Impact of nitrogen annealing on the electrical properties of HgCdTe epitaxial films

The nitrogen annealing of HgCdTe materials grown by molecular beam epitaxy (MBE) was carried out to manipulate their electrical properties. The results show that the annealing temperature, annealing time and cooling process all have significant influences on the electrical properties of HgCdTe materials. Excessive annealing temperature or long annealing time can make voids emerge on the surface of the CdTe passivation layer. Carrier concentration and mobility vary exponentially with annealing time and they reach an equilibrium value determined by annealing temperature over a long annealing duration. Moreover, time constants are given and a longer time is needed for mobility to reach an equilibrium value than carrier concentration. The relationship between equilibrium carrier concentration and annealing temperature is given and the activation energy under nitrogen annealing is calculated as 0.63 eV. For a long cooling duration, Hg vacancies are annihilated by Hg atoms diffusion, which makes carrier concentration lower and mobility higher. In addition, some outlier data were found in this experiment and explained by the combination between Te antisites and Hg vacancies.


Introduction
With the rapid development in the field of infrared detection, HgCdTe, as one of the most important materials, has received extensive attention [1,2]. HgCdTe is the narrow bandgap semiconductor material with a direct band gap and its forbidden bandwidth is adjustable from 0 to 1.6 eV, covering from short-wave infrared to very long-wave infrared regions [3]. It also has the advantages of high quantum efficiency and absorption coefficient [4], which determine the irreplaceable position of HgCdTe material in the field of infrared detectors. HgCdTe photovoltaic detectors use the photovoltaic effect to convert light signals into electrical signals, having the advantages of high target resolution, fast response time, low power consumption and good reliability [5]. The performance of HgCdTe detectors is mainly limited by the dark current of the p-n junction [6,7], which has an impact on the noise level, quantum efficiency and the figure of merit R 0 A [8,9].
The dark current level in the infrared detectors is related to the carrier concentration in the HgCdTe material, and devices with low dark current levels can be obtained by adjusting the carrier concentration through the annealing process. The annealing process is a common post-growth treatment for HgCdTe material. The concentration distribution of defects and dopant elements in HgCdTe can be adjusted through heat treatment [10], thus changing the electrical properties of materials. For infrared detectors in the short and medium wavelength bands, HgCdTe intrinsically doped with Hg vacancies is used as the absorber layer. Adjusting the concentration of Hg vacancies in HgCdTe material through the annealing process can directly affect the carrier concentration of the material and improve the electrical properties of the material, and combined with the mature n-on-p process, infrared detectors can be fabricated to meet the demand. For the fabrication of high-performance infrared detectors, extrinsic p-type doping, such as Cu or Au doping, is commonly used. Cu doping requires a ∼10 16 cm −3 concentration of Hg vacancies to maintain a stable distribution of Cu elements in HgCdTe [11]. Concerning Au doping, the Au element concentration distribution has a strong positive correlation with Hg vacancy concentration and dislocation density [12]. The annealing process can adjust the Hg vacancy concentration and then change the distribution of Au doping concentration. The distributions of Hg vacancies and dopant elements are directly reflected in the carrier concentration of the materials. In this way, it is significant to study the effect of the annealing process on the electrical properties of HgCdTe materials.
Annealing at Hg/Te saturated limits or vacuum conditions is widely used. However, Hg and Te are toxic, and sealing the vacuum annealing tube expends much time. Nitrogen is inert and reacts with virtually nothing under the conditions used for annealing process. Though nitrogen annealing can be cheaper and feasible, there are few reports about it being used for HgCdTe.
In this study, the effect of nitrogen annealing on the electrical properties of HgCdTe was studied by controlling different annealing conditions. The relationships of the carrier concentration and mobility with annealing time are given as well as equilibrium values and characteristic time constants. By fitting the equation of equilibrium concentration varying with annealing temperature, the activation energy of nitrogen annealing is calculated, and the comparison of carrier concentration under different annealing conditions has also been given. Carrier concentration and mobility versus cooling time are shown. In addition, some outlier data were found and explained by the combination between Te antisites and Hg vacancies.

Experiment
The HgCdTe samples used in this study were epitaxially grown on GaAs/Si substrates by molecular beam epitaxy (MBE), with an epitaxial layer of 4 ∼ 6 μm and an average cadmium composition of 0.29. Then all samples were cleaned with a standard bromine/methanol solution for etching and coated with 300 nm CdTe grown by electron beam evaporation. Heat treatment of n-type annealing was carried out to make the carrier concentration reach about 6 × 10 15 cm −3 .
The chamber was evacuated before annealing and then filled with nitrogen. To prevent the influence of the very small amount of air that may stick to the inner wall of the furnace chamber [13], a flow rate of 0.1 slm of nitrogen was continuously injected into the chamber during the heating and constant temperature process so that the residual air could be taken away by the nitrogen gas flow without affecting the temperature of the HgCdTe samples. The chamber was cooled down by using a cooled stream of nitrogen. Heat treatments were carried out at temperatures ranging from 220 to 340°C with annealing times ranging from 2 to 72 h. To investigate the effect of cooling time on the electrical properties of HgCdTe, some samples were gradually cooled at cooling times in the range from 5 to 180 min. The infrared spectrums were measured by Fourier transform infrared spectroscopy (FTIR) at room temperature to obtain the components and thicknesses of the HgCdTe epitaxial layers. After annealing, the surface morphology of the samples was observed by microscopy and uniform carrier concentration and mobility were obtained by Hall measurement at 77 K.

Results and discussion
During the annealing process, interdiffusion at the CdTe/HgCdTe interface occurs and macroscopic voids may appear while grain boundaries in the CdTe passivation layer migrate and fuse [14]. In addition, excessive annealing temperature resulting in Hg atoms escaping from the surface and being carried away by nitrogen can also produce voids. Dark current in the HgCdTe detectors would be increased due to the presence of these voids which leads to the formation of deep energy levels. Therefore, it is necessary to investigate the surface morphology of the CdTe passivation layer after annealing process. Figure 1 shows images of the surface morphology under different annealing conditions. Smooth surface morphology could be maintained in appropriate annealing conditions as shown in figures 1(a) and (b). When the annealing time is long enough to make grain boundaries in CdTe fuse together, voids appear on the surface as shown in figure 1(c), as well as high annealing temperature makes grain boundaries migrate quicker to form them as shown in figure 1(d). We also found that the thicker samples were, the longer the annealing time needed for voids to appear, which is due to more time taken to establish a dynamic equilibrium of interdiffusion [15].
The carrier concentration versus annealing time at different annealing temperatures is shown in figure 2. The curve shapes show that no inversion layer exists for the annealing conditions used in this experiment. The variation of the carrier concentration C can be fitted to the equation [16] C C where C 0 is the initial carrier concentration, C eq is the equilibrium carrier concentration, t is the value of annealing time and τ is the characteristic time constant. In addition, C eq and τ are functions of annealing temperature. For the annealing process of 250°C, the carrier concentration increases smoothly to the equilibrium concentration and the characteristic time constant is about 148 s. For the cases of 280°C and 300°C, equilibrium concentrations are 2.46 × 10 16 cm −3 and 3.18 × 10 16 cm −3 with characteristic time constants of 216 s and 285 s respectively. When samples are annealed at 320°C, the equilibrium concentration increases to 4.12 × 10 16 cm −3 . For all cases, the initial carrier concentration C 0 is about −8.33 × 10 15 cm −3 , which is similar to the carrier concentration of samples after n-type annealing. Figure 3 shows mobility versus annealing time at different annealing temperatures. The mobility decreases with increasing annealing time until reaching an equilibrium value. It seems that the variation of mobility is related to that of the carrier concentration as the decrease in mobility is accompanied by an increase in carrier concentration. The carrier concentration is determined by the Hg vacancies. When the annealing time becomes long, the concentration of Hg vacancy increases, which results in weaker ionizing impurity scattering and makes the mobility decrease, and the equation of mobility has a similar form to the equation (1). Compared with figure 3, it can be observed that mobility takes a longer time to reach an equilibrium value than carrier concentration. For annealing process of 250°C, the time constant of mobility is 255 s, which is 2.3 times larger than that of the carrier concentration. As the temperature increases, their time constants become similar to each other. It is suggested that other defects exist in the material besides Hg vacancies, which are related to the other constituents of HgCdTe [15]. These defects have an impact on the carrier scattering process and they alter during the thermal treatments with a slow generation rate. With the temperature becoming higher, their generation becomes quicker, which makes the time constant of mobility approximate the carrier concentration.
The carrier concentration will change with annealing time due to the generation and recombination of interstitials (Hg i ) and vacancies (V Hg ), which is known as Frenkel generation/recombination. Assume that Hg vacancy is a doubly charged acceptor. where Hg Hg is the Hg atom at substitutional sites, g is the rate at which interstitials and vacancies are generated and k iv is the rate at which interstitials annihilate vacancies. To simulate this process, first-order reaction kinetics is applied to describe the behavior of point defects and only the basic Hg interstitials and cation vacancies are considered. Continuity equations for each species can be given as  Hg t D Hg Hg v Hg iv i Hg where D i is the interstitial diffusion coefficient and D v is the diffusion coefficient for vacancies. The first terms on the right side of equations (3) and (4) where x signifies the mole fraction of CdTe in the Hg 1-x Cd x Te and in this experiment is 0.29. Figure 4(b) compares the equilibrium carrier concentration under different annealing conditions. Annealing at Hg saturated limit is always used to manufacture n-type HgCdTe and the equilibrium value achieved for annealing at Te saturated limit is higher than nitrogen annealing at the same temperature. The activation energy of Te saturated annealing is 0.65 eV, which is approximate to nitrogen annealing. Considering that Te can be toxic and need more costs, it is feasible for nitrogen annealing to adjust carrier concentration. This may improve the process line of HgCdTe devices and thus increase safety and reduce costs. Figure 5 shows the carrier concentration and mobility versus cooling time after annealing at 280°C and cooling time is what samples spend to be cooled down from annealing temperature to about 100°C. As the cooling time becomes longer, the carrier concentration decreases rapidly until the equilibrium concentration is reached, which can be observed as 2.18 × 10 16 cm −3 . The variation in mobility is opposite with the equilibrium value of 392 cm 2 V −1 s −1 . In the cooling process, carrier concentration is determined by the difference in concentration between the Hg vacancies at the annealing temperature and the diffused Hg atoms in the layer. During the cooling duration, Hg vacancies are filled in the HgCdTe epitaxial layer with Hg atoms which have a large diffusion coefficient [20,21], leading to a decrease in carrier concentration. It can be observed that a long cooling duration tends to obtain HgCdTe materials with low carrier concentration and high mobility.
During the experiment, we found some outlier data with long annealing time as shown in figure 6. It is indicated that carrier concentrations are quite similar to the equilibrium carrier concentration of 280°C when the annealing time is less than 48 h. However, it decays below the equilibrium carrier concentration when annealing time reaches 72 h. This situation is the same with the annealing temperature of 300°C as the carrier concentration decreases from an equilibrium carrier concentration of about 3.18 × 10 16 cm −3 to a lower value. It is obvious that high annealing temperature reduces the time for outlier data to appear while it becomes anomalous with annealing time reaching 36 h for 300°C, but for 250°C, no abnormalities are found after 72 h heat treatment.  The carrier concentration is mainly generated by Hg vacancies, however, there are other defects in HgCdTe besides Hg vacancies. Theoretical calculations show that the Te antisites, which means Te on cation sites, are in relatively large concentrations due to the fact that Te atoms are more energetically likely to form Te antisites than Te interstitials [17]. In addition, antisites may exist predominantly in the form of neutral antisite-vacancy pairs in the presence of vacancies. Te antisites can be a source for Te precipitates detected in material quenched following high-temperature HgCdTe substrate growth and annealing [20]. Figure 7 shows the combination between Te antisites and Hg vacancies to form Te precipitates and voids. For high temperature or long annealing time, Te antisites would combine with Hg vacancies, which makes the carrier concentration lower than the equilibrium carrier concentration, and the precipitates nucleate at dislocations and grain boundaries as well as voids formed. Therefore, the carrier concentration decreases and voids appear on the surface morphology besides the reason for grain boundaries fusing.

Conclusion
In this paper, the impact of annealing on electrical properties was studied. The annealing process can make voids emerge on the surface of the CdTe passivation layer by using excessive annealing temperature or long annealing time. The results obtained by Hall measurement show that the annealing temperature, annealing time and cooling process all have an impact on the electrical properties of HgCdTe materials. The variation of carrier concentration and mobility can be fitted to exponential equations. For the long annealing time, the carrier concentration and mobility tend to reach an equilibrium value that is only related to the annealing temperature as well as the characteristic time constant. However, mobility needs a longer time to reach equilibrium value than carrier concentration. The equilibrium carrier concentration shows exponential changes with annealing temperature and the activation energy under annealing of nitrogen atmosphere condition is calculated as 0.63 eV, which is similar to Te saturated limit. During the cooling process, Hg vacancies are filled with diffused Hg atoms in the HgCdTe epitaxial layer. Long cooling duration tends to obtain HgCdTe materials with low carrier concentration and high mobility. Some outlier data are found due to high annealing temperature or long annealing time, which is related to the combination between Te antisites and Hg vacancies. Nitrogen annealing exerts an influence on the electrical properties of HgCdTe and has the advantages of low price, non-toxicity, and low requirement for experimental equipment, which can improve the manufacturing process of infrared detectors.

Acknowledgments
The work was supported by the Key laboratory of Infrared Imaging Materials and Devices of Chinese Academy of Sciences (No. IIMDKFJJ-180).

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).