advances.sciencemag.org/cgi/content/full/5/8/eaaw8337/DC1

Supplementary Materials for

Nitrogen cluster doping for high-mobility/conductivity graphene films with

millimeter-sized domains

Li Lin, Jiayu Li, Qinghong Yuan, Qiucheng Li, Jincan Zhang, Luzhao Sun, Dingran Rui, Zhaolong Chen, Kaicheng Jia, Mingzhan Wang, Yanfeng Zhang, Mark H. Rummeli, Ning Kang, H. Q. Xu*, Feng Ding*,

Hailin Peng*, Zhongfan Liu*

*Corresponding author. Email: zfliu@pku.edu.cn (Z.L.); hlpeng@pku.edu.cn (H.P.); f.ding@unist.ac.kr (F.D.);

hqxu@pku.edu.cn (H.Q.X.)

Published 9 August 2019, Sci. Adv. 5, eaaw8337 (2019) DOI: 10.1126/sciadv.aaw8337

The PDF file includes:

Supplementary Method Fig. S1. Procedure for the growth of Nc-G films. Fig. S2. The continuous nucleation of nitrogen-doped graphene. Fig. S3. The effect of the oxygen-assisted etching-regrowth cycle on suppression of nucleation. Fig. S4. Structural characterization of as-grown large Nc-G single crystals. Fig. S5. Raman characterizations of Nc-G grains. Fig. S6. The isotopic labeling experiment to visualize the growth kinetics of millimeter-sized Nc-G grains. Fig. S7. The reported nμ and σ values function as μ in Nc-G of this work (red) and previous CVD doping strategies (navy blue). Fig. S8. STM images of the clustered nitrogen dopants in graphene lattice. Fig. S9. The STM and STS characterization of single-substitutional nitrogen-doped graphene. Fig. S10. Calculated dissociation energy of C-C-N. Fig. S11. STS measurements of Nc-G films. Fig. S12. N atoms prefer to stay on the edge of a C-N cluster. Fig. S13. C-N clusters without N atoms at the center are more stable. Fig. S14. C-N cluster with flat structure is more stable. Fig. S15. C-N cluster with high ratio of N atoms at the edge is more stable. Fig. S16. A series of triangular shaped C-N clusters with N edges have very low formation energies. Fig. S17. The doping efficiency of Nc-G film. Fig. S18. The large-scale conductivity and transmittance of Nc-G films. Fig. S19. Potential application of Nc-G films. Fig. S20. High electrostatic potential and quasi-bound states induced by nitrogen clusters. Legend for movie S1

Table S1. Mobilities and sheet resistance of previously reported intrinsic graphene. Table S2. Mobilities, conductivity, and stability of previously reported doped graphene. References (45–62)

Other Supplementary Material for this manuscript includes the following: (available at advances.sciencemag.org/cgi/content/full/5/8/eaaw8337/DC1)

Movie S1 (.mp4 format). Demonstration of a flexible touch screen device made from Nc-G film.

Fig. S1: Procedure for growth of Nc-G

The detailed procedure for growing Nc-G, including the annealing, passivation, nucleation,

etching, regrowth, and cooling steps, is shown in fig. S1A and in the section: Materials and

methods (below). Note that two strategies were used during growth to decrease nucleation

density of Nc-G. First, after the annealing step, oxygen was introduced into the chemical vapor

deposition (CVD) chamber to passivate active sites on the Cu foil, in accordance with previous

report (18). Second, after the first nucleation, a selective etching of those nuclei with small

domain size was carried out by introducing oxygen for a second time, which would also

passivate the remaining active sites. Nucleation during the regrowth step that followed was

thereby greatly suppressed (fig. S1B).

Fig. S1. Procedure for the growth of Nc-G films. Drawing of the growth procedure (A), and a

schematic illustration of the oxygen-assisted etching-regrowth cycle for growing Nc-G with

millimeter-sized domain (B). Note that the growth procedure shown in (A) is for a continuous

Nc-G film on Cu foil. The configuration of the graphene (square or hexagonal) is determined by

the lattice orientation of the underlying Cu substrate.

Fig. S2-3: Selective etching by oxygen to suppress graphene nucleation

After passivation of active sites on Cu foil, the nucleation density of graphene is still very high,

presumably as a result of efficient decomposition ability of acetonitrile (ACN) (fig. S2A).

Furthermore, the nitrogen-doped graphene (NG) grains on Cu foil exhibit a broad distribution of

domain sizes. In particular, the maximum domain size exceeds 200 μm, while the smallest is

around 10 μm. Such a broad distribution is attributed to difference in growth rate for each grain.

Nucleation and growth of the nitrogen-doped graphene is fueled by the active carbon species

formed by catalytic decomposition of the carbon source by Cu. Such catalytic capability of Cu is

mainly determined by the morphology of Cu foil. For instance, grain boundaries, dislocation, and

point defects on the Cu foil exhibit strong catalytic ability and thus a higher possibility for

graphene nucleation (19,45-46). Consequently, a spatial difference in catalytic ability leads to a

broad distribution of graphene grain size.

To explore the nucleation time of graphene grains, Cu foil was exposed to a constant flux of

ACN while a small amount of carbon-isotope labeled methane (13

CH4) was introduced in a

sequence of pulses. Because the growth of graphene on Cu is surface-mediated, and the Raman

modes of 12

C and 13

C are separated in wavelengths, thus, the temporal evolution, especially the

nucleation time of each grain, can be visualized by Raman maps of the isotopically labeled

nitrogen-doped grains, as described in previous reports (20). Note that, except for the

introduction of the isotopically labeled methane pulses, the growth parameters were kept the

same as in the normal nucleation of NG grains. In addition, the quantity of 13

CH4 introduced in

the pulses was very limited, to avoid an undesirable perturbation in the overall growth, and was

just sufficient to produce a detectable shift of the G or 2D band in the Raman spectrum.

Consequently, there were two characteristic Raman sources: graphene containing only 12

C,

which was formed entirely from ACN, and graphene containing 13

C and 12

C, which was grown

from a complex gas mixture consisting ACN and 13

CH4. Graphene composed of

13C and

12C

would indicate a G band and a 2D band at lower wavenumbers (20). In detail, a four-pulse

sequence was introduced during the nucleation step. As shown in the corresponding Raman maps

of the 2D band position (fig. S2B, inset), some graphene grains nucleated before the first pulses,

however, the domain size of the grains formed prior to the first pulse was quite different.

Furthermore, for some small nuclei, one or two of the labeling pulses were missing. Such an

observation suggests that the occurrence times of nucleus on the Cu substrate are quite different,

leading to a broad distribution of domain sizes; a nucleus formed earlier exhibiting a larger

domain size. In this regard, the nucleation time for each grain can be inferred from the plots of

domain size as a function of growth time during nucleation, assuming that the growth rate during

nucleation stage is constant and that there is no degradation due to the relatively lower graphene

coverage. The statistics of nucleation time of each grain are presented in fig. S2B, confirming

continuous nucleation of nitrogen-doped graphene.

During the subsequent oxygen etching step, the prominent differences in grain size predict a

highly different capability for grains to survive the etching. In this regard, grains with a larger

size will need a longer time to be totally removed by the oxygen, while smaller ones will be

removed more easily (selective etching). Thus, if etching time is carefully controlled according

to the domain size of Nc-G nuclei, only grains with larger grain size will remain after etching. In

the next step, the ACN and H2 were introduced into the system to initiate the regrowth of

remaining nitrogen-doped graphene nuclei. Furthermore, the oxygen would produce a second

passivation of the remaining active sites on the Cu foil, which is clear of graphene grains after

selective etching (18). When regrowth begins, a large quantity of active species is produced on

the Cu surface. These precursors for graphene are, nevertheless, forced to diffuse towards the

remaining nuclei and fuel the epitaxial growth from the edge, rather than producing new

nucleation centers, as a result of the constant passivation of active sites. Therefore, the overall

nucleation density is maintained during the regrowth. Consequently, the overall number of

as-formed grains is prominently reduced by etching, as evident in scanning electron microscope

(SEM) images of the graphene grains (fig. S3A). In particular, more growth-etching-regrowth

cycles can be used to further reduce the nucleation density (fig. S3B). The reduction of

nucleation density after two selective etching cycles is plotted in fig. S3C. Note that, after the

oxygen etching step, the remaining Nc-G grains exhibit a relatively narrow distribution,

confirming the selective etching effect of oxygen (fig. S3D) (18,47). Note that the average

nucleation density is 0.597 mm-2

, 0.463 mm-2

, 0.914mm-2

for graphene domains grown at 900 oC,

950 oC, 1000

oC, respectively. Although the enhanced growth temperature can result in a higher

nucleation barrier and reduced nucleation density, graphene grains grown at 1000 oC exhibit a

highest nucleation density presumably owing to the enhanced dissociation of acetonitrile, which

would produce more active carbon species for nucleation.

Fig. S2. The continuous nucleation of nitrogen-doped graphene. (A) SEM image of Nc-G

grains on Cu foil, formed after the first nucleation. (B) The statistics of nucleation time of the

grains, inferred from the Raman mapping results. The nucleation time is calculated from the

introduction of ACN to the presence of grains on Cu foil. Note that grains with a smaller domain

size, that is, containing less 13

C pulse, are formed after a relative longer duration, and thus

indicate a longer nucleation time. Inset: Raman map of 2D band position of the isotopically

labeled nitrogen-doped grains transferred onto SiO2/Si substrate, confirming a broad distribution

of domain size.

Fig. S3. The effect of the oxygen-assisted etching-regrowth cycle on suppression of

nucleation. SEM images of Nc-G grains on Cu foil formed after the first (A) and second (B)

oxygen-assisted etching-regrowth cycle (i.e., selective etching cycle). Note that the regrowth step

was stopped before full coverage in order to count the nucleation density. (C) Nucleation density

as a function of the number of growth-etching-regrowth cycles with 50 sccm H2 (green) and 100

sccm H2 (red). (D) Domain size distribution of the as-formed graphene nuclei after first

nucleation (green) and one growth-etching-regrowth cycle (red).

Fig. S4: Single-crystal nature of Nc-G grains with a millimeter-sized domain

Transmission electron microscopy (TEM) was performed to probe the detailed microstructure

and crystallinity of the as-synthesized graphene. The inset of fig. S4A displays an SEM image of

one graphene grain on the TEM grid. Selected-area electron diffraction (SAED) patterns were

obtained in a representative manner from 20 individual regions across the entire domain and

were extensively analyzed, as shown in fig. S4B-I. The distribution of the relative angles of

graphene lattices extracted from the SAED patterns is plotted accordingly (fig. S4A), confirming

the single-crystal character of millimeter-sized NG grains.

Fig. S4. Structural characterization of as-grown large Nc-G single crystals. (A) Histogram

of the angle distribution from the extensive SAED patterns within one millimeter-sized Nc-G

grain on a TEM grid. Inset: SEM image of the grain on the TEM grid (scale bar, 500 μm). (B-I)

Representative SAED patterns of the graphene grain collected across the graphene domain.

Fig. S5: Doping uniformity and quality of the graphitic-nitrogen doped graphene film

Raman spectroscopy was used to study the doping characteristics of Nc-G graphene, such as

doping uniformity and the crystalline quality of Nc-G graphene grains. The Nc-G single crystals

were transferred onto a SiO2/Si substrate using a “dry” transfer method, to avoid the

transfer-related doping (fig. S5A) (48). Note that, a uniform contrast of the transferred grain

indicates the monolayer nature on a large scale, which is also confirmed by Raman spectroscopy.

In contrast to pristine graphene, Raman spectrum of Nc-G (fig. S5B) exhibits a prominent D

band (~1351 cm-1

) and an accompanying D′ band (~1620 cm

-1), which stem from the elastically

scattered photoelectron generated by foreign atoms in a graphene matrix and intravalley double

resonance scattering processes, respectively (49). Furthermore, the observation of a strong 2D

band is indicative of the high quality of nitrogen doped graphene. In comparison to intrinsic

graphene, the blue shifts of the 2D and G bands were also observed, clearly revealing a shift of

the Fermi level as a result of nitrogen doping. Moreover, the corresponding Raman mapping of

defect-induced D band intensity confirms the doping uniformity (fig. S5C).

Fig. S5. Raman characterizations of Nc-G grains. (A) OM image of one Nc-G grain

transferred onto a SiO2/Si substrate. (B) Raman spectra for the Nc-G (red) and intrinsic graphene

grains (green), displaying a prominent D band in Nc-G. The ID/IG ratio ranges from 0.66 to 0.76,

which is also sensitive to the transfer process. (C) Raman map of D band intensity corresponding

to the same area as in (A).

Fig. S6: The growth rate of large Nc-G single crystals

In order to visualize the growth dynamics of the Nc-G grains (including the growth rate), Cu foil

was exposed to a constant flux of ACN and H2, along with a small amount of carbon-isotope

labeled methane (13

CH4) in a specific pulsed sequence. Figure S6A shows a schematic of the

CVD growth of isotopically labeled, millimeter-sized, nitrogen-doped graphene grains using an

oxygen-assisted growth-etching-regrowth strategy. In detail, after oxygen-assisted etching was

used to suppress the nucleation, Cu foil was subsequently exposed to ACN with a small amount

of 13

CH4 introduced in a specific sequence of pulses. The Raman G band position map (fig. S6B,

inset) clearly shows the spatial distribution of isotopic carbon atoms, in which the prominent

striped structure indicates the temporal sequence (marked as times t1, t2, t3, t4) of the 13

CH4

pulses. Therefore, in conjunction with the spatial distribution of the carbon isotopes (marked as

positions x1, x2, x3, x4), the time-dependent growth behavior can be visualized. The Nc-G grains

exhibit a concentric structure, confirming the surface-mediated growth mechanism and the

single-crystal nature of the millimeter-sized grains. The consistency of the rectangular domain

shape during the entire growth suggests the dominance of attachment-limited kinetics during

growth (50). From the time-dependent diameter of the Nc-G grain, the growth rate was

calculated to be around 160 μm/min (fig. S6B), among the best results reported with regarding to

the growth of high-quality large single-crystal graphene (19,50-52). The presence of surface

oxygen, and the higher decomposition rate of ACN, both contribute to the higher growth rate (19).

We believe two reasons that contribute to the improved growth rate. The first one is the low

dissociation energy for the reaction: C-C-NC+CN (0.42 eV), which would produce a large

number of active carbon and nitrogen species to fuel the growth, in contrast with the growth of

graphene by using methane. The second one is the presence of oxygen during the growth, which

would accelerate the growth of graphene by lowering the dissociation energy of carbon

precursors and attachment barrier of active carbon species.

Fig. S6. The isotopic labeling experiment to visualize the growth kinetics of millimeter-sized

Nc-G grains. (A) Schematic of the procedure for growing isotopically labelled millimeter-sized

Nc-G grains by introducing 13

CH4 in a specific pulsed sequence. The (t1, t2, t3, t4) denote times

when the 13

CH4-pulses were introduced. (B) The diameter of the corresponding Nc-G grains as a

function of the growth time. Inset: The Raman G13

-band position map of graphene grains in Fig.

1F. The positions (x1, x2, x3, x4) denote where the 13

CH4 pulses were detected, corresponding to

the times (t1, t2, t3, t4), from which the time evolution of the diameter can be inferred. The

domain-size scales linearly with growth time, showing a constant growth rate of 160 μm/min.

Fig. S7: Improved conductivity of Nc-G in comparison with previously reported doping

strategies.

Incorporation of heteroatoms during CVD growth aim to deliberately enhance the carrier

concentration to improve graphene conductivity (2,23,25-27,53-55). According to the equation:

𝜎 = 𝑛𝜇e (where n is the carrier concentration, μ is the carrier mobility, σ is the conductivity),

graphene conductivity is determined by the carrier mobility and carrier density (concentration) in

working devices. Thus, to present a reasonable comparison of conductivity between Nc-G and

other CVD-doping strategies, both the carrier density and mobility should be taken into

consideration. To attain this, we compared the values of 𝑛𝜇 and 𝜎 from different reports

regarding graphene doping.

We estimated the carrier concentration from the position of charge neutrality points (Dirac points)

from the transfer curve, according to the equation: 𝑛 = −α(𝑉G − 𝑉CNP), where α is related to the

gate capacity (For a 300 nm SiO2, α is 7.2 × 1010

cm-2

V-1

). Note that, as for the carrier mobility,

we directly summarized the reported values from literature, which were obtained using different

fitting models. The values of carrier density and mobility (𝑛𝜇) were usually presented for the

evaluation of conductivity and sheet resistance in applications of transparent conductive films

(13).

Consequently, the values of 𝑛𝜇, 𝜎 and 𝜇, obtained in Nc-G samples and other doping routines,

are presented in fig. S7 for comparison, indicative of significantly improvements in carrier

mobility and conductivity. Clearly, the 𝑛𝜇 value of Nc-G approaches the theoretical value. The

sheet resistance can also be estimated according to the equation: 𝑅s = (𝜎𝑑Gr.)−1, where 𝑑Gr. is

the thickness of monolayer graphene.

Fig. S7. The reported nμ and σ values function as μ in Nc-G of this work (red) and previous

CVD doping strategies (navy blue). The data points are labeled with the reference number from

whence they came in brackets. Theoretical value (orange) (carrier density and carrier mobility) is

obtained according to the previous theoretical work (11). Note that, the carrier mobility values

were obtained using different fitting models. Thus, the nμ values, obtained from literature, are

not the experimental results, which is only used for the comparison of material conductivity (13).

Fig. S8-11: Scanning tunneling microscopy (STM) characterization of Nc-G

STM/STS is an efficient technique for visualizing the atomic-level distribution of dopants in the

graphene lattice and the characterization of local electronic properties, and was used here to

probe the atomic arrangement of dopants in Nc-G films. Large-scale STM image reveals many

bright areas of 2-5 nm in diameter, which can be attributed to the clustering of graphitic-nitrogen

dopants in the hexagonal lattice (Fig. 2A and fig. S8). In sharp contrast to the randomly

distributed individual nitrogens in single-substitutional nitrogen doped graphene, as broadly

observed in previously reports (34,56) and in our results (fig. S9), the lateral dimensions of each

doping center in our Nc-G sample are much larger than that in single-substitutional nitrogen

doped samples. The STS measurements were performed to examine the effect of nitrogen

clusters on the local electronic structures (fig. S11). Similar features of the gap-like depression

near EF and near -400 meV dip relative to the tip, were seen in all of the dI/dV spectra, which

were obtained at points inside a single nitrogen cluster. The gap-like feature can be attributed to

the tunneling current decay near EF caused by graphene electron-phonon coupling (34). The -400

meV feature is associated with the position of Dirac point, confirming the n-type doping induced

by nitrogen-clusters in Nc-G film (34).

In addition, single substitutional nitrogen-doped graphene was synthesized using ammonia gas to

achieve a single-substitutional doping configuration (52). In contrast to nitrogen-cluster doping

in Nc-G films, the atomic-level structure of the ammonia-prepared sample, as shown in STM

images, is totally different, exhibiting a random distribution of single-substitutional N dopants

(fig. S9A), that is consistent with the previous report (34). In particular, the STM images show a

graphene lattice with bright points with lateral dimension less than a few atomic spacings,

indicating single substitution of carbon by nitrogen (34). A strong intervalley scattering tail was

also observed around each dopant in the enlarged STM image of one doping center (fig.S9B,

inset), consistent with previous work (34,57). We also performed detailed STS measurements of

the differential conductance dI/dV to probe the low-energy electronic structure of the

single-point doped graphene (fig. S9B). The depression near -300 meV, relative to the tip,

denotes the position of Dirac point, indicating a clear shift of the Fermi level as a result of

nitrogen doping.

To understand why acetonitrile uniquely contributes to the formation of nitrogen cluster, the

dissociation of C-C-N and C-N groups on Cu(100) surface was calculated. It is found that the

dissociation energy for C-C-NC+CN is only 0.42 eV, while the dissociation energy for

C-NC+N is as high as 2.2 eV (fig. S10). These results clearly show that the CN group

produced by the dissociation of acetonitrile on the Cu(100) surface is highly stable. This result

explains the uniqueness of acetonitrile in the formation of nitrogen cluster in graphene lattice. In

comparison with acetonitrile, NH3, which was broadly used to grow nitrogen doped graphene,

can be dissociated into single N atom only.

Fig. S8. STM images of the clustered nitrogen dopants in graphene lattice. (A to F),

Representative STM image of nitrogen-cluster doping centers observed in Nc-G film deposited

directly on Cu foil. Inset: fast Fourier transform (FFT) of topography shows atomic peaks (outer

hexagon, denoted by yellow cycles) and intervalley scattering peaks (inner hexagon, denoted by

cyan cycles). (B) is enlarged image of (A). The white arrow denotes the presence of cluster of

nitrogen dopants. For the STM measurements, the NG samples were transferred into the

ultra-high vacuum (UHV) system soon after preparation, followed by annealing in vacuum at

∼600°C to remove adsorbates. Scanning conditions: (A and B) 0.002 V, 30.77 nA; (B) -0.002V,

23.86 nA; (D) -0.002 V, 26.25 nA; (E) -0.002 V, 30.78 nA; (F) -0.021 V, 2.75 nA.

Fig. S9. The STM and STS characterization of single-substitutional nitrogen-doped

graphene. (A) STM image of the single-substitutional nitrogen-doped graphene film on Cu foil,

indicating the configuration of individual N-doping center, denoted by white arrows. (B) dI/dV

curves taken on the bright topographic feature (marked in the inset) near single-point doping

center on Cu. Inset: enlarged STM image of the location where the spectrum was taken.

Scanning conditions: (A) 1.0 V, 200 pA; (B) 0.8 V, 200 pA.

Fig. S10. Calculated dissociation energy of C-C-N. The calculated dissociation energy of

C-C-N (A) and C-N (B).

Fig. S11. STS measurements of Nc-G films. (A) An enlarged STM image from fig. S8A and

the locations where the STS spectra was taken. (B) dI/dV curves taken on the bright topographic

features in the nitrogen-cluster doping center, offset vertically for clarity. The red dashed lines

represent the positions of Dirac points. Scanning conditions: (A) 0.002 V, 30.77 nA; (B) -0.002 V,

16.83 nA.

Fig. S12-16: Density functional theory (DFT) calculation of the formation mechanism of

Nc-G films

In this section of the supporting information, density functional theory (DFT) is used in

calculations of the formation energy of various C-N clusters configurations.

Fig. S12. N atoms prefer to stay on the edge of a C-N cluster. (A) C18N3 clusters, forming four

different structures, and their formation energies. Lowest energy in each group is indicated in red.

The CN clusters with N atoms (blue) at the edges have lower formation energies. (B) C12N9

clusters: three different structures and their formation energies. The CN cluster with all N atoms

at the edge has the lowest formation energy. When three of the N atoms are moved to the center

of the cluster, the formation energy increases. The C12N9 cluster with only 3 N atoms at the edge

has the highest formation energy.

Fig. S13. C-N clusters without N atoms at the center are more stable. (A) The structures and

formation energies of C15N6 and C12N9 clusters. By replacing the three C atoms at the center of

C15N6 with N atoms, the CN cluster becomes unstable, which is indicated by the increase of

formation energy. (B) C18N6 and C15N9 clusters. Replacing the three C atoms at the center of

C18N6 with N atoms increases the formation energy of the CN cluster. (C) C7N6 and C6N7

clusters. Replacing the C atom at the center of C7N6 with N atoms leads to C6N7 and increased

formation energy. (D) C7N7 and C6N8 clusters. Replacing the C atom at the center of C7N7 with N

atoms leads to C6N8 and increased formation energy.

Fig. S14. C-N cluster with flat structure is more stable. The structures and formation energies

of CN clusters with and without pentagons. (A) A CN cluster composed of 3 hexagons is more

stable than 3 pentagons. (B) A CN cluster of 6 hexagons is more stable than one composed of 3

hexagons and 3 pentagons. (C) A CN cluster composed entirely of hexagons is more stable than

those composed of 3 pentagons and 4 hexagons.

Fig. S15. C-N cluster with high ratio of N atoms at the edge is more stable. The structures

and formation energies of triangular (black squares) and hexagonal (red circles) CN clusters with

different ratios of N atoms at the edge. For both triangular and hexagonal CN clusters, the

formation energy of the cluster decreases with an increase of the N ratio at the cluster edge. The

formation energy is lowest when all edge atoms are N atoms. Compared to the hexagonal-shaped

CN clusters, the triangular-shaped CN clusters have lower formation energy.

Fig. S16. A series of triangular shaped C-N clusters with N edges have very low formation

energies.

Fig. S17: The tunability of the Fermi level of Nc-G films

The controllability of dopant concentration in Nc-G films can be realized by tuning the growth

temperature of Nc-G film. As reported in previous works, increasing the growth temperature

usually results in a low doping concentration in graphene as a result of the formation energy

(bonding energy) for the C-C bond being higher than for C-N. Consequently, at elevated

temperature, formation of the more stable C-C bond is favored, resulting in a reduced doping

level. Thus, we tune the Fermi level of the as-formed Nc-G by growing Nc-G at different

temperatures (900 oC, 950

oC, and 1000

oC).

The adjustment range of Fermi level is shown to be around 350 meV relative to intrinsic

graphene, as verified by the UV photoelectron spectroscopy (UPS) (fig. S17A). Furthermore,

transport measurements were performed to evaluate the electrical properties of the Nc-G films

grown at various temperatures, as well as to characterize the potential to adjust the Fermi levels.

The transfer-characteristic curves of samples denoted 900 Nc-G, 950 Nc-G, and 1000 Nc-G (fig.

S17B) all exhibit negative charge neutrality points (Dirac points) with different shifts relative to

the zero value, which is the hallmark of different electron-doping levels. The distance between

the charge neutrality point and the zero value corresponds to the Fermi level shift in Nc-G films

compared to pristine graphene. Using the formula (58)

𝐸F = ℏ𝜈F√𝜋𝑛 (S1)

Where 𝜈F denotes the Fermi velocity.

We extract the Fermi level shift (∆𝐸F) in different samples from these curves, revealing that the

∆𝐸F can be tuned from 180 meV (900 Nc-G) to 90 meV (1000 Nc-G), relative to intrinsic

graphene. XPS measurements were also performed to probe the nitrogen concentration indicating

that the atomic concentration of nitrogen dopants can be tuned from 1.4% to 0.6% (fig. S17C).

Raman spectroscopic characterization is also informative with respect to doping level in

graphene. In particular, the D band intensity reflects the concentration of dopants or defects in

the graphene lattice, and the intensity ratio of the D band to the G band (ID/IG) is usually

presented to reflect the doping concentration (45). The Raman spectra of the Nc-G films, formed

at different temperatures (fig. S17D), exhibit a reduction of the D band intensity with elevated

growth temperature with ID/IG ratios of 0.60, 0.25, and 0.15 for 900 Nc-G, 950 Nc-G, and 1000

Nc-G, respectively. ID/IG ratios for single-substitutional nitrogen doped graphene (Ns-G) and

nitrogen doped graphene containing pyridinic and substitutional nitrogen (Np,s-G) are 0.70 and

0.96 respectively. Without using the oxygen-assisted growth, a higher D band intensity was

observed in Np,s-G than that of Nc-G using the oxygen-selective-etching strategy. Furthermore,

it was found that the intensity ratio of the D and D′ peak is reduced in Nc-G indicating an

elimination of defective pyridinic nitrogen after the oxygen-selective-etching step.

In addition, nitrogen-doping induces a noticeable shift of the G band position, which can be used

to directly measure the carrier concentration and Fermi level of doped graphene. In this regard, a

successive blue shift of G band in the 900 Nc-G, 950 Nc-G, and 1000 Nc-G samples were

observed, confirming the dependence of introduced carrier concentration on the formation

temperature (fig. S17E). Thus, all the observations above confirm the possibility in tuning the

doping levels in Nc-G samples through the control over growth temperature.

Nitrogen doping is a method to inject additional electrons into the graphene system, and to shift

graphene Fermi level. Thus, the conductivity would be enhanced in nitrogen-doped graphene to

broaden its electronic application appeal. However, the doping configuration uniquely

determines the doping efficiency, i.e. how much added nitrogen in a graphene system is required

to achieve a desired shift of the Fermi level. In particular, only substitutional doping can induce

n-doping in graphene, as pyridinic nitrogen would disrupt the graphene lattice and usually induce

a contradictory doping effect (p-doping) (14). The doping efficiency is an important parameter in

graphene doping, because the nitrogen atoms in the graphene lattice function as electronic

scattering centers, reducing the quality (i.e. carrier mobility) of the as-formed doped graphene.

Consequently, a method with lower doping efficiency requires a higher quantity of nitrogen

dopant to induce the same number of electrons than one with a higher doping efficiency. In turn,

this would lead to a significant reduction in carrier mobility and conductivity.

Firstly, we compare the nitrogen doping efficiency of Nc-G sample with single-substitutional

nitrogen doped sample (20,52). The previously reported value of nitrogen concentration in

single-substitutional doped sample is 2.0%, higher than the 1.6% in 900 Nc-G (results obtained

from XPS results). However, as evidenced by the G band shift of Raman results (fig. S17D, E),

the ∆𝐸F in single-substitutional doped graphene (~257 meV) is clearly smaller than that in

Nc-G sample (~363 meV). Thus, a lower carrier concentration was obtained in

single-substitutional doped sample (~3.1×10-12

cm-2

) than that of Nc-G (~8.0×10-12

cm-2

),

consistent with the theoretical calculation results (Fig. 1E in the main text). These results clearly

indicate an improved doping efficiency through the clusterization of dopants in Nc-G. To provide

a comparison of doping efficiency in Nc-G with other previously reported CVD-doped graphene

(8, 47-51), the nitrogen concentrations (gained from XPS in our works and previous works)

versus the ∆𝐸F and carrier concentrations are plotted in fig. S17F., confirming the higher doping

efficiency in Nc-G.

Fig. S17. The doping efficiency of Nc-G film. (A) UPS spectra of the 900, 950, 1000 Nc-G, and

pristine graphene films on Cu foil. The Fermi level is shifted by 300, 150, and 50 meV

(compared to pristine graphene) for 900, 950, and 1000 Nc-G, respectively. (B) Transfer curves

of 900 (red), 950 (green), and 1000 Nc-G (orange) devices. (C) XPS spectra of the samples are

identified as in (A). Note that all the films indicate a single peak, corresponding to graphitic

doping. The dopant concentrations of 900, 950, and 1000 Nc-G are 1.4, 1.0, and 0.6%,

respectively. (D) Raman spectra of the samples (identified as in (A)), Ns-G and Np,s-G, all of

which indicate the prominent doping-related D bands. (E) Enlarged Raman G band spectra. The

G band positions of 900, 950, 1000 Nc-G, and Ns-G are 1593, 1591, 1588, 1589 cm-1

respectively. Note that 900 Nc-G exhibits a higher carrier concentration than that of

single-substitutional doped graphene, and a similar ID/IG ratio, indicating that less defects are

required to achieve the same ∆𝐸F in Nc-G. (F) The ∆𝐸F and carrier concentration function as

the nitrogen dopant concentration in Nc-G of this work (red), single-substitutional doped

graphene (blue) and values reported in the literature (yellow). The data points are labeled with

the reference number from whence they came in brackets.

Fig. S18: The large-scale conductivity and transmittance

(1) The large-scale transmittance of Nc-G film.

Optical transmittance is an important parameter in the increased integration of doped graphene

into optoelectronic components such as transparent conductive films, touch screens, and solar

cells (59). To characterize this, as-synthesized Nc-G films were transferred onto quartz substrates

for UV-Vis measurements. Using the same black quartz substrate as reference, the UV-Vis shows

an optical transmittance of 97.7% at 550 nm for Nc-G film, comparable to its intrinsic

counterpart (fig. S18A). Furthermore, multilayer graphene still exhibits good transmittance

(higher than 90%), consistent with other works on fabricating multilayer graphene using a

layer-by-layer transfer method (fig. S18A) (38, 60).

(2) Large-scale conductivity of monolayer and multilayer Nc-G films

During the transfer of graphene, before the removal of the polymethyl methacrylate (PMMA)

layer, the composite structure was heated at 150 oC for an additional hour to minimize the

transfer-related charged impurities. The sheet resistance of large-scale films was characterized

using a four-probe resistance measuring meter to eliminate contact resistance. Four metal probes

were configured in a straight line at intervals of 1 mm, which is suitable for large-area

characterization of film conductivity (fig. S18B, inset). Note that the conductivity value on a

centimeter-sized scale is usually larger than that obtained from the micrometer-sized devices,

presumably owing to unavoidable transfer-related wrinkles on a larger scale. For a better

understanding of the influence of nitrogen doping in enhancing conductivity, the monolayer

intrinsic graphene film, consisting of the same domain size as the Nc-G film, was also

characterized. In contrast to the sheet resistance of intrinsic graphene (~720 Ω sq-1

), a clearly

reduced value of around 155 Ω sq-1

for the Nc-G film was observed, confirming the significantly

enhanced conductivity caused by the nitrogen-cluster doping (fig. S18B).

To enhance the conductivity, multilayer Nc-G films were fabricated by PMMA-assisted

layer-by-layer transfer (38,60). In particular, CVD-grown doped graphene exhibits no screening

of the dopant, which is typical in doped graphene fabricated by post-treatment doping techniques.

Consequently, the sheet resistances of bilayer graphene and trilayer Nc-G films are in the range

of 102 and 58 Ω sq-1

, respectively (fig. S18B), confirming the capability of further reducing sheet

resistance through forming multilayer structures. No change in the sheet resistance was observed

in transferred monolayer, bilayer and trilayer Nc-G after being kept in the air for at least 7 days.

Fig. S18. The large-scale conductivity and transmittance of Nc-G films. (A) UV-vis

transmittance spectra of the monolayer, bilayer and trilayer Nc-G films and the intrinsic

graphene film. (B) The sheet resistance of monolayer intrinsic graphene (green), also containing

1-mm sized domains and continuous monolayer, bilayer (2L) and trilayer (3L) Nc-G film (red),

fabricated by a layer-by-layer transfer technique. Inset: Photograph of graphene film transferred

onto SiO2/Si substrate and the four-probe station for characterizing the sheet resistance. (Photo

Credits: Li Lin, Peking University)

Fig. S19: Potential application of Nc-G films

The large-scale mapping results were presented in fig. S19A, indicating a good conductivity

uniformity, which is important for further applications, such as touch screen. The detailed

fabrication process of touch screen devices is as follows: First, the as-grown Nc-G film on Cu

foil is transferred onto a polyethylene terephthalate (PET) substrate with the assistance of

thermal release tape (Nanjing XFNANO Materials Tech Co., Ltd) using a roll-to-roll transfer

method (38). In our four-wire touch screen device, the Nc-G on foil serves as the bottom panel,

with commercial ITO-PET as the top panel (fig. S19B). Demonstration of touch screen is

presented in fig. S19C and movie S1. Note that the uniformity of the as-transferred Nc-G films,

for instance, is subject to cracking during the transfer, which would strongly influence the

performance of the touch screen. Furthermore, due to their enhanced conductivity, Nc-G films

can also be used as a conductive channel for powering light emitting diode (LED) indicators (fig.

S19D).

Fig. S19. Potential application of Nc-G films. (A) The spatial distribution of sheet resistance

on a 4 cm × 4 cm Nc-G film transferred onto PET. (B) Schematic of a Nc-G film-based touch

screen panel structure. (C) Demonstration of touch screen based on Nc-G. (D) Photograph of

bilayer Nc-G films transferred onto a SiO2/Si substrate to lighten up a green LED indicator. The

sample size is 3 cm × 3 cm. (Photo Credits: Li Lin, Peking University)

Fig. S20: High electrostatic potential and quasi-bound states induced by nitrogen clusters

Fig. S20. High electrostatic potential and quasi-bound states induced by nitrogen clusters.

(A) Electrostatic potential of 3N-cluster doped Nc-G. (B) Energy bandgap, density of states and

the partial charge distribution of valance band maximum (VBM) and conducting band minimum

(CBM) of 3N-cluster doped Nc-G.

Movie S1. Demonstration of a flexible touch screen device made from Nc-G film.

Supplementary Method: transport measurement of Nc-G

(1) Device fabrication:

The Nc-G samples were transferred onto SiO2/Si substrates with marks for alignment,

heat-cleaned, and imaged by atomic force microscopy (AFM) to check for flatness. Next, the

graphene samples were etched into a Hall bar geometry using a PMMA etch mask (PMMA 950K

A2 @ 4000 rpm) with EBL and O2 (RIE). After the samples were patterned, AFM imaging was

performed again to ensure that the channel regions were free of winkles and residues. Finally,

after using EBL to design a PMMA mask, contacts were fabricated on the samples (5 nm Ti and

90 nm Au) using an electron-beam evaporator, followed by a standard metal lift-off technique.

The carrier mobilities of single substitutional nitrogen doped graphene and nitrogen doped

graphene containing a content of pyridinic nitrogen (Fig. 1E) were also characterized.

(2) Sheet resistance measurements:

Four-probe measurements of the sheet resistance at room temperature in a vacuum probe station

were performed using a Keithly Semiconductor Characterization System (Model 4200-SCS) in

the DC current-bias sweep mode. The four-point configuration applies the sweep current bias to

the outside probes and measures the voltage between the inside probes to eliminate contact

resistance.

(3) Nonlinear fitting method for field-effect (FET) mobility:

The carrier density (electrons or holes) in the graphene channel regions ntot can be approximated

by (61)

𝑛𝑡𝑜𝑡 = √𝑛(𝑉𝑔)2 + 𝑛02 (S2)

where n0 represents the density of carriers at the Dirac point (denoted: residual carrier density).

If using the two-terminal method, the metal/graphene contact resistance should be taken into

account during the fitting

𝑅to t = 𝑅contact + 𝑅channel = 𝑅contact + L

W

1

e u√𝑛(𝑉𝑔)2+𝑛02 (S3)

where Rcontact represents the contact resistance, L and W represent the channel length and width,

and μ represents the FET mobility. Ignoring quantum capacitance, the gate-induced carrier

density can be calculated by

𝑛(𝑉𝑔) =𝐶𝑜𝑥∗(𝑉𝑔−𝑉𝑑𝑖𝑟𝑎𝑐)

𝑒 (S4)

In the four-terminal method, contact resistance is eliminated, and Rtot can be written as

𝑅tot = 𝑅channel = 𝐿′

W

1

e 𝑢√𝑛(𝑉𝑔)2+𝑛02 (S5)

where L′is the distance between the two inner probes.

(4) Low temperature magneto-transport measurements:

Shubnikov-de Haas oscillations of Rxx were observed at magnetic fields below 700 mT (Fig. 3A,

inset), while Rxy revealed well-developed plateau-like structures coinciding with Rxx minima at

higher magnetic fields. Low temperature magneto-transport measurements of the graphene

device fabricated on a SiO2/Si substrate show well-developed plateaus at filling factors 2, 6, 10,

and 14 at a quite low magnetic field of 4 T, indicating a high electronic quality (46, 61, 62) (Fig.

3B).

Table S1. Mobilities and sheet resistance of previously reported intrinsic graphene.

Table S2. Mobilities, conductivity, and stability of previously reported doped graphene.

Notes: the method for calculating conductivities is similar with that in fig. S7. The nμ values,

obtained from literature, is not the experimental result, which is only used for reelecting material

conductivity (13).