Bulk-Nanoporous-Silicon Negative Electrode with Extremely High

Jul 2, 2014 - Synthesis of Ultrathin Si Nanosheets from Natural Clays for Lithium-Ion Battery Anodes. Jaegeon Ryu , Dongki Hong , Sinho Choi , and Soo...
0 downloads 0 Views 5MB Size
Download PDF
Letter pubs.acs.org/NanoLett

Bulk-Nanoporous-Silicon Negative Electrode with Extremely High Cyclability for Lithium-Ion Batteries Prepared Using a Top-Down Process Takeshi Wada,*,† Tetsu Ichitsubo,‡ Kunio Yubuta,† Haruhiko Segawa,§ Hirokazu Yoshida,∥ and Hidemi Kato† †

Institute for Materials Research, Tohoku University, Sendai, Miyagi 980-8577, Japan Department of Materials Science and Engineering, Kyoto University, Kyoto 606-8501, Japan § Yokohama R&D Laboratories, Furukawa Electric Co., Ltd., Yokohama, Kanagawa 220-0073, Japan ∥ Metal Research Center, Furukawa Electric Co., Ltd., Nikko, Tochigi 321-1493, Japan ‡

S Supporting Information *

ABSTRACT: We synthesized freestanding bulk three-dimensional nanoporous Si using dealloying in a metallic melt, a topdown process. Using this nanoporous Si, we fabricated negative electrodes with high lithium capacity, nearing their theoretical limits, and greatly extended cycle lifetimes, considerably improving the battery performance compared with those using electrodes made from silicon nanoparticles. By operating the electrodes below the accommodation volume limit of their pores, we prolonged their cycle lifetime. KEYWORDS: lithium-ion battery, silicon, negative electrode, nanoporous structure, dealloying template or depositing a thin film, making the preparation expensive, slow, and difficult to scale up. For the practical application in LIBs, the preparation of such a nanoporous Si by a simple, low cost, and scalable approach is necessary. Thakur et al. prepared two-dimensional porous Si particulates with directional pores by electrochemically etching Si wafers in HFbased solutions and then sonicating them; the resulting spongelike Si anode showed excellent retention capacity of 1000 mAh/ g for over 600 cycles.12 Wu et al. fabricated a three-dimensional network of Si nanoparticles coated with conducting polymer hydrogel, which exhibited an extremely long cycle lifetime of 5000 cycles at 550 mAh/g.13 Yi et al. fabricated nanoscale interconnected Si by a disproportionation reaction using commercial SiO powder, which exhibited high retention capacity of 1459 mAh/g after 200 cycles.14 Recently, Dai et al. developed a bottom-up synthesis of mesoporous Si with very high surface area by reducing Si halogenide precursor with NaK alloy, which is expected to be applied in LIBs.15 In the present work, we designed a negative electrode using freestanding bulk 3DNP structure of Si (3DNP-Si) as the active material. In this network, bulk Si nanoligaments with widths of several hundred nanometers were directly interconnected, forming a 3D framework without need for a template, as

S

ilicon is a promising active material for the negative electrode of Li-ion batteries (LIBs) because it has a theoretical Li capacity of ∼4200 mAh/g, about 10 times that of carbon-based electrodes.1 However, Si electrodes suffer from volume changes of more than 400% with Li insertion and extraction.2 Figure 1a shows a schematic of an electrode made from large particles of Si (typically larger than several micrometers). The large volume change caused by Li insertion pulverizes the conductive network between the active material and the current collector, rapidly degrading the cyclic performance.2 The large elastic strain accompanying Li insertion also significantly slows the kinetics of lithium compound formation.3,4 Many researchers have studied ways to accommodate this volume change by using silicon nanocomposites,5 thin films,6 and nanostructured silicon, including nanotubes,7 nanowires,8 and nanoporous structures.9,10 Silicon-based negative electrodes with porous structures have exhibited improved cyclic performance because the many open channels in such structures act as ideal volume expansion buffers. Cho prepared three-dimensional nanoporous (3DNP) Si by depositing Si particles on SiO2 nanoporous templates; this structure exhibited high capacity of over 2800 mAh/g without significant degradation up to 100 cycles;11 after cycling, the electrode morphology remained unchanged, showing that 3DNP structures can effectively accommodate volume changes. Despite such promising performance from the porous Si-based electrodes, the preparation has generally required several complicated steps, such as fabricating a © 2014 American Chemical Society

Received: April 23, 2014 Revised: July 1, 2014 Published: July 2, 2014 4505

dx.doi.org/10.1021/nl501500g | Nano Lett. 2014, 14, 4505−4510

Nano Letters

Letter

Figure 1. Schematic of the electrode design showing a large Si particle and bulk 3DNP-Si. (a) Electrode made from large Si particle fractures at a particular lithiation capacity. (b) Electrode made from bulk 3DNP-Si has bulky, directly interconnected Si nanoligaments. The many nanopores surrounding the Si nanoligaments should accommodate volume changes up to its porosity, but lithiation exceeding its accommodation limit fractures the Si nanoligament.

operating capacity can be understood by assessing the relation between the porosity and volume changes from lithiation. Dealloying is a well-known approach for preparing 3DNP structures of metals.16 In dealloying, a component of an alloy precursor is selectively dissolved, typically in a corrosive aqueous solution, which causes a 3DNP structure of the remaining element(s) to spontaneously form. The main drawback of dealloying is the applicable elements: those sufficiently stable against oxidization, generally noble metals such as Au,17 Pt,18 Pd,19 and Cu.20 This limitation can be overcome by dealloying in a metallic melt, a method not based on corrosion in an aqueous solution but rather on metallurgical reaction between an alloy precursor solid and a metallic melt. Using this method, the 3DNP structure can form in less noble metals, such as Ti and Fe, without oxidation.21−23 As shown in these studies,21−23 dealloying in a metallic melt is a promising way to fabricate 3DNP structures of less noble materials, including Si. Another advantage of dealloying in a metallic melt is that the nanostructured materials can be prepared using a top-down approach, allowing great volumes of the nanostructure to be prepared from a large precursor ingot. The main step of this method is performed in a common melting furnace, making it simple and expandable to mass production. Figure 2a shows a schematic of producing bulk 3DNP-Si by dealloying in a metallic melt. According to binary phase diagrams,24 Mg−Bi is miscible, whereas Si−Bi is immiscible. On the basis of these atomic interactions, we used the Si−Mg− Bi system, with Si acting as the porous-structure-forming

shown in Figure 1b. This unique structure can increase the total gravimetric capacity more efficiently than can structures that use templates. This 3DNP structure is unlike others because of its continuous network structure of Si nanoligaments. Unlike etched Si with a cellular structure or an agglomerate of Si nanoparticles, this networked structure of Si should somewhat prevent fracture. Namely, this 3DNP-Si has two aspects: (1) aggregates of the nanowires, but such nanowires are tangled, and (2) it can be regarded as a bulk material because of continuous network of nanowires. We expect the many interconnected nanopores surrounding the Si nanoligaments will accommodate the volume change associated with Li insertion. Ideally, this porous structure can accommodate a volume change (ΔV = (VF − VI)/VI, where VI is initial volume and VF is final volume) up to its porosity (Vpore) without fracturing. Furthermore, we expect the large surface area from the continuously connected pores to allow efficient Li-ion/ electron exchange. The advantages of this active material should improve the capacity, operation rate, and cycle life of the lithium-ion battery built from it. Such a bulk 3DNP-Si-based electrode should be fabricated with a simple method that can be extended to industrial mass production. In the present work, we propose a novel, practical way to prepare bulk 3DNP-Si, capable of delivering excellent battery performance, by dealloying a metallic melt, a top-down process based on a simple metallurgical principle. We also confirmed that the dependence of 3DNP-Si electrode cycle lifetime on the 4506

dx.doi.org/10.1021/nl501500g | Nano Lett. 2014, 14, 4505−4510

Nano Letters

Letter

Figure 2. Preparation of bulk 3DNP-Si by dealloying in a metallic melt. (a) Schematic of producing 3DNP-Si by dealloying in a metallic melt. The Si−Mg precursor and Bi melt were selected based on the miscibility of Mg−Bi and the immiscibility of Si−Bi. (b) Schematic of the evolution of the 3DNP structure by dealloying. (i) Initially, the Mg2Si intermetallic precursor has an atomically homogeneous structure. (ii) As the precursor is immersed in the Bi melt, only the Mg atoms dissolve into the Bi melt, leaving Si atoms at the solid/liquid interface to cohere into Si islands. (iii) Bi is etched away to form the 3DNP-Si. The colors of atoms correspond to those in panel a. (c) SEM image and corresponding EDX mapping of Si−Mg precursor. Color of the mapping represents atomic concentration defined by the color bar in the right side of image. (d) SEM image and corresponding EDX maps of sample after immersion in Bi. (e) SEM image and corresponding EDX map of sample after etching. (f) Photograph of the bulk 3DNP-Si powder. (g) TEM image and corresponding selected area diffraction pattern of single grain from the powder.

distribution of Mg and Si with a ∼2:1 Mg/Si concentration ratio. The surrounding matrix was a Mg-enriched phase that contained almost no Si. The X-ray diffraction (XRD) pattern in Supporting Information Figure S1 confirms that the primary grain was the Mg2Si intermetallic compound and the surrounding matrix was mostly hcp-Mg. Figure 2d shows an SEM image and quantitative EDX maps of Si and Bi for the precursor dealloyed by immersion in a Bi melt at 1123 K for 30 min. After immersion, the grains retained almost the same shape and dimensions as the original Mg2Si, but their chemical composition changed. In the grain, the Si concentration fluctuated and a Si/Bi composite structure formed. The surrounding matrix was almost wholly replaced by Bi atoms. Taking advantage of the high passivity of Si,25 we selectively etched the composite in aqueous HNO3 to remove the Bi phase from the Si/Bi composite and the matrix. Figure 2e (and the enlarged version in Supporting Information Figure S2a and b) shows an SEM image and quantitative EDX maps of Si and Bi after etching. These results show a grain with a 3DNP

element, Mg acting as sacrificial element, and Bi acting as the dealloying melt medium. By immersing a Si−Mg precursor in a Bi melt, Mg should selectively dissolve from the Si−Mg alloy into the Bi melt, as suggested by our previous work on other porous metals.21−23 Figure 2b shows a schematic of the evolution of 3DNP-Si by immersing a Mg2Si precursor in a Bi melt. Initially, the Mg2Si precursor intermetallic compound has an atomically mixed structure (i). When the precursor is immersed in the Bi melt, only the Mg atoms dissolve into the Bi melt; the Si atoms left at the solid/liquid interface cohere, growing Si islands in the melt (ii). After solidification by cooling, the Bi elements filling the interspace of Si islands are etched away; upon drying in air, the 3DNP-Si is produced (iii). Figure 2c shows a scanning electron microscopy (SEM) image and quantitative energy-dispersive X-ray (EDX) maps of Si and Mg for the as-prepared Mg72Si28 precursor. The microstructure of the precursor exhibited a typical hypereutectic structure with primary dendritic grains embedded in the eutectic matrix. These results show that the primary grain had a uniform 4507

dx.doi.org/10.1021/nl501500g | Nano Lett. 2014, 14, 4505−4510

Nano Letters

Letter

Figure 3. Performance of a battery with the negative electrode made from bulk 3DNP-Si. (a) Voltage−capacity curves of the bulk 3DNP-Si negative electrode after cycling at a constant-charge capacity of 1000 mAh/g. The rate for all tests was 1 C, apart from the rate of 2 C used for the 3D nanoporous Si only at 1000 mAh/g after 721 cycles. (b) Capacity−cycle curve for negative electrode composed of bulk 3DNP-Si (red circles) and commercial nanoparticle Si (blue circles) under constant-charge capacities of 1000 and 2000 mAh/g; the data were plotted every 20 cycles. (c) Voltage−capacity curves of the bulk 3DNP-Si negative electrode after 1, 50, 100, 300, and 500 cycles in constant-current mode at 0.5 C (equivalent to 1800 mA/g). (d) Capacity−cycle curve for negative electrode composed of bulk 3DNP-Si (red circles) and commercial nanoparticle Si (blue circles) in constant-current mode at (1/2) C; the red dotted line shows an extrapolation by an exponential decay function, and the gray solid circles show data measured in constant-charge-capacity mode at 1000 mAh/g.

Here, we discuss the porosity of the bulk 3DNP-Si from the perspective of changes in the crystal structure. The Mg2Si forms a fluorite-type lattice (Pearson symbol: cF12) with a volume per Si atom of 63.6 Å3.26 Mg atoms are removed from the Mg2Si lattice by dealloying and the remaining Si atoms reconstruct a diamond cubic lattice (Pearson symbol: cF8) with volume per Si atom of 20.0 Å3.26 By assuming the sample undergoes no macroscopic volume change during dealloying, this transformation should introduce 68.6% porosity in the sample, which agrees reasonably with our experimental results. On the basis of the 60.4% porosity of the bulk 3DNP-Si, we calculated the ideal volume accommodation limit (VA,L) to be 253%. For simplicity, we do not consider the volume of the binder and the conductive additive here. This corresponds to a lithiation capacity of ∼2000 mAh/g, assuming a linear relationship between the lithiation capacity and the volume per Si atom of the Si−Li compound.27 To examine the hypothesis of the volume-accommodation mechanism shown in Figure 1b, we tested the porous Si electrode by galvanostatically charging and discharging it in constant-charge-capacity mode (at 2000 and 1000 mAh/g) and in constant-current mode. Figure 3a shows voltage−capacity curves for the bulk 3DNP-Si electrodes after 1, 100, 500, 1000, and 1500 cycles at a constant charge capacity of 1000 mAh/g. The lithiation voltage gradually decreased with cycling, indicating that the electrode slowly degraded. Apart from the initial cycle, during which irreversible

structure, enriched with Si and containing almost no Bi. This etch removed the Bi in the matrix; because the grains were isolated, the resulting material was powdery, as shown in Figure 2f. Figure 2g shows a transmission electron microscopy (TEM) image of a single grain from the powder. This grain comprised many interconnected nanoparticles, faceted single crystals with diameters up to 300 nm. This image shows the interconnected structure of the crystals and the openings between the crystals. From the corresponding selected area diffraction patterns, we identified these crystals as silicon with a diamond cubic structure. Supporting Information Figure S2c shows a highresolution TEM image of a Si nanocrystal, revealing a 2−3 nm amorphous layer covering the Si nanocrystals. Using X-ray photoelectron spectroscopy (XPS), we identified this amorphous layer as SiO2 and SiOx (x < 2) suboxides, as shown in Supporting Information Figure S3. The oxide layer, probably formed during the HNO3 etching, stabilized the bulk 3DNP-Si by preventing it from oxidizing further in air. Using mercury porosimetry, we found the average pore size, porosity, and surface area of the bulk 3DNP-Si to be 400 nm, 60.4%, and 7.6 m2/g, respectively (Supporting Information Figure S4). By changing the temperature of the Bi melt and the immersion time, we tailored the length scale of the 3DNP-Si from several tens to several hundred nanometers (Supporting Information Figure S5). 4508

dx.doi.org/10.1021/nl501500g | Nano Lett. 2014, 14, 4505−4510

Nano Letters

Letter

Figure 4. Accommodation of volume expansion by pores in the 3DNP-Si electrode. (a) SEM micrographs of as-prepared electrode. (b) SEM micrographs of electrode after lithiation up to 1000 mAh/g at a rate of 1 C. (c) SEM micrographs of electrode after lithiation up to 2000 mAh/g at a rate of 1 C. Micrographs were taken with secondary electrons (SE) to reveal the surface morphology and with backscattered electrons (BSE) to detect the element distribution. In the BSE images, the heavier Si-based active material appears in bright contrast and the lighter carbon-based conductive additive appears in dark contrast.

reactions such as film formation at the solid/electrolyte interface occurred, the delithiation capacity was nearly 1000 mAh/g and its Coulombic efficiency (delithiation capacity/ lithiation capacity) was more than 99.5% (Supporting Information Figure S6). Figure 3b summarizes the results from Figure 3a, showing the reversible capacity as a function of cycle number. At 2000 mAh/g, for which the expected volume change is nearly equal to the ideal VA,L, the Si electrode exhibited a lifetime of 220 cycles before its performance began decreasing. However, at 1000 mAh/g, for which the expected volume change is well below the ideal VA,L, the electrode exhibited an extremely long cycle lifetime of over 1500 cycles. Figure 3c shows voltage−capacity curves of the bulk 3DNP-Si electrodes after 1, 50, 100, 300, and 500 cycles in constant current-mode at (1/2) C (equivalent to 1800 mA/g). Figure 3d shows the reversible capacity as a function of cycle number, calculated from Figure 3c. Similar to the results of the constantcurrent mode, these results show large irreversible capacity, probably because of film formation at the solid/electrolyte interface in the initial cycles. The bulk 3DNP-Si electrode had an initial capacity of ∼3550 mAh/g, almost the same as the theoretical value (3579 mAh/g) reported for Si at ambient temperature.28 The capacity faded monotonically with cycling, but remained over 1500 mAh/g after 500 cycles. As expected, the capacity fade rate was much greater than that in constantcharge-capacity mode at 1000 mAh/g. By extrapolating the cycle life data of bulk 3DNP-Si with an exponential decay function, we predict the capacity to fade to