Thin Film IR Detectors

5 downloads 0 Views 212KB Size Report
thermal time constant of the detector, the sensing element was fabricated on a thin ... thickness on the response of pyroelectric IR detectors were studied by ...
04R07036 Total pages 5 Japanese Journal of Applied Physics Vol. 44, No. 2, 2005, pp. 0000–0000 #2005 The Japan Society of Applied Physics

Effects of Membrane Thickness on the Pyroelectric Properties of LiTaO3 Thin Film IR Detectors Chao-Chin C HAN, Ming-Cheng K AO1 and Ying-Chung C HEN1  Department of Biochemical Engineering, Kao Yuan Institute of Technology, Kaohsiung, Taiwan 1 Department of Electrical Engineering, National Sun Yat-Sen University, Kaohsiung, Taiwan (Received July 10, 2004; accepted October 21, 2004; published xxxx yy, zzzz)

High-performance pyroelectric infrared (IR) detectors have been fabricated using lithium tantalite (LiTaO3 ) thin films deposited on Pt(111)/Ti/SiO2 /Si(100) substrates by the diol-based sol–gel method. The thermal isolation of detecting elements was achieved by anisotropic wet etching of the back of the silicon substrate. In order to reduce the thermal mass and thermal time constant of the detector, the sensing element was fabricated on a thin membrane. The effects of membrane thickness on the response of pyroelectric IR detectors were studied by changing the membrane thickness (20–350 mm). As the membrane thickness decreased, the voltage responsivity (Rv ) increased from 4300 up to 8398 V/W, and the specific detectivity (D ) also increased from 1:2  108 to 2:7  108 cmHz1=2 /W. Experimental results reveal that the thermally isolated detectors with membrane thickness of 20 mm exhibit excellent sensitivity. [DOI: 10.1143/JJAP.44.dummy] KEYWORDS: LiTaO3 infrared detectors, thin membrane, voltage responsivity, back etching.

1.

JJA

fabricated.13,14) The influence of membrane thickness on the pyroelectric properties of thin film detectors was investigated, and the voltage responsivity (Rv ), noise voltage (Vn ) and specific detectivity (D ) of detectors were measured using a dynamic analysis system. Although attempts were made to optimize the process parameters, the major efforts in the present study were directed toward improving the thermal isolation from the substrate to obtain high-performance pyroelectric IR detectors.

Introduction

PP RO OF S

The pyroelectric detectors have been widely used for infrared (IR) detection applications based on the pyroelectric effect, i.e., as the detector is exposed to IR light, the pyroelectric material absorbs radiation and its temperature varies, which reduces the polarization (P) and enhances the photocurrent of the detector. The pyroelectric coefficient p is an important parameter which affects the rate of variation of spontaneous polarization with respect to temperature. For the application in pyroelectric imaging arrays, the material should have a large pyroelectric coefficient, low dielectric constant and dielectric loss, and a low heat capacitance.1–3) The most commonly used materials for pyroelectric application include triglycene sulphate (TGS), PbZrTiO3 (PZT), PbTiO3 and LiTaO3 .4–7) Among these materials, the LiTaO3 possesses some excellent characteristics, i.e., large pyroelectric coefficient, small dielectric constant and high Curie temperature,8,9) and is suitable for IR detection applications. In the previous paper, we reported the preparation and properties of LiTaO3 thin films crystallized by rapid thermal annealing (RTA) for application to IR detectors.10) However, some shortcomings were observed for the obtained LiTaO3 thin film IR detectors due to the thermal loss from the pyroelectric film to the substrate. Traditionally, a pyroelectric IR sensor with LiTaO3 film is attached directly to the Si substrate. Because the Si substrate has high thermal conductivity and heat capacity compared with the thin composite film, the thermal energy absorbed in the pyroelectric film will be lost partly to the substrate, thus reducing the pyroelectric current of the detector.11) Consequently, the performance of the pyroelectric IR detector can be further improved by thermal isolation from its substrate through the lowering of heat capacity by etching the back of the Si substrate, so that the induced pyroelectric current can respond fully to the incident IR light.12) In this study, LiTaO3 thin films were prepared on Pt(111)/ Ti/SiO2 /Si(100) substrates by the diol-based sol–gel method, and IR detectors with various membrane thicknesses (d), controlled by the anisotropic etching technique, were

2.

Experimental

The flow diagram of the fabrication process for a LiTaO3 thin film IR detector on a membrane was shown in Fig. 1. The fabrication started with the depositions of Pt(111)/Ti layers and Si3 N4 /SiO2 layers on the front and back of the Si substrate, respectively, as shown in Fig. 1(a). In the second step shown in Fig. 1(b), the LiTaO3 thin films were obtained using a diol-based sol–gel method, in which lithium 2,4pentanedionate, LiC5 H7 O2 (Alfa, 99.5%+ purity) and tantalum isopropoxide, Ta[OCH(CH3 )2 ]5 (Alfa, 99.9%+ purity) were used as precursors and 1,3-propanediol, HO(CH3 )2 OH (Fluka, 99.0%+ purity) was used as solvent. The details of the diol-based sol–gel technique, including the process of synthesizing LiTaO3 sol and RTA-derived film, have been reported previously.10) In the RTA process, the gel films fabricated in each coating step were pyrolyzed in oxygen atmosphere at 300 C for 2 min by rapid thermal processing before final annealing. After applying the multicoating, LiTaO3 thin films were annealed at 700 C for 2 min at a heating rate of 1800 C/min by rapid thermal processing in oxygen atmosphere. The LiTaO3 thin film obtained exhibits the pyroelectric coefficient () of 3:9  108 C/ cm2 K, and the figures of merit (Fv and Fm ) of 3:8  1010 Ccm/J and 3:49  108 Ccm/J. Next, alumina (Al) was evaporated onto the LiTaO3 film as a top contact electrode. Finally, silver (Ag) black, a heat absorption material, was evaporated onto the top contact electrode to assist the absorption of incident IR radiation. The third step shown in Fig. 1(c) was accomplished by patterning the back Si3 N4 and SiO2 layers by reactive ion etching (RIE) and using a buffer oxidation etchant (BOE) solution, respectively. The anisotropic etching step in



Corresponding author. E-mail address: [email protected] [1]

[2]

Jpn. J. Appl. Phys., Vol. 44, No. 2 (2005)

C.-C. C HAN et al.

(a)

(a) (b)

(c)

(d)

JJA

PP RO OF S (b)

Fig. 2. (a) Cross section of etching fixture used to protect front devices during etching in hot KOH, and (b) top view of a detector.

Fig. 1. Fabrication flow diagram of the LiTaO3 thin film IR detector: (a) Pt/Ti bottom electrode deposition, (b) Ag-black/Al/LiTaO3 deposition, (c) back etch-window patterning, and (d) back bulk-silicon etching.

Fig. 1(d) defined the membrane thickness (d). For silicon etching, a 44 wt% KOH solution was used, and an etching rate of 1.4 mm/min for the (100) silicon plane was realized at 80 C. During the etching process, it is necessary to protect the front of the device from the isotropic or anisotropic etchant. The wafer was placed in an etching fixture made of Teflon with O-ring sealing around the wafer edge [Fig. 2(a)]. The wafer is fixed between the Teflon fixture and O-ring that are carefully aligned to avoid mechanical stress in the wafer. Thus, the front of the wafer was protected from the hot KOH etchant by the O-ring and Teflon fixture. The top view of a completed detector is shown in Fig. 2(b). Figure 3 shows the bottom view of a single etched rectangular structure and the cross section of the detector. Upon close inspection, the etch quality of the etched rectangle is found to be very uniform and smooth on the surface and edges [Fig. 3(a)]. The side wall profiles of the etched cross section show an angle of 54.7 from the horizontal surface, as shown in Fig. 3(b). To measure the pyroelectric properties of detectors, various light sources, such as a He–Ne laser (wavelength  ¼ 0:633 mm) and a blackbody radiation furnace (wavelength  ¼ 2:3{3 mm), were focused using a concave lens, mechanically chopped at frequencies from 5 Hz to 1 kHz, and then directly irradiated onto the surface of the sensing electrode. The experimental setup is shown in Fig. 4. The incident power of the IR radiation was measured using a

radiometer (Model R-752, Universal Abrasives, Stanfford, U.K.). Alignment between the light source and the device was achieved by viewing through the objective lens using a charge-coupled device. The pyroelectric voltage and current signals were measured using a lock-in amplifier (Model 7260, EG&G Princeton Applied Research, Princeton, NJ) and monitored by a digitizing oscilloscope (Model HP54502A, Hewlett-Packard), while the detector element was exposed to the incident chopped IR radiation. These measurements were performed at room temperature, in a shielded room. 3.

Results and Discussion

The fundamental performance parameter of a pyroelectric detector is the voltage responsivity (Rv ), defined as the ratio of output voltage induced by the pyroelectric effect to the incident radiant power. During pyroelectric measurement, the radiation energy is absorbed by the Ag-black layer, converted to thermal energy, and then conducted to the active LiTaO3 thin film as well as to the underlying substrate. Figure 5 shows the modulation frequency dependence of Rv for the LiTaO3 thin film IR detectors with various membrane thicknesses (d). Responsivity was constant at low modulation frequencies and decreased at high modulation frequencies. The voltage response of an IR device can be expressed as follows:15) Rv ¼

pAR! pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi G 1 þ ! 2 t 2 1 þ ! 2 e 2

ð1Þ

Here,  is the emissivity, p the pyroelectric coefficient, A the detector area, ! the angular modulation frequency, G the thermal conductance, t the thermal time constant (¼ H=G,

04R07036

Jpn. J. Appl. Phys., Vol. 44, No. 2 (2005)

C.-C. C HAN et al.

[3]

3

Rv (V/W)

10

d= 20 µm d= 50 µm d= 150 µm d= 250 µm d= 350 µm

2

10

10

100

1000

Frequency (Hz)

Fig. 5. The modulation-frequency-dependent Rv of the LiTaO3 thin film IR detectors.

(a)

JJA (b)

50

PP RO OF S Thermal time constant (ms)

40

30

20

10

0

0

Fig. 3. Photographs of back bulk etching observed from (a) back, and (b) cross section. (Bar ¼ 100 mm)

50

100

150

200

250

300

350

Membrane thickness (µm)

Fig. 6. Membrane thickness dependence of thermal time constant (t ) for the detectors.

Chopper

Concave mirror

Blackbody furnace

Detector

Lock-in amplifier (EG&G7260)

Computer

Plane mirror

Fig. 4. Schematic diagram of the experimental setup for voltage responsivity measurements.

where H is the thermal capacity), and e (¼ RC) the electrical time constant. According to eq. (1), Rv is almost constant in the low-frequency region. The Rv value decreases at high frequencies because Rv is inversely proportional to the frequency at a modulation frequency above t 1 . The present results, shown in Fig. 5, are consistent with eq. (1). This tendency is consistent with that reported by Kohli et al.16) As shown in Fig. 5, Rv decreased with increasing thickness of the membrane, due to the increase of thermal loss to the substrate. Figure 6 shows the thermal time constant, t , varied with

the membrane thickness. The thermal time constant is estimated from the measurement of rising and falling time for the observed output waveform. The results in Fig. 6 show that t decreases as the membrane thickness decreases. This is attributed to the larger heat capacity through a membrane of greater thickness, which gives rise to a larger thermal time constant. The Rv values measured at 20 Hz, the frequency at which maximum voltage responsivity occurs, for detectors with various membrane thicknesses are shown in Fig. 7. Rv ðmaxÞ decreased from 8398 to 4300 V/W as the membrane thickness increased from 20 to 350 mm. With increasing membrane thickness, poorer thermal isolation and smaller temperature gradient of the sensing film would be obtained, resulting in a lower output voltage. The noise voltage (Vn ) of detectors with various membrane thicknesses was measured in a frequency range of 5 to 1000 Hz. The frequency dependence of Vn per unit of bandwidth (in units of V/Hz1=2 ) is shown in Fig. 8. It is seen that Vn varies nearly proportionately to f 1=2 , which means that Johnson noise may be dominant in these pyroelectric detectors.17) Figure 8 also shows that Vn increases as the membrane thickness decreases. The sensitivity of a pyroelectric device can be expressed in terms of its specific detectivity (D ¼ A1=2 Rv =Vn ). The frequency dependence of

04R07036

[4]

Jpn. J. Appl. Phys., Vol. 44, No. 2 (2005)

C.-C. C HAN et al.

3

9x10

8

10 3

3

7x10

1/2

D (cmHz /W)

3

6x10

7

10

d= 20 µm d= 50 µm d= 150 µm d= 250 µm d= 350 µm

*

Rv(max) (V/W)

8x10

3

5x10

6

10

3

4x10

3

3x10

0

50

100

150

200

250

300

350

10

100

Membrane thickness (µm)

Fig. 7. Dependence of the maximum voltage responsivity on membrane thickness at 20 Hz.

Fig. 9. The modulation-frequency-dependent specific detectivity of the LiTaO3 thin film IR detectors.

JJA 1E-4

1E-6

1E-7

d= 20 µm d= 50 µm d= 150 µm d= 250 µm d= 350 µm 10

8

3.0x10

PP RO OF S 8

2.5x10

1/2

D*(max) (cmHz /W)

1/2

Vn (V/Hz )

1E-5

1000

Frequency (Hz)

8

2.0x10

8

1.5x10

8

1.0x10

100

0

1000

50

100

150

200

250

300

350

Membrane thickness (µm)

Frequency (Hz)

Fig. 8. Frequency dependence of the noise voltage per unit bandwidth for the LiTaO3 thin film IR detectors.

Fig. 10. Dependence of the maximum specific detectivity on membrane thickness at 100 Hz.

D for the IR detectors is shown in Fig. 9. The results show that D increases with increasing frequency when f < 100 Hz, because Rv is almost saturated and Vn varies with f 1=2 in this frequency range. On the other hand, it is almost constant when f > 100 Hz. This phenomenon probably occurred because both Rv and Vn vary proportionately to f 1=2 , when f > 100 Hz. The D values measured at 100 Hz, at which the maximum D exists, for detectors with various membrane thicknesses are shown in Fig. 10. D ðmaxÞ significantly decreased with increasing thickness of the membrane. The detector with a membrane thickness of 20 mm exhibited the largest D ðmaxÞ value, 2:7  108 cmHz1=2 /W, at 100 Hz. This can be explained by the dependence of the voltage responsivity upon the membrane thickness. As the thickness of the membrane decreased, the output voltage increased, as shown in Fig. 7. It is assumed that the noise level has a slight dependence on membrane thickness. Thus, the detector with the smallest membrane thickness will exhibit the highest D .

350 mm to reduce the thermal mass and thermal time constant. Various membrane thicknesses of 20–350 mm were achieved by anisotropic wet etching of the back of the silicon substrate. With increasing thickness of the membrane, the voltage responsivity (Rv ) decreased from 8398 to 4300 V/W, and the specific detectivity (D ) also decreased from 2:7  108 to 1:2  108 cmHz1=2 /W. The corresponding results show that the detector with a membrane thickness of 20 mm exhibited excellent pyroelectric properties and thus was suitable for application to highly sensitive pyroelectric IR devices.

4.

Conclusion

In this study, LiTaO3 thin films were deposited on Pt(111)/Ti/SiO2 /Si(100) substrates by a diol-based sol–gel process and rapid thermal processing. An IR detector was fabricated on a thin membrane with the thickness of 20–

Acknowledgement This study is partly supported by the National Science Council, Taiwan, R.O.C., under contract no. NSC 92-2216E-110-008. 1) R. W. Whatmore, A. Patel, N. M. Sorrocks and F. W. Aiger: Ferroelectrics 104 (1990) 269. 2) R. Watton: Ferroelectrics 91 (1989) 87. 3) N. Yoshiike, K. Arita, K. Morinaka, K. Hashimoto, M. Kawaguri and A. Nishino: Sens. Actuators A 48 (1995) 73. 4) A. Shaulov and M. Simhony: J. Appl. Phys. 47 (1976) 1. 5) L. Weiguo, S. K. Jong and Z. Weiguang: Thin Solid Films 371 (2000) 254. 6) K. R. Han, H. J. Koo, M. J. Hong and C. S. Lim: J. Am. Ceram. Soc.

04R07036

Jpn. J. Appl. Phys., Vol. 44, No. 2 (2005)

C.-C. C HAN et al.

83 (2000) 971. 7) J. J. Ho, Y. K. Fang, K. H. Wu, W. T. Hsieh, C. W. Chu, C. R. Huang, M. S. Ju and C. P. Chang: IEEE Electron Device Lett. 19 (1998) 189. 8) D. K. Fork and G. B. Anderson: Appl. Phys. Lett. 63 (1993) 1331. 9) L. S. Hung, J. A. Agostinelli, J. M. Mir and L. R. Zheng: Appl. Phys. Lett. 62 (1993) 3071. 10) M. C. Kao, H. Z. Chen, C. M. Wang, Y. C. Chen and M. S. Lee: Appl. Phys. A 79 (2004) 103. 11) F. Y. Chen, Y. K. Fang, C. Y. Shu and J. R. Chen: IEEE Trans. Electron Devices 43 (1996) 2019. 12) K. No, C. G. Choi, D. S. Yoon, T. H. Sung, Y. C. Kim, I. S. Jeong and

[5]

W. J. Lee: Jpn. J. Appl. Phys. 35 (1996) 2731. 13) M. D. Levenson, N. S. Viswanathan and R. A. Simpson: IEEE Trans. Electron Devices 29 (1982) 18. 14) D. P. Kern, P. J. Coane, P. J. Houzego and T. H. P. Chang: Solid State Technol. 27 (1984) 127. 15) R. Takayama, Y. Tomita, K. Lijima and L. Ueda: J. Appl. Phys. 63 (1988) 5868. 16) M. Kohli, C. Wuethrich, K. Brooks, B. Willing, M. Forster, P. Muralt, N. Setter and P. Ryser: Sens. Actuators A 60 (1997) 147. 17) R. W. Whatmore: Rep. Prog. Phys. 49 (1986) 1335.

JJA

PP RO OF S

04R07036