similar investigation by Maeng et al.¹⁹⁵  demonstrated growth of indium oxide films between 175 °C and 250 °C (GPC, 0.7 Å), with the similar aminosilyl precursor [InMe₂{N(SiMe₃)₂}]. Further to this work, the volatile liquid [InMe₂{dmp}], analogous to the aluminium precursor previously discussed, has been trialled with H₂O in depositions found to afford indium oxide at 275 °C with a significant GPC of 0.6 Å.¹⁹⁶ 

Investigations into novel precursors by Kim et al.¹⁹⁷  in 2016 saw the low temperature (90–180 °C) deposition of In₂O₃ with O₃ at a rate of ca. 0.5 Å per cycle with the precursor dimethyl(N-ethoxy-2,2-dimethylcarboxylicpropanamide)indium, [InMe₂{edpa}] (Fig. 13). A subsequent study by Agbenyeke and co-workers¹⁹⁸  saw deposition of the same material using H₂O at 200–300 °C (∼0.8 Å per cycle).

Two novel indium guanidinates of the form [In{(ⁱPrN)₂C(NR₂)}₃] (R=Me, Et), were described by Gebhard et al.¹⁹⁹  (Fig. 13) and were applied to the ALD of indium oxide with H₂O. The monomeric compounds were found to deposit In₂O₃ at a growth rate of 0.4–0.5 Å per cycle at 230–300 °C. A later study into novel indium precursor systems yielded 1-dimethylamino-2-methyl-2-propoxy indium, or [In{dmamp}₃], which, when reacted with O₃ at temperatures between 175–200 °C was found to deposit carbon-negligible In₂O₃ at a GPC of 0.27 Å.²⁰⁰ 

As has been highlighted previously, carbon incorporation by design in the form of carbonates has long been achieved with the use of CO₂ or by plasmolysis or ozonolysis of ligands. Another well-established route towards incorporation of carbon into thin films is via the reaction of alkylated ligand systems with H₂ plasma.²⁰¹  The following overview will attempt to avoid these processes and focus on recent developments in the specific atomic layer deposition of carbon. Examples of ALD precursors used to generate Group XIV materials are summarised in Table 9.

The selection of appropriate methodology for a novel carbon containing precursor with sole aim of introducing carbon into a film is desirable despite the well-established nature of carbon incorporation by other means. This is of particular interest in the deposition of silicon carbide (SiC), an important material for semiconductor applications, with a low wet-etch rate and higher operating temperature than SiN and SiO₂. Current thermal CVD processes for crystalline material require high temperatures >1000–1500 °C, only somewhat mitigated by plasma enhanced CVD, however deposition of amorphous material is possible at temperatures as low as 200 °C.^202,203  To this end, computational studies have attempted to direct research towards this arena, highlighting ethylene (C₂H₄), carbon tetrachloride (CCl₄) and trichloromethane (CHCl₃) as most promising for use alongside silicon precursors silane (SiH₄), disilane (Si₂H₆) and monochlorosilane (SiH₃Cl), though it is acknowledged that the processes occurring at the surface are not likely to be self-limiting.^203,204 

As, once again, the problematic topic of self-limiting ALD reactions is probed, the most notable novel deposition focussed on carbon films is that of graphene, which was reported by Zhang et al. in 2014.²⁰⁵  As seen in conventional graphene CVD, copper foil was used as a substrate, with benzene (C₆H₆) as a pre-aromaticised carbon source. Depositions were carried out with sequential pulses of H₂/Ar plasma at 400 °C. 10 ALD cycles were found to be sufficient to afford the formation of a monolayer graphene sheet of reasonable size.

It is perhaps unsurprising that, as the stalwart of the semiconductor industry, a number of established ALD processes exist towards the fabrication of silicon containing thin-films.¹  A discussion of simple, conventional precursors such as silanes, amidosilanes and chlorosilanes^206–208  is as such beyond the scope of this overview, however a thorough review by Meng et al.²⁰⁹  (2016) of SiN deposition processes provides a good baseline. The development of non-chlorinated precursors is desirable in the first instance, not least due to the corrosive by-products formed.²¹⁰  A study into the impact of aminosilane precursor structure on SiO₂ ALD was carried out by O'Neill et al.,²¹¹  which suggested that a higher degree of amino substitution of silanes resulted in compromised growth rates and ligand incorporation into films, a finding corroborated by Mallikarjunan and co-workers exploring the efficacy of the monoaminosilane, di-sec-butylaminosilane (H₃SiN^sBu₂/DSBAS) with the diaminosilanes bis(tert-butylaminosilane) (BTBAS) and bis(diethylaminosilane), of the formula SiH₂(NRR′)₂ (R=^tBu, R′=H and R=Et, R=Et).^212,213  All three variants described have found use in SiO₂ ALD, with the high growth rates (∼1.6 Å per cycle, O₃) at low (150–300 °C) and moderate (250–300 °C) temperatures respectively for di-sec-butylaminosilane (DSBAS) and bis(tert-butylamino)silane (BTBAS) proving to be of significant interest.^212,214  A number of other amiosilanes have been reported and a computational study by Huang et al.²¹⁵  can be found on the optimisation of precursor design in this field. Other silicon precursors studied for SiO₂/silcon nitride deposition include, among others, the derivatives of silanol or silyl ethers, such as tetraethoxysilane (TEOS), tri(tert-butoxyl)silanol (tBOS), and tris(tert-pentoxy)silanol (TPS), which have been used with H₂O or N₂/NH₃ (plasma) as traditional co-reactants,^216–219  or with trimethylaluminium in a catalysed process known as “rapid ALD”, where multiple layers of material are deposited with each cycle.^{210,220–223}  Interestingly, despite the multilayer growth observed within “rapid ALD”, conformity of high aspect ratio substrates was conserved. These catalysed processes, undertaken in the presence of Lewis acid (TMA)^221,224  or Lewis base (ammonia/pyridine)^225,226  catalysts serve to allow the low temperature deposition of silicon containing films, without the need for aggressive plasma or thermal conditions.²¹⁵ 

Perhaps the most inventive silicon-based precursor development in recent years has been the design and synthesis of a range of cyclic azasilanes by Ju and Strandwitz in 2016²²⁷  (Fig. 15). A series of four ring-strained systems, with a variety of substituents, displayed high reactivity towards SiO₂ substrates with an O₃ co-reagent. The study offers insights into the reactivity of various silicon-based functional groups, with Si–OMe moieties displaying shorter O₃ saturation pulses than their alkyl counterparts.

As group 14 is descended, the prevalence of the 2+ oxidation state increases and due to the demand for both Ge(ii) and Ge(iv) thin films in electronics, processes for the deposition of both valences are desirable. Of these Ge(ii) and Ge(iv) materials, many of the chalcogenides are of importance, as are ternary germanium materials. Germanium telluride (GeTe) and germanium-antimony-telluride (Ge₂Sb₃Te₅) are of great interest as phase change memory materials and as such a number of ALD routes towards them have been developed. A recent review by Harmgarth et al.²⁴⁸  and other publications^{228,249–251}  cover these precursors in greater depth than the scope of this overview, with many of the precursors discussed based on simple germanium halides, alkoxides and amides. There is a relative scarcity of novel precursor development for the atomic layer deposition of germanium chalcogenides, particularly within divalent species, though work in this area is expanding.

Germanium alkoxide precursors have been investigated extensively by Eom et al.^252,253  and others²⁵⁴  for the formation of germanium-antimony-tellurides, with the general conclusion that, though the process is not fully understood, interaction between the germanium precursors and substrate surface was limited as long purge times resulted in little to no deposition, and the preferential formation of Ge(iv)Te₂ as opposed to Ge(ii)Te was a significant hindrance. In an attempt to circumvent this, a combined ALD/CVD process was developed using [Ge(i-C₄H₉)₄] and a reductive H₂ plasma which has been shown to be effective in the deposition of Ge₂Sb₃Te₅.²²⁹  Another recent method towards GeTe films utilises the Ge(ii) amide precursor [Ge{HMDS}₂] and [Te(SiMe₃)₂], which when pulsed alongside methanol to create reactive intermediates in the gas-phase, afforded GeTe thin films. It was found to be necessary to induce gas-phase reactions with methanol due to the lack of reactivity between [Ge{HMDS}₂] and [Te(SiMe₃)₂].²²⁸ 

An interesting precursor for the deposition of both SnS and GeS was developed by Kim and co-workers in 2014, which consists of a single dianionic, racemic cyclic amide ligand with sufficient steric bulk to singly satisfy the coordination sphere of Sn(ii) and Ge(ii). The cyclic amide complex, (Fig. 15) [Ge{(Bu^tN)CH(CH₃)CH(CH₃)(NBu^t)}₂], [Ge{CAMD}₂], was found to deposit with a growth rate of ca. 0.21–0.28 Å per cycle between 50–100 °C after which a decrease in growth was observed.²³⁰ 

GeO₂ was deposited with a divalent, two-coordinate precursor by Perego et al.,²³¹  1,2-bis[(2,6-diisopropylphenyl)imino]acenapthalene germanium(ii), [Ge{dpp-BIAN}] (Fig. 15), and O₃ at temperatures between 185 °C and 225 °C (∼0.5 Å per cycle).^231,255  Further study into novel systems for the deposition of GeO₂ focussed on tetravalent systems, through the somewhat simpler Ge(NMe₂)-chelates (Fig. 15), where coordinative saturation was achieved with N,N′-di-tert-butyl-ethylenediamine or its isopropyl analogue, with growth rates ca. 0.40 Å and 0.31 Å per cycle respectively between 200–330 °C.²³² 

A wide variety of tin(ii) and tin(iv) materials are of increasing interest, with the tin chalcogenides the primary focus of investigations. A 2015 review by Nazarov et al.,²⁵⁶  in addition to a review by Miikkulainen¹⁴⁸  and co-workers, covers many precursors for the deposition of SnO₂ and SnS. Such common and well-established precursors include simple Sn(iv) alkyls, amides and chlorides. With the intermediate position of tin within the carbon group, distinction must be made between the various valences of tin(ii) and (iv) materials, with oxidative control of respective states most difficult to establish in oxygen-rich and highly oxidising environments. This is particularly pertinent in the production of tin oxide materials, with the pure SnO phase considerably less stable than the SnO₂ and SnO_x phases.

Testament to this fact, a number of divalent tin precursors have been trialled in the deposition of tin oxide films, yet have failed to retain the divalent nature of tin within the material. An example of this can be seen with the β-diketonate precursor [Sn{acac}₂], previously used in the production of SnS.^233,234,257  Sn(acac)₂ was found to deposit films of SnO_x within a temperature window of 175–300 °C (∼1 Å per cycle), whilst was used recently by Chang et al.²³⁴  to obtain good stoichiometric control over the production of Sn_xTi_1−xO_y films with titanium isopropoxide and ozone. The same precursor was used with H₂S in further work on SnS deposition to obtain phase control over cubic and orthorhombic SnS films via ALD.²⁵⁸ 

The somewhat subtle nature of tin oxide deposition with respect to oxidation is again exemplified with the N-heterocyclic stannylene precursor investigated by Kim et al.,^230,235  [Sn(ii){(Bu^tN)CH(CH₃)CH(CH₃)(NBu^t)}], the germanium analogue of which was also described. The stannylene was reacted with H₂O₂ to yield SnO₂ films at temperature between 50 °C and 150 °C (∼1.8 Å per cycle),²³⁵  whilst a subsequent group publication sought to avoid the use of unstable H₂O₂, which was thought to decompose within high aspect-ratio substrates at higher temperatures, by using NO as an oxidant, achieving SnO₂ films at a rate of ca. 1.4 Å per cycle at higher temperatures of 200–250 °C.²³⁶  The same precursor was shown in a later study to deposit SnS with an H₂S co-reagent.²³⁰  Work by the same group investigated the application of a second Sn(ii) precursor, the acetamidinato complex [Sn{(ⁱPrN)₂CMe}₂], which shows growth of single-phase SnS at 100–200 °C with a growth rates of ∼0.9 Å per cycle when used with H₂S.²³⁷ 

Oxidative control over the deposition of tin(ii) oxide films was achieved by Han and co-workers²³⁸  using the Sn(ii) precursor bis(1-dimethylamino-2-methyl-2-propoxy)tin [Sn{dmamp}₂]. SnO films grown with H₂O were deposited between 90 °C and 210 °C with phase pure, crystalline films grown at 150 °C and above with a growth per cycle of 0.18 Å. Contrastingly, when ozone was used as co-reagent in a subsequent publication,²³⁹  SnO₂ films were deposited at 100–230 °C exhibiting increasing growth rates with temperature (0.18–0.42 Å per cycle). Further work into the oxidative control of SnO deposition was undertaken by Tupala et al.,²⁴⁰  in a study that applied the long-established stannylene [Sn{HMDS}₂] to atomic layer deposition processes with both H₂O and O₃. Films of SnO were grown with H₂O displaying a variable growth rate (0.05–0.18 Å per cycle) at temperatures of 100–250 °C, whilst films grown with ozone showed lower, even more variable growth rates (0.05–0.11 Å per cycle) at temperatures of 80–200 °C. Importantly, films grown with water were shown to consist of crystalline SnO between 125 °C and 175 °C but contained significant quantities of SnO₂, silicon and nitrogen, and films grown with O₃ were found to consist of SnO₂ and SiO₂, with similar nitrogen contamination.²⁴⁰ 

Despite the use of lead in thin film materials becoming increasingly limited, the manifold oxides (simplified in this study to PbO_x) and sulfides of lead(ii), in addition to ternary structures such as PbTiO₃, have found use in a wide range of optoelectronic devices, solar and sensing applications.^{245,246,259–261}  The heavier lead chalcogenides such as PbSe and PbTe are utilised in thermoelectric, photo optic and other semiconductor applications.^243,262  As with other group 14 elements, simple alkyl and alkoxide complexes of lead and their long-standing uses in deposition are omitted.^245,261,263  Very few recent advances within lead precursor chemistry pertinent to this overview exist, though a brief description of some relevant investigations is given herewith.

PbTe was deposited for the first time in 2014 by Zhang and co-workers²⁴³  via the atomic layer deposition of Pb(THD)₂ and Te(SiMe₃)₂ at growth rates of 0.25 Å per cycle between 150–210 °C, whilst nanolaminates of PbSe/Te were deposited by the same author with the addition of Se(SiEt₃)₂.²⁴⁴  The Pb(THD)₂ complex has also been shown to deposit films of PbS and PbO₂ with H₂S and O₃ as oxidants respectively.^245–247  Further to this, aminoalkoxide Pb(DMAMP)₂, analogous with the Sn(ii) chelate of identical composition, has been successfully used with H₂O on a number of occasions to deposit lead oxides.^241,242 

A number of nitrogen and phosphorous containing films have previously been touched upon in the discussion of advances in the precursor chemistry of lithium, group 13 elements and silicon, and the reader is directed towards more relevant literature in the case of further interest. The deposition of the metalloids and metallic pnictogens arsenic, antimony and bismuth materials are the primary focus of the following subsection, though little significant precursor development has taken place in recent years, in part due to the high efficacy of current technologies.

The evolution and widespread demand within the semiconductor industry for III–V (13–15) materials has seen the development of a range of similar, well-established ALD routes based around alkylated precursors, as mentioned in the group 13 discussions. Arsenide films are all routinely deposited through use of simple metalloid-alkyls, amides or hydrides, such as ^tBuAsH₂,²⁶⁴  EtAsH₂,²⁶⁵  As(NMe₂)₃,²⁶⁶  and, most commonly AsH₃,^267–269  whilst antimonide films often utilise antimony alkylsilyls such as Sb(SiEt₃)₃.²⁷⁰  Materials such as Sb₂O₃ and Sb₂S₃ are also of interest for battery and photovoltaic applications,^271,272  with simple precursors such as Sb(NMe₂)₃ reacting with O₃ or H₂S to afford the desired chalcogenide film.²⁷³  Previous mention has been made of germanium antimony telluride as a phase-change material, with antimony chlorides and alkoxides being used to this end.^228,274 

A number of more varied precursors for the deposition of bismuth containing thin films have been investigated with a view to the deposition of a wide variety of materials, with Bi₂O₃ and Bi₂S₃ desirable due to tunable electronic and thermoelectric properties and uses in supercapacitors and gas sensors.^275–277  More complex materials such as bismuth silicate, bismuth ferrate and bismuth titanates are also of interest for a range of electronic devices.^277,278 

The use of triphenylbismuth [Bi(C₆H₅)₃] is relatively well established within CVD and ALD, readily oxidised to Bi₂Ti₂O₇ with titanium isopropoxide and H₂O, and Bi₂O₃ with O₃ (∼0.23 Å per cycle, 250–320 °C).^277–279  An interesting single-source precursor Bi(CH₂SiMe₃)₃ for bismuth silicate was developed by Harjuoja et al.,^280,281  acting as both bismuth and silicon source when reacted with O₃. Growth rates of 0.4 Å per cycle between 250 °C and 350 °C were observed. For a number of years the chelating [Bi{mmp}₃] system, tris(1-methoxy-2-methyl-2-propoxy)bismuth (Fig. 16), has been used to deposit both Bi₂Ti₂O₇ and Bi_1−x−yTi_xSi_yO_z (BTSO), with reactant pulses of [Ti(OⁱPr)₄] and H₂O,²⁸²  and [Ti(OⁱPr)₄] or [Ti{mmp}₄], [Si(OEt)₄] and H₂O^283,284  or O₃²⁸⁵  respectively.

As is the case for many other ALD processes, there are reports of the use of the β-diketonate ligand THD with H₂O for bismuth thin film deposition, but the process exhibited low growth rates of ca. 0.1 Å per cycle between 270 °C and 300 °C.²⁸⁶  The same precursor, with the considerably more acidic H₂S as a co-reagent, afforded polycrystalline p-type Bi₂S₃ at a GPC of 0.35 Å between temperatures of 175 °C and 250 °C.²⁸⁷  A 2004 study by Vehkamäki and co-workers²⁸⁸  sought to investigate the utility of a range of bismuth amide and thioamidate complexes; homoleptic silylamides [Bi{N(SiMe₃)₂}₃] ([Bi{HMDS}₃]), [Bi{N(SiMe₂Et)₂}₃], [Bi{N(SiMe₂ⁿBu)₂}₃] and [Bi{N(SiMe₂CHCH₂)₂}₃]; homoleptic alkylamides [Bi(NEt₂)₃] and [Bi(NⁱPr₂)₃]; donor functionalised alkylamide [Bi{^tBuNCH₂CH₂NMe₂}₃] and a single thioamidate [Bi{SC(Me)NⁱPr)₃]. The presence of the thioamidate was intended to probe the effect of a conjugated soft sulfurous base on the soft acid bismuth. The commonplace silylamide, [Bi{HMDS}₃] was taken forward to H₂O-oxidised depositions based on thermogravimetric analysis. A narrow temperature window (190–200 °C) was found to produce binary oxide material with poor reproducibility and growth rates between 0.15 Å per cycle and 0.23 Å per cycle. Multicomponent bismuth tantalum oxide films were, however, significantly more reproducible with [Bi{HMDS}₃] as a bismuth source, however the narrow temperature window between volatilisation and decomposition of precursor place large limitations on its further use.²⁸⁸  Findings by Rusek et al.²⁸⁹  (2017), during further studies into a range of bismuth amides as precursors for Bi₂Te₃, are consistent with proving [Bi{HMDS}₃] to be the most promising bismuth source of the compounds studied, however reiterate that more thermally stable bismuth complexes are necessary to carry out >200 °C deposition. Alternative ALD routes based on the reaction of BiCl₃ with [Te(Et₃Si)₂] have been shown to deposit bismuth telluride reliably.^290,291 

An investigation published in 2010 by Hatanpää et al.²⁹²  drew comparisons between a range of established and prospective bismuth precursors: Bi(O^tBu)₃, [Bi(OCMe₂ⁱPr)₃], [Bi(OCⁱPr₃)₃], bismuth β-diketonate, [Bi{THD}₃], and bismuth carboxylate, [Bi{O₂C^tBu}₃]. ALD experiments were undertaken with an H₂O oxidant with all five precursors, with the bismuth alkoxide [Bi(OCMe₂ⁱPr)₃] (m.p. 40 °C) found to be the most successful when used at an evaporation temperature of 85 °C. A maximum growth rate of 0.36 Å per cycle was obtained on an Al₂O₃ surface at 150 °C.²⁹²  A summary of ALD precursors used for Group XV thin films is shown in Table 10.