sensors Article
Particle and Photon Detection: Counting and Energy Measurement James Janesick 1, * and John Tower 2 1 2
*
SRI-Sarnoff, 4952 Warner Avenue, Suite 300, Huntington Beach, CA 92649, USA SRI-Sarnoff, 201 Washington Road, Princeton, NJ 08540, USA;
[email protected] Correspondence:
[email protected]; Tel.: +1-714-377-6223
Academic Editor: Eric R. Fossum Received: 22 March 2016; Accepted: 4 May 2016; Published: 12 May 2016
Abstract: Fundamental limits for photon counting and photon energy measurement are reviewed for CCD and CMOS imagers. The challenges to extend photon counting into the visible/nIR wavelengths and achieve energy measurement in the UV with specific read noise requirements are discussed. Pixel flicker and random telegraph noise sources are highlighted along with various methods used in reducing their contribution on the sensor’s read noise floor. Practical requirements for quantum efficiency, charge collection efficiency, and charge transfer efficiency that interfere with photon counting performance are discussed. Lastly we will review current efforts in reducing flicker noise head-on, in hopes to drive read noise substantially below 1 carrier rms. Keywords: ultra low noise; CCD; CMOS; imagers
1. Introduction Silicon CCD and CMOS imagers have been demonstrated to be exceptional detectors for particle counting and energy measurement for some time. The spectral range where photon counting is possible covers an extensive wavelength range from 0.1 to 1000 nm (1.24 to 12,400 eV), i.e., nIR, visible, UV, EUV and soft X-ray. At the beginning of the EUV range (10 eV) photon energy absorbed by the imager can be determined by using the simple relation [1], EpeVq “ 3.65ni
(1)
where 3.65 is an experimentally determined constant for silicon (eV/carriers) and ni is measured quantum yield (carriers generated/interacting photon). The equation is applicable to photon energies greater than ~10 eV. The formula is not useful for energies less than this because the constant 3.65 wildly fluctuates. Besides photons, this equation is also useful for any particle that ionizes silicon atoms (electrons, protons, muons, etc.). The uncertainty in energy measurement is limited by the detector’s read floor and Fano noise. Fano noise, the variation of charge generated per photon, is found by, FN “ pFni q0.5
(2)
where F is the Fano factor (~0.1 for silicon) and ni is the quantum yield (carriers generated per photon). Physically Fano noise arises within the silicon where a small amount of thermal energy is lost to the silicon lattice (phonons) instead of creating electron-hole pairs. The variation in the loss from pixel to pixel is the Fano noise generated and represents a fundamental noise source in determining the energy of high energy particles [2]. Imagers where Fano noise is greater than the sensor’s read noise are referred to as “Fano noise limited” [2]. Figure 1 plots Fano noise as a function of photon energy (eV) and wavelength (µm)
Sensors 2016, 16, 688; doi:10.3390/s16050688
www.mdpi.com/journal/sensors
Sensors 2016, 16, 688 Sensors Sensors 2016, 2016, 16, 16, 688 688
2 of 17 22 of of 17 17
showing be covered by an imager for aa given read noise floor. For showing the the Fano Fanolimited limitedrange rangethat thatcan can covered imager a given the Fano limited range that can bebe covered byby an an imager for for given readread noisenoise floor.floor. For example, Figure 22 presents aa histogram taken from aa Fano For example, Figure 2 presents a histogram taken from aFano Fanonoise noiselimited limited888um um3T 3T NMOS example, Figure presents histogram taken from noise limited um 3T NMOS pinned pinned photo showing multiple multiple energy energy lines lines from from aa basalt photo diode diode (PPD) (PPD) pixel pixel array array showing basalt target target fluoresced fluoresced with with 5.9 sensor’s read noise is less 22 electrons (e Fano keVMn MnX-rays. X-rays.The The sensor’s read noise is slightly less than 2 electrons (e´ ) allowing Fano 5.9 keV keV Mn X-rays. The sensor’s read noise is slightly slightly less than than electrons (e−−)) allowing allowing Fano limited limited performance to entire X-ray nm). range has limited performance cover thesoft entire soft range X-ray (0.12 range (0.12 nm–10 nm).spectral This spectral has performance to cover covertothe the entire soft X-ray range (0.12 nm–10 nm–10 nm). This This spectral rangerange has been been particularly fruitful for soft X-ray imaging spectrometers used been particularly fruitful for CCD X-ray imaging spectrometers usedin scientific and and space particularly fruitful for CCD CCD softsoft X-ray imaging spectrometers used ininscientific scientific and space applications. applications. The The width width of of each each spectral spectral line line revealed revealed in in Figure Figure 22 is is aa measurement measurement of of the the amount amount of of Fano noise generated. The spectral range for this imager can be further extended into the EUV range Fano noise generated. The spectral range for this imager can be further extended into the EUV range (10 (10 nm–124 nm–124 nm) nm) if if only only photon photon counting counting is is desirable. desirable.
Figure 1. The plots above are used determine “Fano noise limit” an aa given 1. The The plots plotsabove aboveare areused usedtoto to determine the “Fano noise limit” forimager an imager imager with given Figure 1. determine thethe “Fano noise limit” for for an withwith a given read read For read one rms will the soft and into read noise. noise. For example, example, read noise noise ofcarrier one carrier carrier rmscover will cover cover the soft X-ray X-ray and extend extend into the noise. For example, a readaa noise of oneof rms will the soft X-ray and extend into the UVthe at UV at wavelength of 0.03 um (~40 eV). UV at wavelength of 0.03 um (~40 eV). wavelength of 0.03 um (~40 eV).
Photon counting counting and and energy energy histogram histogram generated generated by by aaa 3T CMOS pixel pixel imager imager Figure Figure 2. 2. Photon Photon counting and energy histogram generated by 3T PPD PPD CMOS demonstrating demonstrating Fano-noise Fano-noise limited limited performance performance over over the the entire entire soft soft X-ray X-ray regime. regime.
The The photon photon energy energy for for visible visible (400 (400 nm–700 nm–700 nm) nm) and and nIR nIR (700 (700 nm–1100 nm–1100 nm) nm) wavelengths wavelengths is is only The photon energy for visible (400 nm–700 nm) and nIR (700 nm–1100 nm) wavelengths is only only able one electron-hole pair/photon, able to to generate generate one one electron-hole electron-hole pair/photon, pair/photon, limiting limiting sensing to only photon counting. counting. But But when when able to generate limiting sensing sensing to to only only photon photon counting. But when leaving the visible range into multiple carriers per are allowing their leaving the thevisible visiblerange rangeinto into the UV multiple carriers per photon photon are generated generated allowing their leaving thethe UVUV multiple carriers per photon are generated allowing their energy energy to be determined. The challenge left today is to extend energy measurement into the UV energy to be determined. The challenge left today is to extend energy measurement into the UV and and provide provide photon photon counting counting in in the the visible/nIR visible/nIR wavelengths wavelengths by by reducing reducing the the sensor’s sensor’s read read noise noise floor. floor.
Sensors 2016, 16, 688
3 of 17
to be determined. Sensors 2016, 16, 688 The challenge left today is to extend energy measurement into the UV and provide 3 of 17 photon counting in the visible/nIR wavelengths by reducing the sensor’s read noise floor. The average averagenoise noise floor performance and CMOS is typically shy of The floor for for highhigh performance CCD CCD and CMOS imagersimagers is typically shy of achieving achieving “one” carrier of noise (thisEMCCD excludes EMCCD SPAD detectors). ForFigure example, Figure a3 “one” carrier of noise (this excludes and SPAD and detectors). For example, 3 presents presentstransfer a photon (PT)generated [2,3] curve by512 a 512 × 512 umPMOS × 8 um PPD imager PMOS photon (PT)transfer [2,3] curve bygenerated a 512 um ˆ um um ˆ 8 um PPD CMOS CMOS imager with a read noise floor of 1.08 holes (h+ )at room temperature. It is extraordinary that with a read noise floor of 1.08 holes (h+ )at room temperature. It is extraordinary that the CCD/CMOS the CCD/CMOS imaging community for the most part has achieved approximately “one” carrier of imaging community for the most part has achieved approximately “one” carrier of noise considering noise considering the multitude of solid state phenomena at play at world wide fabrication the multitude of solid state phenomena at play at world wide fabrication foundries for several years foundriesinfor (decades in “one”? the caseIs of CCDs). Butfinal whyoutcome “one”? Is this apparent final (decades theseveral case ofyears CCDs). But why this apparent simply coincidental? outcome coincidental? One alsocarrier wonders why with “one”“one” carriercarrier of noise “one” One also simply wonders further why “one” of further noise along of along signalwith forces the carrier of signal forces the minimum detection limit (MDL) of the detector to be “one” (i.e., minimum detection limit (MDL) of the detector to be “one” (i.e., S/N = “one”). The conundrum S/N = “one”). The conundrum continues are very close butcounting yet so farsingle from continues with why today’s imagers are with verywhy closetoday’s but yetimagers so far from “routinely” “routinely” counting single across visible full photons consistently visible photons consistently imaging arrays. across full imaging arrays.
Figure 3. 3. Photon read noise noise floor floor of of ~1 ~1 hole hole rms. rms. Most high high Figure Photon Transfer curves demonstrating a read performance CCD CCD and and CMOS CMOS imagers imagers are are close closeto tothis thisnoise noiselevel. level. performance
To appreciate the challenge, Figure 4 displays computer simulated histograms for different read To appreciate the challenge, Figure 4 displays computer simulated histograms for different read noise levels for an average signal level of “one” interacting photon/pixel and quantum yield of noise levels for an average signal level of “one” interacting photon/pixel and quantum yield of “one”. “one”. These plots show that it is necessary to have a read noise floor of
