Effect of the aromatic precursor flow rate on the morphology and properties of carbon nanostructures in plasma enhanced chemical vapor deposition

Understanding the effects of the synthesis parameters on the morphology and electrochemical properties of nanocarbon layers is a key step in the development of application-tailored nanostructures. In this paper we used an aromatic carbon as a new kind of precursor for the synthesis of carbon based nanostructures by plasma enhanced chemical vapor deposition (PECVD). Complex precursor molecules enable a new degree of influence over the atomic structure of PECVD synthesized carbons. Here, we report on the synthesis and characterization of the nanostructures resulting from varied flow rates of p-xylene used as carbon precursor. By changing the flow rate of the precursor, three different morphologies with graphitic character were synthesized. The resulting structures were carbon nanofibers (CNF), freestanding carbon nanowalls (fCNW) and interconnected carbon nanowalls (iCNW), formed at flow rates of 3 ml h , between 1 and 3 ml h 1 and less than 1 ml h , respectively. Structural characterization by transmission electron microscopy and Raman spectroscopy indicate a lower defect density for the CNF in comparison to the CNW nanostructures. The electrochemical characterization of the oxygen reduction reaction onset potential and effective surface area feature a significantly higher onset at around 171 mV and an electrochemically active surface area of 0.76 mm 1 for the iCNW compared to 196 mV, 0.61 mm 1 and 0.22 mm 1 for the fCNW and CNF, respectively. The similarities in defect density and differences in activity observed for the iCNW and fCNF suggest that the kind of the defects determines the electrochemical properties. Thus, the iCNW was identified as the most appropriate morphology for further investigations.


Introduction
Carbon materials have long been established as electrode materials in diverse electrochemical applications, such as batteries, fuel cells, supercapacitors, for several reasons. [1][2][3] Graphite, the most famous carbon material with sp 2 hybridization, is low-priced and offers good electrical conductivity and high chemical stability which make it a great electrode material. Regrettable, it lacks porosity limiting its potential to reach high energy densities. 4 Activated carbon is a form of amorphous carbon. It is low-priced as well and possesses a huge surface area making it a perfect material for supercapacitors. 5,6 However, the enhancement of surface area in activated carbon is accompanied with increasing of porosity which leads to poorer electrical conductivity of the layers restricting its widespread utilization in diverse electrochemical devices or presuppose the adding of conductive additives. [7][8][9] Graphene, an aromatic monolayer of carbon atoms, features in return both high electrical conductivity and high surface area making it to a material with promising electrochemical properties. However, graphene has a high tendency to restack and agglomerate due to strong p-p interactions between the sheets. This causes a distinct decrease of active surface area as well as porosity and electrochemical performance. 10 The plasma enhanced chemical vapor deposition (PECVD) method represents a promising approach for stepwise growth of vertically aligned carbon nanostructures of different shapes. Different plasma sources have been used to synthesize vertical nanostructures in literature. Most commonly, direct current (DC), microwave (MW) and radio frequency (RF) induced plasma were used. In DC PECVD, the plasma is generated between two electrodes when a sufficient potential is applied. The substrate is usually placed on the cathode. Limitations arise when the substrate is not electronically conductive. In contrast, MW and RF plasma work without electrodes. Between them, MW plasma has the advantage of working at higher pressures, up to several thousand pascal, and has a high deposition rate. On the other side, RF PECVD is a low temperature deposition method and of advantage when coating temperature sensitive substrates. [11][12][13] The vertical alignment of the nanostructures in all three cases results from the electric eld of the plasma. 14, 15 The application of plasma in PECVD method, compared with CVD, greatly decreases process temperature and allows to improve alignment and direction of the nanomaterial growth. 16 The spectrum of nanostructures fabricated by PECVD includes diamond-like carbon, carbon nanotubes, carbon nanobers, carbon nanowalls and other structures. [17][18][19] Simple aliphatic compounds like CH 4 , C 2 H 2 , CF 4 , C 2 F 6 are usually used as the carbon source. [20][21][22] The PECVD growth process enables the realization of well-dened porous electrode architectures with tunable and controllable morphology, in which the irreversible aggregation of nanostructures, typical for solution-based synthesis procedures, apart from template assisted methods, is prevented. 10 The structural diversity of nanoscaled structures available by PECVD should result in the variation of physical as well as chemical properties typical for non-aligned carbon structures. 1,[23][24][25] This is especially promising for applications that need a high surface to volume ratio, while maintaining a high signal and power transfer efficiency. Examples are electrochemical energy conversion devices, like fuel cells, batteries or capacitors, and sensors. [26][27][28] Therefore, understanding the effect of synthesis parameters on the nanoscale/mesoscale structure, as well as the electrochemical properties of nanostructured carbons, are indispensable preconditions to achieve electrode performance maximization, thus contributing to successful implementation in energy devices and sensors.
In this paper we report on the synthesis and characterization of three diverse carbon nanostructures (CN). The CN were grown using inductively coupled plasma enhanced chemical vapor deposition (IC-PECVD) without a catalyst and with aromatic p-xylene as a carbon precursor. By considering the plasma chemistry of p-xylene, the growth mechanism and the effect of the ow rate on the nanostructure morphology were discussed. Facile adjustment of the precursor ow rate resulted in alternating nanostructures. Carbon nanobers (CNF), freestanding carbon nanowalls (fCNW) and interconnected carbon nanowalls (iCNW) were synthesized, iCNW showed the highest oxygen reduction reaction onset and effective surface area. The inuence of monomer ow rate on the structure, morphology and electrochemical properties was studied by means of Raman spectroscopy, scanning and transmission electron microscopes as well as cyclic voltammetry, respectively, in order to understand the correlation between nano-/mesostructure and electrochemical properties.

Synthesis
Carbon nanostructures (CN) were synthesized in PECVD process using an inductively coupled plasma (ICP)-source according to reference (Hussein and Shoukat et al.) 29,30 and applying p-xylene as carbon precursor. The IC-PECVD process is based on report from Xu et al. 18 A schematic of a modied version is visible in Fig. 1.
The reactor used for the synthesis is made of a quartz reaction vessel with 85 mm in height and with an internal diameter of 70 mm. An open steel cage, to hold the substrate holder, is situated inside the reaction vessel. The substrate, xed on a substrate holder, is mounted parallel to the gas ow direction. The plasma is driven by a 13.56 MHz RF generator (PFG 300 from Hüttinger Elektronik) adjusted to a power supply of 150 W for 20 minutes for all depositions. 150 W was used to ensure a controllable growth speed, because growth speed was observed to increase exponentially with increased power input, while 20 min growth time produced a well recognizable lm and guaranteed high reproducibility. The substrate temperature at these settings was 450 C. The energy is transferred to a cylindrical coil made of 4 mm copper wire wrapped around the quartz vessel in six turns and grounded on the end. The functionality is explained in ref. 25.
The vacuum in the reactor was created by a rotary pump (Alcatel T2060C). The pressure monitoring was done by means of two compact pirani gauges (Pfeiffer PCR260). Process pressure was varied between 4.6 and 7.5 Pa, depending on the precursor supply rate. Argon gas (99.999%) was introduced at 10 sccm by an electronic mass ow controller (Brooks 5850E). The carbon precursor was an aromatic hydrocarbon p-xylene (Sigma-Aldrich, puriss. p.a., $99.0%) which was deaerated for 30 minutes by bubbling with argon. Precursor supply was ensured and adjusted by a syringe pump (Harvard PHD 2000) and a needle valve (Pfeiffer EVR 116). The substrate was chosen to be precleaned silicon wafer pieces coated with titanium nitride. They were mounted on metal plates, which were xed to a cylindrical metal grid in the reactor. The metal plate surface temperature was determined by a k-type thermocouple.

The structural and electrochemical characterization of carbon nanostructures
Scanning electron microscopy (SEM) (Quanta 250 FEG), transmission electron microscopy (TEM) (Zeiss LEO 912), Raman spectroscopy (Olympus BX40) and X-ray Photoelectron Spectroscopy (XPS) (Physical Electronics 5600 ci) were used to study morphology, nanoscale structure and composition of the samples. The Raman spectroscopy plots were evaluated by a Lorentzian t of the D-, G-and D 0 -peak.
The electrochemical properties were investigated with the help of cyclic voltammetry (CV) (Autolab PGSTAT30) in a three electrode setup. The reference electrode was an Ag/AgCl electrode in 3 M KCl solution, a platinum counter electrode and the sample with a diameter of 2.5 mm as working electrode were used. For each sample and investigation, three measurements on different places of the sample were performed.
For the investigation of the ability to catalyze oxygen reduction reaction (ORR), the CV measurements were performed in 0.1 M KOH solution saturated with nitrogen or air at a scan rate of 10 mV s À1 . The effective surface area (A eff ) was determined by CV measurements in solution of 50 mM KCl and 5 mM potassium ferricyanide K 3 [Fe(CN) 6 ] at six scan rates (n) ranging from 5 to 200 mV s À1 . In this context, the examination of A eff has proved to be useful to gain in-depth understanding of specic characteristics of graphene-based materials. 31 The anodic peak current obtained from the CVs was plotted over n 1/2 . Finally, A eff was determined with the help of the resulting slope and the Randles-Sevcik eqn (1.1).
I pa is the anodic peak current, n the number of electrons transferred in the reaction and n the scan rate. D is the diffusion coefficient (7.6 10 À6 cm 2 s À1 ) and C 0 is the bulk concentration of the redox couple K 3 [Fe(CN) 6 ]. 32 The measurements for A eff determination were always performed aer ORR measurements to avoid the possible impurities by iron and inuence of ORR response.

Nanostructures growth & structural characterization
The variation of the liquid precursor ow rate in the range of 0.5 to 5 ml h À1 leads to the formation of three different morphologies, carbon nanobers (CNF), freestanding carbon nanowalls (fCNW) and interconnected carbon nanowalls (iCNW). The growth process does not need a catalyst which was conrmed by parallel synthesis on glassy carbon, glass and titanium nitride coated silicon wafer. For the examination of the ow rate effect on the morphology and the structure height, three samples in each ow range were synthesized. The deposition time and power was 20 min and 150 W for all samples, respectively. SEM images of the corresponding morphologies are shown in Fig. 2. To produce CNF, depicted in Fig. 2a, a ow rate between 3 and 5 ml h À1 was necessary, the resulting height of the bers varies between 2.2 and 6 mm. Fig. 2b shows fCNW growing at a ow rate of 1-2 ml h À1 to a height of 1.4 to 2.2 mm. Below 1 ml h À1 the resulting morphology was iCNW structure, shown in Fig. 2c. iCNW reach a height between 1.1 and 1.6 mm. The height differences are probably caused by accuracy constraints of the feed supply rate. It should be noted that the distinction between fCNW and iCNW is uent. The transition from fCNW to iCNW happens when the vertical sheets connect with each other resulting in a consistent mesh like structure. This means, that from 2 ml h À1 downwards no fundamental structure change happens, except a decreasing of the wall size and simultaneous increasing of the wall number at lower feed ow rates. In Fig. 3 the resulting height of the structure is plotted over the ow rate of the precursor. Below 3 ml h À1 , the dominating CNF nanostructure changes to CNW. While the ow rate changes from 0.5 to 3 ml h À1 , only an increase in height of around 0.5 mm or 33% can be observed. According to Ostrikov et al. 2013, 13 the kind of nanostructure forming depends on supply and consumption of the precursor, whereby the consumption of the precursor depends on the surface temperature, which in turn is inuenced by the power supply. Because the power supply is constant for every synthesis performed, no  changes should occur for a power limited process. The growth of CNW, which happens at ow rates below 3 ml h À1 , does not alter a lot with the changing ow rate indicating power supply limitation. Above 3 ml h À1 , ber growth occurs and the height triples with the ow rate increasing from 3 to 5 ml h À1 . It follows that the ber growth is limited by the precursor supply. The nanostructure of the obtained carbon structures was examined by Raman spectroscopy through investigation of the D-and G-peaks (Fig. 4).
The D-peak, D 0 -peak and the G + D-peak, located at 1325 cm À1 , 1609 cm À1 and 2907 cm À1 , respectively, are caused by the disorder in the material structure. 33 The G-peak located at 1588 cm À1 features the graphitic character of the material, which is responsible for the ability of the material to conduct electricity. The D 0 -peak, which overlaps with the G-peak, has to be subtracted from the G-peak. The size, shape and position of the 2D peak depend on the layers thickness of the material. However, above 5 layers the peak becomes undistinguishable from bulk graphite. 34 CNW are reported to be 2-3 layers thick at the tip but the base, which dominates the Raman response, consists of more than 5 layer. 19,35 The SEM and TEM examinations in Fig. 2 and 6 conrm this, as the visible dimensions are several tens of nm wide at least. For this reason, no further examinations of the 2D were performed. Defects, indicated by the D-peak, represent edges as well as vacancies in the graphitic carbon layer and are responsible for adsorption as well as electron transfer processes. 36 Therefore, defects are very important for electrochemical activity of the material. The results exhibit that all three morphologies are based on sp 2 bonded graphitic carbon. By calculating the I D /I G ratio the relative amount of disorder in sp 2 carbons can be deduced. 34 The calculation results show that iCNW and fCNW with I D /I G ratio of 2.6 possess a similar amount of defects, while CNF, the samples with the highest growth rate, tend to have a I D /I G ratio of less than 1.5 exhibiting a smaller amount of defects. The small width of the D-peaks of 60 to 90 cm À1 (Table 1) is typical for non-amorphous carbons. [37][38][39] To make sure that the carbon nanostructures do not contain a second element, the iCNW samples were investigated by XPS. The XPS plot in Fig. 5 shows a strong C1s peak at 284.6 AE 0.2 eV and a weak O1s peak at 532.6 eV. The dominance of the C1s peak conrms that the nanostructures consist mainly of carbon. The position of C1s peak is typical for sp 2 hybridized carbon. The weak O1s peak could come from the physical adsorption of oxygen or water from air. [40][41][42] A closer look at the sample structure with TEM (Fig. 6) conrms the Raman results, that CNW structures have a higher defect density than CNF structure. The CNF structure (Fig. 6a-c) shows no edges and a smooth surface, while fCNW and iCNW ( Fig. 6d and e, respectively) are vertically grown walls of graphene, which results in a high amount of thin graphene layers with edges. Additionally, the sheets of iCNW structure are twisted and bend in different directions indicating an increased number of atomic defects in the graphene layer. 43 The dark, spherical particles observed in Fig. 6d and e are impurities from previous characterizations, they are not found in uncharacterized samples. The light, spherical particles observed in Fig. 6e are carbon nuclei which we believe to grow on the walls resulting in an irregular surface.
In general, CNF structures encompass a wide range of nanostructures which have a ber like morphology in common. They are classied in three types, parallel, shbone or bamboo like nanostructure. The parallel type, usually called multi walled carbon nanotubes (MWCNT), lacks graphitic edge sites on its lateral surface. Bamboo type has exposed graphitic edge sites at the transition between the links resulting in a bamboo like ber, while shbone type CNF's lateral surface consists of edge sites entirely offering a huge active surface for adsorption and electron transfer. It was, however, in literature reported, that shbone type CNF need a catalyst at the ber tip to grow, which is not present in this research. [44][45][46][47] Close examination of TEM images ( Fig. 6a and b) indicates that bers are likely building from stacked spherical structures implicating a bamboo type nanostructure.
In Fig. 7, schematics of a parallel type CNF (a), a bamboo type CNF with accessible edge sites between the links (b) and the cross-section of a CNW (c) are shown. These schematics were adapted from ref. 48. The small graphene sheets attached to a CNF recognized in Fig. 5c indicates that the sphere growth resulting in CNF and the sheet growth resulting in CNW happen simultaneously at different growth rates and under different constraints.

Plasma chemistry: comparison with conventionally used precursors
In following, we consider chemical reactions which are expected to take place in presence of p-xylene precursor and lead to the varying nanostructure formations. Especially, the formation of CNFs growing at applied conditions without catalyst is of interest, since the growth of carbon nanotube and nanober, irrespective of type, by both CVD and PECVD methods, usually require catalyst. [49][50][51] The two consecutive steps are necessary for growth of aligned nanostructures in PECVD, the formation of nucleation sites and vertical growth under the inuence of an electric eld. Thereby, the carbon source, beside the substrate, is supposed to play a dominating role in nanocarbon synthesis. This results from the strong inuence of the generation process of building units by bond stability in the respective precursor and its carbon content. [52][53][54] By PECVD of conventionally employed hydrocarbons, e.g. C 2 H 2 or CH 4 , and uorocarbons, e.g. CF 4 , CHF 3 , or C 2 F 6 , mainly the different types of reactive carbon dimers (C 2 ) are believed to be the initial building units in carbon nanostructure growth. 53,55 C 2 is generated by dehydration or bond dissociation of precursors through interactions with Ar, Ar ions, metastables or by electron collision and possible radical recombination. Starting from C 2 , the higher mass carbon clusters C n H x + are formed which initiate the formation of critical nuclei and the growth process. 52,53,55,56 The employing of p-xylene as aromatic unit bearing precursor should results in the modication of the nucleation and growth process. The C-H bond at a methyl group (CH 3 ) with the bond dissociation energy lower than 89 kcal mol À1 † is the weakest bond in p-xylene. 57 The C-H bond in benzene ring has dissociation energy of about 113 kcal mol À1 , while the average dissociation energy of a bond between carbon atoms in the benzene ring is estimated to be in the range of 132-138 kcal mol À1 . 57,58 Thus, the stability of aromatic ring is considerably higher than that of the methyl group. We suggest that the process starts with the production of the reactive p-xylyl radicals (Scheme 1), since the initial formation of these radicals was observed by pyrolysis of p-xylene. 59,60 Moreover, the p-xylyl radicals are able to undergo the condensation reaction through elimination of H 2 molecules under formation of polycyclic aromatic (2,6-dimethylanthracene), which can further participate in nucleation and growth process. 60 This reaction path should require plasma of comparatively low energy. For comparison, the dissociation energy of C-H bond in acetylene (HC^C-H) is about 133 kcal mol À1 and that of the rst C-H bond in methane (H 3 C-H) of 105 kcal mol À1 , which is signicantly higher than that of H 2 C-H in p-xylene. 57 According to calculations and mass spectra analysis performed by Koseki et al., 61 the generation of C 8 H 10 + and low-mass ions (Scheme 2), for example by electron collision, is also expected as the second reaction path. This path requires however plasma of higher energy, because of the quite high ionization potential of p-xylene (about 9 eV). The probability of the condensation reaction (Scheme 1) should increase with the rise of p-xylyl radical's concentration. In contrast, the probability of C 8 H 10 + formation and its decomposition to low-mass ions should not change or decrease

Scheme 1
The first possible reaction path in p-xylene PECVD process: formation of reactive p-xylyl radicals and their condensation to polycyclic aromatic 2,6-dimethylanthracene. with a rising precursor ow rate. Thus, we expect that, at higher precursor ow rates, the rst reaction path is more favorable than the second one. We believe that the different kinetics of these possible reaction sequences is the reason for the control of nanostructure formation by precursor ow rate. Also Ostrikov et al. emphasized, 13 that the kind of nanostructure forming depends on supply (formation of reactive building units) and consumption of the precursor (growth process), whereby the consumption of the precursor depends on the surface temperature, which in turn is inuenced by the power supply. Because the power supply is constant for every synthesis performed, no changes should occur for a power limited process. The growth of CNW, which happens at ow rates below 3 ml h À1 , does not alter a lot with the changing ow rate indicating power supply limitation. Above 3 ml h À1 , ber growth occurs and the height triples with the ow rate increasing from 3 to 5 ml h À1 . It follows that the ber growth is limited by the precursor supply.
We assume that the formation of CNF, which possess lower number of defects, is rather resulted from the rst reaction path, while CNW nanostructures with higher defect density and high amount of carbon nuclei are generated mainly from lowmass ions follow second reaction path. The low-mass non aromatic ions should create additional lattice defects.

Electrochemical characterization
To determine the inuence of the morphology of nanocarbon lms on electrochemical characteristics, the activity towards oxygen reduction reaction (ORR), the effective surface area and the capability to heterogeneous electron transfer were investigated. For that purpose, the onset potential E onset of ORR 62 as well as the extent and reversibility of the redox reaction of ferricyanide were examined. Additionally, from the cyclic voltammogram (CV) of ORR, the capacitive part of the current in CV was derived, and used as an indication for different degrees of sample wetting. Fig. 8a shows CVs performed in air (blue) and nitrogen (black) saturated solution for the E onset determination of ORR.
The ORR E onset of this sample is determined as the crossing of the two tangents at À202 mV. A tilting of the curves can be observed to different extent in all measurements. This might result from solution seeped through the CNW outside the measurement area, creating regions with a very high ions transfer resistance. The second reason is possible conductivity limitations of the substrate. The capacitive part of the CV current (CPC) was derived from the difference between the forward and the backward sweep of the CV at À100 mV, as is highlighted in Fig. 8a.
The results of E onset determination plotted over the sample height are shown in Fig. 8b. No correlation between the height of the respective nanostructure and the ORR E onset can be observed, although a clear shi towards positive E onset in case of iCNW is visible. Both fCNW and iCNW should have an abundance of edge plane sites resulting from the vertical growing graphene sheets, which have a strong ability for the adsorption of O 2 and are widely understood to have a higher catalytic and electrochemical activity as graphitic basal planes. 63,64 The iCNW structures additionally have bend and twisted sheets, which indicate atomic structure defects in the sheets. 43 The higher density of atomic structure defects might cause a change in the Fermi level and thereby a lowering of the work function of iCNW. 65,66 The work function lowering is associated with an increase in activity towards ORR. 67,68 The CNF and fCNW structures feature a similar ORR E onset found on average slightly above À200 mV which is apparently more positive than the onset of glassy carbon electrode observed at À220 mV. Because of the fact that the electrochemical properties of glassy carbon are similar to those of basal plane graphite, the higher reactivity of CNF and fCNW towards ORR should originate from a large number of edge planes obtained in these materials. 63 While the edge plane seems obvious for fCNW, their existence in case of

Scheme 2
The second possible reaction path in p-xylene PECVD process: p-xylene ionization and formation of low-mass ions and species. 61 CNF strongly suggests the bamboo type structure for CNF, considering the ndings of TEM analysis.
The CPC was used as an indicator for the degree of sample wetting. In Fig. 9a the inuence of the structure height on the CPC is shown. Fig. 9b is used to examine the relationship between the E onset and the surface area of nanostructure being in contact with the electrolyte.
Besides observation of high value distribution in Fig. 9a, an increase of the CPC with the height of the nanostructures can be noticed for large height differences. Apart from that, iCNW structures seem to have a slightly larger CPC than fCNW probably resulted from smaller pores and thereby higher surface area. CNF have a larger capacitive current CPC indicating a larger surface area resulting from the different morphology and bigger height of nanostructures. A big problem emerging with mesoporosity of carbon structures is the increasing of hydrophobicity, which leads to incomplete wetting of the structure and thereby to deviation in CPC measured. 69 For assessment of relationships between catalytic activity towards ORR and the wetting of carbon surfaces expressed by CPC, the ORR E onset was plotted over the CPC (Fig. 9b). While there is no correlation between ORR E onset and capacitive current CPC observable for CNF and fCNW, iCNW seems to have a better E onset with the higher CPC. Because the height of those iCNW samples is very similar, a higher CPC can implicate a higher degree of wetting. The structure of iCNW, shown in Fig. 6e, features a high number of graphene edges inside the material. These edges are assumed to be the active spots for ORR. 70 A higher degree of wetting should result in more active spots in solution, which also generates a higher CPC and an increased positive shi of E onset , in case of a process with kinetic limitations. For fCNW the inuence of the wetting is negligible, the exposed graphene edges are mostly located at the top of the material. The result is that the active spots of fCNW are in the solution even at a low degree of wetting. No relationship between CPC and E onset can be observed for CNF as well as fCNW indicating that, unlike for iCNW, the amount of active spots is not the limiting factor. Different reaction limiting factors for iCNW compared to fCNW and CNF suggest a fundamental difference in the kind of active spots available.
To evaluate the performance of the three morphologies as an electrode material, the corresponding nanostructures were applied as electrode in CV measurements of redox reaction of potassium ferricyanide K 3 [Fe(CN) 6 ]. Thereby the separation between the peak potentials DE p in CV and the active surface area (A eff ) of electrode were determined. Determining A eff provides a direct parameter for the amount of electrochemical active spots on a sample. Whilst the peak potential difference between the anodic and the cathodic peak DE p give information about reaction reversibility which is 59 mV for a perfectly reversible one electron reaction 71 as well as electron transfer activity. The edge plane pyrolytic graphene with a high number of edge plane sites, which are responsible for fast heterogeneous charge transfer, usually exhibit a small value of DE p . 46 All investigated nanostructures exhibited values below 70 mV for scan rates under 10 mV s À1 evidencing that all materials feature sufficient amount of defects for fast transfer kinetic. The results also show an increase of DE p with increasing scan rate to around 150 mV for 200 mV s À1 . The increase of DE p has been explained as a result of contamination by oxygen and water adsorption at the active spots, which affects the electrode kinetics. 72,73 The results of the A eff measurements are shown in Fig. 10a. The assessment focus is set on the samples inside the marked area, because the one iCNW sample standing out differs signicantly in ORR E onset , CPC, A eff and I D /I G from other samples, which indicate its deterioration caused by some unknown process, like synthesis contamination or pore blockage.
In contrast to the CPC, A eff decreases with rising sample height. This trend denotes the reduction of active spots in samples with larger growth rate. Consequently, a higher growth rate produces a large surface, which yields the high CPC, but fewer active spots, which is reected in a small A eff . For CNF growth this means that the amount of links forming the ber decreases, while the links get larger making the CNF more similar to parallel type rather than bamboo type. Considering the growth mechanism, this evolution could be explained by a decreasing nucleation rate on the ber tips resulting in continued link growth. Decreasing nucleation rate might arise from lower plasma temperature caused by the higher particle density. Analysis of the samples with similar height reveals that CNW have on average a higher A eff than CNF. Moreover, iCNW appear to have a larger A eff than fCNW originating from a higher number in graphitic edges.
When examining an inuence of I D /I G ratio on A eff (Fig. 10b) clear relationships between a higher I D /I G ratio and a higher A eff can be seen conrming the key role of defects as active spots in electron transfer.
The last step is examination of the relationship between the ORR E onset and A eff (Fig. 11). For better comparability, the investigated samples in marked area exclude the likely corrupted iCNW sample and the especially large CNF sample. The different morphologies inside the marked area show a clear correlation between a higher catalytic activity towards oxygen reduction and an increased A eff for iCNW which are the most reactive nanostructures followed by fCNW and then CNF. The resemblance between fCNW and CNF in A eff and E onset values suggests that not the wall growth mechanism is responsible for the increased reactivity towards ORR and higher number of active spots for electron transfer. The strong differentiation of iCNW from fCNW in ORR E onset and A eff supports this view. The higher reactivity of iCNW, especially with respect to ORR, despite of the similar I D /I G values supposes that not only the amount of defects is an important parameter for electrochemical reactions, but that the kind of defects is at least as important for the electrochemical activity of nanocarbon materials. In case of iCNW, the increased amount of atomic defects in the graphitic layer, apparent in their twisted structure, can also cause an increased reactivity caused by a lowering of the work function.

Conclusions
Three different types of nanostructured carbon lms, CNF, fCNW and iCNW, were synthesized by PECVD method through variation of monomer ow rate and comprehensively studied by structural and electrochemical characterization. The results of SEM and TEM analysis indicate two different growth mechanisms for CNF and CNW nanostructures (fCNW and iCNW) also producing different amounts of defects which were identied by examination of I D /I G value from Raman spectra. It was observed that the decreasing of monomer ow rates lead to the increasing amount of defects identied by SEM and TEM as graphitic edge sites. However, the differences in electrochemical activity between CNF and fCNW were small, with the ORR E onset at À196 mV and A eff between 0.46 mm À1 (CNF) and 0.61 mm À1 (fCNW). Moreover, a generally higher electrochemical activity was observed for iCNW, E onset ¼ À171 mV and A eff ¼ 0.76 mm À1 , despite the comparable values of I D /I G with fCNW. In contrast to fCNW, iCNW nanostructures exhibit twisted walls which can originate from atomic defects in graphene sheets. These ndings led us to conclude that the kind of defects is a determining factor for electrochemical activity. Finally, iCNW with interconnected walls were determined as the nanostructures with the best electrochemical activity. The formation of defects and realizing of different structures were discussed with respect to plasma chemistry of p-xylene precursor.
The results of this work demonstrate the potential control over the structure and properties of carbon CN, by adjustment of the precursor ow rate.