3.1. X-ray Photoelectron Spectroscopy
The XPS analysis showed that the pristine powder surface is composed mostly of oxygen-, sulfur- and nitrogen-containing compounds (Table 1
). The main difference between the three investigated powders was related to a different elemental silicon to silicon oxides ratio (Figure 1
). The XPS analysis of modified samples revealed a lower elemental silicon content (21.4 and 2.1% for Si-1 and Si-2 samples, respectively). This variation is the effect of the different silicon oxide layer thickness at the grain surface. The photoelectron effective signal acquisition depth is in the range of 5–15 nm. The detected elemental silicon signal originates from the silicon core, which is present under the top oxide layer. The reduced Siel
signal among modified powders indicate a much thicker oxide coating on those materials. The Si-2 surface oxide layer is thick enough to almost completely block the Siel
signal, indicating the surface oxide layer thickness in the range of photoelectron penetration (~15 nm). The Si-1 powder analysis showed that the Siel
signal was only 5% smaller compared to the pristine powder, thus the very minor oxide thickness buildup occurred during first procedure. It can be also observed that reduction of the native oxide layer during procedure 2, resulted in an absence of SiO in the final product (Si-2).
A large difference in the oxide thickness between Si-1 and Si-2 powders may originate from two different mechanisms: (a) The native oxide layer is so dense and compact, that it almost completely blocks the access of the oxidizing solution to the elemental silicon core, and as a consequence, inhibits the reaction; (b) The traces of HF transferred to the piranha solution during the second procedure lead to silicon oxide dissolution. In that situation the dissolution of SiO2
reveals a fresh Si surface, which is immediately oxidized by the piranha solution. The process continues until all the HF is consumed. The dissolution/oxidation process may proceed with different kinetics at different silicon crystal planes or due to local substrate concentration differences. In that situation, the oxidation/dissolution processes will generate porous oxide structure embedded deeply in the silicon core. As a result of this two opposite reactions a thick oxide layer will be formed in a process similar to Al or Ti oxidation in HF solution [37
The small differences in chemical shift of various forms of silicon-oxygen bonds makes it very difficult to separate the signals from individual compounds/surface groups. The possible determination of small differences in the silicon oxide morphology obtained by modification 1 and 2 will be the scope of further studies.
A small signal (2.7 at.%) related to the presence of nitrogen compounds was found at 402.95 eV, which is the most likely a result of sample contamination during Si NP production. Nevertheless, there was no N KLL line visible in the wide scan, which should be detected even when a small amount of nitrogen is present in the sample. The chemical shift of N 1s (402.95 eV) is close to the region of nitrogen in silicone oxynitride [39
]. Unfortunately, this chemical shift is also in good agreement with other nitrogen compounds bonded with oxygen, e.g., nitrates, nitrated organic compounds, etc. As a result, the direct attribution of the observed signal to a specific nitrogen form was not possible. The overall nitrogen quantity in the modified samples was much lower compared to the untreated material, indicating that the observed compound, whatever its nature, was partially washed away during chemical treatment.
The sulfur contamination originated from the presence of SO4 groups. The magnitude of the sulfur line in the Si-1 sample was slightly higher compared to pristine material due to the chemical modification procedure involving the use of H2SO4. Interestingly the Si-2 sample presented a lower sulfur content.
The Si-2 powder surface was covered with fluorine compounds related to HF utilization in the first part of the modification procedure. The fluorine signal presence originated from Si-F and C-F compounds—probably from partially fluorinated silicon and hydrocarbon compounds or adsorbed SiF62− anions formed during SiO2 dissolution.
3.2. SEM Microscopy
SEM imaging revealed that both of the modified samples contained small nanoparticles with an average diameter of 100 nm (Figure 2
C–F). The main difference between the Si-01 and Si-02 samples can be observed at higher magnifications. The sample modified with piranha solution (Si-01) has well separated NPs (similar to a pristine material—Figure 2
A), while the sample modified with hydrofluoric acid and piranha solution (Si-02) is mostly composed of pristine silicon NPs encapsulated in a dense, agglomerated matrix. This well-visible matrix is additional evidence of the thick silicon oxide layer growth on top of the previously HF-etched Si NPs. The in-situ silicon oxide layers probably interconnect and form a shared oxide matrix that acts as a “glue”. Consequently, the silicon particles are encapsulated in the silicon oxide matrix and thus large, uniform core/shell structures are formed.
It can be assumed that the agglomerated oxide matrix is strongly connected to the Si core and does not undergo separation during slurry preparation, leading to the agglomerated structure presence in the final electrode layer. The effect of agglomeration on the electrochemical behavior may be dual. Larger structures may lead to less uniform mixing with conductive carbon and binder in the slurry. On the other hand, oxide matrix may stabilize the silicon NPs during volume changes. The impact of different oxides structure on the electrode parameters was further evaluated by a series of electrochemical experiments.
The EDX analysis revealed gradual changes of the oxygen content in modified samples. In good accordance with the XPS analysis, the EDX data showed that the Si-1 sample has the oxide content very similar to that of pristine NPs. The slight increase in the oxygen content is related to surface oxidation by piranha solution. The Si-2 sample results showed a major increase in the oxygen content related to the thick oxide layer formation after the modification procedure. The O/Si ratio for Si-2 sample is over two times higher than for pristine NPs and Si-1 material. The ratio, however, is lower that 2.0 (expected for pure SiO2 phase) indicating the presence of elemental silicon underneath the surface oxide layer. The exact content of elemental silicon is hard to determine due to the presence of SiOx, however the estimated oxide thickness (initial particle radius = 50 nm) is close to 6.6, 8.2 and 17.8 nm for the pristine, Si-1 and Si-2 samples, respectively. The estimated oxide thickness is in good agreement with XPS data that revealed very low amount of elemental silicon content in the Si-2 sample. The XPS maximum excitation depth is close to 15 nm. The oxide thickness of Si-2 sample is higher, leading to a lack of Si signal in the XPS analysis.
The EDX analysis revealed small amounts of sulphur (Si-1 and Si-2) and fluorine (only Si-2) contamination due to the presence of residues remaining after the modification procedures. The amount of contaminants was significantly higher in the Si-2 sample (S at. % = 1.33, F at. % = 2.07) compared to the Si-1 sample (S at. % = 0.58). This was probably related to an adsorption of Sulphur- and fluorine-containing species in the thick and porous oxide layer. The comparison with XPS results revealed a much lower sulphur content in the EDX analysis, leading to the conclusion that S-containing residues are present mostly on the surface of the oxide layer. The fluorine concentration was similar in both experiments (XPS and EDX) which suggests a uniform distribution of F-containing species in the whole oxide layer. Detailed results obtained by the EDX analysis can be found in Table 2
and the Supplementary Information (Figures S3 and S4)
Based on XPS and SEM analyses it can be assumed, that two materials produced by different procedures are characterized by distinct morphology. The grains of Si-1 sample (one-step preparation) are covered by thin, dense layer of silicon oxide, while Si-2 sample (two-step procedure) is distinguished by thick porous coating (Figure 3
3.3. Electrochemical Results
Cyclic voltammetry (CV) was performed at a 0.2 mV/s sweep rate. The measurements were conducted before the 1st (freshly assembled), and after 3rd, 5th and 20th galvanostatic cycle. The results show big differences between the CV profiles of the modified powders. During the 1st scan of the Si-1 sample (thin oxide layer), the cathodic part contains only one very small peak at 1.8 V vs. Li/Li+
(R1 at Figure 4
). This peak is probably related to a partial reduction of the surface oxide to lithium silicate. No obvious peaks at the lower potential range (1.7–0.3 V vs. Li/Li+
) were detected indicating a very small SEI generation charge. The further lithium silicate reduction to Lix
and subsequent reduction to Li2
O and Siel
should proceed at 1.3 V and 0.7 V as stated in the literature [36
]. The lack of these peaks indicates that the generated lithium silicate layer is somehow resistant to electro-reduction. The very dense and compact structure of the formed oxide might be the cause of this phenomenon. Silicon oxide is a very poor electronic conductor and restrains the electrolyte decomposition. The formed dense oxide acts as artificial SEI layer, seals the electrode surface and leads to a smaller electrolyte components’ reduction charge during the first cathodic scan. During the first scan, the cathodic part contains a sharp peak starting at ca. 0.2 V vs. Li/Li+
related to the silicon two-phase lithiation process. In further scans, the mechanism of silicon lithiation changes to two separate one-phase reactions reflected by the formation of a R2 peak at higher potential (Figure 4
). During an oxidation scan of the Si-1 sample, two peaks: O1 and O2 (Figure 4
) are clearly visible.
These peaks correspond to well-known reactions of the Li-Si alloy delithiation, proceeding through a Li7Si12 intermediate phase. The reaction related to peak O1 is much slower compared to the subsequent intermediate phase delithiation. The O1 peak is highly broadened, shifted towards positive potentials, and for that reason highly imposed onto the O2 peak. The Li-Si alloy oxidation mechanism is similar during the first 20 cycles, however a small shift towards a positive potential value can be seen between cycle 1 and 3 indicating a decrease of the reaction kinetics during the 3rd and following cycles. The O2 peak potential shift between 5th and 20th cycle of the Si-1 is almost indistinguishable, suggesting a high electrochemical and mechanical stability of Si-1 electrodes.
The CV analysis of the Si-2 sample revealed a completely different electrochemical behavior. During the first scan, two peaks related to a silicon oxide reduction to lithium silicates are visible (R1A and R1B). The large area of the R1B peak (Figure 5
) indicates that a high percentage of lithium silicate was reduced to Lix
. At potentials between 0.3–0.8 V, no obvious peak is visible, however a high cathodic current flow suggests the high electrolyte decomposition and SEI formation rate. The lack of the peak related to the SEI formation may be related to a poor conductivity of the reduced lithium silicate that kinetically hinders the reaction and leads to the severe SEI formation peak broadening. The appearance of R2 peak during the 3rd and further cycles indicates the silicon lithiation through the mechanism similar to that observed for sample Si-1. The oxidation (delithiation) of the Si-2 material proceeds through the intermediate phase formation similar to the Si-1 electrode. The separation of O1 and O2 peaks (Figure 5
) is much greater compared to the Si-1 sample. This difference shows faster kinetics of Li7
formation for the Si-2 sample. Since this reaction is related to delithiation of bulk silicon (core), it should be independent of the surface chemistry (e.g., oxide layer thickens and morphology). We believe that smaller size of the SiEL
core (for the Si-2 sample) resulted in a higher ratio of O1:O2 reaction rates due to lower internal stress generation. This lower stress magnitude originates from the good separation of the Si grains by the oxide matrix [7
The average peak current of the Si-2 electrode is 4–6 times lower compared to the Si-1 sample. The overall reaction rate is much slower in the Si-2 electrode due to a large amount of poorly conductive silicon oxide which limits the electrical conductivity of the active mass. On the other hand, the large amount of silicon oxide can also easily limit the electrolyte access to the silicon core leading to mass transport limitations. Regardless of the limitation mechanism, the Si-2 electrode electrochemical resistance is much higher leading to smaller current density capabilities.
Different factors can affect the electrochemical results obtained for the silicon-based electrodes. The common problems observed for some Li-ion systems are poor electrical conductivity of the material or insufficient electrolyte penetration into the active mass. Those restrictions result in overpotential generation visible as peak flattering, shifting and broadening in CV experiments, as well as in poor capacity obtained for fast (high current) galvanostatic (GA) cycling. It should have marginal effect on the capacity obtained during slow GA cycling because the generated overpotential is directly proportional to the current density.
The third factor affecting the electrochemical results is distinctive for silicon (and few other alloy-type materials). It is related to internal pressure generation during lithium uptake (Si-Li alloy volume change). This pressure generates the overpotential in the range of 60–125 mV/GPa, while Si internal stress induced by lithiation can be as high as 1–3 GPa [16
]. As the result, the lithiation reaction can be slowed down or completely stopped [14
]. The stress-induced overpotential is affected by current density only in small extent and should result in a similar GA capacity drop and CV peak area reduction.
The galvanostatic cycling of the modified materials showed major performance differences (Figure 6
). The initial specific capacity of the Si-1 sample (thin surface oxide) is almost the same as revealed by the pristine material (silicon NP’s). During a few subsequent cycles, the pristine material capacity increased by 12%, suggesting electrode structure changes resulting in an improvement of the electrolyte penetration into the active mass or smaller stress generation. The Si-1 electrode capacity remained almost constant between five initial cycles (increase of only 1.8%) suggesting an initial good electrolyte access to deeper electrode regions and fairly stable electrode macro-structure. After the 10th cycle both: pristine and Si-1 electrode presented close to linear capacity drop, although the capacity retention of the modified sample (Si-1) was much better and reached 83% after 50 cycles, in comparison with 40% for pristine silicon NP’s. The significant difference in the capacity retention between those two materials is the effect of diverse surface oxide characteristics. As indicated by XPS and CV experiments, the Si-1 powder is covered with a thin, but uniform and electrochemically resistant oxide layer. The thickness of the oxide is generally higher for the modified sample which results in a better volume changes buffering, as well as an improved SEI stability. The maximum cycle efficiency (Figure 6
) increased from 96.2% (for pure silicon NP’s) to 99.1% after material modification. A much better SEI stability results in a less electrode degradation and a slower internal resistance buildup during cycling, thus can be assigned as one of the reasons for a better Si-1 capacity retention. The first cycle efficiency (star in a circle in Figure 6
) is slightly lower for the modified sample (71% vs 78%) indicating that some amount of the oxide layer might be reduced during the first lithiation or the specific surface of the Si-1 sample increased after Si NPs modification. The first cycle Coulombic efficiency difference between the pristine and Si-1 modified sample is still relatively small, which can be attributed to the high electrochemical stability of the oxide layer at the modified silicon surface. Because of the high stability and a very high initial specific capacity, the Si-1 material can be directly used as an active material in Li-ion batteries.
The Si-2 electrode behavior during cycling was much different compared to both pristine and Si-1 modified material. The Si-2 electrode reveals an initial capacity close to 550 mAh/g and maintains almost a constant capacity during 50 cycles. The obtained capacity is almost 50% greater compared to graphite. A slight capacity increase occurred up to 30th cycle (14% gain). A comparison between CV and cycling data shows that the lithiation/delithiation reactions kinetic greatly increases during the first 20 cycles as indicated by an over 300% increase in the oxidation peak current in CV experiment, while the specific capacity change during these cycles are only minor. The Si-2 electrode reaction rates (and related peak currents) raise during cycling. This phenomenon is probably related to a couple of factors. One of them is an ongoing transformation of SiO2
(insulator) into Lix
(2.7 S/cm [42
]). It can also originate from electrode structure changes induced by silicon volume variations. After several shrinking/expanding cycles, the silicon-based electrode structure usually becomes loose and a better electrolyte access into silicon grains is possible [11
]. On the other hand, the loose electrode matrix structure leads to unrestricted silicon volume changes. The tight electrode structure of Si-2 sample may therefore limit the electrode capacity during a few initial cycles. After a couple of cycles the loosened structure generates less pressure on the expanding Si-grains and thus the overall capacity increases [7
]. This demonstrates that a high compressive stress generated during the Si particle expansion can completely block the ongoing electrochemical reaction. Lack of oxidation peaks height increase for silicon NP’s and Si-1 materials (Figure S1
and Figure 4
respectively) is an additional indication of its intrinsic looser structure.
Generally, the stress induced lithiation restriction should affect both the galvanostatic cycling capacity and CV peak height. Since the galvanostatic cycling was performed using a small current density, the polarization caused by poor electric conductivity or electrolyte access limitations should have only a minor effect on the obtained capacity. The differences between GA capacity profiles during initial cycles indicates that stress induced lithiation restrictions (observed for the Si-1 sample) were much smaller in the Si-2 electrode, however it was not the only factor responsible for the different electrochemical response of these samples. If the large compressive stress were the only factor, the CV peak current and specific capacity gain in the initial cycles should present a similar magnitude. Thus, the relatively small capacity gain during cycling and the large oxidation peak current increase during CV of the Si-2 sample must originate from a better active layer penetration by the electrolyte. The smaller CV peak height changes for the Si-1 sample indicated that the electrolyte access was sufficient from the beginning of the 1st cycle. The very thick oxide layer formed at the Si-2 material resulted in a decrease of the electrolyte access during the initial cycles. The electrode structure changes induced by Si swelling/shrinking promotes better electrolyte penetration into the active mass after prolonged cycling.
The capacity of Si-2 electrode was almost constant during the first 50 cycles. The experiment was then continued up to 90 cycles (Figure S2
). The electrode retained over 98% of initial capacity after this period. The great capacity retention proves that very thick oxide layer formed at the Si-02 powder during the modification procedure can provide an elastic matrix and thus greatly suppress the electrode degradation induced by silicon volume changes during cycling. Unfortunately, the first cycle efficiency of Si-2 electrode was very low (52%). The main reason for this effect is the severe reduction of the oxide as revealed by CV experiments. The poor initial cycle efficiency is a common problem observed for SiOx
-based electrodes. Nevertheless, the efficiency value recorded for the Si-2 sample is higher compared to the reported in the literature for SiOx
-based electrodes (30–48%) [43
]. Furthermore the Si-2 material maintains a similar or higher specific capacity and capacity retention [43
]. After the initial cycle, the efficiency quickly increases and stabilizes at 99.3% indicating very good electrode structure and SEI stability. Even though the initial oxide reduction is greatly limiting for the practical application of the modified Si-2 material, the use of the pre-lithiation step may greatly improve the material characteristics and applicability in Li-ion batteries. The pre-lithiation process should result in an improvement of several electrochemical parameters. First, after pre-lithiation the SEI forms naturally on the electrode surface due to its low initial potential. Secondly, the Si-oxides are reduced to Si and Li2
O prior to the cell operation, leading to a great first cycle efficiency improvement (lower IRR) and a great reduction of the lithium losses. Several different pre-lithiation techniques had been proposed so far. A high-energy milling of silicon with lithium [45
], a sacrificial lithium electrode utilization [46
] or a stabilized lithium NPs (SLMP) addition into the active mass [47
] were successfully developed at the lab-scale. Unfortunately, the utilization of these methods at the industrial scale is currently very difficult due to the high reactivity of pre-lithiated materials. Nevertheless, the utilization of pre-lithiation should result in first cycle IRR drop, and consequently, in successful application of Si-2 material.