1. A Preliminary Study of MAPS with N Type High Resistivity Epitaxial Layer
Min FU
(co-PhD student of IPHC, Strasbourg, France
and DLUT, Dalian, China)
Dalian University of Technology, Dalian, China

2. OUTLINE
Background
Simulation tool
Simulated structures
Static properties simulation
Transient simulations
Conclusions

3. MAPS Introduction CMOS Monolithic Active Pixel Sensors
for light imaging: late 1980’s
E. R. Fossum, “CMOS image sensors :electronic camera-on-a-chip”, IEEE Trans. On Electron Devices 44 (10) (1997)
Basic pixel electronics schemes (photodiode, 3 or 4 transistors, transfer gate…)  all these elements are still bases of today’s digital cameras

4. From digital cameras to particle tracking
Properties:
Standard commercial CMOS technology
Sensor and signal processing integrated in the same silicon wafer
Signal created in low-doped epitaxial layer (typically ~10-15 μm)
Charge collection mainly through thermal diffusion (~100 ns), reflective boundaries at p-well and substrate
Charge sensing in n-well/p-epi junction
100% fill-factor
High granularity
Low power dissipation
Substantial radiation tolerance
Thinning available as standard post-processing
Only NMOS transistors inside pixels
B. Dierickx, G. Meynants, D. Scheffer “Near 100% fill factor CMOS active pixel sensor”, Proc. of the IEEE CCD&AIS Workshop, Brugge, 1997
Twin - tub (double well), CMOS process with epitaxial layer

5. 300 µm MIMOSA 26 architecture 1152x576 ~ 0.7 Mpixels pitch 18.4 µm
Analog outputs
for test only
Pixel array:
1152x576 ~ 0.7 Mpixels
pitch 18.4 µm
→ Sensitive surface 10.6x21.2 mm2
1152 discriminators
zero-suppression
control (JTAG+DAC)
Memory for binary out
Row sequencer
~3 mm
J.Baudot – Results of MIMOSA 26 – NSS 2009 4

6. Different resistivity epi-layer?
N type or P type?
Fig. 1. Schematic cross section of a MAPS pixel. When a charged particle penetrates through the sensor, there will be some genereated charges along the track. Most of the generated charges are collected by the positive biased N-well, the grounded P-wells and the substrate, respectively, and the rest are diffused and recombined. When the electrons diffuse to the P++ substrate, most of them will be reflected because of the potential barrier between the substrate and the epi-layer. On the other hand, some generated electrons can diffuse from the substrate into the epi-layer. Different detector parameters, including the thickness of the epitaxial layer, the size of a pixel and collecting diodes and number of diodes per pixel, were investigated.

7. Simulation tools
The charge collection efficiency can be evaluated using physical level simulator Sdevice from the Synopsys-SENTAURUS package,
The charge collection is traced as a relaxation process of achieving the equilibrium state after introducing an excess charge emulating passage of the ionising particle
The device is described in three dimensions by a mesh generated using the analytical description of doping profiles and the boundary definition corresponding to the real device,

8. Simulated structures
As the main sensitive region, epi-layer is the key of the study. In the single pixel structure, the thickness of the epi-layer is fixed to 15 µm and the values of resistivity are altered by different dopants and concentrations. Totally, six epi-layers are utilized in the single pixel structure. Until now, only P type epitaxial wafers but no N type ones have been used to fabricate MIMOSA-26 chips, so there are only three matrix structures are selected to evaluate collection performance.
Table I&II. Simulated structures. Six single pixel structure and three matrix.

9. Fig. 2. Structure of a single pixel model
Fig. 2. Structure of a single pixel model. The cross section is achieved by slicing along the central axis of the pixel. The substrate is removed and
only a 0.3 µm P++ doped layer is reserved at the bottom of the epi-layer.
Fig. 3. Structure of a matrix model. 3×3 pixels are built in the middle of the structure. Because the default boundary used in the SD solver is an ideal reflective condition (Neumann condition), the matrix is surrounded by an auxiliary part of silicon and oxide belts to avoid the overestimation of charge collection.

10. Static properties--lifetime
Fig. 4. Simulated doping profiles and electron lifetime . The curves are achieved by slicing along the central axis of the pixel. In the epi-layers, the range of doping concentration is from 1.1×1013 1/cm3 to 1.3×1014 1/cm3 and the resistivity is correspondingly altered from 400 Ω·cm to 100 Ω·cm. But the electron lifetime is just only changed from 9.93 µs to 9.99 µs

11. Static properties—potential&field
Fig. 5. Simulated electric potential profiles. From 5 µm to 10 µm, where the differently doped epi-layers creat different potential values from -0.3 V to 0.9 V. Although the two ends of the curves are fixed to 1.2 V and -0.5 V, the gradients of the potential are extremely distinct in the N type ones.
Fig. 6. Simulated electric field profiles. In the primary part of the epitaxial layer, from 2 µm to 8 µm, the fields in P type pixels are almost zero, but in N type ones, the values are more than 500 V/cm at the same position, especially the 400 Ω·cm one and the 200 Ω·cm one.

12. Fig. 7. Cross-sections of the electric potential distribution in pixels. Top-left is N type 100 Ω·cm; top-middle is N type 200 Ω·cm; top-right is N type 400 Ω·cm; bottom-left is P type 100 Ω·cm; bottom-middle is P type 200 Ω·cm and bottom-right is P type 400 Ω·cm.

13. Fig. 8. Cross-sections of the electric field distribution in pixels
Fig. 8. Cross-sections of the electric field distribution in pixels. Top-left is N type 100 Ω·cm; top-middle is N type 200 Ω·cm; top-right is N type 400 Ω·cm; bottom-left is P type 100 Ω·cm; bottom-middle is P type 200 Ω·cm and bottom-right is P type 400 Ω·cm.

14. Fig. 9. Details of the depletion region in pixels
Fig. 9. Details of the depletion region in pixels. The depleted region is shadowed and surrounded by white lines. Obviously, the depletion regions in the N type epi-layers are much wider than those in the P type ones and the higher resistivity layers can be depleted more easily than the lower resistivity ones.

15. Transient simulation
Fig. 10. Schematic of sub-layers and cubes. In order to simulate the situation under the Fe-55 radiation, a cuboid volume within the central pixel is defined as an event region and separated into 20 sub-layers equally and then each sub-layer is partitioned into 16 cubes whose dimensions are 2.3×2.3×2.3 µm3. Then, there are 320 cubes in total and in each one a 5.9 keV photon is absorbed and converted to 1640 pairs of electrons and holes. Because in the bottom of substrate, the output signals can be neglected (less than the predefined threshold), the bottom 10 unnecessary sub-layers are removed.

16. Simulation — neighbouring diffusion
Fig. 11. The percentage of the seed point o5 is 34%, 43% and 56% when the epi-layer is P type NR, P type HR and N type HR, respectively. In other pixels, the collected electrons
are suppressed about several percents.

17. Simulation — collection time
Table III. Collection time. Because the charge collection mostly depends on the location of ionising radiation incident, four cubes are selected to evaluate the collection time. The first one is sub-layer 01 cube 16, which is the closest one to the o5; the second one is sub-layer 05 cube 07 at the middle of the epi-layer; the third one is sub-layer 06 cube 01 standing for the farthest position in the epi-layer; the last one is sub-layer 10 cube 01 which could be looked as the farthest one in the substrate. Obviously, the N type HR epilayer could reduce the collection time greatly due to its wide internal electric field.

18. Fig. 11. Details of the charge collection efficiency of different epi-layer. The
”Fe-55” and ”SIM” data come from real experiments and the simulations,
respectively. The experimental data are labelled by the top-right axes and the others by the bottom-left ones. The collection efficiency are 73% in NR structures, 88% in P type HR ones and 98% in N type HR ones respectively. It is obvious that the N type HR sensors will be the best one.

19. Conclusions and proposals:
In the static simulations, it can be concluded that the
distributions of electric field in the N type high resistivity epi-layers help the N-wells to collect the electrons. Detailed collection performances of N type HR structures are illustrated by the transient simulation. Firstly, the electron diffusion among the neighboring pixels is suppressed evidently and the collection percentage of the seed point pixel is increased. Secondly, the charge collection time in seed point decreases obviously. Finally, the collection efficiency can be improved.
Overall, it can be concluded that the N type HR epitaxial wafer is one good choice for MAPS detector manufacture and it is reasonable to expect for improved collection performance.

20. merci pour votr attention!
FCPPL 2011
Jinan, China

21. Another choice — Gaussian doping
Preliminary results
Still working on