Read the explanation
The native page stores one dimensional real and imaginary arrays with two hundred fifty six grid points, despite its two dimensional title. In its tunneling reset, the packet center is index sixty four. At momentum parameter three point two, real and imaginary components share a Gaussian envelope and a phase. Compare native width parameters ten and fifty. The source divides both arrays by the square root of their total squared magnitude. Each held density curve therefore has a discrete probability sum of one. On the same scales, a wider envelope spreads that probability over more grid samples and lowers its peak. Both plots use one point eight pixels per index and six thousand pixels per probability. This is exact source discrete normalization, not a continuous integral or proof of physical propagation. The finite grid clips part of the wider packet tail, so the source parameter is not automatically the exact measured standard deviation. No unitary conservation claim follows just from normalizing the arrays. The native barrier width slider controls an inclusive loop over the one dimensional potential array. Its center index is one hundred twenty eight. The start is center minus the floor of half the width, and the end is center plus that same quantity. Both endpoints are included. A native width of five therefore writes indices one hundred twenty six through one hundred thirty, which is five cells. A native width of eighty writes eighty eight through one hundred sixty eight, which is eighty one cells. Each potential rectangle here uses four pixels per occupied cell and forty pixels per assigned potential unit. At native barrier height two point five, both are one hundred pixels tall. The wider rectangle has the exact larger number of source array assignments, rather than an assumed continuous width. This is a causal comparison of what the native slider actually writes. Although some page labels use nanometers and electron volts, these figures teach the source indices and assigned numbers without asserting that its toy engine is calibrated to those physical units. The same native canvas has two different source actions. With the detector probe off, pointer input paints potential. The vertical mapping subtracts pointer height from a baseline at three quarters of canvas height and divides by twenty five. A contact one hundred pixels above that baseline assigns center potential four. The brush writes offsets minus four through plus four around the selected index. Every assigned value is the center value multiplied by the exponential of minus offset squared divided by four. The held nine bars show those exact values at forty pixels per assigned potential unit. The center bar is one hundred sixty pixels high and the outer bars are much shorter. This is a potential profile, so the bar values do not need to sum to one. With the probe on, the same native contact changes detector position instead of painting potential. The original page advances an explicit numerical update and normalizes each frame. Its code does not actually solve the implicit Crank Nicolson system suggested by one comment. Its detector also uses random source rules. These comparisons do not establish a physical measurement law or calibrated quantum experiment.