Engineering Metastability in Atomic Layer Deposition: Polymorph and Valence Control

ABSTRACT

Metastable phases characterized by their higher‐energy states offer promising functionalities for electronic, catalytic, and energy applications. However, their synthesis is often hindered by high formation energy barriers and thermodynamic constraints. Atomic layer deposition (ALD) has attracted significant interest for a wide range of applications, including semiconductors, advanced electronics, and energy‐related applications, owing to its exceptional features, including low‐temperature processing, precise atomic‐scale control, and excellent conformality. Despite these advantages, the inherently low thermal budget of ALD poses significant challenges for the synthesis of metastable phases. This review presents a comprehensive overview of the recent advances in the engineering of metastability via ALD. This review categorizes the manifestations of metastability in ALD into two main directions: polymorphic transformations and valence state control. For polymorphs, strategies, such as temperature modulation, substrate‐induced lattice matching, grain‐size refinement, doping, and solid‐solution formation, enable selective phase stabilization. Approaches for valence control include temperature modulation, the design and selection of the precursor/reactant, and post‐deposition treatments. By linking reaction mechanisms with material phases, this review offers insights into the stabilization of metastable phases and practical design principles for achieving them. These insights will pave the way for new functional materials that surpass conventional thermodynamic limitations and advance next‐generation devices and technologies.

This work presents strategies for stabilizing two types of metastable phases: structural polymorphs and multivalence states via ALD. These approaches focus on lowering the energy barriers required for metastable phase formation, enabling access to metastable states beyond the thermodynamic limits imposed by the low‐temperature conditions of ALD. This review offers practical guidance for achieving controlled phase formation in ALD‐grown films.

1. Introduction

Advances in modern technologies are approaching the performance limits of conventional materials, thereby increasing the demand for new functional materials with enhanced properties and greater versatility. Metastable phases are commonly defined as states that reside at higher free energy than the thermodynamic ground state at a given composition and temperature. In many oxide, nitride, and chalcogenide systems, metastable polymorphs lie within a moderate energy offset—typically on the order of tens of meV per atom—and may be experimentally accessible if kinetic barriers prevent relaxation into the equilibrium phase [1]. In practical film processing, it is also necessary to consider states that behave effectively as metastable within the accessible temperature window. These are phases that do not appear as equilibrium phases under those conditions because the system is kinetically trapped before reaching the ground state or distributed among several closely competing configurations, such as multiple oxidation states of transition‐metal cations. These hard‐to‐form phases differ from conventionally used stable phases in their atomic coordination and bond lengths, and these deviations often translate into superior performance. Metastable phases have been used in various applications, including emerging semiconductors [2, 3, 4, 5, 6, 7, 8, 9], nanoelectronics [10, 11, 12, 13, 14, 15, 16], optoelectronics [17, 18, 19, 20, 21], catalysts [22, 23, 24, 25, 26], sensors [27, 28, 29, 30], and energy storage systems [31, 32, 33, 34, 35]. This suggests that metastable phases can create new opportunities for material technologies.

Despite their potential, the design and synthesis of metastable phases remain challenging. Their formation typically requires bypassing thermodynamically favorable pathways and overcoming substantial energy barriers, often necessitating extreme conditions, such as high temperatures or pressures. However, these conditions are not readily compatible with conventional thin‐film fabrication processes.

Atomic layer deposition (ALD) is a thin‐film growth technique based on self‐limiting surface reactions between precursors and functional groups on the reaction surface. Owing to this self‐limiting mechanism, ALD offers precise thickness control at the atomic level with outstanding uniformity and conformality over complex 3D structures [36]. These unique advantages have made ALD a critical tool for a wide range of applications, including advanced electronics [37, 38, 39, 40, 41, 42, 43, 44], energy storage [45, 46, 47, 48, 49, 50], energy conversion [51, 52, 53, 54, 55], and catalysis [56, 57, 58, 59]. However, the self‐limiting mechanism that enables these advantages also imposes constraints; it is effective only below the thermal decomposition temperature of the precursors, limiting most ALD processes to temperatures typically below 350°C [60, 61], which is significantly lower than the temperatures used in sputtering or chemical vapor deposition. This inherently low thermal budget imposes kinetic limitations on atomic rearrangement and long‐range diffusion, preventing atoms from overcoming the thermodynamic barriers required to form metastable phases. Moreover, the sequential, cycle‐by‐cycle nature of ALD suppresses non‐equilibrium pathways that are often critical in other high‐energy deposition techniques. Consequently, both the narrow ALD temperature window and its self‐limiting reaction mechanism pose significant challenges for the formation of metastable phases. Therefore, ALD may be less favorable than other growth techniques for the realization of metastable phases.

This apparent contradiction raises a fundamental question: Can metastable phases be synthesized via ALD, and if so, what strategies enable their formation? As illustrated in Figure 1a, the solution lies in strategies used to engineer the activation barrier associated with the formation of metastable phases. Recent studies have demonstrated that the synthesis of certain metastable phases via ALD is feasible through the precise control of the process conditions. Furthermore, ALD offers several tools, such as surface chemistry engineering, interface‐mediated templating, and nanoscale confinement, which can be exploited to stabilize metastable phases beyond conventional thermodynamic limits.

FIGURE 1.

(a) Schematic of the engineering of the activation barrier for the selection of metastable phases. (b) Overview of ALD strategies for synthesizing metastable polymorphic and multivalent phases, along with their respective challenges.

In this review, we discuss the strategies for stabilizing metastable phases in thin films using ALD, with a particular focus on phases that are thermodynamically or effectively metastable under the conventional ALD temperature window, yet can be realized and retained through appropriate process design. Before introducing these strategies, we classified the types of metastability observed in ALD‐grown films into two main classes. The first involves controlling the crystal structure, where the materials exhibit polymorphism, that is, the ability to form different crystal structures (such as anatase vs. rutile TiO₂ or monoclinic vs. orthorhombic HfO₂) depending on the reaction pathway. The second focuses on controlling the valence state, in which transition metal cations (e.g., Fe²⁺/Fe³⁺ and Sn²⁺/Sn⁴⁺) can exist in multiple valence states, some of which are metastable. Figure 1b summarizes this classification and outlines the major challenges and representative strategies associated with each direction. For polymorphic control, the main issue lies in overcoming the activation energy barrier under the low‐temperature constraints of ALD. The challenge of valence state control originates from the maintenance of unstable states under typical ALD conditions.

Through this framework, we aim to clarify how metastable phases can be accessed and stabilized during the ALD process, emphasizing that understanding this control is essential for enabling the next generation of high‐performance functional materials.

2. Strategies for Realizing Metastable Polymorphs via ALD

In polymorphic materials, there is only one crystal structure with a thermodynamic ground state, whereas other polymorphs possess higher formation energies. However, certain metastable phases with formation energies within the metastability range (<approximately 70 meV atom⁻¹ above the ground state) [1] can be readily synthesized and observed. Notably, a large number of such polymorphic metastable phases have been successfully synthesized via ALD despite its low‐temperature limits. Representative examples are listed in Table 1 [62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181].

TABLE 1.

Overview of ALD processes for metastable polymorphic phases. This table summarizes, for each material, the precursor/reactant combinations, process windows, substrates, and heteroelements used to stabilize the corresponding metastable polymorph.

Element	Metastable polymorph	Precursor	Reactant	Substrate	Hetero‐element	Growth temperature (°C)	References
Be	Rocksalt‐BeO	BeMe2	H2O	Si	Mg	250	[62]
Ti	Rutile‐TiO2	TiCl4	H2O	Si, SiO2	—	425‐600	[63]
	Rutile‐TiO2	TiCl4	H2O	RuO2	—	400	[64, 65, 66]
	Rutile‐TiO2	TiCl4	H2O	C‐cut sapphire	—	600	[67]
	Rutile‐TiO2	TiCl4	O3	Si, SiO2	—	600	[68]
	Rutile‐TiO2	TiCl4	O3	RuO2	—	250‐450	[69, 70, 71]
	Rutile‐TiO2	TiCl4	O3	R‐cut sapphire	—	350	[72]
	Rutile‐TiO2	TiI4	H2O	SiO2	—	445	[73]
	Rutile‐TiO2	TiI4	H2O2	Si, SiO2	—	455	[74, 75]
	Rutile‐TiO2	TiI4	H2O2	R,C‐cut sapphire	—	275‐375	[74]
	Rutile‐TiO2	TiI4	H2O2	MgO(001)	—	455	[74]
	Rutile‐TiO2	TiI4	O2	Si	—	457	[76]
	Rutile‐TiO2	Ti(OiPr)4	H2O	Si	Nb	300	[77]
	Rutile‐TiO2	Ti(OiPr)4	H2O	RuO2	—	250	[78]
	Rutile‐TiO2	Ti(OiPr)4	O2 plasma	Ru	—	250	[79]
	Rutile‐TiO2	Ti(OiPr)4	O2 plasma	RuO2	—	250	[80]
	Rutile‐TiO2	Ti(OiPr)4	O3	Ru, Ir	—	250	[81, 82, 83, 84, 85, 86]
	Rutile‐TiO2	Ti(OiPr)4	O3	RuO2, SnO2, MoO2	—	250	[87, 88, 89]
	Rutile‐TiO2	Ti(CpMe5)(OMe)3	O2 plasma	TiN	Sn	350	[90]
	Rutile‐TiO2	Ti(CpMe5)(OMe)3	O3	TiN	Sn	350	[91]
	Rutile‐TiO2	Ti(CpMe5)(OMe)3	O3	Ru	—	330	[92]
	Rutile‐TiO2	Ti(CpMe5)(OMe)3	O3	RuO2, MoO2	—	300‐350	[93, 94, 95, 96, 97]
	Rutile‐TiO2	Ti(OEt)4	H2O	Si	—	350	[98]
	Rutile‐TiO2	Ti(NMe2)4	H2O	Si	—	200	[99]
	TiO2‐II	TiCl4	H2O	Si	—	425	[100, 101]
	TiO2‐II	TiCl4	H2O	C‐cut sapphire	—	425	[67]
	TiO2‐II	TiCl4	O3	C‐cut sapphire	—	300‐450	[69, 72]
Mn	ε‐MnO2	Mn(thd)3	O3	C‐cut sapphire	—	186	[102]
Fe	β‐Fe2O3	Fe(Cp)2	O3	ITO	—	200	[103]
Ga	α‐Ga2O3	GaI3	O3	α‐Cr2O3	—	275‐550	[104, 105]
	α‐Ga2O3	GaMe3	O2 plasma	C‐cut sapphire	—	250‐295	[106, 107, 108]
	κ/ε‐Ga2O3	GaI3	O3	Si, SiO2	—	450‐550	[104, 105]
	κ/ε‐Ga2O3	GaMe3	O2 plasma	C‐cut sapphire	—	365	[106]
Zr	Fluorite‐ZrO2	ZrCl4	H2O	Si	—	210‐600	[174, 175, 179]
	Fluorite‐ZrO2	ZrI4	H2O2	Si	—	250‐375	[176, 178]
	Fluorite‐ZrO2	CpZr(NMe2)3	H2O	Si	—	200‐400	[177]
	Fluorite‐ZrO2	Zr(NEtMe)4	O3	TiN	—	225‐300	[180, 181]
Mo	Rutile‐MoO2	Unknown	O3	TiN	Sn	300	[97]
	β‐MoO3	Mo(CO)6	O2 plasma	SiO2	—	160	[109]
	β‐MoO3	Mo(CO)6	O3	Si	—	165	[110]
	β‐MoO3	Mo(CO)6	Mo(CO)6	C‐cut sapphire	—	167	[111]
In	α‐In2O3	In(DMAE‐tBu)	H2O	Si	—	100‐150	[112]
Sn	π‐SnS	Sn(DPFA)2	H2S	Si, SiO2	—	80‐120	[113, 114]
	π‐SnS	Sn(dmamp)2	H2S	Si, SiO2	—	90‐150	[115]
	π‐SnS	Sn(acac)2	H2S	Si, Al2O3	—	80‐120	[116, 117]
	π‐SnS	Sn(Bu)2(diamido)	H2S	SiO2	—	25‐100	[118]
Hf	Fluorite‐HfO2	HfCl4	H2O	Si	—	300‐940	[119, 120, 121, 122, 123, 124, 125, 126]
	Fluorite‐HfO2	HfCl4	H2O	TiN	Al	300	[127]
	Fluorite‐HfO2	HfCl4	H2O	TiN	Sr	300	[128]
	Fluorite‐HfO2	HfCl4	H2O	Si, TiN	Gd	300	[129, 130]
	Fluorite‐HfO2	HfCl4	O3	Si	—	450‐600	[131]
	Fluorite‐HfO2	HfCl4	O3	Si	Pr	325	[132, 133]
	Fluorite‐HfO2	HfI4	H2O	Si	—	600	[122]
	Fluorite‐HfO2	Hf(NMe2)4	H2O	Si	Zr	250	[134]
	Fluorite‐HfO2	Hf(NMe2)4	O2 plasma	W	Ga	270	[135]
	Fluorite‐HfO2	Hf(NMe2)4	O3	TiN	Zr	250‐285	[136, 137, 138, 139]
	Fluorite‐HfO2	Hf(NEtMe)4	H2O	TiN	Al	200	[140]
	Fluorite‐HfO2	Hf(NEtMe)4	H2O	TiN	Si	200‐300	[127, 140]
	Fluorite‐HfO2	Hf(NEtMe)4	H2O	TiN	V	240	[141]
	Fluorite‐HfO2	Hf(NEtMe)4	H2O	TiN	Zr	250	[142]
	Fluorite‐HfO2	Hf(NEtMe)4	H2O	TiN	Zr	260	[143]
	Fluorite‐HfO2	Hf(NEtMe)4	H2O	TiN	Zr	250‐300	[144]
	Fluorite‐HfO2	Hf(NEtMe)4	O2	Si	—	280	[145]
	Fluorite‐HfO2	Hf(NEtMe)4	O2 plasma	Si	Al	250	[146]
	Fluorite‐HfO2	Hf(NEtMe)4	O2 plasma	Si, TaN	Si	200	[147, 148]
	Fluorite‐HfO2	Hf(NEtMe)4	O2 plasma	TiN	La	235	[149]
	Fluorite‐HfO2	Hf(NEtMe)4	O3	TiN	—	200‐220	[150]
	Fluorite‐HfO2	Hf(NEtMe)4	O3	TiN	Al	300	[151]
	Fluorite‐HfO2	Hf(NEtMe)4	O3	TiN	Y	275‐300	[152, 153]
	Fluorite‐HfO2	Hf(NEtMe)4	O3	TiN	Zr	260	[143]
	Fluorite‐HfO2	Hf(NEtMe)4	O3	TiN	Zr	280	[154, 155, 156]
	Fluorite‐HfO2	Hf(NEtMe)4	O3	TiN	La	250	[157]
	Fluorite‐HfO2	Hf(NEtMe)4	O3	TiN	La	265	[158]
	Fluorite‐HfO2	Hf(NEtMe)4	O3	TaN	La	300	[159]
	Fluorite‐HfO2	Hf(OtBu)(NEtMe)3	O3	Rutile‐TiO2	—	250	[160]
	Fluorite‐HfO2	HfCp(NMe2)3	O3	TiN	Y	300	[161]
	Fluorite‐HfO2	HfCp(NMe2)3	O3	TiN	Zr	300‐320	[162, 163, 164]
	Fluorite‐HfO2	Hf(CpMe)2(OMe)Me	O3	TiN	Y	300‐350	[152]
Ta	δ‐Ta2O5	TaF5	H2O	Si	—	400‐450	[165]
	δ‐Ta2O5	TaCl5	H2O	Si, SiO2	—	300‐500	[166, 167]
	δ‐Ta2O5	TaI5	H2O2	Si	—	400	[168]
	δ‐Ta2O5	Ta(OEt)5	H2O, O3	TiN	—	300	[169]
Bi	β‐Bi2O3	Bi(OCMe2 iPr)	H2O	TaN, TiN, Si3N4	—	90‐210	[170]
	β‐Bi2O3	Bi(ph)3	O3	LAO(001)	—	280	[171]
	β‐Bi2O3	Bi(ph)3	O3	STO(001)	—	280	[171]
	β‐Bi2O3	Bi(mmp)3	O3	STO(001)	Fe	200	[172]
	γ‐Bi2O3	Bi(thd)3	H2O	Si	—	200‐300	[173]
	γ‐Bi2O3	Bi(ph)3	O3	Si	—	280	[171]
	δ‐Bi2O3	Bi(mmp)3	O3	YSZ(111)	Fe	200	[172]

The significance of synthesizing metastable polymorphs via ALD extends far beyond academic interest. Several of these phases demonstrated practical utility in advanced technological applications. A representative example is tetragonal ZrO₂, which is used as a high‐k dielectric in commercial dynamic random‐access memory (DRAM) capacitors. While the stable monoclinic phase of ZrO₂ exhibits a low dielectric constant of approximately 15 [182], the ALD‐grown tetragonal phase achieves a significantly high dielectric constant of ∼ 40 [181, 183], enabling the further scaling of DRAM cells. In addition, the orthorhombic phase of HfO₂, a metastable phase exhibiting ferroelectricity, has attracted considerable attention for use in next‐generation memory and logic devices [142, 184, 185, 186, 187].

The successful formation of metastable phases via ALD cannot be attributed to their low relative formation energies. Instead, phase‐selective mechanisms are essential for guiding stabilization. In the following subsections, we explore the primary strategies that enable phase selectivity in ALD‐grown films: the control of the deposition temperature, substrate‐induced lattice matching, grain‐size refinement, and chemical modulation via doping or solid‐solution formation.

2.1. Control of Deposition Temperature

Temperature and pressure are among the most fundamental and effective parameters for selectively controlling the crystal structures of polymorphic materials. However, owing to the intrinsic characteristics of ALD, applying high pressure is practically unfeasible. Moreover, as described in the Introduction, the deposition temperatures in most ALD processes are limited to below approximately 350°C. Nevertheless, even when using the same precursor, the target material may exhibit multiple crystal structures, depending on the specific reaction pathways available. If the energy barriers of these pathways are similar, the resulting crystal structure can vary within the ALD window, as shown in Figure 2. In some materials, low atomic mobility favors the formation of metastable structures even at low temperatures, whereas in others, an increased deposition temperature can provide sufficient thermal energy to induce phase transitions into metastable polymorphs. Representative examples of the temperature‐dependent formation of metastable phases are summarized in Table 2 [63, 68, 73, 74, 75, 76, 98, 99, 100, 101, 104, 105, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 131, 165, 166, 167, 168, 169, 170, 171, 173, 188].

FIGURE 2.

Schematic of the temperature‐dependent engineering of metastable phases in ALD.

TABLE 2.

Representative examples of ALD processes for the formation of metastable polymorphs via temperature modulation. This table summarizes, for each precursor/reactant combination, the growth and annealing temperatures at which metastable and stable phases appear, revealing the temperature window for metastable phase formation and whether it lies below or above that of the stable polymorph.

Element	Metastable polymorph	Precursor	Reactant	Growth temperature (°C)	Annealing temperature (°C)	Temperature window for stable phases (°C)	References
Ti	Rutile‐TiO2	TiCl4	H2O	425‐600	—	165‐400(Anatase)	[63, 68, 98, 100, 101, 188]
	Rutile‐TiO2	TiCl4	O3	600	—	250‐500(Anatase)	[68]
	Rutile‐TiO2	TiI4	H2O	445	—	135‐375(Anatase)	[73]
	Rutile‐TiO2	TiI4	H2O2	455	—	230‐375(Anatase)	[74, 75]
	Rutile‐TiO2	TiI4	O2	457	—	235‐380(Anatase)	[76]
	Rutile‐TiO2	Ti(OEt)4	H2O	350	750	350(Anatase)	[98]
	Rutile‐TiO2	Ti(NMe2)4	H2O	200	500	375‐500(Anatase)	[99]
	TiO2‐II	TiCl4	H2O	425	—	165‐350(Anatase)	[100, 101]
Ga	κ/ε‐Ga2O3	GaI3	O3	450‐550	—	425(Amorphous)	[104, 105]
Mo	β‐MoO3	Mo(CO)6	O2 plasma	160	400	500‐600(α‐MoO3)	[109]
	β‐MoO3	Mo(CO)6	O3	165	300	400‐600(α‐MoO3)	[110, 111]
In	α‐In2O3	In(DMAE‐tBu)	H2O	100‐150	—	200‐250(Bixbyite)	[112]
Sn	π‐SnS	Sn(DPFA)2	H2S	80‐120	—	140‐200(α‐SnS)	[113, 114]
	π‐SnS	Sn(dmamp)2	H2S	90‐150	—	180‐240(α‐SnS)	[115]
	π‐SnS	Sn(acac)2	H2S	80‐120	—	175(α‐SnS)	[116, 117]
	π‐SnS	Sn(Bu)2(diamido)	H2S	25‐100	—	125‐250(α‐SnS)	[118]
Hf	Fluorite‐HfO2	HfCl4	H2O	300‐940	—	226‐600(Monoclinic)	[119, 120, 121, 122, 123, 124, 125, 126]
	Fluorite‐HfO2	HfCl4	O3	450‐600	—	225‐350(Monoclinic)	[131]
	Fluorite‐HfO2	HfI4	H2O	600	—	300‐450(Monoclinic)	[122]
Ta	δ‐Ta2O5	TaF5	H2O	400‐450	—	—	[165]
	δ‐Ta2O5	TaCl5	H2O	300‐500	—	400(β‐Ta2O5)	[166, 167]
	δ‐Ta2O5	TaI5	H2O2	400	—	350(β‐Ta2O5)	[168]
	δ‐Ta2O5	Ta(OEt)5	H2O, O3	300	—	—	[169]
Bi	β‐Bi2O3	Bi(OCMe2 iPr)	H2O	90‐210	—	800(α‐Bi2O3)	[170]
	γ‐Bi2O3	Bi(thd)3	H2O	200‐300	600‐700	300‐500(α‐Bi2O3)	[173]
	γ‐Bi2O3	Bi(ph)3	O3	280	500‐700	400(α‐Bi2O3)	[171]

These examples, such as the formation of metastable π‐SnS (vs. α‐SnS), β‐Bi₂O₃ and γ‐Bi₂O₃ (vs. α‐Bi₂O₃) at low deposition temperatures, highlight the role of kinetic constraints in ALD. At low temperatures, limited atomic diffusion and insufficient surface reconstruction can induce the nucleation of metastable polymorphs during the early growth stages. SnS is a typical example of the phenomenon. In ALD processes using various Sn precursors and H₂S gas, the ALD process at below approximately 150°C favors the formation of the thermodynamically metastable π‐phase (Cubic, Pa‐3), whereas the thermodynamically stable α‐phase (Orthorhombic, Pnma) becomes dominant as the temperature increases [113, 114, 115, 116, 117, 118]. A similar trend was observed in the ALD of In₂O₃ films. In an ALD process employing In(DMAE‐tBu) and H₂O, the metastable rhombohedral phase (R‐3c) is favored at temperatures below ∼160°C, whereas the thermodynamically stable cubic phase (Ia‐3) dominates at higher temperatures [112]. These examples demonstrate that carefully controlled low‐temperature ALD conditions selectively promote the formation of metastable crystal structures.

In contrast, certain metastable phases were formed at elevated deposition temperatures. Notably, unlike metal–organic precursors, halide precursors extend the ALD temperature window to temperatures far exceeding 400°C, thereby enabling the stabilization of certain metastable phases. TiO₂ is typically deposited in the thermodynamically stable anatase phase (Tetragonal, I4₁/amd) under most ALD conditions. However, when halide precursors, such as TiCl₄ or TiI₄, are employed and the deposition temperatures exceed 600°C, the partial formation of the rutile phase (Tetragonal, P4₂/mnm) has been reported [63, 68, 73, 74, 75, 76, 98, 100, 101, 188]. In addition, the TiO₂‐II phase (Orthorhombic, Pbcn), which is typically stabilized only under high‐pressure conditions, has been reported to form at elevated temperatures [100, 101]. Similar phenomena have been observed in ALD‐grown HfO₂ [119, 120, 121, 122, 123, 124, 125, 126, 131], Ta₂O₅ [165, 166, 167, 168], and Ga₂O₃ [104, 105], where the use of halide precursors enables high‐temperature growth that facilitates the formation of metastable polymorphs not observed within the conventional ALD window.

Post‐deposition annealing (PDA) is an effective strategy to stabilize metastable polymorphs. For instance, amorphous MoO₃ films deposited by ALD at temperatures below 165°C using Mo(CO)₆ are transformed into the metastable β‐phase (Monoclinic, P2₁/c) upon annealing at approximately 300°C [109, 110, 111]. Further annealing above 400°C leads to predominant crystallization into the thermodynamically stable α‐phase (Orthorhombic, Pnma). Similarly, Bi₂O₃ films initially deposited in the stable α‐phase (Monoclinic, P2₁/c) at 250–300°C using Bi(thd)₃ or Bi(ph)₃ can be recrystallized into the γ‐phase (Cubic, I23) through PDA at temperatures above 500°C [171, 173]. These examples suggest that the thermal energy supplied through PDA can effectively overcome kinetic barriers, enabling the formation of metastable phases.

2.2. Substrate‐Induced Lattice Matching

Another effective approach for stabilizing metastable polymorphs is to use substrates that provide structural compatibility with the desired phase. In particular, selecting a substrate with a similar lattice structure or matching crystallographic orientation can reduce the energy required for atomic rearrangement, thereby enabling the formation of metastable phases (Figure 3a). This strategy serves as a useful platform for selectively stabilizing metastable polymorphs while suppressing their transition to more stable polymorphs. Table 3 summarizes representative examples in which metastable crystal structures were synthesized via lattice matching between ALD‐grown films and the underlying substrates [64, 65, 66, 67, 69, 70, 71, 72, 74, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 92, 93, 94, 95, 96, 97, 102, 103, 104, 105, 106, 107, 108, 160, 171, 172].

FIGURE 3.

Substrate‐induced pathways for stabilizing metastable phases via lattice or domain matching. (a) Schematic for lowering the activation barrier for stabilizing metastable phases via lattice‐matching. (b) Stabilization of metastable rutile TiO₂ on RuO₂ utilizing the structural similarity between rutile TiO₂ and rutile RuO₂. (c) Schematic of the formation of metastable rutile TiO₂ on Ru via the in situ formation of a rutile RuO₂ interfacial layer. (d) Schematic of the atomic arrangements of rutile TiO₂ (101) and α‐sapphire (012) surfaces.

TABLE 3.

Representative examples of ALD processes for the formation of metastable polymorphs through substrate‐induced lattice matching. This table summarizes cases where lattice‐matching substrates compatible with the desired metastable polymorph are employed to selectively form the target phase within the ALD temperature window.

Element	Metastable polymorph	Precursor	Reactant	Substrate	Growth temperature (°C)	References
Ti	Rutile‐TiO2	TiCl4	H2O	C‐cut sapphire	600	[67]
	Rutile‐TiO2	TiCl4	H2O	RuO2	400	[64, 65, 66]
	Rutile‐TiO2	TiCl4	O3	R‐cut sapphire	350‐450	[69, 72]
	Rutile‐TiO2	TiCl4	O3	RuO2	250‐350	[70, 71]
	Rutile‐TiO2	TiI4	H2O2	R‐cut sapphire	275	[74]
	Rutile‐TiO2	TiI4	H2O2	C‐cut sapphire	375	[74]
	Rutile‐TiO2	TiI4	H2O2	MgO(001)	455	[74]
	Rutile‐TiO2	Ti(OiPr)4	H2O	RuO2	250	[78]
	Rutile‐TiO2	Ti(OiPr)4	O2 plasma	Ru	250	[79]
	Rutile‐TiO2	Ti(OiPr)4	O2 plasma	RuO2	250	[80]
	Rutile‐TiO2	Ti(OiPr)4	N2O plasma	Ru	250	[79]
	Rutile‐TiO2	Ti(OiPr)4	O3	Ru	250	[81, 82, 83, 84]
	Rutile‐TiO2	Ti(OiPr)4	O3	Ir	250	[85]
	Rutile‐TiO2	Ti(OiPr)4	O3	Ru‐Pt alloy	250	[86]
	Rutile‐TiO2	Ti(OiPr)4	O3	RuO2	250	[65, 66, 80]
	Rutile‐TiO2	Ti(OiPr)4	O3	SnO2	250	[87, 88]
	Rutile‐TiO2	Ti(OiPr)4	O3	MoO2	250	[89]
	Rutile‐TiO2	Ti(CpMe5)(OMe)3	O3	Ru	330	[92]
	Rutile‐TiO2	Ti(CpMe5)(OMe)3	O3	RuO2	300‐330	[93, 94]
	Rutile‐TiO2	Ti(CpMe5)(OMe)3	O3	MoO2	320‐350	[95, 96, 97]
	TiO2‐II	TiCl4	H2O	C‐cut sapphire	425	[67]
	TiO2‐II	TiCl4	O3	C‐cut sapphire	300‐450	[69, 72]
Mn	ε‐MnO2	Mn(thd)3	O3	C‐cut sapphire	186	[102]
Fe	β‐Fe2O3	Fe(Cp)2	O3	ITO(001)	200	[103]
Ga	α‐Ga2O3	GaI3	O3	α‐Cr2O3	275‐550	[104, 105]
	α‐Ga2O3	GaMe3	O2 plasma	C‐cut sapphire	250‐295	[106, 107, 108]
	κ/ε‐Ga2O3	GaMe3	O2 plasma	C‐cut sapphire	365	[106]
Hf	Fluorite‐HfO2	Hf(OtBu)(NEtMe)3	O3	Rutile‐TiO2	250	[160]
Bi	β‐Bi2O3	Bi(ph)3	O3	LAO(100), STO(100)	280	[171]
	β‐Bi2O3	Bi(mmp)3	O3	STO(100)	300‐400	[172]
	δ‐Bi2O3	Bi(mmp)3	O3	YSZ(111)	400‐650	[172]

Rutile TiO₂ is one of the most well‐studied systems used to demonstrate this strategy. As shown in Figure 3b, rutile TiO₂ and rutile RuO₂ share the same crystal structure and exhibit similar lattice constants. This structural compatibility enables the selective formation of the rutile phase TiO₂ on RuO₂ substrates even at temperatures below 300°C [64, 65, 66, 70, 71, 78, 80, 93, 94]. Similar results have been reported for other structurally compatible substrates, such as IrO₂ [85], SnO₂ [87, 88], and distorted rutile‐structured MoO₂ [89, 95, 96, 97]. Another example is Bi₂O₃, where deposition on a YSZ (111) substrate leads to the formation of the metastable δ‐phase (Cubic, Fm‐3m) rather than the thermodynamically stable α‐phase, at deposition temperatures above 400°C [172]. This indicates that, although the substrate‐induced lattice matching strategy facilitates phase selection, sufficient thermal energy is still required to drive the phase transformation.

This strategy can be effective even when the ALD‐grown film and underlying substrate have different crystal structures. A representative case involves utilizing the structure of an interfacial layer that forms between the film and the substrate during the ALD process. For instance, metallic substrates, such as Ru or Ir, are known to form surface oxide layers—RuO₂ or IrO₂—upon exposure to strongly oxidizing reactants. These oxide layers adopt a rutile structure and can serve as effective templates for the formation of rutile TiO₂, which is a metastable phase. Consequently, the ALD of TiO₂ on these metal substrates using strong oxidants, such as O₃ or O₂ plasma, promotes the formation of rutile TiO₂ (Figure 3c) [79, 81, 82, 83, 84, 85, 92]. In addition, the selective formation of rutile TiO₂ occurs on Ru–Pt alloy substrates when the Ru content exceeds 50% owing to the preferential formation of a RuO₂ interfacial layer [86].

In certain cases, metastable phases can be stabilized through orientation‐specific matching even when the crystal structures of the film and substrate differ. For example, despite their different structures, rutile and α‐sapphire exhibit structural similarity between the rutile TiO₂ (101) plane and the α‐sapphire (012) plane, as shown in Figure 3d, enabling the formation of rutile TiO₂ at temperatures below 400°C [67, 69, 72, 74]. Similarly, domain‐matching epitaxy has been reported to stabilize other metastable structures, such as TiO₂‐II and ε‐MnO₂ (Hexagonal, P6₃/mmc), on α‐sapphire substrates [67, 69, 72, 102, 189]. These cases demonstrate that the stabilization of metastable phases is not limited to substrates with similar crystal structures but can also be achieved through periodic matching between specific crystallographic planes. These results provide promising opportunities for extending this approach to a broader range of material systems.

Moreover, domain‐matching epitaxy can be used to control the crystallographic orientation of ALD‐grown films. A recent study has demonstrated that, through the orientation engineering of α‐sapphire as a substrate, it was possible to stabilize c‐axis‐oriented rutile TiO₂, which is typically difficult to achieve [94]. This orientation control led to a significant enhancement in the dielectric performance: while the a‐axis‐oriented films exhibited a dielectric constant of ∼80, the c‐axis‐oriented films achieved values as high as 180.

2.3. Grain‐Size Refinement

In thin‐film systems, the total thermodynamic energy includes not only the bulk energy of individual grains, but also contributions from the grain boundary energy, surface energy, and interfacial energy with the substrate (Figure 4a). In the absence of kinetic limitations, a phase that minimizes the total energy is theoretically favored. However, as the grain size decreases, the relative contribution of the grain boundary energy becomes more significant, leading to the preferential formation of metastable phases that are not thermodynamically stable in the bulk form.

FIGURE 4.

Grain‐size‐induced stabilization of metastable polymorphs. (a) Schematic showing that the total energy of a film is the sum of the internal grain energy, surface energy, and grain boundary energy. (b) Calculated phase stability of ZrO₂ polymorphs as a function of grain size, indicating a reversal in stability between monoclinic and tetragonal phases at nanoscale dimensions [190]. Reproduced with permission.^[190] Copyright 2005, Wiley.

This phenomenon is particularly pronounced in materials, such as ZrO₂ and HfO₂. As shown in Figure 4b, when the grain size of the ZrO₂ film decreases (i.e., as the interfacial area increases), the total energy of the thermodynamically stable monoclinic phase can become higher than that of the metastable tetragonal phase, resulting in the reversal of their energetic favorability [190]. At even smaller grain sizes, however, crystallization can be hindered and the film may remain amorphous. Indeed, ALD‐grown ZrO₂ films are observed to form the tetragonal phase over a wide range of growth temperatures and film thicknesses, even without any specific stabilization strategy [174, 175, 176, 177, 178, 179, 180, 181]. However, as the film thickness increases and the grain size grows, these tetragonal films undergo a partial phase transformation into the monoclinic phase [177]. Also, in the case of HfO₂, the tetragonal phase can be maintained when grain growth is suppressed under specific processing conditions. Table 4 summarizes representative examples of ALD‐grown tetragonal ZrO₂ and HfO₂ films, together with the corresponding process windows and film thickness [145, 150, 174, 175, 176, 177, 178, 179, 180, 181].

TABLE 4.

Representative examples of ALD‐grown tetragonal ZrO₂ and HfO₂ films. This table summarizes process windows and film thicknesses where the metastable tetragonal phase is stabilized. In ZrO₂, the tetragonal phase is easily observed over a broad range of growth temperatures and film thicknesses, whereas in HfO₂ it is mainly obtained when grain growth is suppressed.

Element	Metastable polymorph	Precursor	Reactant	Growth temperature (°C)	Film thickness (nm)	Refs.
Zr	Fluorite‐ZrO2	ZrCl4	H2O	210‐600	26‐40	[174, 175, 179]
	Fluorite‐ZrO2	ZrI4	H2O2	250‐375	42‐125	[176, 178]
	Fluorite‐ZrO2	CpZr(NMe2)3	H2O	200‐400	18‐152	[177]
	Fluorite‐ZrO2	Zr(NEtMe)4	O3	225‐300	6‐35	[180, 181]
Hf	Fluorite‐HfO2	Hf(NEtMe)4	O3	200‐220	10	[150]
	Fluorite‐HfO2	Hf(NEtMe)4	O2	280	13	[145]

Compared with ZrO₂, HfO₂ requires a much smaller grain size and film thickness to stabilize the tetragonal phase [191]. Consequently, under typical ALD conditions, HfO₂ tends to crystallize predominantly in the monoclinic phase, whereas a tetragonal phase is rarely observed. However, in HfO₂ films doped with Ce or La, a tetragonal phase has been reported to form when the grain size is reduced to below ∼4 nm [192]. This suggests that the stabilization of the metastable tetragonal structure is feasible in ALD‐grown HfO₂ films by precisely controlling the grain size.

The grain size of HfO₂ films can also be controlled by adjusting the deposition temperature and injection conditions of the oxygen source. For example, in an ALD process using Hf(NEtMe)₄ and O₃, grain growth was suppressed at deposition temperatures below 220°C, resulting in the predominant formation of the tetragonal phase [150]. As the temperature increased, grain growth was promoted, inducing a phase transition toward the monoclinic phase. Similarly, when HfO₂ films were deposited at a relatively high temperature of 280°C with Hf(NEtMe)₄ and O₂, the effective suppression of grain growth led to the stable formation of tetragonal HfO₂ [145]. Accordingly, precise control over grain size is a critical factor in determining the crystal structure of thin films.

2.4. Chemical Modulation via Doping and Solid‐Solution Formation

Another effective strategy for achieving phase selectivity is the incorporation of heteroatomic dopants or the formation of solid solutions. Table 5 summarizes representative cases in which metastable phases were stabilized in ALD‐grown films through such chemical modulations [62, 77, 90, 91, 97, 127, 128, 129, 130, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 146, 147, 148, 149, 151, 152, 153, 154, 155, 156, 157, 158, 159, 161, 162, 163, 164, 193, 194, 195, 196, 197]. For example, in HfO₂, the introduction of dopants or solid‐solution formation induces new polymorphic phases. First‐principles calculations have revealed that doping HfO₂ with various elements (e.g., Si, Ge, Sn, Ti, and Ce) enhances the thermodynamic stability of the tetragonal phase [198, 199]. This effect is primarily attributed to lattice distortion, which becomes more pronounced as the ionic radius of the dopant decreases. Furthermore, doping with trivalent cations introduces oxygen vacancies for charge compensation, further stabilizing the distorted lattice structure. These combined effects lower the free energy of the tetragonal phase, making it more favorable for the formation of the tetragonal phase.

TABLE 5.

Representative examples of ALD processes for the formation of metastable polymorphs through chemical modulation. This table summarizes cases in which metastable polymorphs are stabilized by doping or forming solid‐solution with specific heteroelements, indicating which heteroelements can be used to access each metastable phase.

Element	Metastable polymorph	Precursor	Heteroelement precursor	Reactant	Growth Temperature (°C)	References
Be	Rocksalt‐BeO	BeMe2	Mg(EtCp)2	H2O	250	[62]
Ti	Rutile‐TiO2	Ti(OiPr)4	Nb(OEt)5	H2O	300	[77]
	Rutile‐TiO2	Ti(CpMe5)(OMe)3	SnCl4	O2 plasma	350	[90]
	Rutile‐TiO2	Ti(CpMe5)(OMe)3	ISN‐02(Sn)	O3	350	[91]
Mo	Rutile‐MoO2	Unknown	Sn(dmamp)2	O3	300	[97]
Hf	Fluorite‐HfO2	HfCl4	AlMe3	H2O	300	[127]
	Fluorite‐HfO2	HfCl4	Sr(tBu3Cp)2	H2O	300	[128]
	Fluorite‐HfO2	HfCl4	Pr(thd)3	O3	325	[132, 133]
	Fluorite‐HfO2	HfCl4	Gd(iPrCp)3	H2O	300	[127, 129, 130]
	Fluorite‐HfO2	Hf(NMe2)4	GaMe3	O2 plasma	270	[135]
	Fluorite‐HfO2	Hf(NMe2)4	Zr(NMe2)4	H2O	250	[134]
	Fluorite‐HfO2	Hf(NMe2)4	Zr(NMe2)4	O3	250‐260	[137, 138]
	Fluorite‐HfO2	Hf(NMe2)4	CpZr(NMe2)3	O3	260‐285	[136, 139]
	Fluorite‐HfO2	Hf(NEtMe)4	AlMe3	H2O	200	[140]
	Fluorite‐HfO2	Hf(NEtMe)4	AlMe3	O2 plasma	250	[146]
	Fluorite‐HfO2	Hf(NEtMe)4	AlMe3	O3	300	[151]
	Fluorite‐HfO2	Hf(NEtMe)4	V(NEtMe)4	H2O	240	[141]
	Fluorite‐HfO2	Hf(NEtMe)4	Si(NMe2)4	O2 plasma	200	[148]
	Fluorite‐HfO2	Hf(NEtMe)4	SiH(NMe2)3	H2O	200	[140]
	Fluorite‐HfO2	Hf(NEtMe)4	SiH(NMe2)3	O2 plasma	200	[147]
	Fluorite‐HfO2	Hf(NEtMe)4	SiH2(NEt2)2	H2O	300	[127]
	Fluorite‐HfO2	Hf(NEtMe)4	Y(MeCp)3	O3	275‐300	[152, 153]
	Fluorite‐HfO2	Hf(NEtMe)4	Zr(NEtMe)4	H2O	250‐300	[142, 143, 144]
	Fluorite‐HfO2	Hf(NEtMe)4	Zr(NEtMe)4	O3	260‐280	[143, 154, 155, 156]
	Fluorite‐HfO2	Hf(NEtMe)4	La(iPrCp)3	O2 plasma	235	[149]
	Fluorite‐HfO2	Hf(NEtMe)4	La(iPrCp)3	O3	250‐300	[157, 158, 159]
	Fluorite‐HfO2	HfCp(NMe2)3	Y(Cp)(NR2)2	O3	300	[161]
	Fluorite‐HfO2	HfCp(NMe2)3	ZrCp(NMe2)3	O3	300‐320	[162, 163, 164]
	Fluorite‐HfO2	Hf(CpMe)2(OMe)Me	Y(MeCp)3	O3	300‐350	[152]

In addition, the tetragonal phase has a relatively low energy barrier for the transition to the orthorhombic phase, which is another metastable phase. Accordingly, ALD‐grown HfO₂ films often exhibit a mixture of tetragonal and orthorhombic phases [185]. Interestingly, the orthorhombic phase exhibits ferroelectricity owing to its noncentrosymmetric structure. Park et al. experimentally demonstrated the formation of metastable tetragonal and orthorhombic phases in ALD‐grown HfO₂ films doped with Si, Al, and Gd [127]. Similar observations were reported for doping with other elements, including Al, Si, V, Ga, Sr, Y, La, Pr, and Gd [127, 128, 129, 130, 132, 133, 135, 140, 141, 146, 147, 148, 149, 151, 152, 153, 157, 158, 159, 161].

Owing to the structural similarity between ZrO₂ and HfO₂, Hf_xZr_1‐xO₂ films can form homogeneous solid solutions over a wide compositional range, which facilitates the stabilization of the orthorhombic phase [134, 136, 137, 138, 139, 142, 143, 144, 154, 155, 156, 164]. As the Zr content increases in the Hf_xZr_1‐xO₂ films, the crystal structure transitions from stable to metastable phases. Notably, ferroelectricity emerges when the Zr concentration reaches approximately 30%–50%, where the transition from monoclinic to orthorhombic occurs [142, 143]. Furthermore, ALD‐grown ZrO₂/HfO₂ nanolaminates have been shown to crystallize into the orthorhombic phase, exhibiting ferroelectric behavior [136, 139, 156]. This indicates that nanolaminate structures represent another promising route for stabilizing novel metastable structures.

A strategy has also been proposed to stabilize metastable phases by employing heterostructural materials with the desired crystal structure as a structural framework [62, 77, 90, 91, 97, 162, 163]. Recently, the ALD of thin films incorporating BeO into MgO with a rocksalt structure has been reported to stabilize the highly unstable BeO₆ octahedron [62]. BeO typically crystallizes into the wurtzite phase with stable BeO₄ tetrahedra. Its rocksalt phase is highly unstable, with a calculated energy of 0.483 eV atom⁻¹, which is much higher than the typical range for experimentally accessible metastable phases [1]. Indeed, rocksalt BeO was not experimentally observed prior to this work. In the reported ALD of Be_xMg_1‐xO solid solutions (x ≤ 0.2), the formation of Be–O octahedra was verified and attributed to the observed enhancement in the dielectric constant.

Similar approaches have been employed to stabilize other metastable phases. For example, the metastable distorted rutile phase of MoO₂ was stabilized through the formation of a Sn‐Mo‐O solid solution using rutile‐structured SnO₂ as the structural framework [97]. Additionally, the rutile phase of TiO₂ was stabilized in ALD‐grown solid‐solution films with rutile SnO₂ [90, 91] or NbO₂ [77]. Consequently, these examples highlight the utility of crystallographic templating in which structural similarity enables the stabilization of otherwise inaccessible metastable phases.

3. Materials with Different Valence States

The metastable phases are not limited to structural polymorphisms. Compounds composed of the same elements can also exist in different chemical stoichiometries or valence states, and can be considered metastable. Although materials tend to form in their most thermodynamically stable valence states, alternative valence states can emerge during film growth because of various factors. In particular, the reaction paths in ALD can be influenced by the deposition temperature and specific combinations of precursors and reactants, often enabling the stabilization of the metastable valence states of cations in ALD‐grown films. At the atomic scale, valence control in ALD arises because metal‐containing and anion‐containing species are supplied in separate half‐reactions. During the metal‐precursor pulse, chemisorption and ligand elimination generate a partially oxidized, metal‐rich surface intermediate, whereas the subsequent reactant pulse supplies O, S, Se, or N and further oxidizes this intermediate while removing residual ligands. The steady‐state valence of the growing film is therefore governed by the redox balance between these two half‐cycles, which is tuned by the growth temperature, precursor chemistry, and the type and dose of the reactant. This redox‐balance picture is consistent with vanadium‐oxide ALD using V(III) precursors and different oxidants, where adjusting oxidant strength and sequence systematically shifts the film between more reduced VO_x compositions and more oxidized V₂O₅‐like phases [200].

A wide range of multivalent compounds have been reported using ALD. Table 6 summarizes the representative systems, including oxides (SnO₂‐SnO, MoO₃‐MoO₂, Cu₂O‐CuO, WO₃‐W₂O₃, Fe₂O₃‐Fe₃O₄, MnO‐Mn₃O₄‐MnO₂, V₂O₅‐VO₂‐V₂O₃, CoO‐Co₃O₄, and NiO‐Ni₂O₃) [200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 256, 257, 258, 259, 260, 261, 262, 263, 264, 265, 266, 267, 268, 269, 270, 271, 272, 273], chalcogenides (SnS₂‐SnS, TiS₂‐TiS₃, SnSe₂‐SnSe, CuSe‐Cu₂Se, and Bi₂Se₃‐BiSe) [113, 115, 116, 118, 274, 275, 276, 277, 278, 279, 280, 281, 282, 283, 284, 285, 286, 287, 288, 289, 290, 291, 292, 293, 294, 295, 296, 297, 298, 299, 300, 301, 302, 303, 304, 305], and nitrides (WN‐W₂N, and TaN‐Ta₃N₅) [306, 307, 308, 309, 310, 311, 312, 313, 314, 315, 316, 317]. Additionally, the selective formation of noble metals and their corresponding oxides has been demonstrated using ALD [318, 319, 320, 321, 322, 323, 324, 325, 326, 327, 328, 329, 330, 331, 332, 333, 334, 335, 336, 337, 338, 339, 340, 341, 342, 343, 344, 345, 346]. These behaviors are particularly prominent in transition metal compounds, in which multivalency is inherent.

TABLE 6.

Overview of ALD processes for controlling valence states in metastable multivalent compounds. This table summarizes ALD processes that link accessible oxidation states to specific precursor/reactant chemistries and process windows, providing a concise guide to which multivalent elements and valence combinations have been demonstrated and where opportunities for unexplored valence states remain.

Element	Multivalent compounds	Precursor	Reactant	Temperature (°C)	References
Ti	TiS2	Ti(NMe2)4	H2S	150‐180	[302]
	TiS2	Ti(NMe2)4	H2S plasma	100	[303]
	TiS3	Ti(NMe2)4	H2S plasma	150‐200	[303]
V	V2O3	VO(acac)2	O3	200	[259]
	V2O3+VO2	V(amd)3	H2O2	200	[200]
	VnO2n‐1	VO(OiPr)3	H2O	60‐90	[264]
	V2O4	VO(OnPr)3	CH3COOH	200	[266]
	VO2	V(NEtMe)4	O3	150	[260, 261, 262, 263, 265]
	VO2	V(amd)3	H2O2+H2	200	[200]
	VO2	V(NEtMe)4	H2O	150	[262, 265]
	VO2	VO(OiPr)3	H2O	60‐135	[264, 267]
	V6O13	V(NEtMe)4	O3	150	[261, 262]
	V4O9	V(NEtMe)4	H2O	150	[262]
	V3O7	V(NEtMe)4	O3	150	[262]
	V2O5	VO(acac)2	O3	200	[259]
	V2O5	V(amd)3	O3	200	[200]
	V2O5	V(amd)3	H2O+O2	200	[200]
	V2O5	V(amd)3	H2O2	200	[200]
	V2O5	V(NEtMe)4	O3	150	[265]
	V2O5	V(NEtMe)4	H2O	150	[265]
	V2O5	V(NEtMe)4	O3+H2O	150	[262]
	V2O5	VO(OiPr)3	H2O	135	[267]
	VS	V(NEtMe)4	H2S	200‐300	[305]
	V2S3	V(amd)3	H2S	150‐200	[304]
	VS4	V(amd)3	H2S	225‐250	[304]
Cr	Cr2O3	Cr(acac)3	O3	300‐350	[210]
	Cr2O3+CrO3	CrO2Cl2	H2O	100‐200	[211, 212]
	Cr2O3+CrO3	CrO2Cl2	H2O2	150‐200	[211, 212]
	Cr2O3+CrO3	CrO2Cl2	H2O+H2O2	150	[211]
	Cr2O3+CrO3	CrO2Cl2	CH3OH	150	[211]
Mn	MnO	Mn(EtCp)2	H2O	100‐300	[232]
	MnO	Mn(thd)3	NH3 plasma	180	[235]
	Mn3O4	Mn(tBuAMD)2	H2O	150	[234]
	Mn3O4	Mn(thd)3	H2O	180	[235]
	Mn3O4	Mn(thd)3	O3	≥240	[236]
	Mn3O4	Mn2(CO)10	O3	120‐160	[237]
	Mn3O4+MnO	Mn(thd)3	H2 plasma	180	[235]
	Mn2O3	Mn(tBuAMD)2	H2O	175‐250	[234]
	Mn2O3+MnO	Mn2(CO)10	O3	60‐100	[237]
	Mn5O8	Mn(EtCp)2	O3	150‐180	[233]
	MnO2	Mn(thd)3	O3	140‐230	[235, 236]
Fe	Fe	Fe(tBuAMD)2	H2	250	[203]
	Fe	FeCp(CptBu)	O2 plasma	150‐250	[230]
	FeO	Fe(tBuAMD)2	H2O	250	[203]
	Fe3O4	Fe(2,4‐C7H11)2	H2O2	120	[226]
	Fe3O4	Fe2(OtBu)6	H2O	130‐170	[229]
	Fe3O4	FeCp(CptBu)	O2 plasma	150‐250	[230]
	Fe3O4	FeCp2	O2	400	[231]
	Fe3O4+Fe2O3	FeCp2	O2	350‐500	[227]
	Fe2O3	Fe(2,4‐C7H11)2	O3	120	[226]
	Fe2O3	FeCp2	O2	>500	[227]
	Fe2O3	FeCp2	O3	200	[228]
	Fe2O3	Fe2(OtBu)6	H2O	130‐170	[229]
	Fe2O3	FeCp(CptBu)	O2 plasma	150‐250	[230]
	Fe2O3	FeCp2	O2	400	[231]
	FeS	Fe(amd)2	H2S	80‐250	[287]
	FeS2	Fe(amd)2	H2S plasma	80‐200	[286]
	FeSe	Fe(iPrAMD)2	Se(SiEt3)2	350‐370	[288]
	Fe3Se4	Fe(iPrAMD)2	Se(SiEt3)2	390‐450	[288]
Co	Co	Co(iPrAMD)2	H2	350	[203]
	Co	Co(AMD)2	H2O	180‐270	[206]
	Co	CoCl2(TMEDA)	H2O	225‐300	[208]
	Co3O4	Co(iPrAMD)2	O3	225‐250	[204]
	Co3O4	Co2(CO)8	O3	50	[205]
	Co3O4	CoCp2	O3	150‐280	[209]
	Co3O4+CoO	Co2(CO)8	O3	50	[205]
	Co3O4+CoO	CoCp2	O3	331	[209]
Cu	Cu	[Cu(sBu‐amd)]2	H2 plasma	150	[217]
	Cu	(hfac)Cu(I)(DMB)	H2 plasma+O2 plasma	100	[220]
	Cu	Cu(I)(hfac)(tmvs)	N2	160	[224]
	Cu2O	Cu(PnBu3)2(acac)	H2O + O2	110‐125	[215]
	Cu2O	Cu(dmap)2	O3	80‐140	[216]
	Cu2O	[Cu(sBu‐amd)]2	H2 plasma+O2 plasma	150	[217]
	Cu2O	[Cu(sBu‐amd)]2	H2O	180	[218]
	Cu2O	Cu(dmamb)2	H2O	140‐160	[219]
	Cu2O	(hfac)Cu(I)(DMB)	H2 plasma+O2 plasma	100	[220]
	Cu2O	Cu(dmap)2	H2O	110‐175	[221]
	Cu2O	Cu(hfac)2	H2O	280	[222]
	Cu2O	Cu(OAc)2	H2O	180‐220	[223]
	Cu2O	Cu(I)(hfac)(tmvs)	H2O	220	[224]
	Cu2O	Cu(acac)2	O2+H2O	200	[225]
	Cu2O+CuO	(hfac)Cu(I)(DMB)	O3	100	[214]
	CuO	Cu(acac)2	O3	150‐230	[213]
	CuO	Cu(dmap)2	O3	80‐140	[216]
	CuO	[Cu(sBu‐amd)]2	O2 plasma	150	[217]
	CuO	(hfac)Cu(I)(DMB)	H2 plasma+O2 plasma	100	[220]
	CuO	Cu(I)(hfac)(tmvs)	O3	260	[224]
	Cu2Se	Cu(II) pivalate	(Et3Si)2Se	300	[275]
	Cu2Se	CuCl	(Et3Si)2Se	400	[275]
	CuSe	Cu(II) pivalate	(Et3Si)2Se	165	[275]
	Cu2S	[Cu(sBu‐amd)]2	H2S	130‐150	[276, 277, 278]
	Cu2S	CuAMD	H2S	135	[279]
	Cu2S	Cu(acac)2	H2S	130‐200	[280]
	Cu2S	Cu(dmamb)2	H2S	120‐200	[282]
	Cu1.8S	Cu(thd)2	H2S	175‐280	[284, 285]
	CuS	Cu(acac)2	Solid S	140‐160	[281]
	CuS	Cu(thd)2	H2S	125‐175	[283, 284, 285]
Nb	NbO2	Nb‐TBTEA	H2O	200	[241]
	NbO2	Nb(OEt)5	H2O	170‐300	[243]
	NbO2+Nb2O5	Nb(OEt)5	H2O	170‐300	[243]
	Nb2O5	Nb‐TBTEA	H2O	200	[241]
	Nb2O5	tBuN=Nb(NEt2)3	H2O+O3	150‐325	[242]
	Nb2O5	tBuN=Nb(NMeEt)3	H2O+O3	150‐325	[242]
	Nb2O5	Nb(OEt)5	H2O	170‐300	[243]
Mo	MoO2	MoO2(thd)2	O3	225	[238]
	MoO2	Mo(NMe2)4	O2	150	[239]
	MoOx (2<x<3)	Mo(NMe2)4	H2O+O3	80	[240]
	MoO3	MoO2(thd)2	O3	225	[238]
	MoO3	Mo(NMe2)4	O3	150	[239]
Sn	SnO	Sn(dmamb)2	H2O	100‐210	[244, 245]
	SnO	Sn(dmamp)2	H2O	150‐210	[246, 247, 248, 258]
	SnO	Sn(edpa)2	H2O	70‐300	[249]
	SnO	Sn(iPr2fAMD)2	H2O	220‐240	[250]
	SnO	Sn(N(SiMe3)2)2	H2O	100‐250	[251]
	SnO	Sn(TAA)2	H2O	100‐180	[252]
	SnO2	Sn(edpa)2	O2 plasma	70‐300	[249]
	SnO2	Sn(N(SiMe3)2)2	O3	80‐200	[251]
	SnO2	Sn(NMe2)4	H2O	30‐300	[253, 254]
	SnO2	SnCl4	H2O	300‐600	[255, 256, 257]
	SnO2	Sn(dmamp)2	H2O2	100‐200	[258]
	SnS	Sn(NMe2)4	H2S	100‐240	[292, 295, 297]
	SnS	Sn(η2‐((NBut)CHMeCHMe(NBut)))2	H2S	50‐125	[293]
	SnS	Sn(η2‐MeC(NPri)2)2	H2S	100‐250	[293]
	SnS	Sn(dmamp)2	H2S plasma	210	[298]
	SnS2	Sn(NMe2)4	H2S	140‐180	[292, 295]
	SnS2	Sn(dmamp)2	H2S plasma	150‐240	[294]
	SnSe	Sn(NMe2)4	Se(SiMe3)2	170‐190	[299]
	SnSe	SnCl4	Sn(SiEt3)2	180‐200	[300]
	SnSe	SnEt4	H2Se	250‐400	[301]
	SnSe2	Sn(NMe2)4	Se(SiMe3)2	110‐150	[299]
W	W2O3	W2(NMe2)6	H2O	160‐200	[268]
	WO3	WCO6	H2O	300	[269, 270]
	WO3	W(NtBu)2(NMe2)2	H2O	250‐350	[271, 272]
	WO3	WO2(tBuAMD)2	H2O	150‐270	[273]
	W2N	W(CO)6	NH3	180‐195	[314]
	WxN (1.1≤x≤1.4)	WF6+B2H6	NH3	300	[315]
	WN	WCl5	N2/H2 plasma	250	[317]
	WN1.56	W(CO)(3‐hexyne)3	N2/NH3 plasma	250	[313]
	WN	WCl5	N2/H2 plasma	250	[317]
	W2N	WCl5	N2/H2 plasma	250	[317]
	WN3.32	W(NtBu)2(NMe2)2	N2 plasma	250	[316]
	WxN (1.13≤x≤1.93)	W(NtBu)2(NMe2)2	NH3 plasma	174‐400	[316]
	WxN (1.26≤x≤3.19)	W(NtBu)2(NMe2)2	N2/H2 plasma	250	[316]
	WxN (1.9≤x≤3.76)	W(NtBu)2(NMe2)2	N2/H2 plasma	250‐400	[316]
Ta	Ta2N	Ta(NMe2)5	H2 plasma	225	[307]
	TaN	Ta(NiPr)(NEtMe)3	N2/H2 plasma	400	[306]
	TaN	TaCl5	NH3+Zn	400‐500	[309]
	TaN	TaF5	H2 plasma+NH3	<250	[311]
	TaN	TaF5	H2 plasma+NH3+H2	200	[312]
	Ta2N3	Ta(NMe2)5	N2/H2 plasma	225	[307]
	Ta3N5	Ta(NiPr)(NEtMe)3	N2/H2 plasma	400	[306]
	Ta3N5	TaCl5	NH3	200‐500	[309]
	Ta3N5	TaF5	H2 plasma+NH3	300‐350	[311]
	Ta3N5	TaF5	H2 plasma+NH3/H2 plasma	200	[312]
	TaNx (0.88≤x≤1.35)	TaCp(=NtBu)(NEt2)2	N2/H2 plasma	390‐450	[308]
	TaNx (1≤x≤1.7)	TaF5	NH3	350	[310]
	TaNx (1.53≤x≤1.79)	Ta(=NtBu)(NEt2)3	NH3	240‐280	[308]
Bi	BiSe	Bi(NMe2)3	Se(SnMe3)2	90‐120	[274]
	BiSe2	Bi(NMe2)3	Se(SnMe3)2	150	[274]

The ability to stabilize different valence states within a single‐material system offers significant opportunities for tuning electrical, optical, and chemical properties, thereby increasing the flexibility of material design. A representative case is tin oxide: while SnO₂, the thermodynamically stable phase, behaves as an n‐type semiconductor, the metastable SnO exhibits p‐type conductivity [244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 256, 257]. This polarity contrast in the same material system offers new possibilities of novel electronic devices, including p–n junctions and complementary electronics.

In ALD, several approaches have been employed to selectively control metal valence states, including variations in the deposition temperature, the choice of precursors and reactants, and post‐deposition treatments. In this section, we highlight these strategies and examine representative examples of valence‐controlled metastable‐phase formation via ALD.

3.1. Control of Deposition Temperature

The deposition temperature is a critical factor that determines the valence states of metal cations in ALD‐grown films. Changes in the deposition temperature can alter the precursor reactivity, decomposition behavior, oxidation potential, and volatility of specific elements, potentially resulting in variations in the film composition. Table 7 summarizes the representative systems in which the valence states of the metal ions in the ALD‐grown films varied with the deposition temperature [201, 202, 209, 212, 227, 234, 236, 237, 274, 275, 284, 285, 288, 292, 295, 299, 303, 304, 308, 311, 316]. In certain cases, an increase in the deposition temperature can lead to a higher valence state of the metal ions in the films. For instance, in ALD using FeCp₂ and O₂, Fe₂O₃ and metastable Fe₃O₄ comprising Fe²⁺ and Fe³⁺ ions were formed at relatively low temperatures (350–500°C), whereas only Fe₂O₃ was formed at temperatures above 500°C [227]. A similar trend was observed for the ALD of Bi–Se compounds using Bi(NMe₂)₃ and Se(SnMe₃)₂. The metastable BiSe phase was stabilized in the temperature range of 90–120°C, whereas the stable Bi₂Se₃ phase was observed at elevated temperatures. [274] In another example, using Fe(ⁱPrAMD)₂ and Se(SiEt₃)₂, FeSe was formed at 350°C. However, when the deposition temperature exceeded 370°C, the thermal decomposition of Fe(ⁱPrAMD)₂ reduced the supply of Fe ions, leading to the formation of metastable Fe₃Se₄ [288].

TABLE 7.

Representative examples of ALD processes for controlling valence states in metastable multivalent compounds via the modulation of deposition temperature. This table summarizes how arranging oxidation states as a function of growth temperature makes it clear which thermal regimes favor more reduced or more oxidized valence states in each material system.

Element	Material	Precursor	Reactant	Temperature (°C)	References
Ti	TiS3	Ti(NMe2)4	H2S plasma	100	[303]
	TiS2	Ti(NMe2)4	H2S plasma	150‐200	[303]
V	VS4	V(amd)3	H2S	150‐200	[304]
	V2S3	V(amd)3	H2S	225‐250	[304]
Cr	CrOx (Cr(III)/Cr(VI) = 0.5‐0.7)	CrO2Cl2	H2O	100	[212]
	CrOx (Cr(III)/Cr(VI) = 2)	CrO2Cl2	H2O	150	[212]
	CrOx (Cr(III)/Cr(VI) = 1.2‐1.3)	CrO2Cl2	H2O+H2O2	150	[212]
	CrOx (Cr(III)/Cr(VI) = 10‐12)	CrO2Cl2	H2O+H2O2	200	[212]
Mn	Mn3O4	Mn(tBuAMD)2	H2O	150	[234]
	Mn2O3+MnO	Mn(tBuAMD)2	H2O	175‐250	[234]
	MnO2	Mn(thd)3	O3	140‐230	[236]
	Mn3O4	Mn(thd)3	O3	≥240	[236]
	Mn2O3	Mn2(CO)10	O3	60‐100	[237]
	Mn3O4	Mn2(CO)10	O3	120‐160	[237]
Fe	Fe2O3+Fe3O4	FeCp2	O2	350‐500	[227]
	Fe2O3	FeCp2	O2	>500	[227]
	FeSe	Fe(iPrAMD)2	Se(SiEt3)2	350‐370	[288]
	Fe3Se4	Fe(iPrAMD)2	Se(SiEt3)2	390‐450	[288]
Co	Co3O4	CCTBA	O3	68	[201]
	Co3O4+CoO	CCTBA	O3	80‐138	[201]
	Co3O4+CoO	Co(iPr2DAD)2	O2	125‐250	[202]
	Co3O4	Co(iPr2DAD)2	O2	≥265	[202]
	Co3O4	CoCp2	O3	150‐280	[209]
	Co3O4+CoO	CoCp2	O3	331	[209]
Cu	CuSe	Cu(II) pivalate	(Et3Si)2Se	165	[275]
	Cu2Se	Cu(II) pivalate	(Et3Si)2Se	300	[275]
	CuS	Cu(thd)2	H2S	125‐175	[284, 285]
	Cu1.8S	Cu(thd)2	H2S	175‐280	[284, 285]
Sn	SnS2	TDMASn	H2S	140‐150	[292]
	SnS	TDMASn	H2S	160‐180	[292]
	SnS2	TDMASn	H2S	180	[295]
	SnS	TDMASn	H2S	240	[295]
	SnSe2	Sn(NMe2)4	Se(SiMe3)2	110‐150	[299]
	SnSe	Sn(NMe2)4	Se(SiMe3)2	170‐190	[299]
W	WN	W(NtBu)2(NMe2)2	NH3 plasma	175	[316]
	W2N	W(NtBu)2(NMe2)2	NH3 plasma	400	[316]
	W2N	W(NtBu)2(NMe2)2	N2/H2 plasma	250	[316]
	W3.76N	W(NtBu)2(NMe2)2	N2/H2 plasma	400	[316]
Ta	TaN0.88	TaCp(=NtBu)(NEt2)2	NH3	390	[308]
	TaN1.35	TaCp(=NtBu)(NEt2)2	NH3	450	[308]
	TaN1.53	Ta(=NtBu)(NEt2)3	NH3	240	[308]
	TaN1.79	Ta(=NtBu)(NEt2)3	NH3	280	[308]
	TaN	TaF5	H2 plasma+NH3	<250	[311]
	Ta3N5	TaF5	H2 plasma+NH3	300‐350	[311]
Bi	BiSe	Bi(NMe2)3	Se(SnMe3)2	90‐120	[274]
	BiSe2	Bi(NMe2)3	Se(SnMe3)2	150	[274]

In contrast, the valence states of the metal cations in ALD‐grown films decrease with increasing deposition temperature. In the ALD process using Mn₂(CO)₁₀ and O₃, Mn₂O₃ was formed at low temperatures (60–100°C), whereas Mn₃O₄, which contains both Mn²⁺ and Mn³⁺, was formed at high temperatures (120–160°C) [237]. Similarly, in the ALD of cobalt oxides using CoCp₂ and O₃, stoichiometric Co₃O₄ (with Co²⁺ and Co³⁺) was formed at 150–285°C. However, at 331°C, incomplete oxidation owing to the decomposition of CoCp₂ increased the proportion of Co²⁺, leading to a mixture of Co₃O₄ and metastable CoO [209]. This tendency is particularly notable in materials containing volatile anions, such as S or Se, where higher temperatures can induce anion desorption and promote a reduction in the valence states of metal cations. For instance, in ALD using Sn(N(CH₃)₂)₄ and Se(Si(CH₃)₃)₂, the stable SnSe₂ phase was formed at 110–150°C, whereas the high‐temperature phase SnSe emerged above 170°C [299]. In ALD using V(amd)₃ and H₂S, metastable VS₄ was obtained at 150–200°C owing to the oxidation of V³⁺ by disulfide (S₂ ²⁻) species generated from reactions between H₂S and –SH ligands [304]. However, at higher temperatures (225–250°C), the thermal decomposition of S₂ ²⁻ inhibited this reaction path, and the surface reaction proceeded via only H₂S, leading to the formation of stable V₂S₃ [304].

Changes in the deposition temperature can also alter the chemical nature of the reactants themselves [236, 303]. This effect was demonstrated in the ALD of metastable TiS₃ containing S₂ ²⁻ [303]. In ALD using Ti(NMe₂)₄ and H₂S plasma, TiS₃ was synthesized at 100°C. However, the disulfide species decomposed at 150–200°C, thereby yielding TiS₂ [303]. A similar temperature‐driven transition was observed in the ALD of manganese oxides using Mn(thd)₃ and O₃ [236]. The MnO₂ phase with Mn⁴⁺ was formed at the low temperatures of 140–230°C, whereas above 240°C, the reduced lifetime of O₃ diminished its oxidizing power, leading to the formation of metastable Mn₃O₄, which contains both Mn²⁺ and Mn³⁺ [236].

3.2. Precursor‐Based Control of the Valence State of Films

In ALD reactions based on ligand‐exchange mechanisms, the valence state of the metal center in the precursor is preserved in the resulting film. This occurs when the ligands in the precursor are removed without combustion, thereby preventing further oxidation of the metal ions. This approach has enabled the synthesis of metastable phases of compounds containing multivalent transition metals, such as V, Mn, Co, Cu, Sn, and W (Table 8) [113, 115, 116, 118, 208, 216, 219, 220, 233, 245, 246, 247, 248, 249, 253, 254, 255, 256, 257, 258, 261, 262, 269, 277, 280, 282, 284, 290, 291, 292, 293, 294].

TABLE 8.

Representative examples of ALD processes for controlling valence states in metastable multivalent compounds by tailoring the valence state of the metal in the precursor. This table summarizes how choosing precursors with different formal oxidation states or ligand environments systematically biases the resulting films toward particular valence states, providing practical design rules for precursor selection when targeting metastable oxidation states.

Element	Material	Metal oxidation state in precursor	Precursor	Reactant	Temperature (°C)	References
V	VO2	4+	V(NEtMe)4	O3	150	[260, 261]
Mn	MnO	2+	Mn(EtCp)2	H2O	100‐300	[232]
Co	CoO	2+	Co(BTSA)2(THF)	H2O	75‐250	[207]
Cu	Cu2O	1+	Cu(PnBu3)2(acac)	H2O + O2	110‐125	[215]
	Cu2O	1+	[Cu(sBu‐amd)]2	H2O	180	[218]
	Cu2O	1+	Cu(dmamb)2	H2O	140‐160	[219]
	Cu2S	1+	[Cu(sBu‐amd)]2	H2S	130‐150	[276, 277]
	Cu2S	1+	Cu2DBA	H2S	135	[278]
	Cu2S	1+	CuAMD	H2S	135	[279]
	CuS	2+	Cu(acac)2	Solid S	140‐160	[281]
	CuS	2+	Cu(thd)2	H2S	125‐160	[283]
Sn	SnO	2+	Sn(dmamb)2	H2O	100‐210	[244, 245]
	SnO	2+	Sn(dmamp)2	H2O	150‐210	[246, 247, 248]
	SnO	2+	Sn(iPr2fAMD)2	H2O	220‐240	[250]
	SnO	2+	Sn(TAA)2	H2O	100‐210	[252]
	SnO2	4+	TDMASn	H2O	30‐300	[253, 254]
	SnO2	4+	SnCl4	H2O	300‐600	[255, 256, 257]
	SnS	2+	Sn(acac)2	H2S	80‐175	[116, 289]
	SnS	2+	Sn(dmamp)2	H2S	90‐285	[115, 290]
	SnS	2+	Sn(dmpa)2	H2S	25‐250	[118]
	SnS	2+	Sn(amd)2	H2S	100‐200	[291]
	SnS	2+	Sn(iPr2fAMD)2	H2S	65‐200	[113]
	SnS	2+	Sn(η2 ‐((NBut)CHMeCHMe(NBut)))2	H2S	50‐125	[293]
	SnS	2+	Sn(η2 ‐MeC(NPri)2)2	H2S	100‐250	[293]
W	W2O3	3+	W2(NMe2)6	H2O	160‐200	[268]

Among these systems, tin oxides are representative materials in which the resulting phase—either SnO or SnO_2—is determined by the valence state of the metal in the precursor. Various precursors that contain either Sn²⁺ or Sn⁴⁺ ions are available [244, 245, 246, 247, 248, 252, 253, 254, 255, 256, 257]. Precursors containing Sn⁴⁺, such as SnCl₄, SnI₄, and TDMASn, react with H₂O to form the stable SnO₂ [253, 254, 255, 256, 257]. In contrast, precursors containing Sn²⁺, including Sn(dmamp)₂, Sn(dmamb)₂, Sn(edpa)₂, and Sn(OtAmyl)₂, react with H₂O to yield the metastable SnO [244, 245, 246, 247, 248, 252]. A similar trend was observed in the formation of tin sulfide films [113, 115, 116, 118, 289, 290, 291, 292, 293]. The ALD with a Sn⁴⁺ precursor of TDMASn and H₂S leads to the formation of n‐type SnS₂ at 60–150°C [292], whereas the use of Sn²⁺ precursors, such as Sn(η²‐((NtBu)CHMeCHMe(NtBu)))₂, Sn(dmamp)₂, Sn(dmpa)₂, Sn(ⁱPrAMD)₂, and Sn(ⁱPrFMD)₂, yields metastable p‐type SnS under similar conditions [113, 115, 116, 118, 289, 290, 291, 292, 293].

However, this precursor‐based strategy for controlling valence states may become ineffective at elevated temperatures. At high temperatures, the partial thermal decomposition of the precursor or enhanced reactivity with the co‐reactant may lead to unintended oxidation. For example, in ALD using Sn(dmamb)₂ with Sn²⁺, SnO was formed at 100–150°C, but SnO₂ also appeared in the films deposited at 200°C [244].

Conversely, a certain ALD process intentionally exploit the thermal instability of a precursor. In such case, a thermodynamically unstable precursor self‐decomposes via a disproportionation reaction in an inert atmosphere, allowing metal deposition without a co‐reactant [224]. A notable example is the deposition of metallic Cu using the low‐stability precursor Cu(I)(hfac)(tmvs). Above 60°C, it undergoes the following disproportionation reaction:

$$2\mathrm{Cu}\left(\mathrm{I}\right)\left(\mathrm{hfac}\right)\left(\mathrm{tmvs}\right)=\mathrm{Cu}\left(0\right)+\mathrm{Cu}\left(\mathrm{II}\right){\left(\mathrm{hfac}\right)}_2+2(\mathrm{tmvs})$$ (1)

Through this reaction, the byproducts of Cu(II)(hfac)₂ and tmvs are volatilized at 160°C under a N₂ atmosphere, enabling the deposition of high‐purity Cu metal [224].

3.3. Reactant‐Based Control of the Valence State of Films

The valence state of metal ions in ALD‐grown films can also vary depending on the choice of the co‐reactant. Table 9 summarizes representative ALD processes in which these effects were observed [115, 203, 204, 211, 217, 220, 224, 226, 231, 235, 240, 249, 251, 258, 286, 287, 290, 294, 307, 309, 310, 312, 316, 317]. Such differences are likely attributed to the adoption of different reaction pathways depending on the choice of the co‐reactant. For example, as discussed above, ALD reactions using Sn precursors containing Sn²⁺ with H₂O or H₂S as co‐reactants typically yield SnO or SnS films via the ligand‐exchange mechanism [113, 115, 116, 118, 244, 245, 246, 247, 248, 252, 289, 290, 291, 293]. However, when other reactants, such as O₂ plasma, O₃, or H₂S plasma, are used with the same Sn²⁺ precursors, SnO₂ or SnS₂ films are formed owing to the increase in the valence state of Sn ions (Figure 5a) [249, 251, 294]. Furthermore, the valence states of metal ions in ALD‐grown films can vary, even with changes in the reactant dose supplied during each ALD cycle. For example, in the ALD of iron oxides using Fe(Cp)₂ and O₂ at 400°C, increasing the O₂ pulse time from 1 s to 4 s induced a phase transition from Fe₃O₄ to Fe₂O₃, reflecting an increase in the valence state of Fe ions [231]. In addition, in the ALD of TaN_x using TaF₅ and H₂/N₂ plasma, the valence state of Ta ions in the films has been reported to vary with the N₂/(H₂+N₂) ratio in the reactant plasma [311].

TABLE 9.

Representative examples of ALD processes for controlling valence states in metastable multivalent compounds through the selection of the reactants. This table summarizes how the choice of oxidizing, reducing, or complex‐forming reactants, together with their dose conditions, shifts the balance between competing valence states in the resulting films.

Element	Material	Precursor	Reactant	Temperature (°C)	Refs.
Cr	CrOx (Cr(III)/Cr(VI)>10)	CrO2Cl2	CH3OH	150	[211]
	CrOx (Cr(III)/Cr(VI)=2)	CrO2Cl2	H2O	150	[211]
	CrOx (Cr(III)/Cr(VI)=1.2‐1.3)	CrO2Cl2	H2O2	150	[211]
Mn	MnO	Mn(thd)3	NH3 plasma	180	[235]
	Mn3O4+MnO	Mn(thd)3	H2 plasma	180	[235]
	Mn3O4	Mn(thd)3	H2O	180	[235]
	MnO2	Mn(thd)3	O3	180	[235]
Fe	Fe3O4	Fe(2,4‐C7H11)2	H2O2	120	[226]
	Fe2O3	Fe(2,4‐C7H11)2	O3	120	[226]
	Fe	Fe(tBuAMD)2	H2	250	[203]
	FeO	Fe(tBuAMD)2	H2O	250	[203]
	Fe3O4	FeCp2	O2 (1s pulse)	400	[231]
	Fe2O3	FeCp2	O2 (4s pulse)	400	[231]
	FeS2	Fe(amd)2	H2S plasma	80‐200	[286]
	FeS	Fe(amd)2	H2S	80‐250	[287]
Co	Co	Co(iPrAMD)2	H2	350	[203]
	CoO	Co(iPrAMD)2	O2	250	[203]
	CoO	Co(iPrAMD)2	H2O	150‐250	[204]
	Co3O4	Co(iPrAMD)2	O3	225‐250	[204]
Cu	Cu	[Cu(sBu‐amd)]2	H2 plasma	150	[217]
	Cu2O	[Cu(sBu‐amd)]2	H2 plasma: O2 plasma=3:1	150	[217]
	CuO	[Cu(sBu‐amd)]2	O2 plasma	150	[217]
	Cu	(hfac)Cu(I)(DMB)	H2 plasma: O2 plasma=7:1	100	[220]
	Cu2O	(hfac)Cu(I)(DMB)	H2 plasma: O2 plasma=3:1	100	[220]
	CuO	(hfac)Cu(I)(DMB)	H2 plasma: O2 plasma=1:1	100	[220]
	Cu	Cu(I)(hfac)(tmvs)	N2	160	[224]
	Cu2O	Cu(I)(hfac)(tmvs)	H2O	220	[224]
	CuO	Cu(I)(hfac)(tmvs)	O3	260	[224]
Mo	MoOx (2<x<3)	Mo(NMe2)4	H2O cycles: O3 cycles=1:0, 4:1, 1:1, 0:1	80	[240]
Sn	SnO	Sn(edpa)2	H2O	70‐300	[249]
	SnO2	Sn(edpa)2	O2 plasma	70‐300	[249]
	SnO	Sn(N(SiMe3)2)2	H2O	100‐250	[251]
	SnO2	Sn(N(SiMe3)2)2	O3	80‐200	[251]
	SnO	Sn(dmamp)2	H2O	200	[258]
	SnO2	Sn(dmamp)2	H2O2	100‐200	[258]
	SnS	Sn(dmamp)2	H2S	90‐285	[115, 290]
	SnS2	Sn(dmamp)2	H2S plasma	150‐240	[294]
W	W3.32N	W(NtBu)2(NMe2)2	N2 plasma	250	[316]
	WxN (1.26≤x≤3.19)	W(NtBu)2(NMe2)2	N2/H2 plasma (N2:H2=1.1‐16:1)	250	[316]
	WN	WCl5	N2/H2 plasma (N2:H2=1:3)	250	[317]
	W2N	WCl5	N2/H2 plasma (N2:H2=1:5)	250	[317]
	W2.4N	WCl5	N2/H2 plasma (N2:H2=1:10)	250	[317]
Ta	Ta2N	Ta(NMe2)5	H2 plasma	225	[307]
	Ta2N3	Ta(NMe2)5	N2/H2 plasma (N2:H2=1:1)	225	[307]
	TaN	TaCl5	NH3+Zn	400‐500	[309]
	Ta3N5	TaCl5	NH3	200‐500	[309]
	TaNx (1≤x≤1.7)	TaF5	N2/H2 plasma (PN2/PH2 = 0.014‐4)	350	[310]
	TaN	TaF5	H2 plasma+NH3+H2	200	[312]
	Ta3N5	TaF5	H2 plasma+NH3/H2 plasma	200	[312]

FIGURE 5.

Reactant‐based strategies for controlling valence states in ALD. (a) Schematic of valence state evolution in the ALD of Sn‐based compounds using precursors with Sn²⁺, depending on the type of reactant. (b) Strategy for modulating the valence states of Mo ions in MoO_x via alternating ALD cycles using H₂O and O_3.

Beyond simply selecting a single reactant to control the valence state of metal ions, more precise control of the valence states of metal ions in the films can be achieved by alternating different co‐reactants in the ALD processes [203, 204, 217, 224, 226, 235, 249, 251, 286, 307, 309, 312, 316, 317]. A representative example is the ALD of molybdenum oxides using Mo(NMe₂)₄, where H₂O and O₃ serve as co‐reactants to form MoO₂ or MoO₃, respectively. By adjusting the cycle ratio of H₂O‐based to O₃‐based ALD steps, films with intermediate valence states were systematically synthesized, as shown in Figure 5b [240]. Similarly, in the deposition of copper oxides, a two‐step process, in which metallic Cu is formed from (hfac)Cu(I)(DMB) and H₂ plasma and subsequently oxidized by O₂ plasma, has been proposed to tune the valence states of Cu ions. By varying the ratio of the deposition to the oxidation steps, the film composition can be precisely tuned to span the entire range from Cu to Cu₂O and CuO [224].

3.4. Post‐Deposition Treatment for Tailoring Valence States

Post‐deposition treatments offer a powerful approach to tune the valence state in ALD‐grown films, particularly when the desired metastable valence states cannot be achieved under conventional ALD conditions. By decoupling the deposition and oxidation/reduction processes, such treatments enable the precise modulation of the valence states. There are several key parameters, such as temperature, atmosphere, pressure, and annealing time. These approaches are particularly effective for transition metal compounds, including V, Mn, Fe, Co, Cu, Sn, Nb, and Mo, as summarized in Table 10 [200, 205, 206, 208, 214, 216, 229, 230, 238, 239, 241, 243, 259, 261, 262, 263, 264, 265, 267, 297, 298, 305, 306].

TABLE 10.

Representative examples of ALD processes for controlling valence states in metastable multivalent compounds through post‐deposition treatments. This table summarizes how annealing ambient, temperature, and duration correlate with changes in oxidation state and phase, summarizing general strategies for converting as‐deposited films into desired metastable valence states by thermal or chemical treatments.

Element	Material	Precursor	Reactant	Temperature (°C)	Post‐treatment	Refs.
V	V2O3	VO(acac)2	O3	200	PDA 700 °C H2	[259]
	V2O5	VO(acac)2	O3	200	—	[259]
	V6O13	V(NEtMe)4	O3	150	PDA 450 °C 10 Pa O2	[261]
	VO+VO2	V(NEtMe)4	O3	150	—	[263]
	VO2	V(NEtMe)4	O3	150	PDA 585 °C O2	[263]
	V2O5	V(amd)3	O3	200	—	[200]
	V2O5	V(amd)3	H2O+O2	200	PDA 450 °C N2	[200]
	V2O5	V(amd)3	H2O2	200	PDA 450 °C N2	[200]
	V2O3+VO2	V(amd)3	H2O2	200	PDA 350 °C H2	[200]
	VO2	V(amd)3	H2O2+H2	200	PDA 350 °C H2	[200]
	VO2	V(NEtMe)4	O3	150	PDA 450 °C N2	[265]
	V2O5	V(NEtMe)4	O3	150	PDA 450 °C Air	[265]
	VO2	V(NEtMe)4	H2O	150	PDA 450 °C N2	[265]
	V2O5	V(NEtMe)4	H2O	150	PDA 450 °C Air	[265]
	VO2	V(NEtMe)4	O3	150	PDA 420 °C He/3.7 Pa O2	[262]
	VO2	V(NEtMe)4	H2O	150	PDA 450 °C He/18 Pa O2	[262]
	V6O13	V(NEtMe)4	O3	150	PDA 550 °C He/3.7 Pa O2	[262]
	V4O9	V(NEtMe)4	H2O	150	PDA 356 °C Air	[262]
	V3O7	V(NEtMe)4	O3	150	PDA 560 °C He/48 Pa O2	[262]
	V2O5	V(NEtMe)4	O3+H2O	150	PDA 500 °C Air	[262]
	VO2	VO(OiPr)3	H2O	60‐90	Ar plasma/550 °C vacuum	[264]
	VnO2n‐1	VO(OiPr)3	H2O	60‐90	H2 plasma/550 °C vacuum	[264]
	V2O5	VO(OiPr)3	H2O	135	PDA 300‐500 °C Air	[267]
	VO2	VO(OiPr)3	H2O	135	PDA 500 °C H2	[267]
	VS	V(NEtMe)4	H2S	200‐300	PDA 500 °C H2S	[305]
Fe	Fe2O3	Fe2(OtBu)6	H2O	130‐170	—	[229]
	Fe3O4	Fe2(OtBu)6	H2O	130‐170	PDA 400 °C H2	[229]
	Fe2O3	FeCp(CptBu)	O2 plasma	150‐250	—	[230]
	Fe3O4	FeCp(CptBu)	O2 plasma	150‐250	PDA 385 °C H2	[230]
	Fe	FeCp(CptBu)	O2 plasma	150‐250	PDA 430 °C H2	[230]
Co	Co3O4+CoO	Co2(CO)8	O3	50	—	[205]
	Co3O4	Co2(CO)8	O3	50	PDA 500 °C Air	[205]
	CoO	Co(AMD)2	H2O	180‐270	—	[206]
	Co	Co(AMD)2	H2O	180‐270	Atomic Deuterium 220 °C/Al cap 200 °C	[206]
	CoO	CoCl2(TMEDA)	H2O	225‐300	—	[208]
	Co	CoCl2(TMEDA)	H2O	225‐300	PDA 250 °C H2	[208]
Cu	Cu2O+CuO	Cu(hfac)(dmb)	O3	100	PDA 200‐500 °C Air	[214]
	CuO	Cu(dmap)2	O3	80‐140	—	[216]
	Cu2O	Cu(dmap)2	O3	80‐140	PDA 350‐700 °C N2	[216]
Nb	Nb2O5	Nb‐TBTEA	H2O	200	—	[241]
	NbO2	Nb‐TBTEA	H2O	200	PDA 1000 °C H2	[241]
	Nb2O5	Nb(OEt)5	H2O	170‐300	—	[243]
	Nb2O5+NbO2	Nb(OEt)5	H2O	170‐300	Pulsed laser annealing H2	[243]
Mo	MoO3	MoO2(thd)2	O3	25	PDA 450 °C Air	[238]
	MoO2	MoO2(thd)2	O3	25	PDA 450 °C H2	[238]
	MoO3	Mo(NMe2)4	O3	150	PDA 400 °C O2	[239]
	MoO2	Mo(NMe2)4	O2	150	PDA 400 °C H2	[239]
Sn	SnS	TDMASn	H2S	100	PDA 400 °C vacuum	[297]
	SnS	Sn(dmamp)2	H2S plasma	210	Sn(dmamp)2 feeding 270 °C	[298]
Ta	Ta3N5	Ta(NiPr)(NEtMe)3	N2/H2 plasma	400	—	[306]
	TaN	Ta(NiPr)(NEtMe)3	N2/H2 plasma	400	PDA 850 °C vacuum	[306]

A common approach involves thermal annealing under an oxidative or a reductive atmosphere [200, 205, 206, 208, 214, 216, 229, 230, 238, 241, 243, 259, 261, 262, 263, 264, 265, 266, 267, 297, 298, 305, 306]. Oxidative annealing increases the valence state of metal ions, whereas reductive annealing decreases that of metal ions. For instance, a mixture of VO and V₂O₅ grown by ALD was converted into VO₂ through annealing in O₂ at 585°C [263]. Similarly, Fe₂O₃ films prepared by ALD were reduced to Fe₃O₄ upon annealing in H₂/Ar at 400°C [229]. The incorporation of plasma into the post‐deposition process can further enhance the effectiveness of the valence state control. For instance, amorphous VO_x films with oxidation states close to V₂O₅ were treated with remote H₂ or Ar plasma at room temperature, followed by vacuum annealing at 550°C, resulting in the formation of V_nO_2n‐1 and VO₂ phases, respectively [264].

In some cases, the final valence state is governed by the initial film composition and response to post‐treatment [200, 239, 240, 262, 265]. This interplay enables the highly selective synthesis of multiple phases from a single precursor system. For example, various V₂O₃, VO₂, and V₂O₅ phases have been selectively obtained by using a V(amd)₃ precursor with different oxidants, including O₃, H₂O+O₂, H₂O₂, or H₂O₂+H₂, followed by annealing in N₂ or H₂ [200]. Similarly, MoO_x films grown at 150°C using Mo(NMe₂)₄ and either O₂ or O₃ were post‐annealed in either H₂ or O₂ at 400°C, enabling compositional tuning across a wide stoichiometric range of x = 2.26–3.14 [239].

3.5. Noble Metals and Their Oxides

Noble metals and their corresponding oxides can be selectively grown by ALD, enabling the controlled tuning of their valence states. As illustrated in Figure 6a, the ALD of noble metals typically employs alternating pulses of a metal precursor and an oxidizing reactant. During the injection step of a metal precursor, the metal precursor is adsorbed onto the reaction surface. The ligands of the precursor are partially removed through a reaction with pre‐adsorbed oxygen in the subsurface, reducing the subsurface. In the subsequent injection step of the oxidizing reactant, the oxidizing reactant removes the remaining ligands of the adsorbed metal precursor and re‐oxidizes the subsurface. Through repeated cycles of these two steps, film growth proceeds via alternating surface reduction and oxidation, highlighting the redox‐driven nature of noble‐metal ALD [347, 348, 349].

FIGURE 6.

Redox‐driven ALD mechanism for noble metals and their oxides. (a) Schematic of the phase selection mechanism in the ALD processes of noble metals and their oxides. (b) Schematic illustrating how the balance between oxidation by the oxidant and reduction by the metal precursor governs the phase of the grown films.

In ALD processes involving cyclic redox reactions, the extents of oxidation and reduction during each half‐cycle are not necessarily identical. Oxide films form when the number of oxygen atoms incorporated into the subsurface during the oxidizing reactant pulse exceeds the number of oxygen atoms removed during the subsequent metal precursor pulse. Conversely, if fewer oxygen atoms are incorporated than reduced, metallic films are obtained. The balance between the oxidation and reduction reactions is the key factor determining the final chemical state of the film, as illustrated in Figure 6b.

The redox balance in the ALD process is influenced by several parameters, including the type of metal precursor, oxidizing reactant, and growth temperature. Among these, the growth temperature is generally regarded as the most decisive parameter. As shown in Table 11, ALD processes involving noble metal elements, such as Ru, Ir, Pt, Rh, and Pd, enable the formation of oxides at relatively low temperatures [322, 323, 325, 326, 329, 330, 331, 332, 333, 341, 344, 346]. At higher temperatures, the reduction reaction induced by the precursor becomes dominant, resulting in the formation of metallic films [322, 323, 325, 326, 335, 340, 342, 345]. The critical temperature at which the transition occurs varies with the metal. For Ru and Ir, oxide formation can persist up to relative high temperatures near 270°C [318, 320, 322, 324, 325, 326, 327, 328, 334, 337], whereas for Pt, Pd, and Rh, the transition to metallic films typically occurs at significantly lower temperatures, often below 180°C [339, 341, 344, 346]. The oxidizing reactant also influences phase selection. Stronger oxidants, such as O₃ and O₂ plasma, are more effective than O₂ in promoting oxide formation and can increase the upper temperature limit at which their corresponding oxides are still formed. Furthermore, the dose of the oxidizing reactant, which is determined by the pulse duration or partial pressure [318, 319, 320, 324, 325, 326, 327, 328, 329, 330, 334, 337, 339, 345], can affect the phase selection. The oxide phase is typically formed only when a sufficiently large dose of the oxidant is injected. Although the type of metal precursor has less influence on the phase selection compared with the growth temperature and oxidant, increasing the precursor dose can enhance the reduction reaction during the precursor injection, thereby favoring metal formation [318, 319]. Based on these trends, several studies have demonstrated selective phase control by tuning the dose ratio between the metal precursor and the oxidant [318, 319, 329].

TABLE 11.

Representative examples of ALD processes for noble metals and their oxides. This table summarizes how growth temperature, oxidant strength and dose, and precursor dose control the redox balance in noble‐metal ALD and govern the transition between oxide and metallic films.

Element	Precursor	Oxidant	Phase	Temperature (°C)	Oxidant condition	References.
Ru	Ru(EtCp)2	O2	Ru	270	tO2/tRu <2.5	[318, 319]
	Ru(EtCp)2	O2	RuO2	270	tO2/tRu >3.3	[318, 319]
	Ru(EtCp)2	O2	RuO2	265	Continuous O2 flow	[320]
	Ru(EtCp)2	O2	Ru	300	—	[321]
	Ru(EtCp)2	O2 plasma	RuO2	300	—	[321]
	Ru(EtCp)2	O radical	Ru	300‐340	O* from O2/Ar with 2500W	[322]
	Ru(EtCp)2	O radical	RuO2	200‐260	O* from O2/Ar with 2500W	[322]
	Ru(EtCp)2	O2 plasma	Ru	375‐400	60 scm O2 with 75W	[323]
	Ru(EtCp)2	O2 plasma	RuO2	300‐350	60 scm O2 with 75W	[323]
	dicarbonyl‐bis(5‐methyl‐2,4‐hexanediketonato)Ru(II)	O2	Ru	283	tO2/tRu: 1	[324]
	dicarbonyl‐bis(5‐methyl‐2,4‐hexanediketonato)Ru(II)	O2 plasma	RuO2	283	tO2/tRu: 12	[324]
	Ru(DMBD)(CO)3	O2	Ru	255‐280	O2 2 s	[325]
	Ru(DMBD)(CO)3	O2	RuO2	195‐215	O2 20 s	[325]
	Ru(DMBD)(CO)3	O2 plasma	RuO2	195‐215	O2 plasma 20 s	[325]
	HD(cumene)Ru	O2	Ru	270‐350	200 sccm O2 2 s	[326]
	HD(cumene)Ru	O2	RuO2	200	500 sccm O2 20 s	[326]
	EBBDRu	O2	Ru	225	200 sccm O2 < 10 s	[327]
	EBBDRu	O2	RuO2	225	200 sccm O2 > 45 s	[327]
	EBCHDRu	O2	Ru	225	200 sccm O2 1 s	[328]
	EBCHDRu	O2	RuO2	225	200 sccm O2 10 – 60 s	[328]
	Ru(TMM)(CO)3	O2	RuO2	180	tO2/tRu ∼ 10	[329]
	Ru(DMPD)2	O2	RuO2	185	O2 50 mTorr 30 s or 1 Torr 2 s	[330]
Ir	Ir(acac)3	O3	Ir	200‐225	1s	[331, 332]
	Ir(acac)3	O3	IrO2	165‐185	1s	[331, 332]
	Ir(acac)3	O2	Ir	300	Typically 20 s	[333]
	Ir(acac)3	O3‐H2	Ir	195	Typically 20 s	[333]
	Ir(acac)3	O3	IrO2	195	Typically 20 s	[333]
	(EtCp)Ir(COD)	O2	Ir	230‐290	Low O2/(Ar+O2)	[334]
	(EtCp)Ir(COD)	O2	IrO2	230‐290	High O2/(Ar+O2)	[334]
	(MeCp)Ir(COD)	O3‐H2	Ir	120‐180	—	[335]
	(MeCp)Ir(COD)	O3	Ir	200	—	[335]
	(MeCp)Ir(COD)	O3	IrO2	100‐180	—	[335]
	(MeCp)Ir(COD)	O2	Ir	225‐350	20 sccm O2	[336]
	TICP	O3	Ir	180‐250	3 – 10 s	[337]
	TICP	O3	IrO2	180‐200	30 – 60 s	[337]
	TICP	O2	Ir	200‐320	5 s	[338]
Pt	(MeCp)PtMe3	O2 plasma	Pt	150‐300	Short O3 injection	[339]
	(MeCp)PtMe3	O2 plasma	PtOx	100‐200	Long O3 injection	[339]
	(MeCp)PtMe3	O2	Pt	200‐300	—	[340]
	Pt(acac)2	O3	Pt	140‐200	6 s	[341]
	Pt(acac)2	O3	PtOx	120, 130	6 s	[341]
Pd	Pd(hfac)2	O3	Pd	180‐220	—	[342]
	Pd(thd)2	O2	Pd	180	—	[343]
	Pd(thd)2	O3	PdOx	130‐160	—	[344]
Rh	Rh(acac)3	O2	Rh	250	1‐10 s	[345]
	Rh(acac)3	O3	Rh	200‐220	0.3 s	[342]
	Rh(acac)3	O3	Rh	190	6‐20 s	[346]
	Rh(acac)3	O3	Rh2O3	160‐180	6‐20 s	[346]

Owing to the inherent instability of noble metal oxides, even when oxide films are formed at low temperatures, oxides can be readily reduced to their metallic state upon exposure to reducing agents, such as hydrogen [344, 350, 351] or ethylcyclohexane [352]. In fact, a low‐temperature ALD strategy has been reported in which the precursors O₃ and H₂ are introduced sequentially within a single cycle to efficiently deposit metallic films [344, 350, 351].

4. Conclusions and Outlook

Metastable phases present a powerful avenue for expanding the novel functionality of thin films and unlocking the properties essential for next‐generation electronics, catalysis, and energy‐related applications. However, the inherent challenges in the synthesis of these phases continue to hinder their broader application. Owing to its precise atomic control and excellent conformality, ALD has attracted interest across industries and has emerged as a tool for stabilizing such phases. In particular, as the need for high‐performance materials compatible with 3D integration grows, achieving metastable phases via ALD becomes a critical strategy. Although a low temperature window may hinder the facile synthesis of metastable phases, recent advances in process design and novel mechanistic innovations have steadily overcome these limitations.

This review outlines the recent strategies for realizing metastable phases via ALD. Metastable phases are categorized into two main classes: the selection of polymorphic phases, and the tuning of valence states. For polymorph control, strategies, such as controlling the deposition temperature, substrate‐induced lattice matching, grain‐size refinement, and chemical modulation by incorporating heteroelements, have been introduced. Strategies, such as the control of the deposition temperature, the design and selection of precursors and co‐reactants, and post‐deposition treatments, have been systematically exploited to tune the valence state. Such strategies have extended ALD to the synthesis of diverse metastable materials, such as oxides, chalcogenides, nitrides, and even elemental metals. This demonstrated that ALD is a highly versatile tool for stabilizing metastable phases.

Important opportunities remain. The process window for synthesizing metastable phases remains narrow, and expanding this accessible regime is critical for a broader impact. For instance, the dependence on specific substrates to stabilize certain polymorphs has significant limitations. Approaches that decouple metastability from substrate selection, such as the recent use of sacrificial layers [92], highlight promising directions for addressing this limitation.

To further unlock the synthesis of metastable phases by ALD, it is imperative to employ a rational precursor design tailored to specific reaction pathways and develop a mechanistic understanding of surface reactions via in situ diagnostics. For example, synchrotron‐based in situ X‐ray measurements during ZnO ALD have revealed the early‐stage evolution of nucleation and film coalescence [353]. Likewise, cycle‐resolved in vacuo XPS during Pt ALD has captured transient surface reaction states that define the initial growth chemistry [354]. These real‐time measurements including in situ X‐ray reflectivity, in vacuo XPS, and grazing‐incident wide‐angle X‐ray scattering, reveal transient surface configurations and kinetic constraints that are inaccessible ex situ, providing mechanistic insight into how transient ordering and surface reaction pathways can bias the system toward the stabilization of metastable phases.

Looking ahead, we anticipate that data‐driven approaches will play an increasingly central role. The integration of artificial‐intelligence‐ and machine‐learning‐assisted metastability prediction, tailored to ALD‐specific variables such as precursor chemistry, reactant type, substrate, and temperature window, will enable the identification of kinetic stabilization pathways and provide guidance for experimentally targeting phases that have not yet been accessed by ALD. By combining first‐principles calculations, high‐throughput simulations, and experimentally derived databases with these models, it should become possible to rapidly screen chemistries and process conditions, thereby accelerating the discovery of metastable phases while minimizing costly trial‐and‐error in the laboratory.

These advances will open opportunities not only within the material families already explored, but also in chemistries that have so far remained underrepresented in ALD. For instance, metastable fluorides and phosphides, which often exhibit rich polymorphism and diverse valence states, have yet to be extensively studied in the context of ALD. These materials may present unique opportunities for tailoring optical, electronic, or ionic transport properties, especially in applications such as batteries, catalysis, and optoelectronics. Identifying suitable precursors and reaction schemes for such chemistries represents an important direction for future research. Van der Waals materials represent another frontier where metastable phases may play a pivotal role. The atomically thin nature and weak interlayer interactions of these systems often result in energy landscapes with multiple competing minima, making them ideal candidates for metastable phase stabilization via surface‐controlled processes such as ALD.

With the increasing demand for innovative materials, atomic layer deposition is expected to evolve from a mere deposition technique to a versatile platform for metastable material design, enabling access to exotic phases that surpass thermodynamic constraints and redefining the limits of material discovery. Ultimately, the continued refinement of these strategies will accelerate the integration of metastable phases into next‐generation functional materials, contributing to progress across a wide range of applications.

Conflicts of Interest

The authors declare no conflict of interest.

Data Availability Statement

The authors have nothing to report.