Introduction

Caffeine (1,3,7-trimethylxanthine) is a natural alkaloid, which is found in various kinds of food and beverge that we consume in daily life. It plays an important role in various physiological effects such as relaxation of bronchial muscle, stimulation of the central nervous system, gastric acid secretion and dieresis [1]. However, it can have some adverse effects on health if consumed in excess [2–4]. And it is also considered to be a risk factor for cardiovascular diseases, asthma, kidney malfunction and may also cause depression and hyperactivity [5, 6]. That’s why there have been numerous studies with the target to develop reliable methods for determining caffeine, such as spectrophotometry, chromatography and biosensing [7]. Nevertheless, these methods are generally more expensive, time-consuming and complicated than electroanalytical techniques [7]. Accordingly, interest in electrochemical methods has increased for its simplicity, high sensitivity, good stability, low-cost instrumentation and onsite monitoring [8]. However, the electrochemical determination of caffeine at the more common electrode materials (e.g., metals, glassy carbon) is generally not feasible because the oxidation of caffeine occurs at a very positive potential, thus overlapping with the discharge of the background medium [9]. So various modified electrodes have been developed for the determination of caffeine recently [7, 8, 10].

Nanomaterials have been widely used in analytical chemistry as chemical sensor and biosensor materials [11]. Graphene, a kind of nanomaterial, has attracted attention increasingly since it was discovered in 2004 [12]. It shows promising applications in scientific and technological fields owing to its novel properties such as exceptional thermal and mechanical properties, high electrical conductivity [13–15]. Graphene oxide (GO), the oxidation form of graphene, bears two-dimensional plane and lots of electroactive oxygen containing functional groups and disorder on the basal planes and edges [16–18]. It possesses the properties of large surface area and good chemical stability, too. However, the conductivity of it deceased remarkably compared to graphene. But the excellent solubility and the convenience of preparation and fabrication can make it a nice material for the modification of electrodes.

In this paper, a novel voltammetric sensor of caffeine was fabricated based on GO combined with Nafion modified on glassy carbon electrode (Nafion/GO/GCE). The employment of Nafion has two functions: one intention is the immobilization of GO on GCE surface and the other is its good affinity towards caffeine [19], which would enhance the sensitivity toward caffeine. The characterizations of self-synthesized GO was investigated using spectrum of FT-IR and an atomic force microscopy (AFM), and the fabricated Nafion/GO/GCE were characterized using cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) techniques respectively. The experimental results demonstrated that the Nafion/GO/GCE can improve the oxidation of caffeine intensively. Based on this voltammetric sensor, a sensitive voltammetry of determining caffeine was used in detecting caffeine content in beverages successfully. The analytical results were then compared with that of high performance liquid chromatography (HPLC).

Experimental

Chemicals and reagents

Nafion 117 (5%, w/v in alcoholic solution) was purchased from Alfa (http://www.alfa.com/zh/gh100w.pgm) and was diluted to 0.5% (w/v) with ethanol before use. Caffeine was purchased from Biological and Chemical Reagent Store (Shanghai, China) and used as received. Standard stock solution of caffeine (2.0 × 10−2 mol L−1) was prepared using doubly distilled water and stored at 4 °C. All reagents used for HPLC were of chromatographic pure. Graphite powder (chemically pure) was purchased from Tianjin Yong Da Chemcal Reagent Development Center (Tianjin, China http://www.tjyongda.net). It was grinded in the agate mortar and then was sieved in the sieve with 320 mesh. All other reagents were of analytical reagent grade and used without further purification. The experimental water used was double-distilled.

Apparatus

All the voltammetric measurements were performed using a RST3000 electrochemical system (Zhengzhou Shiruisi Instrument Co. Ltd., Zhengzhou, China http://www.srs666.com) and a CHI650A electrochemical workstation (CHI Instrumental, Shanghai, China http://www.chinstruments.com/schi.html) was used for the EIS measurements. The three-electrode cell consisted of an Ag/AgCl reference electrode, a platinum wire auxiliary electrode and either a bare glassy carbon electrode (GCE) (d = 3 mm) or Nafion/GO/GCE working electrode. All potentials in this paper were given against the Ag/AgCl (3 mol L−1 KCl). All the pH values were measured with a PHS-3C precision pH meter (Leici Devices Factory of Shanghai, China http://www.lei-ci.com), which was calibrated with standard buffer solution (pH 4.00 and 6.86) at 25 ± 0.1 °C every day.

The determination of Caffeine by HPLC was carried out using an Agilent 1200 Series liquidchromatograph with an Agilent C18 (4.6 mm × 150 mm) and an Ultraviolet–Vis detector set at 270 nm. The mobile phase was a methanol/water mixture (30:70, %v/v) at a flow rate of 1.0 mL min−1, while the injection volume was 20 μL. The standard solution of caffeine for linear calibration curve was made using methanol.

Preparation of film-modified GCE

First of all, graphite oxide was synthesized using graphite power mainly according to literatures [20, 21]. Then exfoliation of graphite oxide to GO was achieved by ultrasonication of the dispersion for 40 min (1000 W, 20% amplitude) generally refering to literature [22]. A homogeneous aqueous dispersion of GO (2.3 mg mL−1) was obtained as a result.

The bare GCE was polished carefully with 0.05 μm alumina slurry, and sonicated in absolute ethanol and water, respectively. Finally, the GC electrode was thoroughly rinsed with water and dried in air naturally.

Nafion/GO composite solution was prepared by mixing 0.5% Nafion and GO (the ratio was 10:1 ,v:v) and sonicated for 30 min to get a homogeneous turbid liquid. The Nafion/GO modified GCE was prepared by casting 5 μL Nafion/GO turbid liquid evenly on the GCE surface with a microsyringe and dried at room temperature. For comparison, a Nafion modified GCE (Nafion/GCE) was prepared using the same way. Before use, a new fabricated Nafion/GO/GCE was equilibrated for about 3 h in 0.01 mol L−1 phosphate solution (pH 7.0) and stored in this solution at 4 °C for conserving when it was not used. All measurements were carried out at room temperature.

Results and discussion

Characterization of the GO

The GO used was synthesized in our laboratory. As shown in Fig. 1a, the spectrum of FT-IR of GO showed that the peak at 3428.71 cm−1 attributes to O–H stretching vibration, the peak at 1717.67 cm−1 attributes to C=O stretching vibration, the peak at 1398.92 cm−1 attributes to deformation of O–H, and the peak at 1057.75 cm−1 attributes to vibration of C–O (alkoxy), indicating that GO was successfully prepared. The morphology characterization of GO was performed by atomic force microscopy (AFM) using a mica as matrix on which GO was deposited. It can be seen in Fig. 1b that the GO sheets were almost single-layer. And the average thickness of single-layer GO sheets was approximately 1 nm.

Fig. 1
figure 1

Characterization of GO: a FT-IR spectrum of GO; b AFM image of GO from its dilute aqueous dispersion on freshly cleaned mica

Full size image

The electrode characterizations

The characterizations of fabricated modified electrode were investigated using cyclic voltammetry in 2.0 × 10−3 mol L−1 K3Fe(CN)6 containing 0.1 mol L−1 KCl solution first. Figure 2a showed the voltammograms compared using a bare GCE (cauve a), a Nafion/GCE (curve b) and a Nafion/GO/GCE (curve c) respectively. Figure 2b was the magnified voltammograms of curve b and c. A pair of well-defined redox peaks were observed at the bare GCE. The anodic and cathodic peaks were almost disappeared at Nafion/GCE, showing that Nafion acted as a blocking layer for electron and mass transfer that hindered the diffusion of ferricyanide toward the electrode surface. This was from the electrostatic (negative charge) repulsive effect between Nafion and K3Fe(CN)6. However, the peak current increased and the charging current decreased compared with curve b at Nafion/GO/GCE, showing GO can accelerate the electron transfer of ferricyanide due to its electric conductivity. This investigation also demonstrated that the electrodes were prepared just as design.

Fig. 2
figure 2

a Cyclic voltammograms of different electrodes in 2.0 × 10−3 mol L−1 K3Fe(CN)6 with 0.1 mol L−1 KCl solution. (a): bare GCE, (b): Nafion/GCE, (c): Nafion/GO/GCE. Scan rate: 100 mV s−1. b The magnified voltammograms of curve b and curve c

Full size image

The electrode characterizations were further investigated using electrochemical impedance spectroscopy (EIS), which can give information on the impedance changes of the electrode surface. The Nyquist plots of impedance spectroscopy at bare GCE (curve a), Nafion/GCE (curve b) and Nafion/GO/GCE (curve c) in 5.0 × 10−3 mol L−1 K3/K4Fe(CN)6 (1:1) containing 0.2 mol L−1 KCl solution were showed in Fig. 3. For getting these data, the used open circuit potentials were 0.253 V, 0.240 V and 0.249 V for bare GCE (curve a), Nafion/GCE (curve b) and Nafion/GO/GCE (curve c) respectively. After fitting suitable equivalent circuit and calculation, the electron transfer resistance (Rct) obtained was about 86 Ω, 2713 Ω and 1031 Ω for the bare GCE, Nafion/GCE and Nafion/GO/GCE, respectively. When Nafion was deposited on the surface of GCE, the Rct increased obviously. However, the Rct decreased remarkably after GO was immobilized in the Nafion film modified electrode. This resistance controled the electron transfer kinetics of the redox probe at the electrode, indicating the GO added can reduce the Rct due to its electroconductibility, while Nafion can help GO to attach on the GC electrode surface. And we can obviously know that Nafion/GO composite film was successfully existed on the GC electrode surface just as design.

Fig. 3
figure 3

Nyquist plots of bare GCE (a), Nafion/GCE (b) and Nafion/GO/GCE (c) in 5.0 × 10−3 mol L−1 K3/K4Fe(CN)6 (1:1) with 0.2 mol L−1 KCl solution respectively. Frequency range: 100000 Hz to 0.01 Hz; Perturbation signal: 5 mV

Full size image

Electrochemical response of caffeine at electrodes

Cyclic voltammetry (CV) was employed for investigating the response of caffeine at different electrodes. Figure 4a showed the voltammograms of caffeine (1.0 × 10−5 mol L−1) at the bare GCE (a), Nafion/GCE (b) and Nafion/GO/GCE (c) in 0.10 mol L−1 H2SO4 with a scan rate of 100 mV s−1. As it can be seen, in the potential window between 800 mV and 1600 mV, no redox peak of caffeine was observed at the bare GCE, but an anodic peak of caffeine was observed at both Nafion/GCE and Nafion/GO/GCE under the same conditions. This was likely due to the preconcentration action of Nafion for caffeine at the film surface. Besides, the peak current using Nafion/GO/GCE showed remarkable improvements with respect to that of Nafion/GCE. It was evident that Nafion/GO/GCE had superior electrochemical performances than bare GCE and Nafion/GCE, at least in the present case. The comparison between Nafion/GO/GCE and Nafion/GCE stressed the advantages of GO, which acted an excellent electrochemical oxidation of caffeine. Figure 4b was the comparison of cyclic voltammograms at Nafion/GO/GCE with absence (c’) and presence (c) of caffeine (1.0 × 10−5 mol L−1) in 0.10 mol L−1 H2SO4 at a scan rate of 100 mV s−1. It indicated that the peak came from the oxidation of caffeine indeed, and there was no redox peak of Nafion/GO/GCE in blank solution. The oxidation process of caffeine at Nafion/GO/GCE or Nafion/GCE is irreversible, which is consistent with the reported [9].

Fig. 4
figure 4

a Cyclic voltammograms recorded at the bare GCE (a), Nafion/GCE (b) and Nafion/GO/GCE (c) in the presence of 1.0 × 10−5 mol L−1 caffeine in 0.10 mol L−1 H2SO4 at a scan rate of 100 mV s−1. b Cyclic voltammograms of Nafion/GO/GCE with absence (c’) and presence (c) of 1.0 × 10−5 mol L−1 caffeine

Full size image

From the experiments, we observed that the anodic peak current increased with the time of Nafion/GO/GCE dipping in caffeine solution, and then a stable signal was reached after a certain time. The reason may be that caffeine incorporated with the film firstly when the Nafion/GO/GCE was dipped into the solution. At the same time, the oxidation product of caffeine adhered to the electrode surface may hindered the access of caffeine, and the electrode surface regenerated after a certain time’s stir. So equilibrium condition was achieved at a certain time. As expected, it was quicker at higher concentration to get equilibrium condition, but 400 s was enough even at the lowest analyzed concentration. Accordingly, all the reported data were collected 400 s after the immersion of the Nafion/GO/GCE in the analyzed solution.

Influence of supporting electrolyte and pH

Preliminary experiments were performed to choose the best supporting electrolyte for caffeine. Various types of 0.10 mol L−1 acids including HNO3, HCl, H2SO4, CH3COOH were tested as possible supporting electrolytes. The results allowed choosing 0.10 mol L−1 H2SO4 as the optimal medium in which the best compromise between sensitivity, reproducibility of the signal was achieved.

The pH of sulfate solution can influence the peak potentials and currents of caffeine. As shown in Fig. 5, the peak currents decreased and peak potentials shifted negatively by increasing the solution pH in the range from 0.95 to 2.19 (achieved by diluting 0.10 mol L−1 H2SO4). So pH 0.95 (0.10 mol L−1 H2SO4) was chosen as the most suitable value due to the excellent peak response.

Fig. 5
figure 5

The cyclic voltammograms of Nafion/GO/GCE in H2SO4 with 2.0 × 10−5 mol L−1 caffeine at different pH (a→f): 2.19, 1.97, 1.79, 1.38, 1.21, 0.95; scan rate 100 mV s−1

Full size image

Calibration curve of caffeine

For investigation of the relationship between peak currents and caffeine concentrations, the differential pulse voltammetry (DPV) was performed, which has high sensitivity and excellent separation from background current (Fig. 6a). The optimal experimental parameters were as follows: 50 mV of pulse amplitude; 6 mV of pulse increment; 20 ms of pulse width and 60 ms of pulse period. Data obtained analyzing caffeine standard solutions in 0.10 mol L−1 H2SO4 allowed estimating the functional relationship (Fig. 6b), which is linear in the range of 4.0 × 10−7–8.0 × 10−5 mol L−1. It can be described by the following linear regression equation, with a correlation coefficient R2 of 0.9965:

$$ {i_{\text{p}}}\left( {\mu {\text{A}}} \right) = 0.{796} + 0.{658} \times {1}{0^{{6}}}C $$

Where i p was the oxidation peak current in μA and C was the concentration of caffeine in mol L−1. The detection limit was obtained as 2.0 × 10−7 mol L−1 based on the signal-to-noise ratio of 3 (S/N). Compared with some reported caffeine sensors (Table 1), Nafion/GO/GCE was more sensitive with wider linear range simultaneously. The comparison of this work with reference [7] showed the advantage of GO added. As to reference [8], it can be concluded that GO exhibited electroanalytical performance as excellent as or even better than that of carbon nanotubes. This research suggested that the modified electrode can be a very promising voltammetric sensor for the trace analysis of caffeine.

Fig. 6
figure 6

a Differential pulse voltammetrys at Nafion/GO/GCE in 0.10 mol L−1 H2SO4 with some different caffeine concentrations (a→i): 0 mol L−1, 9 × 10−7 mol L−1, 3 × 10−6 mol L−1, 6 × 10−6 mol L−1, 9 × 10−6 mol L−1, 2 × 10−5 mol L−1, 3 × 10−5 mol L−1, 6 × 10−5 mol L−1, 8 × 10−5 mol L−1. b linear relationship of caffeine concentrations versus peak currents

Full size image
Table 1 Comparison of different caffeine sensors reported with this work
Full size table

Stability and reproducibility of the modified electrode

The Nafion/GO/GCE showed good stability for the determination of caffeine. The relative standard deviation (RSD) of the peak current was 3.2% (n = 5) for 5 successive measurement at the caffeine concentration of 1.0 × 10−5 mol L−1. The storage stability of the modified electrode was also investigated. When not in use, the Nafion/GO/GCE was stored in 0.01 mol L−1 phosphate solution (pH 7.0) at 4 °C. The peak current of caffeine retained 98.0% of its initial response current after a week. The result showed that the modified electrode possessed a nice storage stability and long time-life.

Moreover, a GC electrode was modified for three times under the same preparing conditions and then was performed in 1.0 × 10−5 mol L−1 caffeine. The RSD of the peak current was 3.2% (n = 3), revealing good reproducibility of the modification of the electrode.

Interference

The influence of some potentially interfering species on the determination of caffeine was evaluated in detail. A fixed amount of 2.0 × 10−5 mol L−1 caffeine spiked with various potentially interfering species was evaluated under the same experimental conditions. The results showed that 2000-fold of glucose, 100-fold of ascorbic acid, 50-fold of dopamine, sucrose, 35-fold of epinephrine, 20-fold of uric acid, 2-fold of tannic acid, 3-fold of xanthine and 7-fold of theophylline had almost no influence on the current response of caffeine (signal change below 5%). All these indicated that Nafion/GO/GCE had good selectivity for the determination of caffeine.

Figure 7 showed differential pulse voltammetrys at Nafion/GO/GCE in 0.10 mol L−1 H2SO4 with 2 × 10−5 mol L−1 caffeine and some different theophylline concentrations. Two well-defined peaks appeared at the potentials of 1304 mV and 1436 mV, corresponding to theophylline and caffeine respectively, which offered a promising possibility for the simultaneous determination of theophylline and caffeine.

Fig. 7
figure 7

Differential pulse voltammetrys at Nafion/GO/GCE in 0.10 mol L−1 H2SO4 with 2 × 10−5 mol L−1 caffeine and some different theophylline concentrations (a → h): 0 mol L−1, 2 × 10−5 mol L−1, 4 × 10−5 mol L−1, 6 × 10−5 mol L−1, 8 × 10−5 mol L−1, 1 × 10−4 mol L−1, 1.2 × 10−4 mol L−1, 1.4 × 10−4 mol L−1

Full size image

Real sample analysis

Some beverage samples were tested to evaluate the practical applicability of the method using a freshly prepared Nafion/GO/GCE. The standard addition method was adopted for it allowed minimizing possible matrix effects due to the presence of other components in the real samples. The beverage samples were sonicated to remove the gas content and then were diluted (1 : 201) with the supporting electrolyte to operate in the linear range of the method and to reduce the matrix effect. The values detected were listed in Table 2. The detected caffeine content in cola samples were 5.91 × 10−4 mol L−1 and 5.73 × 10−4 mol L−1, which were in agreement with the regulation of the American Beverage Association [24] in the range of 4.3 × 10−4 mol L−1–8.7 × 10−4 mol L−1 (20 mg–40 mg per 8 oz). Caffeine content of energy drink was in agreement with the declared content (50 mg/250 mL). Standard caffeine was added in the samples and the total contents of caffeine were determined to calculate the recovery (Table 2). And HPLC was employed to test the correctness of the results (Table 2). The detected results obtained from the two methods were compared using t-test under 95% confidence levels. The results showed no significant difference for Cola 2 detection, but there were some difference for the other three samples, which may be caused by the basal component difference of the samples. Whereas, the results obtained from the two methods were generally in accordance with each other, which indicated a potential technique for the quick analysis of caffeine.

Table 2 Determination results of caffeine in beverage samples by DPV and HPLC
Full size table

Conclusions

This study demonstrated a new caffeine voltammetric sensor, Nafion/GO/GCE, which can be applied to the detection of caffeine with excellent sensitivity and selectivity. Nafion/GO/GCE exhibited significant advantages of wide linear range and low detection limit for caffeine compared with previous works. Besides, the modified electrode presented nice reliability and stability, had a superior immunity to some interference and offered a good possibility for extending the technique in routine analysis of caffeine.