PIXEL Slow Simulation Status Report Xin Li 7/30/2008 Update Since the UC/Davis Collaboration • Discovered a mistake and rewrote the code to save time. – Unit error on the step length of random walk. (m to cm). – Code is written in macro instead of compiled style. • After the correction, 10~20min per track – Before the correction, a few hours per track. • Development of a simplified and faster slow simulator. – 3 Seconds per track. • Plan of Building a lookup table which is fast enough to be plugged into STAR software. PIXEL Geometry Model: a chip of 640(x)x640(z) PIXEL array PIXEL size: 30um(x) x 50um(y) x 30um(z) Diode size: 4.5um x 2um x 4.5um 30um Readout electronics layer: 6um Diode layer: 2um Epi layer: 14um Sub layer: 28um 50um y z 30um x Simulation Result After Update Incident angle 45 Incident angle 0 In the sum of collected electrons: 45 0 Contribution from sub: 21% Contribution from epi: 68% Contribution from diode: 11% y z ( much larger contribution from sub (21%) compared to the previous result (1%) due the correction of the step length). Comparison with Experimental and Simulation Result Reference: “Modeling, Design, and Analysis of Monolithic Charged particle Image Sensors” by Shengdong Li, Univ. of California, Irvine Shendong’s experiment and MC comparison STAR test result, from Howard Matis Simplified Slow Simulator •Any ionized electron from any track is independent of each other. •Any electron generated at a specific location has a certain probability to be collected by each different pixel. • One can map out the probability using the real slow simulator mentioned in previous slides. • This map is a function of (x, y, z, theta, phi) , where x, y, z is the origin of the electron, theta, phi are the direction of the first step of random walk during electron diffusion. • Since the step length is very small (10-9m) and direction at every step is totally random in space, the direction of the first step has little effect on the map. Then the map can be only a function of x, y, z. • Segment the central pixel of a chip into fine 3-D grid and generate the map for electron from every grid point The map of electron originated from other pixels can be derived directly from this map through coordinate transformation. • The collection probability of electrons from a track can be obtained from these maps which lead to the final number of electrons collected in each pixel. How does simulator work Establish a coordinate system containing the 640x640 PIXEL array in one chip, with central PIXEL of the chip as origin. Generate electrons in a 3-D grid in the central pixel. calculate collection probability in all pixels on a chip for electron generated in a single grid point. The edge size of the grid unit is 1um along x, y, z axis. Totally there are 30 x 30 x 50 = 45000 grid points. build a look-up table containing the probability distribution for all the grid point in the central pixel. Get collected electron distribution of a track: First find out in which PIXELs the electrons are generated. make coordinate transformation from the central pixel to these pixels and calculate collection probability of every electron in every PIXEL using look-up table. Add them up and obtain the overall collected electron distribution. Code location: /star/u/lixin044/HFT/sample2/ Further Simplification • First we can ignore electrons generated in the diode layer (2um), since electrons will be collected by nwell or absorbed by pwell in this layer. • Second, according to the simulation result, we can ignore electrons generated 19um deep in the sub layer. • So in total 50um thickness (y axis) of a pixel, only need to make samples in 33um (19um sub + 14um epi). If one sample per 1um along x, y, z axis, totally there are 30x30x33=29700 samples. 30um Sampling epi region (33um) sub 50um 19um deep y sub Layer thickness (cm) epi z 30um x Result Comparison between Original and Simplified Slow Simulator Result along x axis Pixel ID Result along z axis Ultimate Simulator •The goal is to build the ultimate slow simulator that can simulate a few hundred tracks per second. • The Simplified Slow Simulator is still not fast enough to fit in STAR software but we can use it to build the ultimate slow simulator. • The ultimate simulator will contains about ten 3-D lookup table of collected electron distribution for 640 x 640 pixel array with track hit on central pixel, make samples as function with parameters r(0~15um), (0~90), (0~360). Here r is the distance from track incident position to the origin (center of the PIXEL), , are the incident angle of the track. • The details will be given in another talk. r Thank you