Supplementary Information A flexible thin-film InGaAs photodiode ...
Supplementary Information A flexible thin-film InGaAs photodiode focal plane array Dejiu Fan1*, Kyusang Lee1*, Stephen R. Forrest1,2,3 1
Department of Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI, USA 2
3
Department of Physics, University of Michigan, Ann Arbor, MI, USA
Department of Materials Science and Engineering, University of Michigan, Ann Arbor, MI, USA *These authors contributed equally to this work.
10×10 photodiode array fabricated on Si and Kapton® substrate The 10×10 thin-film In0.53Ga0.47As p-i-n photodiode arrays are fabricated on both a 500 µm thick semi-insulating Si substrate and a 25 µm thick E-type Kapton® foil. The fabrication processes of 10×10 arrays are similar to that of the 8×100 photodiode array, although the photodiode dimensions are different. For the 10×10 arrays, the top ring contacts have 20 µm/30 µm inner/outer diameters that define the light detection area. Photodiode mesa diameters are 40 µm, with 100 µm pixel separations. The back side linear contacts are 5 µm wide to connect the photodiode rows, and top linear contact columns are also 5 µm wide. A MgF2 (81 nm)/TiO2 (113
1
nm) bilayer anti-reflection coating (ARC) is employed to achieve a maximum EQE at λ = 1550 nm. A scanning electron microscopic image of the array is shown in Fig. S1(a). An optical microscope image of the array with the same dimension fabricated on flexible Kapton® substrate is shown in Fig. S1(b). Maps of the dark current and external quantum efficiency (EQE) of all photodiodes on both Si and Kapton® substrates are shown in Fig. S2(a) and S2(b). The EQE and dark current are measured at -1 V bias with and without 100 µW 1550 nm laser illumination, respectively. The device fabrication yield is 99% for devices fabricated on Si, and 100% for devices on Kapton®. The single malfunctioning device on the Si substrate (indicated in white blank at (3,1) in Fig. S2 (a)) is shorted. Excluding the malfunctioning device, the average EQE at -1V bias and λ = 1550 nm is 88% (1.8 µm thick InGaAs active layer) and 73% (0.85 µm thick InGaAs active layer) for devices fabricated on Si and Kapton® substrates, respectively. The dark current, EQE, and yield of devices fabricated on these two substrates are almost identical over the entire array whether on Si or Kapton®.
Field of view (FOV) measurement The 8×100 p-i-n thin-film InGaAs photodiode array is rolled into a 1cm diameter cylinder as shown in Fig S3. A laser beam illuminates the array as it is rotated around 360o, with measurements taken 36º intervals. The photocurrent of each pixel is collected to map the laser
2
beam output profile as illustrated in the inset of Fig. S3. A 5 mW, 1550 nm wavelength laser is powered by a tunable laser source (Santec TSL-510), and then the light is directed onto the array as described in the text. The measurement at each position is made on the 8×10 photodiodes vicinity of the photodiode with peak intensity to recover the entire laser output profile. The array is rotated and the process is repeated until all 800 photodiode currents are measured.
Detection of stationary objects An imaging experiment is set up to demonstrate detection of various stationary objects around the cylindrical FPA. A 10 mW, 1310 nm wavelength laser beam is shaped using an air-spaced achromatic doublet collimator to illuminate three shadow masks and create three stationary images on the photodiodes array as the laser beam is rotated through 360º. Each of the three 2.4 mm X 2.4 mm masks that cover 64 pixels are positioned ~1 mm from the array. The pixelated image of the letters “O”, “C”, and “M”, are found in Fig. S4(a) with the setup in Fig. S4(b). As shown by Fig. S4(c), the three pixelated letters are imaged on the array by locally mapping the photocurrent of 8 by 8 photodiodes. The image in Fig. S4(c) is blurred possibly due to the incident light leakage reflected between the FPA and the Au-coated shadow masks. Fig. S4(d) shows the measured incident light profile through a single mask slit. The images shown in Fig. S4(c) are then corrected by the cancellation of leakage light using the measured incident light profile. The corrected images are shown in Fig. S4(e). In this experiment, the masks are relatively close to the FPA because we image the objects without focusing. If a microlens array
3
is integrated with the FPA, objects can be detected and imaged even if they are relatively far away from the FPA.
Spectroscopic measurement The 8×100 p-i-n thin-film InGaAs photodiode array is conformally taped on the inner side a 2.54 cm diameter polyethylene terephthalate glycol-modified (PET-G) half tube. The array is positioned 3.4 cm from the grating to achieve a minimum shape mismatch with the focal plane of 30o
the grating system (see Fig. S6). Shape mismatch is defined as σ = 2
∫
( xarray − S )2 dθ , where
−30o
xarray (a function of θ ) is the distance from the grating to the curved photodiode array, S and
θ have the same definition as defined in Eq. (1) in the text. Three collimated diode laser beams (Thorlabs, SM-Pigtailed Butterfly Package Series) at wavelengths of 980 nm, 1310 nm, and 1530 nm are incident on the grating. Air-spaced doublet collimators (Thorlabs, F810FC-1064, F810FC-1310, and F810FC-1550 for 980 nm, 1310 nm, and 1530 nm, respectively) are used to shape the beams. An achromatic doublet (Thorlabs, AC254-075-C) and blaze reflective diffraction grating (Thorlabs, GR13-0610, 1.59 nm/mrad dispersion) are used to focus and diffract the laser beams. The grating, located on a 360◦ rotational stage, is adjusted to achieve a 30◦ beam incident angle. The first order diffracted beams are illuminated on the 4th row of the array. The photocurrents at -1 V bias are collected using a Keithley 2400 SMU. The positions of the collimators and achromatic doublet were adjusted so that the 1530 nm beam is focused to approximately one pixel width (confirmed by measuring the photocurrents of 5 adjacent 4
photodiodes). After intensity peaks are mapped along the 4th row of the array, the laser source and achromatic doublet was changed to continue the measurement until all 100 photodiodes on the 4th row was measured.
Spectrophotometer performance simulation We assume a diffraction grating with blazed at g = 1200 grooves/mm. Both the planar and the curved array have a pixel spacing of 300 µm, and are placed at a distance, L = 3.4 cm, from the diffraction grating. The grating equation gives: (sin θ i + sin θ ) = λ ⋅ g
(S1).
where θi is the light incident angle, θ is the light diffraction angle, and λ is the diffracted wavelength such that:
λ=
sin θi sin θ + g g .
(S2).
Since the first term is a constant depending on the experiment setup, then we have simply:
λ=
sin θ , g
(S3).
although θi has restrictions to make the equation physically meaningful, i.e. λ > 0 . For a planar array, as illustrated in Fig. S6, the wavelength coverage, λc , is,
λc = λmax − λmin
(S4).
where λmax and λmin are the maximum and minimum wavelengths covered by the arrays. 5
Thus, from Eq. (S3):
λc =
sin θ max − sin θ min g
(S5)
where θ max and θ min are the corresponding diffraction angles of λmax and λmin , respectively. Assuming that the grating normal, OC, is at the center of the array, it is obvious that:
θ min = −θ max
(S6)
Thus, Eq. (S5) is simplified to:
λc =
2 sin θ max g
(S7)
Also illustrated in Fig. S6 is that:
tan θ max =
BC BC = OC L
(S8)
Therefore, combine Eqs. (S8) and (S9), the required array length, 2BC , to provide enough coverage for λc is: 2 BC = L ⋅ tan(sin −1 (
λc ⋅ g 2
))
(S9).
For a curved array matching the focal surface of the grating system, the shape of the array in polar coordinates, S , is given by Eq. (1) in the text: S = L ⋅ cos 2 θ
(S10).
The required array length, 2AC , is therefore: θ max
2 AC = −
∫ θ
max
S2 + (
dS 2 ) dθ dθ
(S11).
6
A simultaneous solution of Eqs. (S11), (S10) and (S7) yields the required array length, 2AC as a function of required wavelength coverage, λc .
7
Figure S1. (a) Scanning electron microscopic image of the 10×10 InGaAs photodiode array fabricated on a Si substrate. (b) Microscope image of the array with the same dimensions fabricated on a flexible Kapton® substrate.
8
Figure S2. Dark current (top) and external quantum efficiency (EQE, bottom) maps of 10×10 InGaAs array photodiodes on a (a) Si and (b) Kapton® substrate.
9
Figure S3. Field of view (FOV) measurement setup. The 8×100 p-i-n InGaAs photodiode array is transformed to a 1 cm diameter cylindrical photodiode array. Laser light is guided through optical fiber and normally illuminated on the array. Inset: Schematic illustration of the measurement.
10
Figure S4. (a) Three shadow makes showing pixelated letters “O”, “C”, and “M”. (b) Schematics of the imaging experiment setup. Laser beams illuminate on the array through three makes showing pixelated letters “O”, “C”, and “M” separated by 120º. (c) Photocurrent map on the array showing images of three stationary letters. (d) Incident light profile through a single slit of the shadow mask. (e) Corrected images accounting for light leakage measured in (c).
11
Figure S5. Shape mismatch vs. the position of the array (defined as A in Eq. (1) in the text). When positioned 3.4 cm from the grating, the array achieves minimum shape mismatch with the focal plane, taking into account the diffraction angles from -30◦ to 30◦.
12
Figure S6. Schematic of the geometry used in the wavelength coverage comparison. For a curved (blue) and a planar (orange) array, the required array lengths to cover a wavelength range are 2BC and 2AC , respectively.
13
Department of Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI, USA 2
3
Department of Physics, University of Michigan, Ann Arbor, MI, USA
Department of Materials Science and Engineering, University of Michigan, Ann Arbor, MI, USA *These authors contributed equally to this work.
10×10 photodiode array fabricated on Si and Kapton® substrate The 10×10 thin-film In0.53Ga0.47As p-i-n photodiode arrays are fabricated on both a 500 µm thick semi-insulating Si substrate and a 25 µm thick E-type Kapton® foil. The fabrication processes of 10×10 arrays are similar to that of the 8×100 photodiode array, although the photodiode dimensions are different. For the 10×10 arrays, the top ring contacts have 20 µm/30 µm inner/outer diameters that define the light detection area. Photodiode mesa diameters are 40 µm, with 100 µm pixel separations. The back side linear contacts are 5 µm wide to connect the photodiode rows, and top linear contact columns are also 5 µm wide. A MgF2 (81 nm)/TiO2 (113
1
nm) bilayer anti-reflection coating (ARC) is employed to achieve a maximum EQE at λ = 1550 nm. A scanning electron microscopic image of the array is shown in Fig. S1(a). An optical microscope image of the array with the same dimension fabricated on flexible Kapton® substrate is shown in Fig. S1(b). Maps of the dark current and external quantum efficiency (EQE) of all photodiodes on both Si and Kapton® substrates are shown in Fig. S2(a) and S2(b). The EQE and dark current are measured at -1 V bias with and without 100 µW 1550 nm laser illumination, respectively. The device fabrication yield is 99% for devices fabricated on Si, and 100% for devices on Kapton®. The single malfunctioning device on the Si substrate (indicated in white blank at (3,1) in Fig. S2 (a)) is shorted. Excluding the malfunctioning device, the average EQE at -1V bias and λ = 1550 nm is 88% (1.8 µm thick InGaAs active layer) and 73% (0.85 µm thick InGaAs active layer) for devices fabricated on Si and Kapton® substrates, respectively. The dark current, EQE, and yield of devices fabricated on these two substrates are almost identical over the entire array whether on Si or Kapton®.
Field of view (FOV) measurement The 8×100 p-i-n thin-film InGaAs photodiode array is rolled into a 1cm diameter cylinder as shown in Fig S3. A laser beam illuminates the array as it is rotated around 360o, with measurements taken 36º intervals. The photocurrent of each pixel is collected to map the laser
2
beam output profile as illustrated in the inset of Fig. S3. A 5 mW, 1550 nm wavelength laser is powered by a tunable laser source (Santec TSL-510), and then the light is directed onto the array as described in the text. The measurement at each position is made on the 8×10 photodiodes vicinity of the photodiode with peak intensity to recover the entire laser output profile. The array is rotated and the process is repeated until all 800 photodiode currents are measured.
Detection of stationary objects An imaging experiment is set up to demonstrate detection of various stationary objects around the cylindrical FPA. A 10 mW, 1310 nm wavelength laser beam is shaped using an air-spaced achromatic doublet collimator to illuminate three shadow masks and create three stationary images on the photodiodes array as the laser beam is rotated through 360º. Each of the three 2.4 mm X 2.4 mm masks that cover 64 pixels are positioned ~1 mm from the array. The pixelated image of the letters “O”, “C”, and “M”, are found in Fig. S4(a) with the setup in Fig. S4(b). As shown by Fig. S4(c), the three pixelated letters are imaged on the array by locally mapping the photocurrent of 8 by 8 photodiodes. The image in Fig. S4(c) is blurred possibly due to the incident light leakage reflected between the FPA and the Au-coated shadow masks. Fig. S4(d) shows the measured incident light profile through a single mask slit. The images shown in Fig. S4(c) are then corrected by the cancellation of leakage light using the measured incident light profile. The corrected images are shown in Fig. S4(e). In this experiment, the masks are relatively close to the FPA because we image the objects without focusing. If a microlens array
3
is integrated with the FPA, objects can be detected and imaged even if they are relatively far away from the FPA.
Spectroscopic measurement The 8×100 p-i-n thin-film InGaAs photodiode array is conformally taped on the inner side a 2.54 cm diameter polyethylene terephthalate glycol-modified (PET-G) half tube. The array is positioned 3.4 cm from the grating to achieve a minimum shape mismatch with the focal plane of 30o
the grating system (see Fig. S6). Shape mismatch is defined as σ = 2
∫
( xarray − S )2 dθ , where
−30o
xarray (a function of θ ) is the distance from the grating to the curved photodiode array, S and
θ have the same definition as defined in Eq. (1) in the text. Three collimated diode laser beams (Thorlabs, SM-Pigtailed Butterfly Package Series) at wavelengths of 980 nm, 1310 nm, and 1530 nm are incident on the grating. Air-spaced doublet collimators (Thorlabs, F810FC-1064, F810FC-1310, and F810FC-1550 for 980 nm, 1310 nm, and 1530 nm, respectively) are used to shape the beams. An achromatic doublet (Thorlabs, AC254-075-C) and blaze reflective diffraction grating (Thorlabs, GR13-0610, 1.59 nm/mrad dispersion) are used to focus and diffract the laser beams. The grating, located on a 360◦ rotational stage, is adjusted to achieve a 30◦ beam incident angle. The first order diffracted beams are illuminated on the 4th row of the array. The photocurrents at -1 V bias are collected using a Keithley 2400 SMU. The positions of the collimators and achromatic doublet were adjusted so that the 1530 nm beam is focused to approximately one pixel width (confirmed by measuring the photocurrents of 5 adjacent 4
photodiodes). After intensity peaks are mapped along the 4th row of the array, the laser source and achromatic doublet was changed to continue the measurement until all 100 photodiodes on the 4th row was measured.
Spectrophotometer performance simulation We assume a diffraction grating with blazed at g = 1200 grooves/mm. Both the planar and the curved array have a pixel spacing of 300 µm, and are placed at a distance, L = 3.4 cm, from the diffraction grating. The grating equation gives: (sin θ i + sin θ ) = λ ⋅ g
(S1).
where θi is the light incident angle, θ is the light diffraction angle, and λ is the diffracted wavelength such that:
λ=
sin θi sin θ + g g .
(S2).
Since the first term is a constant depending on the experiment setup, then we have simply:
λ=
sin θ , g
(S3).
although θi has restrictions to make the equation physically meaningful, i.e. λ > 0 . For a planar array, as illustrated in Fig. S6, the wavelength coverage, λc , is,
λc = λmax − λmin
(S4).
where λmax and λmin are the maximum and minimum wavelengths covered by the arrays. 5
Thus, from Eq. (S3):
λc =
sin θ max − sin θ min g
(S5)
where θ max and θ min are the corresponding diffraction angles of λmax and λmin , respectively. Assuming that the grating normal, OC, is at the center of the array, it is obvious that:
θ min = −θ max
(S6)
Thus, Eq. (S5) is simplified to:
λc =
2 sin θ max g
(S7)
Also illustrated in Fig. S6 is that:
tan θ max =
BC BC = OC L
(S8)
Therefore, combine Eqs. (S8) and (S9), the required array length, 2BC , to provide enough coverage for λc is: 2 BC = L ⋅ tan(sin −1 (
λc ⋅ g 2
))
(S9).
For a curved array matching the focal surface of the grating system, the shape of the array in polar coordinates, S , is given by Eq. (1) in the text: S = L ⋅ cos 2 θ
(S10).
The required array length, 2AC , is therefore: θ max
2 AC = −
∫ θ
max
S2 + (
dS 2 ) dθ dθ
(S11).
6
A simultaneous solution of Eqs. (S11), (S10) and (S7) yields the required array length, 2AC as a function of required wavelength coverage, λc .
7
Figure S1. (a) Scanning electron microscopic image of the 10×10 InGaAs photodiode array fabricated on a Si substrate. (b) Microscope image of the array with the same dimensions fabricated on a flexible Kapton® substrate.
8
Figure S2. Dark current (top) and external quantum efficiency (EQE, bottom) maps of 10×10 InGaAs array photodiodes on a (a) Si and (b) Kapton® substrate.
9
Figure S3. Field of view (FOV) measurement setup. The 8×100 p-i-n InGaAs photodiode array is transformed to a 1 cm diameter cylindrical photodiode array. Laser light is guided through optical fiber and normally illuminated on the array. Inset: Schematic illustration of the measurement.
10
Figure S4. (a) Three shadow makes showing pixelated letters “O”, “C”, and “M”. (b) Schematics of the imaging experiment setup. Laser beams illuminate on the array through three makes showing pixelated letters “O”, “C”, and “M” separated by 120º. (c) Photocurrent map on the array showing images of three stationary letters. (d) Incident light profile through a single slit of the shadow mask. (e) Corrected images accounting for light leakage measured in (c).
11
Figure S5. Shape mismatch vs. the position of the array (defined as A in Eq. (1) in the text). When positioned 3.4 cm from the grating, the array achieves minimum shape mismatch with the focal plane, taking into account the diffraction angles from -30◦ to 30◦.
12
Figure S6. Schematic of the geometry used in the wavelength coverage comparison. For a curved (blue) and a planar (orange) array, the required array lengths to cover a wavelength range are 2BC and 2AC , respectively.
13
