Single-step flame synthesis of carbon nanoparticles with tunable structure and chemical reactivity

Abstract

A novel method for the flame synthesis of carbon nanoparticles with controllable fraction of amorphous, graphitic-like and diamond-like phases is reported. The structure of nanoparticles was tailored using a conical chimney with an adjustable air-inlet opening. The opening was used to manipulate the combustion of an inflamed wick soaked in rapeseed oil, establishing three distinct combustion regimes at fully-open, half-open and fully-closed opening. Each regime led to the formation of carbon coatings with diverse structure and chemical reactivity through a facile, single-step process. In particular, the fully-closed opening suppressed most of the inlet air, causing an increased fuel/oxygen ratio and decreased flame temperature. In turn, the nucleation rate of soot nanoparticles was enhanced, triggering the precipitation of some of them as diamond-like carbon (DLC). Surface characterization analyses using Raman spectroscopy, X-ray photoelectron spectroscopy and transmission electron spectroscopy confirmed this hypothesis, indicating a short-range ordered nanocrystalline structure and ∼80% sp³ bonds in the coatings deposited at fully-closed opening. Furthermore, three groups of 5 MHz Quartz Crystal Microbalances (QCMs) coated with soot and DLC, corresponding to each of the three combustion regimes, showed different frequency responses to aqueous ethanol and isopropanol solutions in the concentration range of 0–12.5 wt%. The DLC coated QCMs exhibited relatively constant frequency shift of ∼2250 Hz regardless of the chemical, while the response of soot coated counterparts was influenced by the quantity of heteroatoms in the film. Our method can be applied in chemical sensing for the development of piezoresonance liquid sensors with tunable sensitivity.

---

1. Introduction

Carbon is one of the most remarkable chemical elements, as it is capable of forming a variety of chemical bonds with itself and/or atoms of other elements.¹ Its physicochemical characteristics depend on the structural configuration of the atomic bonds (sp¹, sp² or sp³) and due to these natural peculiarities, carbon can be found in different forms such as diamond, graphite, fullerenes and amorphous carbon.^2,3 The latter is a metastable phase considered as a mixture of highly disordered carbon atoms with different fractions of sp³, sp² and even sp¹ bonding.¹ A major advantage of the amorphous carbon is the ability to exhibit different physicochemical properties by altering the ratio of sp²/sp³ bonds and the quantity of heteroatoms² (e.g. oxygen). For instance, amorphous carbon films that exhibit a short-range ordered nanocrystalline structure and significant sp³ content, known also as diamond-like carbon (DLC), are characterized with enhanced density, wear resistance, chemical inertness and optical transparency.^4–8 On the other hand, the increased amount of sp² bonds can transform the coating into a graphite-like carbon with high porosity, leading to a large specific surface area and improved chemical reactivity.^9–11 Therefore, the amorphous carbon coatings have strong potential for a wide range of practical applications, including electrochemical energy storage,¹² active catalysts for the hydrolysis of cellulose,¹³ chemical sensors,^14,15 photovoltaic solar cells¹⁶ or artificial knee–hip bioimplants.^17,18

For each particular application, the content of sp² and sp³ bonds in the coating along with the quantity of heteroatoms can be adjusted to provide the desirable physicochemical characteristics.² The best way of implementing this concept is through direct activation,¹⁹ carbonization of crosslinked polymers,²⁰ chemical vapor deposition,²¹ pulsed laser deposition,²² ion beam/magnetron sputtering^23,24 or glow discharge RF plasma treatment.²⁵ These techniques are efficient and accurate; however, each of them has specific disadvantages in terms of the deposition rate, film's quality and uniformity, as well as the necessity of expensive equipment (lasers, plasma reactors, chemical chambers, etc.). Furthermore, most of the aforementioned procedures require precise control of the experimental conditions such as vacuum and pressure, which determines the need for specially-designed hermetically sealed chambers. In contrast, the deposition of amorphous carbon films through combustion flame synthesis at atmospheric pressure is a method of fascinating simplicity.^26–35 The flame ensures a chemically reactive environment capable of generating carbon nanostructures in a short and continuous single-step process.³⁶ Moreover, it has been demonstrated that candle flame consists of four major forms of carbon (diamond, graphite, fullerenes, amorphous carbon), which can be successfully identified using anodic aluminum oxide collectors.³⁰ Unfortunately, this approach has not yet been extensively used for industrial purposes, because of a few limiting factors. Firstly, to achieve desirable physicochemical properties of the flame-deposited carbon coatings, additional catalysts or chemical reagents may be required, which complicates the process.^34,37–39 Secondly, up to now, the formation of amorphous DLC from various flame configurations and fuel types has been observed only at high substrate temperatures (above 400 °C), which in turn can limit the applicability of the method to materials with low thermal stability.^27,28,32,40

Here, we present an efficient single-step flame method for the deposition of carbon coatings, whose physicochemical characteristics can be easily manipulated using a conical chimney with an adjustable air-inlet opening. In this study, carbon nanostructures with superhydrophobic or diamond-like properties are derived during the incomplete combustion of rapeseed oil. This is achieved through a precise control of the amount of oxygen involved in the combustion, and subsequently the temperature of the flame, by changing the size of the opening. The major advantage of our approach is the opportunity to tune in situ the fraction of amorphous, graphitic-like and diamond-like phases, allowing for the deposition of carbon coatings with substantially different structure and chemical reactivity. Moreover, the proposed method is catalyst-free and does not require high substrate temperatures (see Section 3.1), which is of crucial importance for its wide practical applicability.

2. Experimental procedure

2.1. Synthesis of the carbon nanoparticles and experimental details

A custom-designed aluminum chimney with an adjustable air-inlet opening, illustrated in Fig. 1, was mounted over an inflamed paper wick soaked in rapeseed oil. The size of the opening was controlled through a circular cover with a diameter of d = 6 cm, which was wrapped around the chimney. This cover was used to tune the inlet oxygen flowing through the narrow 1.5 × 2.5 cm opening, available at the bottom of the chimney. Altering the position of the cover towards the opening allowed in situ manipulation of the combustion process and subsequent synthesis of carbon nanoparticles with different fraction of amorphous, graphitic-like or diamond-like phases.

Fig. 1 Schematic representation of the conical chimney with adjustable opening used to synthesize the carbon nanoparticles. The units are given in centimeters.

Based on the experimental setup, three distinct combustion regimes were established when the opening was fully-open (1.5 × 2.5 cm), half-open (0.75 × 2.5 cm) and fully-closed. The latter cancelled most of the oxygen flow, but since the chimney was not completely sealed (upon closure, a small ∼0.5 cm gap remains between the cover and chimney's bottom); the combustion remained continuous, as there was enough oxygen to make the process self-sustaining. After ignition of the fuel (rapeseed oil), square shaped 2.5 × 2.5 cm microscope glass slides (Fisher Scientific, USA), as well as gold electrode quartz crystal microbalances (QCMs) with a fundamental frequency of ∼5 MHz (SRS, USA) were exposed over the chimney's tip at each regime of combustion, which caused deposition of carbon coatings with various physicochemical characteristics. The film deposition was carried out at burner-to-substrate distance of 7 cm with time duration ranging within 20–60 s, similarly as in our previous work.³⁵ The flame and substrate temperatures were determined using a TP3001 digital thermometer and Kintrex IRT0421 infrared thermal sensor, respectively. For that purpose, the probe of digital thermometer was placed in the flame, keeping similar distance from the burner (the chimney) such as during the film deposition. The flame temperature was recorded after reaching stable temperature reading, which was obtained for ∼60 s. In addition, the substrate temperature at each combustion regime was measured through the infrared thermometer coincidently with the deposition process.

2.2. Surface characterization

After the synthesis of particles and subsequent fabrication of the coatings, their morphology, structure, roughness and chemical composition were investigated using scanning electron microscopy (SEM), X-ray diffraction (XRD), atomic force microscopy (AFM), Raman spectroscopy, X-ray photoelectron spectroscopy (XPS) and transmission electron microscopy (TEM). The SEM experiments were performed using a Hitachi SU-70 Field Emission Scanning Electron Microscope and images were taken at low and high magnifications up to 150k. The XRD measurements were carried out with a Panalytical X'Pert Pro diffractometer operating in Bragg–Brentano mode. An incident X-ray beam was generated with Cu Kα radiation (λ = 1.54 Å) and the samples were scanned from 5° to 60° of 2θ, at 0.03° scan step size, 2° anti-scatter slit, 1° fixed divergence slit and 15 mm mask. AFM images were taken in tapping mode in an area of 1 × 1 μm at a rate of 0.4 Hz with Bruker BioScope Catalyst. Raman spectra of the coatings, deposited at the three distinct regimes, were recorded from 500 to 2500 cm⁻¹ with an acquisition time of 300 s in a Horiba LabRam HR Evolution Confocal Raman spectrometer, using a 20 mW/532 nm He–Ne laser excitation system. TEM was implemented by a Zeiss Libra 120 system operating at 120 kV with a point to point resolution of 0.34 nm. The high-resolution XPS data were collected with a Thermo Fisher ESCALAB 250 X-ray photoelectron spectrometer at a step of 0.1 eV.

2.3. Determination of the physical properties of carbon coatings and their thickness

The electrical resistivity (σ), apparent density (ρ), surface wettability and the thickness of as prepared coatings were determined in several experiments. Four point probe analysis was used for quantitative evaluation of the electrical resistivity of the carbon coatings.⁴¹ For this purpose, an auto-mechanical stage with four equally spaced tungsten metal tips was moved in upward and downward direction. Simultaneously, a high impedance current source was used to supply current through the outer two metal tips, while a voltmeter was measuring the voltage across the two inner probes. The apparent density was defined as a ratio of the mass of the coatings, deposited on 2.5 × 2.5 cm glass slides, towards their volume (thickness). The latter was measured using an optical microscope Nikon eclipse LV100, equipped with a motorized stage ProScan II capable of providing precise focus control by moving the Z-axis in steps as small as 20 nm. Finally, the wettability of the samples was determined through static contact angle (SCA) and contact angle hysteresis (CAH) measurements for droplets of de-ionized water using a Drop Shape Analyzer (DSA 25E, Krüss Germany).

2.4. Proof-of-concept experiments

The hypothesis that the physicochemical performance of carbon coatings, including their chemical reactivity, depends on the ratio of sp²/sp³ bonds and the quantity of heteroatoms² was verified experimentally with nine QCM based chemical sensors. Initially, the QCMs with gold electrode structure, 1 inch diameter and a fundamental frequency of ∼5 MHz were separated in three groups of three sensors. The first group devices was coated with carbon nanostructures through combustion flame synthesis at fully-open, the second at half-open and the third at fully-closed opening, respectively (see Section 2.1). For each combustion regime, the deposition time was appropriately selected in order to ensure as reproducible as possible film thicknesses from device to device. Subsequently, the QCMs were mounted one at a time in a quartz crystal holder connected to a sensor oscillator SRS25, used to ensure continuous crystal oscillations, and a QCM200 digital controller with a built-in frequency counter. The chemical reactivity of the coatings was investigated by measuring the frequency response of each individual sensor in air and after covering the sensing surface with organic solvents such as ethanol and isopropanol (99%, Sigma-Aldrich). These chemicals were dissolved in de-ionized water, leading to aqueous solutions in the concentration range of 0–12.5 wt%. By analyzing the differences in the frequency response between each group of sensors, it was possible to assess whether the reduction of oxygen affects the chemical reactivity of the coatings. The experiments were performed at constant room temperature in an open lab; therefore, any possibility for thermally-induced frequency shifts was avoided.⁴²

3. Results and discussion

3.1. Morphological and structural analysis

Fig. 2 compares the structural and morphological peculiarities of the coatings deposited at three distinct regimes of combustion. The SEM images show that the first two regimes trigger the formation of elongated carbon nanoparticles assembled as islands, separated by micro- and nanoscale pores. Moreover, these particles are disordered, tightly bonded and fused, which corresponds to inherently robust and durable amorphous carbon soot with superhydrophobic properties.³⁵ In contrast, the image in Fig. 2c reveals a significantly different orientation, shape and size of the particles. They occur as grain clusters with an approximate size of 200 nm. A similar surface profile has previously been observed for diamond-like carbon (DLC), fabricated using ion beam⁶ and RF plasma deposition,⁴³ although the graininess achieved with these methods is much more pronounced. The comparative analysis of our SEM results with those reported for ion beam⁶ and RF plasma deposition⁴³ suggests that at the regime of fully-closed opening (Fig. 2c) the uncombusted polyaromatic hydrocarbons may have precipitated as amorphous DLC films. This hypothesis is supported by our XRD measurements (see the ESI†), which indicate a mainly amorphous structure of the fabricated coatings.

Fig. 2 SEM images of carbon coatings deposited at combustion regime with (a) fully-open, (b) half-open and (c) fully-closed opening.

Since such a structural transition has not previously been reported and the reaction mechanism of soot formation in flames is not completely elucidated,^44,45 it is extremely difficult to provide an exact and comprehensive scientific interpretation of our observations. However, a fairly reasonable explanation of the soot-DLC transformation may come up from the detailed kinetic modeling of soot aggregate formation in laminar premixed flames.⁴⁶ According to this model, the soot particle morphology is strongly influenced by the interplay between soot's nucleation, aggregation and initial surface growth that depend on the equivalence fuel/oxygen ratio. In the nucleation region of the flame, the freshly nucleated particles collide, which leads to the formation of fractal aggregates. As the nucleation rate diminishes (at low equivalent ratios i.e. large air fraction) and the surface growth becomes prominent, the aggregated particles acquire a spherical shape.⁴⁶ In contrast, upon enhanced nucleation at relatively constant surface growth rate, the morphology of the clusters formed through collisions of the incipient particles is significantly altered. As a result, the degree of particles' overlap increases vastly, possibly inducing structural changes in the soot. Such a phenomenon is associated with the increased equivalence fuel/oxygen ratio (reduced oxygen content), which increases the fraction of uncombusted polyaromatic hydrocarbons and triggers more intensive nucleation.⁴⁶ In our approach, we manipulate the inlet oxygen flow, and thus the reaction temperature, by changing the size of the opening. At fully-open opening, the substrate temperature after 60 s exposure to the flame at burner-to-substrate distance of 7 cm is ∼160 °C and decreases up to ∼60 °C upon closing the opening (keeping the same exposure time and distance). Simultaneously, by passing from the first to third combustion regime the flame temperature decreases from ∼275 °C to ∼200 °C, respectively. This observation correlates well with the kinetic model of soot formation⁴⁶ and the fundamentals of combustion,⁴⁷ as the reduced oxygen content will result in an increased fuel/oxygen ratio and decreased heat production in the subsequent exothermic chemical reaction. Based on the above considerations, it is likely that at fully-closed opening the nucleation rate in the fume enhances, promoting the morphological changes in the soot i.e. its transformation to amorphous DLC.

3.2. Chemical state analyses

The SEM imaging was followed by Raman spectroscopy and XPS analyses, summarized in Fig. 3 and 4.

Fig. 3 Raman spectra of (a) conventional carbon soot (deposited without chimney) and after modification with chimney at (b) fully-open, (c) half-open and (d) fully-closed opening.

Fig. 4 C 1s photoelectron core level of carbon coatings deposited after modification with chimney at (a) fully-open, (b) half-open and (c) fully-closed opening.

Carbon soot fabricated by the conventional (with no chimney) method has been included too in order to assess how each combustion regime (see Section 2.1) affects the chemical bonds in the coatings. According to Fig. 3, the Raman spectrum of amorphous carbon consists of two distinct bands. The first one at 1596 cm⁻¹ corresponds to an ideal graphitic lattice vibration mode and is denoted as G-band, while the second one, situated at around 1360 cm⁻¹, is associated with the A_1g symmetry mode of disordered graphitic lattice located at the edges of the graphene layer and is called D-band.^34,48 Qualitative characterization of the sp²/sp³ fraction in the coatings can be implemented by analyzing the intensity changes of G- and D-peaks. As seen in Fig. 3a, the conventional carbon soot is characterized with almost equal intensity of both peaks and the I_D/I_G ratio is 0.94. This result correlates well with previously reported I_D/I_G values for amorphous black carbon.^34,48,49 The chimney modification leads to a gradual reduction of the D-peak's intensity, which is an indication of carbon coatings with an increased sp³ content.⁴³ Moreover, at fully-closed opening (Fig. 3d), the I_D/I_G ratio is 0.81, corresponding to a coating with ∼14% less defects compared to the conventional carbon soot. Fig. 4 represents the high-resolution XPS to the C1s for the coatings synthesized at fully-open, half-open and fully-closed opening of the chimney. The deconvolution of the C1s regions is achieved with several peaks corresponding to sp² hybrid form of carbon at 284.8 ± 0.4 eV, sp³ hybridized carbon at 285 ± 0.1 eV, hydroxyl groups (C–O) at 286.5 ± 0.6 eV, carbonyl groups (C=O) at 288.9 ± 0.5 eV and π–π* satellite group at 291 ± 2 eV. Upon reducing the chimney's opening and its subsequent closure, the sp²/sp³ ratio in the coatings is altered significantly; from 1.27 at fully-open to 0.048 at fully-closed opening, corresponding to coatings with ∼80% sp³ content. Moreover, the nanostructures deposited at the third regime do not contain π–π* satellite peak associated with an electronic structure rearrangement of transition between the π bonding and π* antibonding states. Also, Fig. 4c shows a more symmetric shape of the C 1s compared to Fig. 4a and b, which is an additional implication that the electronic structure does not suffer rearrangement effects from sp² hybridized bonds. Furthermore, the oxygenated functional groups in the layers increase by a factor of 2 when switching from half-open to fully-closed opening. These observations are in good agreement with the scan survey of the coatings (see the ESI†).

Finally, the as prepared carbon coatings were examined through high-resolution transmission electron microscopy (HRTEM) and selected area electron diffraction (SAD), shown in Fig. 5.

Fig. 5 HRTEM and SAD images of (a) conventional carbon soot (deposited without chimney) and after modification with chimney at (b) fully-open and (c) fully-closed opening.

The TEM images in Fig. 5 represent the morphological features of the coatings obtained with and without chimney. The conventional soot (without chimney) is characterized with a spherical-like morphology and some lattice fringes could be identified for the larger spheres (>20 nm). In addition, the d-spacing is not homogeneous along the structures, suggesting a short-range order. Smaller spheres do not present any order, indicating mostly amorphous phase. Also, the SAD pattern showed in Fig. 5a exhibits a few continuous rings and some diffused halos on the sample, which are consistent with the HRTEM image. Similarly, after modification with chimney at fully-open air-inlet opening, an onion-like shape of the particles is observed for most of the areas in the soot and the SAD pattern shows mainly two diffused halos corresponding to a short-range order (Fig. 5b). In this case, the two rings can be assigned to graphitic-like nanostructures.

In contrast, the TEM image in Fig. 5c illustrates more crystalline structure of the coating deposited at fully-closed air-inlet opening. Denser structures with large nanocrystals embedded in an amorphous phase are observed. This morphology differs from the sphere-like and the rings in the SAD pattern are well defined. The atomic d-spacings in the coating, synthesized at fully-closed opening, can be matched to both graphitic-like and diamond-like structures with less amorphous phase in overall. A summary of these measurements along with the STD d-spacing for diamond and graphite is presented in Table 1. Although graphitic-like structures are still present, the appearance of more sp³ carbon bonds (see Fig. 4c) along with the evidence of nanocrystalline diamond formation (see Table 1) imply that the regime of fully-closed opening tends to form diamond-like carbon (DLC) structures.

Table 1 Comparison between the d-spacing (Å) calculated from SAD patterns for conventional soot and after modification with chimney at fully-open and fully-closed opening, and STD d-spacings for diamond and graphite

d-Spacing measured (Å)			d-Spacing for some C allotropes
Conventional carbon soot	Fully-open opening	Fully-closed opening	Diamond	Graphite
		3.42		3.4
		2.53
		2.42
2.22		2.28
	2.11	2.05	2.06	2.04
1.89		1.77		1.7
		1.47
1.33		1.36
1.26		1.27	1.26
1.12	1.21	1.15		1.2
		1.01	1.07
		0.96	0.98
		0.94

---

3.3. Physical properties of the carbon coatings

After full characterization of the carbon nanostructures, their electrical resistivity (σ), apparent density (ρ), surface wettability and thickness were defined experimentally. The coatings synthesized at fully-closed opening showed an electrical resistivity of σ = 3.6 × 10⁵ Ω cm, which is two orders of magnitude higher compared to the values for the conventional soot (σ = 1–1.2 × 10³). Furthermore, the apparent density of the latter is calculated to be ρ ∼ 0.04 g cm⁻³, similar to the data reported for superamphiphobic layers based on carbon soot.⁵⁰ In comparison, the short-range ordered DLC nanostructures possess density of ∼0.59 g cm⁻³, indicating decreased porosity. Moreover, the third regime promotes hydrophilic behavior of the coatings determined by low SCA and high CAH of 70° and 20°, respectively. In contrast, the soot exhibits superhydrophobicity with SCA and CAH being 155° and 0.5°, respectively. These values are another evidence for the observed soot-to-DLC transformation, since the sp³-hybridized diamond-like carbon is hydrophilic in nature due to its high surface energy dominated by the covalent character of the sp³ bonds.⁵¹ In addition, all carbon coatings demonstrate linear relationship of their thickness towards the deposition time (see the ESI†). The DLC nanostructures have small thickness of ∼10 μm, while the carbon soot coatings are characterized with much larger thickness of ∼75–125 μm depending on the combustion regime (fully-open or half-open opening). This is attributed to the major differences in the film deposition rate by switching from 1^st to 3^rd combustion regime. At fully-open opening, the deposition rate is ∼1.5 μm s⁻¹ and increases up to 2 μm s⁻¹ at the second regime, which accounts for the reduced oxygen content that degrades the efficiency of combustion and produces more soot.⁴⁷ On the other hand, at fully-closed opening the deposition rate is only about 0.25 μm s⁻¹, meaning that at this stage the combustion process is significantly altered, as it is confirmed by the surface characterization analyses.

3.4. Chemical reactivity assessment

The chemical reactivity of the carbon coatings was assessed through the changes in sensor response of three QCM groups, prior to and after immersion in aqueous ethanol and isopropanol solutions. The choice of these chemicals is related to their practical relevance and harmful impact on the human health when ingested above a certain concentration.⁵² Also, both compounds possess similar density, viscosity and surface tension; therefore, the expected differences in the sensor signal from group to group would be ascribed to the quantity of heteroatoms in the film rather than the physical properties of the liquids. Last but not least, as it is pointed in Section 2.4, the carbon nanostructures were deposited in a way ensuring approximately equal film thickness from device to device; thus minimizing the possibility for thickness (mass loading) induced sensitivity deviations that may compromise the validity of the comparisons.^53,54 For the first two regimes, the film thickness is ∼40 ± 3 μm, while at fully-closed opening it is around 10 ± 1 μm. These values along with the density of carbon materials gave relatively similar mass loading expressed through frequency downshifts within 700–870 Hz. Based on the above considerations, the chemical reactivity of the conventional carbon soot is not considered in the research, since this material is inherently brittle and needs additional stabilization using various stabilizers.³⁵ Thus, the overall massloading on the sensor surface, caused by the stabilizers, would change the QCM's sensitivity, which would compromise the comparative analysis.⁵⁴

Fig. 6 displays the liquid phase frequency response of each QCM group towards aqueous ethanol and isopropanol solutions in the concentration range of 0–12.5 wt%. The first major distinction in the chemical reactivity of the samples is identified upon immersion in de-ionized water (0 wt%). As seen, the QCMs coated with carbon soot (fully-open and half-open opening) decrease their resonance frequency with Δf ∼ 150–250 Hz, corresponding to three to five times lower frequency shift in comparison with the theoretical model for an uncoated QCM.⁵⁵ On the other hand, the dynamic resistance of these sensors (not shown here) remains relatively constant. Such resonance behavior is attributed to a phenomenon called “decoupling of the liquid phase sensor response”. This effect occurs due to the strong reflection boundary at the solid–air interface, arisen from the “air plastron” of the superhydrophobic carbon soot, leading to lower amount of energy interacting with the liquid.⁵⁶

Fig. 6 Frequency response of each sensor group to aqueous solutions of (a) ethanol and (b) isopropanol in the concentration range of 0–12.5 wt%.

Moreover, as evident from Fig. 6, the quantity of heteroatoms in the soot coatings is crucial for their chemical reactivity. The nanostructures fabricated at the first combustion regime have twice more polar functional groups (C–OH) compared to those at the second regime (see Fig. 4). Thus, the first group QCMs exhibits higher sensor response to de-ionized water, as more oxygen atoms available on the sensing surface, more hydrogen bonds would be formed (see Fig. 6a and b). Such a process will increase the overall mass loading on the surface, which will result in additional frequency downshift, according to the Sauerbrey equation.⁵⁷ In addition, the sensors coated with DLC decrease their resonance frequency with more than 1800 Hz after immersion in de-ionized water. This observation correlates well with the increased amount of oxygenated functional groups in the DLC nanostructures. As shown by the XPS analysis, the C–O content in the coatings increases by a factor of two upon passing from half-open to fully-closed opening of the chimney.

The second difference in the chemical performance of the coatings is expressed through the overall resonance behavior of the sensors. For instance, the soot coated QCMs (the first two regimes) demonstrate a non-linear frequency shift when the ethanol concentration increases from 0 to 12.5 wt%. Moreover, there is a large 1.3 kHz step change at 4 wt% followed by a relatively constant sensor signal up to 8 wt%. In comparison, the step change to isopropanol is only 400 Hz and further increase in its concentration causes a proportional quasilinear frequency response. Last but not least, regardless of the chosen chemical, the coatings synthesized at half-open opening exhibit weaker chemical reactivity compared to those deposited at fully-open opening. These results imply that the main chemical reaction is of oxygen–hydrogen type; therefore the coatings with less heteroatoms (oxygen) induce lower sensor signal. Also, isopropanol has more hydrogen molecules compared to ethanol, leading to higher sensitivity of the QCM. As seen, the coatings deposited at the first combustion regime exhibit ∼1.6 times higher sensitivity to isopropanol in comparison with that to ethanol (Δf_C2H5OH ∼ 2350 Hz while Δf_C3H8O ∼ 3750 Hz). Similar trend is observed for the layers precipitated at the second regime (see Fig. 6). Furthermore, the concentration of 12.5 wt% appears to be a threshold, at which the soot loses superhydrophobicity and the resonance frequency does not recover its initial value (baseline). This effect is associated with a wetting state transition from the suspended Cassie–Baxter to the “sticky” Wenzel state due to a chemical and structural modification of the surface caused by ethanol and isopropanol⁵⁸ (see also the ESI†). However, the carbon soot could restore its water repellency by additional hydrophobic chemical treatment, which makes this material appropriate for multiple usages in QCM based gas or liquid sensors.^14,15,58

In complete contrast, the DLC coated QCMs recover their resonance frequency with negligible deviations from the baseline within ±1 Hz and demonstrate relatively constant sensitivity regardless of the chosen chemical. Such behavior is expected since the DLC coatings are chemically inert, meaning that their surface structure remains unaltered upon contact with 12.5 wt% of ethanol or isopropanol. Moreover, this material is smooth on a nanometric scale, which reduces the number of active adsorption sites interacting with the liquid. The root mean square roughness (R_rms) estimated through AFM is R_rms = 37 nm, indicating ∼3–3.5 times smoother surface profile compared to the carbon soot^14,15,35 (see the ESI†). Although DLC is characterized with an increased quantity of oxygenated functional groups, the total number of active sites is reduced. Therefore, the quantity of hydrogen molecules interacting with the layer would be the same despite of the molecular weight of the chemicals (ethanol or isopropanol).

For the sake of completeness, we determined the detection limit of the sensors and the results are summarized in Table 2.

Table 2 Sensitivity and detection limit of each QCM group towards aqueous ethanol and isopropanol solutions

QCM group	Target analyte	ΔC (mg mL^−1)	Δf (Hz)	Δf/ΔC (Hz mg^−1 mL^−1)	Detection limit (mg mL^−1 s^−1)
1	Ethanol	121	2135	17.6	0.85
2		121	1920	15.9	0.94
3		121	401	3.3	4.5
1	Isopropanol	121	3462	28.6	0.52
2		121	3201	26.5	0.57
3		121	292	2.4	6.25

---

The sensitivity is defined as the frequency change Δf towards the target analyte concentration change ΔC and along with the sensor's short-term stability determines its detection limit.⁵⁹ In this study, during the liquid-phase measurements, the resonance frequency was stabilized within ±1–5 Hz s⁻¹. Therefore, the noise level was estimated at its maximum of ±5 Hz s⁻¹, while the signal-to-noise ratio was 3 : 1. Since the capacity of the quartz crystal holder is approximately 1 mL, 12.5 wt% of ethanol and isopropanol are corresponding to ∼121 mg mL⁻¹. As evident from Table 2, the soot coated 5 MHz QCMs yield detection limit to ethanol up to 850 μg mL⁻¹, which is ∼2 times higher resolution in comparison to their uncoated QCM counterparts.⁵⁸ Although our sensors do not provide resolution of ng mL⁻¹ or pg mL⁻¹, further optimization of the signal-to-noise ratio would provide lower detection limit.⁵⁹

4. Conclusions

In this paper, we described a novel and efficient single-step flame method for deposition of carbon coatings with substantially different physical and chemical characteristics. This was achieved using a conical chimney with an adjustable air-inlet opening, mounted over an ignited wick immersed in rapeseed oil. The opening was used for in situ manipulation of the combustion process and subsequent fabrication of nanostructures with different content of amorphous, graphitic-like and diamond-like phases. The SEM and TEM analyses revealed the formation of coatings with distinct morphology; from tightly connected and fused nanoparticles, to grain clusters with approximate size of 200 nm and short-ordered nanocrystalline diamond structure. Such a structural diversity was associated with soot-to-DLC transformation triggered by the increase of the equivalence fuel/oxygen ratio. In addition, the Raman Spectroscopy and XPS showed that the DLC coatings are characterized with low D-band intensity and ∼80% of sp³ bonds. Furthermore, these layers were found to possess hydrophilicity, whereas their soot counterparts exhibited superhydrophobicity with SCA and CAH being 155° and 0.5°, respectively. Also, the coatings deposited at fully-closed opening were thinner, denser and less conductive compared to the soot. Finally, major differences in the chemical reactivity of as prepared nanostructures were observed by analyzing the frequency response of three groups of 5 MHz QCMs, upon immersion in aqueous ethanol and isopropanol solutions. The frequency shift of the soot coated sensors was strongly influenced by the presence of polar C–OH groups, as well as the molecular weight of the chemicals. In contrast, the DLC coated QCMs exhibited relatively constant sensor signal regardless of the analyte, which was attributed to their smooth surface and reduced amount of active adsorption sites. Our investigations are prerequisite for rapid and inexpensive fabrication of carbon coatings with custom physicochemical properties.

Acknowledgements

The authors wish to gratefully acknowledge the research group of Prof. Hadis Morkoc at the VCU Department of Electrical and Computer Engineering for implementing the four point probe analysis. Services in support of the project were provided by the VCU Massey Cancer Center, supported in part with funding from NIH-NCI P30CA016059. Also, the startup support from Virginia Commonwealth University under grant 137422 is greatly acknowledged.

References

1. D. R. McKenzie, Rep. Prog. Phys., 1996, 59, 1611–1664.
2. J. A. Menendez-Diaz and I. Martin-Gullon, Activated carbon surfaces in environmental remediation, ed. T. J. Bandosz, Elsevier, 2006, pp. 1–48, ch. 1.
3. J. C. Anqus and C. C. Hayman, Science, 1988, 241, 913–921.
4. M. Chhowalla, J. Robertson, C. W. Chen, S. R. P. Silva, C. A. Davis, G. A. J. Amaratunga and W. I. Milne, J. Appl. Phys., 1997, 81, 139–145.
5. J. Robertson, Mater. Sci. Eng., R, 2002, 37, 129–281.
6. P. K. Chu and L. Li, Mater. Chem. Phys., 2006, 96, 253–277.
7. C. R. Lin, H. M. Chang and C. K. Chang, Int. J. Photoenergy, 2013, 2013, 612163.
8. N. Dwivedi, S. Kumar, J. D. Carey, R. K. Tripathi, H. K. Malik and M. K. Dalai, ACS Appl. Mater. Interfaces, 2013, 5, 2725–2732.
9. D. Liu, P. Yuan, D. Tan, H. Liu, M. Fan, A. Yuan, J. Zhu and H. He, Langmuir, 2010, 26, 18624–18627.
10. F. Xu, Z. Tang, S. Huang, L. Chen, Y. Liang, W. Mai, H. Zhong, R. Fu and D. Wu, Nat. Commun., 2015, 6, 7221.
11. J. Huang, J. Wang, C. Wang, H. Zhang, C. Lu and J. Wang, Chem. Mater., 2015, 27, 2107–2113.
12. D. S. Su and R. Schlogl, ChemSusChem, 2010, 3, 136–168.
13. S. Suganuma, K. Nakajima, M. Kitano, D. Yamaguchi, H. Kato, S. Hayashi and M. Hara, J. Am. Chem. Soc., 2008, 130, 12787–12793.
14. K. D. Esmeryan, T. A. Yordanov, L. G. Vergov, Z. G. Raicheva and E. I. Radeva, IEEE Sens. J., 2015, 15, 6318–6325.
15. K. D. Esmeryan, V. Georgieva, L. Vergov and J. Lazarov, Bulg. Chem. Commun., 2015, 47, 1039–1044.
16. H. Zhu, J. Wei, K. Wang and D. Wu, Sol. Energy Mater. Sol. Cells, 2009, 93, 1461–1470.
17. R. Hauert, K. Thorwarth and G. Thorwarth, Surf. Coat. Technol., 2013, 223, 119–130.
18. C. A. Love, R. B. Cook, T. J. Harvey, P. A. Dearnley and R. J. K. Wood, Tribol. Int., 2013, 63, 141–150.
19. L. Wei, M. Sevilla, A. B. Fuertes, R. Mokaya and G. Yushin, Adv. Funct. Mater., 2012, 22, 827–834.
20. Z. H. Li, D. C. Wu, Y. R. Liang and R. W. Fu, J. Am. Chem. Soc., 2014, 136, 4805–4808.
21. M. Bojiang and Y. Qingxian, Appl. Surf. Sci., 2012, 258, 4750–4755.
22. G. Matenoglou, G. A. Evangelakis, C. Kosmidis, S. Foulias, D. Papadimitriou and P. Patsalas, Appl. Surf. Sci., 2007, 253, 8155–8159.
23. A. Ibenskas, A. Galdikas, S. Meskinis, M. Andrulevicius and S. Tamulevicius, Diamond Relat. Mater., 2011, 20, 693–702.
24. O. Durand-Drouhin, M. Benlahsen, M. Clin and R. Bouzerar, Appl. Surf. Sci., 2004, 223, 269–274.
25. K. D. Esmeryan, E. I. Radeva and I. D. Avramov, J. Phys. D: Appl. Phys., 2016, 49, 025309.
26. W. A. Carrington, L. M. Hanssen, K. A. Snail, D. B. Oakes and J. E. Butler, Metall. Trans. A, 1989, 20, 1282–1284.
27. L. M. Hanssen, K. A. Snail, W. A. Carrington, J. E. Butler, S. Kellogg and D. B. Oakes, Thin Solid Films, 1991, 196, 271–281.
28. I. Doi, M. S. Haga and Y. E. Nagai, Diamond Relat. Mater., 1999, 8, 1682–1685.
29. M. Qu, J. He and B. Cao, Appl. Surf. Sci., 2010, 257, 6–9.
30. Z. Su, W. Zhou and Y. Zhang, Chem. Commun., 2011, 47, 4700–4702.
31. X. Deng, L. Mammen, H. J. Butt and D. Volmer, Science, 2012, 335, 67–70.
32. H. M. Zhu, S. M. Lei and T. C. Kuang, Adv. Mater. Res., 2013, 750–752, 2100–2103.
33. K. Seo, M. Kim and D. H. Kim, Carbon, 2014, 68, 583–596.
34. B. N. Sahoo and B. Kandasubramanian, RSC Adv., 2014, 4, 11331–11342.
35. K. D. Esmeryan, C. E. Castano, A. H. Bressler, M. Abolghasemibizaki and R. Mohammadi, Appl. Surf. Sci., 2016, 369, 341–347.
36. W. M. Merchan, A. V. Saveliev, L. Kennedy and W. C. Jimenez, Prog. Energy Combust. Sci., 2010, 36, 696–727.
37. K. Saito, A. S. Gordon, F. A. Williams and W. F. A. Stickle, Combust. Sci. Technol., 1991, 80, 103–119.
38. R. L. Vander Wal, L. J. Hall and G. M. Berger, J. Phys. Chem. B, 2002, 106, 13122–13132.
39. R. L. Vander Wal, Combust. Flame, 2002, 130, 37–47.
40. H. Zhu, T. Kuang, B. Zhu, S. Lei, Z. Liu and S. P. Ringer, Nanoscale Res. Lett., 2011, 6, 331.
41. A. P. Schuetze, W. Lewis, C. Brown and W. J. Geerts, Am. J. Phys., 2004, 72, 149–153.
42. K. D. Esmeryan, I. D. Avramov and E. I. Radeva, Sens. Actuators, B, 2015, 216, 240–246.
43. V. S. Yadav, D. K. Sahu, M. Singh and K. Kumar, Proc.–World Congr. Eng. Comput. Sci., 2009, 1, 78–81.
44. O. I. Smith, Prog. Energy Combust. Sci., 1981, 7, 275–291.
45. M. Frenklach, Phys. Chem. Chem. Phys., 2002, 4, 2028–2037.
46. M. Balthasar and M. Frenklach, Combust. Flame, 2005, 140, 130–145.
47. S. P. Burke and T. E. W. Schumann, Ind. Eng. Chem., 1928, 20, 998–1004.
48. A. Sadezky, H. Muckenhuber, H. Grothe, R. Niessner and U. Pöschl, Carbon, 2005, 43, 1731–1742.
49. Y. Gao, Y. S. Zhou, W. Xiong, M. Wang, L. Fan, H. Rabiee-Golgir, L. Jiang, W. Hou, X. Huang, L. Jiang, J. F. Silvain and Y. F. Lu, ACS Appl. Mater. Interfaces, 2014, 6, 5924–5929.
50. M. Paven, P. Papadopoulos, L. Mammen, X. Deng, H. Sachdev, D. Vollmer and H. J. Butt, Pure Appl. Chem., 2014, 86, 87–96.
51. Y. Zhou, X. Song, E. Li, G. Li, S. Zhao and H. Yan, Appl. Surf. Sci., 2006, 253, 2690–2694.
52. D. Jammalamadaka and S. Raissi, Am. J. Med. Sci., 2010, 339, 276–281.
53. D. S. Ballantine Jr, S. J. Martin, A. J. Ricco, G. C. Frye, H. Wohltjen, R. M. White and E. T. Zellers, Acoustic wave sensors: theory, design and physico-chemical applications, Elsevier, 1st edn, 1997, pp. 222–230, ch. 5.
54. G. McHale, M. I. Newton and F. Martin, J. Appl. Phys., 2003, 93, 675–690.
55. K. K. Kanazawa and J. G. Gordon, Anal. Chem., 1985, 57, 1770–1772.
56. G. McHale, P. Roach, C. R. Evans, N. J. Shirtcliffe, S. J. Elliott and M. I. Newton, Proc. IEEE Freq. Control Symp., 2008, 698–704.
57. Z. G. Sauerbrey, Physics, 1959, 155, 206–222.
58. K. D. Esmeryan, G. McHale, C. L. Trabi, N. R. Geraldi and M. I. Newton, J. Phys. D: Appl. Phys., 2013, 46, 345307.
59. L. R. Pardo, J. F. Rodriguez, C. Gabrielli, H. Perrot and R. Brendel, IEEE Sens. J., 2005, 5, 1251–1257.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra06436a

---