Wednesday, November 18, 2026
CS Fusion The compound semiconductor reference
GaN

Why p-Type GaN Is Hard: Magnesium Doping and Activation

p-type GaN is gallium nitride doped with magnesium so that holes, not electrons, carry most of the current. It is hard to make because the Mg acceptor sits roughly 150 to 200 meV above the valence band, so only about 1 % of acceptors ionise at room temperature, and hydrogen from growth passivates Mg until it is removed by annealing.

Key takeaways

  • Magnesium is the only acceptor used in production p-GaN. Its ionization energy is reported at roughly 150 to 200 meV, several times the thermal energy at 300 K (about 26 meV).
  • To reach hole densities of about 1017 to 1018 cm−3, growers add about 1019 to 1020 cm−3 of Mg.
  • Hole mobility is low, typically about 5 to 20 cm²/V·s, so p-GaN resistivity is of the order of 1 Ω·cm or higher.
  • MOCVD-grown Mg-doped GaN is highly resistive as grown, because hydrogen binds to Mg. Amano, Akasaki and co-workers activated it with low-energy electron-beam irradiation (LEEBI) in 1989. Nakamura and co-workers showed in 1992 that thermal annealing in nitrogen works and scales.
  • Too much Mg backfires: above a few × 1019 cm−3, self-compensation and structural defects reduce the hole density.

How is GaN doped p-type?

Doping adds impurity atoms that donate electrons (donors) or accept them (acceptors). A p-type semiconductor has acceptors, which take electrons from the valence band and leave mobile holes behind. For the general picture, see doping and charge carriers, which also covers donor and acceptor levels.

In GaN, an acceptor must replace gallium with a group-II atom. Magnesium on a gallium site (MgGa) is the practical choice. In MOCVD it is supplied as bis(cyclopentadienyl)magnesium, Cp2Mg. Zinc, beryllium and carbon have been studied. Zinc is deeper. Beryllium is predicted to be shallower but tends to self-compensate by also sitting on interstitial sites, where it acts as a donor. Carbon is used in GaN buffers to make them insulating, not p-type.

n-type GaN, by contrast, is easy. Silicon donors sit only about 15 to 30 meV below the conduction band and are almost fully ionised at room temperature. Unintentionally grown GaN is usually n-type already. That imbalance shaped the history of GaN devices.

What is the Mg acceptor ionization energy?

The ionization energy of an acceptor is the energy needed to move an electron from the valence band into the acceptor level, creating a free hole. For Mg in GaN, reported values fall roughly in the 150 to 200 meV range. The exact figure depends on the method (Hall effect versus optical), the Mg concentration and compensation. At high Mg concentrations, the apparent activation energy falls because acceptor states overlap and form a band.

Why that matters: the fraction of acceptors ionised depends on exp(−EA/kT). With kT ≈ 26 meV at 300 K and EA near 170 meV, only a small fraction, often quoted near 1 %, of Mg atoms contribute a hole. The rest sit neutral. That is why p-GaN needs Mg concentrations two orders of magnitude above the hole density it delivers.

Dopant and hostTypeApproximate ionization energyTypical carrier density at 300 KTypical mobility (cm²/V·s)
Si in GaNDonorAbout 15 to 30 meV1017 to 1019 cm−3 electronsAbout 200 to 1,000, falling with doping
Mg in GaNAcceptorAbout 150 to 200 meV1017 to about 1018 cm−3 holesAbout 5 to 20
Mg in AlGaNAcceptorRises with Al content to several hundred meVFalls steeply with AlLow
B in Si (reference)AcceptorAbout 45 meVNear full ionizationUp to about 450
Source: Ioffe Institute NSM archive; S. M. Sze and K. K. Ng, Physics of Semiconductor Devices; S. Nakamura and co-workers, Japanese Journal of Applied Physics (1992). Values vary with concentration, compensation and measurement method, so ranges are shown.

Why does Mg need activation?

Mg-doped GaN grown by MOCVD comes out of the reactor highly resistive, even with plenty of Mg in the crystal. The reason is hydrogen passivation. MOCVD uses ammonia (NH3) as the nitrogen source and often hydrogen as the carrier gas. Atomic hydrogen diffuses into the film and binds to Mg, forming electrically neutral Mg–H complexes. A passivated acceptor cannot create a hole.

Activation is any step that breaks those complexes and drives the hydrogen out. Two methods made history.

  1. LEEBI. In 1989, Hiroshi Amano, Isamu Akasaki and co-workers at Nagoya University reported that low-energy electron-beam irradiation turned Mg-doped GaN into conductive p-type material. They used it to make an early GaN p-n junction LED. LEEBI works only near the surface the beam reaches, and it is slow, so it does not suit wafer-scale production.
  2. Thermal annealing. In 1992, Shuji Nakamura and co-workers at Nichia showed that annealing Mg-doped GaN in nitrogen at about 700 °C and above produced low-resistivity p-type GaN across the whole wafer. They also showed that annealing in ammonia made the film resistive again. That reversibility identified atomic hydrogen from NH3 as the cause of passivation.

Thermal activation in N2 or N2/O2 became the industry standard. Modern recipes typically anneal in the 600 to 900 °C range for minutes, often combined with other steps. MBE-grown p-GaN, which uses no ammonia in plasma-assisted growth, is usually active as grown.

Who solved p-GaN?

Credit is shared. Amano and Akasaki first demonstrated p-type conduction with LEEBI. Nakamura found the thermal route that made it manufacturable and explained hydrogen passivation. Nakamura went on to demonstrate bright InGaN double-heterostructure blue LEDs. In 2014, Akasaki, Amano and Nakamura shared the Nobel Prize in Physics for the invention of efficient blue light-emitting diodes. Both p-type doping and low-temperature buffer layers, also developed by these groups, are cited in that work.

Why more magnesium does not always help

The obvious fix for a deep acceptor is to add more of it. In GaN that works only up to a point. Above a Mg concentration of a few × 1019 cm−3, the hole density typically stops rising and then falls. Proposed reasons include:

  • Self-compensation. Native donors, such as nitrogen vacancies and Mg-related complexes, form more readily as the Fermi level moves toward the valence band and cancel some acceptors.
  • Structural defects. Heavy Mg doping produces pyramidal inversion domains, regions where the crystal polarity flips, and other extended defects that trap or neutralise Mg.
  • Surface roughening and memory effects. Mg tends to accumulate on surfaces and in reactors, which smears doping profiles and complicates sharp p-n junctions.

The upshot is a practical ceiling near 1018 cm−3 holes in GaN at room temperature. Defects and how they interact with dopants are covered further in our piece on defects in GaN.

What poor p-GaN means for devices

  • LEDs. The p-side spreads current poorly, so LEDs use transparent conductive layers such as ITO and careful contact design. Poor hole injection also contributes to electron overflow at high current.
  • Ohmic contacts. GaN’s valence band sits deep, so few metals form low-resistance contacts. Oxidised Ni/Au and Pd-based schemes are common.
  • Ultraviolet emitters. In AlGaN, Mg goes deeper still, which is a main reason deep-UV LEDs have low efficiency. See our article on nitride ultraviolet LEDs.
  • Power transistors. Normally-off GaN HEMTs often use a p-GaN layer under the gate to lift the 2DEG out of conduction at zero bias. Activation quality affects threshold voltage stability. The polarization physics behind it is in our GaN polarization guide.

Workarounds exist. Polarization doping grades AlGaN composition so that bound polarization charge draws in holes without relying on thermal ionisation, as Simon and co-workers reported in Science in 2010. Mg-doped AlGaN/GaN superlattices use band bending to raise the average hole density. Tunnel junctions replace the top p-contact with a second n-layer. For context on GaN’s other properties, see our gallium nitride hub.

FAQ

Why is n-type GaN easy but p-type hard?

Silicon donors are shallow and nearly fully ionised, and native defects lean n-type. The Mg acceptor is deep, hydrogen passivates it, and compensation sets in at high doping.

What temperature is used to activate Mg in GaN?

Nakamura’s 1992 work used nitrogen anneals at about 700 °C and above. Production recipes commonly run in the 600 to 900 °C range, tuned to the layer stack.

Is p-GaN needed in GaN power transistors?

Not for depletion-mode HEMTs. Many enhancement-mode (normally-off) HEMTs use a p-GaN gate layer to shift the threshold voltage positive.

Can hydrogen re-passivate activated p-GaN?

Yes. Exposure to atomic hydrogen, for example from ammonia at high temperature or some plasma steps, can re-form Mg–H complexes and raise resistivity again.

Sources

  • H. Amano, M. Kito, K. Hiramatsu and I. Akasaki, work on p-type conduction in Mg-doped GaN treated with low-energy electron-beam irradiation, Japanese Journal of Applied Physics (1989).
  • S. Nakamura and co-workers, work on thermal annealing of Mg-doped GaN and the role of hydrogen passivation, Japanese Journal of Applied Physics (1992).
  • J. Simon et al., “Polarization-induced hole doping in wide-band-gap uniaxial semiconductor heterostructures,” Science (2010).
  • S. M. Sze and K. K. Ng, Physics of Semiconductor Devices, 3rd ed., Wiley.
  • Ioffe Institute, New Semiconductor Materials (NSM) archive, GaN entry.
Malcare WordPress Security