Letter pubs.acs.org/NanoLett
New Mechanistic Insights on Na-Ion Storage in Nongraphitizable Carbon Clement Bommier, Todd Wesley Surta, Michelle Dolgos, and Xiulei Ji* Department of Chemistry, Oregon State University, Corvallis, Oregon 97331-4003, United States S Supporting Information *
ABSTRACT: Nongraphitizable carbon, also known as hard carbon, is considered one of the most promising anodes for the emerging Na-ion batteries. The current mechanistic understanding of Na-ion storage in hard carbon is based on the “cardhouse” model first raised in the early 2000s. This model describes that Na-ion insertion occurs first through intercalation between graphene sheets in turbostratic nanodomains, followed by Na filling of the pores in the carbon structure. We tried to test this model by tuning the sizes of turbostratic nanodomains but revealed a correlation between the structural defects and Na-ion storage. Based on our experimental data, we propose an alternative perspective for sodiation of hard carbon that consists of Na-ion storage at defect sites, by intercalation and last via pore-filling. KEYWORDS: Na-ion batteries, intercalation, hard carbon, reaction mechanism, defect sites
I
electrochemical performances in accordance with the cardhouse model reported in the early 2000s.19−22 Under this model, the Na-ion storage is attributed to two distinct mechanisms: (1) storage of ions between the graphene sheets inside the TNs and (2) Na-ion/atom filling of the “pores” between the domains. This model further assigns the Na-ion storage between the graphene sheets to the sloping voltage region of a potentiogram while ascribing the storage in the pores to the plateau region, as illustrated in Figure 1. In this study, we tuned the atomic structure of a model hard carbon in order to confirm the electrochemical structure− property relationships given in the card-house model. We
n today’s world, there is an ever-increasing need to move beyond Li-ion batteries (LIBs) for more sustainable and affordable electrochemical energy storage (EES) solutions.1 While LIBs represent the state-of-the-art technology, the relative scarcity and uneven global distribution of lithium reserves greatly hinders their use in large-scale applications such as grid-level EES.2 To date, Na-ion batteries (NIBs) represent one of the most promising alternatives due to the similar chemical and electrochemical properties between Na and Li and the vast abundance of Na resources.3−6 However, some differences between the two types of atoms have presented considerable challenges in the development of NIBs, especially in the case of anode materials.7,8 LIBs utilize a graphite anode, as Li-ions can reversibly form the LiC6 stage-one graphite intercalation compound (GIC). However, Na-ion’s larger radii of 102 pm vs 76 pm of Li-ion, along with its relatively high ionization potential only allows for the reversible formation of the NaC64 GIC.9−12 In pursuing functional NIB anodes, much attention has been devoted to nongraphitizable carbon, often referred to as hard carbon. Unlike highly ordered graphite, which forms ABAB stacked graphene sheets with an interlayer spacing of 0.335 nm, hard carbon is composed of disordered turbostratic nanodomains (TNs) and empty space (referred to as pores hereafter) between these domains. This leads to three distinct chemical environments for the storage of Na-ions: edge/defect sites on pore surfaces, e.g., carbenes, vacancies, and dangling bonds on the edges of TNs, the interlayer space enclosed inside the TNs, and last, the empty pores.13 This specific structure allows for stoichiometry of NaC7.4 or 300 mAh g−1, which is nearly 10-fold that of graphite. In the literature, most studies on hard carbon and nongraphitic carbons14−18 as NIB anodes rationalize their © XXXX American Chemical Society
Figure 1. Visual representation of the card-house model on Na-ion storage in hard carbon. The two distinct phases: intercalation inside TNs and pore filling are seen. Received: May 19, 2015 Revised: July 3, 2015
A
DOI: 10.1021/acs.nanolett.5b01969 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
also observed through Raman spectroscopy by using the equation below (Figure 2b).25
hypothesized that annealing sucrose-derived hard carbon at progressively increasing temperatures would yield the desired “tunable” hard carbon structures by controlling the domain dimensions of the all-important TNs along the axial axis, Lc, and along the ab planes, La. Sucrose-derived carbons were compared to a commercial glassy carbon (Aldrich) that is essentially the “hardest” available hard carbon, as it is typically obtained through pyrolysis at temperatures of over 2000 °C.23 However, as our results will later show, we found some experimental discrepancies that could not be reconciled with the card-house model. This leads us to propose an alternative perspective on the storage of Na-ions in hard carbon materials. In order to examine the Lc and La parameters of the TNs in different samples, we conducted X-ray diffraction (XRD), Raman spectroscopy, and total scattering via neutron diffraction measurements. Surface area and porosity measurements, alongside TEM images, are also included in the Supporting Information (Table S1 and Figures S1−S5). In the XRD patterns, we observed the (002) peaks, indicative of the dspacing between the graphene sheets in the TNs, to be contracting from ∼0.375 nm for pyrolyzed sucrose annealed at 1100 °C (S-1100), to ∼0.371 nm for S-1400, ∼0.370 nm for S1600, and ∼0.352 nm for glassy carbon (Figure 2a). Average Lc
⎛I ⎞ La(nm) = (2.4 × 10−10)λnm 4⎜ G ⎟ ⎝ ID ⎠
where λnm4 is from the wavelength of the laser in nanometers, ID is the integrated D band at 1350 cm−1 attributed to sp2 carbon atoms with defects, and IG is the integrated G band at 1580 cm−1, which arises from planar sp2 configured carbon atoms and is typical in pristine graphene (Figure S6). Results of the Raman spectra show a similar trend as the XRD results, with La values of ∼9.2, 10.5, 13.3, and 19.4 nm for S-1100, S1400, S-1600, and glassy carbon, respectively (Table S2). The scale discrepancy between the XRD and Raman La values has been noted before, as Raman measurements tend to overestimate the La values.26 However, when performing a linear regression of the Raman La values vs the XRD La values, we find an R2 of 0.95, thus confirming the increasing trend in La in the carbon materials (Figure S7). Additional information from the Raman data includes an ID/IG ratio, which can be used to quantify the concentration of defects along the graphene sheets.27,28 These ratios indicate that the S-1100 material has the greatest concentration of defects, while the glassy carbon contains the smallest concentration. While these traditional measurements are insightful, they fall short of revealing local atomic structures. Thus, to probe the short and midrange order of the carbon materials, we performed total neutron scattering measurements and the associated pair distribution function (PDF) analysis. The peaks up to 5 Å represent the short-range correlations of the C−C bonds of one hexagon unit within a sheet of graphene (Figures 2c and S8). Regardless of the annealing temperatures, all the peaks are well aligned and are consistent with the graphene lattice.29 There is an absence of a C−C correlation at 3.35−3.45 Å, which is found in highly ordered, graphitic carbon. This lack of ordering between the graphene sheets supports the existence of turbostratic disorder. The size of the graphene nanodomains can be estimated in the mid/long-range where well-resolved peaks vanish at certain r values. The S-1100 has notable peak features until ∼15−17 Å, the S-1400 exhibits features diminishing around ∼18−21 Å, while the glassy carbon has features that persist past 30 Å (Figures 2c and S9−S11). These data are consistent with XRD and Raman results, as the glassy carbon exhibits the largest domains, while the S-1100 has the smallest. In addition to the size of the graphene sheets in TNs, neutron PDF results also provide information on the sp2 defect concentration present in the TNs. Importantly, we note that the intensity of the PDF peaks differs from sample to sample. In PDF patterns, the area underneath a peak is proportional to the coordination number of that pair correlation.30,31 At 1.42 Å the intensity of S-1400 and glassy carbon is virtually identical, and only S-1100 has a slightly lower intensity. However, as r increases, the difference in peak intensity becomes readily apparent (Figure 2c). In particular, the first three peaks in the S-1100 pattern have lower intensities, meaning that it possesses a significant number of nonhexagonal carbon rings. These defects likely result in a bending of the graphene sheets, which diminishes the long-range ordering and therefore the size of the domains.32 The PDF data show S-1100 has the greatest defect concentration, the glassy carbon has the least, while S-1400 is
Figure 2. (a,b) XRD patterns and Raman spectra of the different carbon materials. (c) PDF results for total neutron scattering, with the inset showing the short-range order.
and La values were estimated through the PDXL software by use of the Scherrer equation on (002) and (100) peaks, respectively.24 The results reveal that the S-1100 is likely to have the smallest TN dimensions with Lc and La values of 1.15 and 2.54 nm, respectively, while these values increase to 1.17 and 3.18 nm for S-1400, 1.19 and 3.50 nm for S-1600, and 1.25 and 6.13 nm for the glassy carbon (Table S2). While the XRD trend is difficult to unambiguously distinguish between the different sucrose carbons, Raman spectroscopy was used as a local probe for more reliable measurements of amorphous materials. The increase in La is B
DOI: 10.1021/acs.nanolett.5b01969 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
which corresponds to the high sodiation potentials in the sloping region.33−36 Furthermore, the “sloping” nature of this region can be attributed to the fact that there is no clear-cut definition of defect sites in amorphous carbon. They range from dangling bonds at the edge of TNs, to monovacancies, divacancies, stone-wales defects, and extreme curvature in graphene sheets; along with the presence of the sp3 “linking” carbons that connect the neighboring TNs. The plethora of different morphological defects indicates that each type may have its own respective sodiation voltage, thereby giving the sodiation of defects a slope-like shape on the potentiogram curve. Furthermore, considering the delocalized nature of electrons of graphene,37 binding of a sodium-ion at defect sites is more energetically favorable, as the defect sites have low energy unfilled molecular orbitals that can effectively store extra electrons. This increases the binding energy with sodium and thus allows the sodiation to happen at higher voltages vs Na+/ Na. We further investigated the sodiation mechanism through the evaluation of kinetic properties using galvanostatic intermittent titration (GITT) measurements (Figure S13).38,39 The Na-ion diffusivity in hard carbon calculated as a function of potential (Figure 3c) shows that diffusion associated with the sloping potentials is much faster than that with the plateau. This suggests that initial sodiation happens on easily accessible sites in the carbon structure. It is reasonable to say that the surface sites of TNs are more accessible than the interlayer space in the TNs. As these sites are progressively sodiated, Na-ions should then diffuse inside the TNs. However, in order to do so, the Na-ions have to overcome a repulsive charge gradient from the previously bound Na-ions on defect sites in order to diffuse inside the TNs. This explains the steep drop in diffusivity in the plateau region of the potentiogram. These kinetic results not only support our hypothesis that the sloping capacity is due to defect sites on the TN surfaces but also give further insights into the sodiation mechanism of the plateau region of the potentiogram curve. This has been the subject of ambiguity, as recently published experimental results have suggested that the plateau region is due to Na insertion into TNs as opposed to the pore filling mechanism suggested by the card-house model.40−42 Similarly, we also observed reversible expansion and contraction of d-spacing due to sodiation and desodiation at the plateau region, i.e., from 0.2 to 0.01 V, while conducting ex situ XRD. This suggests that intercalation occurs at lower voltages (Figure 4a,b). Despite our hypothesis that the plateau region could be attributed to intercalation of sodium-ions in TNs, the newly obtained diffusion results, combined with the rest of our experimental data suggest that the plateau region is in fact a function of both intercalation into TNs as well as pore-filling suggested by the card-house model. When observing the values at low voltages, we can observe that diffusion values reach a minimum at 0.05 V; the same voltage where dQ/dV values reach a maximum. From 0.05 V to the cutoff voltage, the diffusion values gradually increase, while dQ/dV values decrease (Figure 3d). This observation begs the question: if intercalation is the one and only mechanism in the plateau region, why do diffusivity values begin to rise at the very end, displaying an almost U-turn like reversal? If intercalation were the only mechanism, diffusivity values should become progressively lower and lower, as the diffusion length to storage sites keeps
somewhere in the middle. The trend regarding the defect concentration corroborates the Raman spectrum results. Using several techniques, we have demonstrated that TN sizes increase upon annealing and are the largest for glassy carbon. According to the card-house model, capacity obtained in the sloping region, defined as the region above 0.115 V vs Na+/Na of the potentiogram curve, is due to intercalation of sodium-ions in the TNs. With larger TN sizes, the card-house model predicts increasing sloping capacity from S-1100 to S1600. However, the sloping capacity decreases from 120 mAh g−1 to 102 mAh g−1 and to 92 mAh g−1 for S-1100, S-1400, and S-1600, respectively (Figure 3a). One may be concerned about
Figure 3. (a) Sodiation potentiograms for different carbons. Complete electrochemical data in Figures S14−S20 and Table S3−S5. (b) Plot of the sloping capacity vs the ID/IG ratio from Raman spectra. (c) GITT profile and diffusivity as a function of states of charge (inset). The diffusivity values are clearly overestimated as the geometric area of electrodes is used for calculation. Our primary purpose is to demonstrate the trends of evolving diffusivity at different state of charge. (d) dQ/dV plot from 0.12 to 0.01 V with corresponding diffusivity values.
whether it is the decreased d-spacing from S-1100 to S-1600 that causes the lower capacity. As estimated from the XRD patterns, we notice that the d-spacings are only different by 0.005 nm between S-1100 and S-1600. We are aware that this average d-spacing difference is beyond relevant meaning in physics. Instead of being dependent on the size of TNs, the capacity in the sloping region seems to be directly correlated to the defect concentration present. When plotting the slope capacity versus defect concentration in the carbon samples, as expressed by the integrated ID/IG ratio, we observe a linear relationship with an R2 value of 0.90 (Figures 3b and S12). This leads to our hypothesis that the defected carbon sites in hard carbon rather than the TNs are responsible for the capacity in the sloping region. The obtained PDF data also supports this hypothesis as it shows that the S-1100 material has the greatest defect concentration and the largest sloping capacity, while the glassy carbon has the lowest defect concentration and the smallest sloping capacity. Prior computational studies also support this hypothesis as they show that binding energy between Na-ion and carbon is highest at graphene defect sites and vacancy sites, C
DOI: 10.1021/acs.nanolett.5b01969 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
bilayer.34 This is a logical claim, as a sodium-ion between two graphene sheets is more coordinated than one ion interacting with only a single graphene sheet. To summarize, we have revealed some discrepancies with the card-house model on the Na-ion storage in hard carbon. We have demonstrated that the storage mechanism in the sloping part of the potentiogram curve can be better explained through storage at defect sites, as opposed to intercalation between graphene sheets. Furthermore, our additional diffusion, ex situ XRD, and Na-plating experiments lead us to hypothesize that the storage mechanism in the low voltage plateau region is due to the intercalation between graphene sheets and minor phenomenon of Na-ion adsorption on pore surfaces as shown in the graphical representation in Figure 4d. Lastly, we realize that this Letter might be regarded as controversial in the field of sodium-ion storage in amorphous carbon, some of its indications are at odds with long-standing principles. Our sincere hope is to present another angle of perspective to this issue.
Figure 4. (a) Ex situ XRD profiles at various stage of sodiation (S) and desodiation (D) with the (002) peak indicated. (b) Corresponding dspacing plots. (c) Potentiogram of sodiation until Na-metal plating is induced. (d) Potentiogram and schematic of proposed Na-ion threepart storage mechanism. Please note that the figure is meant as a schematic representing the three different types of binding sites as opposed to an accurate representation of the hard carbon structure.
■
ASSOCIATED CONTENT
S Supporting Information *
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.nanolett.5b01969. Surface area and porosity, HRTEM, additional PDF data, complete electrochemical characterization, and experimental methods (PDF)
increasing. However, the continuous drop in diffusivity values is not observed. This suggests that the sodiation mechanism is changing in the voltage range right before reaching the cutoff potential. At such low voltages, storage sites can be characterized by a weak binding energy. Furthermore, the increasing diffusivity values suggest a more facile diffusion than intercalation. Considering the weak binding energy and facile diffusion, we postulate that storage at low voltages is due to the Na-atom adsorption on the sp2 configured pore surfaces. First off, a sp2 graphene surface has been shown in computational studies to be the least energetically favorable binding site for sodium-ions.33−36 Thus, it is logical that storage of sodium-ions on pore surfaces should only occur at low voltages close to that of sodium plating. Additionally, unlike intercalation, storage of on pore surface does not require the expansion of two adjacent graphene layers to make insertion possible, which may account for its faster diffusivity. To test the above hypothesis, it is critical to determine whether such adsorption is of the same potential as the Na plating. When running a half-cell to voltages below 0.0 V vs Na+/Na, we found that the onset of Na-metal plating actually occurs at −0.02 V before reaching a steady value of −0.015 V (Figure 4c). When looking at the potentiogram near 0.0 V, we notice that the slope becomes much steeper starting at 0.04 ± 0.01 V and stays linear until plating begins. We hypothesize that this is a result of continuous deposition of sodium atoms on the pore surfaces until a critical point is reached, and the sodium atoms begin to agglomerate into metallic clusters. Therefore, considering the increase in diffusivity values at the very end of the sodiation process, we find it reasonable to propose that the last storage mechanism that occurs during the sodiation of hard carbon is the deposition of sodium atoms on pore surfaces. This hypothesis of storage on pore surfaces is in good agreement with the original one made by the card-house model and is also supported by the computational results; which show that the binding energy between planar graphene sheets and Na is much smaller than binding on defect sites or in a graphene
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS X.J. acknowledges the financial supports from Advanced Research Projects Agency-Energy (ARPA-E), DOE of the United States, Award number: DE-AR0000297TDD. M.D. gratefully acknowledges the financial support from Oregon State University. Research at ORNL’s Spallation Neutron Source was sponsored by the Scientific User Facilities Division, Office of Basic Energy Sciences of the U.S. DOE. X.J. thanks Dr. Yuliang Cao from Wuhan University for discussion.
■
REFERENCES
(1) Choi, N. S.; Chen, Z. H.; Freunberger, S. A.; Ji, X. L.; Sun, Y. K.; Amine, K.; Yushin, G.; Nazar, L. F.; Cho, J.; Bruce, P. G. Angew. Chem., Int. Ed. 2012, 51, 9994−10024. (2) Thackeray, M. M.; Wolverton, C.; Isaacs, E. D. Energy Environ. Sci. 2012, 5, 7854−7863. (3) Pan, H. L.; Hu, Y. S.; Chen, L. Q. Energy Environ. Sci. 2013, 6, 2338−2360. (4) Kim, S. W.; Seo, D. H.; Ma, X. H.; Ceder, G.; Kang, K. Adv. Energy Mater. 2012, 2, 710−721. (5) Yabuuchi, N.; Kubota, K.; Dahbi, M.; Komaba, S. Chem. Rev. 2014, 114, 11636−11682. (6) Slater, M. D.; Kim, D.; Lee, E.; Johnson, C. S. Adv. Funct. Mater. 2013, 23, 947−958. (7) Bommier, C.; Ji, X. Isr. J. Chem. 2015, 55, 486−507. (8) Chevrier, V. L.; Ceder, G. J. Electrochem. Soc. 2011, 158, A1011− A1014. D
DOI: 10.1021/acs.nanolett.5b01969 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters (9) Doeff, M. M.; Ma, Y. P.; Visco, S. J.; Dejonghe, L. C. J. Electrochem. Soc. 1993, 140, L169−L170. (10) Enoki, T.; Endo, M.; Suzuki, M. Graphite Intercalation Compounds and Applications; Oxford University Press: Oxford, U.K., 2003. (11) Okamoto, Y. J. Phys. Chem. C 2014, 118, 16−19. (12) Wang, Z.; Selbach, S. M.; Grande, T. RSC Adv. 2014, 4, 3973− 3983. (13) Franklin, R. E. Proc. R. Soc. London, Ser. A 1951, 209, 196−218. (14) Li, Y.; Xu, S.; Wu, X.; Yu, J.; Wang, Y.; Hu, Y.-S.; Li, H.; Chen, L.; Huang, X. J. Mater. Chem. A 2015, 3, 71−77. (15) Alcantara, R.; Lavela, P.; Ortiz, G. F.; Tirado, J. L. Electrochem. Solid-State Lett. 2005, 8, A222−A225. (16) Wenzel, S.; Hara, T.; Janek, J.; Adelhelm, P. Energy Environ. Sci. 2011, 4, 3342−3345. (17) Ponrouch, A.; Goni, A. R.; Palacin, M. R. Electrochem. Commun. 2013, 27, 85−88. (18) Wen, Y.; He, K.; Zhu, Y.; Han, F.; Xu, Y.; Matsuda, I.; Ishii, Y.; Cumings, J.; Wang, C. Nat. Commun. 2014, 5, 4033−4042. (19) Thomas, P.; Billaud, D. Electrochim. Acta 2002, 47, 3303−3307. (20) Stevens, D. A.; Dahn, J. R. J. Electrochem. Soc. 2001, 148, A803− A811. (21) Stevens, D. A.; Dahn, J. R. J. Electrochem. Soc. 2000, 147, 1271− 1273. (22) Stevens, D. A.; Dahn, J. R. J. Electrochem. Soc. 2000, 147, 4428− 4431. (23) Hishiyama, Y.; Inagaki, M.; Kimura, S.; Yamada, S. Carbon 1974, 12, 249−258. (24) Warren, B. Phys. Rev. 1941, 59, 693−698. (25) Cançado, L. G.; Takai, K.; Enoki, T.; Endo, M.; Kim, Y. A.; Mizusaki, H.; Jorio, A.; Coelho, L. N.; Magalhães-Paniago, R.; Pimenta, M. A. Appl. Phys. Lett. 2006, 88, 163106. (26) Vázquez-Santos, M. B.; Geissler, E.; László, K.; Rouzaud, J.-N.; Martínez-Alonso, A.; Tascón, J. M. D. J. Phys. Chem. C 2012, 116, 257−268. (27) Ferrari, A. C.; Robertson, J. Philos. Trans. R. Soc., A 2004, 362, 2477−2512. (28) Ferrari, A. C.; Robertson, J. Phys. Rev. B: Condens. Matter Mater. Phys. 2000, 61, 14095−14107. (29) Zhou, P.; Lee, R.; Claye, A.; Fischer, J. E. Carbon 1998, 36, 1777−1781. (30) Egami, T.; Billinge, S. J. Underneath the Bragg Peaks: Structural Analysis of Complex Materials; Elsevier: New York, 2003. (31) Keen, D. A. J. Appl. Crystallogr. 2001, 34, 172−177. (32) Petkov, V.; Difrancesco, R. G.; Billinge, S. J. L.; Acharya, M.; Foley, H. C. Philos. Mag. B 1999, 79, 1519−1530. (33) Datta, D.; Li, J. W.; Shenoy, V. B. ACS Appl. Mater. Interfaces 2014, 6, 1788−1795. (34) Tsai, P.-c.; Chung, S.-C.; Lin, S.-k.; Yamada, A. J. Mater. Chem. A 2015, 3, 9763−9768. (35) Malyi, O. I.; Sopiha, K.; Kulish, V. V.; Tan, T. L.; Manzhos, S.; Persson, C. Appl. Surf. Sci. 2015, 333, 235−243. (36) Shen, H.; Rao, D.; Xi, X.; Liu, Y.; Shen, X. RSC Adv. 2015, 5, 17042−17048. (37) Popov, I. A.; Bozhenko, K. V.; Boldyrev, A. I. Nano Res. 2012, 5, 117−123. (38) Weppner, W.; Huggins, R. A. J. Solid State Chem. 1977, 22, 297−308. (39) Prosini, P. P.; Lisi, M.; Zane, D.; Pasquali, M. Solid State Ionics 2002, 148, 45−51. (40) Ding, J.; Wang, H.; Li, Z.; Cui, K.; Karpuzov, D.; Tan, X.; Kohandehghan, A.; Mitlin, D. Energy Environ. Sci. 2015, 8, 941−955. (41) Lotfabad, E. M.; Ding, J.; Cui, K.; Kohandehghan, A.; Kalisvaart, W. P.; Hazelton, M.; Mitlin, D. ACS Nano 2014, 8, 7115−7129. (42) Ding, J.; Wang, H. L.; Li, Z.; Kohandehghan, A.; Cui, K.; Xu, Z. W.; Zahiri, B.; Tan, X. H.; Lotfabad, E. M.; Olsen, B. C.; Mitlin, D. ACS Nano 2013, 7, 11004−11015.
E
DOI: 10.1021/acs.nanolett.5b01969 Nano Lett. XXXX, XXX, XXX−XXX
