# Fabrication of gallium nitride and nitrogen doped single layer graphene hybrid heterostructures for high performance photodetectors

Sep 2, 2020

### Structural analysis

The structural information of GaN grown on N-SLG substrates by varying the experimental conditions were obtained using XRD as shown in Fig. 1. Dominant diffraction peaks corresponding to GaN were obtained at (left( {1 0 – 1 0} right)), (left( {0 0 0 2} right)) and (left( {1 0 – 1 1} right)) planes. The indexed peaks revealed that GaN exhibit hexagonal crystal structure and match well with the JCPDS data. The peak dominant at (left( {0 0 0 2} right)) plane confirm that the growth of GaN is along c-axis. These amplified peaks exemplify the crystalline quality of the GaN. It is worth to note that the full width at half maximum (FWHM) value estimated for the GaN epitaxially grown on N-SLG substrates at (left( {0 0 0 2} right)) plane is around 0.75°, indicating good growth along c-axis. It can be seen from literature that Chung et. al attained almost similar FWHM (0.81°) for GaN films grown on CVD graphene33. Similarly, Yadav et. al and Sanjay et. al achieved FWHM of around 3–6° for GaN films grown on ZnO and few layer graphene (FLG) respectively34,35. Wong et. al obtained a polytype growth for gallium-based compound materials grown on multi-layer graphene (MLG)36. Hence, it is evident that, irrespective of number of layers (SLG / MLG / FLG), presence of graphene as an intermediate layer has a critical role in enhancing the crystalline quality of GaN.

### Surface morphology analysis

The schematic portraying GaN growth process on N-SLG substrate in a CVD reactor is shown in Fig. 2 which gives a clear picturization about the growth process.

The surface morphology of GaN grown on N-SLG substrates were obtained using SEM as shown in Fig. 3. The growth of GaN was carried out in two different categories by varying the: (a) growth time as 60, 90 and 120 min, referred as sample X, Y and Z respectively and (b) precursor-to-substrate distance as 3, 5 and 7 cm, referred as sample Xʹ, Yʹ and Zʹ respectively.

In respect of growth time of 60 min (Fig. 3a), densely packed GaN flakes were observed on the N-SLG substrates. It may be noted that, initially sparse nuclei islands will be formed during the nucleation stage. These islands get attached to the dangling bonds present in graphene which occurred as an effect of residual oxygen (O) present within the CVD reactor. The oxygen atoms tend to taint the gallium metal (used as gallium precursor) by forming gallium oxide (GaOx) complex. This produces GaOx droplets on the N-SLG substrates during the initial growth process. However, as the growth proceeds, the oxygen atoms are replaced by nitrogen atoms37. As an effect of growth temperature, re-crystallization of islands is favoured, leading to the formation of dense GaN flakes like morphology38. When the growth time was increased to 90 min (Fig. 3b), the nucleation rate increased, resulting in the formation of more dense nucleation layers by adsorbing more and more gallium and nitrogen adatoms. These dense nucleation layers provide ample amount of adsorption sites on the surface and favours more re-crystallization leading to the formation of large nuclei islands. When the growth time was further increased to 120 min (Fig. 3c), the 3D islands became larger, resulting in coalescence of islands. This tends to transform the GaN from 3D islands to quasi 2D films39. However, in respect of precursor-to-substrate distance, only marginal variations in the morphology of GaN films were noticed (Fig. 3(d-f)). This is because, as reported in the literature, gallium vapour concentration decreases with increase in the precursor-to-substrate distance, resulting in shortage or improper supply of gallium species. The results obtained are in good agreement with the literature40.

The energy dispersive x-ray (EDX) spectroscopy maps obtained using SEM were utilized to analyse the material composition of GaN grown on N-SLG surface as shown in Fig. 4. It is seen from the Figure that in respect of all the samples, carbon (C) and oxygen traces were observed in addition to gallium and nitrogen. The carbon and oxygen traces can be attributed either as extrinsic contamination or contamination induced during the GaN growth process. The effect of oxygen contamination occurring during the growth process has been discussed already. The carbon contamination can be due to the decomposition of NH3 into nitrogen atoms as an effect of growth temperature41. This process occurs simultaneously along with the oxygen contamination process and tends to dope nitrogen atoms into the graphene due to the presence of dangling bonds. Therefore, nitrogen doping will benefit in not only replacing the oxygen atoms in the GaOx complex with nitrogen, but also in improving the material properties of graphene. According to literature, nitrogen atoms gets doped into graphene either as graphitic-N or pyridinic-N or pyrrolic-N configuration31. The process of doping tends to improve the electronic properties in graphene. Another advantage in doping is that, when nitrogen atoms get incorporated into the graphene’s honeycomb lattice, it tends to distort the crystal structure by creating disorders42. This is because, the lattice difference between GaN and graphene is around 29.6%. The high strain energy between them lead to the structural deformation in graphene thereby, making it defective. Therefore, doping SLG with nitrogen is also expected to improve nucleation rate during the growth of GaN. Similar phenomena were reported by Y. Gohda et al. using density functional theory (DFT) calculations. He illustrated that GaN grown on graphene tend to break the sp2 C–C bonds partially by forming C-N–C bonding to overcome the in-plane strain in the graphene lattice43. Also, M. Gruart et al.44 observed the formation of rough and imperfectly coalesced GaN clusters along defective regions in the graphene, which is consistent with the results obtained by us.

Therefore, it is concluded that, the above three processes such as: formation of GaOx complex, nitrogen doping into the honeycomb lattice of graphene and defect formation due to lattice difference between graphene and GaN, tend to create carbon, nitrogen and oxygen related point defects such as vacancies, interstitials and substitutional impurities and structural distortions in GaN. However, it is also to be noted that these actions have directly or indirectly contributed towards the nucleation process for the growth of GaN and also helped in improving the material properties.

### Surface composition analysis

Incorporation of nitrogen, carbon and oxygen atoms into graphene cannot be justified based only on SEM and EDX results. Hence, the samples were characterized using XPS to determine the surface composition (Fig. 5). The obtained peaks corresponding to nitrogen, carbon and oxygen were deconvoluted to obtain their functionalities. The carbon peak was deconvoluted into three subpeaks associated to C = C (({text{sp}}^{2}) hybridized), C = N (({text{sp}}^{2}) hybridized) and C–N (({text{sp}}^{3}) hybridized) located at 284.5, 285.8 and 287.5 eV respectively (Fig. 5a). From the carbon spectrum, no peak was detected at 289 eV, emphasizing absence of physisorbed oxygen in the graphene31,45. This concludes that, a substantial quantum of oxygen atoms were replaced with nitrogen atoms either during the growth of GaN or when nitrogen atoms are getting doped into graphene as substitutional impurities, leading to the formation of C–N bonds.

The deconvolution of oxygen spectrum revealed two subpeaks at 531.4 and 532.4 associated to C = O and C–O respectively, which could be attributed to the presence of different oxygen functionalities (Fig. 5b)31,42. The presence of oxygen functionalities could be due to the effect of GaOx complex formation during the initial nucleation stages of GaN.

Nitrogen peak was deconvoluted into four subpeaks associated to pyridinic-N, pyrrolic-N, graphitic-N and amino-N located at 398.4, 400.1, 401.4 and 399.1 eV respectively (Fig. 5c). It is to be noted that, pyridinic-N is ({text{sp}}^{2}) hybridized, where nitrogen atoms specifically get attached to two carbon atoms at the edges or defects of graphene. Whereas, pyrrolic type is ({text{sp}}^{3}) hybridized, in which nitrogen atoms binds into the five-membered carbon ring. Similar to pyridinic-N, graphitic-N is also ({text{sp}}^{2}) hybridized, where nitrogen atoms substitute for carbon atoms within the hexagonal ring31,42,45,46. It is to be noted that, the amount of nitrogen atoms getting doped into SLG will be very low. This is due to high temperature utilized for the growth of GaN which breaks most of the C-N bonds formed42,47. Thus, the elemental traces of nitrogen getting doped into the SLG were observed and confirmed.

### Luminescence characteristics

The luminescence characteristics of GaN samples grown on SLG substrates by altering the time of growth and precursor-to-substrate distance are shown in Fig. 6. It can be seen that, a strong emission peak, known as near band edge (NBE) was observed at around 3.45 eV. The NBE emission is attributed to the radiative recombination of electron from excited state to the ground state48,49.

In addition to NBE, a weak, yet a broad emission peak known as yellow luminescence (YL) was observed at around 2.2 eV. It is to be noted that, irrespective of any growth technique and / or experimental conditions, the presence of YL can be observed usually in undoped and n-doped GaN. The YL is attributed to the radiative transition from a shallow donor to a deep acceptor or due to point defects such as gallium vacancies (left( {{text{V}}_{{{text{Ga}}}} } right)), oxygen substituting gallium (left( {{text{O}}_{{{text{Ga}}}} } right)) or oxygen substituting nitrogen (left( {{text{O}}_{{text{N}}} } right))48,49,50,51,52,53 and surface defects such as excitons bound to shallow or structural defects, stacking faults and screw dislocations54,55,56,57,58,59,60,61,62.

It is worth to note that, the presence of N-SLG underneath GaN has effectively quenched the YL by a greater extent when compared with the YL in GaN without interlayer. However, it was inferred that, the presence of SLG is also a reason for the YL observed in GaN. This is because, carbon atoms get doped into GaN in various forms such as carbon substituting nitrogen (left( {{text{C}}_{{text{N}}} } right)), carbon substituting gallium (left( {{text{C}}_{{{text{Ga}}}} } right)), forming carbon – oxygen complex substituting nitrogen (left( {{text{C}}_{{text{N}}} – {text{O}}_{{text{N}}} } right)) and complex between gallium vacancy and carbon (left( {{text{C}}_{{text{N}}} – {text{V}}_{{{text{Ga}}}} } right)). The YL is a transition either from a shallow donor to a deep acceptor or from a deep donor to a shallow acceptor, which is in good agreement with the literature48,49,50,51,52,53. Therefore, carbon acts as a parasite degrading the material quality of GaN.

Comparing with literature, M. Heilmann et al.25 attained the ratio of intensities between YL and NBE (left( {{text{I}}_{{{text{YL}}}} /{text{I}}_{{{text{NBE}}}} } right)) value at around 2 for GaN micro- and nano-rods grown on SLG substrates, emphasizing the fact that YL dominates. Similarly, H. Yang et al.63 attained the value of ({text{I}}_{{{text{YL}}}} /{text{I}}_{{{text{NBE}}}}) as 5 for GaN grown on reduced graphene oxide (rGO) sheets. However, S.J. Chae et al.64 did not observe YL peak but, observed the donor–acceptor pair (DAP) transitions for GaN grown on FLG, where DAP is attributed to the recombination of electron bound to a shallow donor with a diffuse hole bound to a shallow acceptor. By comparing the above results, S. Sanjay et al.35 attained the value of ({text{I}}_{{{text{YL}}}} /{text{I}}_{{{text{NBE}}}}) at around 0.4 for GaN pyramids grown on FLG substrates, resulting in the suppression of YL peak. In the present scenario, from the Figure, the value of ({text{I}}_{{{text{YL}}}} /{text{I}}_{{{text{NBE}}}}) is estimated at around 0.05 – 0.15, confirming that carbon also contributes to the YL in GaN but, to a very small extent. This can be correlated to the defects in the graphene occurring during the GaN growth process. Thus, the PL results obtained corroborate with SEM data.

From the Figure, it was observed that, irrespective of variations in either growth time or precursor-to-substrate distance, there were no significant change in the peak position of NBE. However, slight variations were observed in relation to the peak intensities of the samples. On comparing the results obtained by S. Sanjay et al.35, it is evident that YL increases with number of layers.

### Electrical and optical studies

Based on the surface morphologies obtained using SEM, the samples Z and Zʹ were preferred for carrying out the current–voltage (I–V) characteristics (Fig. 7). This is because, sample Z represents the transformation in morphology of GaN from 3 to 2D. Also, sample Zʹ portray better surface morphology in comparison to all other samples. Therefore, in the present study, only these two samples were preferred for obtaining the electrical performance.

The electrical response was measured at an applied bias voltage of 0.5 V, both in the presence (bright) and absence (dark) of light as shown in Fig. 7. It was observed from the Figure that the obtained I–V curves were linear and symmetrical, confirming that the carrier density in GaN is high. Since the electron affinity of GaN (left( {{upchi }_{{{text{GaN}}}} = 4.2;{text{eV}}} right)) is less than the work function of Au (left( {phi_{{text{n}}} = 5;{text{eV}}} right)), the contacts formed will be of Schottky type35.

In the case of dark condition, the charge transfer across the metal–semiconductor (MS) interface is to be considered, which is obtained by calculating the barrier height (left( {phi_{{text{B}}} } right)) and ideality factor (left( {upeta } right)) between the MS junctions. The expression for calculating (phi_{{text{B}}}) and ({upeta }) can be given as65

$$phi_{{text{B}}} = left( {{raise0.7exhbox{{{text{KT}}}} !mathord{left/ {vphantom {{{text{KT}}} {text{q}}}}right.kern-nulldelimiterspace} !lower0.7exhbox{{text{q}}}}} right){ln}left( {{raise0.7exhbox{{{text{AA}}^{*} {text{T}}^{2} }} !mathord{left/ {vphantom {{{text{AA}}^{*} {text{T}}^{2} } {{text{I}}_{{text{s}}} }}}right.kern-nulldelimiterspace} !lower0.7exhbox{{{text{I}}_{{text{s}}} }}}} right)$$

(1)

$${upeta } = left( {{raise0.7exhbox{{text{q}}} !mathord{left/ {vphantom {{text{q}} {{text{KT}}}}}right.kern-nulldelimiterspace} !lower0.7exhbox{{{text{KT}}}}}} right) cdot left( {{raise0.7exhbox{{partial {text{V}}}} !mathord{left/ {vphantom {{partial {text{V}}} {partial (ln {text{I}})}}}right.kern-nulldelimiterspace} !lower0.7exhbox{{partial (ln {text{I}})}}}} right)$$

(2)

where ({text{A}}) is the area of Schottky contact, ({text{q}}) is the electronic charge, ({text{V}}) is the applied voltage, ({text{T}}) is the temperature in Kelvin, ({text{K}}) is the Boltzmann constant (left( {1.38{ } times { }10^{ – 23} ;{text{m}}^{2} ;{text{kg}};{text{s}}^{ – 2} ;{text{K}}^{ – 1} } right)), ({text{A}}^{*}) is the Richardson’s constant ((26.8;{text{A}};{text{cm}}^{ – 2} ;{text{K}}^{ – 2}) for GaN) and ({text{I}}_{{text{s}}}) the reverse saturation current derived from thermionic emission model as shown below65

$${text{I}} = {text{I}}_{s} left[ {{text{e}}left( {{raise0.7exhbox{{{text{qV}}}} !mathord{left/ {vphantom {{{text{qV}}} {upeta {text{KT}}}}}right.kern-nulldelimiterspace} !lower0.7exhbox{{upeta {text{KT}}}}}} right) – 1} right]$$

(3)

$${text{I}}_{{text{s}}} = {text{AA}}^{*} {text{T}}^{2} {exp}left( {{raise0.7exhbox{{ – {text{q}}phi_{{text{B}}} }} !mathord{left/ {vphantom {{ – {text{q}}phi_{{text{B}}} } {{text{KT}}}}}right.kern-nulldelimiterspace} !lower0.7exhbox{{{text{KT}}}}}} right)$$

(4)

The barrier height and ideality factor in respect of sample Z and Zʹ were estimated as 0.81 – 0.83 eV and 0.90 – 0.91 respectively. It was observed that, the barrier height and ideality factor values ascertained are in good agreement with the theoretical limit, confirming the MS interface is free from barrier inhomogeneities and interfacial surface states. From the Figure, the dark current (left( {{text{I}}_{{{text{dark}}}} } right)) for sample Z and Zʹ were determined as 0.2 and 0.4 A respectively. Even though, both the samples exhibit high values of dark current, the value exhibited by sample Z is less than that of sample Zʹ, which could be attributed to a poor interface in sample Z (where the morphology transforms from 3 to 2D). Irrespective of type of morphologies, the presence of SLG underneath GaN benefited in improving the dark current.

Similarly, under bright condition, the photo conversion efficiency across the MS interface is to be considered, which is obtained by estimating the photo current (left( {{text{I}}_{{{text{photo}}}} } right)), sensitivity (left( {text{S}} right)), responsivity (left( {text{R}} right)) and detectivity (left( {text{D}} right)) using the following Eqs. 66,67,68.

$${text{I}}_{{{text{photo}}}} left( {text{A}} right) = {text{I}}_{{max}} – {text{I}}_{{{text{dark}}}}$$

(5)

$${text{S}} = left( {{raise0.7exhbox{{{text{I}}_{{{text{photo}}}} }} !mathord{left/ {vphantom {{{text{I}}_{{{text{photo}}}} } {{text{I}}_{{{text{dark}}}} }}}right.kern-nulldelimiterspace} !lower0.7exhbox{{{text{I}}_{{{text{dark}}}} }}} } right) times { }100{text{% }}$$

(6)

$${text{R }}left( {{text{A}}/{text{W}}^{ – 1} } right) = {raise0.7exhbox{{{text{I}}_{{{text{photo}}}} }} !mathord{left/ {vphantom {{{text{I}}_{{{text{photo}}}} } {{text{A}} cdot {text{E}}}}}right.kern-nulldelimiterspace} !lower0.7exhbox{{{text{A}} cdot {text{E}}}}}$$

(7)

$${text{D }}left( {{text{Jones}},left( {{text{or}}} right),{text{cm}};{text{Hz}}^{0.5} ;{text{W}}^{ – 1} } right) = { }left( {{raise0.7exhbox{{text{R}}} !mathord{left/ {vphantom {{text{R}} {left( {sqrt {2 cdot {text{q}} cdot } {text{I}}_{{{text{dark}}}} } right)}}}right.kern-nulldelimiterspace} !lower0.7exhbox{{left( {sqrt {2 cdot {text{q}} cdot } {text{I}}_{{{text{dark}}}} } right)}}}} right)$$

(8)

where, ({text{A}}) is the device active area (left( {1 times 0.1;{text{cm}}^{2} } right)), ({text{E}}) is the light intensity (left( {100;{text{mW}};{text{cm}}^{ – 2} } right)) and ({text{q}}) is the coulombic charge (left( {1.6 times 10^{ – 19} ;{text{C}}} right)).

From the above equations, the photo current (0.6 and 0.8A), sensitivity (150 and 400%), responsivity (150 and 400 ({text{A}}/{text{W}}^{ – 1})) and detectivity (1.08 × 1012 and 2.89 × 1012 Jones in respect of samples Z and Zʹ were determined. It was observed from the Figure that, the photo current got enhanced several times the dark current in case of sample Z and Zʹ respectively. This is because, when the light is illuminated on the samples, electron–hole pairs (excitons) are generated in the GaN. The charge carriers thus generated flow in the graphene through the applied bias voltage resulting in photocurrents. As mentioned earlier, due to the interfacial differences, sample Z exhibit less photo current value when compared to sample Zʹ. It is to be noted that, the presence of a SLG underneath GaN with negligible defects (({text{I}}_{{{text{YL}}}} /{text{I}}_{{{text{NBE}}}}) from PL) and improved material quality (N-SLG) has contributed towards drastic enhancement in photodetector performances.

For practical opto-electronic applications, high speed switching in the photo response is a must. Since the device performance of sample Zʹ is better than that of sample Z, the switching response of the photodetector is carried out only for sample Zʹ as shown in Fig. 8. The switching response of the photodetector for the light under ON and OFF conditions can be fitted using the exponential functions based on the rise and fall time of the photodetector. The rise and fall time consist of two exponentials with different slopes (initial fast response and then a slower response). The exponential functions corresponding to rise and fall time are given as69,70

$${text{I}}left( {text{t}} right)_{{{text{ON}}}} = {text{I}}_{1} left[ {1 – {text{A}}_{1} {exp}left( {{raise0.7exhbox{{ – {text{t}}}} !mathord{left/ {vphantom {{ – {text{t}}} {{uptau }_{1} }}}right.kern-nulldelimiterspace} !lower0.7exhbox{{{uptau }_{1} }}}} right) – {text{A}}_{2} {exp}left( {{raise0.7exhbox{{ – {text{t}}}} !mathord{left/ {vphantom {{ – {text{t}}} {{uptau }_{2} }}}right.kern-nulldelimiterspace} !lower0.7exhbox{{{uptau }_{2} }}}} right)} right]$$

(9)

$${text{I}}left( {text{t}} right)_{{{text{OFF}}}} = {text{I}}_{2} left[ {{text{A}}_{3} {exp}left( {{raise0.7exhbox{{ – {text{t}}}} !mathord{left/ {vphantom {{ – {text{t}}} {{uptau }_{3} }}}right.kern-nulldelimiterspace} !lower0.7exhbox{{{uptau }_{3} }}}} right) + {text{A}}_{4} {exp}left( {{raise0.7exhbox{{ – {text{t}}}} !mathord{left/ {vphantom {{ – {text{t}}} {{uptau }_{4} }}}right.kern-nulldelimiterspace} !lower0.7exhbox{{{uptau }_{4} }}}} right)} right]$$

(10)

where ({uptau }_{1})({uptau }_{4}) are the time constants and ({text{A}}_{1})({text{A}}_{4}) are the numerical constants.

By fitting the above equations, the rise and fall time in respect of sample Zʹ is determined as 10 and 25 ms. When the light was turned ON, the photocurrent was found to rise. Similarly, when the light was turned OFF, the photocurrent was found to get dropped. In the Eqs. (9) and (10), ({uptau }_{1}) represents the transfer of hole from GaN to SLG, ({uptau }_{2}) represent the transfer of charge in GaN, ({uptau }_{3}) represents the lifetime of electrons in GaN before they get transferred to neighbouring graphene film and ({uptau }_{4}) represents the transfer of charge in GaN70.

The photocurrent response of the photodetector is carried out under ON/OFF condition as shown in Fig. 8. The device response remained similar even after switching the device several times. The obtained photo response is found to be faster than that of the photodetectors based on the undoped SLG reported earlier66,71. It is worth to note that response time is dependent on transfer rate of electrons and holes from GaN to SLG. Therefore, the nature of interface between them is crucial to the performance of the devices.

### Hall measurement study

The room temperature Hall measurements were carried out for sample Zʹ of size 10 mm2. It was observed that, the bulk concentration and sheet concentration of the GaN grown on N-SLG substrate was around 7.52 × 1018 cm−3 and 1.51 × 1015 cm−2 respectively. The resistivity and carrier mobility of the sample obtained was around 0.00345 Ω∙cm−1 and 440 cm2 V−1 s−1 respectively. The enhancement in electrical properties can be due to shift in the fermi level above the Dirac point of SLG as an effect of nitrogen doping, which tends to suppress the density of states near the fermi level. This action opens up the bandgap between the conduction band and valence band. In such a condition, SLG behaves as a semiconductor. As mentioned earlier, nitrogen doping regulates the property of graphene by maintaining good conductivity and carrier mobility31. The obtained mobility value is higher than that of the standard mobility value of GaN and also falls within the mobility range for nitrogen doped SLG as reported in literature46,72. The hall measurement data corroborate with I–V results.