Metasurface integrated uncooled silicon germanium oxide microbolometers for long wavelength infrared detection

Abstract

[EMBARGOED UNTIL 6/1/2023] This report presents the design, fabrication, characterization, and noise optimization of metasurface enabled uncooled infrared (IR) microbolometers based on silicon germanium oxide (Si0.344Ge0.602O0.054), (Si0.425Ge0.512 O0.063). The metasurface enabled fabrication of the microbolometer with the supporting arms placed underneath the pixel without disrupting the Fabry-Perot resonant cavity typically used in conventional microbolometers. This allowed fabrication of the microbolometer with longer supporting arms without sacrificing the fill factor and thus increased the voltage responsivity. The longer support arms reduced the thermal conductance between the microbolometer pixel and the underlying substrate. The metasurface also allowed the IR absorption to be engineered to achieve broadband or narrowband response. The measured voltage responsivity, detectivity, and thermal response time were 1.52x104 V/W, 2.5x107 cm Hz1/2/W and 2.01 m seconds to filtered blackbody IR between 2-14 [mu]m. The voltage noise power spectral density (PSD) of the fabricated devices were reduced by annealing the devices in vacuum. In addition, the measured TCR, thermal conductance Gth, and absorptivity were 2.5 percent/K, 4.44x10-8 W/K, and 39 percent, respectively. The voltage noise power spectral density (PSD) of the fabricated devices with a different composition (SiGeO) was optimized by annealing the devices in vacuum and in forming gases. The results demonstrated annealing in vacuum reduced the voltage noise level and shifted the corner frequency to a lower frequency than that of annealing in forming gases. The lowest measured voltage noise in vacuum and in forming gases at the corner frequencies were 1.723 x 10-15 V2/Hz at 17 Hz and 1.91x10-14 V2/Hz at 30 Hz, respectively. The voltage noise before annealing at the same bias current and corner frequency was 5.62 x 10-10 V2/Hz for annealing in vacuum, 4.33 x 10-10 V2/Hz for annealing in forming gases. This is 4.33 x 105 times and 1.27 x 104 times reduction in noise, respectively. The results also demonstrated that annealing in vacuum provided much lower voltage noise reduction than annealing in forming gases. In addition, annealing in vacuum and in forming gases has shifted the corner frequency from 85 Hz to 17 Hz and from 80 Hz to 30 Hz, respectively. The results demonstrated that the voltage noise at 17 Hz after annealing in vacuum is 45.2 times lower than that of annealing in forming gases. The corresponding Hooge's parameters ?, [beta], and Kf for the devices before and after annealing in vacuum were 1.25, 2.23, 3.26x10-10, and 1.02, 2.07, 2.375x10-14 and in forming gases were 1.26, 2.24, 2.435x10-11, and 1.06, 2.09, and 7.078x10-14, respectively. The value of ? is close to 1 after annealing which indicates that the 1/f-noise is dominant at low frequencies. The corresponding 1/f-noise coefficients Kf for the devices that were annealed in vacuum and forming gases were changed from 3.26x10-10 to 2.375x10-14 and from 2.435x10-11 to 7.078x10-14 after 4 hours of annealing at 300 degreesC. The value of Kf started to decrease after each annealing time interval. The decrease in Kf is attributed to the reduction of 1/f-noise. The [beta] value was close to 2 when the device was annealed for 4 hours. This indicates that the voltage noise increased linearly with increasing the bias current. The spectral responsivity and detectivity for the fabricated devices were measured in vacuum and demonstrated tunablility with narrowband between 7 to 16 [mu]m by changing the metasurface disk diameter and periodicity.