Atomic Layer Deposition of Aluminum Sulfide: Growth Mechanism and ...

April 16, 2018 | Author: Anonymous | Category: Documents
Share Embed

Short Description

Dec 15, 2017 - This study describes the synthesis of aluminum sulfide (AlSx) thin films ... Electrochemical performance ...



Cite This: Chem. Mater. XXXX, XXX, XXX-XXX

Atomic Layer Deposition of Aluminum Sulfide: Growth Mechanism and Electrochemical Evaluation in Lithium-Ion Batteries Xiangbo Meng,† Yanqiang Cao,‡ Joseph A. Libera,‡ and Jeffrey W. Elam*,‡ †

Department of Mechanical Engineering, University of Arkansas, Fayetteville, Arkansas 72701, United States Energy Systems, Argonne National Laboratory, 9700 South Cass Avenue, Argonne, Illinois 60439, United States

S Supporting Information *

ABSTRACT: This study describes the synthesis of aluminum sulfide (AlSx) thin films by atomic layer deposition (ALD) using tris(dimethylamido)aluminum and hydrogen sulfide. We employed a suite of in situ measurement techniques to explore the ALD AlSx growth mechanism, including quartz crystal microbalance, quadrupole mass spectrometry, and Fourier transform infrared spectroscopy. A variety of ex situ characterization techniques were used to determine the growth characteristics, morphology, elemental composition, and crystallinity of the resultant AlSx films. This study revealed that the AlSx growth was self-limiting in the temperature range 100−250 °C, and the growth per cycle decreased linearly with increasing temperature from ∼0.45 Å/cycle at 100 °C to ∼0.1 Å/cycle at 250 °C. The AlSx films were amorphous in this temperature range. We conducted electrochemical testing to evaluate the ALD AlSx as a potential anode material for lithium-ion batteries (LIBs). The ALD AlSx exhibited reliable cyclability over 60 discharge−charge cycles with a sustainable discharge capacity of 640 mAh/g at a current density of 100 mA/g in the voltage window of 0.6−3.5 V. for ZnS,34,35 CdS,36 CaS,37 BaS,38 SrS,38 Cu2S,8,39 In2S3,40 WS2,41,42 TiS2,42 PbS,43 SnS,44 GaSx,5,6 Li2S,45 Co9S8,46 NiSx,47 MnS,9 MoS2,48 Al2S3,17 and their ternary derivatives.7,49,50 Among metal sulfides, Al2S3 has been investigated electrochemically and showed great potential as an anode for lithiumion batteries (LIBs),51 boasting a theoretical capacity above 1400 mAh/g (∼4 times of the capacity of graphite anodes in commercial LIBs).51 A previous study of Al2S3 ALD using trimethyl aluminum (TMA) and hydrogen sulfide (H2S)17 reported that exceedingly long H2S purge times of 110 s were required to achieve a growth rate of 1.2 Å/cycle at 150 °C. The Al2S3 growth rate dropped monotonically with decreasing H2S purge time and reached 0.35 Å/cycle at 40s. Typical ALD purge times are in the range of 1−10 s. These very long purge times of 40−110 s for Al2S3 ALD using TMA and H2S are impractical, especially for applications requiring thicker films. This motivated us to evaluate a new process for aluminum sulfide ALD using tris(dimethylamido)aluminum (TDMA−Al) and H2S as precursors. By performing in situ quartz crystal microbalance (QCM), quadrupole mass spectrometry (QMS), and Fourier transform infrared spectroscopy (FTIR) measurements, we explored the growth mechanism and surface chemistry for the

1. INTRODUCTION Metal sulfides are versatile materials and can exhibit exceptional electrical, optical, magnetic, and mechanical properties.1 As a consequence, metal sulfides have been incorporated into a broad range of devices, including heterogeneous catalysts,2,3 energy conversion and energy storage devices,1,4−11 transistors,12 photodetectors,12 and gas sensors.13 Many of these applications require thin film coatings, and numerous coating techniques have been developed for depositing metal sulfide thin films such as chemical vapor deposition (CVD),13 physical vapor deposition (PVD),1,14−16 solution-based methods,1 and atomic layer deposition (ALD).17 Among these techniques, ALD has gained increasing interest recently as a means to deposit metal sulfide functional coatings due to its unique advantages in controlling materials growth and fabricating nanostructures.18,19 Utilizing alternating and self-limiting chemical reactions between gaseous precursors and a solid surface, ALD enables materials synthesis in an atomic layer-bylayer fashion. The self-limiting nature of ALD provides atomiclevel precision over the film thickness and composition and yields exceptionally uniform films over large areas and conformal coatings in complex geometries.19−22 These benefits, as a result, have encouraged researchers to employ ALD coatings for energy devices,10,11,22−28 catalysis,29 medical and biological devices,30,31 plasmonic devices,31 nano- and microelectromechanical systems,32 and novel nanostructured materials.33 To date, metal sulfide ALD processes have been reported © XXXX American Chemical Society

Received: May 26, 2017 Revised: September 25, 2017 Published: October 1, 2017 A

DOI: 10.1021/acs.chemmater.7b02175 Chem. Mater. XXXX, XXX, XXX−XXX


Chemistry of Materials

For the ex situ analysis, ALD AlSx films were deposited on Si wafers and micromachined Si trenches and characterized using a variety of techniques. Aluminum sulfide is extremely sensitive to atmospheric moisture and must be kept in an inert environment at all times. To facilitate air-free transfer of the AlSx films, an Ar-purged glovebag was installed over the exit flange on the ALD reactor. After removal of the samples from the ALD reactor and into the glovebag, the samples were loaded into Ar-filled, sealed glass jars for transport. Spectroscopic ellipsometry (SE; α-SE, J. A. Woollam Co.) was employed, and the ellipsometric data were fitted using a Cauchy model to determine the film thickness and refractive index. The ellipsometer was encased in an Ar-filled glovebag to prevent atmospheric exposure during these measurements. The film morphology was examined using field emission scanning electron microscopy (FE-SEM, Hitachi S4700) equipped with energy dispersive X-ray spectroscopy (EDX). The crystallinity of the as-deposited and annealed AlSx films were determined by X-ray diffraction (XRD, D8 Advance, Bruker). The film composition was measured by a Rutherford backscattering spectrometry (RBS: RBS400, CEA) and X-ray photoelectron spectroscopy (XPS: PHI Quantum 2000). The RBS and XPS measurements were conducted by Evans Analytical Group Laboratories (Sunnyvale, CA). The RBS measurements were performed using a 2.275 MeV He2+ ion beam with backscattering angle of 160° and a grazing angle of 100°. The XPS was equipped with a monochromatic Al−Kα (1486.6 eV) X-ray source, and depth profiling measurements were performed using Ar+ sputtering. The sputter rate was calibrated using SiO2, accounting for a SiO2-equivalent rate of 7.81 Å/min. The analysis area was 1400 × 300 μm. The samples prepared for SEM, XPS, and RBS analysis were capped with ALD ZnS films to protect the ALD AlSx films against air exposure, as successfully practiced in our previous study.45 The ALD ZnS coatings were deposited using alternating exposures to diethyl zinc (DEZ, Aldrich) and H2S with the timing sequence 1−5−1−5 s. To explore the potential for the ALD AlSx to serve as a LIB anode, we deposited ALD AlSx films on nitrogen-doped graphene nanosheet (NGNS) powders (ACS Material, United States) at 100 °C, and the resultant 3D NGNS−AlSx nanocomposites were evaluated electrochemically. The NGNS powders have a surface area of 500−700 m2/g (determined by Brunauer−Emmett−Teller (BET) surface area analysis), consist of typically 1−5 atomic layer graphene nanosheets, and contain 1.0−3.0 atom % nitrogen and 7.0−7.5 atom % oxygen. After 50 cycles of ALD AlSx using the timing sequence 60−60−120− 60 on 59 mg NGNS powder, the resultant NGNS−AlSx sample weighed 195 mg so that the AlSx loading was ∼70 wt %. The NGNS− AlSx was further mixed with Super P carbon black and polyvinylidene fluoride (PVDF, Sigma-Aldrich) using a ratio of 8:1:1 and dissolved in an N-methyl-2-pyrrolidone (NMP, Sigma-Aldrich) solvent for fully homogeneous mixing. The slurry was then cast onto a Cu foil and fashioned into a laminate using a 50 μm doctor blade. The laminate was dried in a furnace at 80 °C within an Ar-filled glovebox for 24 h. Next, the dried laminate was punched into 7/16 in. circular electrodes and subsequently assembled into CR2032 LIB coin cells in an Ar-filled glovebox with H2O and O2 levels below 1 ppm. Li metal was used as the counter/reference electrode; a Celgard 2325 membrane was used as the separator, and 1 M lithium bis(trifluoromethanesulfonyl)imide (LITFSI, Sigma-Aldrich) in 1,3-dioxolane (DOL, Sigma-Aldrich) and 1,2-dimethoxyethane (DME, Sigma-Aldrich) (DOL:DME = 1:1 by volume) was used as the electrolyte. The electrochemical discharge− charge testing was performed on an Arbin 4300 battery tester using either a voltage window of 0.01−3.0 V or a voltage window of 0.6−3.5 V for the NGNS−AlSx electrodes. In addition, control samples were assembled and tested following the same procedures described above, in which the ratio is 8:1:1 for active materials, Super P, and PVDF. The active materials in the control samples consisted of 70 wt % microsized Al2S3 powder (Sigma-Aldrich) and 30 wt % NGNS and were mixed manually. All of the electrochemical testing was performed at room temperature.

AlSx ALD. Next, we employed a series of ex situ characterization tools to analyze the ALD AlSx films to determine their thickness, composition, and microstructure. After establishing this new ALD process, we proceeded to functionalize nitrogendoped graphene nanosheets (NGNS) with ALD AlSx to produce three-dimensional (3D) NGNS−AlSx nanostructured composites. Finally, we evaluated the electrochemical performance of the 3D NGNS−AlSx nanocomposites as an LIB anode material and measured the sustainable charge storage capacity and discharge−charge cyclability.

2. EXPERIMENTAL SECTION The AlSx ALD was performed in a custom viscous flow, hot-walled ALD reactor52 comprised of a 1 m stainless steel tube with an inner diameter of 5 cm. Before each deposition, substrates (i.e., Si(100) wafers and micromachined Si trenches) were first loaded in the reactor. During the ALD, a constant 300 sccm flow of ultrahigh purity Ar (UHP, 99.999%) passed through the flow tube at a pressure of ∼1 Torr. The ALD AlSx was performed using alternating exposures to TDMA−Al (Sigma-Aldrich, United States) and Ar-balanced 1% H2S (Matheson Trigas, United States) with Ar purge periods between each exposure. TDMA−Al normally exists as a dimer, Al2(NMe2)6, but for simplicity, we describe TDMA−Al in monomeric form, Al(NMe2)3. To provide sufficient vapor pressure, the solid Al(NMe2)3 was heated to 80 °C in a stainless steel reservoir, and 50 sccm UHP Ar was diverted through this reservoir during the Al(NMe2)3 exposures. This yielded a partial pressure of ∼0.01 Torr Al(NMe2)3 in the flow tube. The 1% H2S was stored in a pressure-regulated lecture bottle. A series of needle valves was used to deliver 1% H2S pressure pulses of 0.2 Torr during the H2S exposures. The AlSx ALD timing can be described as t1−t2−t3−t4, with t1 and t3 being the exposure times for the Al(NMe 2)3 and H2S, respectively, and t2 and t4 being the corresponding purge times, with all times in seconds (s). The AlSx ALD was systematically investigated using in situ QCM measurements. The QCM studies were conducted using a modified Maxtek Model BSH-150 sensor head and RC quartz crystal sensor (CNT06RCIA, Colnatec). The RC quartz sensors are less sensitive to temperature fluctuations compared to conventional AT quartz sensors, thereby reducing the effects of temperature-induced transients and temperature drift on the QCM measurements. The crystals were sealed within the sensor head using an electrically conducting high temperature epoxy (P1011, Epotek), and the sensor head was modified to provide backside purging of the crystal to prevent deposition on the back.6,45,53 Additional in situ studies were conducted using in situ quadrupole mass spectrometry (QMS) and Fourier transform infrared (FTIR) absorption spectroscopy measurements to explore the surface chemical reactions responsible for the AlSx ALD. The QMS (Stanford Research Systems, Model RGA300) was located downstream from the sample position in a differentially pumped chamber separated from the reactor by a 35 μm orifice. The FTIR (Nicolet 6700 FTIR spectrometer, Thermo Scientific) was operated in transmission mode in a separate ALD reactor as described previously.54 The FTIR reactor utilized gate valves that were closed during the precursor exposures to prevent growth on the IR-transparent KBr windows. Substrates for FTIR measurements were prepared by pressing ZrO2 nanopowder (Aldrich, 25 m2/g) into a stainless steel grid.55,56 The grids were fabricated using photochemical machining (Fotofab, Inc.) and were 50 μm thick with 50 μm bars and 200 μm square openings. ZrO2 is relatively transparent from 4000 to 800 cm−1, the frequency range of interest for identifying surface functional groups, and provides a high surface area to increase the IR absorption signals. The nanopowderfilled grid was mounted onto a stage that could be heated to 500 °C. This stage was then loaded into the FTIR reactor so that the IR beam passed through the center of the grid. During the in situ FTIR measurements, the substrate temperature was maintained at either 100 or 200 °C by the heated stage, and accordingly, the reactor walls also were maintained at 100 or 200 °C to prevent precursor condensation. B

DOI: 10.1021/acs.chemmater.7b02175 Chem. Mater. XXXX, XXX, XXX−XXX


Chemistry of Materials

3. RESULTS AND DISCUSSION 3.1. In Situ QCM Measurements during AlSx ALD. QCM is a powerful tool for evaluating new ALD processes6,45 and can easily detect the mass changes associated with submonolayer quantities of material added or removed from the surface. Figure 1a shows the in situ QCM measurements of

| − (SH)y + Al(N(CH3)2 )3 (g) → | − Sy − Al(N(CH3)2 )(3 − y) + y HN(CH3)2 (g)


| − Sy − Al(N(CH3)2 )(3 − y) + 1.5H 2S(g) → | − (AlS1.5) − (SH)y + (3 − y)HN(CH3)2 (g)


in which “|−” represents the surface and “g” indicates the gaseous phase. The Al compound reacts with y thiol (i.e., −SH) groups and releases y of the three dimethylamido- ligands as dimethylamine (HN(CH3) 2) vapor (eq 1). Next, H2S eliminates the remaining (3−y) dimethylamido- ligands as HN(CH3)2, forming stoichiometric Al2S3 and regenerating the surface thiols (eq 2). This mechanism makes several assumptions: (1) thiol groups are the reactive species for Al(NMe2)3 chemisorption; (2) the only gaseous product is HN(CH3)2; and (3) the stoichiometry is Al2S3. We evaluate these assumptions below. Given eqs 1, 2, and the atomic masses, we define the QCM mass ratio as R=

75 Δm = m1 159 − 45y


The QCM data show R = 0.57 (Figure 1b). Consequently, y = 0.85, meaning that ∼72% of the dimethylamido- ligands remain on the AlSx surface following the Al precursor reaction (eq 1) at 200 °C. By repeating these measurements over the temperature range 100−225 °C (Figure SI-2a), we discovered that nearly 100% of the dimethylamido- ligands remain on the AlSx surface following the Al precursor reaction at 100 and 125 °C. At temperatures ≥150 °C, ∼70% of the dimethylamidoligands remain after dosing the Al(NMe2)3. Figure SI-2a shows in situ QCM measurements performed over the temperature range 100−225 °C to explore the effect of temperature on the AlSx ALD. Assuming a density of 2.32 g/ cm3 for bulk, crystalline Al2S3, we calculated the GPC for ALD AlSx using the QCM data in Figure SI-2a, as indicated by the red symbols in Figure 2. To confirm the QCM data, we prepared ALD AlSx films on Si(100) substrates using 50−600 ALD cycles at temperatures of 100, 150, 200, and 250 °C and measured the film thickness using ex situ spectroscopic ellipsometry (Figure SI-2b). On the basis of these data, we calculated the GPC at each growth temperature, and the results are given by the blue symbols in Figure 2. Both the data from

Figure 1. In situ QCM measurements of ALD AlSx at 200 °C using the timing sequence 5−5−10−5 s. (a) Mass of AlSx film versus time during 60 ALD cycles beginning on ALD Al2O3 surface. (b) Enlarged view of four consecutive ALD AlSx cycles in the regime of constant growth per cycle. Lower traces indicate precursor pulsing during the alternating Al(NMe2)3 (red) and H2S (black) exposures, and the m1, m2, and Δm are described in the text.

mass versus time recorded during 60 AlSx ALD cycles with the timing sequence 5−5−10−5 s at 200 °C. This timing sequence was established using in situ QCM (see Figures SI-1 in Supporting Information) and ensures saturating Al(NMe2)3 and H2S exposures. To establish a uniform starting surface for the AlSx ALD, we deposited an ALD Al2O3 film on the QCM surface using alternating trimethylaluminum (TMA) and H2O exposures with the timing sequence 1−5−1−5 s. Figure 1a reveals that the AlSx ALD growth per cycle (GPC) was larger initially during the first ∼20 ALD AlSx cycles (0−500 s, highlighted orange area) on the Al2O3 surface before gradually decreasing and stabilizing to a constant growth per cycle value. The QCM data show a linear increase in mass versus time at ∼2.9 ng·cm−2·cycle−1 after ∼500 s or ∼20 ALD cycles. An enlarged view of four consecutive AlSx ALD cycles in the stable growth regime is shown in Figure 1b. The QCM mass is given by the blue trace in Figure 1b, while the periods of Al(NMe2)3 and H2S dosing are indicated by the black and blue traces, respectively. The Al(NMe2)3 exposures generate a mass increase m1 = ∼4.7 ng·cm−2·cycle−1 and the H2S exposures decrease the mass by m2 = ∼1.8 ng·cm−2·cycle−1 so that the average net mass change Δm = ∼2.9 ng·cm−2·cycle−1. We can propose a mechanism for AlSx ALD based on these mass changes:

Figure 2. ALD AlSx growth per cycle versus deposition temperature as measured by ex situ spectroscopic ellipsometry (solid symbols) and in situ QCM (open symbols). C

DOI: 10.1021/acs.chemmater.7b02175 Chem. Mater. XXXX, XXX, XXX−XXX


Chemistry of Materials

the spectrum recorded during the previous precursor exposure. As a result, positive absorbance features indicate formation of new surface species, and negative features result from the removal of surface species. Positive features in the ranges of 2781−2929 and 889−1452 cm−1 are seen after the first Al(NMe2)3 exposure. The 2781−2929 cm−1 features are antisymmetric and symmetric C−H stretching modes.57−60 The C−H stretching mode at 2781 cm−1 is a Bolhmann band and results from interaction between the N lone pair orbital and the C−H σ-orbital in N(CH3)2.59−61 This spectral feature indicates that some of the N(CH3)2 on the substrate surface are intact. We also see a negative absorbance at 3732 cm−1 that we assign to the consumption of Al2O3 hydroxyl groups.59 Features in the range of 1106−1452 cm−1 are mainly from CH3 deformation and rocking modes.57−60 In comparison to our previous study for ALD GaSx using Ga2(NMe2)6 and H2S,6 the AlSx ALD did not show any feature at 1575 cm−1 that would indicate NC double bonds from β-hydride elimination in the dimethylamido- ligand59 (i.e., |−N(CH3)2 → |−N = CH2 + CH4 (g)). β-hydride elimination between neighboring methyls to release methane has been reported previously at temperatures above 200 °C.59,57,62,63 We attribute the 889 to 1029 cm−1 features to stretching modes of NC2.57 The first H2S exposure produces negative absorbance features corresponding to the near-complete removal of CH3 and NC2. The difference spectra following consecutive Al(NMe2)3 and H2S exposures are symmetric, indicating that ligand consumption and creation are comparable as expected for the proposed AlSx ALD mechanism (eqs 1 and 2). We also observed small changes in a feature at ∼2400 cm−1, that we assign to S−H stretching.6,64 This S−H feature increases after H2S exposures and decreases after Al(NMe2)3 exposures (Figure 3a, inset). We integrated the NC2 and S−H peaks after each precursor exposure, and these normalized values are shown in Figure 3b. Figure 3b reveals that these spectral changes evolve over the first 5−7 AlSx ALD cycles but then stabilize. We interpret this evolution as the gradual nucleation of the AlSx film on the ALD Al2O3 surface, and this evolution stops once a continuous AlSx film has formed. A similar nucleation behavior was observed during the QCM measurements of AlSx ALD on Al2O3 (Figure 1). In situ FTIR measurements were also conducted at 100 °C (see Figure SI-4 in Supporting Information) and showed a similar ligand exchange as illustrated in Figure 3a. We performed in situ QMS measurements to identify and quantify the gas phase products of the AlSx ALD surface reactions. First, we surveyed masses between m/z = 2−90 to identify the reaction products and found that dimethylamine (DMA, m/z = 45) was the only gaseous product. Figure 4 shows the DMA signals (blue trace) versus time recorded during AlSx ALD at 200 °C using the sequence: 5-s Al(NMe2)3 dose, 50-s purge repeated 4 times, and then 10-s H2S dose, 50-s purge repeated 4 times. For the QMS signals to stabilize after each exposure, we used longer purge times compared to the QCM measurements. Each precursor was dosed four times in succession so that the following three doses would reveal any possible background signals. These background signals are likely cracks of the parent compound formed by electron impact inside the QMS. In Figure 4, it is clear that the m/z = 45 intensity is lowest during the first Al(NMe2)3 dose, followed by a slight increase for doses two and three, and remaining constant during dose four. This pattern is quite reproducible between the three Al(NMe2)3 dosing sequences in Figure 4.

QCM and ellipsometry in Figure 2 indicate that the ALD AlSx GPC decreases with temperature. This trend is similar to our previous study of GaSx ALD using Ga2(NMe2)6 and H2S.6 Comparing the two sets of GPC data in Figure 2, it is evident that the ellipsometry values are somewhat higher than the QCM values. This suggests that the density of the ALD AlSx films is lower than the assumed, bulk value. Indeed, XRD measurements confirmed that the ALD AlSx films are amorphous (Figure SI-3), and XPS elemental analysis detected residual C and N in the films (Figure 6), both of which will reduce the density of the ALD AlSx films. 3.2. In Situ FTIR and QMS Measurements during AlSx ALD. Next, we conducted in situ measurements using FTIR and QMS to explore further the mechanism of AlSx ALD. For the FTIR studies, we loaded a ZrO2 nanopowder-filled grid into the ALD FTIR chamber and heated the ZrO2 nanopowder to 400 °C for 10 min. We then exposed the powder to oxygen (30 pulses with 350 sccm flow, 10 s each, 30 s purge between pulses). This treatment was performed to lower the FTIR background by oxidizing surface hydrocarbon contaminants. After cooling the powder to 200 or 100 °C, we performed 6 TMA/H2O cycles for Al2O3 ALD (30−60−30−60 s timing sequence) to establish a well-defined starting surface. The FTIR studies used 60 s exposures, 60 s purges for the AlSx ALD, and 100% H2S. Twenty-four AlSx ALD cycles were performed, and FTIR spectra were recorded after every precursor exposure. These ALD Al2O3 and AlSx timing sequences were verified to saturate the nanopowder surfaces by extending the exposure times and observing no further changes in the FTIR signals. Figure 3a shows FTIR difference spectra after each precursor exposure for the 1st, 2nd, 3rd, 14th, and 18th AlSx ALD cycles at 200 °C. To generate these difference spectra, we subtracted

Figure 3. (a) In situ FTIR difference spectra recorded after individual Al(NMe2)3 and H2S exposures in the 1st, 2nd, 3rd, 14th, and 18th ALD cycles on an ALD Al2O3 surface at 200 °C. Right-hand spectra show expanded views of the S−H stretching region. (b) Integrated IR absorbance of the NC2 region (red) and S−H region (blue) after each Al(NMe2)3 and H2S exposure. D

DOI: 10.1021/acs.chemmater.7b02175 Chem. Mater. XXXX, XXX, XXX−XXX


Chemistry of Materials

Figure 4. In situ QMS measurements at m/e = 45 during AlSx ALD at 200 °C where the Al(NMe2)3 H2S exposures are separated into 4 separate microdoses to evaluate the degree of saturation of the surface reactions. Lower traces indicate precursor pulsing during the Al(NMe2)3 (red) and H2S (black) exposures as indicated.

We interpret the m/z = 45 signal during the third and fourth doses to be the background signal from Al(NMe2)3 and that the lower initial signal results from the depletion of Al(NMe2)3. This interpretation implies that little or no DMA product is generated (i.e., y ∼ 0 in eq 1). In contrast, the 4 successive H2S doses produce a monotonically decreasing m/z = 45 signal. It is possible that the H2S reaction is not completely saturated in Figure 4, and that additional H2S doses would reduce the m/z = 45 signal to zero. Our interpretation is that all of the m/z = 45 signal during the H2S doses results from DMA formed by the ALD surface reactions. The fact that DMA persists even after four H2S doses suggests that this reaction is relatively slow. The QMS measurements were repeated at a lower growth temperature of 100 °C (Figure SI-5), and the results were very similar to Figure 4 with the exception that the DMA peak decreases more slowly during the H2S dosing, suggesting that the reaction is slower at 100 °C. We assumed pure Al2S3 as the solid phase reaction product in eqs 1 and 2. On the basis of the QMS data discussed above, however, impurities may result if the surface reactions fail to achieve completion due to slow kinetics and incomplete saturation. Some evidence for this is seen in the net FTIR spectra recorded after each complete ALD cycle (Figure SI-6). This figure shows a gradual increase in the NC2 stretch, C−H bend, and C−H stretching regions with increasing ALD cycles consistent with the accumulation of DMA ligands in the film. Figure SI-7 shows the integrated absorbance of the 1450 cm−1 C−H bending mode versus ALD cycles at 100 °C (black symbols) and 200 °C (red symbols). It is clear that the rate of impurity accumulation is ∼2× lower at 200 °C. 3.3. Composition and Morphology of ALD AlSx Films. We utilized RBS and XPS to analyze the composition of ALD AlSx films deposited at 100 and 200 °C. Due to the air-sensitive nature of Al2S3, we applied an ALD ZnS capping layer over the AlSx. ZnS was demonstrated in our previous study45 to protect Li2S from air exposure. The sample deposited at 100 °C consisted of a Si(100) substrate coated with 1250-cycle AlSx and 1000-cycle ZnS, and the sample deposited at 200 °C was a Si(100) substrate coated with 1000-cycle AlSx and 400-cycle ZnS. The two samples were stored under Ar prior to RBS characterization to minimize S exchange with atmospheric moisture. Figure 5 shows the RBS results of the two samples,

Figure 5. RBS measurements of atomic concentration versus depth for films prepared at (a) 100 °C using 1250 ALD AlSx cycles on Si(100) and then capped with 1000 ALD ZnS cycles and (b) 200 °C using 1000 ALD AlSx cycles on Si(100) and then capped with 400 ALD ZnS cycles.

and more detailed results are described in Table 1. It is clear from Figure 5 that both samples comprise distinct ZnS and AlSx layers. Figure 5b reveals that the AlSx deposited at 200 °C is essentially pure and yields an atomic ratio of S/Al ∼ 1.5. In contrast, the AlSx deposited at 100 °C (Figure 5a) shows very high C and N signals and a ratio of S/Al ∼ 1.0. These findings are consistent with the FTIR measurements that showed a much larger accumulation of residual ligands at 100 °C (see Figure SI-6 and SI-7). We also applied XPS depth profiling analysis to the same two samples, and the results are shown in Figure 6. Consistent with the RBS and FTIR analyses, there are many more C and N impurities in the AlSx layer deposited at 100 °C compared to those deposited at 200 °C. On the basis of XPS analysis, the ratio of S/Zn is 0.68 at 100 °C and 0.85 at 200 °C, while the ratio of S/Al is 0.81 at 100 °C and 1.28 at 200 °C. The elemental ratios from XPS are not the same as those from RBS analysis, and this is probably due to the different elementspecific sputtering rates for components in these films. The ratio of C/N is ∼3 in these films, which is higher than the value of 2 expected for residual DME ligands, but this discrepancy might result from differential sputtering. To gauge the importance of the Ar storage used for these samples, we exposed one of the 100 °C films to air and performed XPS depth profiling. As shown in Figure SI-8, the AlSx layer converted almost entirely to Al2O3 because of the air exposure, while the ZnS layer is unchanged. E

DOI: 10.1021/acs.chemmater.7b02175 Chem. Mater. XXXX, XXX, XXX−XXX


Chemistry of Materials Table 1. Composition of ALD AlSx and ZnS Films Determined Using RBS Measurements atomic concentration (atom %) samplesa 100 °C

200 °C


layers layer layer bulk layer layer bulk

1 2 1 2

Zn ZnS AlSx Si ZnS AlSx Si

48.6 ± 1



51.4 ± 1 23.2 ± 1

22.4 ± 1



22.3 ± 5

32.1 ± 5


100 ± 1 49.1 ± 0.5

50.9 ± 1 60 ± 1

40 ± 1 100 ± 1

atomic ratio ZnS1.06 AlS1.04 Si ZnS1.04 Al2S3 Si

Si(100) coated with 1250-cycle AlSx and 1000-cycle ZnS at 100 °C; Si(100) coated with 1000-cycle AlSx and 400-cycle ZnS at 200 °C.

Figure 7. SEM images of ALD AlSx film deposited in high aspect ratio Si trenched substrate using 1250 ALD AlSx cycles and then capped with 1000 ALD ZnS cycles at 100 °C: (a) lower magnification image showing entire trenches covered by the double layers of AlSx and ZnS and higher magnification images showing (b) the top and (c) bottom of the cross sections of the coated trenches.

∼9.4. Figures 7b and c show higher resolution cross-sectional images of the coated Si trenches at the top and bottom. The AlSx and ZnS coatings are easily identified in these images and are uniform and conformal along the trenched Si substrate. It is also disclosed that the GPC for AlSx is ∼0.4 Å/cycle, which is consistent with ellipsometry (Figure 2), and for ZnS the GPC is 1.3 Å/cycle, very close to the value of 1.37 Å/cycle reported previously.68 On the basis of all the information discussed above, the films deposited at 200 °C are composed primarily of Al2S3 with some DMA contamination. At lower growth temperatures (e.g., 100 °C), the surface reactions are sufficiently slow that they do not proceed to completion so that a significant fraction of the dimethylamido- ligands remain in the film where they substitute for S. Consequently, the S/Al ratio is less than 1.5. On the basis of Figures 4 and SI-5, it is likely that longer H2S exposures would reduce the DMA content of the films. 3.4. Electrochemical Application as LIB Anodes. Al2S3 potentially is a very promising anode material for LIBs, but to date, only one study using microsized commercial Al2S3 has been published.51 The researchers showed that the microsized Al2S3 had severe capacity drop in 10 discharge−charge cycles and could not realize long-term cyclability. Using XRD and X-

Figure 6. Atomic concentration versus depth measured by XPS depth profiling analysis for films prepared at (a) 100 °C using 1250 ALD AlSx cycles on Si(100) and then capped with 1000 ALD ZnS cycles and (b) 200 °C using 1000 ALD AlSx cycles on Si(100) and then capped with 400 ALD ZnS cycles.

High-resolution analysis of the Al XPS signals recorded from the AlSx layers helped to determine their chemical states (Figure SI-9). In the case of AlSx deposited at 200 °C, the Al 2p peak is narrow and centered at a binding energy of 74.2 eV, as expected for Al2S3.65 In comparison, the AlSx deposited 100 °C shows a peak at 73 eV and a shoulder at 73.8 eV. The higher binding energy peak is likely Al2S3, while the lower energy peak is consistent with Al−N66 and likely results from residual DMA ligands in the AlSx film. Figure SI-9 also shows that, after oxidation, the Al 2p shifts to a higher binding energy of 76.5 eV, consistent with ALD Al2O3.67 To examine the morphology and conformality of the ALD AlSx films, we deposited a layer of ALD AlSx using 1250 cycles onto trenched Si substrates, followed by a capping layer of ALD ZnS of 1000 cycles at 100 °C (see Figure 7). Figure 7a shows a cross-sectional view of the Si trenches, which are ∼16 μm tall and separated by gaps of ∼1.7 μm, yielding an aspect ratio of F

DOI: 10.1021/acs.chemmater.7b02175 Chem. Mater. XXXX, XXX, XXX−XXX


Chemistry of Materials

consistent with a GPC of 1.4 Å/cycle (see Figure 8b). Evidently, the growth of ALD AlSx on NGNS is much higher than that on Si wafers. This may result from a higher density of surface functional groups on NGNS or from CVD. Figure 8b also shows that the ALD coating is uniform all over the NGNS powder. This is further verified by EDX elemental mapping of carbon, aluminum, and sulfur shown in Figure 8c. The resulting NGNS−AlSx nanocomposites have a loading of ∼70 wt % AlSx. The NGNS−AlSx nanocomposites were synthesized to address the issues encountered in microsized Al2S3 in the following ways: (1) ∼7 nm thick nanofilms of AlSx greatly shorten the diffusion pathways for lithium ions and electrons, (2) the NGNS substrates dramatically improve electrical conductivity, and (3) the porous NGNS provides a flexible scaffold to accommodate volume changes during cycling. To verify the utility of our NGNS−AlSx nanocomposites, we compared the NGNS−AlSx performance with that of a mixture of 30 wt % NGNS and 70 wt % microsized Al2S3 during electrochemical cycling in the voltage window 0.01−3.0 V. In Figure 9a, the discharge capacity of commercial microsized Al2S3 continuously decreased from ∼1200 mAh/g at the first cycle to 150 mAh/g at the 30th cycle. This is consistent with the study by Senoh et al.51 In contrast, the nanostructured NGNS−AlSx exhibited a very impressive performance in the first 6 cycles: starting from a capacity of ∼1000 mAh/g at the first cycle, then improved to ∼1100 mAh/g, and remained until the sixth cycle. Beginning with the seventh cycle, however, the NGNS−AlSx dropped in capacity continuously and remained only ∼200 mAh/g at the 30th cycle. The Coulombic efficiencies (CE) of the nanostructured ALD AlSx and the commercial microsized Al2S3 all are lower than 100%, especially in the first few cycles where CE < 70% for the ALD AlSx and CE < 90% for the microsized Al2S3. These values indicate severe electrochemical irreversibility. Figure 9b illustrates the evolution of the electrochemical irreversibility of the NGNS− AlSx nanocomposites with its discharge−charge profiles. Obviously, the discharge capacities are much higher than the charge capacities in all the cycles. In other words, after the cells were discharged, they could not be fully recharged. To examine the underlying mechanisms, we analyzed the dQ/dV profiles of the NGNS−AlSx. Figure SI-10a shows that there is a reduction peak at around 0.6 V ascribed to the formation of solid electrolyte interphase (SEI) during the first discharge and a strong reduction peak below 0.5 V related to lithium intercalation into NGNS and the formation of LiAl. In particular, the peaks at
View more...


Copyright © 2017 DOCIT Inc.