Timing advances of commercial divalent-ion co-doped LYSO:Ce and SiPMs in sub-100 ps time-of-flight positron emission tomography

Objective. Together with novel photodetector technologies and emerging electronic front-end designs, scintillator material research is one of the key aspects to obtain ultra-fast timing in time-of-flight positron emission tomography (TOF-PET). In the late 1990s, Cerium-doped lutetium–yttrium oxyorthosilicate (LYSO:Ce) has been established as the state-of-the-art PET scintillator due to its fast decay time, high light yield and high stopping power. It has been shown that co-doping with divalent ions, such as Ca 2+ and Mg 2+, is beneficial for its scintillation characteristics and timing performance. Therefore, this work aims to identify a fast scintillation material to combine it with novel photosensor technologies to push the state of the art in TOF-PET. Approach. This study evaluates commercially available LYSO:Ce,Ca and LYSO:Ce,Mg samples manufactured by Taiwan Applied Crystal Co., LTD regarding their rise and decay times as well as their coincidence time resolution (CTR) with both ultra-fast high-frequency (HF) readout and commercially available readout electronics, i.e. the TOFPET2 ASIC. Main results. The co-doped samples exhibit state-of-the-art rise times of on average 60 ps and effective decay times of on average 35 ns. Using the latest technological improvements made on NUV-MT SiPMs by Fondazione Bruno Kessler and Broadcom Inc., a 3 × 3 × 19 mm3 LYSO:Ce,Ca crystal achieves a CTR of 95 ps (FWHM) with ultra-fast HF readout and 157 ps (FWHM) with the system-applicable TOFPET2 ASIC. Evaluating the timing limits of the scintillation material, we even show a CTR of 56 ps (FWHM) for small 2 × 2 × 3 mm3 pixels. A complete overview of the timing performance obtained with different coatings (Teflon, BaSO4) and different crystal sizes coupled to standard Broadcom AFBR-S4N33C013 SiPMs will be presented and discussed. Significance. This work thoroughly evaluates commercially available co-doped LYSO:Ce crystals and, in combination with novel NUV-MT SiPMs, shows a TOF performance that significantly exceeds the current state of the art.


Introduction
Time-of-flight positron emission tomography (TOF-PET) requires dense and fast-decaying scintillation materials with a high light yield to make patients benefit from an increased signal-to-noise ratio (SNR) and a high sensitivity. With the introduction of cerium-doped lutetium oxyorthosilicate (LSO:Ce, Lu 2 SiO 5 :Ce) and lutetium-yttrium oxyorthosilicate (LYSO:Ce, (Lu-Y) 2 SiO 5 :Ce) at the turn of the millennium, this new scintillating material quickly became the state-of-the-art PET scintillator and has defended its position until today (Melcher 2000, Cherry and Dahlbom 2006, Korzhik et al 2007, Lecoq 2016.
In L(Y)SO:Ce, incident γ-photons excite electrons from the valence to the conduction band. After generating secondary charge carriers and those thermalizing and recombining at Ce 3+ centers, optical photons are emitted (Jensen et al 2022). This scintillation signal is characterized by a scintillation rise time of about 70 ps, scintillation decay times of about 40 ns and a light yield of up to 40000 optical photons per deposited MeV (ph MeV −1 ) . Co-doping L(Y)SO:Ce with Ca 2+ or Mg 2+ ions, where the charge of the divalent doping is compensated by Ce 4+ (Chou 2012), has reduced the scintillation rise times to 21 ps and effective scintillation decay times to about 30 ns, but lowered the light yield to 32 000 to 39 000 ph MeV −1 (Blahuta et al 2013, ter Weele et al 2014, 2016, Martinez Turtos et al 2016, 2019, Rusiecka et al 2021, Addesa et al 2022. The change of these properties strongly depends on the actual co-doping fraction. In TOF-PET, the optical photons emitted upon scintillation are commonly digitized using analog siliconphotomultipliers (SiPMs), which consist of several thousands of single-photon avalanche diodes (SPADs) connected in parallel. Creating electron-hole pair cascades, each SPAD is highly sensitive to the signal of a single optical photon incident on the SPADʼs depletion region. The SiPM signal corresponds to the sum of all SPAD signals. The overall TOF-PET system performance and capability to extract precise timing information strongly depends on the number of optical photons being generated in the scintillator upon the impact of a 511 keV γphoton and the simultaneous emission time profile of these optical photons. This link to the achievable coincidence time resolution (CTR) has been described and extensively studied by , Vinogradov (2018)  where τ d,eff and τ r are the effective decay and rise time of the scintillator response, LTE is the light transfer efficiency and ILY is the intrinsic light yield. It has been shown that co-doping L(Y)SO:Ce with divalent ions can enhance its scintillation kinematics, resulting in faster rise and decay time and thus an improved CTR for TOF-PET (Tamulaitis et al 2020). Despite many studies reporting the advantages of co-doping using research samples, a systematic study evaluating the timing properties of commercially available co-doped LYSO:Ce samples is yet to be made. In doing so, it has to be kept in mind that among other scintillation properties and effects linked to the crystal shape and coating, such as the photon transfer time spread (PTS), also SiPM characteristics, such as the single-photon time resolution (SPTR) and the photon detection efficiency (PDE), contribute to the achievable CTR. This study evaluates the timing properties of commercially available LYSO:Ce samples with divalent codoping (Ca 2+ , Mg 2+ ) manufactured by Taiwan Applied Crystal Co., LTD, (TAC) in combination with commercially available state-of-the-art photosensors from Broadcom Inc. We apply state-of-the-art highfrequency (HF) readout electronics (Cates et al 2018, Gundacker et al 2019 to characterize their performance limits and report their CTR and energy resolution (dE/E) with a commercial and system-applicable readout architecture, the TOFPET2 ASIC by PETsys Electronics S.A. (Bugalho et al 2019).

Scintillators
The ingots of the presented LYSO:Ce,Mg and LYSO:Ce,Ca samples were grown in crucibles using the Czochralski technique at the TAC facilities (Chou 2018, Taiwan Applied Crystal Co. LTD 2022. TAC focuses on research and development (R&D) of high-density ceramics and single-crystals, such as LYSO:Ce,Mg and LYSO:Ce,Ca, in various customizable dimensions, i.e. single pixels, matrix assembly, (semi-)monolithic cubes as well as fibers with different lengths and cross sections. This guarantees universities and R&D teams the flexibility to test different designs for PET and other applications. Upon production of the presented samples, a thermal annealing process was used to increase the radiation tolerance and ingot uniformity. The thermal annealing was achieved by placing the double-doped single-crystal ingot in a furnace to increase the yield and luminescence intensity of the overall ingot. The ingot contains 5% to 8% yttrium (Taiwan Applied Crystal Co. LTD 2022). The ingots were cut into smaller units (sizes 2 × 2 × 3 mm 3 , 3 × 3 × 3 mm 3 and 3 × 3 × 19 mm 3 ) and polished (surface roughness 0.02 μm Taiwan Applied Crystal Co. LTD 2022) as depicted in figure 1. On selected pixels, barium sulfate (BaSO 4 ) powder was applied as coating at the TAC facilities. If no coating was applied, the pixels were wrapped in several layers of Teflon tape (see figure 2). The investigated samples were assigned to classes (B, C) with respect to their optical properties by TAC. They are expected to have a light yield of about 34 500 ph MeV −1 and a decay time of no more than 39 ns (Taiwan Applied Crystal Co. LTD 2022 ).

Photosensors and readout electronics
The LYSO:Ce,Mg and LYSO:Ce,Ca pixels were optically coupled to silicon-photomultipliers (SiPMs) from Broadcom (NUV-HD, AFBR-S4N33C013, 3 × 3 mm 2 , 30 μm SPADs) using Cargille Meltmount TM (n = 1.582). The breakdown voltage of these SiPMs was measured to be 26.5 V using a Keithley 2400 Sourcemeter . The SPTR of these photosensors has been determined in this study with a blackpainted lead fluoride crystal (PbF 2 , EPIC crystal, 2 × 2 × 3 mm 3 ) mounted to the SiPM and set up in a coincidence experiment with a reference detector and a 22 Na point source.   finger plot, similar to the approach with laser excitation . The SPTR is the full-width at halfmaximum (FWHM) of the time delay histogram computed with respect to the reference detector, which is filtered for 511 keV events. The measured SPTR amounts to 72 ps (FWHM) at an overvoltage of 37 V. Accounting for the electronic noise of the used HF amplifier, an intrinsic SPTR of 67 ps (FWHM) can be estimated. The signals of the SiPMs were read out by ultra-fast HF electronics (Cates et al 2018, Gundacker et al 2019 with power-optimized timing channels as introduced in Krake et al (2022). Signals were digitized via a Lecroy Waverunner 9404 M-MS (bandwidth 4 GHz, 20 GS s −1 ). In addition, the TOFPET2 ASIC (version 2c) by PETsys Electronics S.A. was used for readout of selected samples to estimate the scintillator performance in combination with scalable and systemapplicable readout electronics (see figure 3). To optimize the timing performance, a pole-zero circuit was implemented in front of the ASIC input stage (C = 100 pF, R = 390 Ω) to stabilize the baseline (Gola et al 2011. The ASIC input stage impedance was configured to 11 Ω (fe_ib1 = 0) (Nadig et al 2020). The leading-edge threshold applied in the scans was vth_t1 = 20 (lsb = 6.66 mV).

Scintillation characteristics
A time-correlated single-photon counting (TCSPC) setup, similar to the one in Gundacker et al (2018), Gundacker et al (2021), was used to precisely measure the rise and decay times of the LYSO:Ce,Mg and LYSO:Ce, Ca samples (see figure 4). A 3 × 3 × 3 mm 3 LYSO:Ce crystal was coupled to a Broadcom AFBR-S4N33C013 SiPM using Cargille Meltmount TM (n = 1.582) and read out by the aforementioned HF electronics, which is the so-called 'start-detector'. The detector was aligned in coincidence with a scintillator sample of interest coupled to a single-SPAD detector, likewise connected to an HF readout board, the so-called 'stop-detector'. Both the timing and energy signal of the start-detector and the timing signal of the stop-detector were digitized by an oscilloscope (Lecroy Waverunner 9404M-MS, bandwidth 4 GHz, 20 GSs −1 ). The impulse response function (IRF) of the TCSPC system was measured using an uncoated and unwrapped ('naked') 2 × 2 × 3 mm 3 PbF 2 crystal manufactured by EPIC crystal, making use of its Cherenkov emission (Kratochwil et al 2021) similar to the method introduced in Gundacker et al (2016) with a lutetium aluminium garnet (LuAG) scintillator. The histogram of recorded time delays between 'start-detector' and 'stop-detector' composed the shape of the IRF of the setup, i.e. its intrinsic time resolution. The IRF can be described by either a sole-Gaussian function (see equation (3) where μ describes the mean time delays within the electronics, or a Gaussian with an exponential tail caused by photon absorption deeper in the SPAD junction (see equation (4)) (Nemallapudi et al 2016) with λ accounting for the exponential contribution of the delay tail and the error function being defined as The scintillation emission was modeled as a sum of exponential functions with multiple rise and decay times multiplied by a Heaviside function (Gundacker et al 2018) where θ represents the point in time equal to the onset of the scintillation emission and ρ i are the weights of a multi-component fit. The convolution of the IRF with the model of the scintillation emission IRF(t) * f (t|θ) was fitted to the histogram of acquired events per measured time difference to determine the rise and decay times τ r,i and τ d,i . The effective decay time can then be computed via where R i are the abundances of the decay times τ d,i . An elaborate description of this method is provided in Gundacker et al (2018.

Coincidence experiments
Coincidence experiments were conducted with both readout circuits, the HF and the TOFPET2 readout (Krake et al 2022, Nadig et al 2022 and figure 3), varying the applied bias voltage and threshold. A 22 Na source with an activity of 2.5 MBq was positioned in between two single detector pixels. The setups were enclosed in a dark chamber ensuring a stable temperature of 16°C. Data were processed matching coincidences within a time window of 7.5 ns. An energy filter was applied ±2σ around the photopeak. The CTR was computed as the FWHM of a Gaussian fit to the resulting coincidence time difference spectrum. The energy resolution was determined via the FWHM of the photopeak in the saturation-corrected energy spectrum, whereby equation (10) was applied as a generic model to compute the energy E in keV from the acquired raw energy value e (see equation (10)). The variables s and a are saturation and amplitude parameters

Scintillation characteristics
The width of the IRF of the TCSPC setup was determined to be σ IRF = 41 ps (FWHM 96 ps) using a 2 × 2 × 3 mm 3 PbF 2 crystal manufactured by EPIC crystal (see figure 5). As depicted in figure 6 for the example of 2 × 2 × 3 mm 3 LYSO:Ce,Ca (class B) and LYSO:Ce,Mg (class B) crystals, a histogram of the time delays acquired between the reference and the single-SPAD detector is plotted and fitted with equation (7), convolved with the measured IRF according to equation (5) and figure 5, to extract the characteristics of the scintillation emission, namely the rise and decay time of the material, using a first-order approach (N = 1) for the rise times and a second-order approach

Timing performance limits
In bias and threshold scans conducted with power-optimized HF readout electronics, 3 × 3 × 3 mm 3 LYSO:Ce, Mg samples wrapped in Teflon tape achieve CTRs of 85 ± 2 ps (FWHM) at an applied bias voltage of 37 V (see table 2 and figure 7). Limiting the crystalʼs base area to 2 × 2 mm 2 to ensure all light is collected improves the CTR to 71 ± 2 ps (FWHM). For scintillators of clinical length (3 × 3 × 19 mm 3 ), the CTR is 117 ± 2 ps (FWHM). Ca-co-doped samples, compared to Mg-co-doping, show the same timing performance within the errors or exceed the timing performance of Mg-co-doped samples, with 2 × 2 × 3 mm 3 LYSO:Ce,Ca samples achieving a CTR of 67 ± 2 ps (FWHM) and 3 × 3 × 19 mm 3 LYSO:Ce,Ca samples achieving a CTR of 108 ± 2 ps (FWHM). No significant difference in CTR is observed between samples from class B and class C. Replacing the Teflon wrapping by BaSO 4 coating improves the CTR to 80 ± 2 ps (FWHM) for 3 × 3 × 3 mm 3 , but worsens it to 133 ± 2 ps (FWHM) for 3 × 3 × 19 mm 3 LYSO:Ce,Mg (class B) (see table 2 and figure 7). The same effects of BaSO 4 coating are observed for samples with Ca-co-doping (see table 2). Figure 8 depicts the CTR measured with a system-applicable readout (TOFPET2 ASIC) as a function of the overvoltage applied, while table 2 reports the best CTR achieved with this readout in comparison to HF electronics. The TOFPET2 input stage impedance was configured to be 11 Ω (fe_ib1 = 0) and a pole-zero filter was employed (C = 100 pF, R = 390 Ω) to guarantee a more stable channel baseline. The leading-edge threshold applied in the scans was vth_t1 = 20 (lsb = 6.66 mV). Changing the readout electronics from HF readout to system-applicable TOFPET2 readout decreases the performance of the 3 × 3 × 3 mm 3 samples (class B) to 151 ± 1 ps and 158 ± 1 ps (FWHM) for Ca-and Mg-doping, respectively. For the crystals with a smaller base area, again with respect to the HF readout, the CTR deteriorates to 132 ps and 133 ps, accordingly. For the 3 × 3 × 19 mm 3 LYSO:Ce,Mg sample, a CTR of 195 ± 1 ps (FWHM) is measured with the TOFPET2 ASIC. In contrast to prior experiments with HF readout electronics, only marginal or no deterioration of the CTR is observed related to coating the samples with BaSO 4 . However, Ca-co-doped samples slightly outperform Mg-co-doped samples. The energy resolution for samples with a 3 × 3 mm 2 base area was ranging from 11.5% to 12.2%, measured with settings optimized for timing performance. The scintillators with a smaller base area (2 × 2 mm 2 ) showed an improved energy resolution of 10.2% to 10.4% at the same timing-optimized settings (see table 2 and figure 9).

Discussion
The measured rise times (single-exponential model) are comparable to values for Ca-co-doped crystal reported in literature ) (see table 1). Switching to a double-exponential model for the rise time computation (see table 3) reveals fast and slow rise time components for both Ca-and Mg-co-doped samples. It has to be noted that the abundance of the slow rise time component is significantly lower than for the fast component for all samples without wrapping or coating. This effect is suppressed once the samples are wrapped in Teflon, resulting in equal abundances of the fast and slow rise time component due to the photon reflections by the wrapping. It is interesting to notice that this slow component was not seen in highly co-doped samples tested in former studies , but is indeed visible in standard L(Y)SO:Ce samples as noted in the same studies. This leads to the conclusion that the amount of co-doping plays a crucial role in suppressing the slow rise time component, which is at the same time an important parameter for fastest timing. In this sense, the precise scintillation emission rise time measurement can be a powerful tool to control the production process and uniformity. The intrinsic emission, however, is masked once the samples are wrapped in Teflon. Naturally, modeling the rise time by a singleexponential function shows a sensitivity to slight differences in the co-doping fractions, as in this case all effects are averaged in one number. However, a double-exponential model produces more stable and comparable results for samples of the same size (see table 3).  Gundacker et al (2021), allows to estimate the errors in a ± 2σ-interval to 3 ns for the fast and 8 ns for the slow decay times. The errors for the effective decay times are only 0.6 ns, thus, the effective decay times are significantly different among Ca-co-doped and Mg-co-doped samples, showing the tendency that Caco-doping exhibits slightly faster scintillation kinematics. This is resembled by the better timing performance measured for the Ca-co-doped samples, for measurements with ultra-fast HF readout electronics as well as for measurements with the TOFPET2 ASIC evaluation kit. Therefore, the effective decay time can be seen as a valid figure of merit. Due to the divalence of both the Mg 2+ and Ca 2+ ions, Mg-and Ca-co-doped samples are expected to show the same effects on scintillation characteristics and timing performance. However, with the slight benefit of Ca-co-doping becoming apparent, the observed difference might be due to different co-doping fractions, which are subject to non-disclosure. As the light yield of the samples could only be assessed relatively, given the proposed experimental setup, this study refrains from reporting it explicitly. It can be noted that the relative light yield in comparison to a small LYSO:Ce crystal (2 × 2 × 3 mm 3 ; Epic-crystal) is the same for both Mg-and Ca-co-doped samples and therefore in the range of 40 000 ph MeV −1 .
Regarding the timing performance of the samples presented in this study, the co-doped samples outperformed LYSO:Ce samples by other producers by more than 10 ps (FWHM) for both the ultra-fast HF readout electronics as well as the system-applicable TOFPET2 ASIC readout. With respect to the achieved CTR, the presented material appears to be similar to the LSO:Ce,Ca crystals investigated in . It seems that the divalent doping fraction is a bit lower, as the slow rise time component is on the edge to be suppressed, but is not completely. Concluding from lower doping fraction that the transparency should be higher for the crystal samples in this study, this further means that the CTR values are expected to outperform Table 1. Scintillation characteristics of the investigated LYSO:Ce,Ca and LYSO:Ce,Mg samples. The tabular compares values from literature to values measured in this study. All rise times reported were determined using a single-exponential model, while decay times reported were determined using either a single-or a double-exponential model according to equation (7). The effective decay time was computed using equation (8). Extracted via Monte-Carlo simulations and considering a ±2σ-environment (95% confidence interval), the errors on the fast and slow decay time component are 3 ns and 8 ns, respectively, with an abundance error of 14% for each component. The error on the effective decay time is 0.6 ns.  other producers for crystals of 20 mm length, which is indeed observed in this studyʼs measurements. Furthermore, we can conclude that the amount of co-doping is to be finely controlled, not only for uniformity aspects, but also for the resulting timing performance in a specific detector assembly. The observed effects of the BaSO 4 -coating are suspected to be due to a change of the angle of total internal reflection on the crystal-air interface, which is caused by applying a diffusive reflector directly on the crystal surface. Similar to a roughened crystal, more light is coupled into the crystal by BaSO 4 and in the case of short crystal lengths (about 10% to 20% more light at a bias voltage of 32 V). The increased light output has a positive effect on the scintillation pulse shape in these smaller crystals, resulting in an improved timing performance. In longer crystals (19 mm), Figure 7. Coincidence time resolution measured as a function of the applied bias voltage using HF readout electronics from Krake et al (2022), Nadig et al (2022). The leading-edge threshold was configured to 80 mV. a All reported energy resolutions have been corrected for the saturation of the photosensor using an exponential model (see equation (10)). The value given states the mean energy resolution of both detector channels for the same settings at which optimal timing performance was achieved.

Rise time a /ps
however, the effect on the light output is less prone (only about 5%) increase. In addition, the diffuse reflection might contribute to a smearing of the scintillation pulse over time due to the longer path the light has to travel inside the crystal, which leads to a deteriorated timing performance. Though other studies also report the effects of BaSO 4 coatings (Gonzalez-Montoro et al 2021), the crystal length at which this effect turns from positive to negative is yet to be determined. Figure 8. Coincidence time resolution measured as a function of the applied bias voltage using the TOFPET2 ASIC evaluation kit. A pole-zero filter was employed to reduce baseline shifts (C = 100 pF, R = 390 Ω). The TOFPET2 input stage impedance was configured to be 11 Ω (fe_ib1 = 0). The leading-edge threshold applied was vth_t1 = 20 (lsb = 6.66 mV). The reported energy resolution of about 10% to 12% is sufficient for system requirements. Operating the SiPMs at higher bias voltages to push their photo-detection efficiency and therefore timing resolution contributes to a higher dark count rate and may also deteriorate the energy resolution. Here, the impact of the electronics on the performance becomes evident, as HF readout allows to push the SiPMs to higher overvoltages, while the highest CTR with the TOFPET2 ASIC is achieved at lower overvoltages (see shifted minima in figures 7 and 8).
Co-doping crystals proves to be an effective method to boost the scintillation characteristics and timing performance of PET detectors. Having been given the opportunity to evaluate the presented crystal samples with SiPMs employing a novel metal-trench (MT) technology, which was co-developed by Fondazione Bruno Kessler (FBK) and Broadcom Inc. (which are already a commercial product Broadcom Inc., 2023b), the timing performance could be pushed even further. With Broadcom NUV-MT SiPMs (3.8 × 3.8 mm 2 , 40 μm SPADs, Figure 10. Optimal coincidence time resolution measured with two LYSO:Ce,Ca crystals (TAC, 2 × 2 × 3 mm 3 , four faces polished, and 3 × 3 × 19 mm 3 , six faces polished) coupled to novel Broadcom NUV-MT SiPMs (breakdown voltage 32 V) using Cargille Meltmount (n = 1.582). The SiPMs were read out by HF electronics and operated at a bias voltage of 50 V. The leading-edge threshold was varied. Table 3. Scintillation characteristics of the investigated LYSO:Ce,Ca and LYSO:Ce,Mg samples. The tabular compares values from literature to values measured in this study. All rise times reported were determined using either a single-or a double-exponential model. breakdown 32 V) and ultra-fast HF readout, a CTR of 70 ps (FWHM) was obtained for a 2 × 2 × 3 mm 3 LYSO: Ce,Ca crystal at an applied bias voltage of 37 V, which is comparable to the performance achieved with the same crystal being coupled to Broadcom AFBR-S4N33C013 SiPMs (CTR of 67 ± 2 ps (FWHM), see table 2). Exploiting a reduced correlated noise due to the MT technology, while maintaining a high PDE, this value could be improved to 56 ± 2 ps (FWHM) at a bias voltage of 50 V (see figure 10). As in both cases, for NUV-HD and NUV-MT SiPMs, the base area of the crystal is significantly smaller than the active area of the SiPM, there should be no difference in photon collection for the two measurements and thus no influence by the crystal-SiPM coverage on the performance. With a clinical LYSO:Ce,Ca crystal (3 × 3 × 19 mm 3 ), as depicted in figures 10 and 11, a CTR of 95 ± 2 ps (FWHM) is achieved at the same bias voltage. Here, the effect of a larger SiPM active area on the CTR should be small, as discussed in Gundacker and Heering (2020). Using the system-applicable and commercially available TOFPET2 ASIC, a CTR of 157 ± 1 ps (FWHM) can be reported for a 3 × 3 × 19 mm 3 LYSO:Ce,Ca crystal and at 42 V applied bias voltage (see figure 11). Compared to standard NUV-HD SiPMs manufactured by Broadcom and their already high PDE of 54% at 420 nm and an overvoltage of 5 V, the NUV-MT SiPMs feature an even higher PDE of 63 % at the same wavelength and their typical point of operation (12 V overvoltage) (Broadcom Inc., 2023a, 2023b. Due to a reduced internal crosstalk by the metaltrench technology, which is expected to additionally improves the CTR , the applied bias voltage can be pushed higher without a decrease in performance, resulting in an even higher PDE.

Conclusion
Fast scintillation materials are one of the key aspects to boost timing in TOF-PET, especially when trying to reach a CTR of 100 ps (FWHM) on system level, which requires to optimize every last component of the detection chain. The presented divalently co-doped and commercially available samples show adequate timing properties to follow this research track, exhibiting state-of-the-art rise and decay times. Uniformity among the crystals will be important to maintain this performance on system scale. The measurements with novel NUV-MT SiPMs indicate that also recent developments on the sensor technology have a large potential to enable reaching this goal, already achieving a CTR of 157 ± 1 ps (FWHM) with LYSO:Ce,Ca crystals of clinical length (3 × 3 × 19 mm 3 ) and commercially available readout electronics (TOFPET2 ASIC), even without any frontend modification. Employing ultra-fast HF readout even allows for a CTR of 95 ± 2 ps (FWHM). This study proves that the goal of 100 ps is indeed feasible, combining efforts from scintillation material development, novel sensor technology and, in the end, emerging electronic front-end designs. Future studies will harvest on these developments for system proof-of-concept detector designs within the framework of the ProtoTOF project.