Surface leakage investigation via gated

When the gate bias is less than or equal to saturated negative gate bias VG ⬉ Vsat and not enough to generate a distinct field-induced depletion region at the M-structure surface, holes

Surface leakage investigation via gated type-II InAs/GaSb long-wavelength infrared photodetectors

G Chen, E K Huang, A M Hoang, S Bogdanov, S R Darvish et al

Citation: Appl Phys Lett 101, 213501 (2012); doi: 10.1063/1.4767905

View online: http://dx.doi.org/10.1063/1.4767905

View Table of Contents: http://apl.aip.org/resource/1/APPLAB/v101/i21

Published by the American Institute of Physics

Related Articles

Unipolar time-differential pulse response with a solid-state Charpak photoconductor

Appl Phys Lett 101, 213503 (2012)

High quality AlN grown on double layer AlN buffers on SiC substrate for deep ultraviolet photodetectors

Appl Phys Lett 101, 192106 (2012)

Polarity inversion and coupling of laser beam induced current in As-doped long-wavelength HgCdTe infrared detector pixel arrays: Experiment and simulation

Appl Phys Lett 101, 181108 (2012)

Quantum mechanical simulation of graphene photodetectors

J Appl Phys 112, 084316 (2012)

Transient photoresponse and incident power dependence of high-efficiency germanium quantum dot

photodetectors

J Appl Phys 112, 083103 (2012)

Additional information on Appl Phys Lett.

Journal Homepage: http://apl.aip.org/

Journal Information: http://apl.aip.org/about/about_the_journal

Top downloads: http://apl.aip.org/features/most_downloaded

Information for Authors: http://apl.aip.org/authors

Surface leakage investigation via gated type-II InAs/GaSb long-wavelength infrared photodetectors

G Chen, E K Huang, A M Hoang, S Bogdanov, S R Darvish, and M Razeghia)

Department of Electrical Engineering and Computer Science, Center for Quantum Devices,

Northwestern University, Evanston, Illinois 60208, USA

(Received 16 October 2012; accepted 2 November 2012; published online 21 November 2012)

By using gating technique, surface leakage generated by SiO2 passivation in long-wavelength

infrared type-II superlattice photodetector is suppressed, and different surface leakage mechanisms

are disclosed By reducing the SiO2passivation layer thickness, the saturated gated bias is reduced to

4.5 V At 77 K, dark current densities of gated devices are reduced by more than 2 orders of

magnitude, with 3071 X cm2 differential-resistance-area product at 100 mV With quantum

efficiency of 50%, the 11lm 50% cut-off gated photodiode has a specific detectivity of 7 1011

Jones, and the detectivity stays above 2 1011 Jones from 0 to500 mV operation bias.V C 2012

American Institute of Physics [http://dx.doi.org/10.1063/1.4767905]

Since Sai-Halasz et al.1 proposed the idea of type-II

InAs/GaSb superlattice (T2SL) in 1970 s, this material

sys-tem has demonstrated its ability to provide sensitive infrared

detection from the short-wavelength infrared region (SWIR)

to the very-long-wavelength infrared region (VLWIR)29

and is moving towards multi-spectral detection.10 However,

the performance of T2SL photodetectors, especially in

long-wavelength infrared region (LWIR), is still limited by

sur-face leakage due to the absence of an effective passivation

technique that can protect mesas from sidewall leakage

caused by chemically and mechanically aggressive focal

plane array (FPA) fabrication steps Such effects become

more severe as FPA pixel sizes scale down for higher

resolu-tion imagers

Surface leakage is believed to originate from the abrupt

termination of the periodic crystalline structure, contamination

from processing, and fixed charges within the passivation

layer, which can generate band bending on mesa-sidewalls

This band bending causes electron accumulation or type

inversion at sidewall surfaces, resulting in a conduction

chan-nel along sidewalls.11,12Various attempts have been made to

understand mesa sidewall surface physics and suppress

sur-face leakage in the long-wavelength infrared photodiodes

Measures to suppress this phenomenon developed in the past

include the double heterostructure12and hybrid graded doping

profile13 in detector designs, use of inductively coupled

plasma (ICP) dry etching,14 and various passivations.15–20

However, current physical understanding of surface leakage

current does not provide sufficient control of its intensity In

addition, SiO2passivation, which is known to generate serious

surface leakage in the long-wavelength infrared region, is still

the dominant and most suitable passivation technique for

T2SL FPA application Therefore, it is crucial to achieve a

deeper understanding of surface leakage physics and suppress

surface leakage specifically in long-wavelength infrared SiO2

passivated photodiodes

Recently, the gating technique has been demonstrated to

have the ability to eliminate surface leakage11in mid-wavelength

infrared (MWIR) Pþ-p-M-Nþ T2SL photodetectors This technique can actively control the sidewall surface band bending through creating a metal-insulator-semiconductor (MIS) structure on the mesa sidewall Gated diodes (GD) demonstrated higher detectivity because of a strongly reduced leakage current at saturated gate bias However, this technique requires very high gate biases to be applied and has not been transferred to the long-wavelength infrared region because for small band gap active regions, the surface leakage in long-wavelength infrared Pþ-p-M-NþT2SL het-erostructure photodetector is more severe and the leakage phenomenon might be more complicated The change of the surface potential is related to the band gap, doping level and the effective mass of the semiconductor.12The Nþ- and

Pþ-contact are large band gap heavily doped semiconductors,21 and the M-structure is a large band gap semiconductor with a much larger effective mass than the p- and Pþ-regions.22 Therefore, for simplicity, the discussion is focused on the p-region while the influence of gate bias on the M-structure,

Pþ-, and Nþ-contact is assumed to be less pronounced than the p-region

Since we already know that applying negative gate bias can realize flat band condition and eliminate surface leak-age,11 positive fixed charges must exist at the T2SL-SiO2 interface or within the SiO2passivation layer When the gate bias is less than or equal to saturated negative gate bias (VG

⬉ Vsat) and not enough to generate a distinct field-induced depletion region at the M-structure surface, holes are accu-mulated at the surface of the p-region, and the space charge region (SCR) is still mainly at the metallurgical junction (Fig.1(a)) Since there is no significant change on the SCR region or type inversion occurring on the surface, the surface leakage current is eliminated Therefore, when VG ⬉ Vsat, the dark current density remains unchanged with respect to the gate bias and is identical to the bulk dark current

When the gate bias is larger than Vsat but smaller than the threshold voltage of inversion (Vsat< VG< VT), fixed charges at the SiO2-T2SL interface or within the SiO2 passiva-tion layer attract electrons, leaving behind an SCR of uncom-pensated ionized acceptor ions near the p-region surface

a)

Email: razeghi@eecs.northwestern.edu.

(Fig.1(b)) The surface generation-recombination (G-R)

cur-rent associated with the surface depletion region is related to

the surface field-induced depletion width, xsd When the mesa

surface is depleted, the surface field-induced depletion width

and hence surface G-R current increases with increasing gate

bias,23in which results in a rising reverse dark current

As gate bias increases above VT, the field-induced

depletion width reaches its maximum value (xsd max) and

type inversion occurs on the surface (Fig 1(c)) Therefore,

there is no further increase in the surface G-R current,23and

the dark current density remains unchanged with respect to

gate bias However, since the surface field-induced depletion

width is very small near the interface between the Pþ- and

p-region, under reverse diode operation bias, electrons in the

p- and Pþ-region can tunnel through the thin field-induced

depletion region and get into the inversion channel

There-fore, dark current density increases dramatically as diode

reverse operation bias increases, which is usually observed

in the long-wavelength infrared SiO2 passivated T2SL

photodetectors

In order to reduce the high saturated gate bias, two

possi-ble options are to apply: a high-k dielectric material or reduce

the dielectric layer thickness The gate bias can be expressed

by the parallel capacitance formula

where V is the saturated gate bias, e0is the permittivity in

vacuum, e is the SiO2dielectric constant, assumed to be 3.9,

r is the SiO2surface charge density, and d is the dielectric

thickness In this paper, we demonstrate high performance

long-wavelength infrared gated diodes with low gate bias

achieved by reducing the SiO2passivation layer thickness

In this work, a long-wavelength infrared T2SL Pþ-p-M-Nþ

heterojunction24was grown on a GaSb substrate with

molec-ular beam epitaxy After a 1.5 lm n-doped InAsSb buffer

layer, the device structure consisted of a 0.5 lm thick

Nþ-contact (nþ 1018cm3), 0.5 lm thick lightly n-doped

M-barrier,22 3 lm thick p-region (p 6  1016cm3),

0.5 lm thick Pþ-contact (pþ 1018cm3) and capped with a

Pþ-type InAs capping layer The superlattice period in the

p-region and Pþ-contact region consisted of 13/7 monolayers (MLs) of InAs/GaSb and 7/11 MLs of InAs/GaSb respec-tively Both regions were doped with beryllium The M-barrier and Nþ-contact superlattice periods consisted of 18/ 3/5/3 MLs of InAs/GaSb/AlSb/GaSb in one period, and both were doped with silicon The material was processed into six dies of single element diodes with diameters ranging from

100 to 400 lm with the same processing procedures as those reported in Ref 25 One die (sample A) was left unpassi-vated for reference and the unpassiunpassi-vated diodes (UPD) were used for electrical and optical characterization The other dies (sample B, C, D, E, and F) were passivated with 10 nm,

20 nm, 60 nm, 300 nm, and 600 nm thick SiO2 dielectric layer, respectively, using plasma-enhanced chemical vapor deposition (PECVD) An additional metal gate was depos-ited on the mesa sidewall of half diodes in those samples, so that they contain both GD and ungated diode (UGD) In the final step, the top contacts were opened using electron cyclo-tron resonance etching All samples were wire-bonded onto 68-pin leadless ceramic chip carriers, loaded into a cryostat, and cooled down to 77 K for characterization

Average I-V characteristics of five different sizes UPDs and UGDs are compared in Figure 2(a) All UGDs exhibit much leakier I-V characteristics than the UPDs because SiO2 passivation layers cause severe surface leakage for diodes in the long-wavelength infrared region UGDs with different SiO2passivation layer thicknesses have the same I-V charac-teristics, which indicate that, above some threshold thick-ness, fixed charges in the SiO2 do not have any effect on surface leakage Therefore, the fixed charges that cause band bending and surface leakage are mainly situated in the SiO2 -T2SL interface or the very thin SiO2 layer near the mesa sidewall Additional surface treatment to reduce the amount

of fixed charges is required before or after SiO2passivation

As shown in Figure2(b), the average dark current den-sities of GDs at VG¼ Vsat are orders of magnitude lower than that of UPDs, and the corresponding saturated gate bias are shown in Table I At100mV operation bias, at which the quantum efficiency (QE) is saturated, GDs exhibit one order of magnitude lower dark current than the UPDs, and more than one order of magnitude lower than the UGDs The

FIG 1 The schematic diagram of gated diode p-region surface condition at (a) accumulation, (b) depletion, and (c) inversion situation.

differential resistance area product at100 mV (RA100 mV)

of UPDs, UGDs, and GDs are reported in TableI The best

RA100 mV value of the GD achieved is 3071 X cm2, which

is 10 times higher than that of UPDs and 500 times

higher than that of UGDs After eliminating the surface

leak-age, the bulk dark current characteristic should be achieved,

and all GDs saturate at the same level, which is confirmed in

Figure2(b) The best saturated gate bias achieved is4.5 V

with 10 nm SiO2 passivation layer The linear correlation

between the saturated gate bias and the thickness of SiO2

(Fig-ure2(c)) follows the Eq (1), and the calculated SiO2 fixed

charged density is estimated to be about 4 1012cm2

The reverse dark current densities at 0.5 V and 1 V

diode operation biases are shown in Figure3as a function of

the applied gate bias According to our previous discussion,

the surface depletion width is believed to be at its maximum

under no applied gate bias and in the case where a positive

gate bias is applied In this scenario, the surface G-R current

will not change with gate bias; the mesa surface is inverted,

and a leakage channel is formed causing the high tunneling

leakage current observed

As gate bias becomes increasingly negative towards

7 V, the reverse dark current density decreases and the

mesa surface begins to be less depleted Since the diode

operation bias is fixed for each individual curve, the decrease

in dark current density seen does not come from bulk current

but a decreasing surface G-R current, which is the dominant dark current mechanism in this region When comparing the dark current between the two curves at the same gate bias, the surface G-R current at 1 V is found to be larger than that at 0.5 V, which is caused by an increasing surface depletion width (xsd) at higher diode reverse operation bias according to Ref.23, which is why the surface G-R current

at1 V is larger than that at 0.5 V

When gate bias is beyond Vsat(9.5 V⬉ VG⬉ 7 V), the reverse dark current density is independent of gate bias

FIG 2 (a) Electrical performance comparison between UPD and UGD with different SiO 2 passivation layer thickness (b) Electrical performance comparison between UPD and GD at saturated gate bias (c) Correlation between gate bias and SiO 2 passivation layer thickness.

TABLE I The differential-resistance-area product at 100 mV (RA 100 mV ), saturated gate bias, and peak detectivity (D*) of UPD, UGD, and GD at satura-tion bias.

D* (Jones) 2.5  10 11 (0 mV) 5.9  10 11 (70 mV) 7.0  10 11 (70 mV) 6.4  10 11 (70 mV) 5.8  10 11 (70 mV) 6.3  10 11 (70 mV) 1.7  10 11 (0 mV)

FIG 3 The correlation between the reverse dark current density and gate bias of sample C at different diode operation bias.

but is found to be dependent on the operation bias, which

means that the surface leakage current is eliminated and the

mesa surface is under accumulation Since the reverse dark

current density at1 V operation bias is one order of

magni-tude higher than that at0.5 volt, this means that the

domi-nant dark current mechanism in this region is the bulk

tunneling current rather than bulk G-R or diffusion current

The peak responsivity (k¼ 9.9 lm) and QE at peak

responsivity of UPDs are shown in Figures4(a) and 4(b) At

100 mV operation bias, the diode exhibits a saturated peak

responsivity of 4.0 A/W and saturated quantum efficiency of

50% at peak responsivity At 77 K, the 50% cut-off wavelength

and 100% cut-off wavelength of this detector are 11 lm and

13 lm, respectively The small bias dependence seen in the

optical behavior is from the doping level of M-barrier, which

causes a small conduction band misalignment between the

M-structured superlattice and the p-region.24,26

Since after passivation, optical measurements are not

possible, the detectivity values for UGDs and GDs are

calcu-lated based on their electrical measurements and the

respon-sivity from UPDs Figure 5 shows detectivities of UPDs,

UGDs, and GDs at different operational bias The detectiv-ities of UPD and UGD decrease very quickly with operation bias because of the noise caused by surface leakage Between 0 and500 mV of operation bias, the detectivity of UPD drops 10 times, and that of UGD drops by 32 times In contrast, in the whole operation range, detectivities of GDs

in all samples are higher than those of UPDs and UGDs, and detectivities of GDs all stay in the level above 2 1011 Jones This is very important for the practical use of this technology because the detector can operate under a much lower bias, reducing noise and power consumption, and has

a much wider operation range for FPA applications Sample

C has the highest detectivities and reaches its maximum detectivity 7 1011 Jones at 70 mV, which is 2.8 times higher than the maximum detectivity of UPD and 4 times higher than the maximum value of UGD Detectivities of other samples are reported in TableI

In summary, we demonstrated the gating technique in long-wavelength infrared photodetectors, and effective sup-pression of the surface leakage generated by SiO2 passiva-tion layer By minimizing the SiO2 passivation layer thickness, we confirmed that the origin of surface leakage is from fixed charges at the SiO2/T2SL interface or within the very thin SiO2passivation layer near the surface, and that a thicker SiO2passivation layer will not further affect surface leakage It means additional surface treatment is required before or after passivation to remove the fixed charges Also, the saturated gate bias of gated diodes has been reduced to

4.5 V, with RA100 mV of 3071 X cm2, and detectivity of

7 1011 Jones at 77 K, which implies that the gating tech-nique has great potential for FPA application Moreover, gated diode offers a much wider operation range than those

of UGDs and UPDs and preserves detectivity above 2 1011 Jones, which is very important in FPA applications Most importantly, the gating technique reveals the evolution of surface leakage with gate bias, which facilitates further research work on surface leakage suppression

The authors acknowledge the support, interest, and encouragement of Dr Fenner Milton, Dr Meimei Tidrow, and Dr Joseph Pellegrino from the U.S Army Night Vision Laboratory and Dr William Clark from U.S Army Research Office

1 G A Sai-Halasz, R Tsu, and L Esaki, Appl Phys Lett 30, 651 (1977).

2 M Razeghi, “Focal plane arrays in type-II superlattice,” U.S Patent No.

6864552 (8 March 2005).

3

A M Hoang, G Chen, A Haddadi, S Abdollahi Pour, and M Razeghi, Appl Phys Lett 100, 211101 (2012).

4 B.-M Nguyen, G Chen, A M Hoang, S A Pour, S Bogdanov, and

M Razeghi, Appl Phys Lett 99, 033501 (2011).

5

S A Pour, E K.-w Huang, G Chen, A Haddadi, B.-M Nguyen, and

M Razeghi, Appl Phys Lett 98, 143501 (2011).

6 B M Nguyen, G Chen, M A Hoang, and M Razeghi, IEEE J Quantum Electron 47, 686 (2011).

7

B M Nguyen, S Bogdanov, S Abdollahi Pour, and M Razeghi, Appl Phys Lett 95, 183502 (2009).

8 M Walther, J Schmitz, R Rehm, S Kopta, F Fuchs, J Fleibner, W Cab-anskib, and J Zieglerb, J Cryst Growth 278, 156 (2005).

9

V Swaminathan, J W Little, and R L Tober, Proc SPIE 6127, 61270R (2006).

10 E K Huang, A Haddadi, G Chen, B M Nguyen, M A Hoang,

R McClintock, M Stegall, and M Razeghi, OSA Opt Lett 36, 2560 (2011).

FIG 4 (a) The peak responsivity (at 9.9 lm) and (b) QE at peak

responsiv-ity at different diode operation bias at 77 K.

FIG 5 Detectivity of UPDs, UGDs, and GDs at saturation gate bias

calcu-lated at different operation bias The calculation of detectivity bases on the

inset equation, with Ri is the responsivity, J is the dark current density, q is

the charge of electron, k b is the Boltzmann constant, T is the temperature,

and RA is the differential resistance-area product.

11 G Chen, B.-M Nguyen, A M Hoang, E K Huang, S R Darvish, and

M Razeghi, Appl Phys Lett 99, 183503 (2011).

12

P.-Y Delaunay, A Hood, B.-M Nguyen, D Hoffman, and M Razeghi,

Appl Phys Lett 91, 091112 (2007).

13 S Bogdanov, B M Nguyen, A M Hoang, and M Razeghi, Appl Phys.

Lett 98, 183501 (2011).

14

E K.-w Huang, D Hoffman, B.-M Nguyen, P.-Y Delaunay, and

M Razeghi, Appl Phys Lett 94, 053506 (2009).

15

A Gin, Y Wei, J Bae, A Hood, J Nah, and M Razeghi, Thin Solid Films

447–448, 489 (2004).

16

A Hood, P Y Delaunay, D Hoffman, B.-M Nguyen, Y Wei, and

M Razeghi, Appl Plys Lett 90, 233513 (2007).

17 E Plis, J B Rodriguez, S J Lee, and S Krishna, Electron Lett 42, 1248

(2006).

18

A Gin, Y Wei, A Hood, A Bajowala, V Yazdanpanah, M Razeghi, and

M Z Tidrow, Appl Phys Lett 84, 2037 (2004).

19 R Rehm, M Walther, F Fuchs, J Schmitz, and J Fleissner, Appl Phys Lett 86, 173501 (2005).

20

F Szmulowicz and G J Brown, Infrared Phys Technol 53, 305–307 (2010).

21 B M Nguyen, D Hoffman, E K Huang, P Y Delaunay, and M Raze-ghi, Appl Phys Lett 93, 123502 (2008).

22 B.-M Nguyen, M Razeghi, V Nathan, and G J Brown, Proc SPIE 6479, 64790S (2007).

23

A S Grove and D J Fitzgerald, Solid-State Electron 9, 783 (1966) 24

B M Nguyen, D Hoffman, P Y Delaunay, E K Huang, M Razeghi, and J Pellegrino, Appl Phys Lett 93, 163502 (2008).

25 A Hood, D Hoffman, B.-M Nguyen, P.-Y Delaunay, E Michel, and

M Razeghi, Appl Phys Lett 89, 093506 (2006).

26

D Hoffman, B M Nguyen, E K Huang, P Y Delaunay, M Raze-ghi, M Z Tidrow, and J Pellegrino, Appl Phys Lett 93, 031107 (2008).