1. Introduction

There is a strong history of delivering innovative and extensive applications of IR imaging in industrial, military, and commercial fields, including: night vision, remote monitoring of facilities and equipment, medical imaging, etc. The research of developing the IR sensors has been versatile, and recently, detectors based on the optical readout method have drawn more and more attention due to their low cost, simple fabrication, and high performance^{[1, 2]}. Following the development of microelectromechanical system (MEMS) technology, several research groups have done much work on optical readable uncooled IR-FPAs and demonstrated varying degrees of success^[3-13]. It is worthy to note that fundamental models of microcantilever FPAs predict that they can match or even exceed the performance of other, more conventional, uncooled IR FPAs^[2]. Optical readable uncooled IR FPAs based on bimaterial microcantilevers can be generally divided into FPAs with Si substrate based on a sacrificial layer release process and substrate-free FPAs based on bulk micromachining technology.

For FPAs with Si substrate, only about 50% of the incident IR energy can reach the cantilever and the efficiency of the infrared absorption is greatly reduced^{[1, 5, 14]}. Moreover, the fabrication of the FPA is very complicated due to the difficulty of releasing the sacrificial layer.

The substrate-free FPA patches up the drawbacks of FPAs with Si substrate and, moreover, it can even image while the FPA is at atmospheric pressure due to the improved thermal isolation realized by eliminating the 2 $\mu$m gap between the cantilever and the substrate, which gets rid of the thermal conductivity of air molecules^{[1, 15, 18]}. However, some negative issues have followed.

(1) The temperature of the substrate-free FPA could not survive to be constant and uniform with the utility of a thermoelectric cooler (TEC) at different environmental temperatures.

(2) The thermal cross-talk phenomenon is obvious.

(3) The minimum resolvable temperature difference (MRTD) is reduced.

For the substrate-free optical readout FPA, the temperature of the cantilever will increase when receiving the infrared radiation. The energy could not be effectively transferred through the supporting frame of the substrate-free FPA to the ambient Si substrate but to the neighboring pixels, so the pixels are not temperature-independent, and can not be isolated from each other^[16-19]. Therefore, when one cantilever absorbs infrared radiation and is deformed, the adjacent cantilever will also deformed. Figure 1 shows the thermal imaging of a traditional substrate-free FPA; we can see that the edge of the finger is blurred (the so-called flipper phenomenon) due to the dependently functional pixels, which affects the thermal imaging quality.

2. Thermal cross-talk analysis

The temperature distribution of the substrate-free FPA both with and without an HSS was simulated using the finite-element method (FEM), as shown in Fig. 2. For the FPA without an HSS, when applying a heat flux on a pixel of the FPA chip, the temperature of the adjacent pixel without heat load also rises after heat balance, as shown in Fig. 2(a), which results in the undesirable deflection of the adjacent cantilevers. The energy of the pixel generated by absorbing heat flux can not immediately transfer to the Si substrate due to the low heat capacity and thermal conductance of the supporting frame of the FPA without an HSS, but transfer through the supporting frame to the adjacent pixel, which results in the deformation of the cantilever of the adjacent pixel. In other words, the pixel is not temperature-independent, and each pixel can not function independently, which is the so-called thermal cross-talk effect. While the supporting frame of the FPA with an HSS is just the same as the Si substrate of the substrate FPA due to its large thermal conductance and heat capacity. So the pixel can function independently, thermal cross-talk effect is eliminated effectively, as shown in Fig. 2(b).

3. Design and fabrication

In order to obtain an FPA with the advantages of the substrate-free and that with a substrate, but avoid the disadvantages of the two simultaneously, we proposed a substrate-free optical readout FPA with an HSS by increasing the conductivity of the supporting frame which acts as the Si base of the FPA with substrate. Energy could then be effectively transferred through the supporting frame of the substrate-free FPA, so the thermal cross-talk effect could be eliminated. As shown in Fig. 3, an FPA of 240 $\times$ 240 pixels array with a 49.5 $\times$ 49.5 $\mu$m² pitch using low stress SiN_x as the structural material and Au as reflector and a second layer in the bi-material region was fabricated successfully. The HSS is fabricated by electroplating copper on the frame of substrate-free FPA. We chose copper as the heat sink material because its heat capacity and thermal conductivity can reach to 390 J/kg$\cdot$K and 401 W/m$\cdot$K, respectively.

Firstly, the substrate-free FPA chip without HSS is prepared, the fabricated processes of the chip have been given in Refs. [14, 19] using Au and low stress SiN_x^[20]. Secondly, the HSS is fabricated on the frame of the FPA chip. Figure 4 shows the basic processes of fabricating the HSS. (1) The electroplating seed layer. Ti/Au was deposited on the frame of the traditional FPA, the thickness of Au is 1500 Å, and 100 Å Ti is deposited as adhesion layer between Au and SiN_x frame, as shown in Fig. 4(a). (2) The second step lithography. Spin thick photoresist and then perform lithography, making the pixel sensitive area covered by the photoresist, then the photoresist on the supporting frame is developed to form a high aspect ratio trench, until the electroplating seed layer is exposed, as shown in Fig. 4(b). (3) Cu electroplating. The FPA chip is immersed in Å solution to electroplate copper to about 8 $\mu$m. The photoresist is removed after electroplating, as shown in Fig. 4(c). (4) Corrosion and release. The device is produced until all the Si substrate is corrupted off and the structure is released using wet etching, shown in Fig. 4(d). The scanning electron microscope (SEM) image of the fabricated FPA with the HSS is shown in Fig. 5.

4. Result and discussion

The FPA chip with an HSS was sealed in a vacuum chamber after fabrication. Commercial FLIR thermal imaging camera was used to measure the temperature distribution of the chip, Figure 6(a) shows the whole profile of the FPA chip with a 240 $\times$ 240 pixel array, Figures 6(b) and 6(c) give the temperature distribution of substrate-free FPA chip without and with a heat sink, respectively. Measurement shows that the temperature difference between the edge and middle of FPA without a heat sink is up to 6 ℃ under 11 ℃ temperature difference between the TEC and the environmental temperature, while the temperature uniformity of substrate-free FPA with an HSS has been greatly improved. The temperature difference between edge and middle of FPA with HSS is below 1 ℃ at the same condition as the above, and it could be better with a thicker heat sink structure. It is noted that the substrate-free FPA with an HSS can be modulated by TEC, namely, the substrate-free FPA with an HSS is temperature-independent, and each pixel can function independently, thermal cross-talk effect existed in substrate-free FPA has be eliminated effectively.

We also demonstrate the use as the IR imager sensor and obtain an IR-image of a hand at room temperature. The principle of the IR imaging system is shown in Fig. 7^[11]. The IR radiation (8-14 $\mu$m) from the target is focused onto the microcantilever array of the FPA chip by using an IR lens. The incident IR energy thus increases the temperature of the microcantilevers and, therefore, leads to the bending deflections in the bimaterial regions due to the mismatch of coefficient of thermal expansion (CTE), as shown in the top right corner insert in Fig. 7. Simultaneously, the microcantilever array is illuminated by a visible parallel beam on the opposite (a light-emitting diode (LED) with 520-540 nm wavelength is used in this work). The deflection of individual reflector is thus proportionally present as a corresponding translation of the diffraction pattern on the back focal plane of the Fourier lens. The required angle-to-intensity conversion is finally achieved by the knife-edge filter, which is located on the back focal plane. The principle of the knife-edge filter is shown in the lower left quarter insert in Fig. 7. As a result, the IR target can be imaged readily by a CCD camera.

In order to visually show the performance of the FPA with an HSS, the thermal imaging of hot substances is performed. Figure 8(a) shows the picture of the imaging object (sheet metal), the sheet metal is heated to 100 ℃ by a hot plate, and the thermal imaging of substrate-free FPA with an HSS using the proposed optical readout system is shown in Fig. 8(b). It is noted that the edge of the thermal imaging is vivid, which demonstrates that the flipper phenomenon caused by thermal cross-talk effect is eliminated. In addition, it is worthwhile pointing out that the light dark difference exists in the thermal imaging due to the different emissivity caused by the colour of the imaging object.

Response sensitivity and noise equivalent temperature difference (NETD) of the system have also been measured experimentally. The response sensitivity of the substrate-free FPA with an HSS measured experimentally is 13.6 grey/K(F/# $=$ 0.8), and the experimental NETD of the IR imaging system can be expressed as:

${\text{NETD}} = {I_{{\text{noise}}}}/(\Delta I/\Delta {T_{\text{s}}}),$

(1)

where ${I_{{\text{noise}}}}$ is the average gray level of the total noise of system, $\Delta {\text{ }}$I/$\Delta {\text{ }}$T is the average response sensitivity of the system, $\Delta {\text{ }}$I is the average response gray level when the IR object's temperature changes $\Delta {\text{ }}$T. ${I_{{\text{noise}}}}$ can be expressed as:

$I_{\rm noise} =\sqrt {(I_{\rm 1noise} ^2+I_{\rm 2noise} ^2+\cdots +I_{\rm Nnoise} ^2)/N},$

(2)

where ${I_{{\text{Nnoise}}}}$ is the gray level of the noise of $N$th thermal image without IR objects, during this experiment, we have taken more than 40 pictures for the average noise value. The experimental result showed that the average ${I_{{\text{noise}}}}$ was 8 gray and $\Delta {\text{ }}$I/$\Delta {\text{ }}$T is 13.6 grey/K. Thus, the NETD of the system is 588 mK. It is noted that although the response was reduced due to the increase of the conductivity of the supporting frame, the substrate-free optical readout FPA with an HSS could be used as a sensor with more reliable performance, and the response could also be further improved to reach excellent performance by optimizing the structure, which has broad application prospects in the field of IR technology.

5. Conclusion

In this paper, a substrate-free optical readout focal plane array(FPA)with an HSS is demonstrated. The HSS is fabricated by electroplating copper on the frame of substrate-free FPA, which effectively overcomes the drawbacks of the thermal cross-talk effect that exists in substrate-free FPA. The response and NETD of the imaging system is 13.6 grey/K, and 588 mK, respectively, and further optimization of the FPA with an HSS is underway.