3.1. Characterization of Pd/MWCNT
XRD analysis method was used to determine the crystallographic structure of Pd/MWCNT. Figure 1 demonstrates the XRD spectrum of Pd/MWCNT. The diffraction peak at 25.7° 2ϴ angle is attributed to the (0 0 2) plane of the hexagonal carbon originating from MWCNT (JCPDS card no. 00-041-1487). The diffraction peaks observed at 2ϴ angles of 39.7°, 45.9°, 67.3°, 81.0°, and 85.4° are indexed cubic Pd (1 1 1), Pd (2 0 0), Pd (2 2 0), Pd (3 1 1) and Pd (2 2 2) facets and It showed very high compatibility with JCPDS card no. 01-087-0641. In addition, the peaks detected at 43.1° and 53.4° 2ϴ angles correspond to the (1 1 0) and (1 1 2) planes of tetragonal PdO, respectively (JCPDS card no. 01-075-0584). The presence of a small amount of PdO in the catalyst structure may be due to the oxidation of Pd during the analysis carried out in an open atmosphere and the incomplete reduction of K2PdCl4. The crystal size of Pd/MWCNT was calculated as 4.06 nm by the Scherer equation.
The surface morphology and elemental composition of Pd/MWCNT were analyzed by SEM-EX method. As can be seen from Figs. 1a and b, the Pd nanoparticles were homogeneously dispersed on the MWCNT. Figure 1c-e shows SEM images of Pd/MWCNT at 100 nm scale. It can be clearly seen that the nano-sized Pd particles are dispersed on the MWCNT without agglomeration. The elemental mapping results of C, Pd, and O containing Pd/MWCNT are given in Fig. 1f-h, while the overlay images of these elements are presented in Fig. 1i. The homogeneous distribution of Pd nanoparticles was also confirmed by the elemental mapping photograph of Pd/MWCNT (Fig. 1h). EDX analysis was performed to confirm the elemental composition of Pd/MWCNT (Fig. 1j). The EDX result of Pd/MWCNT shows that the catalyst consists of carbon, oxygen, and palladium. In addition, elemental percentages by mass for C, O, and Pd were determined as 83.63%, 7.99%, and 8.44%, respectively. The peak of oxygen observed in the EDX spectrum is due to PdO.
The morphology, particle size, and distribution of Pd/MWCNT were investigated by TEM. The image of the MWCNT network is clearly visible from Fig. 3. It has been determined that Pd nanoparticles have spherical shapes and nanometric scales. In addition, Pd adhered to both the outer and inner walls of the MWCNT and showed a homogeneous distribution on the support material. The sizes of the particles observed in each TEM image were determined with ImageJ software and the corresponding particle size distribution histograms are given in the inset. While it was observed that Pd nanoparticles were dispersed in a narrow size range namely 4–8 nm, the average particle size of 300 randomly selected particles was found to be 4.51 nm.
3.2. Electrochemical Evaluation
Before RSM experiments, the optimum Pd loading ratio was determined among 0.5%, 1%, 3%, 5%, 7%, 10%, and 20% Pd/MWCNTs and FAEO performances were recorded in 0.5 M H2SO4 + 1M HCOOH solution and at 100 mVs− 1 scan rate (Fig. 4). It was observed from Fig. 4a that all Pd loading ratio catalysts exhibited a distinct hydrogen adsorption-desorption region in the potential range of -0.2 and 0 V [22, 23]. In the potential range of 0 V and 0.6 V, no electrochemical activity occurred and this potential range was explained by the double-layer region [24, 25]. The areas of the obtained voltammograms increased with increasing metal loading rates and the highest voltammogram area was reached with 20% Pd loading. This is explained by the number of electrochemically active sites increasing with the increase of Pd content. Also, the peak observed at a potential of about 0.49 V in the reverse scan was attributed to the reduction of PdO [26, 27].
Figure 4b and Fig. 4c show voltammograms of Pd/MWCNT in 0.5 M H2SO4 + 1 M HCOOH solution in terms of mass and specific activity, respectively. At low Pd loadings namely 0.5% and 1% Pd loading rates, no obvious formic acid dehydrogenation peak was observed in the forward scan. Rather, a gradual electrooxidation of formic acid resulting from dehydration and dehydrogenation of formic acid was observed. The peak observed in the reverse scan indicates overall FAEO on the clean electrode surface [28, 29]. The low peak intensity observed at 0.47 V potential in the reverse scan for 0.5% and 1% Pd-containing catalysts indicates that these catalysts do not have enough electrochemically active surface area for FAEO. The fact that these catalysts do not have a PdO reduction peak in Fig. 4a supports this approach. Specific activities for 3%, 5%, 7%, 10%, and 20% Pd/MWCNT were determined as 0.18, 0.68, 0.22, 1.53, and 3.91 mA cm− 2 while mass activities were calculated as 48.1, 58.4, 26.7, 252.0, and 184.0 mA mg− 1 Pd (Fig. 4b and c). 20% Pd/MWCNT and 10% Pd/MWCNT exhibited the highest specific activity and mass activity for FAEO, respectively. When the Pd loading ratio was increased from 3–10%, the mass activity increased 4.25 times, while when it was increased to 20%, a sharp decrease in mass activity attributable to agglomeration was observed. Therefore, RSM experiments for FAEO on Pd/MWCNT were carried out with 10% Pd/MWCNT.
Table 2
Electrocatalytic features of Pd/MWCNTs
Catalysts | Specific Activity (mA cm− 2) | Mass Activity (mA mg− 1 Pd) | Peak Potential (V) | Onset Potential (V) |
%3 Pd/MWCNT | 0.18 | 48.1 | 0.10 | -0.06 |
%5 Pd/MWCNT | 0.68 | 58.4 | 0.11 | -0.07 |
%7 Pd/MWCNT | 0.22 | 26.7 | 0.18 | -0.07 |
%10 Pd/MWCNT | 1.53 | 252.0 | 0.14 | -0.08 |
%20 Pd/MWCNT | 3.91 | 184.0 | 0.14 | -0.09 |
3.3. Optimization of electrode preparation conditions
GCE was modified with 10% Pd/MWCNT under the conditions given in Table 2 to determine the optimum electrode preparation conditions for FAEO and electrochemical measurements were performed in 0.5 M H2SO4 + 1 M HCOOH solution. Voltammograms obtained from CV measurements performed at determined experimental points for the optimization of Vs, td, and tu independent parameters are given in Figure S1, and the relevant specific activity values were analyzed with the Design Expert 7.0 software. Experimental results from Design Expert 7.0 show that FAEO on Pd/MWCNT could be well explained by a quadratic model. Accordingly, the specific activity values for FAEO on Pd/MWCNT in terms of coded and actual values are given in Eqs. 3 and 4.
Specific activity for FAEO =
+ 2.20+(1.15*A) +(1.89*B)+ (1.06*C) +(1.06*C*B)+(0.74*C*A)+(1.24*B*A)-(0.46*C2)+(0.94*B2)-(0.72*A2) (3)
Specific activity for FAEO =
+ 0.91837+(0.045185*Vs)-(0.085626*tu)+(0.029042*td)+(7.53977E-003*Vs*tu)+(0.00801246*Vs*td)+( 0.00215921*tu*td)-(0.020506*Vs2)+( 0.00107839*tu2)-(0.00190070*td2) (4)
Table 3
ANOVA table for FAEO on Pd/MWCNT/GCE
Source | Sum of | | Mean | F | p-value | |
| Squares | df | Square | Value | Prob > F | |
Model | 89.32662 | 9 | 9.92518 | 12.22955 | 0.0003 | significant |
A-(td) | 13.25186 | 1 | 13.25186 | 16.3286 | 0.0024 | |
B-(tu) | 35.64878 | 1 | 35.64878 | 43.92552 | < 0.0001 | |
C-(Vs) | 11.33059 | 1 | 11.33059 | 13.96126 | 0.0039 | |
AB | 12.34217 | 1 | 12.34217 | 15.20771 | 0.0030 | |
AC | 4.406342 | 1 | 4.406342 | 5.429382 | 0.0420 | |
BC | 8.929708 | 1 | 8.929708 | 11.00296 | 0.0078 | |
A2 | 1.436475 | 1 | 1.436475 | 1.769988 | 0.2129 | |
B2 | 2.421992 | 1 | 2.421992 | 2.984316 | 0.1148 | |
C2 | 0.58864 | 1 | 0.58864 | 0.725307 | 0.4143 | |
Residual | 8.115733 | 10 | 0.811573 | | | |
Lack of Fit | 4.536979 | 5 | 0.907396 | 1.267754 | 0.4005 | not significant |
Pure Error | 3.578754 | 5 | 0.715751 | | | |
Cor Total | 97.44235 | 19 | | | | |
Statistical analysis of the quadratic equation (Eq. 1) proposed by Design Expert 7.0 was evaluated by analysis of variance and the ANOVA of the Pd/MWCNT modified GCE electrode is presented in Table 2. The F value and p-value of the quadratic model proposed by the software were determined as 12.23 and 0.0003, respectively. The model p-value being less than 0.05 indicates that the proposed model is significant in the 95% confidence interval. In addition, the p-values of model parameters A, B, C, AB, AC, and BC were determined as 0.0024, 0.0001, 0.0039, 0.0030, 0.0420, and 0.0078, respectively. In this direction, it was concluded that the effects of Vs, tu, and td parameters and their binary interactions on FAEO were statistically significant. The p-values of the terms A2, B2, and C2 in the model are 0.2129, 0.1148, and 0.4143, respectively, and are not statistically significant as they are greater than 0.1. However, these model terms were seen to make a positive contribution to the model due to their closeness to the 0.1 p values and were not removed from the model. In this study, statistical parameters such as adequate precision, lack of fit, PRESS and regression coefficient (R2), and model fit plots were used to test the suitability of the model. The lack of fit F value found as 1.27 in the ANOVA test is statistically insignificant, indicating that the proposed model is in good agreement with the experimental data. In addition, PRESS and adequate precision values for the model were determined as 82.84 and 15.5, respectively. The relatively low predicted residual sum of squares (PRESS) value and adequate precision value greater than 4 indicate that the proposed model has good predictive power in the investigated parameter range. In addition, the R2 value for the quadratic model was determined as 92%, indicating that the model has the power to explain 92% of the experimental data.
The plots provided by Design Expert 7.0 are shown in Fig. 5 to test the compatibility of the proposed quadratic model with the experimental data. The distribution of actual and estimated specific activity values for FAEO on Pd/MWCNT modified GCE is given in Fig. 5a. As can be seen from Fig. 5a, the experimental points are scattered around the diagonal. The scattering of the experimental data around the diagonal reveals that the proposed model represents the experimental points well [30]. Figure 5b shows the internally studentized residual vs. normal probability distribution. It is desired that the residues be distributed around the straight line determined by the software, and the distribution observed in Fig. 5b fits this description quite well. The very good distribution of the residuals around the straight line is attributed to the unsystematic random distribution of the errors [31]. The distribution of residuals versus estimated values were shown in Fig. 5c. As can be seen from Fig. 5c, the residues are randomly distributed between the reference values indicated by the red lines. Randomly distributed points within the reference limits are explained by the non-constant variance [2]. The distribution of residuals according to the experimental order given in Table 1 is presented in Fig. 5d to test systematic experimental errors. It was determined that the residuals were not distributed in a certain order according to the experiment run number. The fact that the residuals are not dispersed within a trend indicates the absence of systematic errors and that the specific activity values for FAEO on Pd/MWCT modified GCE are time-independent. The model parameters and Fig. 5 show parallelism with each other, indicating that the proposed model has good predictive power within the studied parameters.
Figure 6 shows the binary interactions of the independent parameters Vs, td, and tu. The binary interaction of tu and Vs parameters for FAEO on Pd/MWCNT/GCE is presented in Fig. 6a. Specific activity increased from 1.25 mA cm− 2 to 5.66 mA cm− 2 when tu was increased from 1 min to 60 min under conditions where Vs and td were 5.25 µl and 20.5 min, respectively. In addition, it was observed that the rate of increase in specific activity after 30.5 min tu value was 3.2 times that of 0-30.5 min. It was concluded that the inability to provide a homogeneous catalyst slurry at low tu values and the ability to obtain a uniform catalyst slurry after 30.5 min tu value are the main reasons for the increased specific activity. The nanoparticles well dispersed in Nafion increased the charge transfer efficiency. The 3d response surface plots of Vs and td are presented in Fig. 6b. When the amount of catalyst slurry transferred onto the GCE was increased from 0.5 µl to 5.25 µl, the specific activity increased from 0.30 mA cm− 2 to 1.98 mA cm− 2 (tu=30.5 and td=20.5 min). After this value, when Vs was increased up to 10 µl, the specific activity continued to increase slowly until 2.80 mA cm− 2 and reached a constant value. The increase in the specific activity for FAEO as Vs increased was attributed to the presence of more active sites for the reaction on the Pd/MWCNT modified GCE. The nearly constant specific activity observed at catalyst slurry amounts greater than 8 µl Vs may be due to mass transfer limitation because of excess catalyst slurry. In Fig. 6c, the effect of the td parameter on the specific activity was examined under the conditions of 30.5 min tu and 5.25 µl Vs. By increasing the drying time of the Pd/MWCNT/GCE electrode system from 1–30 minutes, the specific activity increased from 0.32 mA cm− 2 to 2.62 mA cm− 2 and remained constant until the 40th minute.
Optimization of the parameters tu, Vs, and td was carried out to maximize the specific activity for FAEO using Design Expert 7.0 software. It is aimed to maximize the specific activity at minimum tu, td, and Vs values, and the suggested 10 solutions as a result of numerical optimization are given in Table 4. With these inputs, the optimum values for the parameters Vs, td, and tu were determined as 1.84 µl, 45 min, and 37.05 min, respectively, while the model response value at these points was determined as 2.50 mA cm− 2. At these optimum conditions, the specific activity for FAEO on Pd/MWCNT was 2.67 mA cm− 2 with a deviation of 6.83% (Fig. 7a). Consistent with the fit tests of the model, it was concluded that the proposed quadratic model for FAEO on the Pd/MWCNT/GCE electrode system made a good estimation in the investigated parameter range.
Table 4
Numerical optimization solutions for FAEO on Pd/MWCNT/GCE
Number | Vs (µL) | tu (min) | td (min) | Specific Activity (mA cm− 2) | Desirability |
1 | 1.84 | 45 | 37.05 | 2.50 | 0.945 | Selected |
2 | 1.84 | 45 | 37.27 | 2.50 | 0.945 | |
3 | 1.84 | 45 | 36.57 | 2.50 | 0.945 | |
4 | 1.84 | 45 | 37.68 | 2.49 | 0.944 | |
5 | 1.84 | 45 | 37.85 | 2.50 | 0.944 | |
6 | 1.84 | 45 | 36.21 | 2.50 | 0.944 | |
7 | 1.84 | 45 | 36.07 | 2.50 | 0.944 | |
8 | 1.85 | 45 | 38.75 | 2.49 | 0.944 | |
9 | 1.85 | 45 | 35.29 | 2.50 | 0.944 | |
10 | 1.87 | 45 | 39.96 | 2.50 | 0.943 | |
The results of CV, CA, and EIS performed under optimum conditions are presented in Fig. 7. Figure 7a shows CV profiles of 50 cycles at 100 mV s− 1 scan rate in 0.5 M H2SO4 + 1 M HCOOH. In the forward scan, the peak observed at 0.22 V potential is attributed to formic acid dehydrogenation [32], while the shoulder peak detected at 0.55 V potential indicates that FAEO occurs via the dehydration pathway [33]. It is known that the peaks observed at high potentials for FAEO indicate poisoning caused by carbonaceous compounds accumulating on active sites [34, 35]. In this direction, although the shoulder peak at 0.55 V potential indicates catalyst poisoning, the high intensity of the peak caused by formic acid dehydrogenation indicates a high toxicity tolerance of Pd/MWCNT. The peak at 0.29 V potential in reverse scan shows overall FAEO on the clean electrode surface [23]. The peak intensity in the forward scan (2.67 mA cm− 2) is greater than the current density in the reverse scan (2.27 mA cm− 2), confirming the high CO poisoning tolerance of Pd/MWCNT. Specific activities for FAEO on Pd/MWCNT were determined as 2.67 mA cm− 2 and 0.00023 mA cm− 2 for Cycle 1 and Cycle 50, respectively. In the chronoamperometric curves of the Pd/MWCNT, initial current drops due to CO poisoning were observed in the first 20 seconds and were consistent with the CV results. After 20 seconds, the Pd/MWCNT exhibits an almost constant current (Fig. 7d). Pd/MWCNT showed a long term stability at optimum conditions, and CA and CV results of Pd/MWCNT show parallelism. Figure 7c shows the EIS curves of Pd/MWCNT in 0.5 M H2SO4 + 1 M HCOOH solution at a scan rate of 100 mV s− 1. As the radius of the EIS profiles decreases, the charge transfer resistance decreases and the electrocatalytic activity increases [36]. In this direction, the charge transfer kinetics and electrocatalytic activity of Pd/MWCNT increased with the increment of applied potential. It was concluded from the trend of the impedance profile that FAEO on Pd/MWCNT proceeds through charge transfer kinetics.
Synthesis procedures and surface properties of anode catalysts for high electrocatalytic performance are of primary importance. However, it was noted that the modification of the electrode system prepared with catalysts facilely increased the electrocatalytic activity. An extremely limited number of studies have been reported in the literature on the optimization of electrode preparation conditions to increase electrochemical activity. With the electrode systems Pt/CNT/GCE and Pd/CNT/GCE previously reported by our working group, electrocatalytic performances close to those of complex synthesis procedures and more expensive catalyst systems were observed [31, 37]. It is seen from Table 5 that electrode preparation parameters are not specified in detail in most of the studies reported in the literature. Geng et al. prepared dual-phase PdCu by a one-pot wet-chemical method. The specific activity for FAEO was reported as 1.91 mA cm− 2 with the electrode prepared under 12.4 µg cm− 2 Pd load, 5 µL Vs and 15 min tu electrode preparation conditions [38]. The Pd/FeP anode catalyst was prepared via the thermal annealing method by Bao et al., and they reached a specific activity of 1.75 mA cm− 2 for FAEO at 5 µL Vs, 141.5 µg cm− 2 Pd load and 30 min tu parameter values [39]. In another study, Hu et al. have been reported that specific activities of PdAg/C, Pd-Au/C, Pd-Cu/C, Pd-Pt/C, Pd-Ru/C, and Pd/C at 50 µL Vs and 10% Pd loading rate were 0.8, 1.8, 3.2, 3.3, 1.5 and 0.9 mA cm− 2, respectively. The researchers attributed the higher specific activity of bimetallic catalysts than Pd/C to the electronic state alteration of the Pd surface with the addition of the second metal [40]. In addition, the specific activities of trimetallic electrocatalysts namely PdSnAg/C (Vs=2 µL and Pd load = 40 µg cm− 2) ve Pd70Ag20Ni10 (Vs=5 µL, tu=10, td=30, and Pd load = 10 wt.%) for FAEO was reported as 3.59 and 1.6 mA cm− 2, respectively [41, 42]. The Pd/MWCNT/GCE electrode reported in the current study showed higher specific activity than many of the studies reported on FAEO in the literature. It was concluded that optimization of the preparation conditions of the electrodes using the reported catalysts will have a booster effect on the electrocatalytic activity for FAEO. It has been concluded that these benefits will make a positive contribution to the commercialization of DFAFCs.
Table 5
Literature on electrochemical measurement conditions for DFAFC anode catalyst
Catalysts | Solution CH2SO4/HCOOH | Vs (µL) | Pd load (µg cm− 2) | tu (min) | td (min) | Specific Activity (mA cm− 2) | Refs. |
PdCu/C | 0.5 M/0.5 M | 5 | 12.4 | 15 | n.a | 1.91 | [38] |
Pd/FeP-250 | 0.5 M/0.5 M | 5 | 141.5 | 30 | n.a | 1.75 | [39] |
Pd@F-Gr | 0.1 M/0.5 M | 4 | 113 | n.a | n.a | 2.16 | [43] |
Pd-Pt/C | 0.5 M/0.5 M | 50 | 10 (wt.%) | n.a | n.a | 3.3 | [40] |
Pd-Au/C | 0.5 M/0.5 M | 50 | 10 (wt.%) | n.a | n.a | 1.8 | [40] |
Pd-Ru/C | 0.5 M/0.5 M | 50 | 10 (wt.%) | n.a | n.a | 1.5 | [40] |
Pd/C | 0.5 M/0.5 M | 50 | 10 (wt.%) | n.a | n.a | 0.9 | [40] |
PdBi | 0.5 M/0.5 M | 5 | 20 | n.a | n.a | ca. 4 | [44] |
PdSnAg/C | 0.5 M/0.5 M | 2 | 40 | n.a | n.a | 2.59 | [42] |
Pd70Ag20Ni10 | 0.5 M/0.5 M | 5 | 10 (wt.%) | 10 | 30 | 1.6 | [41] |
Pd/MWCNT | 0.5 M/1 M | 1.84 | 10 (wt.%) | 45 | 37.05 | 2.67 | This work |