Kenneth G. Lathama,
Aditya Rawalb,
James M. Hookb and
Scott W. Donne*a
aDiscipline of Chemistry, University of Newcastle, Callaghan, NSW 2308, Australia. E-mail: scott.donne@newcastle.edu.au; Fax: +61 2 4921 5472; Tel: +61 2 4921 5477
bNMR Facility, Mark Wainwright Analytical Centre, University of New South Wales, Sydney, NSW 2052, Australia
First published on 22nd January 2016
The incorporation of nitrogen into hydrothermal carbon with (NH4)2SO4 is shown to have a significant influence on its chemical composition and surface characteristics. This in turn boosts the pseudo-capacitive behavior of hydrothermal carbons and their overall electrochemical stability. A combination of X-ray photoelectron spectroscopy, Fourier transform infra-red spectroscopy (FTIR) and scanning electron microscopy (SEM), yielded insights on the influence of nitrogen doping on surface functionalities. 1- and 2-D solid state NMR established the molecular-level structure of both doped and non-doped hydrothermal carbon. Cyclic voltammetry and electrochemical impedance spectroscopy has established the electrochemical behaviour of these hydrothermal carbons, indicating that nitrogen doping enhances not only the capacitance but also the stability of the hydrothermal carbons.
Although pure carbon materials have been readily adopted, in recent years the incorporation of heteroatoms into the carbon matrix has been shown to increase performance in a number of traditional applications (i.e., supercapacitors).7 This has led to the widespread development of doped carbon materials incorporating nitrogen, phosphorus, sulphur and various transition metals into almost every allotrope of carbon.8–14 In terms of sheer number of investigations, nitrogen has become one of the most widely examined dopants, with nitrogen-doped carbon finding use as a pH-responsive adsorbent, fuel cell electrode material, CO2 absorber, as well as a material with extremely high conductivity.11,15–17
One specific application of doped carbon is in electrical double-layer capacitors, often called supercapacitors, where the application of doped carbon electrodes utilizing oxygen and nitrogen has been shown to significantly increase capacitance.7,18,19 These electrochemically active heteroatoms increase the overall capacitance by providing sites in which highly reversible redox reactions can occur (pseudo-capacitance).7 These sites, typically in the form of functional groups attached to the carbon surface; i.e., –C–O– or C–NH2, are not always beneficial, with acidic surface functionalities (carboxylic, lactonic and phenolic) being inclined to decrease the electrochemical stability of the carbon electrode.1 Thus selectivity is required when producing functionalized carbon to utilize the benefit of pseudo-capacitance for supercapacitors.
Currently, a number of approaches have been developed for the purpose of selectively doping nitrogen into carbon structures, such as electric arc, pyrolysis, hydrothermal and pre- or post-treatments with nitrogen sources.20,21 Of these, hydrothermal carbonization (HTC) is one of the most promising techniques in carbon materials research. In comparison to other doping methods, HTC offers an environmentally friendly approach to the formation of highly functionalized, doped carbon structures from renewable sources, as carbonization occurs in water at substantially lower temperatures (180–300 °C) and autogeneous pressures in closed systems.22,23 As a result, greenhouse gas production is significantly reduced and wet untreated biomass can be utilized, representing a significant energy saving as pre-treatment drying is not required. Furthermore, chemical and structural morphology, as well as surface functionality, can be easily controlled through the manipulation of the treatment conditions and solution pH.24,25 Finally, nitrogen doping of hydrothermal carbons is easily achieved through the addition of soluble nitrogen compounds to the hydrothermal solution prior to carbonization.15,24
While HTC has been utilized on a wide range of feedstock materials, including glucose and cellulose, as well as increasingly more chemically complex feedstock materials (e.g., wood, food and animal waste),4,26,27 sucrose has received minimal attention despite its ability to carbonize under hydrothermal conditions at a rate faster than any other feedstock.28 Additionally, sucrose is relatively pure, widely available, inexpensive and a renewable feedstock, giving it a distinct advantage over biomass in the production of carbon for applications that are susceptible to the variability in functionalization from different feedstocks, such as supercapacitors.29
At this time, significant insights from a number of studies have been made into the structure of hydrothermal carbon.30,31 These studies employed 13C-enriched glucose coupled with advanced solid state NMR techniques to elucidate the underlying carbon structure formed from HTC. They found that hydrothermal carbon produced under standard conditions (180 °C, neutral pH, 4.5 hours) resulted in the formation of a five membered furan-based carbon structure, remarkably different to the six-membered arene structure formed under pyrolysis.32 The predominately furan structure of hydrothermal carbon was found to be formed via furan–furan linkages on the α-carbon, as well as through sp2 or sp3 type carbon groups to produce a highly cross-linked carbon structure.30 Previous studies have examined briefly the effect of nitrogen doping on HTC via 13C solid state NMR, demonstrating that there is a significant increase in aromatic domains in the form of pyrazine and pyrrole structures dispersed within the furan network.31
Although these studies are an important foundation for understanding hydrothermal carbonization, there is still the need to discover specifically how nitrogen is incorporated into the carbon structure of hydrothermal carbon from sucrose. Additionally, nitrogen doping of HTC has been obtained with glycine as the nitrogen source, which prompted the question as to whether the more readily available ammonium salts could also be used in this context.
In this work, we report the investigation into the chemical structure and morphology of hydrothermally carbonized sucrose, with and without the addition of ammonium sulfate, through the use of 13C and 15N solid state NMR spectroscopy, X-ray photoelectron spectroscopy, scanning electron microscopy, as well as FTIR spectroscopy. Additionally, as hydrothermal carbon tends to have very low surfaces areas (10–20 m2 g−1) but are rich in surface functional groups, any capacitance produced from the materials under electrochemical cycling will be predominately pseudo-capacitive in nature. As a result, cyclic voltammetry and electrochemical impedance spectroscopy have also been employed to probe this effect.
350 mL of the 0.2 M sucrose and 0.2 M sucrose + 0.2 M ammonium sulfate solutions were reacted in a 410 mL Teflon-lined stainless steel bomb reactor in an oven at 200 °C for 4 hours. The 15N containing solution was reacted in a 23 mL Teflon-lined stainless steel bomb reactor in an oven at 200 °C for 4 hours. After this time, the reactor was left to cool naturally in the oven before filtering the resultant suspension through a 0.45 μm filter to collect the solid product.
To remove any residual compounds physisorbed on the surface, the hydrothermal carbon was subjected to Soxhlet extraction with acetone as a solvent at a rate of one cycle per 30 minutes for 24 hours. The remaining solid product obtained after washing with acetone was denoted HC-S from the pure sucrose feedstock, HC-SAS from sucrose plus (NH4)2SO4, and HC-15N from sucrose plus enriched (15NH4)2SO4.
The surface area of each material was determined using a Micromeritics ASAP 2020 Surface Area and Porosity Analyzer. Samples were degassed under vacuum at 200 °C for 24 h prior to analysis. An adsorption isotherm was then measured over the partial pressure (P/Po) range 0.05–0.30 using nitrogen as the adsorbate at 77 K from which the specific surface area was determined using the linearized BET (Brunauer–Emmett–Teller) isotherm.
For solid state NMR analysis, the samples were packed into 4 mm zirconia rotors with Kel-F caps and spun up to 12 kHz MAS in a 4 mm HX double resonance MAS NMR probehead. The 13C{1H} and 15N{1H} NMR experiments were carried out on a Bruker AVANCE III 300 spectrometer, with a 7 Tesla wide-bore super conducting magnet, operating at frequencies of 300 MHz, 75 MHz and 30 MHz for the 1H, 13C and 15N nuclei, respectively. The 90° pulse lengths for the 1H, 13C, and 15N nuclei were set to 3.2 μs, 3.5 μs and 5.4 μs, respectively. In all cases, 1H decoupling was achieved using SPINAL64 decoupling with a field strength of 72 kHz. The quantitative 13C CP-MAS NMR spectra of materials was acquired at 12 kHz MAS using the MultiCP technique33 with a spin echo prior to acquisition to suppress baseline distortion. The quantitative 15N NMR spectrum was acquired by direct polarization (DP-MAS) NMR with recycle delays of 300 s. Dipolar dephasing sequences were used to identify the non-protonated 13C and 15N species, with 40 μs and 90 μs of gated 1H decoupling periods respectively for each nucleus prior to signal acquisition. Additional spectral editing of the 13C and 15N NMR spectra was achieved with chemical shift anisotropy (CSA) recoupling34 with CSA recoupling times of 47 μs and 40 μs for 13C and 15N, respectively. Two dimensional (2D) heteronuclear 1H–13C and 1H–15N correlation (HETCOR) spectra were acquired at 10 kHz MAS with frequency-switched Lee-Goldburg (FSLG) 1H–1H homonuclear decoupling at a field strength of 80 kHz. 1H to 13C and 1H to 15N polarization transfers were achieved using Hartman-Hahn cross-polarization contact times of 0.2–2 ms – the 0.2 ms contact time probes two to three 1H–13C bond distances while this shorter contact time correlates to one bond 1H–15N distances. Detection of species with greater than one 1H–15N bond distance from the 15N, employed a 2 ms contact time.
Resistivity measurements were conducted on each material by compressing the hydrothermal carbon between two conductive plates in an insulating ceramic-lined stainless steel die and measuring the resistance with respect to compaction pressure. The resistivity of the material (ρ; Ω cm) was calculated using:
![]() | (1) |
For the EIS experiments, the hydrothermal carbon working electrode was initially held at −0.3 V and allowed to equilibrate for 5 minutes. After this time period had elapsed, an EIS experiment was conducted using a 10 mV RMS ac excitation signal over the frequency range from 20 kHz down to 0.1 Hz. When this EIS experiment had been completed the potential was then stepped cathodically 25 mV, where the electrode was again allowed to equilibrate for another 5 min. It was noted that the working electrode current had effectively returned to ∼0 mA after this time. At this point another EIS experiment was completed. This process was repeated until −0.9 V was reached. To ascertain the capacitance and quantitatively compare the behaviour of each hydrothermal carbon, an equivalent circuit was fitted to the data using complex non-linear least squares (CNLS) regression. The equivalent circuit used here is shown in Fig. 1. The equivalent circuit consists of (i) a series resistance (Rs) which takes into account the resistivity of the electrode and electrolyte, which is commonly attributed to the high frequency end of the semicircle; (ii) a charge transfer resistance (Rct) reflecting the faradaic processes the hydrothermal carbon undergoes at the electrode–electrolyte interface; and (iii) two constant phase elements, the first (CPEdl) in parallel with Rct represents the double layer capacitance at the electrode–electrolyte interface, while the second (CPEp) represents the mass transport features of the system. Each constant phase element is defined by two parameters, σ and m, with the impedance given by:
![]() | (2) |
![]() | (3) |
![]() | ||
Fig. 1 Equivalent circuit used for fitting the electrochemical impedance spectroscopy (EIS) of both hydrochars, where Rs and Rct are resistors and CPEdl and CPEd are constant phase elements. |
Since hydrothermal carbons can be considered to have rough, porous electrode–electrolyte interface, where m < 1, ωmax equals the angular frequency when −Z′′ is at a maximum.
The XPS survey scans of both samples identified the elemental composition (at%) of the surface to be 81.38% carbon and 18.61% oxygen for HC-S (Table 1). On the other hand, HC-SAS had an identical level of carbon (81.20%) but decreased oxygen content (12.77%) due to the incorporation of nitrogen (5.37%), as well as a small quantity of sulphur from SO42− (0.66%). To further quantify the surface functionality of both hydrothermal carbons, high resolution region scans were taken of the C1s, O1s and N1s peaks and synthetic peaks fitted under each scan for analysis.
Functionality (BE; eV) | HC-S | HC-SAS | |
---|---|---|---|
Carbon | C![]() |
34.02 | 38.09 |
C–C, C–Hx (284.4) | 28.80 | 30.62 | |
C–O– (286.2–288.0) | 9.62 | 5.96 | |
C![]() |
7.07 | 4.98 | |
COO− (288.0–289.2) | 1.61 | 1.33 | |
CO3 (289.0–291.6) | 0.28 | 0.39 | |
C–N | — | 5.31 | |
Oxygen | O![]() |
7.45 | 6.48 |
–O–C (532.9) | 11.13 | 6.18 | |
Nitrogen | Quaternary nitrogen (401.6) | — | 0.51 |
Pyrrolic, pyridonic and lactam (400.2) | — | 1.44 | |
Amine, amide and nitrile (399.3) | — | 1.95 | |
Pyridinic and imines (398.4) | — | 1.42 |
Elemental analysis | ||||||||
---|---|---|---|---|---|---|---|---|
XPS total | wt% | at% | at% (–H) | |||||
HC-S | HC-SAS | HC-S | HC-SAS | HC-S | HC-SAS | HC-S | HC-SAS | |
Carbon | 81.40 | 81.37 | 65.95 | 63.51 | 46.32 | 44.99 | 74.86 | 72.26 |
Oxygen | 18.58 | 12.66 | 29.50 | 25.43 | 15.56 | 13.52 | 24.44 | 25.92 |
Nitrogen | — | 5.31 | — | 5.86 | — | 3.56 | — | 5.72 |
Hydrogen | — | — | 4.55 | 4.47 | 38.12 | 37.73 | — | — |
For the C1s peak (Fig. 2), six peaks were fitted based on literature values, while keeping the FWHM consistent. These peaks represented CC (284.5 ± 0.5 eV), C–C, C–Hx (285.1 ± 0.1 eV), C–O– (285.8 ± 0.3 eV), C
O (287.5 ± 0.3 eV), COO− (289.2 ± 0.3 eV) and CO3− (290.6 ± 0.4 eV), although an additional peak was fitted in HC-SAS (Fig. 2(b)) for C–N (286.9 ± 0.6 eV).12,35,36 For HC-S, two main peaks representing C
C and C–C/C–Hx, and four smaller peaks consisting of the main carbon/oxygen functionalities were present. These smaller peaks decrease in the order C–O–, C
O, COO− and CO3−, revealing that the oxygen groups on the surface consisted mainly of C–O– and C
O, with a relatively small proportion of carboxylic acid and carbonate groups. In contrast, HC-SAS has a slightly higher level of C
C bonding, while having similar levels of C–C/C–Hx. The oxygen functionalities on HC-SAS followed a similar pattern to HC-S, although their relative quantities have decreased. Overall, the peaks under the C1s XPS spectrum were confirmed by the FTIR data, with peaks at 3100–3700 (O–H and N–H stretching), 2971, 2923, 856, 798 and 757 (C–H stretching and bending), 1600 (C
C stretching), 1510 (C
C stretching), 1360 (C–O stretching), 1205 (C–O–Ar stretching), and 1090 (C
C–H bending) cm−1 in both samples (Fig. 3).
To evaluate the decrease in carbon/oxygen functionalities by increase in nitrogen, a C–N peak was fitted under the C1s spectra of HC-SAS utilizing a few assumptions. Firstly, it was assumed that all nitrogen groups were bound to carbon, as N–O and NO groups were not present in the O1s and N1s XPS, FTIR and NMR spectra (see later) of HC-SAS. Secondly, all nitrogen groups were chemically bound, as physisorbed compounds would have been removed through post treatment washing, and finally, the C–N peak was limited to a maximum area of 5.37%. As a result, the fitted C–N peak indicated a loss in single bonded C–O groups, suggesting that nitrogen is incorporated where these functionalities would normally be located, which is similar to previous findings on N-doped hydrothermal carbon and consistent with the loss of furan groups to pyrrole under nitrogen doping.31
The O1s XPS spectra for both HC-S and HC-SAS (Fig. 2) consisted of a broad peak with no clear indication of components. As a result, the generally accepted deconvolution scheme consisting of two components were fitted corresponding to the single bonded (–O–C) and double (OC) bonded oxygen groups.35,37 For HC-S, the fitted peaks indicate that the surface consists predominately of single bonded oxygen groups, whereas HC-SAS contained identical peaks due to the decrease in C–O– groups and an increase in double bonded oxygen incorporated into SO42−. These fitted O1s peaks are in good agreement with the C1s data.
For nitrogen, the broad N1s XPS peak for HC-SAS was analyzed by fitting with five peaks representing the major functionalities located in the N1s region38–40 (Fig. 4). These are imines and pyridinic groups (398.4 eV), amine, amide and nitrile groups (399.3 eV), pyrrolic, lactam and pyridonic groups (400.2 eV), quaternary nitrogen (401.6 eV) and NOx (405 eV). The lack of a peak at 405 eV for NOx (Fig. 4) indicates that all nitrogen atoms at the surface are bound to carbon. The remaining synthetic peaks show that the surface consists of predominately amine/amide groups, followed by equal ratios of five- and six-membered carbon/nitrogen functionalities, and lastly quarternary nitrogen. In the corresponding FTIR data (Fig. 3), there was a broadening towards lower wavenumbers of the peak at 1600 cm−1, due to presence of a CN stretching peak. Additionally, peaks at 1420 and 1090 cm−1 are from C–N–C bonding stretches consistent with ring and terminal nitrogen groups such as in pyrrole and amides.41
Overall, the XPS and FTIR analysis indicate a wide range of functional groups are present on the surface of both HC-S and HC-SAS. In comparing the results here to previously reported hydrothermal carbons,24 the effect of the acetone wash was not apparent in the FTIR data, although a decrease in the total nitrogen and oxygen content on the surface was observed. Examining this further revealed that HC-SAS had the highest percentage loss of oxygen, although the nitrogen content was the least affected, suggesting that surface bound nitrogen groups are inherently chemical bound to carbon surface. In fact this could suggest that nitrogen bonding to carbon under hydrothermal conditions is ideally more favourable than its oxygen counterpart. A final point to make is that with the addition of (NH4)2SO4, carbonization increases, specifically CC bonding.
The SEM images of both hydrothermal carbons display the typical spherical morphology found when polysaccharides are hydrothermally carbonized (Fig. 5).42,43 Interestingly, HC-S displays two very distinct sphere sizes, ∼1 μm and 4–6 μm, suggesting that two different reaction pathways occur for sucrose. The appearance of these smaller spheres was hinted at in a previously published paper,24 although in our work it appears that the acetone wash has effectively separated the two vastly different sphere sizes. In contrast, HC-SAS tends to display a narrower range of spherical sizes from 2–3 μm. The differences in sphere sizes could possibly account for the difference in the specific surface areas of the two hydrothermal carbons; i.e., 123.7 m2 g−1 for HC-S compared to 94.1 m2 g−1 for HC-SAS.
![]() | ||
Fig. 5 Scanning electron microscope (SEM) images of (a) HC-S and (b) HC-SAS at 10![]() |
The reason why two sizes occur in HC-S, but only one size with the inclusion of nitrogen in HC-SAS is currently unclear, although it may be as a result of the differing reaction rates of the monomers of sucrose (glucose and fructose), under hydrothermal carbonization.42,44 What is clear is that the reaction mechanism for nitrogen-doped hydrothermal carbon produces a material with a vastly different morphology to its non-doped counterpart.
For HC-SAS it was observed that the oxygen content was diminished in line with the increase in nitrogen, a result previously indicated in the XPS, and providing additional confirmation that the incorporation of nitrogen into the hydrothermal structure forms in-place of oxygen groups. Comparing the total nitrogen uptake to other N-doped hydrothermal carbons from the literature, the ratio between total nitrogen in the initial solution and nitrogen incorporated into the hydrothermal carbon is slightly lower than other studies,7,16,23,25,31,45 although most of these studies have utilized nitrogen containing precursors (e.g., chitosan, glycine). At this stage it is unknown whether this is a factor of saccharide, longer reaction times or the nitrogen dopant as previous studies tended to examine glucose at longer reaction times (>5.5 hours).15,31,45 Nevertheless, the elemental analysis indicates that reasonable levels of nitrogen doping can occur with readily accessible, cheap ammonium salts under hydrothermal carbonization.
Fig. 6(a) and (d) display the quantitative 13C NMR of HC-S and HC-SAS, respectively. The spectra in both cases are broad with overlapping peaks, which illustrates the complex and disordered molecular structure of the hydrothermal chars. The main features in both spectra have been alphabetically labelled and assigned to molecular structures shown in Fig. 7.
The assignment of the different spectral features is elucidated through additional spectrally edited NMR experiments described in the following paragraphs. The primary differences in the spectra are in the enhanced intensities for the aromatic (site labelled f) and NCH (site labelled m) carbon species in HC-SAS as compared to HC-S material. Concomitantly there is a reduced signal intensity at alkyl (sites labelled a–c), furanic (site labelled i) and ketonic (site labelled j) carbon species in the HC-SAS as compared to the HC-S material. The quantitative 13C NMR spectra of HC-S (Fig. 6(a)) is very similar to that previously reported for hydrothermal carbon synthesized from glucose,30 suggesting the formation of similar hydrothermal carbon structures from the use of either glucose or sucrose. This similarity enables us to assign the main spectral features of Fig. 6(a) as belonging to furan ring structures, indicated by the strong intensity of NMR signals at 150 ppm, 120 ppm and 110 ppm (i, f and e, respectively), along with interconnecting linkages between furan/terminal groups at 25 and 44 ppm (b and c). Additional peaks a and j are assigned to alkyl, ketone and acidic functionalities, respectively, which form part of the linker groups between the ring structures.
The broad linewidth of the NMR spectra presents a fundamental challenge to the elucidation of all the possible functional moieties present. However, distinguishing the spectral signatures of protonated carbon species from the non-protonated/methyl carbon species was facilitated by gated 1H decoupling, (Fig. 6(b) and (c), respectively), for the HC-S material. A strong intensity for protonated carbons is detected in the regions 100–155 ppm (peaks labelled d, f and g) and between 6–60 ppm (peaks labelled a, b and c). The CSA suppression scheme implemented on top of the C–H selection (narrow lines in Fig. 6(b)) suppresses the peaks d, f and g which must therefore have a large CSA, confirming that they are all sp2 hybridized carbon species, as expected. This is an important test to confirm that all the sucrose has indeed transformed into the hydrothermal carbon. Briefly, the anomeric carbon in sucrose has a chemical shift of ∼105 ppm which overlaps with the spectral range of aromatic or sp2 hybridized carbons. Thus the peak labelled d could potentially be a residual fragment from the sucrose molecule. However, since the anomeric carbon is sp3 hybridized and has a smaller CSA, if d were indeed the anomeric carbon, it would have survived the CSA filter, which is not the case. On the other hand, peaks at a, b and c, in the aliphatic region (which have a smaller CSA than the sp2 carbons) are only partially suppressed by the CSA suppression scheme. The strong peak intensity at 130 ppm, labelled f in Fig. 6(b), confirms the presence of significant amounts of aromatic carbon species. While the presence of low concentrations of aromatic carbon species has previously been identified in hydrothermal carbons,30 the peak intensities in the current spectra indicate that they form a significant component of the product. The weak signal detected at 150–155 ppm, labelled g in Fig. 6(b), cannot be from either the protonated furanic carbon (chemical shifts of ∼140 and ∼105 ppm) or protonated aromatic carbon (chemical shift of ∼130 ppm) species, nor can it be from the ipso phenolic carbon which is non-protonated. A protonated alkene, bonded to a quaternary aliphatic carbon is the only candidate which is assignable to site g. Evidence for the presence of the quaternary aliphatic carbon is indeed seen in quaternary/methyl carbon selection in Fig. 6(c). Here the signal at 30 ppm (labelled site h) is distinguished from overlapping signals of the methyl protons (labelled site a) by the CSA suppression (narrow line Fig. 6(c)). The quaternary carbon has a larger CSA than the methyl carbon and is therefore preferentially suppressed. Additionally, absence of signal intensity at ∼105 ppm from the CSA suppressed spectra (Fig. 6(c) thin line) confirms the absence of any anomeric carbon species and indicates the total conversion of the primary sugar structure into the condensed carbon. Overall, the HC-S carbon is composed of furan and aromatic ring structures with significant aliphatic functionality and a branched structure as seen from the presence of quaternary carbon species.
The spectrally edited 13C NMR of the protonated and non-protonated/methyl carbon species in HC-SAS (Fig. 6(c) and (f)) shows significant differences from that of the HC-S. In the first instance, there is a greater concentration of the protonated aromatic carbon species relative to the protonated aliphatic carbon species. In addition to the NMR peaks seen for the HC-S material, the presence of nitrogen results in multiple additional peaks at sites labelled k, l, m, n, o, p, q and r in the HC-SAS, which overlap with the peaks seen in the HS-C material. This indirectly indicates that nitrogen is incorporated into many distinct sites within the carbon structure. The assignment of these structural carbon sites was carried out in combination with the 15N NMR results and is discussed below.
Total conversion of the sucrose to HC-SAS was deduced from the absence of any anomeric carbon signal at ∼100 ppm, similar to that found for HC-S. The CSA suppression in Fig. 6(f) (narrow line) does not suppress any of the peaks between 10–35 ppm, indicating that these are methyl species and that there are no quaternary aliphatic carbon species present in HC-SAS. The signal at 108 ppm (labelled d) was assigned to CH on the basis of furan rings in HC-S, as well as the HC-SAS material. However, a signal for a non-protonated sp2 carbon is also observed in Fig. 6(f) at 108 ppm (site labelled n) which overlaps with site d and is absent/reduced in intensity in the spectrum of the HC-S material. To account for this signal, an alternative bonding arrangement not present in HC-S must be considered, wherein the bottom of a furan ring is bonded to two additional five-membered rings, forming a six-membered aromatic linkage (sites labelled o, p, n), an arrangement that has not been reported previously. In addition to the presence of the nitrogen functionality, HC-SAS is seen to be distinct from the HC-S in that the structure is considerably more aromatic and condensed with reduced branching/functionalization of the ring structures.
Based on the assignments of the protonated nitrogen species, several sites in the carbon spectrum can be assigned. In Fig. 6(d), site p is assigned to the carbon attached to the protonated imdazolium nitrogen, sites q and r are assigned to the protonated carbon and non-protonated carbon next to a pyridinic nitrogen, respectively.
Spectra of the non-protonated nitrogen species are shown in Fig. 8(d) and contain the expected peak from the pyridinic nitrogens labelled “6”. Additionally, it also shows significant intensity of non-protonated nitrogens that overlap with those of the protonated nitrogens. The site at 280 ppm (labelled “5”) can be assigned to the non-protonated nitrogen in the imidazolium ring. The site between 190–220 ppm (labelled “4”) has the same chemical shift as 4, but a smaller CSA of approximately 6 kHz, and it is potentially a site similar to 4 but connected to an alkyl group instead. The signals between 90–190 ppm all have a similar CSA of approximately 3.5 kHz. Consequently, these sites are assigned to non-protonated anologues of the sites 3, 2 and 1. Additionally, the carbon signal at m in Fig. 6(d), is unlikely to be due to an OCH group since it would be expected to be present in the HC-S material as well. Instead, if the peak at site 2 in the non-protonated 15N NMR spectrum also has contributions from proline-like amides, then it is possible to assign m to the Cα-like carbon site in proline with peaks in the 60–70 ppm region. This indicates that the formation of the unsaturated pyrroles may be preceded by that of saturated pyrrolidine rings.
In comparison to the HC-S, the 1H–13C HETCOR spectrum of HC-SAS (Fig. 9(b)) shows a very different pattern of connectivity. Under identical acquisition conditions to those used to acquire data for HC-S, the alkyl carbon species of HC-SAS show no/weak correlation peaks to the aromatic/pyrrolic/pyridinic protons, while the corresponding carbons show a relatively weak cross peak to the alkyl protons. Similarly, the ketone carbon only shows a cross-peak to the aliphatic protons, indicating that the ketones do not functionalize the aromatic structures to a significant degree. Thus the 1H–13C HETCOR spectra indicate that HC-SAS is formed with a relatively high proportion of condensed aromatic/pyrrolic/pyridinic structures accompanied by a much lower degree of branching and functionalization as compared to HC-S.
2D 1H–15N HETCOR spectroscopy provides an additional dimension of structural information about HC-SAS. With cross-polarization contact times of 0.2 ms (Fig. 10(a)) the correlation spectrum of HC-SAS shows only the protonated nitrogens and their directly bonded protons. Thus no correlation peaks are observed for the non-protonated pyridinic nitrogens, while the protonated imidazole, imide, pyrrole and amide nitrogens (site labelled 4, 3, 2, and 1) show correlation peaks to their bound protons. The dominant signal of the pyrrole nitrogen (∼150 ppm) has the strongest intensity correlations to the 1H species at ∼10 ppm. The relatively high 1H chemical shift is consistent with a hydrogen-bonded N–H proton. In comparison, the imidazole N–H proton has a chemical shift of ∼15 ppm, indicating a very strong hydrogen bonding interaction. There is a possibility that the hydrogen bonding interactions may be occurring among spatially proximate nitrogens, however at 3 wt% incorporation of nitrogen, it is unlikely that there is a significant a degree of neighboring nitrogens to form such interactions. It is more likely that these interactions occur with the oxygen bearing ketonic and furanic species in the material. To observe protons within a ∼0.5 nm distance (three to four bond distance) of the nitrogens, a 2D 1H–15N HETCOR spectrum with a longer cross-polarization contact time of 2 ms was acquired and is shown in Fig. 10(b). Here strong correlations between non-protonated pyridinic and imidazole nitrogens (labelled sites 6 and 5) and the expected aromatic protons two bonds away are observed, along with much weaker correlations to the alkyl protons. This result corroborates the conclusions from the 1H–13C HETCOR analysis and confirms that the aromatic rings in HC-SAS are not as extensively functionalized as the HC-S. On the other hand, the imidazole, pyrrole and amide species do show strong correlation with the alkyl protons, indicating that these species are functionalized or branched to a greater extent than the aromatic ring species.
The 1D and 2D solid state NMR experiments thus reveal fundamental structural insights about the nature of hydrothermal carbon. Firstly, the structure of HC-S is similar to that proposed in previous studies,30 thus the underlying mechanism is most likely very similar. Secondly, there is a dramatic change in the carbon speciation when a relatively small amount of nitrogen from an ammonium salt is introduced into the hydrothermal reaction and is seen to give rise to a distribution and speciation of nitrogen that is different to that incorporated from the use of glycine.31 Based on existing 15N NMR chemical shifts and chemical shift anisotropies,46 a more diverse set of nitrogen speciation can be discerned in HC-SAS. Importantly, the analysis of molecular connectivities reveals that HC-SAS consists of two distinct domains; i.e., a condensed aromatic domain and a branched aliphatic domain. As opposed to the “homogenous” structure for HC-S, the aromatic and aliphatic domains of HC-SAS may form “heterogeneous” domains with segregation on the nanometer scale. This conclusion is based on the fact that molecular proximity on a sub-nanometer scale of ∼0.5 nm would result in a 2D 1H–13C NMR spectrum for HC-SAS which would be not unlike the HC-S. On the other hand, molecular separation on a >10 nm scale would yield spectra where there are effectively no cross-peaks between the aromatic and aliphatic domains, which thus gives an ∼1 nm scale for the domain sizes. The “heterogenous” structure also indicates that during the synthesis of HC-SAS, the two domains form by a competing set of reactions. The NMR analysis allows the assembly of model partial structures of HC-S and HC-SAS, as shown in Fig. 7(a) and (b), which reflect the branched “homogeneous” structures of HC-S, in contrast to the condensed “heterogenous” structures of HC-SAS.
Capacitance produced from the electrical double layer is linked to the specific surface area of the electrode material, with increases in surface area generally providing increases in capacitance.1 Increasing the surface area of carbons is performed through activation techniques that apply a physical (e.g., CO2, steam) or chemical (e.g., KOH, H3PO4, NaOH) activation agent and high temperatures (600–1100 °C).47,48 This process increases the porosity of the carbon structure through the removal of both carbon and various functional groups within the structure. Although activated carbons typically achieve higher capacitances than non-activated carbon, the activation process is counter-productive to electrochemically examining functional groups that provide pseudo-capacitance on carbon surface. This is because the removal of electrochemically active groups cannot be avoided in the activation process. Additionally, the structural differences in the parent material (HC-S/HC-SAS) may influence the formation of porosity from activation creating an additional factor. This would further detract from the objective to examine the effect of pseudo-capacitance on hydrothermal carbon. As a result, both hydrothermal carbons were not activated.
The effects of nitrogen doping can be assessed from the comparison of electrochemical measurements of the doped and non-doped materials. To ascertain the potential capacitance and stability increase of nitrogen doping on hydrothermal carbon, HC-S and HC-SAS were studied using cyclic voltammetry (CV) in the potential range −0.3 to −0.9 V vs. SCE, and electrochemical impedance spectroscopy (EIS) using a three electrode cell in an alkaline electrolyte (1 M KOH). Fig. 11 shows the CV plots of both hydrothermal carbons at a rate of 10 mV s−1. The CV curve of HC-S shows a poor rectangular shape, implying a high electrical resistivity of the hydrothermal carbon (10.9 × 104 Ω cm), with a decrease in area after extended cycling leading to a decreased capacitive behaviour and hence lower specific capacitance (Table 2).
Cycle Number | Capacitance (F g−1) | |
---|---|---|
HC-S | HC-SAS | |
1 | 54 | 61 |
250 | 50 | 60 |
500 | 39 | 60 |
750 | 38 | 60 |
1000 | 34 | 60 |
To highlight the effect of pseudo-capacitance, a box representing an ideal EDLC with no pseudo-capacitance behaviour is drawn over the CVs. The area outside this box shape is generally due to redox active species and can be seen between −0.7 and −0.9 V and −0.3 and −0.4 V on HC-S. This is attributed to reduction reactions on oxygenated functionalities resulting in the removal of such groups from the carbon surface. This in turn decreases the pseudo-capacitance contribution from HC-S and thus its overall capacitance over 1000 cycles. Conversely, the CV curves of HC-SAS display a more uniform rectangular shape and do not suffer from the decrease in capacitance that is apparent in HC-S. Directly comparing the two CV curves, the influence of nitrogen doping can be easily seen as the pronounced reduction reactions with oxygen are not apparent.
The resistivity of HC-SAS is also lower compared to HC-S (2.6 × 104 Ω cm). It must also be noted that the surface area of HC-SAS (94.1 m2 g−1) is lower than HC-S (123.7 m2 g−1), thus the increase in capacitance is not due to EDLC behaviour. This suggests that nitrogen doped hydrothermal carbon is both inherently more stable and produces higher pseudo-capacitance than its non-doped counterpart. Furthermore, the reduction in oxygenated functionalities on the surface from nitrogen incorporation is electrochemically beneficial.
With regards to which species contributing to pseudo-capacitance, unlike other systems employing pseudo-capacitance (e.g., polyaniline) where discrete peaks related to the redox processes of electrochemically active species can be observed,49 these are not apparent in HC-S and HC-SAS. This makes it impossible to ascertain which surface functionalities or structures are responsible for the increase in both stability and capacitance in HC-SAS.
Electrochemical impedance spectroscopy (EIS) was employed to further examine the behaviour and performance of both hydrothermal carbons. The impedance spectra of both hydrothermal carbons is displayed in Nyquist plots (Fig. 11(c) and (d)) for scans where the CV window was significantly different (at −0.6 and −0.7 V) and the vertex potentials (−0.3 and −0.9 V). The impedance spectra of both hydrothermal carbons is displayed in Nyquist plots (Fig. 11(c) and (d)) for scans where the CV window was significantly different (at −0.6 and −0.7 V) and the vertex potentials (−0.3 and −0.9 V). In both hydrothermal carbons, each EIS spectra contains (i) a significantly depressed semicircle situated just above the Z′ axis in the high frequency region representing double layer capacitance, and (ii) a straight line with decreasing frequency representing mass transport. The substantial differences in behaviour observed between both samples over the potential range, qualitatively reflect changes in the double layer capacitance, charge transfer process at the electrode–electrolyte interface and redox reactions at the surface. These variations are from variations in the electrochemical processes occurring at the hydrothermal carbon electrode, as the result of the applied ac potential accessing different functionalities on the surface of both hydrothermal carbons.
To ascertain the capacitance and quantitatively compare the behaviour of each hydrothermal carbon, an equivalent circuit (Fig. 1) was fitted to the data with the results displayed in Table 3. In terms of series resistance (Rs), there is very little change between both hydrothermal carbons, averaging 11.2 ± 0.4 Ω over the entire series, this is due to the 45% conductive carbon black (CCB) added to each electrode. As Rs indicates the entire electrodes resistivity (CBB + HC), not just the hydrothermal carbon. This accounts for the difference between Rs and the CV/resistivity measurements. The charge transfer resistance (Rct) on the surface of the hydrothermal carbons increased for HC-SAS over HC-S but also varied significantly. This implied that nitrogen may be inhibiting charge transfer reactions at the selected potentials, which in the case for HC-SAS would most likely be inhibiting the reduction of surface oxygen groups or the formation of hydrogen adatoms at lower potentials.
Sample | Potential (V) | CPEct | Rct (Ω) | Cdl (mF) | Rs (Ω) | CPEd | ||
---|---|---|---|---|---|---|---|---|
σ1 | m1 | σ2 | m2 | |||||
HC-S | −0.3 | 907 | 0.682 | 1.0 | 0.05 | 11.9 | 13.6 | 0.501 |
−0.6 | 131 | 0.468 | 1.4 | 0.04 | 11.6 | 8.9 | 0.542 | |
−0.7 | 97 | 0.393 | 1.6 | 0.02 | 11.4 | 9.7 | 0.580 | |
−0.9 | 1018 | 0.656 | 1.5 | 0.03 | 12.1 | 19.9 | 0.439 | |
HC-SAS | −0.3 | 712 | 0.661 | 5.6 | 0.12 | 11.0 | 16.7 | 0.711 |
−0.6 | 522 | 0.707 | 2.4 | 0.21 | 11.0 | 12.7 | 0.670 | |
−0.7 | 10 | 0.335 | 6.6 | 41.39 | 10.9 | 10.5 | 0.737 | |
−0.9 | 10 | 0.382 | 4.8 | 26.62 | 11.0 | 9.7 | 0.702 |
Cdl represents the double layer capacitance of the hydrothermal materials. Generally, this does not account for redox processes involved in pseudo-capacitance as the charge transfer on these groups is slower than double layer charge storage. However, if the kinetics of the charge transfer processes is comparable to double layer charge storage, they can be accounted for in Cdl. HC-S has a very low Cdl, which indicates that (i) there is very little double layer capacitance, and (ii) that the capacitance observed in HC-S is from redox reactions that are slower than the double-layer. As HC-SAS has a lower surface area than HC-S, it would be expected that Cdl would also be lower. However, Cdl is actually higher in HC-SAS, which indicates that the charge transfer rate of the redox active groups on HC-SAS are significantly faster than those on HC-S. As the biggest difference is the inclusion of nitrogen groups, this suggests that nitrogen groups have a significantly faster charge transfer rate than oxygen groups.
CPEd represents the straight line in the Nyquist plots, ideal capacitive behaviour is a vertical line with m = 1. As m is significantly less than one, this specifies there is restrictive ion transport through the porous microstructure for both the hydrothermal carbons. In regards to capacitance, the EIS correlated with the conclusions from the CV data in which HC-SAS had an increased capacitance over HC-S.
![]() | (4) |
![]() | (5) |
![]() | (6) |
![]() | (7) |
![]() | (8) |
![]() | (9) |
![]() | (10) |
For the oxygen functionalities, these reactions are only quasi-reversible in KOH electrolytes,50 leading to the gradual decrease in capacitance noted for HC-S. These reactions could also be completely irreversible due to the loss of gaseous CO from (5) and CO2 from (4) if these reactions were pushed to higher potentials. Additionally, the furan rings that dominates the structure of HC-S is not likely to contribute electrochemically, other than to increase the resistance of the material.
In HC-SAS, the nitrogen functionalities are fully reversible, leading to the increased stability observed under cycling. Nitrogen groups are only going to be lost from primary amines, thus the high level of pyridines, pyrrols, and secondary and tertiary amines are going to remain stable throughout cycling and not be lost as gas. Additionally, the incorporation of nitrogen into the structure has an additional benefit of removing furan groups, as well as decreasing other oxygen functionalities, from the hydrothermal structure. Overall, the incorporation of nitrogen into hydrothermal carbon is seen to increase the stability of the carbon electrode as well as increase its capacitance.
This journal is © The Royal Society of Chemistry 2016 |