Tuning Properties of Partially Reduced Graphene Oxide Fibers upon Calcium Doping

Abstract

The arrangement of two-dimensional graphene oxide sheets has been shown to influence physico-chemical properties of the final bulk structures. In particular, various graphene oxide microfibers remain of high interest in electronic applications due to their wire-like thin shapes and the ease of hydrothermal fabrication. In this research, we induced the internal ordering of graphene oxide flakes during typical hydrothermal fabrication via doping with Calcium ions (~6 wt.%) from the capillaries. The Ca²⁺ ions allowed for better graphene oxide flake connections formation during the hydrogelation and further modified the magnetic and electric properties of structures compared to previously studied aerogels. Moreover, we observed the unique pseudo-porous fiber structure and flakes connections perpendicular to the long fiber axis. Pulsed electron paramagnetic resonance (EPR) and conductivity measurements confirmed the denser flake ordering compared to previously studied aerogels. These studies ultimately suggest that doping graphene oxide with Ca²⁺ (or other) ions during hydrothermal methods could be used to better control the internal architecture and thus tune the properties of the formed structures.

1. Introduction

Graphene oxide (GO) is a form of graphene having a unique collection of oxygen rich groups on its surface (-COOH, -OH, =O, -O-) [1,2]. Because of oxygen rich moieties it is a promising material for sensing in biology and medicine: high-contrast bio-imaging and bio-sensing applications [3,4], e.g., biosensors (glucose, mechanical stress, magnetic field) [3,4], anticancer therapies [5,6], as well as for flexible electronic applications, e.g., supercapacitors [7], FET transistors [8,9], electrical wires [10], or as active elements in mechanical energy harvesters [11]. Well reduced fibers with flake ordered composition are reaching superior physical properties with electrical conductivity of σ = 10⁶ S·m⁻¹, thermal conductivity of 1557 W·m⁻¹·K⁻¹, tensile strength of 1.9 GPa, and a Young’s modulus of 309 GPa [12]. A stronger connection between flakes changes the flake arrangement and alters the above-mentioned properties in comparison to previously studied graphene oxide structures formed by hydrothermal methods [2,13]. Their electrical conductivity strongly depends on the quality of the reduction process and flake ordering and can be described by variable range hopping models [14,15].

In this article, we were interested in the change of magnetic, mechanical, and electric properties of graphene oxide fibers prepared via the hydrothermal method with additional doping by divalent Ca²⁺ ions. The addition of Ca²⁺ in a coagulation bath while fibers are formed during a wet spinning method results in increased flakes tendency for the formation of a denser network with Calcium ions acting as bridges between the individual flakes [16]. This method is one of the three known among the dry spinning method [17,18] and hydrothermal method (used here) [2] for the preparation of micro-fibers. In particular, we recently showed [13] that electron paramagnetic resonance (EPR) is an effective method that allows distinguishing multiple paramagnetic contributions of magnetic susceptibility, e.g., Pauli (conduction electrons) and Curie (localized paramagnetic centers, e.g., sp³ defects), and could be used to study the internal ordering of flakes. The source of the EPR signal, among different carbon materials, is related to the conduction electrons, e.g., anthracite [19], or localized paramagnetic states on the surface or edges, e.g., GO [20] or graphene [21]. However, most of the aforementioned centers are present in different samples, due to preparation processes, making the synthesis method and ion doping a crucial aspect to evaluate.

Graphene-based samples can show complicated magnetic behavior due to zigzag edges [22,23], which can give rise to the edge states magnetism and basal-plane sp³ defects formed by -OH groups (~1–1.2 µ_B) and adatom-induced magnetism [24,25]. We will discuss the Ca²⁺-doped fibers and compare the results to different carbon samples previously reported in the literature: graphene [21,26], GO in the form of aerogels [13,20], as well as with other carbon materials [27,28] which magnetic properties depend on particle size [29]. We will finally show the competitive mechanical and electrical response of the fibers, as well as the suitability of Ca²⁺-doping for tuning physico-chemical properties in reduced graphene oxide (rGO) fibers.

2. Materials and Methods

2.1. Graphene Oxide Preparation

Graphene oxide was purchased from the NANOPOZ company (Poznań, Poland). Flake diameters in the range of 0.9–46 µm, with a maximum number of flakes with a diameter of ~2 µm were obtained, the same as in previous reports [2]. Briefly, flake size distributions were obtained by manually measuring sizes using ImageJ^® from the sizes of around 200 flakes from multiple scanning electron microscopy images.

2.2. Reduced Graphene Oxide Fiber Preparation

GO suspension of 4 mg/mL was injected to the opened at both sides’ capillaries (with inner diameter of 0.6 mm), then sealed by melting, placed in autoclave and heated for 2 h at 180 °C, where the suspension coagulated and formed a hydrogel (Figure 1a). The hydrogel was left to dry in air at 21 °C (Figure 1a, inset). The preparation method was previously described when preparing partially reduced graphene oxide aerogels [2], except here the container was changed to Calcium containing capillaries. After the formation of xerogel, final drying was performed at 70 °C for 48 h (without further annealing at higher temperatures).

2.3. Scanning Electron Microscopy (SEM) and Energy Dispersive X-ray Spectroscopy (EDX)

EDX measurements were performed at 5 kV. The preparation of cross-sections for SEM analysis was carried out using a focused ion beam system JIB-4000 (JEOL) with Ga⁺ ions. The current and voltage of the ion beam during preparation were 23 pA to 60 nA and 30 kV, respectively.

2.4. Vibrating Sample Magnetometry (VSM)

DC magnetic susceptibility measurements were performed using the VSM Quantum Design PPMS system in a temperature range of 4 to 300 K, using a maximum applied magnetic field of ±1 T. Zero-Field-Cooled (ZFC) and Field-Cooled (FC) lines were taken for magnetic field 0.3 T (3 kOe) in a temperature range from 4 to 300 K. The value of the magnetic field was selected in order to easily compare it with the EPR measurements.

2.5. Electron Paramagnetic Resonance (EPR)

The spectroscopic continuous wave (CW) EPR measurements were performed with a RADIOPAN SE/X-2547 (9 GHz) spectrometer equipped with a RCX661A TM110 resonator and an Oxford CF935 cryostat allowing measurements in a temperature range of 4.2–300 K. The modulation amplitude was 0.05 mT, the microwave power was 11.38 mW (without saturation effects), and the microwave frequency was recalculated for each measured point to match exactly 9 GHz. The number of points per spectra: 1024, accumulations: 2, time per one point: 120. The number of spins was estimated by a direct comparison method with DPPH standard which was earlier calibrated with a primary standard—copper sulphate pentahydrate monocrystals. The EPR relaxation measurements were conducted with a Bruker ELEXSYS E580 EPR Spectrometer equipped with an EN4118X-MD4 resonator in a temperature range of 5–100 K. Field-sweep echo-detected (FSED) spectra were obtained from X-band pulse experiments. The shot repetition time was set at 186 µs, π/2 pulse was set to 20 ns for T_m, T₂, and FSED measurements, and 34 ns for inversion magnetization experiments to obtain the maximum magnetization inversion.

2.6. Electric Conductivity Measurements

Room temperature electric conductivity measurements were performed with a 4-point method using Keithley 2636 B SMU. The distance between the voltage electrodes was 5 mm, and the diameter of rGO fiber used was d = 50 µm.

2.7. Nanoindentation

The uniaxial tensile tests of rGO fibers were performed using the Agilent T150 UTM tensile machine. Fibers with a typical diameter of 91 ± 13 µm and a length of 4.5 mm were glued to the supporting frame and analyzed by the scanning electron microscope (Quanta FEG 250, FEI) to determine their dimensions—each of the fifteen fibers’ diameters were measured at five sites. The analysis allowed identifying defects in fibers’ morphology, like kinks or bend/displacement. Such defects correlate with lower tensile strength measured afterwards by the means of a uniaxial tensile test. Extension for each fiber was realized with a strain rate of 10⁻⁴ s⁻¹ to enable a slow relaxation of fibers during the measurement. Nanoindentation tests were performed using a Triboindenter (TI950 Hysitron) with a Berkovich tip. Mechanical values were extracted using the Oliver–Pharr method [30]. Fibers were immobilized on metallic plates by commercially available silver paste. Samples were investigated at the maximum load of 1000 µN using a single signal, partial load-displacement function, and nano dynamical mechanical analysis (nanoDMA).

3. Results and Discussion

3.1. Fabrication of Reduced Graphene Oxide (rGO) Fibers

Initially, we injected a solution of 4 mg/mL graphene oxide flakes inside capillaries containing calcium ions (Figure 1a). We then subjected such capillaries to a hydrothermal treatment and then dried them out on the Teflon tape (Figure 1a, inset). The fibers were left in an ambient atmosphere for drying, which resulted in the collapse of the hydrogel pore structure and led to the formation of xerogel with a shape resulting from the capillary container, initially just after drying of about 600 µm diameter, decreased after a full drying cycle to ~50 µm (Figure 1b and Figure S1, Supporting Videos 1 and 2), and typical length decreasing from ~8 cm to ~5 cm. Further drying at 70 °C (48 h) removed remaining moisture/H₂O from the structure but did not influence the shape or physical dimensions of the obtained fibers. The post-hydrothermal drying treatment leads to 9–10 times shorter diameters and 20–30% shortened lengths with final surface roughness Rq equal to 62.8 ± 3.3 nm in a reproducible manner, as reported previously [31].

The surface of the fabricated fibers under scanning electron microscope (SEM) shows partially reduced graphene oxide flakes lying on each other in a co-axial arrangement (Figure 1b and Figure 2f). The reduction protocol is the same as used in our previous studies, thus we expected the produced fibers to contain the same defects [13,31]. The EDX (Figure 1c) analysis of the fibers was done at a relatively low acceleration voltage (5 kV) reducing the depth of penetration into the sample. The analysis was made for the three main elements in multiple spots (3 × 8 grid) and showed the average elemental fiber composition of C: 62, O: 32, and Ca: 6 wt.% (C/O = 1.94). As expected, quantitative maps of C, O, and Au show roughly homogeneous carbon and oxygen distribution over the entire fiber (Figure S2).

Due to obvious limitations of the SEM technique in assessing internal structure of partially reduced graphene oxide fibers, focused ion beam milling (FIB) was used. It allowed the visualization of the pseudo-porous structure inside the fiber (Figure 2a–c and Figure S4). The structure located inside the rectangular milled square in Figure 2b and presented as magnified in Figure 2a was used for the estimation of the pore surface to the surface of the selected square. Surface porosity is defined as the ratio of the overall pore area in the square to the entire square surface. The surface porosity across multiple images on average comes to ~20%. It should be mentioned that the ion beam could damage the pore structure and this data should be treated with caution, as coming from pseudo-pore structure. The porous structure is visible only when the ion beam irradiates the surface perpendicular to it, while in the other case, the flat surface is visible (Figure 2f). Figure 2d shows a pore size statistic made inside of the rectangular milled element visible in this image. The distribution’s profile can be approximated by the lognormal function. The mean pore size is 360 ± 445 nm (from 240 elements). The minimum pore diameter is 50 nm and the largest is 3150 nm.

3.2. Electrical Properties

Improved, due to Ca²⁺-doping, layer stacking is visible in electrical measurements (DC). Current vs. voltage plot shows a linear relationship and a specific conductivity in the order of 20–40 S·m⁻¹, while undoped it is in the range of 2–7 S·m⁻¹ (Figure 3, Figures S5 and S6). It is known that electrical conductivity measured on a single GO flake is in the order of 0.05–2 S·m⁻¹ [32], while for a fully reduced graphene oxide aerogel it is reaching 528 S·m⁻¹ [33]. Our previous results obtained for partially reduced aerogels showed a conductivity of 1.7 S m⁻¹ [13]. For comparison, conductivity of dried GO straps (known as GO paper) is in the order of 1.2 × 10⁻⁴ S·m⁻¹ (Figure S5). The electrical conductivity in such systems is caused by a variable range hopping with the highest energy necessary for hopping between the flakes ($\sigma ~e^{-{\left(T_0\right)}^{\frac{1}{2}}}$). Electrostatically assembled inter-flake links by Ca²⁺ decrease the energy necessary for this process by modification of the structural arrangement, causing a stronger overlap of the flakes. This mechanism can be found in systems were conductive particles are randomly connected to each other [34] with an energy barrier located in between the flakes [35]. DC measurements performed by Rani et al. [36] on compressed powders of graphite, expanded graphite, reduced graphene oxide, and carbon nanotubes showed the increase of conductivity with the increase of pressure which was explained by forming of effective contacts between grains [37,38].

3.3. Mechanical Properties

Pristine reduced graphene oxide fibers due to having a collapsed pore structure in comparison to aerogels are much more resistant to damage. Individual performance depends on the size of the graphene flakes used, as well as their arrangement during drying and reduction [10]. To characterize the basic mechanical properties, the ultimate tensile strength was measured (Figure 4a). For the tensile strength measurements, all fiber diameters were estimated by previous SEM investigation to find the average pore diameter, which later was used as a normalization parameter (Figure 4a). The average fiber diameter taken for this measurement was 91 ± 13 µm and the length of the obtained fibers reached 5 cm. Only the fibers exhibiting proper morphology were evaluated (some omitted fibers are shown in Figure S7).

The stress–strain curve of rGO fibers indicates the brittle nature of the material without a significant plastic deformation region. The elasticity modulus calculated from the linear segment of the stress–strain curve was 360 ± 200 MPa. The fracture occurs at strain 2.1% ± 1.1%, corresponding to the stress of 4.7 ± 1.4 MPa. The value of fracture stress is about one order of magnitude lower than in the case of carbon nanofibers reinforced by GO [39] or poly(lactic acid)/GO/stearic acid composites [40], which is due to reinforcing fillers (nanotubes). Nanoindentation tests performed on the samples showed similar mechanical values to the ones reported in the literature for porous fibers, E = ~3.3 GPa and H = ~66.7 MPa. The lack of recovery on the single indentation tests (Figure 4c) suggests the highly plastic surface of the fibers, which is congruent with the SEM images and studies after FIB milling. Depth studies (Figure 4d) show the compressive regime of the nanofibers, above ~800 nm, which accounts for the plastic deformation of the sample. After that point, elastic response is observed with a 40% recovery, which is attributed to the porosity of the samples. Results for pristine fibers on the mechanical analysis reported elsewhere, are in proximity to the results obtained here: ultimate tensile strength of 29.4 MPa at 8.6% of elongation [41]. Fibers obtained by the dry spinning method show some spread of parameters: tensile strength is in the rage of 39.2–1450 MPa, elongation is in the rage of 1.1–20%, and a Young modulus of 1.9–282 GPa [42,43].

3.4. Magnetic Properties

Due to the different sources of a magnetic signal in plain defects ad-atoms or edge dangling bonds (edge states) the magnetism can show a complicated character. Ferro- and antiferromagnetism were theoretically predicted in graphene [44,45]. Moreover, the presence of vacancies (sp³ defects) caused by irradiation [46,47,48,49] or surface doping with hydrogen H⁺ ions and fluorine in the form of XeF₂ molecules [50,51,52,53,54], or adatoms like -OH groups [24,55,56,57,58,59], moisture [35] or contamination by Mn²⁺ [60] or Fe²⁺ ion, can be source of additionally appearing exchange interactions [61,62,63,64,65]. Above that, an interplay between localized (point defects) and delocalized (conduction electrons) paramagnetic centers take place which depends on the particle size, and according to Ćirić et al. [66,67], the EPR signal comes from conduction electrons for large flakes and from localized centers for smaller than 1 μm² flakes. Smaller particles (<30 nm) which have more defects, and for this reason they are more reactive, anchor 6.6 times more iron ions than larger, micrometre sized particles on the edges [60].

Magnetization vs. magnetic field dependence of partially reduced graphene oxide fibers measured by vibrational magnetometer are in the range of ±1 T (10 kOe) (Figure 5a). As expected for paramagnetic sample, saturation magnetization grows with decreased temperature. The Zero Field Cooling (ZFC) and Field Cooling (FC) curves show differences in temperatures below 100 K indicating weak ferromagnetic interactions. The field of 3 kOe (Figure 5b) was chosen to easily compare the results with those from EPR in which the resonance signal was recorded at 9 GHz.

EPR spectroscopy shows a single signal at a g-factor of 2.0025 (300 K) which could be decomposed to two Lorentzian lines (Figure 6a). Normalized to 300 K, the EPR intensity obeys the Curie low, approximated with the equation $\frac{167.6\pm 7.8}{T}$ reaching in 4.2 K value of 30. On the other hand, better fits could be achieved by the addition of a constant positive magnetic susceptibility, approximated by the equation $\frac{167.6\pm 7.8}{T}+1.76\pm 0.4,$which can be further interpreted as additional contribution of Pauli paramagnetism stemming from conduction electrons (Figure 6c). The presence of both localized and delocalized centers in different proportions was already shown in GO [66], reduced GO similar carbon material: Panich et al. [64], Shames et al. [68], Augustyniak-Jabłokow et al. [27].

The total number of spins was estimated as $3\times {10}^{18}$ spins/g. Figure 6b,d shows the temperature dependences of g-factors and linewidths. The above g-factor is of a free radical in the entire temperature range indicating small positive spin–orbit coupling, as expected for radicals or non-interacting sp³ defects located on the flakes surface. The difference in g-factors of both components, which become significant below 100 K, cannot be directly explained by ZFC/FC curves in Figure 5b. It is usual that EPR signals in carbon samples can be decomposed into two lines, further changes in g-factors are small, and differences appear naturally when having different signal sources. There is no premise in EPR results presented in Figure 6, which could explain the bifurcation in ZFC/FC.

Pulse EPR measurements confirm the presence of two lines-species in the sample (Figure 6). Two lines are visible in a field-swept echo experiment, suggesting inhomogeneous line broadening in both cases (Figure 7). Notice that the broader CW-EPR signal is shown for comparison. The lines have widths of 1.06 mT and 0.32 mT. The integral intensity of the first line is larger by a factor of 3.2.

Further, the spin–lattice T₁, phase memory time T_m, and spin–spin relaxation time T_2echo^* were recorded in the temperature range of 5–100 K (Figure 8). The spin–lattice time can be decomposed into two components, which rates at 5 K have 0.5 MHz (2 µs) and 0.06 MHz (16.6 µs), increasing with temperature to 1.25 and 0.6 MHz at 100 K, respectively (Figure 8a). The relaxation process is dominated by a direct process until around 70 K (linear function) after which a higher order Raman relaxation process becomes more dominant. Both radicals can be described by equations${\mathrm{T}}_1^{-1}={\mathrm{A}}_0+{\mathrm{A}}_1\times \mathrm{T}+{\mathrm{A}}_2\times {\mathrm{T}}_5$, where the first component is temperature independent and not related to the relaxation processes, A₁ is the direct Raman relaxation, and A₂ is multi phonon Raman relaxation (

Table 1. Components of different relaxation processes for two paramagnetic centers.

T1A−1	T1B−1	T1GO−1 [20]
A0 = 0.51 s−1	A0 = 0.06 s−1
A1 = 4.4 × 10−3 K−1·s−1	A1 = 9.0 × 10−4 K−1·s−1	A1 = 4.7 × 10−4 K−1·s−1
A2 = 2.8 × 10−11 K−5·s−1	A2 = 4.2 × 10−11 K−5·s−1	A2 = 6.6 × 10−9 K−5·s−1
R2 = 0.9995	R2 = 0.9957

). In comparison, dried GO paper [20] showed spin relaxation T₁ time of 52 ms at 5 K, while here it is 17 µs (T₁⁻¹ was 535 Hz at 100 K). The difference lies in the oxygen reduction process, which produces a larger number of surface defects and causes 3000 times faster electron spin relaxation. Moreover, a direct relaxation process in GO paper dominates up to 100 K, which as it seems can be treated as an indicator of a small number of defects. It seems that the reduction process increases the number of surface defects and by that the relaxation rate, which further starts to be dominated by a higher order Raman processes already at lower temperature.

In the wider picture, electron relaxation studies performed on other carbon materials show following results: Panich et al. reported graphene oxide heavily doped with Mn²⁺ ions showing T₁ (from saturation measurements): <10 ns, T₁⁻¹ = 100 MHz [64], for nanographites doped with Fe^2+,3+ T₁ < 10⁻⁹ s [60], Rao et al. showed graphene nanoribbons two spin lattice relaxation processes: T₁⁻¹ ~0.014 MHz and 0.1 MHz [69]. For graphene, T₁⁻¹~T₂^−1*~3 MHz was reported at 100 K [21] (and 0.7 kHz at room temperature to 1.45 kHz at helium temperature). In partially reduced graphene oxide (prGO) with an average flake size of ~2 µm the relaxation times were 17.5 MHz (T₁⁻¹ = T₂⁻¹) for which the reduction caused the increase of the spin–lattice relaxation rate T₁⁻¹ by a factor of ~5.8 [13]. For Ca-doped fibers, the T₁⁻¹ is lower at 100 K around 1.25 MHz, compared to undoped fibers mentioned above at 17.5 MHz.

The spin–spin times $T_2^{*}$ were fitted by a single exponential function$V\left(t\right)=A\times \exp (-\frac{t}{T_2^{*}})$ from the free induction decay. The phase memory time was evaluated at each temperature from the echo intensity decay with the increase of the dwell time between two pulses as well as the mono-exponential function used of the form$V\left(2\mathit{\tau }\right)=A\times \exp (-\frac{2\mathit{\tau }}{T_m})$. T₁ is much longer than T_m, 27 times longer in 5 K, which suggests that T_m is dominated by molecular motions.

The phase memory and spin–spin relaxation rates grow with the increase of temperature, as suggested by the broadening of the EPR line (Figure 8b). The spin packets are ~12.2 times narrower than the line width (T₂^* = 35.8 ns, T_m = 219.1 ns). The motional narrowing of the spin packets (decrease of the relaxation rate) corresponds to the decrease of the EPR line (T₂^−1*), which may be caused only by motional narrowing of spin packets. As far as it goes for the tendency, the decrease of the phase memory relaxation rate is similar to the one observed in graphene oxide [20] and it seems natural for this sample. The difference in the case of GO lies in the presence of oxygen rich groups and their re-orientations under the influence of a magnetic field. This process can be caused by motion of the centers due to the increase of temperature or by an exchange interaction between individual resonance frequencies of the paramagnetic centers.

4. Conclusions

In summary, we used calcium containing capillaries during hydrothermal fabrication of graphene oxide dispersion to form Ca²⁺ ions doped reduced graphene oxide fibers. Formed fibers appeared to have lateral alignment of GO flakes around the fiber axis with stronger inter-flake connections developed by cross-linking points of divalent Calcium ions. Doped fibers exhibit tensile strength in the order of 4.7 MPa and an elastic modulus of 360 MPa. Nanoindentation tests performed on site of the fiber showed the following values: E = ~3.3 GPa and H = ~66.7 MPa. The changed flake ordering caused by Ca²⁺ ions increased electrical conductivity and decreased the electron spin–lattice relaxation rate $T_1^{-1}$. The electron spin relaxation rate increased in comparison to the results obtained previously for hydrothermally produced GO paper, whereas the specific conductivity also increased to ~20–40 S·m⁻¹ (compared to 2–7 S·m⁻¹ undoped rGO fibres/films and 1.2 × 10⁻⁴ S·m⁻¹ for GO paper). The sample is paramagnetic, exhibiting two EPR signals resulting from two different sp³ defects (-OH groups), other structural defects, as well as conduction electrons. The relaxation T₁ rates can be described by a direct process and multiphoton Raman scattering for temperatures reaching 100 K. Thermal reduction increases the T₁ relaxation rate. Removing the oxygen surface groups makes the formation of local phonons easier, which ends with a T⁵ relaxation order dependence.

Our results point to an easy way of tuning the properties of hydrothermally formed graphene oxide-based structures by replacing containers with calcium, which could bridge the flakes by inter-connections and enhance the flake–flake interactions, thus forming more conductive fibers for multiple applications, including electronics, bio-electronics, and cell-culture applications.