Evaluation of Metal/Indium-Tin-Oxide for Transparent Low-Resistance Contacts to P-Type Gan
Total Page:16
File Type:pdf, Size:1020Kb
Evaluation of metal/indium-tin-oxide for transparent low-resistance contacts to p-type GaN Wenting Hou, Christoph Stark, Shi You, Liang Zhao, Theeradetch Detchprohm, and Christian Wetzel* Future Chips Constellation and Department of Physics, Applied Physics, and Astronomy, Rensselaer Polytechnic Institute, Troy, New York 12180, USA *Corresponding author: [email protected] Received 12 March 2012; revised 27 June 2012; accepted 8 July 2012; posted 16 July 2012 (Doc. ID 164558); published 2 August 2012 In search of a better transparent contact to p-GaN, we analyze various metal/indium-tin-oxide (ITO) (Ag/ITO, AgCu/ITO, Ni/ITO, and NiZn/ITO) contact schemes and compare to Ni/Au, NiZn/Ag, and ITO. The metal layer boosts conductivity while the ITO thickness can be adjusted to constructive trans- mission interference on GaN that exceeds extraction from bare GaN. We find a best compromise for an Ag/ITO (3 nm ∕ 67 nm) ohmic contact with a relative transmittance of 97% of the bare GaN near 530 nm and a specific contact resistance of 0.03 Ω ·cm2. The contact proves suitable for green light-emitting diodes in epi-up geometry. © 2012 Optical Society of America OCIS codes: 230.3670, 310.7005. 1. Introduction While contact resistance and current spreading Our society’s drive for energy efficiency places high are known to improve with increasing thickness of relevance on the identification of low-resistance the metal stack, the transmittance decreases. For ohmic contacts to wide bandgap group-III nitrides p-type GaN, semitransparent Ni/Au contacts are most commonly used. Reported specific contact for use in light-emitting diodes [1,2] (LEDs) and −3 third-generation solar cells [3]. In particular, trans- resistance values mostly range from 4.4 × 10 to p 5 × 10−4 Ω ·cm2 [9,10], with the exception of only parent ohmic contacts to -type GaN are of topical −6 2 concern. Throughout the group-III nitrides, hole con- one report of a value as low as 4 × 10 Ω ·cm [11]. duction is the more limited one compared to electron A spectral transmittance around 70% to 80% in the transport [4–6]. The high binding energy of the com- visible spectrum has been reported [9,10]. ITO mon Mg acceptor [7], its propensity for structural layers, on the other hand, have shown very high defect generation, and low hole mobility require transparency with transmittance above 90% in the large-area hole injection close to the optically active visible spectrum, but non-ohmic contact behavior p [12], with specific contact resistance in the range region. This can best be achieved by -contact −1 2 schemes that provide both low contact resistance of 10 Ω ·cm [13]. In an attempt to improve both and high optical transparency in a specified spectral contact resistance and transparency, we here jointly region. Furthermore, the large binding energy of study both in several metal/ITO contacts schemes— — p electrons at the valence band maximum of GaN namely Ag/ITO, AgCu/ITO, Ni/ITO, NiZn/ITO to - limits suitable p-contact metals to those with the GaN and compare them to standard Ni/Au, NiZn/Ag highest work function [8]. and ITO contacts. 2. Experiments 1559-128X/12/235596-05$15.00/0 (0001) Ga-face p-type GaN layers, 200 nm thick, © 2012 Optical Society of America were grown by metal-organic vapor epitaxy on 5596 APPLIED OPTICS / Vol. 51, No. 23 / 10 August 2012 unintentionally doped GaN (u-GaN) templates, 4 μm 1.0 Ni/Au (5 nm/5 nm) thick, on c-plane sapphire. Free hole concentrations 17 −3 NiZn/Ag (5 nm/200 nm) of 4 × 10 cm at room temperature were achieved 0.5 ITO (200 nm) 19 −3 using Mg doping at a concentration of 10 cm .On Ag/ITO (3 nm/67 nm) top, a p-GaN, 10 nm thick, Mg-doped (up to 1020 cm−3) contact layer was grown. As substrates for 0.0 transmittance measurements, float glass slides, n Current (mA) 0.2 mm thick, and -GaN on double side polished -0.5 (DSP) sapphire were used. Contact metal stacks were prepared by means of an e-beam evaporation and subsequent rapid thermal annealing on both -1.0 -2 -1 0 1 2 templates using each of the following schemes (nota- Voltage (V) tion in sequence of deposition): Ni/Au (5 nm ∕ 5 nm) Fig. 1. (Color online) Current-voltage characteristics of different with a 1 min at 500 °C anneal in oxygen ambient; contact schemes on p-GaN. NiZn/Ag (5 nm ∕ 200 nm) with a 1 min at 550 °C an- neal in oxygen ambient; ITO (200 nm, 100 nm, and 67 nm), each with a 1 min at 550 °C anneal in oxygen The current-voltage characteristics between a pair ambient; Ag/ITO (3 nm ∕ 130 nm, 3 nm ∕ 67 nm), of the respective contact schemes to p-GaN are AgCu/ITO (3 nm ∕ 130 nm), Ni/ITO (3 nm ∕ 130 nm, shown in Fig. 1 (5 μm contact spacing). The relative 1 nm ∕ 60 nm, 3 nm ∕ 100 nm, and 3 nm ∕ 67 nm), and optical transmittance of the same contact layers on NiZn/ITO (3 nm ∕ 130 nm), each with a 1 min at 550 °C glass is shown in Fig. 2 and Fig. 3. anneal in oxygen ambient; Ni/ITO (3 nm ∕ 130 nm) We find that after annealing, the Ni/Au contacts with a 1 min at 450 °C anneal in oxygen ambient. (Fig. 1, black solid squares) readily show ohmic −3 2 We define relative transmittance of the layer stack behavior with ρc 2.2 × 10 Ω ·cm . In our experi- on a substrate by the transmitted power in ratio with ence, however, this value can vary strongly with min- the power transmitted through a bare substrate. We or details of the deposition and annealing process, determine the ratio experimentally by simultaneous source metal purity, and surface roughness of the measurement of both samples in parallel beams of a p-GaN. For example, minor C contamination of the spectrophotometer. To properly account for multiple reflection effects in transmittance, we tested the Ni/ITO 3 nm/130 nm metal stacks on both kinds of substrates, the common 0 200 nm ITO glass slide and GaN on DSP sapphire. 100 annealed 550 C 1 min 80 Ni/ITO 1 nm/60 nm 3. Results 60 Ni/ITO 3 nm/130 nm 0 The specific contact resistance (ρc) of lateral contacts Ni/Au after anneal annealed 450 C 1min is generally measured using the transmission line 40 method [14]. In this model, the ohmic contact is de- Ni/Au before anneal posited on a thin film layer, whose total resistance is 20 much lower than that of the supporting substrate un- (%) Transmittance Relative on glass derneath [14]. The resistance values of contact pairs 0 R 300 400 500 600 700 800 with different pad spacing pp are analyzed as a Wavelength (nm) function of spacing, and ρc and the sheet resistance can be extrapolated from the data. Typically, these Fig. 2. (Color online) Transmittance of contact films on 0.2 mm glass slides for various film stacks. prerequisites are not met when analyzing p-contacts to full LED structures. Due to the limitations of hole transport in GaN, the high resistance of the p-type n 100 GaN layer makes shunt currents through the -type ITO (100 nm) layers underneath very likely and so distorts the in- Ni/ITO (3 nm/100 nm) terpreted value of ρc. Under low enough applied test 90 voltage, however, the pn-junction between the paral- lel layers does not turn on and blocks parallel conduc- 80 tion through the n-side of the device [15,16]. We therefore limit our analysis of ρc to the 1 V range. ITO (67 nm) Furthermore, inaccuracies in the layer and contact 70 Ag/ITO (3 nm/67 nm) geometry have a strong effect on values derived from Ni/ITO (3 nm/67 nm) on glass p (%) Transmittance Relative contacts on poorly conducting layers, such as -type 60 GaN [17]. We therefore paid particular attention to 300 400 500 600 700 800 accurately account for the actual contact geometry Wavelength (nm) within 5%. This results in an error of ρc less than Fig. 3. (Color online) Transmittance of contact films on 0.2 mm 10% for our results. glass slides for ITO film with and without a thin Ni or Ag layer. 10 August 2012 / Vol. 51, No. 23 / APPLIED OPTICS 5597 Ni source from the graphite crucible has shown to in- line) again is only less than 5% smaller than for bare crease ρc by more than an order of magnitude. In ITO 100 nm (black solid line) and 67 nm (blue, short addition, we find that Ni/Au contacts on p-GaN sur- dashed line) in the visible spectrum. Apparently, the faces, that have been roughened to increase light additional metal layer of Ni or Ag can indeed signif- extraction, show higher ρc and tend to induce current icantly lower ρc below that of the bare ITO contact crowding. For such conventional contacts, a spectral while maintaining a rather high transmittance. transmittance around 70% to 80% in the visible spec- The specific contact resistance of an Ni/ITO trum has been reported [9,10]. We here find a rela- (3 nm ∕ 130 nm) film decreases when lowering the tive transmittance of 40% before annealing (Fig. 2, annealing temperature from 550 °C to 450 °C. The blue open circles) and 75% after annealing (Fig. 2, relative transmittance of an Ni/ITO (3 nm ∕ 130 nm) magenta solid circles) at a wavelength of 500 nm.