The chemical physics of sequential infiltration synthesis—A thermodynamic and kinetic perspective

Sequential infiltration synthesis (SIS) has emerged over the past decade¹ as a technique to grow hybrid organic-inorganic materials and rapidly fabricate structured inorganic films. SIS relies upon the diffusion of metal-organic vapor precursors into the bulk volume of polymers. Due to attractive chemical interactions with certain polymer functional groups, precursors can exhibit prolonged residence with the polymer. This enables SIS to realize the entrapment of the inorganic metal oxide reaction products within the polymer. This technique works across multiple length scales—whether transforming the properties of macroscale materials or functionalizing nanoscale devices formed by lithography or molecular self-assembly.² 

The development of SIS emerged from the application of atomic layer deposition (ALD) to polymer films. In ALD, alternating pulses of vapor precursors are used to grow thin conformal layers of inorganic materials on solid substrates. Each precursor irreversibly reacts with the substrate chemically in a self-limiting fashion, presenting alternating surface chemistry appropriate for reaction with the subsequent precursor. Polymers, composed of entangled molecular chains with associated free volume, are not dense solids, and the precursors used in ALD can often diffuse through them substantially. The diffusivity of precursors enables SIS in which inorganic material is intentionally deposited within the polymeric phase. While SIS often utilizes ALD precursors and is often conducted in the reactors designed for ALD, the processing parameters required for uniformity and reproducibility differ substantially (Fig. 1). In typical ALD processes, precursor pulses are brief (<2 s) as they are required only to provide sufficient exposure to saturate the surface chemical groups of the growth substrate. In SIS, by contrast, pulse pressures are comparatively larger to provide enough precursors to infiltrate a 3D volume, and exposures are comparatively longer to allow a complete diffusion of the precursors into the polymer phase. The diffusive nature of precursor transport within polymers, in contrast to simple saturation on solid substrates, means that the distribution of the precursors within the film is sensitive to diffusion time, precursor vapor partial pressure, temperature, interactions between the polymer and precursor, and polymer microstructure.

Researchers have referred to the process of infiltrating metal-organic precursors into polymers by a variety of different names, such as sequential infiltration synthesis (SIS), vapor phase infiltration (VPI), sequential vapor infiltration (SVI), and multipulse vapor infiltration (MPI). In some implementations, the precursor vapors are delivered in one long pulse. In others, the precursors are delivered in many shorter pulses. Some implementations involve sealing off the growth chamber to contain a large static dose of pressure, while others rely upon long durations of transient precursor flow that is constantly pumped. The different terminology and variations in process parameters should not obscure the fact that these all describe the same fundamental phenomena. The cumulative duration of exposure and partial pressure of the precursor vapor as well as subsequent purge dictate the diffusion of precursor infiltration into and out of the polymer bulk. In this perspective, all referenced literature will be called SIS for simplicity.

SIS emerged as a distinct concept as a consequence of attempts to apply ALD to polymeric substrates by the ALD research community. The diffusion of precursors beyond the near surface of polymers was observed as early as 2005, when the George group noted that polymers, including polymethyl methacrylate (PMMA), could uptake trimethyl aluminum (TMA) by absorption within their free volume.⁹¹ However, the crucial role of interactions between precursors and polymer functional groups was not discussed in these early studies. Indeed, the diffusivity of precursors into polymer films was considered a problem in some early studies. In experiments considering PMMA as a masking layer for area-selective ALD for patterned TiO₂ growth, researchers found that titanium tetrachloride (TiCl₄) diffused and reacted deeper into PMMA films than titanium isopropoxide (TIP).^3,4 Unless the PMMA films were of sufficient thickness, precursors would diffuse through the entire film and react with the masked silicon substrate. Therefore, the diffusion of ALD precursors into and reaction with polymer films were viewed as challenges to be overcome. With careful control, ALD largely on the surface—but not intentionally within polymer films—is possible and has its own technological utility, as has been reviewed previously.⁵ 

In the years since the diffusion of ALD precursors through polymers was observed, SIS has been developed to intentionally enable the direct incorporation of metal oxide or metallic materials within polymers, yielding hybrid materials with novel properties. It should be noted that SIS is distinct from molecular layer deposition (MLD) in which purely organic or organic-inorganic hybrid materials are applied as sequential monolayers using self-limiting surface reactions in which the organic components are incorporated in the vapor phase precursors and are retained in the resulting thin films.⁶ SIS-derived hybrid materials may further be subsequently subjected to plasma or thermal annealing treatments to remove the polymer constituents entirely, resulting in purely inorganic metal or metal oxide structures that retain the form of the original polymer morphology.^7,8 Due to the versatility of SIS, researchers have utilized SIS for a wide host of applications with more on the horizon. The demonstrated application spaces for SIS have been recently reviewed in detail.^9–11 A survey of SIS literature organized by use inspiration is tabulated in Table I. The first use inspiration category in Table I lists reports that investigate the fundamental physical chemistry that underpins SIS phenomena or otherwise quantifies processing parameters.

As the application space for SIS widens, the functional implementations of the technique are at risk of outpacing the development of a firm theoretical and experimental physicochemical basis for SIS. Attempts at a complete description of the varied and complex phenomena at play during SIS have only just begun.²⁶ Complete descriptions are complicated by the disparate length and time scales involved in the SIS process and the wide range of interaction strengths introduced by choice of metal-organic precursor, polymer structure, polymer functional group, temperature, and other experimental design variables.

The goal of this perspective is to concisely present the state-of-the-art in SIS chemical physics to introduce a conceptual framework of this developing topic to unfamiliar researchers. As such, this paper focuses on the fundamental physical and chemical phenomena that are at play in SIS. The perspective is structured as follows:

• Introduction of the conceptual underpinnings of the physical and chemical phenomena involved in SIS.

• Examination of the experimental “phase space” available in SIS process design.

• Description of the many characterization techniques from which one may deduce the fundamental phenomena of SIS.

• Survey of simulation and calculation methods applied to SIS.

• Perspective on the outstanding gaps in knowledge and potential avenues for future research.

Thermodynamics govern the interactions between the chemical precursors and the polymer chains. A clear understanding of the thermodynamics of the system facilitates the rational design of an SIS process.

The first thermodynamic process in SIS is the change of phase that takes place when the precursor vapor molecule leaves the headspace of the reactor chamber and dissolves into the polymer matrix. If the precursor is insoluble in the polymer, SIS cannot occur. If the precursor can dissolve into the polymer, then the solubility will dictate the equilibrium concentration of the precursor in the polymer. The concentration of the precursor in the film can impact the density of the resulting SIS material. The simplest classical thermodynamic model for solubility is Henry’s law,

where C is the concentration of the precursor in the polymer, S is the solubility coefficient of the precursor in the polymer, and p is the partial pressure of the precursor in the vapor phase. The van’t Hoff equation describes the temperature dependence of S,

where k is Boltzmann’s constant, T is the temperature, S₀ is a constant, and ΔH_s is the partial molar enthalpy of dissolution. These basic expressions do not capture the full complexity of the polymer matrix, which consists of free volume elements (also called microvoids) that contribute to the solubility of infiltrating precursors. These are treated in “dual-sorption theory” or Langmuir models that add contributions from the free volume in addition to the polymer chain volume.⁹⁰ 

These simple expressions provide three possible routes to maximizing the equilibrium solubility of the precursor in the polymer: raising the partial pressure of the precursor vapor, raising the temperature of the system, and choosing precursor/polymer pairs with the smallest ΔH_s. The accessible range of partial pressure for a given precursor is limited to its vapor pressure at the SIS temperature, in order to prevent condensation of the precursor into the liquid phase on the sample and reactor wall surfaces. Engineering of the polymer free volume may also enhance the solubility of precursor molecules within a polymer film. Temperature will have an impact on this parameter beyond that reflected in the solubility enhancement from the van’t Hoff equation. This is especially true in the vicinity of phase transition temperatures for the polymer (glass, crystallization, etc.).

The solvation of precursor molecules into a polymer film enables the retention of the precursor within the film during the removal of the partial pressure of the vapor phase by purging via vacuum pumps. These solvated molecules may not be interacting with functional groups along the polymer, but are arrested in their movement by the energetic barriers to site-hopping imposed by their local configurational environment. The molecules must change phase back from their solvated state to the gas phase during this purging process, the kinetics of which will be discussed later. The transient entrainment of precursor molecules within the polymer network enables material growth in polymers inert to strong interactions with organometallic precursors, for example, polystyrene (PS)²⁷ and polyethylene (PE) that lack polar groups.⁹¹ If the purge time between metal-organic precursor exposure and oxygen-bearing precursor exposure is sufficiently short, the two reactants can mix and react in the polymer phase at the exclusion of substantial mixing in the vapor phase. In this case, the time allowed for desorption of the precursor from the polymer phase to the vacuum is insufficient to completely remove the molecule from the polymer. In a sense, this can be considered a localized chemical vapor deposition (CVD) chemistry within the polymer. However, the outgassing of solvated molecules in inert polymers is expected to be rapid, resulting in an extremely low density of the precursor in the film over practical time scales for the introduction of the second precursor without gas-phase mixing occurring. This yields slow growth rates in such processes, leading to studies of polymer modification with oxidative processes such as plasma treatments to modify the polymer prior to SIS and thereby introduce favorable precursor-polymer interactions.^92–95 

With precursor molecules solvated within the polymer film, the next thermodynamic consideration is the interaction between the precursor and the chemical moieties built into the polymer chain. There are a number of possible interactions that a precursor molecule may have with the polymer. These interactions may lead to the formation of a reversible complex between the precursor and polymer functional group or lead to complete electron transfer yielding a covalent bond between the precursor and polymer functional group. Depending on the interactions of the precursor molecule, the resulting SIS product may be covalently linked to the polymer backbone in a hybrid material or may exist as a nanocomposite without chemical bonds between the polymer and nucleated metal or metal oxide.

Research into SIS has provided ample evidence for the existence of a noncovalent attractive interaction between the aluminum atom of TMA and the C=O bond of the ester carbonyl in PMMA.^21,23,24,28 The aluminum in the TMA monomer is sp² hybridized with the three methyl groups bound by the hybrid orbitals formed from the 3s and two 3p orbitals. The remaining empty 3p orbital results in a strong Lewis acid,^96,97 which is capable of accepting the lone pairs of the Lewis basic C=O to form a Lewis adduct [Fig. 2(a)]. The vapor-phase TMA molecules can be present in equilibrium with a dimer form, where each Al is bound to four methyl groups, at the temperatures and pressures used in SIS, though the stability and interactions of this dimer within polymers are unkown.²⁶ Density functional theory (DFT) has been used to estimate the energy of this TMA-PMMA Lewis adduct—the energy of the complex formed between the TMA molecule and a PMMA monomer was calculated to be −0.477 eV at 300 K.¹⁶ The Lewis adduct is reversible, which enables the interaction to be broken and reformed as the precursor diffuses through the polymer film and encounters multiple adduction sites.

When water (or other oxygen sources such as ozone,³⁰ oxygen plasma,³¹ or hydrogen peroxide³⁷) is introduced into a polymer with adducted or trapped organometallic precursors supplied during the first half-cycle of SIS, it encounters a different chemical environment than in the case of ALD. In the SIS case, the water can react directly with molecular TMA that still has all three methyl groups, in contrast to a monomethyl- or dimethyl-aluminum terminated surface as in ALD. The interaction of the organometallic precursor with its environment may also change its reactivity (e.g., by weakening bonds), enabling growth even under conditions that are not amenable to ALD. For example, while the lowest reported temperature for ALD growth of In₂O₃ using trimethyl indium and water is 200 °C,⁹⁸ SIS using these same precursors was demonstrated by our group at 80 °C in PMMA—conditions under which no ALD was observed.¹⁶ 

If the organometallic precursor and polymer functional group are sufficiently reactive, a substrate-precursor reaction can occur simultaneously with diffusion that directly covalently links the metal to the polymer. For example, TMA and PMMA form a reversible Lewis adduct at low temperatures, but at temperatures in excess of at least 150 °C, the two are proposed to undergo an irreversible pericyclic reaction to form PMMA-Al acetate and ethane as a byproduct, as shown in Fig. 2(b).^25,28

The choice of organometallic precursor and polymer functional group can also affect the balance between reversible complexes and covalently reacted groups with the polymer. Direct covalent reactions with a particular precursor can be achieved at lower temperatures through selection of a polymer functional group with a lower energy barrier. Polyacrylic acid (PAA) differs from PMMA in that it has carboxylic acid instead of methyl ester moieties. Recent evidence from the Parsons group points to the direct covalent incorporation of TMA in PAA at significantly lower temperatures (as low as 60 °C) relative to PMMA. This may be attributed to the presence of an acidic proton, which facilitates the reaction, generating methane as a byproduct.²⁸ Careful selection of the functional group and process temperature can, therefore, determine the balance between reversible complexes and direct covalent bond formation during organometallic precursor exposure. This balance is crucial for controlling the uniformity and depth of deposition in polymer films and bulk polymers. In the limit of rapid covalent bond formation that meets or exceeds the kinetics of diffusion, there is a possibility that further diffusion into or out of the underlying polymer film will be impeded as the polymer free volume becomes filled by the covalently bound precursor.

Designing SIS reactions a priori will require detailed understanding of the reaction pathways and interaction energies between polymer functional groups and metal-organic precursors. The reaction pathways of various organic functional groups with TMA were recently studied by the Knez group with a combined experimental and computational effort using substituted phenyls with hydroxyl, amino, and nitro groups.⁹⁹ This study found energetically favorable and spontaneous reversible complex formation between each of these three functional groups and TMA but found energy barriers to the reactions that would form a covalent bond and liberate a TMA methyl ligand. The calculated energies for the various steps are provided in Table II, and a representative energy-reaction coordinate map for the 4-hydroquinone example is shown in Fig. 3.

The most significant changes to the chemical landscape of the hybrid film occur between the first cycle and subsequent SIS cycles. In the first exposure of a metal-organic precursor, the precursor can only interact with functional groups intrinsic to the polymer itself (or residual contaminants such as H₂O). However, after the first full SIS cycle, which includes exposure to the oxygen source—typically water—the film contains both intrinsic polymer functionalities (which may or may not themselves be chemically changed by the first SIS cycle) and metal oxyhydroxides that are now incorporated. These incipient nuclei are often expected to exhibit hydroxyl terminated surfaces, in analogy to a propagating metal oxide ALD surface after water dosing, which may be highly reactive with the organometallic precursor provided in the second SIS cycle. Figure 4 shows the infrared (IR) absorbance at 3016 cm⁻¹, after background subtraction, ascribed to the methane reaction product expected during some part of the SIS of In₂O₃ in PMMA using TMIn and water.¹⁶ There is a marked difference in the intensity of this signal between the first and second TMIn exposure (cycles 1a and 2a), which is attributed to the reaction of TMIn with nucleated indium oxyhydroxide. In the first exposure, no methane is generated, as the TMIn is only reversibly complexed/adducted to the carbonyl in the PMMA backbone. In contrast, the methane observed during the second SIS cycle is strong evidence for the irreversible reaction of indium oxyhydroxide nuclei with TMIn, which may occur in addition to further complexation with carbonyls.

Early work in SIS recognized the value of seeding the growth of materials other than Al₂O₃ with a single cycle of TMA/H₂O. The SIS of Al₂O₃ affords aluminum oxyhydroxide seeds with high density and selectivity that are subsequently amenable to reaction with a wide range of organometallic ALD precursors that are freed from the requirement of direct interaction with the polymer phase. Al₂O₃ SIS was used to seed tungsten growth in polymers by the George group in 2008.¹⁰⁰ Tungsten and tungsten oxide, along with SiO₂, have also been grown on Al₂O₃ seeds that were selectively deposited by SIS in the PMMA domains of PS-b-PMMA block copolymers.^12,36 This seeding process leverages the well-studied process space of TMA/H₂O SIS at the cost of significant alumina incorporation in the resulting deposit, which may be deleterious to the functional properties of the final hybrid composite.

SIS can be governed by interactions within the polymer film beyond functional groups in the polymer itself. Films prepared by spin coating, for example, contain residual solvent molecules, which can be retained in the film. The choice of solvent can, therefore, impact SIS if the solvent has a significant Lewis basic character or is otherwise prone to interaction with the intended organometallic precursor. In a study of ZnO SIS in the resist material SU-8, researchers found significant infrared signal of C=O peaks (bonds that are absent from the SU-8 material) from the cyclopentanone solvent used as solvent for spin-coating.³⁵ The presence of this solvent was detected even after baking in excess of typical procedures. The C=O functionality was proposed to bond with diethyl zinc (DEZ) through the reaction scheme shown in Fig. 5.

Generally, these solvent-driven seeding strategies are less controlled than those involving the (self-limited) interactions between polymer functional groups and metal-organic precursors. The nonuniform distribution of solvent molecules with the polymer film or bulk is subject to kinetic control through diffusion that varies strongly with film depth and process conditions, and this can lead to nonuniform SIS.

Beyond the solvents used in spin coating, many polymer samples exposed to ambient humidity will also absorb water. PMMA, for example, can absorb 2% w/w water under ambient conditions,¹⁰¹ while other polymers of interest may be significantly more hygroscopic. The majority of water is expected to be purged from the polymer films during inert gas purging under the low vacuum conditions of a warm ALD/SIS reactor. However, residual water within the film may react directly with many incoming organometallic precursors. Of course, the presence of water within the film can be intentionally exploited in analogy to the spin-coating solvent discussed above, by applying a controlled water predose in situ, as was done to boost the SIS of ZnO by DEZ in PMMA.⁴⁹ As a form of solvent seeding, these SIS reactions are generally less uniform and also influenced by the kinetics of small molecule diffusion in polymers.

Polymers that would be considered inert based on their chemical structure are sometimes observed to support SIS, albeit with much lower growth rates. In recent work by Perego et al., the growth rate of Al₂O₃ SIS in PS thin films was observed to be nearly independent of the purge time between TMA and H₂O exposures.²⁷ This evidence suggests that the dominant mechanism for growth of metal oxide in the PS is not steric entrapment of either reactant vapor within the film, as in such a case purge time dependence would be expected. Rather, reactive defects within the polymer film are proposed to behave as nucleation sites. These reactive defects may in part be attributed to polar groups at the chain ends present in some polymerization chemistries. The ultralow density of these reactive defects would explain the slow growth of Al₂O₃ in inert polymers such as PS. The possible role of chain-end chemistry may be revealed in the future by studying PS of varying molecular weight (M_w). At lower M_w, there would be a higher density of chain ends for a given film thickness, and a higher nucleation rate may be observed. A 2010 paper studied Al₂O₃ SIS in plasma-treated and pristine bulk and thin film polystyrene.¹⁰² They saw growth in both the plasma-treated samples and in the pristine bulk sample (sample taken from a commercial Petri dish), which had pristine oxygen content as measured by X-ray photoelectron spectroscopy (XPS). They also observed much slower nucleation in the pristine thin film polystyrene, which was shown to be essentially oxygen-free by XPS. The accumulation of growth in the “defect-free” thin film polystyrene was ascribed to oxidation caused by repetitive exposure to the TMA and H₂O. The processes by which inorganic deposits nucleate via SIS in such inert polymers require further study.

Diffusion governs the transport of precursor molecules from the free film surface throughout the polymer volume. The fundamental science of small molecule diffusion through polymeric films has a long history in the context of membranes for gas separations.^103–105 The transport of dissolved small molecules in polymers is called “solution-diffusion” and is driven by the chemical potential gradient imposed by the concentration gradient of penetrant precursor molecules in the film. This chemical potential gradient arises from the energetics of the various interactions described above.¹⁰⁶ 

At the simplest level, the transport of the organometallic vapors through the polymer can be modeled as a purely diffusive process. Fick’s second law describes how the concentration of the diffusing species changes in space and time,

where n is concentration (volume density of precursor molecules), t is time, and D represents the diffusivity of the precursor molecule in the polymer defined by the following equation:

where k is Boltzmann’s constant and T is temperature. ΔH_D is the energy barrier for the penetrant precursor molecule to hop from one site to another, and D₀ is the frequency of site-hopping attempts of the penetrant molecule. A recent review by Losego and Leng solves this equation for the most common boundary conditions of a thin film on a gas-impermeable substrate and a fiber, yielding expressions for the spatial and temporal concentration of penetrant molecules.² 

Introducing reversible complex formation between the penetrant molecules and polymer functional groups brings a second level of complexity to modeling the diffusion in SIS. By approximating the polymer as a region with a volumetric density of reactive sites c₀, transport through the polymer can be described as

Here, n is the volume density of precursor molecules, and θ is the fraction of occupied reactive sites and is governed by the reversible complex rates of formation, k⁺, and breaking, k⁻,

Combining these two equations yields the following expression for transport of dissolved molecules within the polymer:

This equation is the diffusive equivalent of transport in a chromatography column, where reversible absorption/desorption slows the transport of molecules through the medium. If the concentration of available sites is in local equilibrium with the local density and the solubility of the penetrant molecule is low, the fraction of occupied sites can be approximated as

so that

In the dilute limit, the transport process can be well-described by conventional diffusion with an effective diffusivity term depending on the ratio of the reversible complex formation/breaking rates. These effects can be captured in a reaction-limited diffusivity D^* defined by the following expression:

Beyond the dilute approximation, the quasiequilibrium approximation leads to a diffusivity that itself depends on the fraction of the occupied reactive sites within the polymer. In each of these cases, the approximations yield identical Fickian diffusion profiles, just with diffusivity values reduced by the strength of reversible interactions. Further extending these concepts to irreversible reactions between precursors and functional groups can begin with approximations such as those developed for reactive transport in porous materials. These ideas are thoroughly developed in the following references: Refs. 107 and 108.

The influence of complexation energy and formation/breaking rates on diffusivity is well-illustrated by the case of Al₂O₃ SIS vs In₂O₃/Ga₂O₃ SIS. Trimethyl aluminum, trimethyl indium, and trimethyl gallium (TMA, TMIn, TMGa) were used along with water for these three materials respectively, such that the ligands were identical and only the metal varied. These three precursors were found to form isostructural complexes with the carbonyl moiety of PMMA, but the energy of the TMA interaction was calculated to be ∼3× greater than either of the other two materials.¹⁶ As a consequence, the depth of In₂O₃ infiltration in a thick PMMA film under similar conditions is significantly deeper than Al₂O₃, highlighting how the design of precursor/polymer energetics influences SIS.

Finally, the long residence time of precursor molecules within the polymer raises the possibility of accessing a regime where both coreactants do not mix in the gas phase but are simultaneously present within the polymer. Effectively speaking, this describes a chemical vapor deposition (CVD)-like growth within the polymer itself. Being CVD-like, the reaction takes place wholly between the two molecules without the polymer playing a role, other than enabling the mixing to take place within the polymer and not the vapor phase.

The diffusion of small molecules in polymers has been extensively studied in the literature. For instance, Barrer carried out systematic studies on various polymers involving gases such as hydrogen, oxygen, and carbon dioxide, establishing correlations between the size of the molecule and the activation energy of the diffusion process. In the case of water, a more relevant example in the context of SIS, diffusivities of 1.7 × 10⁻⁹ cm²/s at 60 °C and 5 × 10⁻¹⁰ cm²/s at 45 °C have been reported for Kapton polyimide and polyacrylonitrile (PAN) with a reported activation energy of 5.4 kcal/g mol for the Kapton polyimide case.^109,110 Beyond water, though, there is a scarcity of data for traditional ALD/SIS precursors. However, if we take a value of 10⁻⁹ cm²/s as a reference and we consider a micrometer-thick polymer layer, the resulting characteristic diffusion time (in the absence of strong interactions or reactions with the polymer) is on the order of 10 s. Calculating the diffusivity of SIS precursors in different polymer systems is an extremely useful endeavor. A comprehensive study of the diffusivity of TMA within PMMA was conducted by the Perego group using in situ spectroscopic ellipsometry (SE),²⁷ a technique described in Sec. IV A 2. At 90 °C for a 14 kg/mol PMMA polymer, the diffusivity of TMA was calculated to be 1.2 × 10⁻¹² cm²/s. The three order of magnitude difference from the case of ∼10⁻⁹ cm²/s for water highlights the strong role of noncovalent interactions in SIS precursor transport within polymers.

A number of material parameters affect the diffusivity of SIS precursors within polymers independent of the thermodynamics of a given precursor/polymer interaction, as outlined below. It is important to emphasize that all the models above ignore the impact that inorganic clusters growing inside the polymer as part of the SIS can have on both the solubility and the diffusivity of the polymer. The idea that, after a number of SIS cycles, precursor transport within the polymer is greatly impeded agrees well with the experimental observation of dense crusts forming near the surface of polymers after a sufficiently large number of SIS cycles.

The polymer free volume arises from entropic interactions in the entangled chains of polymers, which can change with the material’s processing history. The polymer chain volume combined with the free volume yields the specific volume (volume/mass), which is the inverse of density. Starting with a glassy polymer in which segmental motion of the polymer chains relative to each other is suppressed, heating causes the volume of the polymer chain to increase due to thermal expansion. The free volume does not increase with temperature until the glass transition temperature (T_g), at which point the free volume fraction v_f0 (free volume/specific volume) obeys the following equation:¹¹¹ 

where Δα is the difference in the thermal expansion coefficient above and below T_g. By raising the temperature of the film above T_g and rapidly quenching it, some of the excess free volume from the rubbery state can be frozen in, such that the polymer is less dense than it would be at slower cooling rates. These thermal treatments can, in principle, enhance the diffusivity of precursors in polymers by increasing the number of hopping pathways along which they have to diffuse. More detailed theory of solvated molecule diffusion as a function of glassy/rubbery state can be found in the following review: Ref. 112.

The polymer free volume can be locally perturbed in the interphase regions near the substrate and near the free surface, which can affect the diffusivity of metal-organic compounds in these regions.^113,114 In many systems, this effect is small. For example, in PMMA, the vacuum-polymer interphase region that deviates from bulk PMMA was measured to be ∼2 nm.¹¹⁵ 

Developments in polymer synthesis have opened up a massive field of possibilities in polymer chain architecture. Copolymers, bottlebrush polymers, star polymers, and more are possible. There is also a vast range of monomers available, and modifications to monomer groups accessible, with modifications that may directly alter interactions with SIS precursors. More indirect modifications may perturb the free volume of the polymer with implications for SIS precursor diffusivity.

By controlling the volume fraction of methyl methacrylate (MMA) in a PS-r-PMMA random copolymer, the density of reactive sites relative to inert PS sites can be tuned. In a study of such a system, Caligiore et al. found that the diffusion coefficient of TMA through this copolymer was modestly increased over that measured for pure PMMA, at MMA fractions exceeding 10%.²⁹ The SIS deposition rate of Al₂O₃ was linear with the MMA fraction, as expected, and the porosity and domain size of the annealed metal oxide did not apparently differ across this wide range of MMA fractions. This study thus shows that the density of reactive sites in a polymer can modestly impact the diffusivity of SIS precursors.

In another example, the influence of polymer free volume on the kinetics of precursor transport through thin films was concisely studied by comparing poly-n-methacrylate polymers.²⁰ This series of polymers consists of alkyl chains of increasing length on the ester bond of the molecule. The presence of this alkyl group lowers the glass transition temperature of the molecule by disrupting entanglement of the polymer.¹¹⁶ Thus, at a given reaction temperature, the longest chain polybutyl methacrylate polymer has the largest free volume of the group. Using quartz crystal microgravimetry (QCM), the Jur group found that polymers with increasing free volume elements (longer pendant chains and lower T_g) allow more rapid diffusion of TMA under equivalent SIS conditions (Table III). However, sufficiently long pendant alkyl groups can themselves crystallize. Crystallization of polymer components has been observed to change the distribution of metal oxide products during SIS, likely due to changes in the diffusion pathways within the film.^52,53 Therefore, more studies are needed to generalize trends based on such modifications to chain architecture.

While the poly-n-methyl methacrylate isothermal study provides a clear comparison of analogous polymers with disparate T_g, more care must be taken when ascribing changes in SIS behavior across temperatures to T_g. Infiltrated molecules can plasticize polymers, promote chain motion, and lower the effective T_g from what is measured for the polymer by ex situ. The degree to which this plasticization affects diffusivity must be considered on a case-by-case and cycle-by-cycle basis. Temperature simultaneously affects functional group reactivity at the same time as it increases chain motion. Since the true glassy/rubbery transition of polymers under infiltration is not generally known, care must be taken in ascribing temperature-dependent trends in SIS studies to particular phenomena.

The molecular weight (M_w) of the polymer will also influence the diffusive environment of vapor precursors due to the variable volume density of polymer chain ends. A polymer with a larger M_w will have fewer chain ends, which have a larger configurational free volume than monomer units within the chain itself. The glass transition temperature of PMMA, for example, increases as a function of M_w due to the influence of chain-end density.¹¹⁷ This is true across polymers and is described by Flory-Fox theory.¹¹⁸ A higher density of chain ends leads to more diffusive pathways for vapor precursors, and therefore, SIS in the low-diffusion time limit is expected to yield metal or metal oxide at greater depths in a lower M_w polymer film. The diffusivity of TMA in homopolymer PMMA at 90 °C was calculated to be 3.75 times greater when the molecular weight was decreased from 14 to 3.9 kg/mol.²⁷ M_w can also affect the density of the resulting SIS product. Al₂O₃ grown in 200 nm 15 kg/mol PMMA thin films by SIS at 90 °C had an average effective refractive index (n_eff) of 1.396 (porosity 26%), whereas 350 kg/mol PMMA produced an n_eff of 1.374 (porosity = 29%).⁸² This effect is subtle—an order of magnitude difference in molecular weight yields only a 3% change in porosity of the resulting metal oxide product. Nevertheless, this highlights how engineering of the polymer free volume environment and the resulting diffusivity of the precursor through the polymer can influence the properties of the resulting hybrid material. The M_w of the polymer also dictates the density of chain ends, which can have different chemistries as discussed previously. We note, however, that the interaction between organometallic precursors and polymer functional groups affords tunability that is orders of magnitude larger than the reported polymer molecular weight effects alone.

The sensitivity of diffusivity to precursor-polymer energies, along with the nanoscale chemical templates of block copolymers, will enable the precise control of multiple materials to form heterostructures. The first report, by the Segal-Peretz group, demonstrating this concept achieved a hexagonally closed-packed array of ZnO nanowires with Al₂O₃ tips using a PS-b-PMMA template.¹⁷ This was achieved using brief exposures of TMA with a long purge, followed by longer exposures of DEZ with a briefer purge, and completed with a H₂O exposure. Because the diffusivity of TMA is exceptionally slow, the vertical depth to which the PMMA domains were infiltrated was limited. Long purges following the TMA enabled much of this infiltrated material to desorb. DEZ can bypass the complexed TMA-MMA zones by way of the PS domains, which the SIS precursors³⁴ do not. By exploiting the differing thermodynamics and kinetics of multiple precursors within a polymer, a variety of heterostructures ranging from linear junctions to core-shell structures may be realized.

The range of materials that have been deposited by ALD is vast and spans much of the periodic table.^119,120,137 ALD processes have been developed for the deposition of pure elements, oxides, nitrides, carbides, sulfides, and more. The library of materials grown by SIS to date is, by contrast, far more limited. The periodic tables for ALD and SIS in Fig. 6 highlight this disparity. The current published library of SIS materials includes Al₂O₃, ZnO, TiO₂, In₂O₃, Ga₂O₃, VO_x, SnO_x, SiO₂^*, W/WO_x^*, Sn^**, and Mo^**, where (^*) indicates that the material was grown on initial Al₂O₃ nuclei grown by TMA/H₂O and (^**) indicates that the metal complexes directly and irreversibly binds to the polymer with no “B” reactant. Although some reports describe the SIS oxide with the standard thermodynamically stable chemical formula, in other reports, a general MO_x nomenclature is used, as the oxidation state of the metal may not be definitively known. Here, we use the naming from the original reports. SIS of Al₂O₃ in PMMA from TMA and H₂O is the most well-studied SIS system by far. ZnO and TiO₂ are the two next highest in publication frequency with other SIS material processes restricted to a handful of reports. SIS SiO₂ and W/WO_x have only been achieved with a seed cycle of Al₂O₃, as previously mentioned.^12,36 While this seeding method is useful in providing reactive nuclei for subsequent cycles, the resulting materials will have some Al₂O₃ content, which may impact functional properties. The development of processes for the direct SIS of functional materials is also desirable.¹⁶ A thorough understanding of the chemical physics involved in the varied aspects of SIS will be required as researchers attempt to fill out more of the periodic table in pursuit of novel functional hybrid materials.

The present SIS periodic table is primarily constrained by the limitations of polymeric substrates compared to typical inorganic ALD substrates. Many polymers are not thermally stable in excess of 200 °C and, therefore, cannot survive the high temperatures required for some ALD precursors. Metal precursors must be able to dissolve into the polymer and have a reasonable diffusivity in order to enable SIS and be able to coordinate with functional groups on the polymer. With these considerations in mind, certain classes of ALD chemistries can be evaluated as more challenging:

Platinum group metals and oxides: These materials typically require temperatures in excess of 300 °C and O₂ as a coreactant.^121,122 One issue with highly oxidizing precursors at higher temperatures may be oxidative damage to inert copolymer domains leading to a lack of selectivity. However, Pt processes as low as room temperature have been demonstrated with more complex coreactant schemes.¹²³ 

Transition metals: The most common ALD recipes for materials such as Fe and Ni use metal amidinate precursors and H₂ as a coreactant. These processes require similarly high temperatures in excess of 300 °C and are poor candidates for SIS.

Metal nitrides: These materials require ammonia as a precursor and typically very high temperatures in excess of 400 °C and H₂ or plasmas as coreactants.

Other classes of materials that can be considered more viable include the following:

Metals using metal hexafluorides: Currently, the only report of direct metal SIS is tungsten using WF₆ and silane/disilane.^12,36 Mo and Nb are two other common metals deposited with the same basic chemistry, which are amenable to low-temperature processing. One potential issue is the oxidation of these metals during the polymer removal step.

Metal sulfides: Sulfide ALD chemistry is similar to oxide ALD chemistry and is amenable to similarly low temperatures. These materials are a potentially productive route for SIS materials development. By replacing H₂O with H₂S, the existing periodic table of oxides may be duplicated in sulfides.

Metal fluorides: Similar to the metal sulfides, metal fluorides may be deposited by ALD with appropriate metal precursors and either HF or HF-pyridine vapors. A variety of ALD metal fluorides have been grown at 150 °C,¹²⁴ a temperature compatible with many polymers.

If the parameter space of the ALD periodic table is formidable, filling out the SIS periodic table will be an even more daunting challenge as processes may be developed along a tremendous number of design axes. The parameter space in SIS process design extends along at least the dimensions described in Fig. 7.

The choice of metal-organic precursor, co-reacting precursor, and polymer functional group is the primary means to control the energetics of the interactions. A recent paper from the Segal-Peretz group demonstrates the first tin oxide SIS recipe using the precursor tetrakis(dimethylamido)tin(iv) (TDMASn) with a hydrogen peroxide coreactant in P2VP polymers.³⁷ The tetrakis(dimethylamido) ligands were previously conjectured² to be too bulky to permit substantial diffusion for SIS. By switching from the carbonyl in PMMA to the pyridine in P2VP, researchers observed significant entrapment of the TDMASn and conversion to tin oxide with the more highly oxidizing hydrogen peroxide. The absence of the SIS reaction in PMMA with this precursor, in addition to the need for a stronger “B” component, highlights the need for compatibility across components along with the appropriate process temperature to achieve new SIS materials.

Within a given set of precursors and functional groups, advances in polymer synthesis open a wide variety of structural modifications to the polymer. Active and inactive groups can be combined in copolymers, in blocks, random, or gradient structures, to impose mesostructured or unstructured spacing between interactive groups. Novel polymer architectures may be able to control the mean distance between SIS nuclei in unanticipated ways. The duration of vapor exposure and the length of purge between coreactant exposures offer routes to control the distribution of the inorganic product throughout the sample. The temporal changes to the organic/inorganic film as the SIS process proceeds provides yet another axis upon which SIS can be manipulated. The first cycle introduces inorganic nuclei into the bulk, which present distinct reactive sites for subsequent cycles. The inclusion of this material also changes the free-volume pathways and the diffusivity of subsequent exposures. Thus, changing the first cycle determines how all other cycles progress. By changing the purge time in the first cycle, as an example, the density and depth of infiltration of the seeds can be changed. Finally, the co-reactant compatibility with the polymer template must be considered as an additional challenge. For example, transient and/or reactive coreactants including ozone and nearly all plasmas may not persist deep into thick polymer films or may alter the polymer backbone and/or functional groups.

Experimental characterization of SIS chemistry and physics is essential to develop a firm theoretical framework for the various phenomena at play in order to improve our intuition and predictive capacity while exploring this vast SIS variable space.² Researchers have used a number of experimental techniques in concert to span the disparate length and time scales involved in SIS. Examples of pertinent data acquired: the depth and density gradient of an inorganic deposit from the film surface to substrate, the kinetics of vapor-phase diffusion through the polymers, the strength of intermolecular attractions and identity of participating polymer functional groups, the bonding between organic and inorganic components, and the structural evolution of the deposited phases across length scales. In situ methods are preferable in many cases as they capture the dynamics of SIS processes and conditions during half-cycles that are incompatible with ambient environments. However, ex situ techniques also provide valuable information for understanding the SIS process. These ex situ methods include spectroscopies, microscopies, and reciprocal-space scattering methods. Reciprocal-space methods probe structural correlations across a wide range of length scales, as summarized in Fig. 8. Some reciprocal-space techniques, which are most simply implemented ex situ, may also be conducted in dynamic in situ fashion with integrated experimental apparatus such as those described in Ref. 125. In this section, the most relevant SIS characterization techniques are described in the context of the types of information about SIS that they provide. Existing approaches that have yet to be substantially applied to SIS science may fill gaps remaining in our present understanding of SIS. We highlight the potential of several of these methods to motivate future SIS studies.

Fourier transform infrared (FTIR) spectroscopy offers a unique window into the functional groups and interactions of chemical processes during SIS. FTIR spectroscopy can measure the characteristic infrared absorbance of polymer functional groups, residual solvents and impurities, SIS precursors, adducts, and byproducts as well as bulk modes of the resulting inorganic species. Commercial FTIR spectrometers may interface with commercial and custom SIS chambers and can operate either in the transmission mode or reflection mode. In the transmission mode, the polymer thin film must be cast on a relatively IR-transparent substrate, such as an undoped Si wafer. In the reflection mode, a relatively smooth and highly IR-reflective metallic surface such as an evaporated gold film on a Si wafer is ideal. In either case, reference to a bare substrate is preferred for complete calibration. Reference to bare substrates upon reactant exposure further enables straightforward identification of the IR absorbances of SIS precursors in the headspace during extended exposures. Some SIS/FTIR setups use gate valves that are shut during precursor exposures to prevent growth on the IR-transparent windows. However, this approach limits the temporal resolution of kinetic information and precludes the measurement of vapor-phase SIS precursors and byproducts. An alternative method to prevent growth on the windows is to purge the volume between the SIS substrate and the windows with an inert gas to create a diffusion barrier.

FTIR spectra are used to track identifiable absorption features that may appear, disappear, or shift during SIS processing. New absorption features indicate ligands associated with the vapor-phase or infiltrated precursor, inorganic incorporation, or byproduct formation. It is often convenient to plot the FTIR data as difference spectra by subtracting the previous spectrum (i.e., the spectrum recorded prior to the precursor exposure) from the current spectrum. In this way, negative features signal the consumption of species and positive features reveal the creation of new species. Permanent negative features may correspond to functional groups that are consumed in the course of irreversible chemical reactions between the precursor and polymer. Transient peak shifts and negative features (i.e., bleaches) may indicate reversible interactions between organometallic precursors and polymer functional groups.

One example of such a peak shift is the perturbation of the carbonyl in PMMA due to interaction with TMA.^{16,21–24,28} Figure 9(a) shows this peak shift from 1729 cm⁻¹ to 1670 cm⁻¹, which is attributed to the Lewis acidic TMA drawing electron density away from double-bonded oxygen of the carbonyl. Both the intensity and magnitude of the peak shift reveal information on the SIS process. By tracking the intensity of the shifted peak over the course of precursor exposure, information on the kinetics of infiltration may be inferred.²³ The ratio of absorbance of the shifted peak and the original peak measures the fraction of functional sites that participate in adduct formation. In sufficiently thin films and for sufficiently long exposures to ensure a homogeneous distribution of the precursor, this fraction provides insight into the equilibrium constants of the complex. If the density of functional groups in the polymer is known, then the density of complexed TMA can be further quantified. The change in wavenumber or magnitude of the peak shift provides some information into the strength of the complex formed between a functional group and the precursor. The stronger the complex, the more red-shifted the perturbed peak may be. Since the strength of the interaction between the precursor and polymer functional group plays a significant role in the diffusivity of the precursor through the film, the magnitude of peak shift can provide insight into optimal processing temperatures and exposure times. For example, in a recent comparative study of the group 13 metal alkyl precursors TMA, trimethyl indium (TMIn), and trimethyl gallium (TMGa), the C=O peak was ∼20 cm⁻¹ further red-shifted during exposure to TMA vs TMIn or TMGa, suggesting a stronger adduction.¹⁶ Density functional theory simulation, discussed in more detail below, suggested that the enthalpy of TMA-MMA association was ∼3× more negative than the corresponding adduction with TMIn or TMGa. This was correlated with significantly slower diffusion kinetics for TMA into PMMA films as measured by the time-resolved PMMA C=O bleach and recovery upon purging.

The same approach can be applied for new peaks that appear upon exposure, which correspond to precursor infiltration, such as the two associated with TMA in polyethersulfone films in Figs. 9(b) and 9(c). The integrated Al-CH₃ rock and CH₃ stretch belonging to TMA within the film reveal vastly quicker kinetics for infiltration than for purging,⁵⁸ as is also the case in PMMA.^23–25 Looking forward, detailed time-resolved comparisons of the kinetics of precursor peaks and shifted peaks due to precursor complexes may inform the complex balance of diffusing solvated precursor, complexed precursor, and irreversibly bonded precursor in a single experiment.

In situ FTIR can also be used to track irreversible chemical changes made to the infiltrated polymer matrix by the SIS precursors. If the precursor directly reacts with a functional group, the spectral features associated with that group would be significantly altered or disappear entirely. This appears to occur in the case of Al₂O₃ SIS into polyethylene oxide (PEO) (Fig. 10). When TMA is dosed, the ether signature at 1100 cm⁻¹ disappears and does not recover after exposure to water.²² This suggests that TMA can cause polymer chain scission by breaking the C—O—C backbone of PEO.

Spectroscopic ellipsometry (SE) is a nondestructive optical measurement by which changes in refractive index and film thickness film thickness (Ångstroms to hundreds of nanometers in range) can be deduced from fitting to an optical model. SE relies on wavelength-dependent changes in phase angle and intensity of circularly polarized light as it impinges on a sample. More detailed modeling can be performed if the measurement is repeated at multiple incident angles [i.e., variable angle SE (VASE)]. These data are compared to the simulated values derived from a model optical stack that includes the thickness, complex refractive index, and optionally the roughness and porosity of each layer in the stack. SE has been widely applied to ALD studies on planar solid samples and can be readily integrated into SIS reactors by means of optically transparent windows.^126,127 As discussed above for FTIR, it can be convenient to isolate these windows using gate valves or an inert gas purge.

In situ SE enables evaluation of the hybrid film throughout the SIS process. In the example of Al₂O₃ SIS in the membrane-forming polymer polyethersulfone at 110 °C (Fig. 11), the kinetics of swelling/deswelling of the film during TMA exposure and purging (I and II) are observed. The film swells and the refractive index drops during exposure as TMA molecules diffuse into the polymer film. The inverse process happens during purging, though at the end of II, the film has still not recovered its initial thickness or refractive index. Figure 11 reveals that infiltration is much faster than deswelling. The film is exposed to TMA once again in III with a brief purge in IV. The water exposure in V shows a much more rapid change. The thickness drops slightly, but there is a substantial drop in refractive index, indicative of the rapid diffusivity and reactivity of water with TMA to form Al₂O_3. The decrease in refractive index upon H₂O exposure is counterintuitive, given the relatively high refractive index of 1.65 for ALD Al₂O₃, but may relate to the formation of aluminum hydroxide, Al(OH)₃, which has a lower refractive index of 1.57. Figure 12 highlights the role of purge time in the continuity of the metal oxide deposit upon polymer removal. As purge time increases, the density of TMA molecules in the film decreases upon the introduction of water. When the continuous PMMA phase is removed, the metal oxide shows an apparent porosity with increasing purge time.

In situ SE is a direct complement to in situ FTIR as they probe physical and chemical changes to the polymer film, respectively. Whereas in situ FTIR only tracks the evolution of solvated precursor that interacts with chemical functional groups, in situ SE may track the total solvated precursor content in the film including the precursor that is not interacting with functional groups. These two in situ techniques can be used together to provide a fuller picture of the kinetics of precursor transport and reactions with polymers.

Quartz crystal microgravimetry (QCM) can be used to infer the mass of a film affixed to an oscillating quartz crystal by measuring the change in resonant frequency of the crystal. These frequency measurements are sufficiently sensitive to detect the deposition of submonolayers during ALD. In SIS, QCM can be used to determine the kinetics of vapor infiltration and purging by measuring the mass of the affixed polymer film over time. Figure 13 displays QCM data recorded during repeated precursor pulses of a single TMA exposure into 50-nm PMMA films at different temperatures.²⁵ The mass changes induced by these TMA pulses are shown in Fig. 13(b) and reveal that the mass gain per pulse is maximized at 100 °C. The increase in mass uptake between 70 °C and 100 °C can be attributed in part to the increase in solubility of TMA with increased temperature [Eq. (1)], along with the transition from a glassy state to a rubbery state. In other QCM experiments from the same paper (not shown), the mass loss after 85 min of purging at 70 °C was 60% from the exposure peak, whereas at 150 °C, the analogous mass loss was only 30%. While the 150 °C loss occurred over 15 min of purging, the mass loss from the 70 °C sample showed no sign of plateauing over the observed measurement. This is consistent with one-component irreversible incorporation of TMA in PMMA at high temperatures mentioned above. The long and slow purging of TMA at low temperatures is likewise consistent with the slow breaking of the reversible carbonyl/TMA complex as time-resolved through FTIR studies.^23,24 QCM provides gravimetric evidence of physical changes to the film during SIS, similar to how SE provides optical evidence of such physical changes. As such, both techniques complement the chemical information derived from FTIR spectroscopy.

Quadrupole mass spectroscopy (QMS) of product gases is a common method for identifying the chemical byproducts produced during ALD.^128,129 To the best of our knowledge, QMS has only been used to study SIS processes in Ref. 45. The authors measured the methane generated by TMA SIS in polyurethane foams over successive cycles before water exposure and observed an increase in the methane product. This is consistent with the picture of methane generation after the first half-step occurring on existing —OH nuclei but not occurring through the reversible complex formed with polymer functional groups. QMS could be used to identify the product gases and resolve open questions in the chemical evolution of metal-organic precursors within polymer films. For example, a recent combined FTIR/ab initio density functional theory study showed that at sufficiently high temperatures (T > 150 °C), TMA in PMMA will react via a pericyclic reaction with the O of the carbonyl to directly form Al—O linkages and release ethane as a gaseous byproduct.²⁸ In contrast, ALD reactions of TMA on surface hydroxyls are known to produce exclusively methane. QMS of the gases downstream of the reaction chamber could, in future studies, complement FTIR to confirm the existence of ethane and further validate the chemical mechanisms. QMS may also be used to determine which chemical processes take place in subsequent SIS cycles in which Al—O and potentially hydroxyls are present within the film. If ethane is confirmed to be a reaction product for direct incorporation at high temperature, the presence/absence of ethane in exposures after the first cycle will further reveal the chemical environment of reactions in mixed polymer/inorganic samples.

Cross-sectional electron microscopy offers direct visualization of SIS samples from the substrate to surface. This technique involves fracturing a sample and mounting it orthogonally to the electron beam. SIS samples can be fractured by hand, but this may introduce artifacts to the cross-sectional surface due to plastic deformation of the sample. Submerging the sample in liquid nitrogen while fracturing can promote brittle fracture of the sample by being below its glass transition. Techniques such as focused ion beam (FIB) milling⁶⁵ can achieve a cleaner surface, but these are time-consuming and require specialized equipment. Charging effects and electron beam damage further complicate the imaging of SIS samples, particularly before polymer removal.

Scanning electron microscopy (SEM) yields both topographic information of the cross-sectioned surface and atomic number contrast (Z-contrast). Both of these make direct visualization of inorganic material incorporation into the polymer matrix possible, as shown in Fig. 14. The Z-contrast can be enhanced using a backscattered electron detector. Multiple papers have reported the use of electron-beam based spectroscopy such as energy dispersive spectroscopy (EDS), which relies on elementally characteristic x-rays emitted following electron bombardment to spatially resolve the metal distribution in SIS thin films.^26,80,81,130 These cross-sectional EDS measurements can be difficult to quantify as the interaction volume of the incident electrons can be hundreds of nanometers to micrometers, often larger than the film thickness itself. Truly quantitative cross-sectional EDS may be realized in transmission electron microscopy (TEM) with ultrathin samples prepared by microtome or FIB techniques, as the thin samples reduce the interaction volume enabling nanometer spatial resolution.

Top-surface metrology methods such SEM [and atomic force microscopy (AFM), discussed below] can only reveal two dimensions of the three-dimensional distribution of hybrid material that result from SIS. Extending to the third dimension via transmission electron microscope (TEM) tomography can reveal depth-dependent details about SIS. Tomography involves the acquisition of many TEM images of a thin film at different angles, followed by software reconstruction of those slices into a three-dimensional object. However, the sample must be prepared on a specialized TEM-transparent substrate such as a lithographically prepared silicon nitride membrane.¹³¹ In Fig. 15(a), the blue regions represent Al₂O₃-infiltrated PMMA domains in a lamella-forming PS-b-PMMA thin film. The boundary between the two domains is drawn at sharp changes in contrast. This method allows for the identification of BCP defects in the bulk of the film (i.e., below the film surface), such as the lamellar junction highlighted in Figs. 15(b)–15(d). In this example, the SIS Al₂O₃ acts as a contrast agent to improve image quality in TEM tomography for the identification of buried self-assembly defects. However, we envision that TEM tomography can be used in the future as a real-space imaging modality to measure, for example, density gradients that emerge from diffusion-limited infiltration in SIS.

Atomic force microscopy (AFM) can reveal detailed topographic and nanomechanical information of SIS sample surfaces. This is of particular interest in the study of SIS in phase-separated block copolymers. In such systems, where SIS is restricted to one phase, the impact of inorganic material incorporation in confined domains relative to bulk homopolymers is of interest. In Fig. 16, AFM was used to assess the change in topography and Young’s modulus of PMMA cylinders constrained by a PS matrix. The researchers found an increase in modulus and in the variance of the modulus, with successive Al₂O₃ SIS cycles. The mechanical properties of the PS matrix were relatively unchanged.

Time-of-flight secondary ion mass spectroscopy (TOF-SIMS) can be used to analyze the composition of SIS samples. This technique involves the pulsed bombardment of the sample surface with an ion beam and the analysis of the ejected secondary ions of the sample by mass spectroscopy. By calibrating the etch depth of each ion bombardment, the concentration of the different materials within the sample can be determined throughout the thickness of the sample. This can yield information about the extent of the SIS modification and thus indirectly informs the diffusivity of the metal-organic precursor. An example of this technique applied to Al₂O₃ SIS in PMMA is shown in Fig. 17.

X-ray photoelectron spectroscopy (XPS) is a technique that can quantitatively determine the composition and electronic and chemical states of atoms at the surface of a sample. XPS involves the irradiation of the sample with an x-ray source and the energy-resolved collection of electrons emitted by the sample. The energy of the emitted electrons reveals information about the chemical linkages and electronic structure, and their intensities reveal information on the elemental composition and ratios of different chemical bonds. When applied to SIS, XPS can reveal how the nucleation and growth of inorganic materials affect the polymeric chemical structure in the hybrid film. Figure 18 shows one such example. After annealing, XPS can be used to assess the degree of hydroxylation (—OH surface termini) as a function of SIS parameters and annealing conditions and can also determine how well carbon is removed by the annealing process. XPS can be used in conjunction with ion-beam etching to depth-profile SIS-processed samples. However, care must be taken in the analysis of XPS depth-profiling data as sputtering rates are element-specific and this can create artifacts.

While XPS can provide information on surface chemical linkages, X-ray absorption spectroscopy (XAS) techniques can provide information on the coordination environment of specific elements over the bulk SIS sample. XAS is usually performed in synchrotron facilities, as it involves tuning the X-ray energy over a specific atomic absorption edge, which allows core electrons to be excited. Typically, XAS analysis is broken down into two separate areas based on the energy range above the absorption threshold: X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS). These are collectively termed X-ray absorption fine structure (XAFS) analyses. For SIS of metal oxides, XAFS can directly probe the oxygen coordination environment of the infiltrated metal cation within the polymer itself.

One such example is shown in Fig. 19, where we compare the Zn-K EXAFS obtained from one SIS cycle of ZnO in thermoplastic polyurethane (TPU) using DEZ and H₂O. In Fig. 19, we compare two processing conditions that only differ on the purge time between DEZ and H₂O: a one-minute purge time leads to an EXAFS that is consistent with those of both the ALD ZnO and the ZnO powder references included in Fig. 19. In contrast, the six-minute purge time sample has a markedly different spectrum. The oscillations observed in Fig. 19 come from the impact that the chemical environment around Zn atoms has on the wave function of the free electron. If we carry out the Fourier transform, we obtain information on such coordination environment. While the fitting of the spectra is beyond the scope of this perspective, the comparison between the 1-min and 6-min purge times showed that the 1-min purge time yielded a ZnO product with coordination similar to the bulk wurtzite ZnO structure. In contrast, a 6-min purge time yielded substoichiometric contributions from the intermetal Zn—Zn linkage. In particular, the Zn—Zn signal of the second coordination sphere is suppressed, suggesting that the deposit is composed of small zinc clusters, rather than stoichiometric ZnO nanoparticles.

Purge time is hypothesized to have such a dramatic effect on the zinc coordination environment through tuning the concentration of the DEZ precursor in the polymer. If water is introduced into the system after a brief purge, there is a relatively high concentration of DEZ in the polymer. The coexistence of the two components solvated in the polymer allows for a CVD-like reaction, generating nanoparticlelike ZnO with the stoichiometric wurtzite structure. As the purge time is increased, the density of solvated DEZ decreases due to diffusive loss to the gas phase and vacuum pump. While the change in Zn coordination may be caused by a simple decrease in connectivity of the clusters due to the lower density of DEZ, it could be that the retained DEZ is reversibly or covalently linked to the polymer chain, changing the chemistry that occurs upon reaction to water.

Prior studies on the coordination environment of the ZnO clusters infiltrated into polymer films have shown significant deviations in the second coordination sphere with respect to the expected wurtzite structure, indicating the formation of small clusters composed of at most a few Zn atoms. Moreover, this change in structure correlates with changes in the electronic properties of the inorganic phase as measured using photoluminescence.⁴⁹ EXAFS thus enables to study the local electronic structure of metals/metal oxides within polymer networks and the important roles that SIS processing parameters can have on the structure and properties of materials.

Grazing incidence small angle x-ray scattering (GISAXS) allows for the identification of periodic or semiperiodic structures, such as the phase-separated morphologies of block copolymers, in thin films.¹³² In the context of SIS, GISAXS can be used to determine how confined polymer blocks—both SIS-amenable and SIS-inert—are changed by the introduction of inorganic materials. For example, in Fig. 20, GISAXS data from a symmetric lamella-forming PS-b-PMMA film are displayed in the pristine phase-separated state and after five cycles of Al₂O₃ SIS.¹⁴ In the pristine film, where the two blocks are perfectly symmetrical, reciprocal space peaks associated with a 27.6 nm pitch are observed. A Bragg rod predicted for this pitch at q_y = 0.0556 is missing, which is expected, as it is forbidden by the symmetry of the block copolymer. After SIS, this missing peak appears due to the swelling of the PMMA phase by Al₂O₃ growth, which breaks the symmetry of the equally sized polymer blocks. The contrast of the scattered signal increases dramatically due to the greater Z-contrast provided by the selective Al₂O₃ deposition. The pitch of the block copolymer is unchanged, as evidenced by the lack of movement in the q-space features, which implies that some compression of the PS has occurred. Finally, diffuse scattering at higher q values appears, which is related to nanometer-scale Al₂O₃ domains that are introduced into the PMMA.

From the two static GISAXS snapshots in this example, information was extracted about the selectivity of the SIS process, the deformation of the polymer after SIS, and the structure of the deposited hybrid material. GISAXS and other SAXS data have been published on a number of polymer-precursor pairings.^36,54,57 At the time of writing, only examples of static ex situ GISAXS data have been published. However, it is, in principle, possible to interface a GISAXS experiment to an SIS chamber via x-ray transparent windows. With in situ GISAXS, the temporal resolution of a measurement would be limited only by the time needed to acquire sufficient signal for reliable data, along with considerations of damage prolonged x-ray beam exposure can have on samples. With high-intensity light sources, in situ GISAXS would enable dynamic measurements of the mesostructural evolution of block copolymers throughout the SIS process during precursor diffusion, purging, and inorganic material formation from one to many cycles.

Pair distribution function (PDF) analysis is another X-ray scattering technique that considers both the Bragg and diffuse contributions to X-ray scattering to extract information about the short-range and medium-range microstructure of materials even in the absence of long-range order.¹³³ In contrast to X-ray absorption, which provides information on the coordination environment of a specific element, a PDF analysis of a high-energy X-ray diffraction pattern extracts the distribution of distances between any and all pairs of atoms in the sample. The analysis of low-Z materials such as polymers using this technique has been well established in the literature, with the peaks at low-r values reflecting the interatomic distances within monomers and the medium range providing information on the crystallinity of the polymer.¹³⁴ Moreover, the combination of PDF with reverse Monte Carlo techniques provides a way of reconstructing the short- and medium-range order in such materials.¹³⁵ 

A key challenge for the direct application of PDF to SIS is the simultaneous presence of an inorganic phase embedded into the polymer film. While the acquisition of PDF data from such samples is straightforward, the challenge is how to uncouple contributions from both polymer and inorganic deposit during the PDF analysis. However, this problem is greatly simplified through the development of reaction chambers with in situ high-energy X-ray measurement capabilities.¹²⁵ In situ PDF would enable probing the impact of the dissolution of the SIS precursors on the polymer structure, including the dynamics of absorption and desorption during purge times. More importantly, it would facilitate the isolation of the contribution of the inorganic clusters from the polymer background. PDF analyses can conceivably inform how solvated precursors alter the polymeric microstructure and resolve the distribution of metal-oxygen distances within amorphous deposits.

Research into the fundamental processes governing SIS will increasingly require the complementary information that the discussed characterization techniques offer. Combining in situ FTIR with in situ ellipsometry and/or QCM gravimetry can bring together the dynamics of chemical processes with physical/diffusive processes in SIS. The combination of such techniques can allow researchers to further assess the role of temperature-dependent precursor-polymer interactions on transport. In the coming years, x-ray techniques will shed light on the precise microstructure of hybrid SIS materials within polymers. X-ray techniques will reveal how processing can determine microstructures, yielding novel functional materials. There are many more electrical, mechanical, and chemical characterization methods used to assess the functional properties of SIS materials that are not discussed here.

Density functional theory (DFT) is a powerful method, and relatively simple implementations for SIS systems can generate quantities that are useful in process development. In the simplest implementation, a single monomer and a metal-organic precursor are brought together and allowed to equilibrate. This type of simulation can predict whether a reversible complex will form between the polymer and metal-organic molecule. If a complex does form, then the precursor is likely a viable candidate for SIS. For a given polymer-precursor pair, the binding energy, equilibrium distance, and equilibrium constant, along with perturbations to the IR spectrum of the initial molecule, can all be extracted.^16,25 The next degree of complexity is to describe the reaction coordinate–energy map of potential chemical reactions between the precursor and functional group, as in work discussed earlier by the Knez group.⁹⁹ 

A further step in complexity for DFT-modeling of SIS is introducing the second coreactant (H₂O, O₃, H₂O₂, etc.) to the reversible polymer/precursor complex and monitoring the possible reaction pathways and reaction coordinates that occur. Recently described SIS chemistries show growth at conditions that ALD/CVD does not occur. DFT investigations of the reaction in isolation, on a model Si surface, and in complex with a monomer may reveal reaction promotion or catalysis caused by the reversibly complexed configuration. Future DFT studies of SIS processes could include the interactions of multiple polymer functional groups with a single SIS metal precursor molecule or the interaction of more than one SIS metal precursor molecules on a single polymer functional group. Other simulation categories that, to the best of our knowledge, have not been employed in the context of SIS include molecular dynamics, reactive molecular dynamics, and multiscale modeling, which have potential as powerful tools to inform the holistic aspects of the SIS process across time and length scales.

Sequential infiltration synthesis has emerged as a technique to grow hybrid organic/inorganic components and templated inorganic nanostructures with implications across a wide range of research and application areas. Fundamental physical chemical research into the phenomena involved will enable a deeper understanding of this technique, and a wide range of experimental and simulation approaches are available to realize a more thorough framework for developing and analyzing SIS processes. Careful consideration of the thermodynamics and transport involved may open the SIS materials library beyond metals and metal oxides to include functional nitrides, sulfides, and other compounds.

In the coming years, researchers can focus on delineating between the different classes of precursor/polymer interactions that enable temporal colocation of reactants in the polymer at the exclusion of the vapor phase. We anticipate that as new functional materials, such as conductive oxides, catalytically active oxides, photoluminescent materials, and more, are integrated into the SIS materials library, the role of kinetic processing and precursor/polymer interactions may open avenues to tune the microstructure and functional properties of materials. Routes that we have identified for such research include comparing different precursor ligands with the same metal center and engineering of the local environment of interacting functional groups both sterically and chemically. Recent evidence from comparing trimethyl-metal systems reveals that polymer functional groups may act catalytically to enable reactions below the expected temperature limits. In this context, materials discovery by machine learning methods¹³⁶ may prove useful in combinatorial screening of precursor-functional groups in silico.

This work was supported as part of the Advanced Materials for Energy-Water Systems (AMEWS) Center, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Contract No. DE-AC02-06CH11357. This research used, in part, resources of the Advanced Photon Source (Sector 9-8BM), an Office of Science User Facility operated for the U.S. Department of Energy (DOE) Office of Science by Argonne National Laboratory.