Laser-induced graphene was studied as an alternative, carbonaceous material for electrochemical (bio)sensing applications. Its promising characteristics described by us earlier [15, 16] indicate that it may be not only an alternative but a superior transducer material for electrochemical point-of-care sensors. Not much is known about the effect of energy input during the fabrication in correlation to electroanalytical performance, only anecdotal data are available describing obtainable micromorphologies. By systematically studying process parameters that influence the energy input and correlating it to micromorphologies, Raman, and especially electroanalytical characteristics, this knowledge gap was sought to be filled by the study.
Understanding of scribing parameters’ effects on laser-induced graphene
We created LIG electrodes with a circular working area (d = 3 mm, A = 0.071 cm2) connected via a bridge to an electrical contact area for electrochemical testing (see Fig. 1c). To assess the influence of available laser instrument parameters, the power- and speed settings as well as spatial pulse density were systematically varied.
The electrodes were judged by mechanical integrity and visual appearance, represented by the heatmap overview for a medium pulse density setting (Fig. 2). Representative different visual appearances are shown in the inset. Electrodes of homogenous texture that would withstand bending without delamination were created by a suitable combination of power and speed (area of green color in Fig. 2). On the contrary, too much energy input resulted in brittle electrodes that would peel off the substrate easily. This was the case when the power setting was not ideal and hence a bit too high or the speed a bit too low (red and orange color). At low power and high-speed settings, the substrate was not carbonized or only partially (brown colors). When the energy input was just at the lower limit necessary for carbonization, only part of the desired shape was carbonized and instead triangular shapes appeared (see the second electrode from bottom in the inset of Fig. 2). This odd phenomenon stems from the carbonization starting at randomly located but energetically favorable nucleation points on the substrate. Those initially carbonized islands then promote carbonization in the immediate vicinity through increased absorbance which causes the lines of converted LIG to become longer with each consecutive sweep and create the observed triangular patterns with the tip facing upwards. The tips face downwards when the laser scanning direction is reversed (i.e., going from bottom to top regarding the y-direction).
While all green-marked laser settings generated visually similar electrodes, conductivity, and electrochemical activity differed. Cyclic voltammograms (CVs) were recorded on a series of electrodes in whose fabrication all parameters but one were kept constant to demonstrate the influence of a single parameter. The settings of 1000 × 1000 PPI and a constant power of 30% were chosen, because at this point, the speed could be varied on a wide range while producting functional electrodes (see Fig. 2). The resulting CVs in Fig. 3a show that peak-to-peak separation (ΔEp, Fig. 3b) and sheet resistance (Fig. 3c) both increase with scribing speed. To remove the influence of lead resistance from CV analysis, the leads were painted with conductive silver paste. Consequently, ΔEp values dropped overall and were influenced less by the scribing speed. Apparently, the high resistance of the LIG leads causes significant potential drop (iR-drop) between electrode working area and connecting clamp, which distorts the shape of the voltammograms. Longer leads caused higher drop (resistivity factor) but also larger electrode surfaces, higher concentration of redox species, or increased scanning speed in potential sweeping experiments (current factor). Strategies to prevent iR-drop therefore include the use of small electrodes and low redox species concentrations, if application of conductive paint is to be avoided. The lead dimensions of a designed electrode are usually dictated by practical reasons and therefore offer little room for adjustment. However, with optimized laser settings (see below), the sheet resistance of LIG could be reduced to as low as 10–20 Ω sq.−1 which greatly reduced the iR-drop problem (the value of 10–20 Ω sq.−1 was not a specific target, but it was the best achievable). Therefore, it was possible to make an all-LIG electrode without additional processing steps, like painting the leads, when the right scribing conditions are used.
Furthermore, the influence of laser pulse density was investigated. The laser cutter used in this study permitted only certain locked pulse density settings. Hence, 500 × 500 PPI, 1000 × 1000 PPI, and 1000 × 2000 PPI were compared (x- by y-direction). The heatmap in Fig. 2 is collected at 1000 × 1000 PPI which resulted in a particularly good set of electrodes at a broader range of settings. Some power/speed combinations at the higher pulse density of 1000 × 2000 PPI yielded even better conductivity and probably electron transfer. However, these electrodes were sometimes prone to delamination and the design-space for power and speed was narrower at the increased pulse density (see Fig. S3). Therefore, the medium setting of 1000 × 1000 PPI was used in this work to create LIG electrodes for sensing applications. Furthermore, in terms of output, scribing the same pattern at 2000 PPI in y-direction takes twice as long compared to 1000 PPI. Power/speed heatmaps recorded at other pulse density settings can be found in Fig. S3 along with a comparison of CV performance among the best electrodes obtained at different pulse density settings in Fig. S4. Electrodes created with the lowest spatial pulse density option of 500 × 500 PPI ranked lowest in conductivity and were deemed less useful for electrochemical measurements.
It should be mentioned that the exact results of Fig. 2 (and the other maps) are likely not directly translatable to machines with a different maximum laser power or different size of focused beam. Even the size of the desired pattern or placing the substrate too close to the limit of the scribing area can influence the carbonization outcome. However, from experience with different flatbed laser systems, we have found that optimal power/speed conditions always follow more or less the green diagonal region in Fig. 2. Specifically, as a rule of thumb, it seems that low power, low speed, and high spatial pulse density will create LIG electrodes with high conductivity and generally good electron transfer behavior at the electrode-electrolyte interface.
Based on these findings, the setting of 1% power, 10% speed, and 1000 × 1000 PPI (1/10/1000 × 1000) was selected as a very conductive, electrochemically active, and mechanically sturdy LIG material. Unless different settings are specifically mentioned, these settings were used to produce electrodes for all following electrochemical tests, and scribing took about 1 min per electrode. It should be pointed out though that many of the settings within the green range of the heat map (Fig. 2) can be suitable for a user’s electroanalytical needs and be created at higher throughput.
To demonstrate the variation in chemical sensing performance within the same pulse density setting, three electrode types made with increasing energy input were chosen from the green area in Fig. 2: 1/10, 25/40, and 60/100 (the first number denotes %power, the second %speed), and SWV responses were recorded to standards of [Fe(CN)6]3− in concentrations between 1 and 100 μM (Fig. 4). The sensitivity (see Fig. S16) decreases from type 1/10 over 25/40 to 60/100 which correlates with rising sheet resistance of the different LIG types: (26 ± 0.6) Ω, (49 ± 1.5) Ω, (57 ± 6) Ω for 1/10, 25/40, and 60/100, respectively. While all three electrode types appeared visually homogenous, the microstructure of 1/10 appears most uniform and flat while 25/40 shows fibrous structures of LIG and gaps. This appearance is even more pronounced in LIG type 60/100.
One concern about electrode manufacturing reproducibility regarded the location of the polyimide substrate in the machine when being scribed, since the lens carrier might be slower in some regions than in others due to available acceleration distance. Electrodes were thus scribed in several relative locations on the engraving table and peak-to-peak distances were measured. Also, 1- × 1-cm squares of LIG were made in the same locations to measure the sheet resistance. It was found that substrate location had no significant impact on performance in CV (Fig. 5a), measured as peak-to-peak separation. All electrodes showed around 100-mV peak-to-peak separation but the—overall low—sheet resistance values dropped from 33 Ω at the top left location to 23 Ω in the bottom right position (Fig. 5b). We conclude that while the effect of location on sheet resistance may be kept in mind, for most electrochemical experiments, no special attention needs to be paid as to where the substrate is placed during electrode manufacture.
It is known that the presence of doping atoms such as nitrogen can increase the overall conductance of graphene-like materials . It was therefore investigated if such doping could be achieved by simply changing the gas environment during the scribing process. Therefore, the laser cutter was flooded through a port in the chamber with either nitrogen or argon during operation. While this simple approach admittedly did not create a perfect gas atmosphere, it can quickly be realized in any lab. The LIG surfaces prepared under argon and nitrogen became very hydrophobic with water contact angles of 170° and 150° respectively, while the sessile drop spread completely on LIG prepared under ambient atmosphere. The gas environment influenced the microstructure (Fig. S9) and also the Raman spectrum (Fig. S10) but oddly seemed not to cause significantly different outcomes in cyclic voltammetry (Fig. S11). In terms of large-scale fabrication, this indicates that the ambient atmosphere is sufficient to produce high-quality electrodes. More in-depth investigations regarding the influence of lasing atmosphere on spectral characteristics and hydrophobicity of LIG were published by Li et al. and Mamleyev et al. [12, 25]
Infrared reflectance spectra of the Kapton substrate and LIG are displayed in Fig. S5A. The features of polyimide in the fingerprint region between 600 and 1800 cm−1 completely disappeared and the overall transmission dropped profoundly across the whole spectrum after carbonization. Both observations confirm the chemical transformation to carbonaceous LIG.
The carbon content increased from roughly 68% in Kapton to above 93% in LIG (Fig. S5B), indicative of carbonization, while contents of hydrogen and nitrogen decreased from 3 and 7% to values below 1%, likely being released as gases during the scribing process. The oxygen content dropped to 5% after scribing, which points toward the presence of oxygen-containing groups in the carbon lattice of LIG, which were also found via XPS analysis (Fig. S7). The Raman spectrum of LIG features the characteristic D, G, and 2D peaks known from graphene-like materials (Fig. S6).
SEM micrographs of porous LIG of the type 1/10/1000 × 1000 are shown in Fig. 6. A pattern of horizontal trenches is visible at low magnification (Fig. 6b), which was created when the pulsed laser beam passed over the substrate in successive lines from top to bottom with a pitch of about 25 μm. Given this pitch and the beam diameter of approximately 125 μm, each location on the surface specified by the pattern was irradiated an average number of 18 times because of laser beam overlap in x- and y-directions (see also Fig. S8).
As seen in the cross-sectional view in Fig. 6d, a significant part of the Kapton substrate, around 100 μm, remains intact after lasing and serves as a support for the more delicate coral-reef-like LIG structure which has an average height of 27 ± 3 μm as determined by SEM.
SEM pictures of LIG obtained at other scribing conditions revealed significantly different morphologies. This observation has already been disclosed by others, e.g., Tiliakos et al. identified five different morphic groups of LIG with differences in Raman spectra, electrical conductivity, and wettability . They used a galvanometric laser processing unit which permitted a high degree of variation in scan rate and pulse frequency. On the other hand, Duy et al. reported the creation of long LIG-fibers (LIGF), by decreasing the pulse frequency while at the same time using a smaller beam diameter, effectively reducing the beam overlap . The commercial flatbed laser processing unit used by their group is similar to the one in this study and we obtained LIGF by reducing pulse densities to 500 PPI but still using a beam diameter of 125 μm (Fig. S12). Apparently, the fibrous structure was also created when a certain beam overlap occurs, in this case approx. 4 times. In fact, even at 1000 × 1000 PPI, brush-like LIG structures could be observed, given the right power/speed combination, as seen in Fig. 4f above. Correlating the morphology to the electroanalytical performance in Fig. 4e, it can be deduced that large brush-like structures are not favorable for electroanalysis as likely looser structures lead to a higher resistance and hence worse electroanalytical performance. A possible gain in overall surface area hence does not translate here into better electrodes for analysis.
The electrochemically active surface area (ESA) of LIG electrodes, determined with the voltammetric method via the Randles-Sevcik equation, was 0.107 cm2 (about 1.8 times the geometrical surface area AGEO, Fig. 7b) and the calculated effective heterogeneous electron transfer coefficient (k0,eff) for the [Fe(CN)6]3−/4− couple was 0.003 cm s−1. For comparison, commercial screen-printed carbon electrodes exhibited a lower calculated ESA of 0.9 times AGEO and a comparable k0,eff of 0.002 cm s−1. However, the Randles-Sevcik relationship—like the Nicholson method for the determination of k0,eff—is only strictly applicable to smooth electrode surfaces with planar semi-infinite diffusion of dissolved redox species. Therefore, the values of ESA and k0,eff reported here should be regarded as an estimate rather than accurate. The specific surface area of LIG 1/10/1000 × 1000 determined by nitrogen adsorption isotherm analysis was approx. 330 times the geometrical surface area. Obviously, this does not correlate to the determined ESA. We assume that the electrochemical experiment solution may not enter all of the pores and cavities due to hydrophobic pouches and that some areas are not conductively connected well enough, which leads to this dramatic difference in measurements.
Table 1 compares the values of ESA, k0,eff, and sheet resistance to previously published data for LIG or very similar material (LSG). The values we report here are in the spectrum of previously reported results.
Finally, electrochemical impedance spectroscopy revealed extremely low impedance values compared to commercial screen-printed carbon electrodes of the same geometrical size (Fig. S13) We conceive that especially this characteristic will make LIG an interesting material for EIS sensors as also demonstrated by us and other groups previously [29, 31, 32].
Voltammetric applications of LIG electrodes
The electrochemical behavior of various molecules of interest on LIG electrodes was investigated by CV to test the broad applicability of this electrode type for chemical sensing (Fig. S15). We observed that [Ru(NH3)6]3+, a representative for molecules that undergo outer-sphere electron transfer, was just as easily detected as [Fe(CN)6]3−, which is classified as a more surface-dependent redox species (Fig. S15A) . The detection of dopamine (DA) can generally be hindered by the presence of ascorbic acid (AA) or uric acid (UA) which oxidize at very similar potentials. In Fig. S15B, the peaks of DA, AA, and UA are sufficiently separated on LIG as also shown by our collaborators earlier . This beneficial effect might be partly explained by the transition from a planar semi-infinite diffusion regime to thin-layer diffusion, as Compton’s group has pointed out in the past about the modification of glassy carbon electrodes with nanomaterials . Figure S15C and D demonstrate the detection of p-nitrophenol and paracetamol, which could be direct analytes of interest, while a CV of the common redox mediator methylene blue (MB) is shown in Fig. S15E. MB adsorbs easily onto LIG, as indicated by the very low peak separation and increased currents in consecutive scans and, in fact, Rathinam et al. already suggested LIG powder as adsorbent for MB in water treatment . Since MB is also used as an electron mediator in biosensors, the strong adsorption may be beneficial in that case.
Generally, an overall high background capacitance due to the large inherent surface area can be observed on LIG electrodes, paired with an excellent ability for oxidation and reduction reactions of inner and outer-sphere electroactive species. We observed that the current response of different redox species was significantly larger on LIG compared to commercial screen-printed carbon or glassy carbon electrodes. This may be caused by not only the porous nature of LIG but also the apparent presence of many reactive edge sites (see the large D peak in the Raman spectrum of Fig. S6) which may cause this improved electrocatalytic activity . This demonstrates the overall utility of LIG as sensitive electrode material for analytical applications.
Finally, we compared the detection of [Fe(CN)6]3− on the chosen LIG electrode type using chronoamperometry (CA), cyclic voltammetry (CV), and square-wave voltammetry (SWV). Detection via CV below a concentration of 50 μM was not possible (Fig. 8c, d), the lowest detectable concentrations with CA and SWV were 25 μM and 5 μM respectively. Upwards, the response was linear to the highest tested concentration of 500 μM with CA and CV, while SWV allowed a linear calibration up to 100 μM with signals increasing less at higher concentrations (only linear part of calibration displayed in Fig. 8f).
The SWV calibration of K3[Fe(CN)6] on LIG electrodes in Fig. 8e, f can be compared to data from commercially available screen-printed carbon electrodes (Dropsens, DRP-110) in Fig. S14. Our LIG electrodes express a similar limit of detection (LOD and LOQ with SPE = 1.0 and 3.0 μM) although the screen-printed electrodes show less background current.