Co-porphyrin functionalized CVD graphene ammonia sensor with high selectivity to disturbing gases: hydrogen and humidity

A CVD graphene on thermally oxidized Si substrate was purchased from Graphene Platform (http://grapheneplatform.com/jp/). The oxide thickness was 90 nm. After cleaning in acetone and 2-propanol, electrodes were formed by electron-beam evaporation of Ti/Au with a shadow mask. A schematic diagram of the device structure is shown in Fig. 1(a). To functionalize the CVD graphene, CoOEP saturated chloroform solution was spin-coated onto the graphene at 2000 rpm for 60 s on a graphene.³⁸⁾ Optical microscope images of the graphene before and after CoOEP functionalization are shown in Figs. 1(b), 1(c). The inset of Fig. 1(c) shows the SEM image of CoOEP functionalized graphene. These images confirmed that the small disk-like patterns of CoOEP were successfully functionalized.

Zoom In Zoom Out Reset image size

The size and morphology of CoOEP was characterized by AFM. The top views of AFM are shown in Figs. 2(a) and 2(b). As shown in Fig. 2(a), small particles were observed and we consider that they were multi-layer graphene domains.³⁹⁾ On the other hand, it can be seen from Fig. 2(b) that the CoOEP crystal are functionalized on the graphene. A comparison of the height distribution before and after functionalization is shown in Fig. 2(c). The shift of height distribution peak after functionalization indicates that CoOEP was functionalized to graphene, which induces the surface roughness.

Zoom In Zoom Out Reset image size

Gas sensing measurement setup is schematically shown in Fig. 3. The concentration of NH₃ and humidity of the test gas were controlled using MFC and flow meter. Dry air-based gas was used to measure the gas sensing properties. In some experiments, the concentration of NH₃ was confirmed using gas chromatography.

Zoom In Zoom Out Reset image size

To investigate the elemental composition and chemical state of CoOEP functionalized graphene, X-ray photoelectron spectroscopic (XPS) measurements were used. The XPS spectra were acquired using Quantera SXM (ULVAC-PHI) equipped with monochromatic AlKα source. As shown in Fig. 4(a), the wide-scan of CoOEP functionalized graphene showed the N1s, Co2p3/2, and Co2p1/2 peaks, with the binding energy of 398 eV, 780 eV and 795 eV, respectively. Furthermore, in order to investigate the chemical states of elements, narrow scans were performed on the C1s, N1s and Co2p. After deconvolution, it was revealed that the C1s spectra of CoOEP functionalized graphene contains five peaks at 283.9, 285.3, 286.3, 287.7, 289.0 eV as shown in Fig. 4(b). One at binding energy 283.9 eV features the C=C sp2 bonds in the graphene network.⁴⁰⁾ The peak at 285.3 eV is attributed to the C–C sp3 bonds.⁴⁰⁾ Two components with binding energies at 287.7, and 289.0 eV are attributed to the presence of PMMA residues (e.g. C=O, O=C–O).⁴¹⁾ Also, the peak at 286.3 eV is attributed to pyrrole rings in CoOEP.⁴²⁾ In previous works, it was reported that two peaks appeared in the N1s peak in the non-metal porphyrin.^43–46) When cobalt ion is inserted into porphyrin, these two peaks collapse into a single peak.^43,46) The binding energy of this single peak is attributed to Co–N bond. The binding energy was reported as 398.8 eV.⁴⁶⁾ This value corresponds reasonably well with our experimental data shown in Fig. 4(c). Figure 4(d) shows the result of deconvolution of Co2p XPS spectra. From this data, Co was present in the state of Co²⁺, and it was revealed that the Co²⁺ is coordinated to porphyrin.^43,46)

Zoom In Zoom Out Reset image size

Figure 5(a) shows the typical time dependence of sensor response to target gas. The sensor response was defined as

$$\begin{eqnarray}&&\left({\rm{Sensor}}\,{\rm{Response}}\right)=\displaystyle \frac{R-{R}_{0}}{{R}_{0}}=\displaystyle \frac{{\rm{\Delta }}R}{{R}_{0}},\end{eqnarray} \tag{ 1 }$$

where R was the time-dependent resistance and R₀ was the initial resistance at the time when the device started to be exposed to the test gas. The drain voltage (${{V}}_{0}$) was fixed at 0.01 V and the operating temperature of the sensor was kept at room temperature.

Zoom In Zoom Out Reset image size

Figure 5(b) shows the comparison of sensor responses to 2 ppm NH₃ between the CoOEP functionalized graphene and a non-functionalized graphene. The sensor response of CoOEP functionalized graphene was approximately six times greater than that of the non-functionalized graphene. The improvement of the sensor response originates from a difference in adsorption site to NH₃ between CoOEP functionalized graphene and non-functionalized graphene. In previous reports, non-functionalized graphene sensors were n-doped by physical adsorption of NH₃ to the graphene and chemical adsorption to the defects of the graphene.^{30,31,33,47,48)} On the other hand, it has been reported that a Co atom of Co-porphyrin combined with NH₃ as a ligand.⁴⁹⁾ Therefore, CoOEP functionalized graphene has greater adsorption sites than non-functionalized graphene and is more strongly n-doped during NH₃ mixture gas exposure. It also has been reported that the graphene in the atmosphere is p-doped due to the influence of oxygen, water and SiO₂ substrate.^33,50) As a result, the resistance of graphene is increased by n-doping.

Figure 5(c) shows time-dependent sensor responses of CoOEP functionalized graphene for various NH₃ concentrations ranging from 0.04 to 0.79 ppm. The concentration of NH₃ was determined from gas chromatography. Even at the lowest NH₃ concentration of 0.04 ppm, a clear sensor response of 0.23% was observed.

Figure 5(d) shows the NH₃ concentration dependence of the sensor response. The relationship is fitted by the empirical function: 2.7x^0.835, where x is the NH₃ concentration in ppm unit. Exhaled breath of healthy human contains NH₃ with concentrations ranging from 0.32 to 1.08 ppm.⁸⁾ Thus, our sensor showed enough sensitive to NH₃ for breath NH₃ detection.

Since human breath contains several-hundred-ppm H₂ and both H₂ and NH₃ are reductive, H₂ can be disturbing gas for accurate detection of NH₃. Figure 6 shows the selectivity of the CoOEP functionalized graphene sensor against H₂. In the experiment of Fig. 6(a), the concentration of NH₃ was changed from 0 to 1 ppm during the time-dependent measurement, while the concentration of H₂ was fixed at 400 ppm. As shown in Fig. 6(a), our sensor did not show any response to H₂ even under disturbing atmosphere where H₂ with concentration of four hundred times greater than NH₃ concentration was included. On the other hand, our sensor successfully recognized low concentration (1 ppm) NH₃. In the experiment of Fig. 6(b), the concentration of H₂ was changed from 0 to 500 ppm while the concentration of NH₃ was fixed at 1 ppm. As shown in Fig. 6(b), since the sensor response were consistent in all cases, it was confirmed that H₂ did not affect the sensor response to NH₃. Therefore, we demonstrated the successful recognition of NH₃ with almost no-sensitivity against H₂ in CoOEP functionalized graphene sensors.

Zoom In Zoom Out Reset image size

In addition to the effect of H₂, we also need to consider the effect of humidity. Human exhaled breath contains tens of thousands of parts per millions of water in absolute humidity.⁵¹⁾ It is necessary to develop gas sensors which detect the test gas under high humidity in order to carry out the breath diagnosis. Figure 7(a) shows the comparison of sensor responses to 1 ppm NH₃ between the CoOEP functionalized graphene and a non-functionalized graphene under high humidity atmosphere: relative humidity (RH) 80%. It is confirmed that CoOEP functionalized graphene can detect low concentration NH₃ even under high humidity. This result suggests that CoOEP preferentially adsorbs NH₃ rather than water. In addition, compared with the results in Figs. 5(b)–5(c), the recovery characteristics was improved. This is because when the exposure of the NH₃ mixture is stopped, NH₃ desorption occurs and the graphene is p-doped by water adsorption. Figure 7(b) shows the sensor response to NH₃ under rapid humidity changes. The blue data shows the sensor response when the gas with 50% RH was exposed and the orange data shows the sensor response when the gas with 50% RH mixed with 1 ppm NH₃ was exposed. As shown in the blue line, in the case of humid air, graphene was p-doped by water. On the other hand, when the mixture gas of NH₃ and humid air was exposed, the effect of humid air was first appeared and then the effect of NH₃ was appeared. From these date, it is confirmed that CoOEP functionalized graphene sensor can detect NH₃ even under rapid humidity changes. Figure 7(c) shows the result of long-term gas sensing measurement: 6 h under high humidity atmosphere. The black solid line represents the sensor response toward NH₃. The orange dotted line represents the NH₃ concentration measured by gas chromatography. We demonstrated that the sensor of this study show good robustness against the humidity for a long time.

Zoom In Zoom Out Reset image size

We consider that there are two key factors for gas sensing. First, the graphene is p-doped before exposing NH₃ to the sensor. In a previous report, it was reported that graphene on SiO₂ was p-doped in air.⁵⁰⁾ This effect is caused by the oxygen anions, which are generated by hole doping to graphene. These anions are stabilized by the presence of water.⁵⁰⁾ Moreover, it has also been reported that graphene is p-doped by functionalization of CoOEP. In a previous report, 0.6 electrons were redistributed from the graphene to CoOEP using the dispersion corrected DFT in VASP.⁵²⁾ The charge transfer is caused by a large wave function overlap between Co atom's d-electrons and graphene's $\pi$-electrons.⁵³⁾ It has been reported that in order to reduce the overlap, the interface electronic structure of metal-graphene complexes was modified and push some electrons into metal from graphene.^53,54) In fact, it has been reported that graphene is chemisorbed on Co metal which acts as a donor for Co metal.^55,56) This phenomenon is called "push-back" effect.⁵⁴⁾

Next, the graphene is n-doped by NH₃ adsorption on CoOEP. We consider a hybridization between the Co d-orbital and NH₃ HOMO. In previous reports, for isolated Co-porphyrin molecule, Co-porphyrin has a cobalt in a divalent state (d₇ configuration: ${\left({{\rm{d}}}_{xz}\right)}^{2}{\left({{\rm{d}}}_{yz}\right)}^{2}{\left({{\rm{d}}}_{xy}\right)}^{2}{\left({{\rm{d}}}_{{z}^{2}}\right)}^{1}{\left({{\rm{d}}}_{{x}^{2}-{y}^{2}}\right)}^{0}$^52,57–59)). Overall, the number of valence electrons of the Co-porphyrin is 15:7 electrons from Co²⁺ and 8 electrons from the porphyrinate moiety.⁵⁹⁾ The ${{\rm{d}}}_{{x}^{2}-{y}^{2}}$ of Co is split by the coupling with the symmetric orbitals of the ligand: 4N.⁵⁹⁾ In the case of Co-porphyrin on graphene, due to the asymmetry induced by the graphene substrate, the d-orbital of Co is broadening by coupling with graphene p_z orbital.^52,58) In this state, when NH₃ approaches as an axial ligand of the Co-porphyrin, the NH₃ HOMO and the Co ${{\rm{d}}}_{{{z}}^{2}}$-orbital are hybridized and the orbital are split. The interaction of the split orbital with the graphene p_z orbital causes a change in the electronic structure of Co-porphyrin-graphene complexes.

In order to check the coupling of electron orbitals, we have performed first-principles density functional calculation with SIESTA.⁶⁰⁾ A 5 × 5 × 1 k-point Monkhorst–Pack grid was taken in the Brillouin zone. The local density approximation (Perdew/Zunger⁶¹⁾) and the mesh cut off energies of 300 Ry were used. The structure has been relaxed until the forces exerted on all atoms were less than 0.05 eV Å⁻¹. Figures 8(a)–8(c) show the optimized geometries of graphene/Co-porphyrin, graphene/Co-porphyrin/NH₃, Co-porphyrin, respectively. Since the interaction between graphene and the central part of CoOEP is important to explain the sensing mechanism, only the central part of CoOEP was considered in this calculation [Fig. 8(c)]. Figure 8(d) shows the projected density of states (PDOS) for graphene/Co-porphyrin. It was confirmed that parts of d_xy and ${{\rm{d}}}_{{z}^{2}}$ orbitals of Co atom were located above the Fermi energy level. The d_xy orbital interacts with the orbitals of N atoms in the pyrrole group, causing the energy split. Meanwhile, the ${{\rm{d}}}_{{z}^{2}}$ orbital interacts with the carbon-p_z orbital of graphene. Figure 8(e) shows the PDOS after a NH₃ adsorption to the graphene/Co-porphyrin. It was confirmed that the d_z² orbital of Co is split [arrow portion of the Fig. 8(e)] because of the interaction with the N atom's orbital of the NH₃. Figure 8(f) shows the comparison of the PDOS of graphene before and after the NH₃ adsorption. The fermi energy shifts with respect to the dirac point, which means that n-doping is induced for graphene²⁰⁾ [Fig. 8(g)].

Zoom In Zoom Out Reset image size

The performance of our sensors was compared with previous reports. As shown in Table I, NH₃ sensors have been fabricated using pristine graphene and Pt functionalized graphene. However, a high selectivity to NH₃ in comparison with disturbing gases has not been achieved. It is reported that the influence of disturbing gases such as H₂ and humidity on the sensors can be suppressed by using CoWO₄ and CuBr, which are known to detect NH₃ selectively. In this work, the excellent selectivity was verified by using CoOEP, and in addition, it was shown that NH₃ detection was possible under the rapid humidity change situation.