Progress in MOCVD growth of HgCdTe heterostructures for uncooled infrared photodetectors

Infrared Physics & Technology 49 (2007) 173–182 www.elsevier.com/locate/infrared

Progress in MOCVD growth of HgCdTe heterostructures for uncooled infrared photodetectors A. Piotrowski

b

a,*

, P. Madejczyk b, W. Gawron b, K. Kłos a, J. Pawluczyk a, J. Rutkowski b, J. Piotrowski a, A. Rogalski b

a VIGO System SA, 3 S´wietliko´w Str., 01-389 Warsaw, Poland Institute of Applied Physics, Military University of Technology, 2 Kaliskiego Str., 00-908 Warsaw, Poland

Available online 9 August 2006

Abstract This paper describes the significant progress in the development of metalorganic chemical vapour deposition of Hg1xCdxTe (HgCdTe) multilayer heterostructures on GaAs/CdTe substrates for uncooled infrared photodetectors. The paper focuses on the interdiffused multilayer process (IMP). The optimum conditions for the growth of single layers and complex multilayer heterostructures have been established. One of the crucial stages of HgCdTe epitaxy is CdTe nucleation on GaAs substrate. Successful composite substrates have been obtained with suitable substrate preparation, liner and susceptor treatment, proper control of background fluxes and appropriate nucleation conditions. Epiready (1 0 0) GaAs wafers with 2–4 disorientation towards h1 0 0i and h1 1 0i have been used. Due to the large mismatch between GaAs and CdTe, both (1 0 0) and (1 1 1) growth may occur. Generally, layers with orientation (1 0 0) show superior morphology compared to (1 1 1), but they are also characterized by hillocks. The benefits of the precursors, ethyl iodine (EI) and arsine (AsH3), for controlled iodine donor doping and arsenic acceptor doping at dopant concentrations relevant for HgCdTe junction devices are summarized. In situ anneal seems to be sufficient for iodine doping at any required level. In contrast, efficient As doping with near 100% activation requires ex situ anneal at near saturated mercury vapours. Finally, the multilayer fully doped heterostructures for photovoltaic devices operated at room temperature have been fabricated. The special attention is focused on the improvement in multijunction LWIR photovoltaic detectors. The performance of photodiodes is also presented.  2006 Elsevier B.V. All rights reserved. Keywords: MOCVD growth; HgCdTe ternary alloy; Room temperature photovoltaic devices

1. Introduction Historically, crystal growth of Hg1xCdxTe (HgCdTe) has been a major problem mainly because a relatively high Hg pressure is present during growth, which makes it difficult to control the stoichiometry and composition of the grown material [1–3]. The wide separation between the liquidus and solidus, leading to marked segregation between CdTe and HgTe, was instrumental in the development of all the bulk growth techniques to this system. In addition *

Corresponding author. E-mail address: [email protected] (A. Piotrowski).

1350-4495/$ - see front matter  2006 Elsevier B.V. All rights reserved. doi:10.1016/j.infrared.2006.06.026

to solidus–liquidus separation, high Hg partial pressures are also influential both during growth and post-growth heat treatments. Early experiments and a significant fraction of early production were done using a quench-anneal or solid-state recrystallization process. Bridgman growth was attempted for several years near the mid-70s last century. At the same time, solvent growth methods from Te-rich melts were initiated to reduce the grown temperature. One successful implementation was the travelling heater method up to 5-cm diameter [4]. The bulk HgCdTe crystals were initially used for any types of infrared photodetectors. At present they are still

174

A. Piotrowski et al. / Infrared Physics & Technology 49 (2007) 173–182

used for some infrared applications [5] such as n-type single element photoconductors, SPRITE detectors and linear arrays. The epitaxial techniques offer, in comparison with bulk growth techniques, the possibility to grow large area epilayers and sophisticated device structures with good lateral homogeneity, abrupt and complex composition and doping profiles, which can be configured to improve the performance of photodetectors. The growth is performed at low temperatures, which makes it possible to reduce the native defects density. Among the various epitaxial techniques, the liquid phase epitaxy (LPE) is the most matured method [6]. At present the vapour phase epitaxy (VPE) growth of HgCdTe is typically done by nonequilibrium methods; metalorganic chemical vapour deposition (MOCVD), [7] molecular beam epitaxy (MBE), [8] and their derivatives. The great potential benefit of MBE and MOCVD over the equilibrium methods is the ability to modify the growth conditions dynamically during growth to tailor band gaps, add dopants, prepare surfaces and interfaces, add passivations, perform anneals, and even grow on selected areas of a substrate. The growth control is exercised with great precision to obtain basic materials properties comparable to those routinely obtained from equilibrium growth. Earlier works on the fundamental issues of MOCVD growth and doping have been directed toward receiving of high quality cryogenically and thermoelectrically cooled HgCdTe photodiodes [7,9–11] and non-equilibrium devices [12,13]. This paper presents recent advances in MOCVD growth of HgCdTe in our laboratory, with emphasis on these aspects of technology for the growth of heterostructure photovoltaic devices operated at room temperature [14] in both middle wavelength (MW) and long wavelength infrared (LWIR) spectral ranges. The special attention is focused on the improvement in multijunction LWIR photovoltaic detectors. The performance of these detectors is also presented.

Carrier gas H2 EI DEZn DMCd

Liner purge

2. Hg1xCdxTe MOCVD growth One of the important advantages of MOCVD system is the possibility to use low-cost composite substrates (GaAs, Si, sapphire) as the viable alternative to costly CdZnTe substrates. This system does not require high vacuum for growth, is easier serviceable technique with larger throughput. Donor and acceptor doping at medium and high level, essential for near room temperature devices, is simpler with MOCVD. We did the Hg1xCdxTe epitaxial growth in the horizontal reactor of AIX-200 Aixtron MOCVD unit (Fig. 1). It was equipped with additional mercury bath. Two separated gas inlet channels are provided to prevent premature gas reactions and dust formation. Hydrogen is used as a carrier gas. Dimethylcadmium (DMCd), diisopropyltelluride (DIPTe) and diethylzinc (DEZn) used as precursors are held in temperature-stabilized baths. Ethyl iodine (EI) is used as a donor and arsenic (AsH3) as acceptor dopant sources. DMCd, DEZn and EI are delivered through the upper channel while DIPTe and AsH3 are delivered through the lower channel over the quartz container where elementary mercury is held. Hg1xCdxTe and CdTe were grown at 360–410 C with mercury source kept at 200– 220 C. Higher temperatures are used for reactor thermal cleaning. Aixtron’s gas foil rotation technique has been used for better composition uniformity onto 200 substrates. Reactor pressure of 500 mbar was typically used. Hydrogen from the upper channel dilutes mercury saturated hydrogen from the lower channel. Therefore, decrease in the mercury partial pressure over the wafer must be taken into account and it can be altered by changing the lower to upper channel flow ratio. In practice, the flow is selected separately for each phase; for example 100/500 sccm or 1200/1200 sccm flow ratios have been used. In this notation the first number is mass gas flow in the upper channel and the former is mass gas flow in the lower channel.

Pressure control

Quartz reactor Quartz liner

Partition

Wafer Rotating satellite

Hg

Graphite susceptor

thermocouple AsH3 DIPTe Carrier gas H2

Wafer thermocouple Hg IR heater

Wafer IR heater Vacuum pump

Fig. 1. Schematic drawing of Aix-200 II–VI reactor.

A. Piotrowski et al. / Infrared Physics & Technology 49 (2007) 173–182

CdTe buffer growth on GaAs is probably the most critical stage in HgCdTe growth due to lattice mismatch. For GaAs and CdTe, the mismatch is approximately 14.3%. 2–10-lm thick CdTe buffer neutralizes so huge mismatch. The buffer plays also a role of Ga diffusion barrier [15]. Careful chemical and thermal treatment of the reactor must be carried out after each growth run to prevent residual deposits of HgCdTe on substrate that adversely affect nucleation stage of growth. Different phases on the edges and in the middle of the wafer were frequently observed for poorly cleaned reactor. Epiready (1 0 0) GaAs wafers with 2–4 disorientation towards h1 0 0i and h1 1 0i have been used. Due to the large mismatch between GaAs and CdTe, both (1 0 0) and (1 1 1) growth may occur. It mostly depends on substrate disorientation and preparation, nucleation conditions and growth temperature. Cd or Te substrate treatment just before growth results in (1 0 0) and (1 1 1) orientation, respectively. Generally, layers with orientation (1 0 0) show superior morphology compared to (1 1 1) but they are also characterized by hillocks. Height of hillocks is typically twice higher compared to the layer thickness. Origin of hillocks is not fully understood at present. Hillocks are created during CdTe buffer growth and HgCdTe growth only enlarging them. Epilayers with hillocks are practically useless for device fabrication. Various ways to prevent hillocks formation have been tried: Zinc nucleation layer, different Cd/Te ratios, different substrates orientations have been used, but we were able to grow hillocks-free layers only on some substrates from one supplier. Therefore, most of the devices were based on (1 1 1) layers. Direct Hg1xCdxTe growth is difficult due to different thermodynamic properties of HgTe and CdTe [7,16]. We applied interdiffused multilayer process (IMP) for Hg1xCdxTe growth to avoid this problem. IMP gives the possibility for controllable growth of heterostructures with complex composition and doping profiles. HgTe and CdTe layers are deposited one after another with period 100–200 nm and homogenized by interdiffusion during growth. Any composition can be achieved this way by proper selection of HgTe and CdTe layer growth times. Detailed information about IMP growth of Hg1xCdxTe epilayers are gathered in two recently published papers [17,18]. The composition uniformity across the wafer has been evaluated by mapping of infrared transmission spectra. The transmission spectra are close to the theoretical ones. This is an evidence of good in-depth homogeneity. The composition uniformity over the wafer as low as ±0.1% is achieved. 3. Properties of Hg1xCdxTe layers Electrical properties of nominally undoped Hg1xCdxTe layers are determined by native acceptors (metal vacancies) and uncontrolled background doping. Vacancy concentration strongly depends on reactor temperature, vapour pres-

175

sure and material composition. The native doping level after growth at 360 C with mercury temperature 200 C is approximately 5 · 1016 cm3. This is much below the expected level of >1 · 1017 cm3 for this temperature conditions [19]. We suspect that it is due to partial annihilation of vacancies during cool down. It could be influenced by hydrogen presence in the layer. Post-growth annealing in mercury vapours can change the vacancy concentration. Vacancies may be practically eliminated by a prolonged, low temperature (200 C) annealing at near saturated mercury pressures. Such annealing reveals the background doping level about Nd  Na = (1–3) · 1015 cm3 for our MOCVD system. Ex situ anneal in sealed quartz ampoules in mercury vapours is typically used to maintain isothermal conditions. We developed more convenient ex situ anneal technique in which a wafer after growth was transported into reusable quartz container. Temperatures up to 400 C and Hg pressures up to 2 bars could be used. Vacancy and background doping is not sufficient for advanced infrared devices that require donor and acceptor doping at medium and high doping levels. Iodine and arsenic have been used as foreign dopants. Both are well behaved, stable and slowly diffusing dopants. To be active dopants they must occupy Te sites. Metal-rich conditions are favourable for iodine incorporation and are required for arsine. Therefore they have been introduced during CdTe phase with Cd/Te ratio >1. Most of doping experiments have been carried out for (1 1 1) Hg1xCdxTe layers. 3.1. Iodine doping Ethyl iodine is a highly effective precursor without any memory effects [20]. Cd/Te ratio of 1.1–1.3 has been used during CdTe phase to improve efficiency of I incorporation and activation. It allows doping concentration between 1 · 1017 and 1 · 1018 cm3 for the compositions of about 0.3. It appears that iodine incorporation efficiency in HgCdTe from EI exhibits a strong dependence on the precise orientation of the substrate [18,21]. Fig. 2 shows the concentration of iodine in Hg1xCdxTe (x  0.20) layers, measured by SIMS, for a number of misorientations from the (1 0 0) plane toward (1 1 1)B. During growth, the EI partial pressure was maintained at constant values marked in the figure. The HgCdTe layers were grown on CdZnTe [21] and GaAs substrates (our results). We can see that the iodine concentration varies over a wide range form mid-1015 cm3 for (1 0 0)8 ! (1 1 1)A to mid-1018 cm3 for the (2 1 1)B orientation. The variation in dopant incorporation with crystallographic orientation is well known for several dopants in GaAs and other III–V semiconductors. The iodine occupies the Te sublattice sites substitutionally. The Te sides on the (2 1 1)B surface provide a more stable and favourable adsorption site with three bonds from the underlying group II atoms while the

176

A. Piotrowski et al. / Infrared Physics & Technology 49 (2007) 173–182

1019

Iodine concentration (cm-3)

(211)B 1018

(111)B

1017

1016

EI partial pressure 1.81x10-6 atm. (Ref. 9) -6 8.5x10 atm. (our results) 1015 -20

0 20 40 Misorientation from (100) in degrees

60

Fig. 2. Iodine incorporation in Hg1xCdxTe (x  0.20) vs. misorientation from the (1 0 0) plane toward (1 1 1)B measured by SIMS. During growth, the EI partial pressure was maintained at constant values: 1.81 · 106 atm (j, after Ref. [9]) and 8.5 · 106 atm (m, our results).

(1 0 0) surface provides a weaker adsorption site with two bonds. Good reproducibility of I doping has been observed. Fig. 3 shows the relationship between 300 K electron concentration and alloy composition. It is clearly shown decrease of iodine incorporation with composition increasing. Preliminary experiments with increasing Cd/Te ratio have not improved EI doping noticeably. However, slight improved efficiency of doping has been observed with ex situ annealing at 360 C at near saturated mercury pressures. This effect may be related to incomplete iodine

homogenization during growth and in situ annealing, since iodine diffusion is very slow (noticeably slower then Cd diffusion into HgTe). Heterostructure photovoltaic IR detectors require heavily doped n+-layers with relatively high x composition (x > 0.45). It is expected that such layers should have resistivity as small as possible to provide good electrical contact. Until now, the Hg1xCdxTe layers with composition x > 0.45 doped with EI doses above 10 ppm had electron concentration lower than 7 · 1017 cm3 and sheet resistance above 10 X. These parameters indicate that significant part of iodine atoms are not located in Te lattice and are not electrically active. The transport properties of iodine-doped HgCdTe layers have been also characterized extensively in our laboratory. In Fig. 4, the 77 K mobilities obtained as a function of electron concentration for iodine-doped samples are shown. We note that the mobilities of electrons in samples fabricated in our laboratory are comparable with previously published values for layers doped using iodine as a vapour from the solid element [22]. Expected decreasing of mobility with increasing doping is due to ionized impurity scattering mechanism. Temperature-dependent lifetime measurements have been carried out using the steady-state photoconductivity technique. Experimentally measured data for Hg1xCdxTe (x  0.30) samples with electron concentrations about 1016 cm3 agree well with a theoretical model that include Auger 1 and radiative recombination mechanisms [3]. 3.2. Arsenic doping For acceptor doping the most widely used dopant in HgCdTe is arsenic. In MOCVD there have been extensive studies of arsenic doping with a variety of precursors. The most popular is arsine (AsH3) used also in our laboratory.

1019

10

6

x = 0.20 — 0.24 T = 77K

Mobility at 77K (cm2/Vs)

Electron concentration (cm-3)

EI: 10 ppm Cd/Te = 1.5

1018

10

17

105

104 after Ref. 22 our results

Annealing ex situ in situ

10

103 14 10

16

0.2

0.3

0.4 0.5 Composition x

0.6

0.7

Fig. 3. Room temperature electron concentration for iodine-doped HgCdTe layers vs composition x. HgCdTe layers were grown with Cd/Te ratio of 1.5.

1015

1016

1017

1018 -3

Carrier concentration at 77 K (cm )

Fig. 4. 77 K mobilities of iodine-doped Hg1xCdxTe (x = 0.20–0.24) layers vs. electron concentration. The layers have been doped using ethyliodine (d) and iodine as a vapour from the solid element (m, after Ref. [22]).

A. Piotrowski et al. / Infrared Physics & Technology 49 (2007) 173–182

1018

Hole concentration at 77 (cm-3)

As less toxic liquid precursors such as tertiarybutylarsine (TBAsH2), phenylarsine (PhAsH2) and dimetyhylaminoarsenic (DMAAs) became commercially available. Arsine incorporation into Te sites is very sensitive for metallic rich conditions. This assures conditions for arsenic to enter Te sites where it acts like an acceptor. Otherwise arsenic can show amphoteric behaviour. The arsenic incorporation rate and doping efficiency strongly depend on crystallographic orientation. Fig. 5 shows the efficiency of arsenic doping as a function of AsH3 partial pressure for two crystallographic orientations: for the (1 0 0) and (1 1 1) ones. Different orientations were obtained by different nucleation conditions. The difference in doping efficiency for two orientations is clearly evident, with near an order of magnitude higher incorporation rate in (1 0 0) orientation. This strong polarity dependence can be qualitatively understood by considering that the two Cd and As atoms are jointly involved in the incorporation process, because the adsorbed species originate from the adduct DMCd-AsH3. The detailed explanation of this phenomenon is presented in Ref. [23]. p-Type doping using AsH3 precursor was achieved over the range of 3 · 1015–5 · 1017 cm3 in IMP growth, where the As-precursors are injected in the CdTe growth cycle under Cd-rich conditions. For example, Fig. 6 shows the hole concentration of arsenic doped (1 1 1)Hg0.72Cd0.28Te layers at 77 K vs. arsine concentration in reactor. The HgCdTe layers were grown with Cd/Te ratio of 1.5. Thermal annealing is required to enhance activation of the As acceptors. As Fig. 6 indicates, an ex situ isothermal annealing in high mercury vapour (360/350 C) significantly affects the acceptor doping by increasing hole concentration. It is much more difficult to achieve heavily doped x-low HgCdTe materials. The growth time of CdTe cycle (the active time of As doping) is considerably shorter than HgTe cycle for x-low layers. As a result, the relative arsenic incorporation is worse in epilayers with x-low composi-

177

x = 0.28 Cd/Te = 1.5

1017

1016 Annealing ex situ in situ

1015 -2 10

1

10 10-1 100 102 AsH3 concentration in reactor (ppm)

Fig. 6. 77 K hole concentration of arsenic doped (1 1 1)Hg0.72Cd0.28Te layers vs. arsine concentration. The HgCdTe layers were grown with Cd/Te ratio of 1.5.

tions than in x-high ones. So far, after postgrowth annealing, concentrations of up to 3 · 1017 cm3 have been achieved. Fig. 7 shows that hole concentration of 2 · 1017 cm3 in (1 1 1)Hg0.83Cd0.17Te layer can be achieved under the following growth conditions: Cd/Te ratio equal to 5 and arsine fraction of 4 ppm in the growth zone (mercury temperature of 220 C). Higher doping concentrations (>5 · 1017 cm3) have been achieved in higher x, postgrowth annealing materials (x > 0.25) with Hg temperature of 220 C and arsine fraction of 10 ppm. Early studies of arsenic doped MOCVD HgCdTe layers using TBAsH2 and PhAsH2 precursors indicated that the carrier lifetimes were significantly lower than those obtained in As-doped HgCdTe grown by Hg-rich liquid phase epitaxy (LPE) [22,24]. It was suggested [25] that As-doping of CdTe with AsH3 causes the incorporation of As–H pairs in addition to As. It is likely that As–H pairs are incorporated also with TBAsH2 and PhAsH2 precursors

1019 18

x = 0.28—0.33

10

18

Annealing

(100) layers

10

17

Hole concentration (cm-3)

SIMS As concentration (cm-3)

10

ex situ in situ

10

(111) layers

10

16

1015 -2 10

17

As: 4 ppm x = 0.167 — 0.175 16

10 100 102 10-1 Arsenic concentration in the reactor (ppm)

Fig. 5. Arsenic incorporation in Hg1xCdxTe (x = 0.28–0.33) vs. AsH3 partial pressure, measured by SIMS, for the (1 0 0) and (1 1 1) orientations.

10

0

1

2

3 4 5 Cd/Te ratio

6

7

8

Fig. 7. Room temperature hole concentration for arsenic doped (1 1 1)Hg1xCdxTe (x  0.17) vs. Cd/Te ratio maintained during growth.

178

A. Piotrowski et al. / Infrared Physics & Technology 49 (2007) 173–182

The photovoltaic devices of conventional design operated at higher temperatures suffer from

Carrier lifetime at 80 K (s)

10-4

10

10

x = 0.30 T = 80 K

-5

• poor quantum efficiency, and • low differential resistance.

-6

10-7

10

-8

10-9 1014

DMAAs TBAsH2 PhAsH2 AsH3 17

15 10 1016 10 80 K Hall concentration (cm-3)

10

18

Fig. 8. Comparison of 80 K lifetimes in As-doped Hg1xCdxTe (x  0.30) layers from DMAAs, TBAsH2, PhAsH2, and AsH3. The experimental data for HgCdTe layers grown with DMAAs, TBAsH2, and PhAsH2 precursors are taken from Ref. [9]; our results (d) concerns HgCdTe laters doped with AsH3. The solid line represents traditional radiative limit for Hg0.70Cd0.30Te (without including reabsorption effects).

since they are substituted AsH3 in MOCVD HgCdTe growth. Although As–H complexes are expected to be electrically neutral, they are likely to be recombination centres in HgCdTe and are probably an important factor in the lower lifetimes observed in As-doped films grown with the TBAsH2 and PhAsH2 precursors. Unlike TBAsH2 and PhAsH2, dimethylaminoarsenic (DMAAs) – another precursor used for p-type HgCdTe doping, [9,25,26] has no As–H bonds and it is therefore expected that As–H complexes will not be incorporated in the As-doped films [27]. Fig. 8 compares lifetime measurements reported for Asdoped Hg1xCdxTe (x  0.30) films grown with DMAAs, TBAsH2, PhAsH2, and AsH3 precursors. The data taken at 80 K are plotted as a function of the Hall concentration. It is clear that the lifetimes of the films doped with DMAAs show higher values of lifetimes than those measured in films doped with TBAsH2, PhAsH2, and AsH3. However, our experimental data obtained using AsH3 precursor coincide well with theoretical predictions (solid line) for traditional radiative limit for Hg0.70Cd0.30Te (without including reabsorption effects). These data provide confidence that the carrier lifetimes in As-doped MOCVD layers with AsH3 precursor are close to values reported in As-doped Hg-rich LPE grown HgCdTe. 4. Uncooled HgCdTe devices with improved performance The class of photon detectors is sub-divided into different types depending on how the electric or magnetic fields are developed. Initial efforts to produce near room temperature HgCdTe photon detectors have concentrated mainly on optimization of photoconductors and PEM detectors. More information can be found in previously published papers [14,28–30].

The second factor is especially crucial for LWIR devices. Only charge carriers that are photogenerated at a distance shorter than the diffusion length from junction can be collected. The absorption depth of LWIR radiation (k > 5 lm) is longer than the diffusion length. Therefore only a limited fraction of the photogenerated charge contributes to the quantum efficiency. The calculations show that the ambipolar diffusion length in an uncooled 10.6 lm photodiode is less than 2 lm while the absorption depth is 13 lm. This reduces the quantum efficiency to 15% for single pass of radiation through the detector. The resistance of the p–n junction is very low due to a high thermal generation. In the materials with a high electron to hole mobility ratio, the resistance is additionally reduced by ambipolar effects. As a result, the preamplifier noise and noise of parasitic resistances may exceed the thermal generation-recombination noise. Thus, the performance of conventional devices is very poor, so they are not usable for practical applications. This section presents the recent progress in Hg1xCdxTe multilayer heterojunction (MLHJ) photodiodes. The investigated P+–P–p–N+ and n+–p–P+ structures are operated in middle wavelength (MW) and long wavelength (LW) spectral bands, where lower case letters refer to the absorber region while upper case letters refer to higher Cd mole fraction layers (wider band gaps). 4.1. MWIR devices An important aspect of MW MLHJ is a mesa P-doped/ p-doped/N-doped device structure that includes a widebandgap cap layer over a narrow-bandgap base (active) layer. The use of heterostructures restricts the volume of the absorber, and hence, the volume for dark current generation, while the higher energy gaps either side of the absorber limits minority carrier injection. The n-type absorbing base region was deliberately doped with iodine at a level of about 1015 cm3. The N+-type back region was doped with iodine at a level of about 5 · 1017 cm3. The p-type layer was doped with arsenic at a level of about 5 · 1017 cm3. Thus, the structure consists of five Hg1x CdxTe layers with type P+/P+/P/p/N+, composition x = 0.30/0.35/0.35/0.25/0.45 and thickness 0.5/0.5/1/5/5lm, respectively. Fig. 9 shows SIMS depth profile data and doping across the typical MLHJ structure. The standard photolithographic techniques were used for the device fabrication. The mesas were etched to depth of 8 lm, that is, into the N+-region, using a Br/glycol mixture. The mesa’s were passivated by MOCVD deposited CdTe. As an ohmic contact, Au was used for p-type side, while In—for the n-type side. In some cases, the ion milling

A. Piotrowski et al. / Infrared Physics & Technology 49 (2007) 173–182

(111) orientation Concentration (cm-3 )

1019 As (SIMS)

1018

Nd-a

Na-d

1017 10

I (SIMS)

16

1015 P+ P

p 0.25

Composition x= 0.3

N

0

0.35 0.35

2

4

+

CdTe 1

0.45

6 8 10 Thickness (μm)

12

14

Fig. 9. SIMS depth profiles and doping concentration in the structure.

was applied before In deposition. The test devices had the following diameters: 450 lm, 200 lm, and 130 lm. Fig. 10 shows the spectral responses at temperatures 300 K, 200 K and 77 K. Peak internal quantum efficiency was changed within temperature. The curves are classical in shape with no unexpected features. The cutoff wavelength decreases with temperature decreasing as the bandgap increases and agrees well with the room-temperature transmission edge of the base layer. The expected filtering of the photodiode response spectrum by the N+-type layer is clearly evident in the sharp increase in response at 2.3 lm. The maximum quantum efficiency value of up to 60% (without antireflection coating) for 4-lm cutoff wavelength device at 200 K is observed. The dark current of devices is thermally generated (diffusion limited devices – ideality factor is close to one). The reverse biased characteristics of the diodes exhibit negative resistance and soft breakdown at relatively low voltages (see Fig. 11). The negative resistance is associated with the Auger suppression that occurs when the devices are reverse biased, thereby reducing the carrier density in the p-type region. It is difficult to define a breakdown voltage because the reverse current increases continuously. To explain the cause of breakdown in reverse current, the

I–V characteristics at different temperatures were measured. The current depends strongly on temperature and the temperature coefficient of current at temperature above 200 K is positive, what indicates that thermally generated current dominate. At T < 200 K the weak dependence of current on temperature with a negative temperature coefficient of current implies the band-to-band tunnelling process. The dark current transport mechanisms are also exhibited by the behaviour of the dynamic differential resistance as a function of reverse bias voltage and operating temperature, as shown in Fig. 12. The maximum dynamic resistance, obtained at low reverse bias voltages, increases slightly with temperature decreasing and moves to zero bias voltage. At temperature 77 K the maximum value corresponds to forward bias voltage. These observations indicate that tunnelling near the surface at the circumference of the junction is very likely the cause of reverse current breakdown. Fig. 13 plots the R0A product against cutoff wavelength at 295 K and 180 K for three groups of backside-illuminated HgCdTe photodiodes: MBE, LPE and MOCVD. We can see that HgCdTe photodiodes with cutoff wavelengths less than 4 lm fall short of the limits calculated with the traditional radiative recombination equations. 20 10 128K

Current (mA)

1020

0

187K 211K

-20 231K

-30 -40 -0.3

-0.2

-0.1 Voltage (V)

0

0.1

Fig. 11. Current–voltage characteristics of MW MLHJ device operated at different temperatures.

35000

Dynamic resistance (Ω)

Responsivity (A/W)

165K

-10

2.5 2.0 300 K 1.5

200 K 77 K

1.0 0.5 0 2.0

179

2.5

3.0 3.5 4.0 Wavelength (μm)

4.5

5.0

Fig. 10. Spectral dependence of current responsivity for MW MLHJ device operating at 77 K, 200 K and 300 K.

30000

231 K 187 K 128 K

25000 20000 15000 10000 5000 0 -0.3

-0.2

-0.1 Voltage (V)

0

0.1

Fig. 12. Dynamic differential resistance vs. bias voltage for MW MLHJ at 128 K, 187 K, and 231 K.

180

A. Piotrowski et al. / Infrared Physics & Technology 49 (2007) 173–182 107 T=180 K 106 T=295 K

R0A (Ωcm2)

105

MBE LPE MOCVD MOCVD our lab

104 10 10

3

2

101 100 1.5

2.0

2.5 3.0 3.5 4.0 Cutoff wavelength (μm)

4.5

5.0

Fig. 13. R0A data for MWIR HgCdTe photodiodes compared with calculated R0A due to diffusion current with only radiative (solid curves) and radiative and Auger 1 (dashed curves) (after Ref. [31]). Open circle points represent our data.

At their respective temperatures, the R0A for all devices are limited by diffusion current, as determined both by the temperature dependence of R0A as well as by the shape of the current–voltage characteristics. The R0A curves in Fig. 13 are calculated for ideal thin-base limits set by diffusion current from an n-type absorber layer. The solid curves assume that only radiative recombination is present, while the dashed curves assume both radiative and Auger 1 recombinations. The open circle points represent our data. The satisfactory consistency between the theoretical curve and experimental data confirms high performance of MLHJ HgCdTe photodiodes fabricated in our laboratory. 4.2. LWIR devices The present generation of uncooled LWIR devices is based on photovoltaic devices [14,32–37]. The problems of poor quantum efficiency and large series resistance have been solved through adoption of sophisticated heterojunction architectures of photovoltaic devices in combination with the methods of reduction of physical size of an active element. An efficient improvement was development of multiple heterojunction photovoltaic devices to increase the voltage

responsivity of the devices. Practical realization of the multiheterojunction device that is consisted of a structure based on backside-illuminated n+–p–P photodiodes (symbol ‘‘+’’ denotes strong doping, capital letter—wider gap) have been presented in several papers [33–36]. The individual detector elements were prepared by a combination of conventional dry etching, angled ion milling, and angled thermal evaporation for contact metal deposition (see Fig. 14). p-Type active layer with thickness of approximately 3 lm and doping levels of about 1016 cm3 were grown on GaAs substrates using MOCVD and in situ As doping. The delineation trenches in the epilayer were wet chemical etched using Br/ethylene glycol or Br/HBr solutions to a depth of 2/3 of the thickness of epilayer. The etch was followed by shallow ion beam milling using a Kaufman-type ion gun. The sample was placed at 45 to the direction of the Ar+ beam so that only one wall of the trench was exposed to the beam. The ion beam milling results in n-type conversion and the formation of n+–p junctions on the uppermost surface and on one side of the trench walls. Due to the grading of the epilayer composition and doping with depth, the final structure approximates an n+–p–P+ photodiode. Cr/Au metallization was required to provide external contact to the device and to short-circuited N+–P+ junction formed at the base of trench, effectively connecting side-by-side diodes in series. As a final step, the devices were passivated by thermal evaporation of 200 nm of CdZnTe followed by 300 nm of ZnS. For an 3-lm thick epilayer, the minimum practical multijunction period using existing wet etching technology is 10 lm. The devices are manufactured as single devices optimized for any wavelength within the 2–16 lm range, with active area sizes from a few micrometers to a few millimetres. Linear arrays up to 120 elements and small 2D arrays are manufactured as custom devices. Fig. 15 shows the room temperature spectral response of HgCdTe multiheterojunction devices monolithically immersed to the CdZnTe lens. Generally, these room temperature devices have responsivities that are comparable to, or better than, photoelectromagnetic devices operating under the same conditions. Peltier cooled devices exhibit performance that is comparable to photoconductors operating at the same wavelength and temperatures. However,

Fig. 14. Backside-illuminated LWIR multiple heterojunction device.

A. Piotrowski et al. / Infrared Physics & Technology 49 (2007) 173–182

Detectivity (cmHz1/2/W)

10

12

10

11

D*BLIP

10

2TE cooling Opt. immersion

10

10

9

2

3

4

5

6 7 8 9 Wavelength (μm)

10 11 12

Fig. 15. Measured detectivity of multiheterojunction uncooled HgCdTe detectors. Optical immersion has been used to improve the performance.

181

low input resistance, the junction capacitance and the photodiode series resistance determine the RC time constant. Very short RC time constant is expected in optically immersed photodiodes with a very small area of an active region. With these improvements, photodiodes can be use for gigahertz range detection of IR radiation. Optical immersion has been used almost exclusively for single element devices. The use of a single immersion lens to a large array is problematic in view of optical aberrations and the large lens size. The problem can be solved by implementation of a small size 2D array monolithically integrated with microlenses matrix. The small size of active element (7 · 7 lm) is beneficial for good collection of photogenerated charge carriers especially in LWIR devices. The individual elements can be accessed individually or connected in series. 5. Conclusions

in contrast to photoconductors, multiheterojunction detectors can be used at both low and very high frequencies. Heterodyne experiments indicate that the response time of LWIR devices at a wavelength of 10.6 lm is only of about 1 ns. Without optical immersion, photovoltaic detectors are sub-BLIP devices with performance close to the g–r limit. Well designed optically immersed devices, when thermoelectrically cooled with two-stage Peltier coolers, show detectivities up to 1011 cm Hz1/2/W at 5 lm, closely approaching the BLIP limit. Situation is less favourable for LWIR photovoltaic detectors. Detectivities exceeding 1 · 109 cm Hz1/2/W and 6 · 109cm Hz1/2/W have been measured with uncooled k = 8.5 lm non-immersed and optically immersed devices, respectively. Optically immersed 10.6 lm PV devices cooled with two-stage Peltier cooler with detectivities >4 · 109 cm Hz1/2/W has been also measured. Despite of all improvements (advanced architecture, optical immersion, Peltier cooling) they show detectivities below the BLIP limit by almost one order of magnitude. The devices are especially promising as uncooled 7.8–9.5-lm detectors that can be used for thermal imagers. Detectivities about 109 cm Hz1/2/W at k  9 lm enable achieving the thermal resolution better than 0.1 K for staring thermal imagers operating with f/1 optics. The frequency response of photovoltaic devices is limited by transport of photogenerated charge carriers through the absorber region and by RC time constant. Transport through the absorber region is a combination of diffusion and drift. The p-type Hg1xCdxTe is the material of choice for an absorber of a fast photodiode due to a large diffusion coefficient of electrons. The diffusion transit times are 100 ps for extrinsic p-type 2-lm thick Hg0.82Cd0.18Te absorber. Further reduction can be achieved with thinner absorber. Drift transport in reverse biased devices can further reduce the transit time. The main limitation of response time is usually the RC time constant. For transimpedance preamplifier with a

Practical implementation of the advanced photodetector architecture requires well established epitaxial technology. The paper presents recent progress in the MOCVD growth of HgCdTe multilayer heterostructures on GaAs/CdTe substrates for IR photodetectors operated near room temperature. This technique has been selected for its inherent versatility (low growth temperature, ability to grow layered structures with complex composition and doping profiles while maintaining sharp interfaces). MOCVD makes possible to use low-cost and high quality substrates (GaAs, sapphire and silicon) and has the potential for a large-scale cost-effective production. Physical properties of MOCVD HgCdTe heterostructures on GaAs substrates exceed those on costly CdTe substrates. HgCdTe heterostructures have been grown on 200 (1 0 0)GaAs. Reproducible n- and p-type doping at the low, intermediate and high level (1015–1018 cm3) has been achieved with stable iodine and arsenic dopants. The dopants are easily activated during growth. The transport properties of MOCVD layers are comparable with those published for LPE epilayers. The experimental data provide confidence that the carrier lifetimes in MOCVD layers are close to theoretical limits determined by radiative and Auger recombinations. Final part of the paper presents the recent progress in Hg1xCdxTe MLHJ photodiodes grown by MOCVD and operated at high temperatures in both MW and LW spectral bands. The present generation of HgCdTe devices operated at higher temperature is based on photovoltaic devices. The problems of poor quantum efficiency and large series resistance of LWIR devices have been solved through adoption of sophisticated multiheterojunction architectures in combination with the methods of reduction of physical size of an active element. These devices are especially promising as uncooled 7.8–9.5-lm detectors that can be used for thermal imagers. Initial results are encouraging, indicating the potential for achieving 109 cm Hz1/2/W at k  9 lm.

182

A. Piotrowski et al. / Infrared Physics & Technology 49 (2007) 173–182

Acknowledgement This work was supported by the Polish Ministry of Science and Information Technology in the frame of the Project No. PBZ-MIN-009/T11/2003. References [1] W.D. Lawson, S. Nielson, E.H. Putley, A.S. Young, Preparation and properties of HgTe and mixed crystals of HgTe–CdTe, J. Phys. Chem. Solids 9 (1959) 325–329. [2] P. Capper, Bulk growth techniques, in: P. Capper (Ed.), Narrow-gap II–VI Compounds for Optoelectronic and Electromagnetic Applications, Chapman and Hall, London, 1997, pp. 3–29. [3] A. Rogalski, K. Adamiec, J. Rutkowski, Narrow-Gap Semicondcutor Photodiodes, SPIE Press, Bellingham, 2000. [4] L. Colombo, R.R. Chang, C.J. Chang, B.A. Baird, Growth of Hg-based alloys by the travelling heater method, J. Vac. Sci. Technol. A6 (1988) 2795–2799. [5] P. Norton, HgCdTe infrared detectors, Opto-Electron. Rev. 10 (2002) 159–174. [6] P. Capper, T. Tung, L. Colombo, Liquid phase epitaxy, in: P. Capper (Ed.), Narrow-gap II–VI Compounds for Optoelectronic and Electromagnetic Applications, Chapman and Hall, London, 1997, pp. 30– 170. [7] S.J.C. Irvine, Metal-organic vapour phase epitaxy, in: P. Capper (Ed.), Narrow-gap II–VI Compounds for Optoelectronic and Electromagnetic Applications, Chapman and Hall, London, 1997, pp. 71– 96. [8] O.K. Wu, T.J. deLyon, R.D. Rajavel, J.E. Jensen, Molecular beam epitaxy of HgCdTe, in: P. Capper (Ed.), Narrow-gap II–VI Compounds for Optoelectronic and Electromagnetic Applications, Chapman and Hall, London, 1997, pp. 97–130. [9] P. Mitra, F.C. Case, M.B. Reine, Progress in MOVPE of HgCdTe for advanced infrared detectors, J. Electron. Mater. 27 (1998) 510–520. [10] M.B. Reine, A. Hairston, P. O’Dette, S.P. Tobin, F.T.J. Smith, B.L. Musicant, P. Mitra, F.C. Case, Simultaneous MW/LW dual-band MOCVD HgCdTe 64 · 64 FPAs, Proc. SPIE 3379 (1998) 200–212. [11] C.D. Maxey, J.P. Camplin, I.T. Guilfoy, J. Gardner, R.A. Lockett, C.L. Jones, P. Capper, M. Houlton, N.T. Gordon, Metal-organic vapor-phase epitaxial growth of HgCdTe device heterostructures on three-inch-diameter substrates, J. Electron. Mater. 32 (2003) 656–660. [12] C.T. Elliott, New infrared and other applications of narrow gap semiconductors, Proc. SPIE 3436 (1998) 763–775. [13] C.T. Elliott, Photoconductive and non-equilibrium devices in HgCdTe and related alloys, in: P. Capper, C.T. Elliott (Eds.), Infrared Detectors and Emitters: Materials and Devices, Kluwer Academic Publishers., 2001, pp. 279–312. [14] J. Piotrowski, Uncooled operation of IR photodetectors, OptoElectron. Rev. 12 (2004) 111–122. [15] W.E. Tennant, C.A. Cockrum, J.B. Gilpin, M.A. Kinch, M.B. Reine, R.P. Ruth, Key issues in HgCdTe-based focal plane arrays: an industry perspective, J. Vac. Sci. Technol. B 10 (1992) 1359–1369. [16] L.M. Smith, J. Thompson, Metal organic chemical vapor deposition (MOCVD) of cadmium telluride, mercury telluride and cadmium mercury telluride, Chemtronics 4 (1989) 60–70. [17] A. Piotrowski, P. Madejczyk, W. Gawron, K. Kłos, J. Pawluczyk, M. Grudzien´, J. Piotrowski, A. Rogalski, Growth of MOCVD HgCdTe heterostructures for uncooled infrared photodetectors, Bull. Pol. Ac.: Technol. 53 (2005) 139–149.

[18] P. Madejczyk, A. Piotrowski, W. Gawron, K. Kłos, J. Pawluczyk, J. Rutkowski, J. Piotrowski, A. Rogalski, Growth and properties of MOCVD HgCdTe epilayers on GaAs substrates, Opto-Electron. Rev. 13 (2005) 239–251. [19] H.R. Vydyanath, Donor and acceptor dopants in Hg1xCdxTe alloys, J. Vac. Sci. Technol. B9 (1991) 1716–1723. [20] P. Mitra, Y.L. Tyan, T.R. Schimert, F.C. Case, Donor doping in metalorganic chemical vapor deposition of HgCdTe using ethyl iodine, Appl. Phys. Lett. 65 (1994) 195–197. [21] P. Mitra, F.C. Case, M.B. Reine, R. Starr, M.H. Weiler, Doping in MOVPE of HgCdTe: orientation effects and growth of high performance IR photodiodes, J. Cryst. Growth 170 (1997) 542–548. [22] C.D. Maxey, I.G. Gale, J.B. Clegg, P.A.C. Whiffin, Doping studies in MOVPE-grown CdxHg1xTe, Semicond. Sci. Technol. 8 (1993) S183– S196. [23] L. Svob, I. Cheze, A. Lusson, D. Ballutaud, J.F. Rommeluere, Y. Marfaing, Crystallographic orientation dependence of As incorporation in MOVPE-grown CdTe and corresponding acceptor electrical state activation, J. Cryst. Growth 184/185 (1998) 459–464. [24] D.D. Edwall, L.O. Bubulac, E.R. Gertner, p-Type doping of metalorganic chemical vapor deposition-grown HgCdTe by arsenic and antimony, J. Vac. Sci. Technol. B 10 (1992) 1423–1427. [25] M.J. Bevan, M.C. Chen, H.D. Shih, High-quality p-type Hg1xCdxTe prepared by metalorganic chemical vapor deposition, Appl. Phys. Lett. 67 (1995) 3750–3752. [26] P. Mitra, Y.L. Tyan, F.C. Case, R. Starr, M.B. Reine, Improved arsenic doping in metalorganic chemical vapor deposition of HgCdTe and in situ growth of high performance long wavelength infrared photodiodes, J. Electron. Mater. 25 (1996) 1328–1335. [27] B. Clerjaund, D. Cote, L. Svob, Y. Marfaing, R. Druilhe, Solid State Commun. 85 (1993) 167. [28] J. Piotrowski, W. Gawron, Z. Djuric, New generation of near-roomtemperature photodetectors, Opt. Eng. 33 (1994) 1413–1421. [29] J. Piotrowski, A. Rogalski, Photoelectromagnetic magnetoconcentration, and Dember infrared detectors, in: P. Capper (Ed.), NarrowGap II–VI Compounds and Electromagnetic Applications, Chapman & Hall, London, 1997, pp. 506–525. [30] J. Piotrowski, A. Rogalski, Uncooled long wavelength infrared photon detectors, Infrared Phys. Technol. 46 (2004) 115–131. [31] M.B. Reine, Photovoltaic detectors in HgCdTe, in: C.T. Elliott, P. Capper (Eds.), Infrared Detectors and Emitters: Materials and Devices, Kluwer Academic Publishers., Boston, 2001, pp. 313–376. [32] J. Piotrowski, W. Gawron, Extension of longwavelength IR photovoltaic detector operation to near room-temperatures, Infrared Phys. Technol. 36 (1995) 1045–1051. [33] C. Musca, J. Antoszewski, J. Dell, L. Faraone, J. Piotrowski, Z. Nowak, Multiheterojunction large area HgCdTe long wavelength infrared photovoltaic detector for operation at near room temperature, J. Electron. Mater. 27 (1998) 740–746. [34] J. Piotrowski, Z. Nowak, J. Antoszewski, C. Musca, J. Dell, L. Faraone, A novel multiheterojunction HgCdTe long-wavelength infrared photovoltaic detector for operation under reduced cooling conditions, Semicond. Sci. Technol. 13 (1998) 1209–1214. [35] W. Gawron, A. Rogalski, HgCdTe buried multijunction photodiodes fabricated by the liquid phase epitaxy, Infrared Phys. Technol. 43 (2002) 157–163. [36] J. Piotrowski, M. Grudzien´, Z. Nowak, Z. Orman, J. Pawluczyk, M. Romanis, W. Gawron, Uncooled photovoltaic Hg1xCdxTe LWIR detectors, Proc. SPIE 4130 (2000) 175–184. [37] J. Piotrowski, P. Brzozowski, K. Jo´z´wikowski, Stacked multijunction photodetectors of long wavelength radiation, J. Electron. Mater. 32 (2003) 672–676.