Aluminum nitride (AlN) is an insulator that has shown little promise to be converted to a semiconductor via impurity doping. Some of the historic challenges for successfully doping AlN include a reconfigurable defect formation known as a DX center and subsequent compensation that causes an increase in dopant activation energy resulting in very few carriers of electricity, electrons, or holes, rendering doping inefficient. Using crystal synthesis methods that generate less compensating impurities and less lattice expansion, thus impeding the reconfiguration of dopants, and using new dopants, we demonstrate: (a) well behaved bulk semiconducting functionality in AlN, the largest direct bandgap semiconductor known with (b) substantial bulk p-type conduction (holes = 3.1 × 1018 cm−3, as recently reported in our prior work), (c) dramatic improvement in n-type bulk conduction (electrons = 6 × 1018 cm−3, nearly 6000 times the prior state-of-the-art), and (d) a PN AlN diode with a nearly ideal turn-on voltage of ∼6 V for a 6.1 eV bandgap semiconductor. A wide variety of AlN-based applications are enabled that will impact deep ultraviolet light-based viral and bacterial sterilization, polymer curing, lithography, laser machining, high-temperature, high-voltage, and high-power electronics.

Next generation high-power electronic and optoelectronic devices demand further advances in semiconductor materials for deep ultraviolet (UVC) optoelectronics, high-performance power conversion, and radio frequency (RF) devices. For example, multi-kilovolt switching elements necessary for power grid applications must presently be constructed from many devices because of insufficient breakdown characteristics of even the best wide bandgap semiconductors known today. Significantly higher performance can be achieved with ultrawide bandgap (UWBG) semiconductors like Ga2O3 but at the cost of poor thermal conductivity (27 W/m K), causing inadequate power dissipation.1 AlN (Eg = 6.1 eV) is an extreme bandgap material (EBM) with superb thermal conductivity (319 W/m K) that, if made to conduct electrically, could revolutionize high-power and high-temperature applications and could operate under extreme environments and high radiation.2 The extreme bandgap of AlN results in far lower intrinsic carrier concentrations at high-temperatures resulting in reduced leakage currents for high-temperature operation. The larger bandgap/critical field also allows higher blocking voltage in power devices and the generation of more energetic photons in the deep UVC range approaching 200 nm wavelengths suitable for water purification, sterilization, and biological threat neutralization. The critical electrical field for AlN is 15.4 MV/cm,1 much higher than all other power device relevant wide bandgap semiconductors including GaN (4.9 MV/cm), β-Ga2O3 (9 MV/cm), and SiC (3.1 MV/cm).1 High quality bulk AlN substrates (dislocation density of ∼102–104 cm−2) are already available for epitaxial growth of low defect density device structures.3 AlN's high thermal conductivity also affords simplified heat extraction schemes.1 The only limitation up to now in AlN-based devices was the lack of doping relegating AlN to insulating applications, though some success has been achieved with heterojunctions, such as the successful demonstration of a n-AlN/p-diamond diode.4 Recently, we achieved bulk p-type AlN with hole concentration up to 3.1 × 1018 cm−3.5 Until this report for n-type AlN, no more than 1015 cm−3 bulk electron concentration has been shown in the literature.6,7 Once doping capability is achieved, the next evolutionary step is to demonstrate a PN homojunction diode, the basic core structure of power diodes, transistors, LEDs, lasers, photodiodes, and many more devices.

Previously, the major limitation in achieving AlN-based electronic devices was the lack of a suitable dopant, making its theoretical potential for power switches irrelevant and resulting in GaN, β-Ga2O3, and SiC being used for power devices instead. P-type conductivity of AlN was a major challenge where reports of surface conductivity via carbon doping were shown but no bulk experimental conductivity had been demonstrated.8 Beryllium (Be) is the best possible p-type candidate in AlN because the Be–N bond energies and atomic radius of Be closely match with Al as shown in Table I and shown drawn to scale in Fig. 1. Recently, we successfully achieved p-type Be doped AlN films with hole concentrations up to 3.1 × 1018 cm−3 via the improved growth kinetics of metal modulated epitaxy (MME) demonstrating high quality films at unconventionally low substrate temperatures.6 MME utilizes low substrate temperatures during growth to lower contamination normally resulting from gaseous outgassing and uses multiple parameters to control the surface chemistry and kinetics to facilitate proper incorporation of the dopants on the cation site. MME has also demonstrated the highest known hole concentrations for p-type GaN.9–11 The MME AlN:Be p-type films were successfully applied to novel p-AlN/i-GaN/n-GaN heterojunction Schottky, junction barrier Schottky (JBS), and PIN diodes.12 In this later work, it was found that ultra-high vacuum purity was key. Low background pressures during growth resulted in higher hole concentrations and less compensation.

FIG. 1.

The drawn-to-scale crystal structure of AlN showing the incorporation of: Be (atomic radius of 112 pm) as a p-type substitutional impurity replacing Al (atomic radius of 118 pm); Si (atomic radius of 111 pm) as an n-type substitutional impurity substituting Al; N (atomic radius of 56 pm); DX center formation of Si by transitioning to a deep state to compensate n-type AlN; O acting as a donor by replacing N; O-DX center formation compensating n-type AlN.

FIG. 1.

The drawn-to-scale crystal structure of AlN showing the incorporation of: Be (atomic radius of 112 pm) as a p-type substitutional impurity replacing Al (atomic radius of 118 pm); Si (atomic radius of 111 pm) as an n-type substitutional impurity substituting Al; N (atomic radius of 56 pm); DX center formation of Si by transitioning to a deep state to compensate n-type AlN; O acting as a donor by replacing N; O-DX center formation compensating n-type AlN.

Close modal
TABLE I.

Atomic radii of the III-nitride host atoms and theoretical/experimental p-type and n-type dopants.30 

ElementAtomic radius (pm)
Al 118 
Be 112 
Si 111 
48 
56 
ElementAtomic radius (pm)
Al 118 
Be 112 
Si 111 
48 
56 

For n-type AlN, Si is the best theoretical dopant as its atomic radius closely matches with Al as shown in Fig. 1 and in Table I. Si as a substitutional impurity in AlN results in a 6% theoretical relaxation of the nearest N bonds.13 Promising results have been previously shown in literature for n-type doping of AlN near surface regions by ion implantation.14,15 However, no more than 1015 cm−3 bulk electron concentrations, suitable for high drift field regions in devices, were shown in n-type doping of bulk AlN films.6,7 This disparity in doping results, with one technique showing viability and another not, suggests that Si itself is not the problem but some other defect/impurity related species is the impediment and other doping methods may be more appropriate. Si is a shallow donor in GaN with activation energy of ∼17 meV but its activation energy in AlGaN increases with Al content from 24 meV for Al0.85Ga0.15N to 211 meV for Al0.96Ga0.04N.16,17 The problem with doping AlN with Si can be understood by considering the atomic locations in which Si sits in the crystalline lattice. The solubility limit of a dopant depends on the formation energy of the dopants. Formation energy further depends on the atomic radii matching, the bond strength of the host vs dopant atoms, and favorable geometric configurations of the dopant in the lattice. The activation energy of Si in AlN is above 200 meV due to the formation of Al vacancies,18 high threading dislocations trap electrons15,18 and DX center formation. The DX center forms when Si captures a secondary electron with a geometric rearrangement including a 2% contraction of 3 basal positioned Si–N bonds and the breaking of the c-axis Si–N bond as shown in Fig. 1, causing a transition from shallow to deep state.19,20 In the DX configuration shown in Fig. 1, the Si atom stays close to the substitutional Al-site but shifts downward as the c-axis bond breaks resulting from the contraction of the nearest N bonds.13 Complicating issues, the Al-vacancy forms a complex with Si resulting in self-compensation of the doping at high Si doping levels.18 Likewise, oxygen is a donor in AlN at low concentrations resulting in a 4% theoretical elongation of the nearest N bonds. However, as shown in Fig. 1, at higher doping concentrations, oxygen also forms a DX center where the 0.19 nm basal AlN bonds shift to asymmetric lengths of 0.182, 0.182, and 0.175 nm and result in the diagonal displacement of O toward the open space in the crystal. This DX center reconfiguration forms a deep state at higher O concentrations compensating n-type AlN.13,21 Furthermore, this self-compensation was found to increase in magnitude with an increase in threading dislocations.18 Thus, by growing the AlN films under highly crystalline conditions (reduced threading dislocations density) compensation of Si doping in AlN can be reduced. Vacancies, particularly Al vacancies, act to increase the likelihood of DX center formation by allowing easier reconfiguration due to larger voids in the otherwise dense crystal structure and by forming Si-vacancy and O-vacancy complexes that also act as deep centers.21 

Metal Modulated Epitaxy (MME) is a cyclic molecular beam epitaxy (MBE) derivative, which operates in an ultra-high vacuum, extremely pure, low impurity outgassing environment limiting background carbon, and oxygen to values typically in the 1015 to low 1017 cm−3 range. MBE or metal organic chemical vapor deposition (MOCVD) for III-Nitrides operates well above the desorption temperatures increasing epitaxy chamber outgassing of impurities and naturally creating exponentially higher concentrations of vacancies, Nvacancies, as governed by the vacancy production equation,

[Nvacancies]=[NAtomic]CeEformation/kT,

where NAtomic is the atomic concentration of missing element, C is the number of equivalent configurations of the vacancy, and Eformation is the energy needed to form the vacancy including the net energy required from breaking and reconfiguring atomic bonds.22 Contrarily, MME operates well below the desorption temperatures and at metal-rich surface conditions minimizing the vacancy production, especially for Al. MME compensates for the lower growth temperature with extremely metal-rich surface chemistry that virtually eliminates the harmful Al vacancies and allows easier surface bond breakage and, thus, long surface adatom diffusion lengths. For example, consider the surface diffusion equation,

Surfacediffusivity=[a2(ω4)]eϕ/KT,

where a is the hopping distance, ϕ is the energy barrier for hopping, and ω is the vibrational frequency. Given the metal-N barriers for hopping can be 15 times larger than for metal–metal surface hopping barriers,23 this barrier height discrepancy results in metal-rich surfaces having 5–6 decades longer diffusion lengths than semiconductor bond rich surfaces like ammonia-based MBE or MOCVD. Even with 500 degrees difference in temperature between MOCVD and MME, the much lower hopping barrier height, 0.1–0.2 eV, results in a longer surface diffusivity for MME than N-rich MOCVD because the N-Rich barriers are ∼10–15 times higher. We do note, however, that MME and MBE do not benefit from the enormous gas phase diffusion that more than compensates for the lack of surface diffusion owing to the enormous bond energy of N-rich surfaces, and, thus, MOCVD dominates present III-nitride production. This difference in surface diffusion length is evident in MME (and most MBE) surface morphologies vs that of MOCVD. MME and metal-rich MBE tend to show surfaces where spiral hillocks form around dislocations as the step flow growth is interrupted by the surface void found at dislocations.24 Conversely, MOCVD morphology is governed not by adatom diffusion but by gas phase diffusion and conformally covers regions of dislocations resulting in a flat surface even in the presence of dislocations that disrupt step flow growth of atoms on the surface. When paired with MME's metal-rich surfaces that increase adatom diffusion lengths for higher quality, MME can reduce Al-vacancy concentrations known to pair with silicon and oxygen to form deep centers and rob AlN of electrons via DX center formation.21 

Finally, we note that DX center formation requires geometric rearrangements of the dopants. As the lattice expands at the extremely higher temperatures (1400–2200 °C for many growth methods25,26 compared to 600–700 °C here) the 5.27 × 10−6 thermal expansion coefficient predicts an ∼2%–6% differential increase in c-axis elongation enhancing the likelihood of atomic rearrangements resulting specifically from the c-axis bond breakage, a requirement for DX center formation. In short, MME provides a high purity, low outgassing environment absent of Al vacancies with long adatom diffusion lengths and yields a denser crystal that is expanded to a lesser degree, making it less prone to crystalline rearrangement.

The atomic radii matching of Si and Al in AlN, and the optimal MME growth kinetics make a strong case to investigate n-type Si doped AlN films. N-type AlN in combination with the previously achieved p-type AlN:Be MME films5 completes the essential components to demonstrate homojunction AlN diodes.

The AlN:Si films and AlN homojunction diodes were grown in a Riber 32 plasma-assisted molecular beam epitaxy (PAMBE) system via MME on hydride vapor phase epitaxy (HVPE) AlN on sapphire templates from MSE Supplies. Two-in. diameter AlN on sapphire wafers obtained from MSES was first piranha (3:1 volume ratio of H2SO4:H2O2) cleaned for 1 min at 150 °C followed by 5:1 volume ratio of de-ionized water to hydrofluoric acid (DI H2O:HF) for 30 s. The cleaned wafers were later backside metalized with 2 μm Tantalum for uniform heating during growth. The backside metalized wafers were then diced into 1 × 1 cm2 templates. The metalized and diced AlN templates were subsequently solvent cleaned (acetone clean at 45 °C for 20 min, 3 min methanol clean, DI water rinse, and blown dry with nitrogen), followed by a 10-min piranha (3:1 volume ratio of H2SO4:H2O2) clean at 150 °C to remove organic solvents. The templates were then ex situ chemically cleaned in a 10:1 volume ratio of DI H2O:HF for 25 s to partially remove the surface oxides followed by DI water rinse and dried with nitrogen.

The AlN templates were immediately loaded into an introductory chamber with a base pressure of ∼10−9 Torr and thermally outgassed at 200 °C for 20 min. Later, the templates were moved through an analytical chamber into the growth chamber and outgassed at 850 °C for 30 min.

Al, Be, and Si fluxes were supplied from standard effusion cells. A Veeco UNI-bulb radio frequency (RF) nitrogen plasma source was used to supply nitrogen plasma during growth at a RF plasma power of 350 W and a flow rate of 2.5 sccm. The RF plasma power and flow rate were kept constant for all the growths. The MBE growth chamber base pressure was ∼5 × 10−11 Torr and the beam equivalent pressure (BEP) of the nitrogen plasma was ∼1.2 × 10−5 Torr. The growth rate was 700 nm/h for the AlN:Be films and 1.40 μm/h for the AlN:Si films. However, MME has previously demonstrated growth rates as high as ∼10 μm/h.27 The MME open/close shutter cycle scheme for the AlN:Be and AlN:Si samples is given in Table II. A STAIB Instruments RH20S 20 kV Reflection High Energy Electron Diffraction (RHEED) gun was used in combination with k-Space Associates kSA 400 analytical RHEED system to monitor in situ surface morphology and to calculate the run-time growth rates of the films. The AlN:Be films were grown at a III/V ratio of 1.3 and at a substrate temperature of 700 °C using precisely controlled excess metal coverage to compensate for the loss of adatom mobility at low temperatures. Al, N, and Be atoms hopping on a metal terminated surface need only to break weak metallic bonds that are substantially smaller than the strong AlN semiconductor bonds of a stoichiometric AlN surface. The dead time (see Ref. 9 for MME tutorial and detailed description of MME terms) was kept at 8.5 s to consume the metal and dopants in each cycle and not let the dopants diffuse vertically during growth. The metal-rich conditions result in smooth surface morphology of the films while the low substrate temperature helps in limiting Be-diffusion in the growth direction. The AlN:Si films were grown under high crystalline MME growth conditions not suitable for proper p-type doping at a substrate temperature of 800 °C and at a III/V ratio of 1.3.

TABLE II.

Description of the MME p-type AlN:Be and n-type AlN:Si films.

Sample IDMME shutter cycle open/closed times (s/s)Growth time per MME shutter cycle (s)
AlN:Be 5/10 6.5 
AlN:Si 21/11 27.3 
Sample IDMME shutter cycle open/closed times (s/s)Growth time per MME shutter cycle (s)
AlN:Be 5/10 6.5 
AlN:Si 21/11 27.3 

Pt/Pd/Au (10 nm/10 nm/100 nm) contact stacks were deposited in a Denton Explorer e-beam evaporation chamber for both n- and p-type AlN films, for Hall measurements (in the van der Pauw configuration), and for device characterization. The contacts were subsequently annealed under purified nitrogen inside a MILA-3000 rapid thermal annealing (RTA) furnace at 800 °C for 1 min p-type AlN films, and at 875 °C for 1 min for n-type AlN:Si films. Details of the effects of contact annealing on p-type AlN are published elsewhere.5 Secondary ion mass spectroscopy (SIMS) of a Si doped calibration sample was performed at Evans Analytical Group (EAG). A state-of-the-art Hall measurement tool, M91 FastHall Controller from Lake Shore Cryotronics Inc., was used for four-point resistivity and Hall effect measurements. The FastHall station has a 1 T magnet, a resistance measurement range of 1 mΩ–1 GΩ, and a mobility measurement range of 10−2–106 cm2/V s.

First, the Si incorporation into AlN was calibrated via SIMS. 150 nm thick MME AlN:Si layers of various Si doping were grown at a growth temperature of 800 °C, III/V ratio of 1.3, and MME O/C shutter cycles of 21 s/11 s. SIMS results were then used to guide the doping of thicker films used for Hall analysis. Specifically, 500 nm AlN:Si films were grown via MME on MSES HVPE AlN on sapphire templates at a substrate temperature of 800 °C, III/V ratio of 1.3, and MME O/C shutter cycles of 21 s/11 s with Si SIMS determined concentrations in the range of 5 × 1017–7 × 1019 cm−3 as summarized in Table III.

TABLE III.

Description of MME grown AlN:Si films at a growth temperature of 800 °C with their SIMS and Hall concentrations.

Sample IDSIMS concentration (cm−3)Hall concentration (cm−3)Electrically active donor ratioHall mobility (cm2/V s)
N4591 5 × 1017 Not measurable Not measurable Not measurable 
N4592 3 × 1018 9 × 1017 ± 1 × 1015 0.3 63 ± 0.1 
N4595 8 × 1018 5 × 1017 ± 1 × 1015 0.06 370 ± 0.7 
N4596 3 × 1019 1 × 1018 ± 5 × 1015 0.03 273 ± 2 
N4597 7 × 1019 6 × 1018 ± 5 × 1015 ∼0.08 17 ± 0.1 
Sample IDSIMS concentration (cm−3)Hall concentration (cm−3)Electrically active donor ratioHall mobility (cm2/V s)
N4591 5 × 1017 Not measurable Not measurable Not measurable 
N4592 3 × 1018 9 × 1017 ± 1 × 1015 0.3 63 ± 0.1 
N4595 8 × 1018 5 × 1017 ± 1 × 1015 0.06 370 ± 0.7 
N4596 3 × 1019 1 × 1018 ± 5 × 1015 0.03 273 ± 2 
N4597 7 × 1019 6 × 1018 ± 5 × 1015 ∼0.08 17 ± 0.1 

A metal stack of 10 nm Pt/10 nm Pd/100 nm Au was chosen as contacts to the AlN:Si films for Hall and resistivity measurements. The contacts were deposited at the corners of very large 1 × 1 cm2 samples via lithography and lift-off. First, the samples were cleaned via acetone, isopropanol (IPA), DI water, and dried with nitrogen followed by dehydration bake at 100 °C for 5 min. Subsequently, NR9-1500PY negative photoresist (PR) was spin coated at 3000 rpm for a dwell time of 40 s and at a ramp rate of 5 s followed by a pre-exposure bake at 150 °C for 60 s. The PR spin-coated and baked samples were then exposed under 365 nm ultraviolet light at a dose of 350 mJ/cm2 followed by a post-exposure bake at 100 °C for 60 s. The PR spin coated and exposed samples were then developed in RD6 for 10 s followed by a 1:1 ratio of buffered oxide etch (BOE):DI water clean for 30 s. The 10 nm/10 nm/100 nm Pt/Pd/Au contacts were deposited inside a Denton Explorer e-beam evaporator at a deposition rate of 0.1 nm/s with a background pressure of ∼1 × 10−6 Torr followed by a lift-off in acetone for 20 min. The samples were finally rinsed via IPA and DI water and dried with nitrogen. The lithography and lift-off process resulted in van der Pauw configuration for contact current–voltage linearity checks and Hall measurements.

After deposition of the contacts, the samples were then annealed in a MILA-3000 rapid thermal annealing (RTA) furnace. The annealing time was 1 min with a ramp-up and ramp-down time of 60 s each at 875 °C under a nitrogen environment.

A separate p-type sample, N4492, doped at 7 × 1018 cm−3 Be grown at a substrate temperature of 700 °C and MME Open/Closed cycle of 5 s/10 s was used for circular transmission line measurements (CTLMs) for contact resistance comparison of individual films vs device contacts.

For the PN diode N4633, first a 1 μm n-type AlN:Si film with Si doping of 8 × 1018 cm−3 was grown at a substrate temperature of 800 °C and MME Open/Closed cycle of 21 s/11 s on a ∼4 μm HVPE AlN on a sapphire template from MSES Inc. Then, a 200 nm AlN:Si film with Si doping of 5 × 1017 cm−3 “i-layer” corresponding to an unmeasurably low doping as shown in Table III was grown under the same conditions. This was followed by a 200 nm AlN:Be p-type film with Be doping of 7 × 1018 cm−3 grown at a substrate temperature of 700 °C and MME Open/Closed cycle of 5 s/10 s similar to Ref. 5.

After growth, 100 μm diameter quasi-vertical devices were fabricated on the sample using ICP plasma etching. The same metal stacks used for the individual layers above were used for the p- and n-type contacts except they were annealed at 950 °C under nitrogen environment for 1 min. The higher annealing temperature was determined by iterative cycles of anneals at a lower temperature, current measurement, and subsequent higher temperature anneals until the performance degraded. The higher rapid thermal annealing temperature for these devices seems to be related to the different metal coverage of the device mask compared to the contact study mask and likely is a result of AlN's transparency in the optical heated annealer.

Size (strain) dictates that Si is the best donor dopant atom substituting the Al atom in AlN. The atomic radius of Si (111 pm) closely matches with the atomic radius of Al (118 pm).28 The atomic radius matching of Si with Al in AlN in combination with the capability of MME to surpass the solubility limit of dopants in III-nitride materials via improved growth kinetics (non-equilibrium growth via rapid synthesis) was utilized to investigate Si doped AlN films. First, the Si incorporation into AlN was calibrated via secondary ion mass spectroscopy (SIMS). Multiple 150 nm thick MME AlN:Si layers of various Si doping were grown and SIMS results were then used to guide the doping of thicker films used for Hall analysis. Specifically, 500 nm AlN:Si films were grown with Si SIMS determined concentrations in the range of 5 × 1017–7 × 1019 cm−3 as summarized in Table III.

A metal stack of Pt/Pd/Au was deposited via lithography at the corners of very large 1 × 1 cm2 AlN:Si samples in a van der Pauw configuration for contact current–voltage linearity checks and Hall measurements. The use of large samples ensures that the measured properties are global properties and not merely local anomalies. After deposition of the contacts, the samples were then annealed via a rapid thermal annealing (RTA) furnace. The effect of the annealing process on the electrical contact properties of the samples was studied by investigating its I–V characteristics through a four-point probe measurement. I–V characteristics for the Pt/Pd/Au contacts on a representative MME grown films are shown in Fig. 2. The annealed N4595 AlN:Si film in Fig. 2(a) critically crosses zero current at zero voltage indicating that thermal voltages or piezoelectric offsets are not present. Also, the post-annealed AlN:Si contacts are highly linear. The post-annealed AlN:Si film (N4595) shown in Fig. 2(a) has ∼5 orders of magnitude higher current than a control undoped AlN film (N4436) shown in Fig. 2(b), which proves the increased conductivity was a result of the Si doping.

FIG. 2.

Current–voltage characteristics of the Pt (10 nm)/Pd (10 nm)/Au (100 nm) contacts on the annealed (a) N4595 AlN:Si film (b) N4436 AlN control undoped film.

FIG. 2.

Current–voltage characteristics of the Pt (10 nm)/Pd (10 nm)/Au (100 nm) contacts on the annealed (a) N4595 AlN:Si film (b) N4436 AlN control undoped film.

Close modal

The conductivity of the AlN:Si samples was investigated through Hall measurements. The contact resistance of the AlN:Si films was in the kΩ range, which is well within the measurement capability of the Lake Shore Hall tool. However, Hall measurements of the lowest doped AlN film, N4591, could not be performed due to a National Institute of Standards and Technology (NIST) “F-factor” symmetry coefficient of less than 95% in the various measured contact resistances.29,30 Hall measurements of the AlN:Si films in the Si doping range of 5 × 1017–7 × 1019 cm−3 show reliable results, F > 99% with electron concentrations in the range of 9 × 1017–6 × 1018 cm−3 as listed in Table III. The 6 × 1018 cm−3 electron bulk concentration in AlN is ∼6000 times higher than the previously reported state-of-the-art.6,7

Also shown in Table III is the electrically active donor ratio, defined as Hall concentration/SIMS concentration. This fraction must take into account compensation but also the natural doping efficiency of a degenerately doped semiconductor defined by ND+=ND1+gDe(EfED)/kT, where ND+ is the ionized (electrically active) doping concentration, gD is the spin degeneracy factor (2 for AlN), ED is the donor energy, Ef is the Fermi energy, ND is the substitutional impurity concentration, and kT is the thermal energy. This natural doping efficiency ranges from 1 when the Fermi energy is low (lightly doped, not applicable for our doping ranges), 1/3 when the Fermi energy is at the donor energy to vanishingly small values when Ef is even slightly above ED at high doping concentrations. While the observed doping efficiency values (∼0.3–0.03) are well within the expected values for this naturally occurring donor efficiency, the complicating role of compensation and impurity band formation is not known at this time. In particular, MME has been shown to result in significant impurity band formation,10 leading to higher doping efficiency than can be explained by an isolated dopant energy, requiring an integral of the above equation over the impurity band with a weighting by the dopant's density of states. We further note that sample N4595 has a lower electron concentration as compared to N4592 even though N4595 has a higher SIMS concentration. This can only occur if compensation is more significant in N4595 than N4592 or if the donor energy is concentration dependent. The latter is possible as the donor energy, ED, for best matching the experimental results for N4592 is ∼0.055 eV while for samples N4595, N4596, the best match is for ED = 0.115 eV. Finally, we note that N4597 can only be explained by impurity band formation (which is common to MME doping) because the measured electron concentration is higher than predicted from an isolated donor state theory and doping efficiency thus increases. Future temperature dependent transport studies must account for compensation, transport inside the impurity band, as well as in the conduction band but are beyond the present scope of this paper.

The van der Pauw resistivity analysis assumptions include a contact resistance that is less than the film resistance and, thus, can be averaged out by the van der Pauw method. Given these contacts' resistances are still high compared to the film resistance, this assumption is not valid, and a contact voltage drop added to the bulk resistivity voltage drop appears in the van der Pauw measurement making the resistivity (and the corresponding mobility) measurements merely estimates. Thus, we provide only preliminary electron mobilities that need additional verification. We note that since the current and voltage are measured from different contacts for the carrier concentration determination in Hall measurements, this contact voltage drop effect did not degrade the carrier concentration determination and all reported carrier concentrations have an uncertainty of less than 0.5%.

Figure 3 shows circular transmission line measurements (CTLMs) for N4492, a planar p-type film (not a device) with a constant outer radius of 200 μm and gaps of 25, 35, 45, 55, 65, and 75 μm. These are much smaller gaps than used for the Hall measurements (∼1 cm) of the p-type AlN:Be film from Ref. 5, and, thus, the current levels are significantly higher. The contacts of p-type film N4492 are highly linear carrying a significant current of ∼0.4 mA.

FIG. 3.

P-contacts transmission line measurement (PTLM) of N4492 AlN:Be film.

FIG. 3.

P-contacts transmission line measurement (PTLM) of N4492 AlN:Be film.

Close modal

For validation of the p-type nature of the AlN:Be films and the n-type nature of the AlN:Si films, an AlN homojunction diode showing a turn-on voltage comparable to the semiconductor bandgap is desired. In this regard, an AlN PN diode (N4633) was grown.

Figure 4 shows circular transmission line measurements (CTLMs) for n- and p-type contact layers of the PN diode structure. Both n- and p-type contacts all show a linear trend. However, the current levels of both the n- and p-type contacts for the N4633 AlN diode are repeatably lower for multiple fabricated devices than the n-type N4595 and p-type N4492 films (non-device) presumably due to an anomaly from the ICP tool during etching and annealing during the fabrication process, specifically, when the PN diode was etched a dark tint formed in the wafer from an unknown origin, suggesting that the plasma etching of Be and Si doped AlN needs further optimization. This color change (presumed to be contamination) could not be removed and could introduce a combined increase in the contact resistance of the homojunction AlN diode by 3–4 orders of magnitude for two contacts. The current of the n-type layer of the device N4633 (etched) was ∼2–3 orders of a magnitude lower than the n-type film N4595 (unetched) for the same contact patterns. Also, the current of the p-type layer on the device N4633 (processed but not etched) was lower by ∼<1 order of magnitude as compared to the p-type film N4492 (unprocessed). We attribute this reduction in current through the p-type contacts of the device N4633 to higher annealing temperature than optimal for the p-type contact layer necessary to optimize the worse conducting n-type contact. Future studies will use individual anneals for n- and then p-type contacts but were not pursued here in this initial study due to the complexity of optimal temperatures found based on mask metal coverage (see Sec. III).

FIG. 4.

(a) n-contacts transmission line measurement (NTLM) and (b) p-contacts transmission line measurement (PTLM) of the AlN homojunction PN diode.

FIG. 4.

(a) n-contacts transmission line measurement (NTLM) and (b) p-contacts transmission line measurement (PTLM) of the AlN homojunction PN diode.

Close modal

While each of the CTLM measurements in Figs. 3 and 4 shows Ohmic behavior with linear correlation coefficients exceeding 99.9%, when plotting the results in the standard way described by Klootiwijk and Timmering,31 which uses simplifying assumptions or in the manner of Cohen and Gildenblat,32 which includes the full Bessel function fitting necessary for CTLM analysis, neither result in fits of sufficient quality to reliably extract the sheet resistance or contact resistance. At present, it appears that the CTLM fitting failure results from either a finite thickness limitation not accounted for in the CTLM theory or the rare condition that the contact resistance exceeds the sheet resistance violating the assumptions implicit in the CTLM analysis. This is not surprising as contacts to a 6.1 eV semiconductor are expected to be challenging but does indicate a need for future improvements in contact technology to AlN.

Nevertheless, the forward diode response was nearly ideal except for the high series resistance owing to the aforementioned contact issues with the fabricated device. Figure 5 shows the linear and semilog current density–voltage (JV) characteristics of the N4633 AlN PN diode. The turn-on voltage in the linear and semilog plots is ∼6 V, which is in line with expectations for a 6.1 eV AlN semiconductor. While a clear 6 orders of magnitude of rectification is shown, a low breakdown voltage and very high series resistance are also evident by the high current density tail on the semilog plot and the soft turn on in the linear plot. The current density of this sample can be further improved by 3–4 orders of magnitude by optimizing the fabrication process of the device to match that of the previously processed films. Still, the small reverse to large forward current sweep shows 6 orders of magnitude rectification in this AlN homojunction diode, which may be further improved for achieving even higher performance high-power devices by optimizing the growth9 and fabrication12 conditions.

FIG. 5.

(a) JV and (b) semilog JV characteristics of five equal dimension AlN homojunction PN diodes from N4633.

FIG. 5.

(a) JV and (b) semilog JV characteristics of five equal dimension AlN homojunction PN diodes from N4633.

Close modal

In conclusion, 6000 times higher bulk AlN electron concentrations are achieved as compared to literature. With this successful experimental achievement of both n-type AlN:Si films and p-type AlN:Be films, state of the art AlN homojunction PN diodes are demonstrated. The evidence of 6 orders of magnitude rectification with the proper turn-on voltage of ∼6 V for a 6.1 eV AlN semiconductor offers ultimate confidence that the pioneering doping results shown are in fact real. A new semiconducting AlN era has, thus, emerged with AlN no longer being simply an insulator. This study demonstrates near-future exciting promise for AlN-based DUV optical emitters and detectors, high-power/voltage/temperature, and high-frequency switching devices capable of operation in extreme radiation and heat environments.

This work was supported by the Office of Naval Research (ONR) Multidisciplinary University Research Initiatives (MURI) Program entitled, “Leveraging a New Theoretical Paradigm to Enhance Interfacial Thermal Transport In Wide Bandgap Power Electronics” under Award No. N00014-17-S-F006 administered by Dr. Mark Spector and Mr. Lynn Petersen. This work was also in part supported by the Air Force Office of Scientific Research under Award No. FA9550-21-1-0318 administered by Dr. Ali Sayir.

The authors have no conflicts to disclose.

H.A. and W.A.D. conceived and designed the project and experiments. H.A. performed the growth, characterization and fabrication of the samples and devices, and prepared the manuscript. Z.E. and C.M.M. helped with preparation of the templates and the manuscript. S.L. contributed to annealing of the PN diode.

The data that support the findings of this study are available within the article.

1.
J. Y.
Tsao
,
S.
Chowdhury
,
M. A.
Hollis
,
D.
Jena
,
N. M.
Johnson
,
K. A.
Jones
,
R. J.
Kaplar
,
S.
Rajan
,
C. G.
van de Walle
,
E.
Bellotti
,
C. L.
Chua
,
R.
Collazo
,
M. E.
Coltrin
,
J. A.
Cooper
,
K. R.
Evans
,
S.
Graham
,
T. A.
Grotjohn
,
E. R.
Heller
,
M.
Higashiwaki
,
M. S.
Islam
,
P. W.
Juodawlkis
,
M. A.
Khan
,
A. D.
Koehler
,
J. H.
Leach
,
U. K.
Mishra
,
R. J.
Nemanich
,
R. C. N.
Pilawa-Podgurski
,
J. B.
Shealy
,
Z.
Sitar
,
M. J.
Tadjer
,
A. F.
Witulski
,
M.
Wraback
, and
J. A.
Simmons
,
Adv. Electron. Mater.
4
,
1600501
(
2018
).
2.
M.
Feneberg
,
R. A. R.
Leute
,
B.
Neuschl
, and
K.
Thonke
,
Phys. Rev. B
82
,
075208
(
2010
).
3.
Hexatech Inc. (2021).
4.
C. R.
Miskys
,
J. A.
Garrido
,
C. E.
Nebel
,
M.
Hermann
,
O.
Ambacher
,
M.
Eickhoff
, and
M.
Stutzmann
,
Appl. Phys. Lett.
82
,
290
(
2003
).
5.
H.
Ahmad
,
J.
Lindemuth
,
Z.
Engel
,
C. M.
Matthews
,
T. M.
McCrone
, and
W. A.
Doolittle
,
Adv. Mater.
33
,
2104497
(
2021
).
6.
M. L.
Nakarmi
,
K. H.
Kim
,
K.
Zhu
,
J. Y.
Lin
, and
H. X.
Jiang
,
Appl. Phys. Lett.
84
,
3769
(
2004
).
7.
T.
Ive
,
O.
Brandt
,
H.
Kostial
,
K. J.
Friedland
,
L.
Däweritz
, and
K. H.
Ploog
,
Appl. Phys. Lett.
86
,
024106
(
2005
).
8.
K.
Kishimoto
,
M.
Funato
, and
T.
Kawakami
,
Appl. Phys. Express
13
,
015512
(
2020
).
9.
H.
Ahmad
,
K.
Motoki
,
E. A.
Clinton
,
C. M.
Matthews
,
Z.
Engel
, and
W. A.
Doolittle
,
ACS Appl. Mater. Interfaces
12
,
37693
(
2020
).
10.
B. P.
Gunning
,
C. A. M.
Fabien
,
J. J.
Merola
,
E. A.
Clinton
,
W. A.
Doolittle
,
S.
Wang
,
A. M.
Fischer
, and
F. A.
Ponce
,
J. Appl. Phys.
117
,
045710
(
2015
).
11.
H.
Ahmad
,
T. J.
Anderson
,
J. C.
Gallagher
,
E. A.
Clinton
,
Z.
Engel
,
C. M.
Matthews
, and
W.
Alan Doolittle
,
J. Appl. Phys.
127
,
215703
(
2020
).
12.
H.
Ahmad
,
Z.
Engel
,
M.
Zia
,
A. S.
Weidenbach
,
C. M.
Matthews
,
B.
Zivasatienraj
,
M. S.
Bakir
, and
W. A.
Doolittle
,
Semicond. Sci. Technol.
36
,
125016
(
2021
).
13.
L.
Gordon
,
J. L.
Lyons
,
A.
Janotti
, and
C. G.
van de Walle
,
Phys. Rev. B
89
,
085204
(
2014
).
14.
M.
Hayden Breckenridge
,
Q.
Guo
,
A.
Klump
,
B.
Sarkar
,
Y.
Guan
,
J.
Tweedie
,
R.
Kirste
,
S.
Mita
,
P.
Reddy
,
R.
Collazo
, and
Z.
Sitar
,
Appl. Phys. Lett.
116
,
172103
(
2020
).
15.
M. H.
Breckenridge
,
B.
Pegah
,
Q.
Guo
,
B.
Sarkar
,
D.
Kchachariya
,
S.
Pavlidis
,
J.
Tweedie
,
R.
Kirste
,
S.
Mita
,
P.
Reddy
,
R.
Collazo
, and
Z.
Sitar
,
Appl. Phys. Lett.
118
,
112104
(
2021
).
16.
F.
Mehnke
,
X. T.
Trinh
,
H.
Pingel
,
T.
Wernicke
,
E.
Janzen
,
N. T.
Son
, and
M.
Kneissl
,
J. Appl. Phys.
120
,
145702
(
2016
).
17.
W.
Götz
,
R. S.
Kern
,
C. H.
Chen
,
H.
Liu
,
D. A.
Steigerwald
, and
R. M.
Fletcher
,
Mater. Sci. Eng. B: Solid-State Mater. Adv. Technol.
59, 211 (1999)
18.
I.
Bryan
,
Z.
Bryan
,
S.
Washiyama
,
P.
Reddy
,
B.
Gaddy
,
B.
Sarkar
,
M. H.
Breckenridge
,
Q.
Guo
,
M.
Bobea
,
J.
Tweedie
,
S.
Mita
,
D.
Irving
,
R.
Collazo
, and
Z.
Sitar
,
Appl. Phys. Lett.
112
,
062102
(
2018
).
19.
R.
Zeisel
,
M. W.
Bayerl
,
S. T. B.
Goennenwein
,
R.
Dimitrov
,
O.
Ambacher
,
M. S.
Brandt
, and
M.
Stutzmann
,
Phys. Rev. B
61
,
R16283
(
2000
).
20.
S.
Petit
,
R.
Jones
,
M. J.
Shaw
,
P. R.
Briddon
,
B.
Hourahine
, and
T.
Frauenheim
,
Phys. Rev. B
72
,
073205
(
2005
).
21.
G. A.
Slack
,
J. Phys. Chem. Solids
34
,
321
(
1973
).
22.
C. G.
van de Walle
and
J.
Neugebauer
,
J. Appl. Phys.
95
,
3851
(
2004
).
23.
A. D.
Arulsamy
and
K. K.
Ostrikov
,
Phys. Lett. A
373
,
2267
(
2009
).
24.
B.
Heying
,
E. J.
Tarsa
,
C. R.
Elsass
,
P.
Fini
,
S. P.
Denbaars
, and
J. S.
Speck
,
J. Appl. Phys.
85
,
6470
(
1999
).
25.
W. H.
Chen
,
Z. Y.
Qin
,
X. Y.
Tian
,
X. H.
Zhong
,
Z. H.
Sun
,
B. K.
Li
,
R. S.
Zheng
,
Y.
Guo
, and
H. L.
Wu
,
Molecules
24
,
1562
(
2019
).
26.
S.
Hasan
,
A.
Mamun
,
K.
Hussain
,
D.
Patel
,
M.
Gaevski
,
I.
Ahmad
, and
A.
Khan
,
MRS Adv.
6
,
456
(
2021
).
27.
B. P.
Gunning
,
E. A.
Clinton
,
J. J.
Merola
,
W. A.
Doolittle
, and
R. C.
Bresnahan
,
J. Appl. Phys.
118
,
155302
(
2015
).
28.
E.
Clementi
,
D. L.
Raimondi
, and
W. P.
Reinhardt
,
J. Chem. Phys.
47
,
1300
(
1967
).
29.
L. J.
van der Pauw
,
Semiconductor Devices: Pioneering Papers
174–182 (1991).
30.
NIST, Nanoscale Device Characterization Division (2010).
31.
J. H.
Klootwijk
and
C. E.
Timmering
, in
IEEE 2004 International Conference on Microelecronic Test Structures
(IEEE Xplore,
2004
), Vol. 17, p.
247
.
32.
S. S.
Cohen
and
H. S.
Gildenblat
,
Metal-Semiconductor Contacts and Devices
, 13th ed. (
Academic Press
,
Orlando, FL
,
1986
).