Solid-state supercapacitors with rationally designed heterogeneous electrodes fabricated by large area spray processing for wearable energy storage applications.

Supercapacitors are in demand for short-term electrical charge and discharge applications. Unlike conventional supercapacitors, solid-state versions have no liquid electrolyte and do not require robust, rigid packaging for containment. Consequently they can be thinner, lighter and more flexible. However, solid-state supercapacitors suffer from lower power density and where new materials have been developed to improve performance, there remains a gap between promising laboratory results that usually require nano-structured materials and fine-scale processing approaches, and current manufacturing technology that operates at large scale. We demonstrate a new, scalable capability to produce discrete, multi-layered electrodes with a different material and/or morphology in each layer, and where each layer plays a different, critical role in enhancing the dynamics of charge/discharge. This layered structure allows efficient utilisation of each material and enables conservative use of hard-to-obtain materials. The layered electrode shows amongst the highest combinations of energy and power densities for solid-state supercapacitors. Our functional design and spray manufacturing approach to heterogeneous electrodes provide a new way forward for improved energy storage devices.

Supercapacitors with approximately ten times higher power denpan class="Chemical">sities but lower energy densities than Li ion batteries can be coupled with batteries to meet peak power demands, or coupled with energy-harvesting systems to store intermittently generated electricity from renewable sources over short time periods1. Most commercial supercapacitors use liquid organic electrolytes such as tetraethylammonium-tetrafluoroborate in acetonitrile23. Although these electrolytes provide relatively high ionic mobility and fast charge/discharge kinetics, potential electrolyte leakage raises safety and environmental concerns that requires cell designs with rigid and robust packaging to contain the liquid electrolytes. Solid-state supercapacitors that replace liquid electrolytes with an ion conducting gel or polymer membrane do not require rigid packaging and consequently can be thinner, lighter and provide greater design freedom. They are thus potentially attractive for applications such as future wearable electronics and conformal energy storage systems. However, the main weaknesses of solid-state supercapacitors are reduced ion mobility and reactivity that undermine the key supercapacitor advantage of high power density4. Efforts to reduce these limitations include the development of solid-state electrolytes with progressively higher ion mobility567, more electrochemically reactive electrodes8910 and higher ionic and electrical conducting electrodes1112.

Here, for the first time, we take a different approach based on the use of layered electrodes in which the material and/or morphology in each layer is different (Fig. 1(a)). The arrangement is depan class="Chemical">signed to exploit the inherent advantages of each material and morphology optimally at each position in the electrode. While very thin structured electrodes (100–600 nm) have been reported to exhibit excellent performance13, we report a layered electrode of a commercially practical thickness (~33 μm) that exhibits amongst the best reported charge storage performance for a solid-state device. Specific capacitance and power performance usually decrease as electrodes become thicker because of decreasing material utilisation through the electrode thickness, especially when ionic diffusion is limited such as the case in solid-state supercapacitors; but thicker electrodes are desirable to reduce the overall current collector area and increase capacitance per area14. The focus here is to maximise the utilisation of active materials in relatively thick, structured electrodes for solid-state supercapacitors.

Figure 1

Schematic diagrams of (a) the heterogeneous three-layer electrode for solid-state supercapacitors; (b) the spray deposition arrangement used to fabricate solid-state supercapacitors; and (c) an as-sprayed 16 cm x 9.5 cm flexible solid-state supercapacitor to demonstrate scalability of the fabrication technique.

The electrodes were manufactured using a scalable spray technique that enabled a discrete, heterogeneous three-layer electrode structure to be manufactured over large areas. Because of the discrete layered structure and the different intrinsic response of each layer, a transition in the electrode charge storage dynamics, from charge storage primarily due to redox reaction at slow scan rates (<500 mV s−1) to solely high electric double layer (EDL) capacitance at fast scan rates (500 mV s−1) was clearly resolved.

Design rationale for the electrode structure

Laboratory synthesised meso-porous anatase pan class="Chemical">TiO2 particles (p-TiO2) have recently been studied for energy conversion and storage applications, such as solar cells and supercapacitors1516, but this is the first time that p-TiO2 has been applied for solid-state supercapacitors. These meso-porous TiO2 particles made by a template-based route are potentially attractive because of their high surface area, but are available currently only in comparatively small quantities. We mixed p-TiO2 with multi-wall carbon nanotubes (MWNTs) in a stable suspension that was sprayed directly onto a 1M H2SO4 treated H+ ion conducting Nafion membrane to form the first, 650 nm thick layer (Layer 1) of a three-layer electrode, as shown schematically in Fig. 1(a). The rationale was to exploit optimally the high surface redox reactivity of p-TiO2 at the membrane interface where there was the highest concentration of H+ ions, and so to use this hard-to-obtain material both efficiently and conservatively.

A much thicker (Layer 2, 32 μm) of smaller and non-porous, readily available commercial anatase pan class="Chemical">TiO2 nanoparticles (c-TiO2, 20 nm in size), again mixed with MWNTs, was then sprayed on top of Layer 1. The rationale was to provide an inter-connected MWNT network decorated with the c-TiO2 nanoparticles that exploited both the redox reaction possibilities of TiO2 and the EDL capacitance of the inter-connected MWNT scaffold decorated with c-TiO2 nanoparticles (since MWNTs are well-known to provide EDL capacitance in the presence of H+ 17). A critical aspect was that for Layers 1 and 2, ion conducting ionomer was co-sprayed along with the TiO2 and MWNTs so that during in-situ drying at deposition, the TiO2 attached to the MWNT scaffold, and both were coated with ionomer12. The ionomer coating was intended to promote ion mobility throughout the electrode in the solid-state, while the inter-connected MWNT network provided high electrical conductivity.

Finally, Layer 3 (~200 nm in thickness) comprising low-defpan class="Chemical">ect few-layer graphene sheets (400 S cm−1 18) made by shear exfoliation of graphite in deionised water, was sprayed on top of Layer 2 with the intent to decrease the contact resistance between the electrode and the subsequently added current collector. Some graphene sheets moved into the porous MWNT scaffold to connect to the MWNTs, and the edge planes exposed by the graphene sheet also contributed some further EDL capacitance19202122. Cu current collectors were then pressed onto Layer 3 after the electrode was dried, so the layer of graphene sat at the interface of Layer 2 of the electrode and the current collector itself. To check for any Cu reaction or corrosion effects, indium tin oxide (ITO) coated glass current collectors in an identical arrangement were also studied, and the consistent electrochemical results between Cu and ITO coated glass current collectors confirmed that any underlying Cu reaction with the electrode or electrolyte could be neglected, and as previously shown17.

Spray processing

Fabrication of the symmetric solid-state supercapacitor with elpan class="Chemical">ectrodes each comprising of up to three discrete layers was realised by spray atomisation and deposition of up to three different suspensions in sequence, in a single operation.

The three types of aqueous-based suspensions of elpan class="Chemical">ectrode materials were prepared by sonication at 600 W and 20 kHz for 30 min. Fig. 1(b) shows the spray apparatus where multi-nozzles sprayed consecutively the three types of aqueous-based suspensions of electrode materials into the three layers of the electrode, onto an H+ion conducting Nafion membrane, maintained at 100 °C on a heated vacuum stage. The nozzles moved in a pre-programmed zig-zag pattern along X and Y directions at 20 mm s−1 to spray an area of up to 20 cm × 20 cm. Any fugitive water from the spray suspension evaporated continuously on the heated vacuum-chuck stage as the electrodes formed. The membrane with one side covered with the three-layer sprayed electrode was then flipped and the other side sprayed using the identical procedure, to directly form a solid-state supercapacitor with no need for any subsequent re-immersion in liquid electrolytes. No binders were needed for any of the layers.

To allow comparison with electrodes of pan class="Chemical">similar materials23, the average loading mass of the electrodes was determined as 1.06 mg cm−2 (with an error estimation of ±1.1% based on measurements of 50 samples), with coin cell test area of 1.13 cm2 and 1.2 cm diameter. Larger electrodes of 4.5 cm × 4 cm were also investigated to show scalability, and gave consistent areal capacitances to within ±2.7% variation of the smaller cells. Figs. 1(c) and S1 in the Supplementary Information (SI) show an even larger 16 cm × 9.5 cm solid-state supercapacitor fabricated by the same spray method. The spray process is capable, so far, of making 200 nm-70 μm thick and up to 1 m × 20 cm large area electrodes using a drum coater variant24.

Results and Discussion

Morphology, electrode structure and surface area characterisation

In order to quantify the benefits of the layering approach (Fig. 1(a)), various other layered and non-layered electrodes that contained the same materials but in different, but comparable arrangements were fabricated identically, with the same thickness to within a ±2.4% variation, and are shown in Tables 1 and 2.

Table 1

A summary of the types of electrode fabricated to the same ±2.4% thickness variation and their corresponding areal capacitance per electrode in a solid-state supercapacitor cell arrangement.

Electrode	Description	Capacitance (mF cm−2)
		5 mV s−1	100 mV s−1	500 mV s−1	2000 mV s−1
Ec	Randomly mixed c-TiO2 + MWNT	91.7	46.3	N/A	N/A
Ep	Randomly mixed p-TiO2 + MWNT	53.3	18.8	N/A	N/A
Epc	Randomly mixed p-TiO2 + c-TiO2 + MWNT	60.1	28.6	N/A	N/A
E2	Two-layer electrode (A layer of p-TiO2 + MWNT directly on the ion conducting membrane, and then a layer of c-TiO2 + MWNT)	247.6	65.9	11.5	3.5
E3	Three-layer electrode (A layer of p-TiO2 + MWNT directly on the ion conducting membrane, a layer of c-TiO2 + MWNT, and then a layer of graphene)	272.5	77.1	20.2	11.8

The mass ratio of TiO2 : MWNT was kept the same for all electrodes.

Table 2

A summary of the types of electrode fabricated to the same ±2.4% thickness variation and their corresponding gravimetric capacitance per electrode in a solid-state supercapacitor cell arrangement.

Electrode	Description	Capacitance (F g−1)
		5 mV s−1	100 mV s−1	500 mV s−1	2000 mV s−1
Ec	Randomly mixed c-TiO2 + MWNT	86.3	43.7	N/A	N/A
Ep	Randomly mixed p-TiO2 + MWNT	50.2	17.7	N/A	N/A
Epc	Randomly mixed p-TiO2 + c-TiO2 + MWNT	56.9	26.9	N/A	N/A
E2	Two-layer electrode (A layer of p-TiO2 + MWNT directly on the ion conducting membrane, and then a layer of c-TiO2 + MWNT)	237.4	62.7	10.9	3.3
E3	Three-layer electrode (A layer of p-TiO2 + MWNT directly on the ion conducting membrane, a layer of c-TiO2 + MWNT, and then a layer of graphene)	265.9	73.3	19.0	11.2

The mass ratio of TiO2 : MWNT was kept the same for all electrodes.

Fig. 2(a) is a scanning electron microscopy (SEM) image of the ~200 nm pan class="Chemical">p-TiO2 particles with well-defined 20–50 nm pores. Fig. S2 is an X-ray diffraction (XRD) pattern of the same p-TiO2, showing crystalline anatase. Fig. S3 is an SEM image of the top-view of the sprayed [p-TiO2+ MWNT] layer, showing the p-TiO2 crystals with some pores covered by ionomer, needed to transport H+ ions. In the p-TiO2-based Layer 1 of the electrode, a relatively low fraction of MWNTs (weight ratio p-TiO2 : MWNT = 12:1) was used to maximise the p-TiO2 fraction, while maintaining electrical conductivity. Fig. 2(b) is a top-view SEM image of the sprayed, thicker [c-TiO2 + MWNT] (Layer 2), where the c-TiO2 was ~20 nm in size. A larger fraction of MWNTs (weight ratio c-TiO2 : MWNT = 1:2) was used to promote EDL capacitance in Layer 2. Fig. 2(c) is an SEM image of the cross-sectional view of Layers 1 and 2 together, and Fig. 2(d) shows a magnified cross-section of Layer 1 only, showing a p-TiO2 crystal retained by a MWNT, and other crystals more deeply embedded.

Figure 2

Top-view SEM images of (a) p-TiO2 crystals in powder form and (b) the sprayed [c-TiO2 + MWNT] layer; cross-sectional SEM images of (c) the first [p-TiO2 + MWNT] layer and the second thicker [c-TiO2 + MWNT] layer together, and (d) magnified [p-TiO2 + MWNT] layer; TEM images of (e) sprayed graphene layer on a holey-carbon grid; (f) magnified top right corner of (e) where some graphene sheets overlapped the edge of the holey-carbon film; and (g) magnified straight graphene edges in (f).

The surface area of the p-TiO2 powder was measured by the Brunauer-Emmett-Teller (BET) method as 233 ± 1.2 m2 g−1 compared with 50 ± 1 m2 g−1 for the pan class="Chemical">c-TiO2 powder. Here, we were also able to measure the BET surface areas of the as-sprayed free-standing electrodes of [p-TiO2 + MWNT] and [c-TiO2 + MWNT]. Fig. S4 shows an image of a 2.8 cm × 2.3 cm free-standing electrode of [c-TiO2 + MWNT] peeled carefully from the substrate for the BET measurement. For a fair comparison, the electrode thickness was kept the same and the weight ratio between TiO2 and MWNTs in both films was kept the same at 1:2. The electrode specific surface areas were 143 m2 g−1 and 98 m2 g−1, the pore volumes were 0.44 cm3 g−1 and 0.26 cm3 g−1, and the average adsorption/desorption pore sizes were 12 nm and 9 nm for the [p-TiO2 + MWNT] and [c-TiO2 + MWNT] free-standing electrodes respectively. Therefore, p-TiO2 contributed more strongly towards an overall higher electrode surface area than c-TiO2.

To characterise the graphene sheets in Layer 3, Fig. S5 is a Raman sppan class="Chemical">ectrum of a drop of aqueous suspension of the exfoliated graphene sheets, showing a relatively weak narrow D band, a dominant narrow G band indicative of few defect graphene made by exfoliation, and a 2D band suggestive of 4 to 7 layers of graphene18. Holey-carbon transmission electron microscopy (TEM) grids were sprayed simultaneously with Layer 3, and Fig. 2(e) is a wide-field TEM image showing full coverage. The multi-layer graphene sheets (red circle) were typically ~360 nm × 200 nm, consistent with the literature18. The graphene sheets (green circle in the top right corner in Fig. 2(e)) were magnified in Fig. 2(f), showing straight edges with ordered fringes, which were further magnified in Fig. 2(g) and manual counting suggested 4–17 layers after spraying25.

To characterise together Layers 2 and 3 of the three-layer electrode (E3 in Table 1), X-ray photoelpan class="Chemical">ectron spectroscopy (XPS) depth profiling using Ar+ ions sputtering through the top Layer 3 was used to investigate chemical composition changes through Layer 3, down into Layer 2. Fig. S6 shows the atomic % profiles of C, O and Ti versus etch depth. As expected, the profiles showed that the C concentration decreased and the Ti concentration increased with increasing depth, supporting the intent of a multi-layer structure where the layers had different compositions. The C concentration was stable at ~200 nm, agreeing with the estimate of ~200 nm thick graphene Layer 3 measured by a stylus profilometer. Since the top graphene layer may not be entirely continuous and can be permeable to incident X-rays, some Ti signal was inevitably detected even before all of the graphene was etched away, but Figure S7 shows that as expected the Ti 2p peak intensity increased as etch depth further increased.

Electrochemical characterisation

Fig. 3(a) shows the cyclic voltammetry (CV) curves of the solid-state supercapacpan class="Chemical">itor using the three-layer electrodes E3 up to 500 mV s−1. The main pair of redox reaction peaks between 0.13 and 0.67 V shown in Fig. 3(a) was due to the reversible redox reaction between -OH functional groups of TiO2 and H+ ions from the H2SO4 treated Nafion membrane, and residual H2SO4 from the sprayed suspension, according to26272812:

Figure 3

(a) CV curves of a solid-state supercapacitor using three-layer electrodes E3 at scan rates from 5 to 500 mV s−1; (b) corresponding curves of current normalised by the square root of scan rate as a function of voltage for the same supercapacitor; (c) detailed XPS Ti2 spectra for pristine TiO2 nanoparticles; (d) detailed XPS Ti2 spectra for electrode E3 after one CV cycle at 5 mV s−1; (e) areal capacitance of electrode E3 as a function of maximum cell voltage at 5 mV s−1; and (f) current per unit weight using electrodes E3 as a function time at 1.5 V.

where X+ indicates the protons and/or alkali metal cations (e.g. Na+, Li+ and K+) in the electrolyte28.

Fig. 3(b) shows the same CV data with current normalised by the square root of scan rate as a function of voltage for the same elpan class="Chemical">ectrode, which can be used to better resolve low scan rate behaviour when a wide range of scan rates is used17. The small peak at ~1.3 V at 5 mV s−1 in Fig. 3(b) arose from the interaction of the mobile H+ ions (in the ionomer coating in the electrode) and the negatively charged F− species in the ionomer29. This peak became negligible at scan rates faster than 5 mV s−1, which represents a more typical range used for supercapacitors30.

Fig. 3(b) also shows a pair of redox reaction peaks between 0.13 and 0.67 V associated with the -OH functional group of TiO2 as shown in Eq. (1). To confirm the proposed redox reaction, Fig. 3(c,d) show the detailed XPS Ti2 sppan class="Chemical">ectra of the pristine TiO2 nanoparticles and the electrode E3 after one CV cycle at 5 mV s−1 respectively. Both spectra show peaks at 459 and 465 eV corresponding to Ti4+ (Ti2) and (Ti2) respectively31. The electrode after a single CV cycle showed a peak at 455 eV corresponding to Ti3+ 31, showing that preparation of the aqueous-based solution for spraying and/or the redox reaction during charge and discharge increased the hydroxyl group concentration on the TiO2 surface12. A similar change in oxidation state has also been shown by in-situ X-ray absorption spectroscopy (XAS) in titanium carbide in a supercapacitor electrode32. Additionally, Figure S8 shows the detailed XPS O1 spectrum for the same electrode after the CV cycle. The peaks at 532.5 and Fig. S8 534.0 eV corresponded to Ti-O-Ti and Ti-OH, respectively33, confirming the presence of -OH groups on Ti.

Fig. 3(e) shows the areal capacitance in relation to maximum cell voltage for the solid-state supercapacitor upan class="Chemical">sing the three-layer electrodes E3, which is another method to investigate the presence of any irreversible reactions within the cell34. The areal capacitance continued to increase almost linearly with increasing cell voltage from 0.3 to 1.5 V, indicating negligible irreversible reactions over this range3435. To confirm further, electrode current as a function of time was also monitored at a constant voltage of 1.5 V, and as shown in Fig. 3(f), the current density was small and relatively stable at ~0.1 A g−1 over 2 hrs, consistent with negligible H2 evolution or additional parasitic reactions32.

To study the benefits of the layered electrode, the elpan class="Chemical">ectrochemical properties of solid-state supercapacitor cells using the different types of electrode arrangement were compared, as summarised in Tables 1 and 2. All electrodes were of the same thickness ±2.4%. Fig. 4(a) shows the CV curves of the solid-state supercapacitors using [c-TiO2 + MWNT] electrodes Ec. The capacitance estimated from the CV curves for this [c-TiO2 + MWNT] electrode Ec was 91.7 mF cm−2 (86.3 F g−1) at 5 mV s−1 according to the detailed estimation method in the Methods section. Compared with other TiO2 and carbon nanotube hybrid electrodes from the literature, an electrode of porous TiO2 on carbon nanotubes coated on a carbon paper exhibited a capacitance of 145 F g−1 at 5 mV s−1 in a liquid 0.5 M H2SO4 electrolyte36, and a layer-by-layer (LbL) assembled 760 nm thin electrode of sub-8 nm positively charged TiO2 particles and negatively charged MWNTs exhibited a capacitance of 262 F g−1 at 1 mV s−1 in a liquid 1 M H2SO4 electrolyte37.

Figure 4

CV curves of a solid-state supercapacitor using (a) [c-TiO2 + MWNT] electrodes Ec; (b) [p-TiO2 + MWNT] electrodes Ep; and (c) randomly mixed [p-TiO2 + c-TiO2 + MWNT] electrodes Epc. All electrodes were of the same thickness to within ±2.4%. The mass ratio of TiO2 : MWNT was the same for all electrodes.

Here, the capacitance of the Ec elpan class="Chemical">ectrode arose due to (i) the wetted, intimate interface between the TiO2 nanoparticles and the electron conducting MWNT scaffold, and with the ion-conducting ionomer that coated them both12; and (ii) the supply of H+ ions from the dilute H2SO4 used for spraying and the redox reaction of the hydroxyl groups on TiO2 nanoparticles, shown by the XPS spectrum in Fig. S8. The mobility of H+ ions was provided by the negatively charged SO3− end groups in the ionomer coating. The mobile H+ ions were charge compensated by the relatively high electron mobility in the inter-connected MWNT that formed the electrode38, reflected by the relatively high current densities in Fig. 4(a)39.

Fig. 4(b) shows the CV curves of the solid-state supercapacpan class="Chemical">itors using [p-TiO2 + MWNT] electrodes (Ep). The mass ratio of TiO2 : MWNT was kept the same for both electrodes in Figs. 4(a,b). The CV curves exhibited more prominent redox reaction peaks than electrode Ec, showing that the porous p-TiO2 with a higher surface area contributed proportionally more redox reactions to charge storage than the non-porous c-TiO240. However, the absolute current densities for electrode Ep in Fig. 4(b) were lower than for electrode Ec in Fig. 4(a). Four point probe measurements of electrodes Ep and Ec gave electrical conductivities of 0.22 S cm−1 and 3.4 S cm−1 respectively, suggesting that even though the p-TiO2 had a higher specific surface area than c-TiO2, the nearly one order of magnitude larger p-TiO2 particles inhibited the electrical connectivity of the MWNT network41.

Fig. 4(c) shows the CV curves of a solid-state supercapacpan class="Chemical">itor using randomly mixed [p-TiO2 + c-TiO2 + MWNT] electrodes (Epc). Comparing the CV curves for the different electrodes shows that the three-layer electrode E3 in Fig. 3(a) exhibited more prominent redox peaks than the randomly mixed electrode Epc, despite comprising largely the same electrochemically active materials. To understand how electrode structure gave such distinct differences, Trasatti’s method42 for each electrode was used to deconvolute surface and diffusion-controlled contributions to capacitance. The method relies on the assumption that the surface and diffusion-controlled contributions are governed by different kinetics, and respond differently to increasing scan rates43. Fig. 5(a) shows 1/C (C = areal capacitance) for the electrode E3 as a function of the square root of the scan rate v in the 5–100 mV s−1 region, where both surface and diffusion-controlled contributions were significant43. The intercept of the linear region of the plot with the 1/C axis estimated the total capacitance possible for E3 at an infinitely slow rate as 333.3 mF cm−2. Then, to extract the surface-controlled contribution to capacitance, Fig. 5(b) shows C as a function of v−1/2. The intercept of the linear region of the plot with the C axis estimated the surface charge at a scan rate of infinity as 62.5 mF cm−2. In comparison, Figs. 5(c,d) estimated the total capacitance for Epc as 74.2 mF cm−2, the surface charge at a scan rate of infinity as 34.1 mF cm−2. The lower surface charge of Epc than E3 likely arose from a lower active surface area because the smaller c-TiO2 may have blocked some of the pores in the larger p-TiO2 (the size of the c-TiO2 was 20 nm and the pores in p-TiO2 was 20–50 nm). These negative synergistic interactions between the two types of TiO2 were avoided in the discrete layered electrode structure.

Figure 5

Trasatti’s method for the three-layer electrode E3: (a) inverse capacitance as a function of square root of scan rate; (b) capacitance as a function of inverse square root of scan rate; and, Trasatti’s method for the randomly mixed [p-TiO2 + c-TiO2 + MWNT] electrode Epc: (c) inverse capacitance as a function of square root of scan rate; and (d) capacitance as a function of inverse square root of scan rate.

Tables 1 and 2 summarise both the areal and gravimetric capacitance of the different electrodes in a solid-state supercapacpan class="Chemical">itor. Tables 1 and 2 suggest that if the entire electrode consisted of [p-TiO2 + MWNT] only, the specific areal and gravimetric capacitance of the electrode would be relatively low (53.3 mF cm−2 and 50.2 F g−1 at 5 mV s−1 for a ~33 μm thick electrode) because, as previously shown, the comparatively large p-TiO2 particles (200 nm) prevented the formation of a well-inter-connected MWNT network throughout the electrode. This is because the MWNTs after sonication and spraying were typically ~500 nm in length, while the p-TiO2 particle size was ~200 nm: when the particle and MWNT length scales were similar, the percolating MWNT network was more restricted, as shown by the electrical conductivity measurements. Consequently, a proportion of the p-TiO2 particles remained electrically isolated within the electrode and could not contribute to charge storage.

On the other hand if the same electrode conpan class="Chemical">sisted entirely of [c-TiO2 + MWNT] only, the specific areal and gravimetric capacitance of the electrode might be expected to increase (91.7 mF cm−2 and 86.3 F g−1 at 5 mV s−1 for the same electrode thickness) because these smaller TiO2 nanoparticles (20 nm) were more readily electrically connected into the MWNT network. However, while this may be the case, as the early CV curves showed, these non-porous c-TiO2 particles did not contribute as much pseudo-capacitance as the high surface area porous p-TiO240.

Instead, by placing the p-TiO2 at the interface between the H+ treated pan class="Chemical">Nafion membrane and the rest of the electrode only (Layer 1 in the schematic diagram in Fig. 1(a)), with a lower fraction of MWNTs, the higher redox reactivity of the p-TiO2 could be exploited, without undermining connectivity in the majority of the rest of the electrode. Then using the c-TiO2 nanoparticles and a higher fraction of MWNTs (Layer 2 in Fig. 1(a)) for the rest of the electrode ensured a high surface area, inter-connected MWNT network and a relatively high conductivity pathway from the p-TiO2 to the current collector. For both layers, the ionomer coating facilitated H+ movement and EDL contributions to capacitance. Critically, the high redox reactivity in Layer 1 and the generation of additional H+ ions to promote EDL capacitance combined to give an overall high capacitance of 247.6 mF cm−2 (237.4 F g−1) at 5 mV s−1 for the same electrode thickness. Capacitance was further increased after adding a third layer of graphene, as now described below.

Overall, the variations in capacitance measurements for each elpan class="Chemical">ectrode type in Tables 1 and 2 were in the range ±3–6%, and were thus significantly smaller than the differences in capacitance among the different electrodes. We note that if the differences in electrochemical behavior were controlled primarily by underlying additional reactions, the results between the five different electrode arrangements of the same materials would not be so marked since all contained the same various materials. The strong differences between the electrode arrangements show that the arrangement of materials and electrode structure are the dominant effects on electrochemical response.

To further assess the proposed positive synergistic effpan class="Chemical">ects in the layered arrangement and to estimate the utilisation of the active materials in the solid-state supercapacitor arrangement, the electrodes were also tested in a standard three-electrode configuration using a liquid 1 M H2SO4 electrolyte, Pt counter electrode and Ag/AgCl reference electrode44. Fig. 6(a) shows the CV curves of the [c-TiO2 + MWNT] electrode Ec with a parallelogram shape, indicating fast charge/discharge kinetics typical for supercapacitors using a liquid electrolyte. The estimated capacitance for the electrode Ec in a liquid electrolyte from Fig. 6(a) was 92.5 mF cm−2 (87.1 F g−1) at 5 mV s−1 and 64.1 mF cm−2 (60.4 F g−1) at 100 mV s−1. By comparison with capacitances of 91.7 mF cm−2 (86.3 F g−1) at 5 mV s−1 and 46.3 mF cm−2 (43.7 F g−1) at 100 mV s−1 for the same electrode Ec in the solid-state supercapacitor configuration, the average utilisation of active materials in the solid-state for the composite electrode Ec was estimated as 86%.

Figure 6

(a) CV curves of a [c-TiO2 + MWNT] electrode Ec and (b) CV curves of a three-layer electrode E3, both in a three-electrode configuration using Pt as the counter electrode, Ag/AgCl as the reference electrode and 1 M H2SO4 as the liquid electrolyte.

The three-layer electrode E3 was also tested using the same three-electrode configuration. Fig. 6(b) shows the corresponding CV curves for E3 with more obvious redox reaction peaks again confirming the better performance of the layered electrode, and also a parallelogram shape showing fast charge/discharge kinetics. The capacitance of E3 estimated from Fig. 6(b) was 277.8 mF cm−2 (270.8 F g−1) at 5 mV s−1 and 99.4 mF cm−2 (94.6 F g−1) at 100 mV s−1. Again by comparison with capacitances of 272.5 mF cm−2 (265.9 F g−1) at 5 mV s−1 and 77.1 mF cm−2 (73.3 F g−1) at 100 mV s−1 for the same electrode E3 in the solid-state configuration, the average utilisation of active materials in the solid-state was estimated similarly as 88%.

Fig. 3(a) also showed that the CV curve of the solid-state supercapacpan class="Chemical">itor using the three-layer electrodes E3 became more parallelogram-shaped at a relatively fast scan rate of 500 mV s−1, suggesting redox reactions no longer had time to occur27, leaving only residual EDL capacitive behaviour at the interface between the MWNTs, TiO2 and the ionomer coating. Fig. 7(a) shows the CV curves of the solid-state supercapacitor using electrodes E3 up to extremely fast scan rates of 2000 mV s−1, with the curves maintaining an approximate parallelogram shape. Similar fast charging behaviour has been shown in exfoliated graphene electrodes in a liquid electrolyte of 1-butyl-3-methyl-imidazolium tetrafluoroborate in acetonitrile45. In contrast, Fig. 7(b) shows a more distorted CV curve, even at 500 mV s−1, for the solid-state supercapacitor using the two-layer electrodes E2 (i.e. without the graphene layer, otherwise identical to E3, as summarised in Tables 1 and 2) indicative of restricted charge/discharge kinetics at fast scan rates for E246. The charging and discharging kinetics of electrode E3 were improved because some few-layer graphene sheets infiltrated into the MWNT scaffold during fabrication before the fugitive carrier (dilute H2SO4) evaporated completely. The few-layer graphene enhanced electrical connectivity of the MWNT scaffold and its connection to the current collector, as shown schematically in Fig. 8. The many few-layer graphene sheets also likely made a contribution to EDL capacitance since their exposed edge planes can provide up to an order of magnitude higher EDL capacitance than the basal planes provided by the graphene surface and MWNTs1920212247.

Figure 7

(a) CV curves of a solid-state supercapacitor using three-layer electrodes E3 from 500 to 2000 mV s−1; (b) CV curve of a solid-state supercapacitor using two-layer electrodes E2 at 500 mV s−1; (c) Capacitance of the electrodes E2 and E3 in relation to scan rate; galvanostatic charge/discharge curves of a solid-state supercapacitor using electrodes E3 at (d) 1 mA cm−2 and (e) 3 mA cm−2; and (f) Nyquist plot of a solid-state supercapacitor using electrodes E3 from 105 to 0.01 Hz.

Figure 8

Schematic diagram of the [c-TiO2 + MWNT] structure (a) before spraying graphene Layer 3; and (b) after spraying graphene Layer 3.

The dependence of capacitance on scan rate for both the three-layer electrode E3 and two-layer elpan class="Chemical">ectrode E2 is shown in Fig. 7(c). Capacitance was reduced but relatively stable at scan rates above 500 mV s−1 because there was no redox reaction energy storage contribution, only residual EDL capacitance. There was a 42% decrease in capacitance for E3 compared with a 70% decrease for E2 as the scan rate increased from 500 mV s−1 to 2000 mV s−1 (summarised in Tables 1 and 2), noting that a scan rate of 2000 mV s−1 is amongst the highest used for supercapacitors484950, showing that the three-layer electrode E3 with a graphene layer had a greater ability to maintain residual EDL capacitance than the two-layer electrode E2 without the graphene layer, again indicating that the few-layer graphene contributed to the connectivity of the MWNT scaffold, and to its efficient connection to the current collector.

For the three-layer electrode E3 in partipan class="Chemical">cular, there was a resolvable transition from redox active behaviour at slow scan rates below 500 mV s−1, dominated by p-TiO2, to EDL capacitive behaviour above 500 mV s−1, which was particularly enabled by the graphene. Conventional randomly mixtures of electrode materials integrate together the different storage contributions, masking the intrinsic behaviour of the constituent materials (typically metal oxides and C-based materials). Capacitance in these electrodes decreases comparatively quickly as 2000 mV s−1 is approached because the insulating effect of the metal oxides on electrical conductivity becomes exposed as their appreciable contribution to energy storage fades51. However, in E3 the discrete graphene Layer 3 was able to facilitate a distinct EDL component (through exposed edge planes and by efficient connections to the MWNT scaffold in Layer 2) even at the fastest scan rates. Similar effects have been reported in both hybrid nanoporous gold/MnO2 films of 100 nm thick, where MnO2 contributed redox-based pseudo-capacitance and nanoporous gold provided EDL capacitance13, and also in a pseudocapacitive MoN <100 nm thick film coated on a Ti substrate with a H4SiW12O40-H3PO4-poly(vinyl alcohol) (SiWA-H3PO4-PVA) solid-polymer electrolyte52. Here, we have shown similar effects but in much thicker electrodes comprising discrete layers.

To further assess the effect of the pan class="Chemical">graphene Layer 3 in the electrode, galvanostatic charge/discharge and electrochemical impedance spectroscopy (EIS) were used. Fig. 7(d) shows the galvanostatic charge/discharge curves of the solid-state supercapacitor using three-layer electrodes E3 at a current density of 1 mA cm−2, with the non-linear response again due to the redox reaction (which strictly is non-capacitive in nature) associated with TiO21226. Fig. 7(e) shows the performance of electrode E3 at a higher current density of 3 mA cm−2. The more linear response exposes again the underlying EDL behaviour of electrode E3 at high charge/discharge current densities, as previously discussed. The estimated capacitance per electrode for E3 from the linear part of the discharge curve was 246.2 mF cm−2 (231.8 F g−1) at 1 mA cm−2 and 64.6 mF cm−2 (60.8 F g−1) at 3 mA cm−2. The IR drop is related to the internal resistance i.e. the sum of the electrolyte ionic resistance, electrode resistance and interfacial resistance345. The IR drop in Fig. 7(e) was 0.036 V, lower than 0.08–0.2 V commonly reported for solid-state supercapacitors tested in similar conditions53.

Fig. 7(f) shows a Nyquist plot from a solid-state supercapacitor upan class="Chemical">sing the three-layer electrodes E3. The intersection point of the best-fit curve to the data with the real axis at high frequency represented the series resistance (R), which includes the contact resistance between the electrode material and current collector, and the resistance of the electrolyte54. R was estimated as 3.5 Ω for the three-layer electrode E3, lower than 28 Ω for the two-layer electrode E2 in Fig. S9, and also lower than R for other solid-state supercapacitors (e.g. 4 Ω for an electrode of MoS2 on C cloth with a LiCl-poly(vinyl alcohol) (PVA) gel electrolyte54), showing that as intended the graphene layer reduced the contact resistance between the electrode material and current collector.

The semi-circle diameter of the Nyquist plot at high frequency represented the charge transfer resistance (R) of the elpan class="Chemical">ectrode55, and was estimated at 8.5 Ω for the three-layer electrode E3, lower than 42 Ω for the two-layer electrode E2 in Fig. S9, and also lower than R for other solid-state supercapacitors (e.g. 10.7 Ω for an activated carbon-TiO2 hybrid electrode with a H3PO4-PVA gel electrolyte56). Therefore, adding a small amount of graphene as Layer 3 effectively reduced R (contact resistance of the interface between electrode and current collector) by 88%, and also reduced R (through-plane electrode resistance) by ~80%, supporting the idea that some of the sprayed few-layer graphene mixed into the meso-porous electrode structure immediately on deposition. Furthermore, R for electrode E3 without the current collector, measured by careful electrical connection directly to the graphene layer, was ~7 Ω, suggesting that if made more robust, the graphene layer could act directly as the current collector.

The volumetric energy and power densities of the three-layer electrode E3 were 23.3 mWh cm−3 and 380 mW cm−3 at 1 mA cm−2, and 6.1 mWh cm−3 and 1400 mW cm−3 at 3 mA cm−2, respectively. The energy and power densities compare favorably with the literature57585960 as shown in Fig. 9(a). For example, although a reduced graphene oxide (RGO)/polypyrrole (PPy) interdigital electrode using a PVA-H2SO4 gel electrolyte exhibited a higher power density of ~10000 mW cm−3 at a similar energy density of ~8 mWh cm−3 at a similar high current density, the electrode exhibited a lower maximum energy density of 13.2 mWh cm−3 compared with 23.3 mWh cm−3 for E3 at a low current density of ~1 mA cm−2 60.

Figure 9

(a) Volumetric Ragone plot; (b) gravimetric Ragone plot of the three-layer electrode E3 in a solid-state supercapacitor configuration, and comparison with similar electrodes from the literature; and (c) cycling performance of the solid-state supercapacitor using three-layer electrodes E3 at 100 mV s−1.

The gravimetric energy and power densities of the three-layer electrode E3 were 45.7 Wh kg−1 and 1.1 kW kg−1 at 1 mA cm−2, and 19.0 Wh kg−1 and 4.2 kW kg−1 at 3 mA cm−2, respectively. Fig. 9(b) shows the Ragone plot of gravimetric energy and power densities per electrode, with the three-layer electrode E3 again providing a competitive performance to other solid-state supercapacitor electrodes616263.

However, gravimetric energy and power densities per elpan class="Chemical">ectrode frequently do not give a realistic indication of the performance of an assembled cell, since full cells also contain current collectors, separators etc., and often volumetric normalisation is preferred23. In this study, the total volume of the solid-state supercapacitor cell including current collectors, electrodes and treated Nafion membrane was 0.0199 cm3. The volumetric energy and power densities of the cell were then estimated as 2.2 mWh cm−3 and 35.8 mW cm−3 at 1 mA cm−2, and 0.6 mWh cm−3 and 127.5 mW cm−3 at 3 mA cm−2, respectively. Again comparing with the literature, at a similar current density of 3 mA cm−2, MnO2-TiN nanotube hybrid arrays using a PVA-KOH-KI-ethylene glycol (EG) gel electrolyte exhibited 0.7 mWh cm−3 and 115 mW cm−3 64, comparable to our performance; and porous poly(3,4-ethylenedioxythiophene)(PEDOT) coated TiN nanotube array using a PVA-H2SO4-EG gel electrolyte exhibited 2.26 mWh cm−3 and 250 mW cm−3, due to its high surface area of the porous material65. However, this PEDOT-based nanopore array was fabricated through HF corrosion of TiN65, suggestive of scalability problems for nearer industrial-scale processing. In contrast, the performance of electrode E3 was delivered through a process that can be easily scaled and operated for a wide range of materials.

When cycling at 100 mV s−1, the electrode E3 maintained 90.2% capacitance after 10,000 cycles on the bench top (conpan class="Chemical">tinuously exposed to ambient air and moisture with no packaging), offering encouraging potential in, for example, future wearable electronic applications.

Conclusions

A symmetric solid-state supercapacitor upan class="Chemical">sing three-layer electrodes was fabricated to exploit optimally the inherent advantages of the active materials. Porous TiO2 placed at the interface between the electrode and the ion conducting membrane/separator contributed significant redox behaviour and charge storage behaviour, especially at slow scan rates. At the other side of electrode, at the interface with the current collector, exfoliated graphene reduced both contact and charge transfer resistance and increased EDL capacitance through exposed edge planes. The spray atomisation and deposition of the suspensions in layers onto the heated membrane facilitated rapid drying and the formation of a 3D inter-connected MWNT scaffold decorated with TiO2 nanoparticles, both of which were coated with the ion-conducting ionomer. The resulting intimate interface between the TiO2, MWNTs and ion-conducting ionomer helped realise both redox-based and EDL contributions to capacitance. This rational design of a three-layer electrode was intended to maximise the contribution of each type of active material to improve overall performance, and to use hard-to-obtain active materials most effectively. By careful comparison of different electrode structures comprised of identical materials, the three-layer arrangement was shown to be the optimal arrangement of these particular materials. This paper has given a specific demonstration of a more general idea that placing different materials at different positions in structured electrodes, rather than using random mixtures as widely practised, can improve performance. This approach may thus find applications in other energy storage and conversion devices such as Li-ion batteries and fuel cells.

Methods

Before spray, the pan class="Chemical">Nafion membrane was pre-treated by immersing in 1 M H2SO4 at 60 °C for 30 min-a standard procedure to exchange perfluorosulfonate (SO2F) groups along the main polymer chain for SO3−H+ groups66. Porous TiO2 single crystals were synthesised by: TiF4 was dissolved (20 mM to 400 mM) in water in a 125-ml-volume autoclave lined with Teflon to which 180 mM 1-methylimidazolium tetrafluoroborate was added. 650 mg silica template was added to 50 ml TiF4 solution. A transparent mixture was formed and kept at 120 °C for 10 hr in an oven. After that, the precipitate (TiO2 particles with the silica templates) was formed at the bottom of the Teflon reactor and was collected. The silica template was then selectively etched in 2 M aqueous NaOH at 80 °C for 60 min in a polypropylene beaker. The TiO2 product was collected by centrifugation (3,000 rpm for 60 min) and washed several times in water and ethanol. Graphene was synthesised by: an aqueous suspension of 5 mg ml−1 graphite powder with 0.1 mg ml−1 NaC was prepared in a sonication bath (Ultrawave U1250D, 200 W, 30–40 Hz) for 51 hr. The resultant suspension was centrifuged at 3,000 rpm for 150 min. The supernatant containing graphene was collected18.

For spraying, three aqueous suspenpan class="Chemical">sions were prepared by sonication at 600 W and 20 kHz for 30 min: (a) 1 mg ml−1 p-TiO2, 0.083 mg ml−1 MWNTs, 0.1 wt% sodium dodecylbenzenesulfonate (SDBS) and 50 wt% Nafion ionomer in 0.5 M H2SO4; (b) 2 mg ml−1 MWNTs, 1 mg ml−1 c-TiO2, 0.1 wt% SDBS and 50 wt% Nafion ionomer in 0.5 M H2SO4; and (c) 5 mg ml−1 graphene, 0.1 mg ml−1 NaC in deionised water. The stable suspensions were pumped at 3 ml min−1 into multiple nozzles, atomised using compressed air at 310 kPa and sprayed onto a Nafion membrane (70 μm) consecutively on a heated vacuum stage at 100 °C.

The weight of electrodes was measured by a microbalance (Sartorius) with 0.01 mg acpan class="Chemical">curacy and electrode thickness using a Dektak 6M stylus profilometer (Veeco Instruments Inc). The conductivity of the electrodes was measured by using a standard four-point probe configuration and a Keithley 220 programmable current source meter on the electrodes deposited on Si wafers, with measurements repeated eight times on each electrode. Porous TiO2 powders were characterised by XRD (Cu radiation, λ = 1.5 A). Exfoliated graphene was investigated using Raman spectroscopy (Horiba LabRAM ARAMIS) with a 532 nm wavelength laser. The BET (Micromeritics Gemini V) specific surface areas were measured by using N2 adsorption and desorption at 77 K. A full isotherm was recorded from relative pressures of 0.01 to 0.95 then back from 0.95 to 0.01. The BET surface area was calculated using the data between relative pressures of 0.05–0.3.

The surface chemistry was analysed by XPS in an ion pumped Thermo Scientific K-Alpha 128-channel detecpan class="Chemical">ting analyser equipped with an Al K X-ray source. The XPS analyser operated at a constant pass energy of 100 eV for wide scans and 20 eV for detailed scans. The etching of the samples for depth profile measurements was performed with Ar + sputtering at 1000 eV. The surface morphology of the electrodes was examined by SEM (JEOL 6500F at 10 kV), and high resolution TEM (JEOL 2100F at 200 kV). Electrochemical testing of solid-state supercapacitors was performed using a Reference 600/EIS300 Gamry potentiostat/galvanostat with a combination of CV, galvanostatic charge/discharge and EIS.

As a full cell can be treated as two capacitors in series, the capacitance of one cell C was calculated according to34:

in which C1 and C2 are the capacitances of individual electrodes. In this type of symmetric solid-state supercapacpan class="Chemical">itor, assuming two equal capacitors in series67, the capacitance of the electrode C is related to the capacitance of the cell C by:

Specific capacitance was calpan class="Chemical">culated from both the CVs and galvanostatic charge/discharge curves. For the CVs, the specific gravimetric capacitance was estimated by integrating the area under the current-potential curve and then dividing by the sweep rate, the mass of film electrode and the potential window according to68. For the calculation of specific gravimetric capacitance, the total electrode mass including TiO2 nanoparticles, MWNTs and ionomer gel was used, according to:

where C is the specific gravimetric capacitance (F g−1), m is the mass of one elpan class="Chemical">ectrode (g), v is the scan rate (V s−1), V − V represents the potential window (V), and I is either the charging or discharging current (A). For specific areal capacitance, the area of one electrode was used for the estimation.

In the galvanostatic charge/discharge process, the specific gravimetric capacitance was estimated from the slope of the linear part of the discharge curve according to Eq. 5, where the discharge current I is normally used and t is the corresponding discharge time (s) from a voltage V 69:

For estimating specific areal capacitance from the galvanostatic charge/discharge process, the area of one electrode was used.

The volumetric and gravimetric energy densities E and power denpan class="Chemical">sities P per electrode were estimated from69:

where x is the volume and mass of one electrode, and t is the discharge time. The volumetric energy density E and power density P per supercapacitor cell were also estimated from:

where v is the volume of one supercapacitor cell including the current collectors, electrodes and Nafion membrane.

Additional Information

How to cite this article: Huang, C. et al. Solid-state supercapacitors with rationally depan class="Chemical">signed heterogeneous electrodes fabricated by large area spray processing for wearable energy storage applications. Sci. Rep. 6, 25684; doi: 10.1038/srep25684 (2016).