DLTS Study of Defects in HgCdTe Heterostructure Photodiode

Abstract

Deep-level transient spectroscopy (DLTS) measurements were performed on HgCdTe heterostructure photodiode grown by metal-organic chemical vapor deposition (MOCVD) on GaAs substrate. In order to extract defects from individual layers of the heterostructure, three consecutive etchings were performed. In the first experiment, the N⁺/T/p/T/P⁺/n⁺ structure was chemically etched to the N⁺ bottom contact to obtain a mesa-type detector. Six localized defects were extracted across the entire photodiode. In the second experiment, the bottom contact was made to the p-type absorber. Two localized defects were found in the p/T/P⁺/n⁺ structure. In the third experiment, the top layers were removed and N⁺/T/p type detector was made—the cap contact was made to the p-type absorber, the bottom to the N⁺ layer. Five defect levels were identified, three of which overlap with the first experiment. A deep-trap level located at 183 meV above the top of the valence band was identified within the absorber bandgap. This energy coincides with the activation energy determined from the Arrhenius plot for the dark currents. A defect at the level of ∼0.5E_g suggests that the dark current at low reverse bias voltage is dominated by the Shockley–Read–Hall (SRH) mechanism.

Introduction

The growth of a high-quality Hg_1−xCd_xTe epitaxial layer for applications in infrared (IR) detectors requires the knowledge and control of many experimental factors causing material defects. Among those influencing the morphology of the layer, one can mention the type of substrate, the orientation of the substrate, the temperature of the substrate during growth, and specific properties resulting from the deposition technique. The high density of internal defects in Hg_1−xCd_xTe epitaxial alloys is considered to be one of the most serious issues in the use of this material for large-format and color infrared matrices.^1,2 There are many structural defects in HgCdTe, including point defects related to metal vacancies, misfit dislocations,^2,3,4 threading dislocation,³ stacking faults and twins,^3,5 surface crater defects^6,7 and pyramidal hillocks.^1,8 The presence of defects in semiconductors causes disturbances in their band structure, as a result of which additional levels are created in the band gap and energy bands creating trapping centers for carriers. Their presence reduces the lifetime and mobility of carriers, due to the Shockley–Read–Hall (SRH) generation-recombination mechanism, which determines the usefulness of the material for optoelectronic device fabrication. Deep level transient spectroscopy (DLTS) is one of the methods used to characterize electrically active defects in semiconductors.⁹ It is widely used to characterize deep traps in silicon,¹⁰ SiGe alloys,¹¹ II-VI compounds^12,13 and III-V compounds.¹⁴ It is able to distinguish between majority and minority-carrier traps.⁹ It allows us to determine the activation energy of a deep level, its capture cross section, concentration and can distinguish between traps and recombination centers.

DLTS uses a semiconductor structure in which there is a depletion region, typically a Schottky diode and p–n junction. This work, in contrast to the papers showing the results obtained for single junctions,^15,16 presents DLTS results for the HgCdTe heterostructure photodiode grown by metal-organic chemical vapor deposition (MOCVD) on GaAs substrate. Since the complex heterostructure of the analyzed photodiode contains several depletion regions, the experiment was divided into three stages, described later in the paper.

Measurement Technique

The DLTS signal is obtained from the capacitance transient which originates from the filling and emptying of the defect levels by applying electrical pulses to the sample. Under ideal conditions, the capacitance transient C(t) of the deep level is a simple exponential:⁹

$$ \Delta C\left( t \right) = \Delta C_{0} e^{{\frac{ - t}{\tau }}} $$

(1)

with $\Delta C\left( t \right) =$ where $C\left( t \right) - C_{\infty },$ $C_{\infty }$ is the steady state capacitance $\left( {t = \infty } \right)$, $t$ is the time and $\tau$ is referred to as the time constant. The emission rate is the inverse of the time constant, $e_{n} = 1/\tau$.

The capacitance of the junction is measured at two points of time, $t_{1}$ and $t_{2}$, and at different temperatures after applying a voltage pulse $V_{{\text{p}}}$. The difference between the times $t_{1}$ and $t_{2}$ is called the rate window. In the DLTS method used double “box-car” was used to define the rate window.

$$ C\left( {t_{1} } \right) - C\left( {t_{2} } \right) = \Delta C_{0} \left[ {\exp \left( {\frac{{ - t_{1} }}{\tau }} \right) - \exp \left( {\frac{{ - t_{2} }}{\tau }} \right)} \right] $$

(2)

where $\Delta C_{0}$ is the change in capacitance.

To obtain the energy and apparent cross section, the Arrhenius plot is based on the equation for the emission rate:

$$ \ln (T^{2} /e_{n} ) = \left( {\ln \frac{{T^{2} }}{{\sigma_{n} \upsilon_{{{\text{th}}}} \frac{{g_{0} }}{{g_{1} }}N_{c} }}} \right) + \frac{{E_{A} }}{kT} $$

(3)

where $\frac{{g_{0} }}{{g_{1} }}$ is degeneracy ratio ($g_{0}$—when traps are empty, $g_{1}$—when traps are occupied), and $E_{A}$ is the energy barrier for electron emission, which for defects with non-thermally activated capture cross section and an entropy factor equal to zero could be linked to energy level $E_{t}$ of defects with: $E_{A} = E_{C} - E_{t}$, $k$ is the Boltzmann constant, $T$ is the absolute temperature, $N_{c}$ is the density of states in the conduction band, $v_{{{\text{th}}}}$ is the thermal velocity, and $\sigma$ is the capture cross section. We used a complete hardware and software system by Semetrol. The sample was mounted in a helium closed-cycle cryostat allowing the temperature to be maintained in the range of 50 to 300 K.

Device Design and Fabrication

The Hg_1−xCd_xTe detector was fabricated with an N⁺/T/p/T/P⁺/n⁺ configuration by metal-organic chemical vapor deposition (MOCVD). The HgCdTe epilayer was grown on 2-inch, epi-ready, semi-insulating (100) GaAs substrate after a 3-μm-thick CdTe buffer layer.^17,18 The N⁺/T/p/T/P⁺/n⁺ structure consists of four main layers: a heavily In-doped N⁺ contact layer with a composition of x = 0.523 and a thickness of ∼10 μm. Then we have a p-type As-doped absorber with a composition of x = 0.370 and a thickness of ∼3.3 μm. Next, there is a wide-band gap, heavily As-doped, 2-μm-thick P⁺ layer with a composition significantly larger than that of the absorber (x = 0.536). The 1-µm-thick n⁺ layer with a composition lower than that of the absorber (x = 0.171) and donor concentration of about 10¹⁸ cm⁻³ is on the top of the entire structure. The parameters of the transient T layers are conditioned mainly with the interdiffusion processes during the MOCVD growth.

The detector was designed to be bottom-illuminated through the N⁺ layer and optimized for operation at a temperature of 230 K. The cut-off wavelength ($\lambda_{c}$) is conditioned by the Cd composition of the active layer and operating temperature. At 230 K, $\lambda_{c}$ determined at 50% of the maximal signal is about 3.25 μm (see Fig. 1).

Fig. 1

Current responsivity for backside illuminated HgCdTe N⁺/T/p/T/P⁺/n⁺ photodiode operating at selected temperatures.

Figure 2 shows the current density–voltage (J–V) and capacitance–voltage (C–V) characteristics of the N⁺/T/p/T/P⁺/n⁺ photodiode. A strongly asymmetric nature of the J–V curves was observed, indicating excellent rectifying behaviors, which means that the current flows through the p–n junction in one direction only. C–V measurements were also performed to define the depletion width of the diode. The value of the junction capacitance at zero bias is of the order of 30 pF, which proves that the measurement can be made using the DLTS method.

Fig. 2

Current–voltage characteristics (a) and capacitance–voltage characteristics (b) for HgCdTe N⁺/T/p/T/P⁺/n⁺ photodiode operating at selected temperatures.

Experiment

DLTS studies were carried out on mesas with an area of 400 μm × 400 μm. Detector mesas were defined by standard optical photolithography and wet chemical etching with Br:HBr (1:100) diluted in deionized water (50:50:1 Br:HBr:H₂O) solution. Cr/Au contacts were deposited on the top and bottom contact layers.

In order to extract defects from individual layers of the heterostructure, three consecutive etchings were performed for the same epitaxial layer. In the first experiment, the etching was performed through the entire N⁺/T/p/T/P⁺/n⁺ structure to the N⁺ bottom contact (Fig. 3a). In the second experiment, the bottom contact was made to the p-type absorber—p/T/P⁺/n⁺ structure (Fig. 3b). In the third experiment, the top layers were removed and a N⁺/T/p type detector was made—the cap contact was made to the p-type absorber and the bottom to the N⁺ layer (Fig. 3c).

Fig. 3

Three chemical etching procedures for DLTS measurement: (a) the N⁺/T/p/T/P⁺/n⁺ structure was chemically etched to the N⁺ bottom contact; (b) the bottom contact was made to the absorber—p/T/P⁺/n⁺ structure; (c) electrical contact between p-type absorber and N⁺ layer—N⁺/T/p structure.

Results

Figure 4 shows calculated band profiles for the analyzed photodiode operating at zero and 0.2 V reverse bias voltage, and the temperature of 230 K. Calculations were made by commercially available APSYS software. The calculated band diagram for the N⁺/T/p/T/P⁺/n⁺ HgCdTe heterostructure shows three junctions (see Fig. 4a). The highest voltage is deposited at the junction between the absorber and bottom contact layer and extends deep into the absorber layer. It follows that the DLTS method can be used for such a structure, because the measured transient capacitance can only be attributed to changes in the junction capacitance associated with the HgCdTe absorber. To confirm this, the experiment was divided into three parts (see Fig. 4b and c). In the second experiment, two junctions are visible: p–P⁺ and P⁺–n⁺, which are not present in the third experiment. The third experiment contains only one junction (N⁺–p), which is the dominant junction for the entire photodiode.

Fig. 4

Calculated band diagrams for HgCdTe heterostructure detector N⁺/T/p/T/P⁺/n⁺ (a) and division of the heterostructure into single p-n junctions (b) p/T/P⁺/n⁺ structure; (c) N⁺/T/p structure. The orange line is the valence band (VB) and the blue line is the conduction band (CB). Calculations were made at a temperature of 230 K (Color figure online).

DLTS measurements were performed for two frequencies, 100 kHz and 300 kHz, with the rate window (RW) values of 144 s⁻¹ and 240 s⁻¹ respectively, and filling pulse width $t_{p} =$ 1 ms. In the first experiment, the voltage applied to the device during the measurement (bias voltage) was set to V_R = −1.2 V and the voltage applied to fill traps (fill pulse) was set to V_p = −0.4 V and V_p = 0.4 V. In two successive experiments, the bias voltage was set to V_R = −1 V and the fill pulse was set to V_p = −0.2 V and $V_{{\text{p}}} =$ 0.2 V. The applied reverse bias voltage has determined the width of the space charge region in the p-type absorber. The negative fill pulse caused the trap levels to be filled with electrons in the space charge region (see Fig. 5a). After the negative pulse ceased, the reverse voltage was restored and electrons were emitted into the conduction band (CB). The applied positive fill pulse filled the trap levels with holes (see Fig. 5b), and after the cessation of the pulse, holes were emitted into the valence band (VB).

Fig. 5

Scheme of filling the traps in the p-type absorber at (a) negative fill pulse, (b) positive fill pulse.

DLTS spectra performed for three experiments is shown in Fig. 6. Six localized defects were found across the entire N⁺/T/p/T/P⁺/n⁺ photodiode (first experiment shown in Fig. 6a). DLTS using 300-kHz capacitance measurements revealed one trap with a positive fill pulse (peak B) and four traps with a negative fill pulse (peaks A, D, E, F). DLTS spectra using 100-MHz capacitance measurements were also taken and compared with those at 300 kHz. One trap was revealed at a frequency of 100 kHz (peak C) which does not overlap with the measurement at 300 kHz. Since for the entire N⁺/T/p/T/P⁺/n⁺ photodiode it is difficult to clearly determine which of these traps are localized in the absorber and whether they are electron traps or hole traps, DLTS measurement was performed for the isolated junctions. Two localized defects were found in the second experiment performed for the p/T/P⁺/n⁺ structure (Fig. 6b). One trap (peak E) overlaps with the first experiment and is not present in the third one (N⁺/T/p structure). This suggests that it most likely comes from one of the upper layers etched off for the third experiment. Five defect levels were identified in the third experiment—N⁺/T/p structure (Fig. 6c), and three of them (peaks A, B, and C) overlap with the first experiment. The traps associated with these peaks can be considered to lie in the absorber layer. A and C are assigned to the electron traps (revealed with a negative fill pulse) while B is assigned to the hole trap (revealed with a positive fill pulse).

Fig. 6

DLTS signal for first (N⁺/T/p/T/P⁺/n⁺ structure) (a), second (p/T/P⁺/n⁺ structure) (b) and third (N⁺/T/p structure) experiment (c). The red and blue lines were designated at negative V_p, and the green line was designated at positive V_p (Color figure online).

Moreover, the shape of the A, B, C, and E DLTS peaks suggests point defects, probably related to mercury vacancies or impurity atoms which occupy anion or cation sites.^17,18 DLTS peaks from point defects are narrow and symmetrical, and their extremes do not change position with increasing time needed to fill traps, while for extended defects the DLTS signal line is always widened and the position of the signal maximum remains almost constant when changing the length of the filling pulse.

Arrhenius plots for the entire and divided structure are presented in Fig. 7. DLTS parameters such as capture cross sections for carriers and activation energies are shown in Table I. Three activation energies for peaks A, B, and C obtained from the first experiment overlap with the activation energies from the third experiment. Activation energies indicate that these defects are located within the absorber’s bandgap (E_g = 354 meV at 77 K and E_g = 385 meV at 300 K). B is estimated to have an energy level at E_V + 183 meV and a hole-capture cross section of 4.1 × 10^–12 cm². A (C) has a level at E_C − 283 meV (E_C − 267 meV), an electron capture cross section of 5 × 10^–16 cm² (2.6 × 10^–15 cm²), averaged over two experiments. The classical trap is characterized by a large capture cross section in the range of 10^–15–10^–12 cm², then the cascading process is dominant.¹⁹ The others traps, with capture cross sections in the range of 10^–16–10^–15 cm², are neutral.²⁰

Fig. 7

Arrhenius plot of emission rate for first (N⁺/T/p/T/P⁺/n⁺ structure) (a), second (p/T/P⁺/n⁺ structure) (b) and third (N⁺/T/p structure) experiment (c).

Table I The trap energies and capture cross sections found by the DLTS method

As mentioned earlier, the E peak comes from one of the upper layers. Considering the activation energy equal to 479 meV, it is most likely located within the bandgap of the wide-gap P⁺ layer (E_g = 625 meV at 77 K and E_g = 616 meV at 300 K). Activation energy equal to 618 meV for the F peak, which is visible only in the first experiment, is close to the bandgap of the P⁺ layer and is much larger than the energy gap for other photodiode layers. Activation energies higher than the energy gap can be explained by two effects. First, the activation energy is not an energy level in the band structure but an enthalpy change. Second, the cross section of the capture can also be thermally activated.

Figure 8 shows the Arrhenius plot of the dark current measured at 0.1 V reverse biased N⁺/T/p/T/P⁺/n⁺ photodiode. The estimated thermal activation energy of ∼183 meV is close to 0.5E_g, what suggests that the dark current at low reverse bias is dominated by the SRH mechanism. This activation energy coincides with the DLTS trap located at the energy level of E_V + 183 meV.

Fig. 8

Current density as a function of 1/kT for U = −0.1 V and activation energy obtained from the slope of the curve.

Conclusions

The paper presents the possibilities of DLTS analysis of a complex heterostructure containing several depletion regions. Performing three different experiments separating individual junctions of the heterostructure allows determining which layer the trap comes from. Three peaks A, B, and C with activation energies of E_C − 283 meV, E_V + 183 meV, and E_C − 267 meV were detected within the absorber band gap. The deep trap level located at the energy of 183 meV above the top of the valence band coincides with the activation energy determined from the Arrhenius plot for the dark current. The activation energy corresponds to ∼0.5E_g, which means that the dark current at low reverse bias is dominated by the SRH mechanism. The determined electron capture cross section for this trap is 4.1 × 10^–12 cm². The E peak with the activation energy of 479 meV is most likely located within the bandgap of the wide-gap P⁺ layer.