Observable Channels of Light The Closure Criterion and Its Robustness Test

SRE defines light as the mutually non-annihilating topological residue on the causal intersection of joint execution chains. The definition is in hand; the next question is practical: can this thing be read out? This paper examines three candidate channels one by one—decay rate, interference contrast, and the loop sign product—and reaches one negative and one positive conclusion. The first two are structurally unusable. Both demand a continuous scale: the rate channel needs a difference of decay rates, the contrast channel needs a coherence length. SRE contains no continuous scale, so any scheme that requires one degenerates into measuring a parameter that was inserted by hand. Two numerical facts are given: separating two exponentials requires a dimensionless span above $3$, and the theory cannot supply that span; the measured contrast decay length is fixed by the per-edge transmission rate and is independent of the criticality parameter. The third works. The closure criterion is a sign product taking only $+1$ or $-1$; it needs no calibration, no fitting, and is immune to weights. Six families of controls, 23 cases in total, all pass, and one by-product emerges: on a closed background complex, the number of defects must be even. SRE 把光定义为联合执行链因果交集上的互不湮灭拓扑残余。定义有了,接下来的问题很实际:这个东西能不能被读出来? 本文逐条检验三条候选通道——衰减速率、干涉对比度、闭合符号积——并给出一个否定加一个肯定的结论。 前两条在结构上不可用。它们都要求一个连续尺度:速率通道要一个衰减率差,对比度通道要一个相干长度。SRE 不含任何连续尺度,凡是需要它的方案都会退化成在测量自己输入进去的参数。本文给出两条数值证据:双指数分离需要无量纲跨度达到 $3$ 以上,而理论给不出那个跨度;对比度的衰减长度实测由逐边传输率决定,与临界性参数无关。 第三条可用。闭合判据是一个符号乘积,取值只能是 $+1$ 或 $-1$,不需要标定、不需要拟合、不受权重影响。本文用六组对照检验它的稳健性,共 23 个情形全部通过,并给出一个副产品:在闭合的背景复形上,缺陷数必为偶数。