Preamble

Characterizations of a Novel 3D Trench-Column Sensor with Internal Gain Manwen Liu, 1, 2, 3, Kuo Ma, Huimin Ji, 1, 2, 3 De Zhang, Yanwen Liu, Zheng Li, Zhihua Li, 1, 2, 3, and Jun Luo 1, 2, 3, 1 State Key Laboratory of Fabrication Technologies for Integrated Circuits, Chinese Academy of Sciences, Beijing 100029, China Institute of Microelectronics, Chinese Academy of Sciences, Beijing 100029, China School of Integrated Circuits, University of Chinese Academy of Sciences, Beijing 100049, China Department of Modern Physics and State Key Laboratory of Particle Detection and Electronics, University of Science and Technology of China, Hefei 230026, China Over the last few years, 3D radiation sensors have been developed in High Energy Physics (HEP) experiments and other fields such as microdosimetry in proton and ion therapy. Due to their unique architectural design, 3D sensors offer radiation tolerance and fast signal response at relatively low bias voltages, making them one of the most promising sensor technologies for spatiotemporal (4D) tracking in extreme environments. Thin 3D sensors feature lower capacitance (and thus lower noise) and reduced material budget but suffer from limited charge collection. To address this, the IMECAS group has proposed a novel 3D trench-column sensor with internal gain, where charge multiplication is primarily enabled by the ultra-narrow radius of N columnar electrodes. These sensors were manufactured on 8-inch wafers using IMECAS CMOS processing technologies, achieving an aspect ratio of 70:1 via the deep reactive ion etching (DRIE) process. In this paper, we present the electrical characteristics, charge collection and time resolution of the sensor with a 35 pixel cell.

Through Current-Voltage ( ) and Capacitance-Voltage ( ) measurements, we investigate the depletion and breakdown characteristics of the sensors. Charge collection and time resolution have been evaluated using infrared transient current technology (TCT) and Sr-source measurements. The results confirm a substantial internal gain in the proposed sensor.

Keywords

3D trench-column sensors, Charge collection, Internal gain

INTRODUCTION

3D radiation sensors were proposed by Sherwood Parker and his collaborators in 1997 [ ], distinguished from planar sensors by vertical electrodes that are perpendicular to the wafer surface and extend into the substrate. The smaller in- terelectrode distance endows these sensors with advantages including low full depletion voltage, fast time response, and radiation hardness.

Pioneering research on 3D sensors initially focused on 9

single-sided fabrication processes [ ], while groundbreaking, faced challenges in electrode alignment and charge collection

uniformity. "Active edge" technology was adopted in single- 12

sided 3D sensors to minimize dead peripheral regions [ 3 ]. 13

Subsequently, institutions such as CNM (Barcelona) [ ] and FBK (Trento) [ ] independently developed more streamlined double-sided 3D schemes, resolving earlier process bottle- necks and improving the reproducibility of the fabrication.

This technological advancement was notably applied in high- energy physics (HEP), with the ATLAS Insertable B-Layer (IBL) standing as the first HEP application of 3D sensor tech- nology [ ]. However, its deployment required overcoming challenges in scaling to large-area arrays for particle detec- tion.

Small-pitch 3D pixels, tailored for high-luminosity LHC

upgrades [ 7 , 8 ], mitigate mechanical fragility via Si-Si direct 25

wafer-bonded substrates [ ]. Meanwhile, 3D-trenched detec-

tors, developed under initiatives like INFN TIMESPOT [ 10 ], 27

further enhance timing resolution by optimizing trench depth and electrode spacing [ However, because of the low electric field regions between the same types of electrodes, the signal response of the 3D

column sensor is not uniform. In addition, the capacitance 32

of the trench sensor is much larger than that of planar sen- sors due to the depth of the electrodes and the small spac- ing between them [ To mitigate these limitations, Zheng Li in BNL (USA) proposed a novel 3D device struc- ture, where a column electrode is surrounded by a trench elec- trode [ ]. This design yields a much more homogeneous electric field with nearly pure radial dependence, eliminat- ing potential saddle points and low-field regions in the sensor center—common issues in conventional 3D column sensors.

Additionally, the capacitance of the BNL sensor is consid- erably lower than that of trench sensors because of its col- umn collecting electrode. In particular, in all of the afore- mentioned studies, internal gain was not incorporated into the design stage due to manufacturing technology constraints.

SENSOR DESCRIPTION AND SIMULATION We designed and fabricated a novel 3D trench-column sen- sor with internal gain by decreasing the widths of the col- umn and trench electrodes, and optimizing the doping strat- egy [ ]. The basic structure consists of an N column elec- trode surrounded by a P trench electrode, as illustrated in . The P electrode penetrates the 30 epitax- ial layer, while the tip of the N electrode is approximately away from the highly doping substrate to prevent early breakdown.

In this work, we investigated a 5 5 array of 3D trench- column sensors with a pixel size of 35 and a

trench width of 1 µ m . In order to conveniently perform sen- 59

sor characterizations, a temporary metal layer was deposited to connect the electrodes. As shown in Figure , the gold- colored region in the layout represents the temporary metal layer: the large pad on the right is connected to the P trench electrode, the three small pads on the left are each connected to five N column electrodes, and the ten pads along the top and bottom are each connected to a single N column elec- trode.

Layout of the 5 5 array 3D trench-column sensor with a pixel size of 35 and a trench width of 1 The basic cell was simulated using Sentaurus TCAD and the cross-sectional electric field distribution at a bias volt- age of 80 V is shown in Figure . Figure presents the 1D electric field at depths of 15 and 25 for dif- ferent bias voltages. The maximum electric field around the tip of the N electrode exceeds 200 kV/cm, where carriers can acquire sufficient energy to generate secondary electron-

hole pairs through impact ionization. This process enables 75

avalanche multiplication, serving as a gain mechanism for the 76

sensor [ Electric Field [kV/cm]

Bias Voltage = 10 V Bias Voltage = 20 V Bias Voltage = 40 V Bias Voltage = 60 V Bias Voltage = 80 V

Electric Field [kV/cm] Depth 15 Radius [ Radius [ elec- trode and the longitudinal middle of the P electrode for different bias voltages.

MEASUREMENT SETUP Four types of measurements were performed: Current- Voltage ( ), Capacitance-Voltage ( ), infrared Tran- sient Current Technology (TCT), and Sr-source measure- ments. measurements were conducted to characterize leak- age current and breakdown voltage. The test system com- prised a Keithley 2470 High Voltage SourceMeter and a Keithley 6482 Dual-Channel Picoammeter coupled to an Apollowave alpha-200CS probe station. During testing, the sensor was placed in a dark box on a Peltier-cooled thermal chuck, maintained at 20 C. Negative high voltage was applied via a probe needle to the pad connected to the P trench electrode, while leakage current was measured through another probe needle on the pad connected to five N column electrodes. measurements were used to analyze the sensor de- pletion behavior.

The setup included a Keithley 4200-scs semiconductor characterization system paired with a PW-800 probe station. The sensor was housed in a dark box on an ordinary chuck at room temperature. Negative high voltage was applied to the P trench electrode pad and capacitance was measured on the pad connected to five N column elec- trodes.

Charge collection and time resolution were evaluated us- ing TCT and -source measurements.

As shown in Fig- , the sensor was glued using double-sided conductive tape and then electrodes were wire bonded to the signal pad of the USTC amplifier board [ ], which was designed to mea- sure the time resolution of Low Gain Avalanche Detectors (LGADs) [ ]. Here we improved the transimpedance on the readout board, so the gain of the amplifiers is increased by a factor of 1.7, which is high enough to measure the expected small signals from the 3D sensor.

The TCT set-up is shown in Figure where the cus- tomized infrared laser with a wavelength of 1060 nm was used and the full width at half maximum (FWHM) of the well-focused laser spot was less than 11 . The laser per- pendicularly passes through the sensor from the top side, and the center of the laser was adjusted to be between the trench and column electrodes.

The set-up of TCT measurement system. The set-up of beta-particle measurement using a Sr radia- tive source is the same as described in detail elsewhere [

A Hamamatsu Photonics K.K. (HPK) Type 3.1 LGAD [ 22 , 121

] was used as the trigger and reference, whose timing res- olution is 60.62 ps (20 C, 180 V) and 49.71 ps (-30 C, 150

V). The amplifier boards with the sensor and the mechani- 124

cal tools, providing precise alignment of all components, are placed in an environmental chamber. The measurement is completed at 20 C and -30 C. The low temperature is neces- sary to reduce noise. On the other hand, the charge collection at different temperatures when the bias voltage is the same can also estimate whether the gain occurs or not indirectly,

because the impact ionization coefficient depends on the tem- 131

perature. In this measurement, multi-pixels are wire bonded, including the 15 pixels of raw 2, 3, 4 and the 2 pixels in the bottom right corner in Figure RESULTS AND DISCUSSION Leakage Current and Capacitance The measured curve at 20 C is shown in Figure The leakage current increases rapidly at a few volts, then flats at the level of several tens of nA. After 82 V, the current be-

gins to increase significantly, and the sensor eventually breaks 140

down. Figure displays the curve at room tem- perature. The test is performed using an AC signal amplitude of 30 mV and frequency of 1 MHz. The curve demonstrates two slops which relate to two different depletion regions: the trench-column region depletes at 2 V whereas the under-column region depletes at around 10 V.

Leakage Current [A] Capacitance [fF] Bias Voltage [V] Bias Voltage [V] ) curve at 20 C. (b) The capacitance versus bias voltage ( ) curve at room temperature.

Results of TCT measurements laser intensities when the bias voltage is 84 V. The numbers after "Laser" in the legend represent different laser intensities.

The smaller numbers correspond to the larger laser intensities.

It shows that the amplitude of averaged waveforms increases with the increase of intensity because of the more energy de- position in the sensor. The averaged waveforms at different bias voltages are shown in Figure and the amplitude in- creases with the bias voltage obviously.

The collected charge of each event is calculated by the in- tegration of the waveform divided by the calibration constant Amplitude [V] Amplitude [V] Laser50 Laser60 Laser70 Laser80 Laser82 Laser85 Laser86 Time [ns] Time [ns] of the amplifier board. Then, the distribution of charge is fitted with a gaussian function to extract the most probable value (MPV) as the collected charge under the corresponding measurement conditions. Figure shows the distribution of the collected charge at 84 V and Laser50, and the red line is the gaussian fitting line. The relationship of the collected charge and bias voltage is shown in Figure . When the bias voltage is in the range of 5-20 V, the slope of the curves increases slowly. That is because the sensor is gradually fully depleted and in the meantime the velocity of carriers also in-

creases with electric field ( < 20 kV/cm). Next, the charge 169

almost remains stable until 40 V, because the drift velocity of carriers is saturated, which does not increase with the elec- tric field any more. When the bias voltage is above 40 V, the charge increases more and more rapidly. Figure shows the collected charge scaled to the collected charge at 20 V.

It has an exponential dependence on the bias voltage, which indicates that the gain occurs before breakdown.

Regarding time resolution, the time-of-arrival (TOA) of

the sensor and the synchronization signal of the laser are 178

extracted using the Constant Fraction Discriminator (CFD) method (30 for 3D sensor and 50 for laser), which is an

efficient way to minimize the effect of time walk. A gaussian 181

function is applied to fit the TOA difference ( TOA). The time resolution ( ) can be calculated by equation and it is equal to the "Sigma" of the gaussian fitting function regard- less of the time resolution of the laser.

σ 3D = �

Reference voltage, and there are also three stages in the curve. Firstly, the time resolution decreases with the bias voltage because the sensor gradually fully depletes and the velocity of carri- ers increases with the electric field at low bias voltage. After the carrier velocity became saturated in almost all areas of the sensor, the time resolution became stable. Due to the gain oc- curring above 40 V, the time resolution decreases further, and it can reach around 31 ps. In this test, the laser is focused at the fixed position, so the time resolution is mainly dominated

Entries Entries Mean: 7458.22 Mean_Err: 0.97 Sigma: 30.72 Sigma_Err: 0.71 Entries: 1032 MPV: 82.88 Sigma: 0.07 Entries: 1032 Collected charge [fC] TOA [ps] Charge [fC] Laser50 Laser50 Bias Voltage [V] Bias Voltage [V] Time Resolution [ps] Rise Time [ns] Laser50 Laser50 Bias Voltage [V] Bias Voltage [V] -axis refers to the number of events per bin. (b) Distributions TOA. The "Sigma" is the standard deviation of the distribution of the TOA difference and the "Sigma Err" is the fitting error of the standard deviation. (c) The collected charge as a function of bias voltage. (d) Scale the collected charge to the collected charge at 20 V. (e) The time resolution as a function of bias voltage. (f) The rise time resolution as a function of bias voltage. by the jitter, excluding the contribution of the location of im- pact within cell [ ]. The rise time (10 ) of the aver- aged waveforms is also calculated. Figure shows the rise time as a function of bias voltage, and it varies around 1300 ps. The larger rise time and the shape of averaged waveforms indicate that the bandwidth of the amplifier board is not high enough.

Results of Sr-source measurements The 3D trench-column sensor is measured at 20 C and -30 C when the bias voltage is 84 V. The distributions of charge in Figures are fitted with a landau function con- voluted with a gaussian function to extract the most probable value (MPV) as the collected charge at 20 C and -30 C, re- spectively. Considering that the most probable value (MPV)

energy loss of the minimum ionizing particel (MIP) in 30 µ m 211

silicon is about 0.3 fC, the gain is about 4.17 at 20 C and 6.37 at -30 C, respectively. At the same bias voltage, the charge is higher in low temperature, which is related to the larger impact coefficient [ The time-of-arrival (TOA) of the 3D trench-column sensor and the reference are also extracted by the constant fraction descriminator (CFD) method (30 for 3D sensor and 50 LGAD). And the distributions of TOA are fitted by a gaus- sian function and are shown in Figure . The time resolution of 3D trench-column sensor is 231.89 10.18 ps C, 84 V) and 182.45 7.07 ps (-30 C, 84 V), respec- tively. With about 1.5 times larger gain at -30 C, the time res- olution reduces about 49.44 ps. Therefore, the better time res- olution is expected in the second batch of 3D trench-column sensors with a thicker active thickness. Besides, the present amplifier board needs to be optimized for time measurement of 3D sensors with less internal rise time than LGADs.

Entries Entries Mean: -240.70 Mean_Err: 12.22 Sigma: 236.23 Sigma_Err: 9.99 Entries: 537 MPV: 1.25 Sigma: 1.20 Entries: 537 TOA [ps] Entries Entries Mean: -238.01 Mean_Err: 9.31 Sigma: 189.10 Sigma_Err: 6.81 Entries: 516 MPV: 1.91 Sigma: 1.97 Entries: 516 Collected charge [fC] TOA [ps]

84 V and 20

C. (b) Distributions of TOA of the sensor working at

84 V and 20

C. (c) Distributions of collected charge of the sensor working at 84 V and C. (d) Distributions of TOA of the sen- sor working at 84 V and C. The "MPV" is the most probable value of the distribution of collected charge. The "Entries" on the -axis refers to the number of events per bin. And the "Sigma" is the standard deviation of the distribution of the TOA difference and the "Sigma Err" is the fitting error of the standard deviation.

CONCLUSION

In conclusion, both the TCT measurements with the infra- red laser (1060 nm) and the beta-scope measurements with 90 Sr-electrons confirm that the internal gain occurs as pre- dicted by the TCAD simulation. The TCT tests show that the largest collected charge at 84 V scaled to collected charge at

20 V is 28.68

0.07 and the time resolution at the fixed mea- surement position is 30.72 0.71 ps. The Sr-source tests

show that the gain for the minimum ionizing particle (MIP) is 237

about 4.17 at 20 C, 84 V and 6.37 at -30 C, 84 V. The gain is larger in low temperature, which is related to the larger impact coefficient. However, the time resolution for the min-

imum ionizing particle (MIP) need to be investigated further. 241

Anyway, the development of novel 3D trench-column sensors in this study provides a feasible technology to provide the in- ternal gain without adding the extra gain layer, like LGADs, promising advances in spatiotemporal (4D) tracking with low material in extreme experimental environments.

ACKNOWLEDGMENTS

This work is supported by the National Key R D Program of China under Grant 2023YFF0719600, General Program of National Natural Science Foundation of China under Grant

12375188. This work was partially carried out at the USTC

Center for Micro and Nanoscale Research and Fabrication.

We are also grateful for the support with equipment and tech- S. Parker, C. Kenney, and J. Segal, Nucl. Instrum. Meth. A , 328 (1997) C. Kenney, S. Parker, J. Segal, and C. Storment, IEEE Trans.

Nucl. Sci. , 1224 (1999) C. Kenney, S. Parker, and E. Walckiers, IEEE Trans. Nucl. Sci. , 2405 (2001)

[4] G. Pellegrini, M. Lozano, M. Ullán, R. Bates, C. Fleta, and 280

D. Pennicard, Nucl. Instrum. Meth. A 592 , 38 (2008) . 281

A. Zoboli, M. Boscardin, L. Bosisio, G.-F. Dalla Betta, C. Piemonte, S. Ronchin, and N. Zorzi, IEEE Trans. Nucl. , 2775 (2008)

[6] C. Da Via, M. Boscardin, G.-F. Dalla Betta, G. Darbo, C. Fleta, 285

C. Gemme, P. Grenier, S. Grinstein, T.-E. Hansen, J. Hasi, 286

C. Kenney, A. Kok, S. Parker, G. Pellegrini, E. Vianello, and 287

N. Zorzi, Nucl. Instrum. Meth. A , 321 (2012) S. Terzo, M. Boscardin, J. Carlotto, G.-F. Dalla Betta, G. Darbo, O. Dorholt, F. Ficorella, G. Gariano, C. Gemme,

G. Giannini, S. Grinstein, A. Heggelund, S. Huiberts, A. Kok, 291

O. Koybasi, A. Lapertosa, M. E. Lauritzen, M. Manna, R. Men-

dicino, H. Oide, G. Pellegrini, M. Povoli, D. Quirion, O. M. 293

Rohne, S. Ronchin, H. Sandaker, M. A. Abdulla Samy, B. Stugu, and L. Vannoli, Front. Phys.

Volume 9 - 2021 (2021), 10.3389/fphy.2021.624668

[8] M. Meschini, A. Cassese, R. Ceccarelli, L. Viliani, M. Di- 297

nardo, S. Gennai, D. Zuolo, A. Messineo, S. Parolia, A. Ebrahimi, D. Pitzl, G. Steinbrück, G. Dalla Betta, R. Mendicino, G. Alimonti, C. Gemme, M. Boscardin,

nical personnel by the IMECAS and the USTC. In addition, 254

this work performed in the framework of the CERN-DRD3 collaboration - WP2 project Novel silicon 3D-trench pixel de- tector fabricated on the 8-inch wafer utilizing CMOS process- ing technologies.

AUTHOR CONTRIBUTIONS Manwen Liu and Kuo Ma contributed equally to this work.

Device concept, design and manufacture were performed by Manwen Liu, Huimin Ji, Zheng Li, Zhihua Li and Jun Luo.

Measurements and data collection were conducted by Kuo Ma and De Zhang. Huimin Ji carried on the TCAD simula- tion. All authors contributed to the data analysis. The first draft of the manuscript was written by Manwen Liu and Kuo Ma. Yanwen Liu, Zheng Li, Zhihua Li and Jun Luo partici- pated to the draft writing, review and editing. Manwen Liu, Zhihua Li and Jun Luo were responsible for draft editing, su-

pervision, project administration and funding acquisition. 270

DECLARATIONS The authors declare that they have no conflict of interest.

BIBLIOGRAPHY and S. Ronchin, Nucl. Instrum. Meth. A (2020), G.-F. Dalla Betta, M. Boscardin, G. Darbo, R. Mendicino,

M. Meschini, A. Messineo, S. Ronchin, D. Sultan, and 304

N. Zorzi, Nucl. Instrum. Meth. A , 386 (2016) R. Mendicino, G. T. Forcolin, M. Boscardin, F. Ficorella, A. Lai, A. Loi, S. Ronchin, S. Vecchi, and G.-F. Dalla Betta, Nucl. Instrum. Meth. A , 24 (2019)

[11] F. Borgato, D. Brundu, A. Cardini, G. M. Cossu, G. F. 309

Dalla Betta, M. Garau, L. La Delfa, A. Lai, A. Lampis, A. Loi, M. M. Obertino, G. Simi, and S. Vecchi, Front. Phys.

Volume (2023), 10.3389/fphy.2023.1117575

[12] A. Lampis, F. Borgato, D. Brundu, A. Cardini, G. Cossu, 313

G.-F. Dalla Betta, M. Garau, L. La Delfa, A. Lai, A. Loi, M. Obertino, G. Simi, and S. Vecchi, J. Instrum. , C01051 (2023)

[13] C. Da Vià, G.-F. Dalla Betta, and S. Parker, Radiation Sensors 317

with 3D Electrodes (1st ed.) (CRC Press, 2019). G.-F. Dalla Betta and M. Povoli, Front. Phys.

Volume 10 - 2022 (2022), 10.3389/fphy.2022.927690 Z. Li, Nucl. Instrum. Meth. A , 90 (2011) M. Liu, H. Ji, W. Cheng, L. Zhang, Z. Li, B. Tang, P. Zhang,

W. Xiong, T. Vickey, E. G. Villani, Z. Li, D. Zhang, and J. Luo, 323

“Design, fabrication and initial test of a novel 3D-Trench sen- 324

sor utilizing 8-inch CMOS compatible technology,” (2025), Y. Okuto and C. Crowell, Solid-State Electron. , 161 (1975)

W. Maes, K. De Meyer, and R. Van Overstraeten, Solid-State Electron. , 705 (1990) J. Ge, C. Li, D. Zhang, Y. Yang, A. Wang, X. Yang, H. Liang, and Y. Liu, Nucl. Instrum. Meth. A , 167222 (2022)

[20] G. Pellegrini, P. Fernández-Martínez, M. Baselga, C. Fleta, 332

D. Flores, V. Greco, S. Hidalgo, I. Mandi´c, G. Kramberger, D. Quirion, and M. Ullan, Nucl. Instrum. Meth. A (2014) C. Li, X. Yang, J. Ge, T. Wang, X. Zheng, Y. Sun, and Y. Liu, Nucl. Instrum. Meth. A , 167008 (2022) X. Yang, S. Alderweireldt, N. Atanov, M. Ayoub, J. B. G. da Costa, L. C. García, H. Chen, S. Christie, V. Cindro, H. Cui, G. D’Amen, Y. Davydov, Y. Fan, Z. Galloway, J. Ge,

C. Gee, G. Giacomini, E. Gkougkousis, C. Grieco, S. Grin- 341

stein, J. Grosse-Knetter, S. Guindon, S. Han, A. Howard, Y. Huang, Y. Jin, M. Jing, R. Kiuchi, G. Kramberger, E. Kuw- ertz, C. Labitan, J. Lange, M. Leite, C. Li, Q. Li, B. Liu, J. Liu, Y. Liu, H. Liang, Z. Liang, M. Lockerby, F. Lyu, I. Mandi´c, F. Martinez-Mckinney, S. Mazza, M. Mikuž, R. Padilla, B. Qi, A. Quadt, K. Ran, H. Ren, C. Rizzi, E. Rossi, H.-W. Sadrozin- ski, G. Saito, B. Schumm, M. Schwickardi, A. Seiden, L. Shan, L. Shi, X. Shi, A. S. C. Ferreira, Y. Sun, Y. Tan, A. Tri- coli, G. Wan, M. Wilder, K. Wu, W. Wyatt, S. Xiao, T. Yang, Y. Yang, C. Yu, L. Zhao, M. Zhao, Y. Zhao, Z. Zhao, X. Zheng, and X. Zhuang, Nucl. Instrum. Meth. A , 164379 (2020) Z. Galloway, V. Fadeyev, P. Freeman, E. Gkougkousis, C. Gee, B. Gruey, C. Labitan, Z. Luce, F. McKinney-Martinez, H.- W. Sadrozinski, A. Seiden, E. Spencer, M. Wilder, N. Woods,

A. Zatserklyaniy, Y. Zhao, N. Cartiglia, M. Ferrero, M. Man- 356

durrino, A. Staiano, V. Sola, R. Arcidiacono, V. Cindro,

G. Kramberger, I. Mandi´c, M. Mikuž, and M. Zavrtanik, Nucl. 358

Instrum. Meth. A , 19 (2019) G. Kramberger, V. Cindro, D. Flores, S. Hidalgo, B. Hiti,

M. Manna, I. Mandi´c, M. Mikuž, D. Quirion, G. Pellegrini, 361

and M. Zavrtanik, Nucl. Instrum. Meth. A 934 , 26 (2019) . 362

C. R. Crowell and S. M. Sze, Appl. Phys. Lett. , 242 (1966)