Tuning Material Properties of Oxides and Nitrides by Substrate Biasing during Plasma-Enhanced Atomic Layer Deposition on Planar and 3D Substrate Topographies.

Oxide and nitride thin-films of Ti, Hf, and Si serve numerous applications owing to the diverse range of their material properties. It is therefore imperative to have proper control over these properties during materials processing. Ion-surface interactions during plasma processing techniques can influence the properties of a growing film. In this work, we investigated the effects of controlling ion characteristics (energy, dose) on the properties of the aforementioned materials during plasma-enhanced atomic layer deposition (PEALD) on planar and 3D substrate topographies. We used a 200 mm remote PEALD system equipped with substrate biasing to control the energy and dose of ions by varying the magnitude and duration of the applied bias, respectively, during plasma exposure. Implementing substrate biasing in these forms enhanced PEALD process capability by providing two additional parameters for tuning a wide range of material properties. Below the regimes of ion-induced degradation, enhancing ion energies with substrate biasing during PEALD increased the refractive index and mass density of TiO x and HfO x and enabled control over their crystalline properties. PEALD of these oxides with substrate biasing at 150 °C led to the formation of crystalline material at the low temperature, which would otherwise yield amorphous films for deposition without biasing. Enhanced ion energies drastically reduced the resistivity of conductive TiN x and HfN x films. Furthermore, biasing during PEALD enabled the residual stress of these materials to be altered from tensile to compressive. The properties of SiO x were slightly improved whereas those of SiN x were degraded as a function of substrate biasing. PEALD on 3D trench nanostructures with biasing induced differing film properties at different regions of the 3D substrate. On the basis of the results presented herein, prospects afforded by the implementation of this technique during PEALD, such as enabling new routes for topographically selective deposition on 3D substrates, are discussed.

Introduction

The ubiquity of thin-films in emerging nanotechnology applications,[1] together with device scaling using 3-dimensional (3D) architectures,[2,3] place stringent requirements on nanoscale processing techniques used in thin-film deposition. Atomic layer deposition (ALD) is a well-known method for obtaining ultrathin films with precise growth control.[4] It is a cyclic deposition process based on sequential and self-limiting reactant dose steps for synthesizing thin-films in a layer-by-layer manner. The self-limiting surface reactions during the separate reactant exposures lead to excellent film uniformity on large area substrates and unparalleled conformality on high-aspect-ratio 3D nanostructures. In conventional thermal ALD, the energy needed for surface reactions to proceed is provided by substrate heating. An alternate source for the required energy can be obtained by using an energy enhanced ALD method known as plasma-enhanced atomic layer deposition (PEALD).[5] In PEALD, the substrate is exposed to species generated by a plasma during one of the reactant exposure steps of an ALD cycle.[5,6] These species constitute a mixture of highly reactive atomic and molecular neutrals (plasma radicals), photons, ions and electrons. The highly reactive plasma species generated during PEALD can provide an alternate source of the energy required for film growth. The high reactivity of the plasma species therefore, allows access to a parameter space such as low temperature deposition environments. Tuning material properties during film growth at a particular temperature is typically performed by varying either the substrate material or the reactants needed for film deposition. For PEALD at any given temperature, the various parameters of plasma operating conditions (e.g., reactant gas, flow rate, pressure, power, direct/remote source configuration, etc.) allow greater freedom in processing conditions for tuning the growth and material properties of thin-films.[6]

Apart from the contribution of highly reactive plasma radicals toward film growth during PEALD, the ions generated by the plasma can also play a significant role during film deposition.[5,6] Ion-surface interactions during plasma exposure are a characteristic feature of plasma-enhanced deposition techniques.[7−10] Additional energy can be provided to the deposition surface by the kinetic energy of ions impinging on the substrate. While plasma radicals can enable film growth through their high reactivity, the additional energy delivered to the deposition surface by ion bombardment can influence a wide range of material properties of the growing film. Examples reported in the literature include tailoring of the optical refractive index, electrical conductivity, mechanical stress, mass density, crystalline structure, morphology, surface roughness etc.[9,11,12] The limit to which ion-surface interactions can modify such properties depends on the energy, mass, reactivity, impingement rate (or flux) and dose (or net flux integrated over time) of ionic species impinging on the film surface.[7−9,13,14] On the basis of the values of these parameters, various physical, chemical or combined physicochemical processes can occur at the surface and subsurface regions of the film as the ions transfer energy and momentum during deposition. The use of an inert gas (e.g., argon) leads to predominantly physical processes which include, but are not limited to, surface or bulk atom displacements, subsurface implantation of incident ions or displaced surface atoms, desorption of adsorbed surface impurities from reactor background, etc.[8,9,14] The use of reactive gases (e.g., oxygen or hydrogen) involves an additional chemical or physicochemical component where the reactive ions can undergo chemical reactions forming, either stable products leading to film growth, or volatile products leading to film removal.[8,10] However, the processes and trends of property modulation also depend on many inherent characteristics of the growing film such as the bond energy, bond type (i.e., ionic, covalent or metallic), crystallization temperature, free energy difference between crystalline and amorphous phases or a combination of several of these parameters.[14−17] Given the multitude of factors that can play a role during ion-surface interactions, it is not a trivial or straightforward task to anticipate the trends in film characteristics as a result of varying the ion characteristics during deposition. Furthermore, the effects of ion-surface interactions have been extensively investigated in the literature for conventional flux-controlled processes,[18] such as physical vapor deposition (PVD) and plasma-enhanced chemical vapor deposition (PECVD), that lead to continuous film growth. In the surface-controlled process[18] of PEALD where self-limiting film growth occurs by sequential doses of a precursor and plasma activated coreactant, ion-surface interactions only occur during the plasma exposure step of a PEALD cycle. Consequently, only the characteristics of those ionic species generated during plasma activation of the coreactant can affect film properties in PEALD. These effects may or may not be different from those occurring in continuous growth process like PECVD in which both the precursor and coreactant contribute species to the mixture of ions formed during deposition. The first experimental investigation on the role of ions during PEALD was reported by Profijt et al.[19,20] in previous work conducted within our group. They looked into the effects of enhancing oxygen ion energies during PEALD on the growth and material properties of aluminum oxide, titanium oxide and cobalt oxide.[19,20] The films were deposited in a home-built remote plasma ALD tool equipped with substrate biasing. In a remote plasma source configuration, the ion energy can be controlled during plasma exposure by varying (or biasing) the substrate potential, either by adjusting the impedance between the substrate and ground, or by applying a voltage signal on the substrate.[19] Enhancing ion energies with substrate biasing during the plasma exposure step of PEALD was observed to have significant effects during film growth that were material or process-specific. Increasing bias voltages during oxygen plasma exposure increased the growth rate and oxygen content of aluminum oxide films while lowering their mass density.[19] The residual stress of the deposited films was also observed to be altered from tensile to compressive with the use of substrate biasing. The growth rate, density and oxygen content of titanium oxide films deposited with substrate biasing during oxygen plasma exposure showed a similar behavior as aluminum oxide.[19,20] However, an additional effect was observed in case of titanium oxide where its crystalline phase could be gradually tailored from anatase to rutile with increasing bias voltages.[19,20] For cobalt oxide, the growth rate decreased and the film density increased at higher bias voltages while the films became slightly oxygen deficient.[19] The aforementioned results were demonstrated for film deposition only on planar substrates having a small area (1 × 1 in. pieces of Si wafer) and by applying substrate biasing for the entire duration of the plasma exposure step.

In this work, we continued the research on the role of ions during PEALD by analyzing the effects of enhanced ion energies on the material properties of films grown on both planar and 3D substrate topographies. The ion energy was enhanced during plasma exposure using a commercial remote plasma ALD system (Oxford Instruments FlexAL) equipped with radio frequency (RF) substrate biasing. The effect of enhancing oxygen ion energies during PEALD of titanium oxide was reinvestigated in this system on large area substrates (200 mm) using a different titanium precursor with additional material characterization, compared to the previous work conducted by Profijt et al.[19,20] This provided further insight on the effects of ion energy control during PEALD of titanium oxide. Additionally, the effects of varying the dose of higher energy ions on the growth and material properties of titanium oxide was also investigated. The effects of ion-surface interactions on material properties during PEALD were investigated for two more oxides, namely hafnium oxide and silicon oxide. Furthermore, the scope of our research on the role of ions during PEALD was expanded by looking into the effects of ion-surface interactions for another group of materials, namely, the nitrides of titanium, hafnium and silicon deposited using different plasmas (i.e., hydrogen, argon/hydrogen, and nitrogen). The oxides and nitrides of these three elements form dielectric or conductive materials that are used in a variety of applications (see Supporting Information). The diverse range of applications that are heavily reliant on the properties of these materials highlights the necessity for precise control over those properties during film deposition. Enhancing ion energies with substrate biasing during PEALD on planar substrates was observed to have pronounced effects on the growth and material properties of the aforementioned thin-films. A comprehensive analysis of these properties was undertaken by characterizing the growth rate, mass density, refractive index (for dielectric materials), resistivity (for conductive materials), residual stress, and surface roughness of the films deposited on planar substrates. The role of ion energy control on film microstructure and crystallinity was also investigated. The effect of enhanced ion energies on the thickness uniformity of materials deposited on large area planar substrates (200 mm Si wafer) was also investigated compared to previous work carried out on smaller substrates.[19,20] Owing to the material and/or process specific effects of substrate biasing observed in previous work, an empirical investigation comprised of an extensive characterization of film properties spanning six different materials was undertaken in this work. It was observed that substrate biasing can enhance the versatility of PEALD processes by providing two additional parameters (magnitude, duration/duty cycle of bias) for tuning a wide range of material properties. Furthermore, biasing during PEALD on 3D trench nanostructures effectively delineated the role of directional ion bombardment by inducing differing film properties at different (planar and vertical) regions of the 3D substrate. These results demonstrate the potential of substrate biasing during PEALD in enabling routes toward topographically selective[3] processing on 3D substrates.

Applying Substrate Bias during PEALD Cycles

In a remote plasma ALD reactor configuration, the plasma source generating radicals and ions is located at a distance away from the substrate stage thus allowing the plasma and substrate conditions to be varied quite independently of each other. The ion flux impinging on the substrate can be increased by increasing the source power. When the substrate is exposed to the plasma, a positive space charge layer called the sheath is formed between the plasma and the substrate due to the difference in mobilities of the heavy ions and light electrons in the plasma. The voltage across the sheath (ΔVsheath) is the difference between the plasma potential (Vp) and the substrate potential (Vsub). The sheath voltage repels electrons from the substrate into the plasma and accelerates ions toward the substrate resulting in so-called ion bombardment. At sufficiently low pressures, the ion mean free path is larger than the plasma sheath thickness, so the ions are accelerated over the full sheath width without undergoing collisions. This collisionless plasma sheath condition leads to the highly directional or anisotropic nature of ion bombardment. Typical sheath thicknesses for processing plasmas are on the order of 10–4–10–2 m, and therefore, the sheath does not follow the profile of 3D structures (e.g., trenches and vias) with micro- or nanoscale dimensions. As a result, directional ions accelerated through a collisionless sheath will mainly collide with substrate surfaces parallel to the sheath, for example, the planar top and bottom surfaces of a 3D trench nanostructure and not its vertical sidewalls. The energy (E) of directional ions impinging on planar substrate surfaces is proportional to the sheath voltage, as shown in equation where q is the charge of an ion (assuming an electropositive plasma). When the substrate is placed on a grounded reactor table, it is at zero potential (Vsub = 0 V), so the sheath voltage equates to the plasma potential. Equation indicates that the ion energy can be enhanced by increasing the sheath voltage. If the system is equipped with substrate biasing, that is, by tuning the impedance between substrate table and ground or by connecting the substrate table to an additional power source, then the substrate potential can be set or biased to have a nonzero value (Vsub ≠ 0 V).[19] When a sinusoidal radio frequency (RF) voltage signal is applied to the substrate through a blocking capacitor, it acquires a negative average or DC offset value such that the net flux of ions and electrons to the substrate stage over one RF cycle is zero. The negative DC offset or time-averaged substrate bias voltage (Vsub = −⟨Vbias⟩) can be increased by increasing the amplitude of the bias signal applied to the substrate table. This causes ΔVsheath to increase which in turn enhances E, thus allowing the ion energy to be controlled during plasma exposure.

To perform PEALD with substrate biasing, an existing process can be modified by applying a bias on the substrate table during the plasma exposure step. The step sequences of the PEALD processes without and with substrate biasing are shown in Figure . The substrate bias can be applied for the whole duration (bias duty cycle = 100%, Figure b) or a part of the duration (0% < bias duty cycle <100%, Figure c) of the plasma exposure step. For example, during a plasma exposure step of 10 s, the bias can be active for all 10 s (Figure b) or half of this duration (5 s) by applying it in an interleaved manner either at the beginning, middle or end (Figure c) of the plasma exposure step. Varying the duration of the applied bias translates to changing the dose or fluence[13,14,21] (i.e., particle flux integrated over time) of higher energy ions impinging on the substrate. Substrate biasing during PEALD can also be implemented in many other configurations (see Figures S1 and S2).

Figure 1

Step sequence for two-step [AB] plasma ALD cycles (a) without any substrate biasing during plasma exposure where the substrate table is grounded (source plasma “ON”, substrate bias “OFF”, bias duty cycle = 0%), (b) with substrate biasing during plasma exposure where the bias is applied for the full duration of the plasma exposure step (source plasma “ON”, substrate bias “ON”, bias duty cycle = 100%), and (c) with interleaved substrate biasing where the bias is applied for a fraction of the plasma exposure step (source plasma “ON”, substrate bias “ON”, 0% < bias duty cycle <100%). A refers to the precursor dose step; B is the coreactant or plasma gas exposure step, and n is the number of ALD cycles.

Experimental Details

PEALD Process Conditions

Film deposition using PEALD was performed in an Oxford Instruments FlexAL reactor equipped with substrate biasing (Figure ). The reactor consists of a radio frequency (RF) power supply (13.56 MHz, up to 600 W) connected to a water-cooled copper coil wrapped around a cylindrical alumina tube to generate an inductively coupled-plasma (ICP). This remote RF-ICP source generates the radicals and ions during the plasma exposure step of PEALD. For the specific FlexAL configuration, an additional external RF power supply (13.56 MHz, up to 100 W) can be connected to the reactor table that allows for substrate biasing (up to ⟨Vbias⟩ ≈ −350 V at ∼10 mTorr) during plasma exposure. Both RF power sources were connected to the system via automated matching networks consisting of inductive and capacitive components. The FlexAL system provides a readout for ⟨Vbias⟩ when applying an RF bias signal during plasma exposure. Additionally, an oscilloscope was connected to the reactor table via a high-voltage probe from which sinusoidal RF bias voltage waveforms developing on the table could be measured. These oscilloscope measurements were used to independently verify ⟨V⟩ readouts from the FlexAL system (within ±5 V). A base pressure in the reactor chamber of ∼10–6 Torr was obtained using a turbo pump. A butterfly valve in front of the turbo pump controlled the effective pumping speed and functioned as an automated pressure controller. The chamber wall temperature was set to 150 °C while the substrate stage temperature was set between 150 and 500 °C. All substrates underwent a 30 min heating step prior to commencing deposition in order to ensure substrate temperature stabilization. The precursor delivery lines were heated to 70 °C to prevent precursor condensation. The oxides and nitrides of Ti, Hf, and Si were deposited without and with substrate biasing. The precursors, plasma gases and other PEALD process conditions for depositing the six materials are shown in Table . Existing PEALD processes for these six materials[22−28] were used for the runs without any biasing where the substrate table was grounded during the plasma exposure step. For the runs performed with substrate biasing, the PEALD processes were modified by applying an RF bias signal to the substrate table during the plasma exposure step. PEALD step sequences without (Figure a) and with substrate biasing (Figure b and c) were implemented. Details regarding PEALD process conditions are further outlined in the Supporting Information.

Figure 2

Schematic of an Oxford Instruments FlexAL system equipped with substrate biasing. A radio frequency inductively coupled-plasma (RF-ICP) source acts as the remote plasma generator allowing control over the flux of radicals and ions impinging on the substrate. A second RF power supply (RF-Bias) connected to the substrate table enables substrate biasing which allows control over the ion energy. Both RF power sources are connected to the system via automated matching units (AMU) consisting of inductive and capacitive components.

Table 1

PEALD Process Conditions for the Materials Deposited with and without Substrate Biasing during Plasma Exposure

material	TiOx	TiNx	HfOx	HfNx	SiOx	SiNx
precursor	TDMATa		TDMACpHb		BDEASc	DSBASd
bubbler temperature (°C)	60		60		50	40
precursor delivery	bubbled with 100 sccm Ar		bubbled with 100 sccm Ar		vapor drawn
stage temperature (°C)	150 and 300	200	150	450	200	500
precursor dose time (ms)	200	200	400	400	175	500
precursor reaction step (s)					1	3
precursor purge time (s)	3	3	2	2	3	2
plasma gas	O2	Ar + H2	O2	H2	O2	N2
plasma gas flow (sccm)	100	10 + 40	100	100	100	100
plasma pressure (mTorr)	9	6	15	30	15	11
RF-ICP power (W)	200	100	400	100	200	600
plasma exposure time (s)	10	10	8	10	5	20
plasma purge time (s)	3	4	3	4	2	3
⟨Vbias⟩ or average bias voltage (V)	0 to −254	0 to −255	0 to −280	0 to −210	0 to −295	0 to −103
bias during plasma	all 10 s, last 5 s	last 5 s	last 4 s	all 10 s, last 5 s	all 5 s	last 10 s

TDMAT – Ti(NMe2)4.

TDMACpH – CpHf(NMe2)3.

BDEAS – SiH2(NEt2)2.

DSBAS – SiH3N(sBu)2.

Thin-Film Characterization on Planar and 3D Substrates

Unless otherwise stated, the planar substrates used for depositing films with and without substrate biasing were single side polished (SSP) c-Si (100) substrates having a thin native oxide layer (∼1.5 nm) for dielectric materials and SSP c-Si (100) substrates having a thick thermal oxide layer (∼450 nm) for conductive materials. Film thicknesses between ∼20 and ∼80 nm were deposited for material characterization on planar substrates (except for SiO films which were ∼200 nm). The optical properties and film thickness of the deposited layers on c-Si substrates were measured by means of spectroscopic ellipsometry (SE) using a J.A. Woollam M2000D rotating compensator ellipsometer (1.2–6.5 eV). SE measurements were also used to determine the thickness uniformity of films deposited on 200 mm planar c-Si wafers with and without substrate biasing by mapping film thickness at several locations across the wafers. For SiO and SiN films deposited with and without substrate biasing, the optical model used consisted of a silicon substrate, ∼1.5 nm native oxide and the deposited layer parametrized with a Cauchy dispersion relation.[27,28] A similar three layer optical model was used for analyzing the ellipsometry data of TiO and HfO films grown with and without biasing where the deposited layer was parametrized using two Tauc–Lorentz oscillators.[19,29] For TiN and HfN, different models were used for the films deposited with and without biasing. The optical model used for the films grown without biasing consisted of a silicon substrate, ∼450 nm thermal oxide and the deposited layer parametrized using one Drude, one Lorentz, and one Tauc–Lorentz oscillator.[26] The TiN and HfN films deposited with biasing were parametrized using one Drude and two Lorentz oscillators.[29]

The root-mean-squared (RMS) surface roughness and mass density of the films deposited on planar substrates were analyzed using X-ray reflectometry (XRR) performed using a Bruker AXS D8 Advance system in the grazing incidence geometry with a Cu Kα X-ray source (radiation wavelength of 0.154 nm).[30] The thicknesses of the deposited films were also obtained with XRR which were in good agreement with the corresponding values obtained using SE. The residual stress levels of the deposited films were determined by means of wafer-curvature measurements at room temperature. A KLA-Tencor FLX 2320 system was used to determine the curvature of 3 in. double side polished (DSP) c-Si (100) wafers before and after deposition of the films. Based on the difference in curvature, the residual stress of the films deposited with and without substrate biasing was calculated using the Stoney equation.[19,30] A sign convention was used where positive values represented tensile and negative values represented compressive residual stress. Rutherford backscattering spectrometry (RBS) and elastic recoil detection (ERD) measurements were used to determine the composition (stoichiometry) of films deposited on planar SSP c-Si substrates with ∼1.5 nm native oxide. The RBS and ERD measurements with subsequent data simulations were performed by the company Detect 99 using a 1.8–2 MeV helium-ion beam. The areal densities of the elements were determined from raw data simulations. The crystallinity of films deposited on planar c-Si substrates was analyzed using grazing incidence X-ray diffraction (XRD). The measurements were performed in a PANalytical X’pert PRO MRD with a Cu Kα X-ray source (radiation wavelength of 0.154 nm) operated at an incidence angle of 0.5°. A JEOL ARM 200 transmission electron microscope (TEM) operated at 200 kV was used to analyze the microstructure of films deposited on planar TEM windows. These windows consisted of ∼15 nm Si3N4 membranes coated with 5 nm of SiO2 grown using ALD. This ensured a SiO2 starting surface while maintaining transparency to the electron beam. Both bright-field TEM (BF-TEM) and high-angle-annular-dark-field scanning TEM (HAADF-STEM) modes were employed to characterize the samples in plan-view. Electrical sheet resistance measurements of conductive films deposited on c-Si substrates with ∼450 nm thermal oxide were performed at room temperature using a Signatone four-point probe (FPP) in combination with a Keithley 2400 Source Measurement Unit (SMU) that acted as both current source and voltage meter. The electrical resistivity was obtained from the slope of the generated I–V curve and the film thickness deduced from the SE measurements.

For investigating material properties on 3D substrate topographies, films were deposited with and without substrate biasing during PEALD on coupons containing high aspect ratio trench nanostructures (width ∼100 nm, height ∼450 nm, AR = 4.5:1) and analyzed with cross-sectional TEM. These 3D nanostructures were created[28] by first depositing a thick SiO2 film on a Si wafer using PECVD, that was subsequently etched into trench structures. The SiO2 trench structures were then coated with a SiN layer using high-temperature CVD, on which a SiO2 layer was deposited using ALD. Coupons containing these trench nanostructures were prepared and provided by Lam Research. A JEOL 2010F ultrahigh-resolution scanning TEM at 200 kV (Nanolab Technologies) was employed to obtain cross-sectional images of the films deposited on the 3D trench nanostructures. The films were coated with a layer of spin-on epoxy to protect them from damage during sample preparation for TEM cross-sectional imaging. The samples were then placed on a Cu TEM grid, after which an energetic ion beam was used to mill and polish the samples at 30 kV, 100 pA and 5 kV, 40 pA, respectively. The deposited film thickness was measured at three regions of several trench nanostructures in the sample, namely at the planar top and bottom regions together with the vertical bottom-side region (see Figure S3) by counting pixels in the TEM image. Conformality was determined by taking the ratio of film thicknesses at the bottom-side and bottom of the trench to that at the top of the trench. The conformality values reported are the average of the results obtained across several trench nanostructures with the same aspect ratio. For obtaining wet-etch rates (WER) of silicon nitride films at planar and vertical surfaces of the 3D substrates, coupons containing the trench nanostructures with films deposited with and without substrate biasing were dipped in a dilute hydrofluoric acid solution (HF:H2O = 1: 100) for 30 s. Two samples for TEM cross-sectional imaging were prepared from the same coupon, one before and one after the chemical wet-etch treatments. TEM measurements were conducted at the three aforementioned regions across several trench nanostructures of the same as-deposited and post wet-etch samples. The WERs at the aforementioned trench regions were determined by comparing the as-deposited and post wet-etch film thicknesses at those regions. The WER values reported are the average of the results obtained across several trench nanostructures with the same aspect ratio. Uncertainties reported for the values were based on both the accuracy of the measurement and the variation between measurements conducted across several trench nanostructures of the same sample. Potential depletion of the etchant inside the trench was not taken into account in these experiments.

Results

PEALD of Oxides on Planar Substrates

Titanium Oxide (TiO)

The growth and material properties of TiO films deposited on planar substrates at 300 °C were investigated for ⟨Vbias⟩ between 0 and −254 V applied during the O2 plasma exposure step. Figure shows these properties in terms of the growth per cycle (GPC), refractive index, mass density, residual stress and surface roughness expressed as a function of ⟨Vbias⟩ applied during plasma exposure. The solid symbols are for films deposited with ⟨Vbias⟩ applied for the entire duration of the 10 s plasma exposure step (bias duty cycle = 100%, Figure b), while the hollow symbols denote films deposited with ⟨Vbias⟩ applied in an interleaved manner, that is, during the last half (5 s) of the 10 s plasma exposure step (bias duty cycle = 50%, Figure c). Table outlines additional growth and material properties from RBS analysis for the films deposited with and without substrate biasing. The growth properties are outlined in terms of the number of titanium and oxygen atoms deposited per nm2 per cycle (GPC [Ti] and GPC [O], respectively) while the material properties are outlined in terms of the O/Ti ratio and impurity content ([H] and [C] at. %). The crystalline properties of the TiO films are depicted in Figure , which shows grazing incidence XRD patterns for films deposited at 300 °C with ⟨Vbias⟩ ranging from 0 V to −254 V. Figure a denotes patterns for ⟨Vbias⟩ applied for the entire 10 s O2 plasma exposure step while Figure b shows results for ⟨Vbias⟩ applied during the last half (5 s) of the 10 s plasma exposure step. The microstructure of the TiO films can be visualized in Figure which shows plan-view HAADF STEM images for films deposited on planar TEM windows at 300 °C with and without substrate biasing.

Figure 3

(a) Growth per cycle (GPC), (b) refractive index, (c) mass density, (d) residual stress, and (e) RMS surface roughness of titanium oxide films deposited at 300 °C expressed as a function of the average bias voltage, ⟨Vbias⟩, applied for the entire duration and last half (5 s) of the 10 s O2 plasma exposure step.

Table 2

GPC (Film Thickness Per Cycle and Number of Atoms Deposited per nm2 per Cycle) and Elemental Composition of Titanium Oxide Films Deposited at 300 °C with and without Bias Voltages Applied during the O2 Plasma Exposure Stepa

			RBS				ERD
⟨Vbias⟩ (V)	bias duration during 10 s O2 plasma step (s)	GPC (Å/cycle)	GPC [Ti] (# Ti at. per nm2 per cycle)	GPC [O] (# O at. per nm2 per cycle)	O/Ti	[C] at. %	[H] at. %
0	no bias	0.48 ± 0.02	1.35 ± 0.03	2.72 ± 0.05	2.0 ± 0.1	<d.l.b	3 ± 7
–152	all 10	0.61	1.62	3.47	2.2	<d.l.	3
	last 5	0.55	1.50	3.07	2.0	<d.l.	3
–205	all 10	0.70	1.71	3.85	2.3	<d.l.	3
	last 5	0.60	1.61	3.67	2.3	<d.l.	3

For the films deposited with biasing during plasma exposure, data are shown for average bias voltages, ⟨Vbias⟩, applied during the whole (10 s) and the last half (5 s) of the 10 s O2 plasma step. Typical uncertainties are given in the first row.

Values below detection limit (d.l.) of 8 at. % for [C].

Figure 4

Grazing incidence X-ray diffractograms for titanium oxide films deposited at 300 °C with average bias voltages, ⟨Vbias⟩, ranging from 0 to −254 V applied for the (a) entire duration and (b) last half (5 s) of the 10 s O2 plasma exposure step. Peaks corresponding to the anatase and rutile phase are denoted with “A” and “R”, respectively.

Figure 5

Plan-view high angle annular dark-field (HAADF) STEM images for titanium oxide films deposited at 300 °C with (a) 0 V, (b1 and b2) −102 V, (c1 and c2) −152 V, (d1 and d2) −205 V and (e1 and e2) −254 V average bias voltages, ⟨Vbias⟩. Images (b1), (c1), (d1), and (e1) denote ⟨V⟩ applied for the entire duration of the 10 s O2 plasma exposure step while images (b2), (c2), (d2), and (e2) represent ⟨Vbias⟩ applied for the last half (5 s) of the 10 s plasma exposure step.

No Substrate Bias

In the absence of substrate biasing during plasma exposure, (i.e., grounded substrate or ⟨Vbias⟩ = 0 V, bias duty cycle = 0%, Figure a) the film had a GPC of 0.48 ± 0.02 Å (Figure a) which is similar to the growth rate reported by Wei et al.[31] for TiO deposited at the same temperature using TDMAT and O2 plasma. The values obtained for the refractive index and mass density were 2.47 ± 0.03 and 3.7 ± 0.2 g/cm3 (Figure b and c), respectively, which is comparable to the values reported in the literature for the anatase phase of TiO.[32] This film was found to have a tensile residual stress (Figure d) which was also reported for TiO deposited at 300 °C in previous work.[33] An O/Ti ratio of 2.0 ± 0.1 (Table ) was observed for this film indicating the formation of stoichiometric TiO2 for deposition without any ⟨Vbias⟩. Low impurity contents in terms of [H] and [C] (3 ± 7 at. % and <8 at. %, respectively) were obtained in all films (Table ). The XRD pattern for the film deposited without biasing (Figure a) shows diffraction peaks that correspond to the anatase phase of TiO2. The intense and narrow (101) peak indicates the presence of a highly crystalline TiO2 film with large anatase crystal grains. This is confirmed by the complementary information from the TEM image for the film deposited without biasing (Figure a), which shows a compact or void-free microstructure with lateral grain sizes on the order of ∼150 nm.

Effect of Bias Magnitude

For the 10 s bias duration condition (bias duty cycle = 100%, Figure b), the GPC showed a monotonic increase with increasing ⟨Vbias⟩ and reached a value of 0.80 ± 0.02 Å at −254 V (Figure a). This is also reflected in the GPC [Ti] and GPC [O] values which increased to 1.71 ± 0.03 Ti and 3.85 ± 0.05 O at./nm2, respectively, at −205 V applied for all 10 s of plasma exposure (Table ). The trends for refractive index, mass density and residual stress can be categorized into two stages composed of an initial buildup to a maximum followed by a gradual decrease with increasing ⟨Vbias⟩. In the first stage, the refractive index and mass density reached peak values of 2.51 ± 0.03 and 4.1 ± 0.2 g/cm3, respectively, as ⟨Vbias⟩ was increased to −152 V (Figure b and c). The corresponding behavior of the residual stress for TiO in the first stage consisted of a change from tensile to compressive on applying a bias to the substrate, denoted by negative values on the vertical axis of Figure d. This stress evolution from tensile to compressive is similar to that frequently reported in the literature for thin-film deposition using energetic particle bombardment.[19,34−37] The compressive stress reached a maximum as ⟨Vbias⟩ was increased to −152 V, similar to the peaking of refractive index and mass density with biasing at the same voltage. The maximization of compressive stress for deposition using energetic particle bombardment has been reported in the literature to be related to the formation of bombardment induced point defects, such as interstitials in the film bulk.[34−37] Under conditions of intense ion bombardment during film growth, the implantation of the energetic ions or the formation of subsurface interstitial atoms resulting from ion-surface interactions that push deposited atoms into the film bulk (i.e., recoil implantation or forward sputtering of knock-on atoms generated by ion-surface collisions) can increase atomic packing of the growing film.[34−37] This raises compressive stress as well as the mass density which scale with the concentration of those point defects.[34−37] The O/Ti ratio for the film deposited using −152 V applied for 10 s of plasma exposure was observed to increase to 2.2 ± 0.1 (Table ) indicating the formation of an oxygen rich film during energetic oxygen ion bombardment. The excess oxygen can be speculated to be present as interstitial species which could stimulate an increase in compressive stress while also contributing toward elevating mass density and refractive index observed when increasing ⟨Vbias⟩ up to −152 V. Applying and steadily increasing ⟨Vbias⟩ to −152 V cause the anatase diffraction peaks in the XRD patterns (Figure a) to gradually decrease in magnitude while new peaks corresponding to the rutile phase of TiO2 appeared. This is similar to the results obtained in previous work conducted in our group where the crystalline phase of TiO films, grown using different precursors[19,20] in a home-built plasma ALD tool equipped with substrate biasing, could be transformed by varying ⟨Vbias⟩ during the O2 plasma step. TiO is known to have a higher refractive index and mass density in the rutile compared to the anatase phase.[32] Therefore, besides the effect of a higher interstitial content on material properties, the observed increase in GPC, refractive index, material density and compressive stress could also be caused by phase transformation, or more specifically, the formation of denser rutile crystallites in the mixed phase (anatase + rutile) TiO films obtained using ⟨Vbias⟩ up to −152 V (Figure a). The deposited films were observed to retain a compact and void-free microstructure for ⟨Vbias⟩ up to −152 V (Figure b1 and c1). However, based on estimates of lattice fringe areas in magnified TEM images (see Figure S4b1 and S4c1), lateral grain sizes of ∼20 nm were observed in the bias deposited films which were nearly an order of magnitude smaller than the grains formed without biasing.

For ⟨Vbias⟩ beyond −152 V applied for all 10 s of plasma exposure, the GPC (Figure a), GPC [Ti], GPC [O], and O/Ti ratio (Table ) continued to increase but the refractive index and mass density entered their second stage where they rapidly decreased, reaching values of 2.27 ± 0.03 and 3.7 ± 0.2 g/cm3, respectively, at −254 V (Figure b and c). Furthermore, increase in ⟨Vbias⟩ beyond −152 V also initiated the second stage in residual stress behavior. The compressive stress began to decrease, similar to the trends in refractive index and mass density as ⟨Vbias⟩ was increased to −254 V (Figure d). The anatase diffraction peaks in the XRD patterns completely disappeared and only rutile diffraction peaks remained (Figure a). The intensities and widths of these peaks became lower and broader, respectively, on increasing ⟨Vbias⟩ beyond −152 V indicating an increase in the degree of disorder or amorphization of the film microstructure. This onset of amorphization induced by highly energetic ion bombardment has also been reported in the literature[14,37,38] and it could play a role in lowering the refractive index and mass density of the films. Furthermore, the relaxation of compressive stress on increasing ⟨Vbias⟩ beyond −152 V suggests that the yield strength of the material could have been exceeded under such intense ion bombardment, which could then lead to plastic deformation.[14,36,39,40] These aforementioned trends in material properties were in line with the appearance of voids in the film microstructure for TiO deposited with ⟨Vbias⟩ greater than −152 V (Figure d1 and e1). The void fraction increased with increasing ⟨Vbias⟩, which could also play a role in compressive stress relaxation together with decreasing refractive index and mass density. The surface of TiO films initially became smoother with biasing as the RMS roughness decreased at −102 V (Figure e). This could be due to a combined effect of phase transformation, material densification and/or coalescence at grain boundaries due to enhanced mobility of surface species induced by energetic ions[37] during film growth. The roughness increased with further increase in ⟨Vbias⟩ to −254 V, which could be attributed to the onset of void incorporation or ion bombardment induced surface damage.[37] The trends in RMS surface roughness obtained from XRR were corroborated with additional measurements performed using atomic force microscopy (see Figure S5 and S6). Altogether, the results above demonstrate how the growth and material properties of TiO films deposited using TDMAT and O2 plasma can be tuned by altering the magnitude of ⟨Vbias⟩ applied with a 100% duty cycle during the plasma exposure step of PEALD.

Effect of Bias Duration/Duty Cycle

When ⟨Vbias⟩ having the same magnitude but different duty cycle was applied during the O2 plasma exposure step (i.e., the last 5 s of the 10 s plasma exposure, bias duty cycle = 50%, Figure c), differences were observed in the variation of TiO film properties. The GPC again showed a trend of increasing with ⟨Vbias⟩, but at a slower rate than the 10 s bias condition reaching a comparatively lower value at −254 V (Figure a). This was also the case for GPC [Ti] and GPC [O] values (Table ), which increased to comparatively lower values at −205 V applied for the shorter 5 s of plasma exposure. The refractive index and mass density of the films also increased with biasing for the shorter 5 s condition, but were consistently higher than the corresponding values for films deposited using the same ⟨Vbias⟩ applied for a longer duration of 10 s (Figure b and c). The refractive index and mass density for the 5 s bias condition reached higher maximum values at a higher ⟨Vbias⟩ of −205 V compared to the corresponding peaks at −152 V obtained using the 10 s bias condition (Figure b and c). The residual stress of the TiO films again changed from tensile to compressive for the 5 s bias condition and also reached a higher maximum at a higher ⟨Vbias⟩ of −205 V compared to the maximum obtained at −152 V using the 10 s bias condition (Figure d). The O/Ti ratio for the film deposited using −205 V applied for the shorter duration of 5 s was 2.3 ± 0.1 (Table ) which again indicates the formation of an oxygen rich film during energetic oxygen ion bombardment. Since compressive stress tends to scale with the concentration of point defects such as interstitials,[34−37] the higher fraction of excess oxygen in the film grown with −205 V applied for the shorter duration of 5 s could explain why it had a higher compressive stress, mass density and refractive index compared to the film grown with −152 V applied for 10 s. It is worth noting that the use of −102 V for the shorter duration of 5 s yielded TiO that was approximately stress-free in nature as evidenced by the insignificant value of −25 ± 50 MPa measured for that film. Furthermore, TiO films deposited using the 5 s bias condition retained a compact and void-free microstructure up to a higher ⟨Vbias⟩ of −205 V (Figure b2, c2, and d2) compared to the −152 V ceiling for obtaining similar void-free films using the 10 s bias condition (Figure b1 and c1). Increasing ⟨Vbias⟩ beyond −205 V using the 5 s bias condition led to a decrease in both refractive index and mass density together with compressive stress relaxation (Figure b, c and d), similar to the trends observed for the 10 s bias condition beyond −152 V. This can again be attributed to the incorporation of voids in the film microstructure (Figure e2) and/or the onset of plastic deformation[14,36,39,40] for TiO deposited at −254 V using the 5 s bias condition. The TiO film surface initially became smoother with biasing using the 5 s condition, similar to the 10 s condition. However, the lowest RMS roughness was now at −152 V applied for the last 5 s (Figure e) compared to −102 V applied for all 10 s of plasma exposure. Further increase in ⟨Vbias⟩ beyond −152 V using the 5 s bias condition increased surface roughness. Similar to the delayed changes observed in the aforementioned properties, the anatase to rutile phase transformation using the 5 s bias condition also occurred at a slower rate compared to the 10 s bias condition (Figure b). It is worth noting that for the 10 s bias condition, using −152 V leads to a combined anatase and rutile mixed phase with the most intense R(110) peak (Figure a) whereas for the 5 s condition, a similar mixed phase with the most intense R(110) peak is observed at −205 V (Figure b). It seems that for PEALD of TiO with substrate biasing, optimum material quality in terms of a high refractive index, high density and compact microstructure is obtained for films with a mixed phase content having the highest R(110) peak intensity and no voids. Varying the duration of the applied bias translates to changing the dose or fluence (i.e., particle flux integrated over time) of energetic ions impinging on the substrate. The dose or fluence of energetic ions has also been reported in the literature as a parameter that can have significant effects on material properties.[13,14,21] Altogether, these results effectively demonstrate how the growth and material properties of TiO films deposited using TDMAT and O2 plasma can be tuned, not only by varying the magnitude of ⟨Vbias⟩, but also by changing the duration or duty cycle of ⟨Vbias⟩ applied during the plasma exposure step.

Effect of Substrate Bias at Lower Deposition Temperature

Figure shows selected area electron diffraction patterns and plan-view HAADF STEM images for TiO films deposited at a lower temperature of 150 °C with and without substrate biasing applied during 10 s of O2 plasma exposure. For the film deposited without any ⟨Vbias⟩, the diffuse rings in the electron diffraction pattern (Figure a1) and the absence of crystal lattice fringes in the plan-view HAADF STEM image (Figure a2) clearly indicate the amorphous nature of the film. However, the appearance of sharp rings (Figure b1) and crystal lattice fringes (Figure b2) indicate the presence of polycrystalline material in the film deposited with ⟨Vbias⟩ of −205 V. This is confirmed by the appearance of diffraction peaks corresponding to the rutile phase of TiO in the grazing incidence XRD pattern for the film deposited with −205 V at 150 °C (see Figure S7). Profijt et al.[20] reported a similar result for TiO deposited at 200 °C using Ti(CpMe)(NMe2)3 as the precursor. Their film deposited without biasing was amorphous, while applying −100 V bias during the whole duration of the O2 plasma exposure yielded crystallinity, evidenced by the R(110) diffraction peak in the XRD pattern. The results obtained in this work using a different precursor (TDMAT) effectively reproduce the phenomenon of inducing crystalline material formation with substrate biasing at a comparatively lower temperature that would typically yield amorphous films.

Figure 6

(a1, b1) Selected area electron diffraction patterns and (a2, b2) plan-view high angle annular dark-field (HAADF) STEM images for titanium oxide films deposited at 150 °C. Panels (a1) and (a2) are for the film deposited without substrate biasing (⟨Vbias⟩ = 0 V) while panels (b1) and (b2) are for the film deposited with substrate biasing (⟨Vbias⟩ = −205 V) applied during 10 s of the O2 plasma exposure step.

Hafnium Oxide (HfO)

The growth and material properties of HfO films deposited on planar substrates at 150 °C were investigated for ⟨Vbias⟩ between 0 and −280 V applied during the O2 plasma exposure step. Figure shows these properties in terms of the GPC, refractive index, mass density, residual stress and surface roughness expressed as a function of ⟨Vbias⟩ applied during plasma exposure. The crystalline properties of the HfO films can be analyzed in Figure which shows grazing incidence XRD patterns for films deposited with ⟨Vbias⟩ ranging from 0 to −280 V. Additional information regarding the crystallinity and film microstructure can be obtained in Figure which shows selected area electron diffraction patterns and plan-view HAADF STEM images for HfO deposited on planar TEM windows with and without substrate biasing.

Figure 7

(a) Growth per cycle (GPC), (b) refractive index, (c) mass density, (d) residual stress, and (e) RMS surface roughness of hafnium oxide films deposited at 150 °C expressed as a function of the average bias voltage, ⟨Vbias⟩, applied during the O2 plasma exposure step.

Figure 8

Grazing incidence X-ray diffractograms for hafnium oxide films deposited at 150 °C with average bias voltages, ⟨Vbias⟩, ranging from 0 to −280 V applied during the O2 plasma exposure step. Peaks corresponding to lattice planes in monoclinic hafnium oxide are indicated.

Figure 9

(a1, b1) Selected area electron diffraction patterns and (a2, b2) plan-view high angle annular dark-field (HAADF) STEM images for hafnium oxide films deposited at 150 °C. Panels a1 and a2 are for the film deposited without substrate biasing (⟨Vbias⟩ = 0 V) while panels b1 and b2 are for the film deposited with substrate biasing (⟨Vbias⟩ = −204 V) applied during the O2 plasma exposure step.

For deposition without any bias voltage, the film had a GPC of 1.03 ± 0.02 Å and a refractive index of 2.04 ± 0.03 (Figure a and b) which are comparable to the values reported by Sharma et al.[25] for HfO deposited using TDMACpH and O2 plasma at the same temperature. The film was observed to have a mass density of 9.0 ± 0.2 g/cm3 (Figure c). The lack of diffraction peaks in the XRD pattern (Figure ) indicate the formation of amorphous HfO deposited without biasing. This is confirmed by diffuse rings in the electron diffraction pattern (Figure a1) and absence of lattice fringes in the HAADF STEM image (Figure a2). A void rich amorphous matrix is also revealed in the HAADF STEM image which could explain the low mass density and refractive index of this film when compared to bulk HfO.[41,42] The film had a tensile residual stress of 770 ± 50 MPa (Figure d) which is comparable to that reported for HfO in previous work.[30]

When substrate biasing was implemented during plasma exposure, the GPC showed a slow increase with ⟨Vbias⟩ up to −204 V (Figure a). Further increase in ⟨Vbias⟩ to −280 V led to a rapid increase in GPC. The refractive index and mass density showed a two stage behavior where they initially increased with ⟨Vbias⟩ reaching a maximum of 2.09 ± 0.03 and 9.7 ± 0.2 g/cm3, respectively, at −152 V (Figure b and c) which are closer to the values for bulk HfO.[41,42] Applying and steadily increasing ⟨Vbias⟩ led to the appearance of diffraction peaks in the XRD patterns (Figure ) corresponding to lattice planes of monoclinic HfO. The phenomenon of crystalline material formation with substrate biasing at low temperature discussed previously for TiO was also effectively demonstrated in case of HfO. Increasing ⟨Vbias⟩ to −152 V led to higher diffraction peak intensities and narrower peak widths indicating the presence of a higher crystalline fraction and formation of larger crystal grains.[43] Therefore, the observed increase in GPC, refractive index and mass density with ⟨Vbias⟩ can be initially attributed to the formation of crystalline material at −113 V, followed by an increase in crystalline content and/or grain size at −152 V. The increase in refractive index and mass density with biasing could in principle, also be due to a reduction in void content observed for the film grown without biasing. A similar increase in film crystallinity, refractive index and mass density of HfO films was reported by Sharma et al.[25] due to an increase in substrate temperature. Therefore, the results obtained in this work demonstrate that enhancing the ion energy during PEALD of HfO provides an alternate route for tuning such properties. Unlike the previous case of TiO or any other material investigated in this work, the residual stress of HfO showed a three stage behavior. In the first stage, the tensile residual stress of the HfO films increased with ⟨Vbias⟩ to 1960 ± 50 MPa at −113 V (Figure d). This is in contrast to the trend exhibited by TiO, which showed a change in stress from tensile to compressive when applying a similar ⟨Vbias⟩. The difference in stress behavior could be due to the difference in initial microstructure of the two films grown without biasing. TiO deposited at a higher temperature (300 °C) yielded a highly crystalline and void-free film, whereas HfO deposited at a lower temperature (150 °C) consisted of a void rich amorphous matrix. Therefore, when applying a similar ⟨Vbias⟩ around −100 V, TiO changed from one void-free crystalline state to another, whereas HfO transitioned from a disordered and porous amorphous matrix to a more ordered crystalline structure, with perhaps a lower void content. Further increase in ⟨Vbias⟩ led to the second stage in stress behavior where HfO went from a tensile maximum at −113 V to a compressive maximum at −152 V (Figure d). This may again be related to another change in film microstructure. In this case, HfO transitioned from a less crystalline film with small grains at −113 V to a comparatively more crystalline film with larger grains and possibly, an even lower void content at −152 V. The surface roughness of the films also showed a two stage behavior with substrate biasing. The films initially became smoother on increasing ⟨Vbias⟩ as the RMS roughness decreased to a minimum at −152 V (Figure e). This could again stem from coalescence at the grain boundaries[25,37] due to enhanced mobility of surface species induced by energetic ions during film growth, similar to the case of TiO.

For ⟨Vbias⟩ beyond −152 V, the refractive index and mass density entered their second stage where they now decreased with substrate biasing (Figure b and c). On the contrary, increasing ⟨Vbias⟩ to −204 V led to a more intense and narrower (−111) diffraction peak in the XRD pattern (Figure ). This, combined with the decrease in refractive index and mass density of HfO at −204 V suggests the formation of a more crystalline film with larger grains and a porous microstructure. This is confirmed by the electron diffraction pattern and HAADF STEM image in Figure . Sharp rings corresponding to monoclinic HfO (Figure b1) and crystal lattice fringes (Figure b2) reaffirm polycrystalline material formation using ⟨Vbias⟩ of −204 V, while the dark regions (Figure b2) provide evidence of a void rich microstructure. Application of ⟨Vbias⟩ above −204 V led to a gradual decrease in peak intensities and broadening of peak widths in the XRD patterns (Figure ). This indicates a decrease in crystallinity or material amorphization[14,37,38] under intense ion bombardment. The trend in residual stress entered its third stage for ⟨Vbias⟩ exceeding −152 V due to the onset of compressive stress relaxation (Figure d). A rise in RMS roughness was observed on increasing ⟨Vbias⟩ to −280 V (Figure e) implying that the trend for surface roughness had transitioned to its second stage. These trends for HfO deposited using ⟨Vbias⟩ beyond −152 V are similar to those observed earlier for TiO films grown using high ⟨Vbias⟩. As a result, the degradation of TiO properties due to increase in film void content, plastic deformation[14,36,39,40] and ion bombardment induced surface damage[37] at high ⟨Vbias⟩ could also hold true for HfO deposited using similar conditions. Further analysis regarding the effects of biasing on HfO microstructure, composition and deposition at a higher temperature will be reported in a subsequent publication.

Silicon Oxide (SiO)

Figure shows the GPC, refractive index, mass density, residual stress and surface roughness of SiO films deposited at 200 °C as a function of ⟨Vbias⟩ applied during the O2 plasma exposure step. For the deposition performed without biasing during plasma exposure the film had a GPC of 1.02 ± 0.02 Å, refractive index of 1.45 ± 0.03, mass density of 2.3 ± 0.2 g/cm3 and RMS surface roughness of 1.0 ± 0.2 nm, similar to those reported in previous work by Dingemans et al.[27] In contrast with the tensile residual stress observed for TiO and HfO deposited without biasing, a compressive residual stress of −96 ± 50 MPa was observed for SiO deposited without biasing, similar to that reported previously for SiO deposited at the same temperature.[30]

Figure 10

(a) Growth per cycle (GPC), (b) refractive index, (c) mass density, (d) residual stress, and (e) RMS surface roughness of silicon oxide films deposited at 200 °C expressed as a function of the average bias voltage, ⟨Vbias⟩, applied during the O2 plasma exposure step.

When implementing substrate biasing during the O2 plasma exposure step, the GPC showed a monotonic decrease with increasing ⟨Vbias⟩ and reached a value of 0.62 ± 0.02 Å at −295 V (Figure a). This is in contrast to the trends in GPC observed earlier for TiO and HfO but similar to the monotonic decrease in GPC observed for cobalt oxide reported by Profijt et al.[19] It serves to demonstrate how the trends in growth properties as a function of ⟨Vbias⟩ can be material specific. The refractive index of SiO was observed to increase slightly to 1.47 ± 0.03 by applying ⟨Vbias⟩ of −111 V, but did not increase beyond this value with further increase in biasing. The mass density was observed to remain fairly unchanged with biasing with a slight increase to 2.4 ± 0.2 g/cm3 using ⟨Vbias⟩ of −295 V. The residual stress of the SiO films remained compressive when deposited with substrate biasing and reached a maximum value of −280 ± 50 MPa at −111 V. Further increase in ⟨Vbias⟩ did not lead to any significant change in the compressive stress. The RMS surface roughness also remained fairly constant when using ⟨Vbias⟩ to deposit SiO films. Unlike the significant variations in growth and material properties observed for TiO and HfO previously, the results for SiO showed a significant variation in the growth rate but a very small change in material properties as a function of ⟨Vbias⟩ during plasma exposure. Further analysis of film OH content, optical transmission losses, wet-etch resistance, etc., will be reported in a subsequent publication that will provide more insight on the effects of substrate biasing during PEALD of SiO.

PEALD of Nitrides on Planar Substrates

Titanium Nitride (TiN)

The growth and material properties of TiN films deposited on planar substrates at 200 °C was investigated for ⟨Vbias⟩ between 0 V and −255 V applied during the Ar + H2 plasma exposure step. Figure shows these properties in terms of the GPC, resistivity, mass density, residual stress, and surface roughness expressed as a function of ⟨Vbias⟩ applied during plasma exposure. Table outlines additional growth and material properties from RBS analysis for the films deposited with and without substrate biasing. The growth properties are outlined in terms of the number of titanium, nitrogen and oxygen atoms deposited per nm2 per cycle (GPC [Ti], GPC [N], and GPC [O], respectively) while the material properties are outlined in terms of the N/Ti ratio and impurity content ([C], [Ar], and [H]). Note that both Ti and N species in the deposited TiN films originated from the precursor itself while the reductive Ar + H2 plasma resets the surface by the formation of amino ligands to enable precursor adsorption in the subsequent cycle.[23] The crystalline properties of the TiN films are depicted in Figure which shows grazing incidence XRD patterns for films deposited with ⟨Vbias⟩ ranging from 0 V to −255 V. The microstructure of the TiN films can be observed in Figure which shows plan-view HAADF STEM images for films deposited on planar TEM windows with and without substrate biasing.

Figure 11

(a) Growth per cycle (GPC), (b) resistivity, (c) mass density, (d) residual stress, and (e) RMS surface roughness of titanium nitride films deposited at 200 °C expressed as a function of the average bias voltage, ⟨Vbias⟩, applied during the Ar + H2 plasma exposure step.

Table 3

GPC (Film Thickness per Cycle and Number of Atoms Deposited per nm2 per Cycle) and Elemental Composition of Titanium Nitride Films Deposited at 200 °C with and without Average Bias Voltages, ⟨Vbias⟩, Applied during the Ar + H2 Plasma Exposure Stepa

		RBS							ERD
⟨Vbias⟩	GPC	GPC [Ti] (# Ti at. per	GPC [N] (# N at. per	GPC [O] (# O at. per		[O]	[C]	[Ar]	[H]
(V)	(Å/cycle)	nm2 per cycle)	nm2 per cycle)	nm2 per cycle)	N/Ti	at. %	at. %	at. %	at. %
0	0.47 ± 0.02	1.22 ± 0.03	0.76 ± 0.10	0.92 ± 0.05	0.63 ± 0.06	28 ± 5	<d.l.b	0 ± 0.06	10 ± 7
–130	0.35	1.46 ± 0.01	1.18 ± 0.04	0.10 ± 0.02	0.80 ± 0.02	3 ± 2	<d.l.	0.20 ± 0.05	3 ± 3
–255	0.34	1.18	1.17	0.06	1.0	2	12 ± 3	0.77	2

Typical uncertainties are given in the first and second rows unless otherwise stated.

Values below detection limit (d.l.) of 8 at. % for [C].

Figure 12

Grazing incidence X-ray diffractograms for titanium nitride films deposited at 200 °C with average bias voltages, ⟨Vbias⟩, ranging from 0 to −255 V applied during the Ar + H2 plasma exposure step. Peaks corresponding to lattice planes in cubic titanium nitride are indicated.

Figure 13

Plan-view high-angle annular dark-field (HAADF) STEM images for titanium nitride films deposited at 200 °C with (a) 0 V, (b) −87 V, (c) −130 V, (d) −187 V, and (e) −255 V average bias voltages, ⟨Vbias⟩, applied during the Ar + H2 plasma exposure step.

For film deposition without any biasing, a GPC of 0.47 ± 0.02 Å was obtained (Figure a) which is comparable to that reported for TiN deposited using TDMAT and Ar + H2 plasma at this temperature.[24] The film had a resistivity of 1960 ± 60 μΩcm and a mass density of 3.9 ± 0.2 g/cm3 (Figure b and c, respectively). Compared to the resistivity and density of bulk TiN, the higher resistivity and lower density of the TiN film grown without biasing hint toward the presence of impurities[23,24] and/or void incorporation at grain boundaries.[44,45] This is partly confirmed by the high [O] and [H] impurity contents of 28 ± 5 at. % and 10 ± 7 at. %, respectively, for this film leading to a nonstoichiometric N/Ti ratio of 0.63 ± 0.06 and a GPC [O] value of 0.92 ± 0.05 O at./nm2 (Table ). The GPC [O] is significant when compared to the GPC [Ti] and GPC [N] values (Table ). No inert argon gas or detectable carbon impurities were incorporated in this film (Table ). For deposition involving highly reactive materials such as Ti, Hf, Nb, etc., the partial pressure of the vacuum species (e.g., background water, oxygen) can influence the amount of impurity incorporation (e.g., up to tens of an atomic percent of oxygen) in the deposited film.[26,36,46,47] Assuming a partial pressure of ∼10–8 Torr for background impurities (for the base pressure of ∼10–6 Torr in the FlexAL system) corresponds to an impingement rate of about ∼0.5 impurity atoms per nm2/s on the growing film surface.[36,48] A plasma exposure time of 10 s can therefore, lead to the incorporation of ∼0.5 impurity at./nm2 in the film (assuming a sticking coefficient of 1) which is of the same order as the GPC [O] value obtained for the TiN film grown without biasing. Another reason for the significant [O] content could be due to oxidation of the film upon exposure to the environment. The film was observed to have a tensile residual stress (Figure a) similar to that reported for TiN in previous work.[49] The XRD pattern for this film (Figure ) shows diffraction peaks that correspond to cubic TiN. The fairly small peak intensities and broad peak widths indicate the presence of small crystal grains. This was confirmed by complementary information in the magnified plan-view HAADF STEM image of the film (see Figure S8a). It showed small lateral grain sizes of the order of ∼5 nm based on estimates of lattice fringe areas. A loosely packed microstructure consisting of crystalline grains with porous regions in between was also revealed. The small grain size together with intergranular voids observed in the TEM image could also contribute toward the high resistivity and low density of the TiN film deposited without biasing.

When substrate biasing was applied during the last half (5 s) of the 10 s Ar + H2 plasma exposure step, the trends in GPC, resistivity, mass density, and residual stress again consisted of two stages as observed earlier for the transition metal oxides. In the first stage, the GPC decreased with substrate biasing and reached a minimum value of 0.32 ± 0.02 Å at ⟨Vbias⟩ of −187 V (Figure a). This decrease in GPC for TiN is unlike the trends in GPC observed for the transition metal oxide films discussed previously, but similar to that observed for SiO. This again illustrates how the trends in growth properties as a function of ⟨Vbias⟩ can be material specific. The reduction in film GPC coincided with a 9-fold decrease in GPC [O] with biasing at −130 V (Table ). The large decrease in GPC [O] outweighed the comparatively smaller increase in GPC [Ti] and GPC [N] at −130 V (Table ). As a result, a large reduction in the film [O] content to 3 ± 2 at. % and a simultaneous increase of the N/Ti ratio to 0.8 ± 0.03 (Table ) was observed when ⟨Vbias⟩ was increased to −130 V. The [H] content also decreased significantly to 3 ± 3 at. % while any carbon impurities that may have been incorporated using −130 V still remained below the detection limit (Table ). The significant decrease in [O] and [H] impurity contents and improvement in TiN stoichiometry was reflected in an order of magnitude lower resistivity and a higher mass density of 134 ± 5 μΩcm and 5.3 ± 0.2 g/cm3 (Figure b and c), respectively, at −130 V. Energetic ion bombardment during deposition of TiN[50] and other nitrides[11,51] has been reported to improve material purity by removing adsorbed oxygen/OH species from the growing film surface originating from background impurities in the vacuum environment. Therefore, the large reduction in [O] and [H] impurity content by energetic ion bombardment during Ar + H2 plasma exposure with biasing can contribute toward lowering GPC, while improving film density and electronic conductivity. The residual stress behavior of the TiN films in the first stage changed from tensile to compressive on applying ⟨Vbias⟩ to the substrate (Figure d). The compressive stress reached a maximum as ⟨Vbias⟩ was increased to −130 V, analogous to the minimum in resistivity and maximum in mass density obtained at the same ⟨Vbias⟩. The stress evolution from tensile to compressive for deposition of TiN using energetic particle bombardment could be due to the interplay of several factors discussed in the literature.[44,45,52] A part of the reason could be due to the change in film composition when increasing ⟨Vbias⟩ up to −130 V leading to the formation of cleaner TiN films that have a higher bulk density compared to pure TiO2 or contaminated TiNOH. Another factor that could play a role is inert argon gas incorporation, even below 1 at. %, at interstitial positions which has been reported in the literature to induce significant compressive stress.[44,52] This is in agreement with the small but measurable quantity of 0.20 ± 0.05 at. % for [Ar] in the highly compressive TiN film deposited using −130 V (Table ). Analysis of film crystalline properties revealed that applying ⟨Vbias⟩ initially led to more intense and narrower (111) and (220) peaks while suppressing the (200) peak, as observed in the XRD pattern for −87 V (Figure ). This suggests the presence of a more crystalline material with larger grains which was corroborated by the magnified HAADF STEM image of the film deposited using −87 V (see Figure S8b). It showed increased lateral grain dimensions of the order of ∼10 nm and a more closely packed film microstructure with a reduced void fraction. However, further increase in ⟨Vbias⟩ to −130 V led to a gradual decrease in intensity and broadening of the (111) and (220) peaks in the XRD pattern while the (200) peak slowly reappeared (Figure ). Although this seems to suggest a decrease in the crystalline order or grain size, the TEM image of the film deposited at −130 V (Figure c) revealed a very compact TiN film with a void-free microstructure that retained grain dimensions of ∼10 nm (see also Figure S8c). The presence of a void-free microstructure with large crystal grains for TiN deposited using −130 V could also be another factor behind the high electronic conductivity, mass density and compressive stress discussed earlier for this film. The discrepancy between XRD and the rest of the results could be due to the highly compressive stress measured in this film which can also play a role in diffraction peak broadening[43−45] of XRD patterns besides a decrease in crystal grain size and/or volume fraction.

For ⟨Vbias⟩ greater than −130 V, the GPC continued to decrease until a voltage of −187 V beyond which it entered its second stage and started to rise (Figure a). The GPCs for films deposited using ⟨Vbias⟩ of −130 V and −255 V were comparable to each other (Table ). The values for GPC [Ti], GPC [N], and GPC [O] decreased when ⟨Vbias⟩ was increased from −130 to −255 V (Table ). Although the N/Ti ratio became unity when ⟨Vbias⟩ was increased to −255 V, which seemingly suggested growth of stoichiometric TiN, the [C] and [Ar] contents became significantly large reaching values of 12 ± 3 at. % and 0.77 ± 0.05 at. %, respectively, at that voltage (Table ). The elevated [C] and [Ar] impurities could partly explain the rise in film GPC when ⟨Vbias⟩ was increased from −187 V to −255 V. The [H] content did not change significantly when ⟨Vbias⟩ was increased beyond −130 V (Table ). The trends in resistivity and mass density also entered their second stage when using ⟨Vbias⟩ larger than −130 V with the resistivity becoming larger and the mass density decreasing. Furthermore, increase in ⟨Vbias⟩ beyond −130 V also initiated the next stage in residual stress behavior with a relaxation of the compressive stress (Figure d), similar to the trends in resistivity and mass density. For the (111) and (220) diffraction peaks in the XRD pattern, the intensities and widths became significantly smaller and wider, respectively, while the (200) peak became more prominent on increasing ⟨Vbias⟩ beyond −130 V (Figure ). These observations were in line with the onset of amorphization induced by highly energetic ion bombardment,[14,37,38] similar to the cases of TiO and HfO discussed earlier, and could contribute in lowering the electronic conductivity and mass density of the TiN films. Furthermore, the relaxation of compressive stress on increasing ⟨Vbias⟩ beyond −130 V also indicates that the yield strength of the material could have been exceeded under intense ion bombardment leading to plastic deformation.[14,36,39,40] These aforementioned trends in material properties were concomitant with the reappearance of voids in the film microstructure for TiN deposited with ⟨Vbias⟩ greater than −130 V (Figure d and e). The void fraction increased with increasing ⟨Vbias⟩ which could also play a role in compressive stress relaxation together with decreasing conductivity and mass density. Other factors that could contribute toward relieving compressive stress include bubble formation from excess argon incorporation[44,53] and change in composition to TiNC owing to the significant carbon impurity content seen for the film deposited using −255 V. The RMS roughness of the TiN film surface seemed to remain fairly constant (around ∼1 nm) when using ⟨Vbias⟩ during the plasma exposure step.

Hafnium Nitride (HfN)

Figure shows the GPC, resistivity, mass density, residual stress and surface roughness of HfN films deposited at 450 °C as a function of ⟨Vbias⟩ applied during the H2 plasma exposure step. All films were deposited with ⟨Vbias⟩ applied for the entire duration of the 10 s plasma exposure step (solid symbols) apart for a film deposited at ⟨Vbias⟩ of −210 V applied during the last half (5 s) of the 10 s plasma exposure step (hollow symbols). For the deposition without substrate biasing during plasma exposure, the film had a GPC of 0.35 ± 0.03 Å and a resistivity of (90 ± 7) × 104 μΩcm (Figure a and b) which are comparable to the values reported by Karwal et al.[26] for HfN deposited using TDMACpH and H2 plasma at the same temperature. The low film conductivity was attributed to a high oxygen impurity content of 15 at. % that got incorporated in the film mainly during deposition due to background water present in the ∼10–6 Torr vacuum environment of the ALD system,[26] similar to what was speculated previously for TiN. The HfN film deposited without biasing was also observed to have a mass density of 10.1 ± 0.2 g/cm3, tensile residual stress of 1786 ± 50 MPa and an RMS surface roughness of 2.5 ± 0.2 nm (Figure c, d and e).

Figure 14

(a) Growth per cycle (GPC), (b) resistivity, (c) mass density, (d) residual stress, and (e) RMS surface roughness of hafnium nitride films deposited at 450 °C expressed as a function of the average bias voltage, ⟨Vbias⟩, applied for the entire duration and last half (5 s, only for ⟨Vbias⟩ = −210 V) of the 10 s H2 plasma exposure step.

For the 10 s bias duration condition (bias duty cycle = 100%, Figure b), the GPC showed a monotonic increase with increasing ⟨Vbias⟩ (Figure a). This is in contrast to the initial dip followed by a gradual rise in GPC as a function of ⟨Vbias⟩ observed earlier for TiN, but similar to the corresponding trends for TiO and HfO. However, the film resistivity, mass density and residual stress showed a two stage behavior with increasing ⟨V⟩, similar to the trends seen before for TiN. The use of substrate biasing during the plasma exposure step was observed to trigger a drastic improvement in the electronic properties of HfN through a reduction in film resistivity by 2 orders of magnitude (Figure b). For the 10 s bias duration condition, increasing ⟨Vbias⟩ led to an initial decrease in film resistivity that reached a minimum value of (33 ± 7) × 102 μΩcm at −130 V. The mass density initially increased slightly with biasing and reached a peak at the same voltage (Figure b). In the initial stage, the residual stress changed from tensile to compressive on applying ⟨Vbias⟩ during plasma exposure. The compressive stress reached a maximum of −2004 ± 50 MPa on increasing ⟨Vbias⟩ to −130 V, similar to the minimization of resistivity and peaking of mass density at the same voltage. The RMS surface roughness of the films was observed to decrease slightly with biasing (Figure e). For ⟨Vbias⟩ beyond −130 V applied for 10 s of plasma exposure, an increase in film resistivity, reduction in mass density and relaxation of the compressive stress (Figure b–d) signaled a transition to the second stage of HfN property variation with substrate biasing. Given the similar trends in material properties for PEALD of TiN and HfN with substrate biasing, factors that could contribute toward the initial rise in film conductivity, mass density and compressive stress include a reduced oxygen content, improved stoichiometry (i.e., N/Hf ratio closer to 1), enhanced grain size or crystalline content, reduced void content and grain boundary density, all induced by energetic ion bombardment. Similarly, for ⟨Vbias⟩ beyond −130 V, the onset of ion bombardment induced amorphization,[14,37,38] plastic deformation,[14,36,39,40] void creation or carbon impurity incorporation could play roles in lowering film conductivity, mass density, and compressive stress.

Using ⟨Vbias⟩ of the same magnitude but different duration/duty cycle during plasma exposure (bias duty cycle = 50%, Figure c) created a difference in the way that the growth and material properties of HfN were altered, similar to the case observed earlier for TiO. The growth rate of the HfN film deposited using −210 V applied during the last half (5 s) of the 10 s H2 plasma exposure step was significantly lower than the corresponding value obtained using the longer bias duration of 10 s (Figure a). A lower resistivity and higher mass density were obtained for the film deposited with the shorter 5 s bias duration compared to the corresponding values of the film grown with the longer 10 s bias duration for the same ⟨Vbias⟩ magnitude of −210 V (Figure c). Similarly, the compressive stress at −210 V was higher when implementing the shorter bias duration of 5 s during H2 plasma exposure (Figure d). The surface was also smoother in case of the film grown using 5 s of bias (Figure e). These results again demonstrate how altering the bias duration/duty cycle leads to a varying dose or fluence (i.e., particle flux integrated over time) of energetic ions impinging on the substrate, which can provide an alternative route for tuning material properties[13,14,21] during PEALD. Further analysis on the effects of ion energy control with substrate biasing during PEALD of HfN will be reported in a subsequent publication.

Silicon Nitride (SiN)

The growth and material properties of SiN films deposited on planar substrates at 500 °C was investigated for ⟨Vbias⟩ between 0 and −103 V applied during the last half (10 s) of the 20 s N2 plasma exposure step. Figure shows these properties in terms of the GPC, refractive index, mass density, residual stress and surface roughness expressed as a function of ⟨Vbias⟩ applied during plasma exposure. Table outlines additional growth and material properties from RBS analysis for the films deposited with and without substrate biasing. The growth properties are given in terms of the number of silicon and nitrogen atoms deposited per nm2 per cycle (GPC [Si] and GPC [N], respectively) while the material properties are outlined in terms of the N/Si ratio and impurity content ([O], [C], and [H] at. %).

Figure 15

(a) Growth per cycle (GPC), (b) refractive index, (c) mass density, (d) residual stress, and (e) RMS roughness of silicon nitride films deposited at 500 °C expressed as a function of the average bias voltage, ⟨Vbias⟩, applied during the N2 plasma exposure step.

Table 4

GPC (Film Thickness Per Cycle and Number of Atoms Deposited per nm2 per Cycle) and Elemental Composition of Silicon Nitride Films Deposited at 500 °C with and without Average Bias Voltages, ⟨Vbias⟩, Applied during the N2 Plasma Exposure Stepa

		RBS					ERD
⟨Vbias⟩	GPC	GPC [Si]	GPC [N]		[O]	[C]	[H]
(V)	(Å/cycle)	(# Si at. per nm2 per cycle)	(# N at. per nm2 per cycle)	N/Si	at. %	at. %	at. %
0	0.14 ± 0.02	0.55 ± 0.02	0.79 ± 0.04	1.4 ± 0.1	2 ± 2	<d.l.b	4 ± 3
–103	0.21	0.53	0.91	1.7	10	8 ± 3	6

Typical uncertainties are given in the first row unless otherwise stated.

Values below detection limit (d.l.) of 8 at. % for [C].

In the absence of substrate biasing during plasma exposure, the film had a GPC, refractive index, and mass density of 0.14 ± 0.02 Å, 1.96 ± 0.03, and 2.9 ± 0.2 g/cm3 (Figure a–c), respectively. These values are comparable to the results reported in our previous work[28] for SiN deposited using DSBAS and N2 plasma at the same temperature. The residual stress of this film was found to be compressive in nature (Figure d) similar to that reported previously for SiN deposited at a comparable temperature.[54] The nature of the stress for SiN is unlike the tensile stresses observed earlier for the transition metal compounds deposited without biasing but similar to the compressive stress obtained for SiO. An N/Si ratio of 1.4 ± 0.1 was observed for this film (Table ) indicating the formation of nearly stoichiometric Si3N4 (N/Si ratio of 1.33) for deposition without any biasing. Low impurity contents in terms of [O], [C], and [H] (2 ± 2 at. %, < 8 at. % and 4 ± 3 at. %, respectively) were measured for this film (Table ) indicating the formation of high quality SiN.

For deposition with substrate biasing, the GPC showed a monotonic increase with ⟨Vbias⟩ (Figure a). Although GPC [Si] remained fairly unchanged, the increase in GPC could be partly explained by an increase in GPC [N] that reached a value of 0.91 ± 0.04 N at./nm2 at −103 V (Table ). This indicated the formation of a nitrogen rich film with increase in ⟨Vbias⟩, which was corroborated by a higher N/Si ratio of 1.7 obtained for deposition using −103 V (Table ). The refractive index and mass density showed a monotonic decrease with ⟨Vbias⟩ (Figure b and c). The trends in refractive index and mass density were accompanied by a relaxation of the compressive stress as a function of ⟨Vbias⟩ (Figure d). The film surface became slightly smoother as the RMS roughness decreased with increasing ⟨Vbias⟩ (Figure e). The impurity contents increased with ⟨Vbias⟩ as observed by the elevated [O], [C], and [H] values at −103 V (Table ). The simultaneous decrease in refractive index, mass density, and compressive stress together with an increase in the impurity content indicate that energetic ion bombardment through substrate biasing during PEALD of SiN degrades its material properties on planar substrates. These monotonic trends in refractive index, mass density and residual stress of SiN deposited using relatively small ⟨Vbias⟩ (≤−100 V) are unlike any of the trends observed previously for other materials investigated in this work. It suggests that even the slightest increase in ion energies during N2 plasma exposure could induce plastic deformation,[14,36,39,40] excess nitrogen content, N2 gas bubble formation,[14,55] and void incorporation, which could then be factors responsible for film quality degradation. This sort of material degradation was also observed in the other materials (except SiO) but only when using comparatively higher ⟨Vbias⟩ (≥−130 V), indicating the presence of higher ion energy thresholds for degradation in those materials.

Thickness Uniformity on 200 mm Substrates

The thickness uniformity of films on large-area planar substrates (200 mm c-Si wafers) was investigated for all six oxides and nitrides deposited with and without substrate biasing. Note that the results obtained were not from optimized precursor dose and plasma exposure times intended for attaining the best uniformities but serve to demonstrate the role of substrate biasing on thickness uniformity during PEALD on large-area substrates. The thickness nonuniformity was defined as the ratio of the standard deviation in thicknesses measured at several points across the wafer to the average thickness of all data points measured. The values obtained are expressed as percentages in Table . The results show that for most cases the thickness uniformity of films on 200 mm substrates improved when using substrate biasing during PEALD, except for HfO and SiN, where the uniformity remained fairly unchanged. For TiO deposited with biasing, the change in thickness nonuniformity was also observed to depend on the duration or duty cycle of the bias applied during O2 plasma exposure. The use of a longer bias duration (10 s) increased nonuniformity to 2.8%, whereas the same ⟨Vbias⟩ applied for a shorter bias duration (5 s) decreased nonuniformity to 2.0% relative to the nonuniformity of the film grown without biasing (2.3%). The two transition metal nitride films deposited without biasing had the highest thickness nonuniformities on 200 mm planar substrates (7.5% for TiN, 9.4% for HfN). The use of substrate biasing during PEALD of these two nitrides led to the most significant improvements in thickness uniformity among all the materials. For TiN deposited using −187 V, the thickness nonuniformity decreased to 6.6%, while for HfN deposited using −210 V (applied for the entire duration of the 10 s H2 plasma exposure step), the thickness nonuniformity decreased to 7.7%. Interestingly, when HfN was deposited using the same ⟨Vbias⟩ but a lower bias duration of 5 s, the thickness nonuniformity decreased to an even lower value of 6.7%. These results demonstrate that both the ion energy and the dose of higher energy ions can play roles in altering film uniformity.

Table 5

Thickness Uniformity of Films Deposited on 200 mm Planar c-Si Wafers without and with Average Bias Voltages, ⟨Vbias⟩, Applied during Plasma Exposure Steps

	No substrate bias (%)	With substrate bias (%)	⟨Vbias⟩ (V)
titanium oxide	2.3	2.8a, 2.0b	–205
hafnium oxide	4.6	4.6	–152
silicon oxide	1.1	0.7	–200
titanium nitride	7.5	6.6	–187
hafnium nitride	9.4	7.7a, 6.7b	–210
silicon nitride	3.5	3.6	–103

Substrate biasing applied for the entire duration of the plasma exposure step.

Substrate biasing applied for the last half of the plasma exposure step.

PEALD on 3D Substrates

The cross-sectional TEM image of a TiO layer on 3D trench nanostructures is shown in Figure . The film was deposited at 150 °C using −205 V applied during the entire duration of the 10 s O2 plasma exposure step. Film conformalities of 84% and 100% were measured at the bottom-side and bottom regions of the trench, respectively (see Figure S9). As discussed before, an increase in film GPC was observed during PEALD of TiO with substrate biasing on planar substrates. This can cause film GPCs at the planar top and bottom surfaces of the 3D trench nanostructure to exceed the GPC at the vertical bottom-side region and thereby yield a lower conformality at that region of the 3D substrate. Furthermore, the effect of enhanced ion energies is also visible in Figure in terms of the different morphologies of the TiO film at planar and vertical surfaces of the 3D trench nanostructures. Previously, it was shown that for PEALD of TiO on a planar substrate at 150 °C, an amorphous film was obtained when the deposition was performed without substrate biasing (Figure a1 and a2). However, using −205 V during the 10 s O2 plasma exposure step resulted in the formation of polycrystalline TiO (Figure b1 and b2) in the rutile phase (see Figure S7) on a planar substrate. As a result, the TiO film regions formed at planar top and bottom surfaces of the 3D trench nanostructures in Figure can be assumed to consist of polycrystalline grains formed in the rutile phase, while the TiO film deposited at the vertical sidewall surface regions can be considered to be amorphous. This can be explained by the directional or anisotropic nature of ions when they travel through a nearly collisionless plasma sheath[56] before impinging on a substrate. Using substrate biasing makes the directional ions impinge on the planar surface regions of the 3D trench nanostructures with much more energy than on its vertical sidewalls, thereby inducing differing film properties at different regions of the 3D substrate.

Figure 16

Cross-sectional TEM image of titanium oxide film on 3D trench nanostructures (AR = 4.5:1) deposited at 150 °C with an average bias voltage (⟨Vbias⟩ = −205 V) applied during the entire duration of the 10 s O2 plasma exposure step. The presence of crystalline material (rutile) on planar surfaces and amorphous material on vertical sidewall surfaces are indicated.

The cross-sectional TEM image of a HfO layer on 3D trench nanostructures is shown in Figure . The film was deposited at 150 °C using −204 V applied during the O2 plasma exposure step. Film conformality of 71% and 80% was measured at the bottom-side and bottom regions of the trench, respectively, (see Figure S10) which is comparable with the results reported by Sharma et al.[25] for HfO deposited on such 3D trench nanostructures. Similar to the result for TiO discussed in the previous section, the role of energetic ion bombardment with substrate biasing during deposition is again observed in Figure where the deposited HfO film has different morphologies on the planar and vertical surfaces of the 3D trench nanostructures. Earlier on, it was shown that PEALD of HfO on a planar substrate at 150 °C yielded an amorphous film when no ⟨Vbias⟩ was applied during plasma exposure (Figure a1 and a2). The use of −205 V during the O2 plasma exposure step led to the growth of polycrystalline HfO (Figure b1 and b2) with a monoclinic crystal lattice structure (Figure ). Consequently, the HfO film region at the planar top and bottom surfaces of the 3D trench nanostructures in Figure can be expected to consist of polycrystalline monoclinic grains while the HfO film grown at the vertical sidewall surface regions can be assumed to be amorphous. This can again be attributed to the role of directional ions traveling through a nearly collisionless plasma sheath that bombard planar surface regions of the 3D trench nanostructures with more energy than the vertical sidewalls when implementing substrate biasing. The results obtained for HfO effectively reproduce the phenomenon seen earlier for TiO, where enhancing the energy of directional oxygen ions during film deposition induced simultaneous growth of different phases (crystalline and amorphous) of the same material at different surfaces (planar and vertical) of a 3D substrate.

Figure 17

Cross-sectional TEM image of hafnium oxide film on 3D trench nanostructures (AR = 4.5:1) deposited at 150 °C with an average bias voltage (⟨Vbias⟩ = −204 V) applied during the O2 plasma exposure step. The presence of crystalline material (monoclinic) on planar surfaces and amorphous material on vertical sidewall surfaces are indicated.

Cross-sectional TEM images of SiN layers on 3D trench nanostructures are shown in Figure . The images for as-deposited films grown at 500 °C without any substrate biasing and with −103 V during N2 plasma exposure are shown in Figure A and B, respectively. Conformalities of 37% and 56% were obtained at the bottom-side and bottom of the trench structures, respectively, for the film deposited without any ⟨Vbias⟩ (Figure A), similar to the values reported by in our previous work[28] for SiN deposited on such 3D substrates. The bottom-side film conformality improved slightly to 39% while that at the bottom improved significantly to 78% for the film deposited using −103 V during N2 plasma exposure (Figure B). Note that some significant corner tapering was observed for the SiN film deposited with −103 V. The quality of these SiN films was analyzed by measuring their etch resistance in dilute hydrofluoric acid solution (HF:H2O = 1:100). The as-deposited films underwent a 30 s wet-etch treatment in dilute HF and cross-sectional TEM images of the films were measured after the etch treatment. Figure A′ and B′ show the post wet-etch TEM images of the films deposited without any substrate biasing and with −103 V applied during N2 plasma exposure, respectively. The wet-etch rate (WER) values were determined by comparing the as-deposited and post wet-etch film thicknesses at three different regions (top, bottom-side, bottom) of the 3D trench nanostructures which are outlined in Table . The SiN film deposited without biasing seemed to be highly etch resistant at the planar top and bottom regions of the trench as indicated by the small or insignificant WER values (≤1 ± 1 nm/min, Table ) at those regions. These low WERs indicate the formation of high quality SiN films on planar top and bottom surfaces of the 3D trench nanostructures for deposition without any substrate biasing. This is also in agreement with the high mass density, refractive index (Figure b, c) and low impurity content (Table ) of SiN deposited without biasing on a planar c-Si substrate, as discussed earlier. The WER observed at the bottom-side region of the trench (3 ± 1 nm/min, Table ) is relatively higher than those at the two planar trench regions, indicating a reduced HF-etch resistance of SiN at the vertical trench sidewalls. However, the value is small in absolute magnitude indicating an overall high HF-etch resistance at both planar and vertical trench surfaces for the film deposited without substrate biasing. For the SiN film deposited with −103 V during N2 plasma exposure, the film regions located at the planar top and bottom surfaces of the trench exhibited very high WERs and were completely removed after the etch treatment (Figure B′). These high WERs indicate the degradation of SiN films on planar top and bottom surfaces of the 3D trench nanostructures for deposition with substrate biasing. This is corroborated by the lower refractive index, mass density (Figure b, c) and higher impurity content (Table ) observed earlier for SiN deposited on planar substrates using −103 V during N2 plasma exposure. However, SiN film regions located at the vertical sidewalls of the trench nanostructures exhibited low WERs (∼2 ± 1 nm/min, Table ), similar to that for the film deposited without biasing, and remained selectively at the trench sidewalls after the wet-etch treatment (Figure B′). This can again be explained by the combined factors of high energy and directionality of ions traveling through a nearly collisionless plasma sheath when ⟨Vbias⟩ is applied to the substrate. These energetic ions collide with much more energy on the horizontal top and bottom surfaces of the 3D trenches than the vertical sidewalls. Consequently, the SiN films growing at the top and bottom trench regions are selectively degraded in comparison to those growing at the sidewalls. As a result, the film regions obtained at those sidewalls for deposition with substrate biasing retain the high quality and HF-etch resistance inherent for films formed at the same vertical surfaces for deposition without any ⟨Vbias⟩.

Figure 18

Cross-sectional TEM images of (A, B) as-deposited and (A′, B′) post wet-etch (in 30 s dilute HF) silicon nitride films grown on 3D trench nanostructures (AR = 4.5:1) at 500 °C. Images (A) and (A′) are for films deposited without substrate biasing (⟨Vbias⟩ = 0 V), while images (B) and (B′) are for films deposited with substrate biasing (⟨Vbias⟩ = −103 V). Film conformality at the bottom-side and bottom regions of the trench are indicated as a percentage of film thickness at the top of the trench.

Table 6

Wet-Etch Rates of Silicon Nitride on 3D Trench Nanostructures (AR = 4.5:1) for Films Deposited at 500 °C without (⟨Vbias⟩ = 0 V) and with Substrate Biasing (⟨Vbias⟩ = −103 V) Applied during N2 Plasma Exposurea

	wet-etch rate (nm/min)
⟨Vbias⟩ (V)	top	bottom-side	bottom
0	0 ± 1	3 ± 1	1 ± 1
–103	completely etched	2 ± 1	completely etched

The wet-etch rates are reported for silicon nitride films located at planar (top, bottom) and vertical (bottom-side) regions of the 3D substrate topographies after 30 s dip in an etchant solution of dilute hydrofluoric acid (HF:H2O = 1:100). Typical uncertainties are given in the first row.

Discussion

⟨Vbias⟩ Regimes during PEALD for Tuning Properties

The results obtained in this work demonstrate how substrate biasing during PEALD can have prominent effects on the growth or material properties of oxide and nitride thin-films. In the case of the four transition metal compounds, the trends in material properties as a function of ⟨Vbias⟩ mostly showed a two stage behavior composed of an initial buildup (or decrease) followed by a gradual fall (or increase) with increasing ⟨Vbias⟩. These trends observed for the large range of ⟨Vbias⟩ investigated in this work enabled identification of regimes for material property improvement (i.e., higher refractive index, conductivity, and mass density) and degradation in PEALD processes. ⟨Vbias⟩ up to approximately −150 V during O2 plasma exposure and −130 V during H2 or Ar + H2 plasma exposure improved material properties of the transition metal oxide and nitride films, respectively. Applying ⟨Vbias⟩ beyond these values showed material property degradation. For silicon oxide and silicon nitride, a one stage behavior was observed where material properties improved slightly for the former and instantly degraded for the latter as a function of substrate biasing. Besides varying ⟨Vbias⟩ magnitude, reducing the duration of substrate biasing during plasma exposure extended the ⟨Vbias⟩ regime for improving TiO material properties to about −200 V. These results indicate that regimes exist not only for the ion energy but also for the dose of higher energy ions. Consequently, both the magnitude of ⟨Vbias⟩ and the duration/duty cycle of the applied bias can be used as tunable parameters for influencing ion-surface interactions by modifying the ion energy and dose/fluence of higher energy ions, respectively. It is interesting to note that even though an H2 plasma generates lighter ions (e.g., H3+, H2+) than an Ar + H2 plasma (e.g., ArH+, Ar+), similar trends in material properties were observed for the two transition metal nitrides when grown using these two plasmas with substrate biasing. This will be investigated in future work on the basis of growing a transition metal nitride using these plasmas with substrate biasing and measuring the corresponding ion characteristics (e.g., energy, flux, etc.) and material properties. Furthermore, the trends in material property variation reported for substrate biasing during conventional flux-controlled deposition processes (PECVD or PVD) are similar to those observed in the surface-controlled PEALD processes of this work for all films[44,57−60] except SiN. Enhancing ion energies during PEALD of SiN with relatively small ⟨Vbias⟩ (<−100 V) lowered mass density, refractive index, and compressive stress while simultaneously increasing WER which are opposite to the trends reported for SiN PECVD or PVD processes with substrate biasing.[61,62] Even for substrate biasing during PEALD, the mass density of TiO as a function of ⟨Vbias⟩ increased for films in this work but decreased for films deposited in previous work reported by Profijt et al., who used different Ti precursors.[19,20] The unique trends in property variation with substrate biasing signify how the effects of controlling ion-surface interactions can be highly material and/or process specific. It therefore, necessitates further empirical investigation of other materials and process conditions for PEALD with substrate biasing.

As mentioned in the Introduction, several inherent characteristics of the material deposited using varying ion energies or doses of higher energy ions can also influence the final film properties. It is beyond the scope of this work to provide a detailed account correlating all the characteristics with the observed results, so a brief discussion is made based on two factors. For instance, the bond energies are higher for oxides relative to nitrides for the six materials investigated in this work (see Table S1). This could play a role in the transition metal oxides showing a higher ⟨Vbias⟩ threshold than the corresponding nitrides before the onset of material degradation (i.e., reduction of mass density, increase in film void content and amorphization). Indeed, Si–O has the highest and Si–N the lowest bond energy among the six materials which could also be a factor behind SiO material properties remaining fairly unchanged even at large ⟨Vbias⟩ (>−200 V), while SiN films degraded at low ⟨Vbias⟩. Furthermore, the four transition metal compounds investigated in this work have been reported to form in a crystalline phase for PEALD processes between 100 and 250 °C (see Table S2). PEALD processes for SiO and SiN reported in the literature so far have yielded only amorphous films since these materials require much higher temperatures for crystallization (>900 °C, see Table S2) either during deposition or postdeposition annealing treatment. The high crystallization temperature could be a reason why the silicon compounds showed a one stage behavior compared to the two stage behavior of the transition metal compounds whose crystalline properties (phase, volume fraction) varied with enhanced ion energies.

Prospects for Substrate Biasing during PEALD

On the basis of the results obtained in this work, the implementation of substrate biasing during PEALD can provide several opportunities for enhancing the capabilities of atomic scale processing. Controlling the energy and/or dose of ions during the plasma exposure step can enable the tuning of a wide range of material properties for the same precursor/plasma reactants and deposition temperature. Furthermore, the crystalline phase of certain materials such as the transition metal compounds can be obtained using low temperatures at which PEALD without substrate biasing typically yields amorphous films. This also eliminates the requirement of specific substrate surfaces to deposit a high temperature crystalline phase material at a low temperature, for example, the need for a RuO2 surface to deposit rutile phase TiO2 at 250 °C.[63] This phase is known to have a high dielectric constant which is a desired property for semiconductor device applications.[63] Furthermore, mixed phase anatase + rutile TiO2 is reported to have a higher photocatalytic activity[64] than the individual phases while films with small grains and/or a porous microstructure (i.e., high surface to volume ratio) have a higher sensitivity for gas sensing.[65] During PEALD of TiN in this work, the Ti–N bonds in the deposited film originated from Ti–N bonds present in the TDMAT precursor. Substrate biasing during the subsequent reducing Ar + H2 plasma step altered material composition by lowering [O] content that could have originated from background impurities in the vacuum environment. The chemical composition could also be modified at high ⟨Vbias⟩ by an increase in [C] content. While a low [O] and [C] content yielded more conductive TiN, an increase in [C] content could have an influence on the material work function.[66] The decrease in [O] content enabled by substrate biasing during a reducing plasma could in principle, be applicable for varying the composition of metal oxide films deposited with precursors containing metal–O bonds. Recently, it was shown that substoichiometric TaO films could be obtained with PEALD using Ta(OC2H5)5 as the precursor (that already contained Ta–O bonds), followed by a reducing Ar + H2 plasma.[67] This process scheme led to the incorporation of oxygen vacancies in the deposited metal oxide. In this regard, it can be speculated that enhancing ion energies during the reductive plasma step could potentially confer in situ control over the amount of generated oxygen vacancies. As a result, PEALD with substrate biasing could be a method for growing substoichiometric metal oxide layers with a tunable composition, relevant for emerging applications such as resistive nonvolatile memory devices (e.g., RRAMs).[68]

Substrate biasing during PEALD has thus far been implemented in the second step of a two-step PEALD process, that is, during the B step of an (AB) process where n is the number of cycles. It could also be implemented in multistep PEALD processes, such as during the third C step of an (ABC) cycle or an [(AB)C] supercycle, where x is the number of AB cycles completed before the C step is performed and n is the number of supercycles (see Figures S1 and S2).[18] Furthermore, substrate biasing during PEALD has so far been carried out using reactive plasmas where enhancing ion energies result in combined physicochemical process during film growth. Using an inert plasma (e.g., Ar) in the C step of the aforementioned multistep cycles and applying a bias only at that step could in principle, lead to purely physical effects during energetic ion bombardment. Such purely physical processes could induce differences in the trends of material property variation as a function of substrate biasing. Other variants of multistep PEALD supercycles are used to deposit layered film stacks (or nanolaminates[69]) that rely on the differences in properties of the individual layers. These differences are generally procured by growing at least two different materials alternately for the stacked layers. Since enhanced ion energies during PEALD can tailor the properties of a given material, stacked layers with different properties using the same material could in principle, be created by growing one layer without (Figure a) and the next layer with substrate biasing (Figure b or c). This technique could also be employed for growing films on sensitive substrates that cannot withstand the effects of enhanced ion energies. Deposition with biasing can be carried out only after initially performing some PEALD (or even thermal ALD) cycles without any biasing on such substrates. This could in principle, lead to energetic ion-surface interactions on the initial layer of the deposited film and not on the substrate (provided the initially grown layer is of a sufficient thickness). For stacked layers comprised of different materials, substrate biasing during PEALD can enable better control over the properties of those layers, for example, by reducing tensile stress of HfO2 layers in HfO2/Al2O3 bilayers to prevent delamination.[30] Film growth with steadily increasing/decreasing bias voltages with each subsequent cycle could also enable tailoring of interface properties, for example, formation of a gradually changing instead of an abrupt interface.

Furthermore, substrate biasing during plasma exposure can also be implemented as pre- or postdeposition treatment steps which, based on the reactants used and the treatment duration, can modify surface or subsurface regions (i.e., a few monolayers below the surface). This can enable in situ surface functionalization that can either promote or inhibit film growth. For the case of substrate biasing applied in an interleaved manner in the plasma step (i.e., during a part of the plasma exposure time), another aspect that could potentially influence material properties is variation of the moment at which the bias is applied. For instance, a 50% bias duty cycle can be implemented in a 10 s plasma step by applying it for 5 s during either the first, middle or last half of the plasma exposure time. Simultaneous ignition of the plasma and the bias could lead to the cracking of bulky ligand species (e.g., amino, cyclopentadienyl, isopropoxide, etc.) remaining on the surface after precursor adsorption by the impact of high energy ions. This could lead to the incorporation of decomposed ligand species in the growing film and elevate film impurity content. Therefore, a means to prevent this could be to apply the bias near the end of the plasma exposure step as used in this work which should, in principle, lead to a cleaner or more ligand-free surface before the impingement of energetic ions.

Substrate biasing during PEALD of the two transition metal oxides on 3D trench nanostructures led to the simultaneous formation of different material phases at different surface orientations during growth of the same material in the same deposition run. This indicates how controlling the energy of directional ionic species in a collisionless plasma sheath allows for selective processing during PEALD on 3D substrates. A similar effect was also observed in case of SiN where enhanced ion energies during film growth induced selective degradation of material properties for the same material at different surface regions of a 3D substrate. Different categories of selectivity during film growth have been previously defined in terms of area-, phase-, microstructure-, or chemical composition-selective deposition.[70] Recently, a new approach for carrying out selective deposition on 3D substrates was reported by Kim et al.[3] Selective surface treatment of planar regions in 3D trench nanostructures was performed using directional ions that enabled selective anisotropic deposition of Pt only on the vertical sidewalls of the trenches. This approach for growing films in an area-selective manner on 3D substrates having different geometries or topographical orientation was termed as topographically selective deposition. In this regard, the results obtained in this work demonstrate how ion energy control with substrate biasing during PEALD offers an alternative pathway for topographically selective deposition. Based on the aforementioned selective growth categories, substrate biasing during PEALD on 3D substrates can be considered to induce microstructure (or chemical composition) selective growth of SiN and phase selective growth of TiO and HfO at different surface regions of the 3D substrates. Implementing such variants of selective growth on 3D substrates therefore, enables different routes for conducting topographically selective deposition. The selectivity attribute imparted by the directional impingement of high energy ions could in principle, be tuned (attenuated) by controlling (lowering) the directionality and/or energy of the ions. When substrate biasing is implemented in a collisional plasma sheath that can be obtained at high pressure conditions (e.g., > 100 mTorr), both the directionality and the energy of the ions are affected. At these conditions, the ion mean free path becomes smaller than the thickness of the plasma sheath causing the ions to experience collisions with gas-phase species while crossing the sheath. These collisions alter the vertical trajectory of the ions while transferring both charge and energy to neutral gas-phase species such as the reactive plasma radicals. As a result, the energy of ions impinging on planar substrate surfaces can become considerably lower than the voltage drop across the plasma sheath. The gas-phase collisions could also lead to the impingement of both ions and radicals on the vertical surfaces of a 3D structure. Depending on the pressure and magnitude of substrate bias applied during plasma exposure, the energy of the ions and radicals colliding with sidewalls could in principle, be higher than the energy of the same species interacting with the sidewalls in a nearly collisionless plasma sheath. This could have direct implications on the conformality and material properties of films deposited on 3D substrates. Therefore, substrate biasing during PEALD on 3D substrates with highly collisional plasma exposures offer new avenues for investigation.

Conclusions

The effects of modifying ion-surface interactions during PEALD on the growth and material properties of oxide (TiO, HfO, SiO) and nitride (TiN, HfN, SiN) thin-films have been investigated using a commercial 200 mm remote plasma ALD system equipped with RF substrate biasing. The magnitude of ⟨Vbias⟩ and the duration/duty cycle of the applied bias during plasma exposure were demonstrated as parameters for influencing ion-surface interactions by modifying the ion energy and dose/fluence of higher energy ions, respectively. Controlling these parameters had significant material and/or process specific effects that enabled tailoring of a wide range of properties during film growth. These include, but are not limited to, the mass density, optical refractive index, electrical resistivity, residual stress, crystalline properties (e.g., volume fraction, phase, grain size, etc.), void fraction, surface roughness, thickness uniformity (on large area planar substrates), and chemical composition, which are summarized in Figure .

Figure 19

Schematic illustration representative of the material properties and process control enabled by substrate biasing during PEALD on planar and 3D substrate topographies.

As long as ion energies remained below the regimes for ion-induced degradation, enhancing ion energies by increasing ⟨Vbias⟩ led to an improvement in the mass density for all materials except SiN, which degraded with substrate biasing. A higher mass density was generally accompanied by an increase in the optical refractive index of dielectric films. Growth of a denser material with biasing also indicated the presence of a compact film with a void-free microstructure while the growth of underdense material generally led to an increase in the film void content. Enhancing ion energies with substrate biasing significantly reduced the electrical resistivity of the two transition metal nitrides. This was also accompanied by a significant reduction of the oxygen impurity content in TiN while implementing a high ⟨Vbias⟩ elevated the film carbon content. These serve to demonstrate how substrate biasing can be used for tuning the chemical composition of materials. The residual stress could be tailored from tensile under no bias conditions to compressive with substrate biasing for the transition metal compounds whereas that for the silicon compounds remained compressive as a function of ⟨Vbias⟩.

Substrate biasing during PEALD on 3D trench nanostructures effectively depicted the role of directional ion bombardment by inducing distinct film properties at different surface orientations (planar and vertical) of the 3D substrate. For TiO and HfO at low temperature, biasing led to phase selective growth of crystalline material on planar surfaces and amorphous material on vertical sidewalls of the trenches. For SiN, biasing led to microstructure selective degradation of film regions growing at planar surfaces of the trenches but not those growing at vertical sidewalls. These results demonstrate how substrate biasing during PEALD can enable topographically selective growth control on 3D substrates. The insights obtained in this work reveal numerous opportunities afforded by this technique for advancing the practical applications of atomic scale processing. The effects on other material systems, material properties and processing conditions will be reported in future investigations.