Integrating sphere

An integrating sphere is an optical cavity with diffusely reflecting interior walls and ports for illumination, samples or detectors. Repeated wall reflections mix the incident radiant flux, reducing sensitivity to where it first strikes. Baffles prevent a receiver from directly viewing an unwanted source or illuminated patch; port size and wall reflectance remain part of the response model.^[ref]

An original encounter model illustrates the effect of openings. Suppose a well-mixed reflection retains fraction $\rho$ and fraction $f$ of the reflected light escapes through ports before the next wall encounter. The integrating-sphere port-loss factor $1-f$ combines with wall reflectance to give a survival fraction $q=\rho (1-f)$. If the first wall encounter carries energy $U$, successive encounters carry $U,qU,q^{2}U,\dots$, so their summed incident energy is $U/(1-q)$.

With stipulated $\rho =0.98$ and $f=0.02$, this sum is $25.25U$. Enlarging the openings to $f=0.05$ reduces it to $14.49U$. These are repeated encounters with the same initially supplied energy, not optical gain: absorption and escape together still account for that original energy. The example counts incident wall encounters from the first one and does not assert a calibrated detector multiplier.

An actual sphere measurement compares its detector signal with a calibrated input or reference. Coating wavelength dependence, ports, sample substitution and temporal response can affect the conversion. The primary guide discusses reflectance/transmittance arrangements and their measurement geometry.^[ref] An integrating sphere is therefore useful apparatus, rather than a guarantee that every outgoing component has been measured without correction.

The integrating-sphere port-loss factor $1-f$ is the fraction of a well-mixed reflected field that next encounters the wall rather than an absorbing opening, where $f$ is total port area divided by the internal spherical area. It is a geometric retention factor, not wall reflectance. With uniform wall reflectance $\rho$, successive wall encounters retain fraction $q=\rho (1-f)$. The primary sphere derivation explicitly separates the port fraction from reflectance and generalizes to nonzero port reflectances.^[ref]

An original energy-accounting example starts with energy $U$ incident on the wall after the input beam's first impact. At each encounter fraction $1-\rho$ is absorbed; fraction $\rho f$ reflects and escapes; fraction $q$ reaches the wall again. Summing the geometric series givesThe two fractions sum to one. For stipulated $\rho =0.96$ and $f=0.04$, about $0.5102U$ is ultimately absorbed and $0.4898U$ escapes. Thus a four-percent port fraction does not mean that only four percent ever exits: repeated reflections give repeated opportunities. This is the integrating-sphere escape probability in the stated ideal encounter model.

The input convention matters: these fractions start at a wall encounter and omit any input beam that escapes without striking that wall. Real ports can contain reflective samples or detectors, and baffles change direct paths. A sphere experiment can vary an aperture while monitoring a stable injected radiant flux, but its response must include those surfaces and the receiver field of view. The factor $1-f$ alone is not a detector calibration or a universal collection efficiency.

The integrating-sphere escape probability is the chance that a photon in a specified integrating sphere model eventually leaves through a port rather than being absorbed by the wall. It depends on both the port area fraction $f$ and wall reflectance $\rho$, as well as where the model begins. In the ideal uniformly diffuse model with nonreflecting ports, the integrating-sphere port-loss factor is $1-f$ and the survival probability between successive wall encounters is $q=\rho (1-f)$.^[ref]

Start immediately before a wall encounter. The photon is absorbed there with probability $a=1-\rho$, reflects and escapes before the next wall encounter with probability $e=\rho f$, or returns to the wall with probability $q$. These satisfy $a+e+q=1$. Under the same well-mixed encounter rule each time, eventual escape at encounter $n\ge 1$ has probability $q^{n-1}e$. Summing the geometric series givesSimilarly, $P_{\mathrm{abs}}=(1-\rho )/(1-q)$ and $P_{\mathrm{esc}}+P_{\mathrm{abs}}=1$. These are an original probability derivation of the guide's repeated-reflection bookkeeping, not an additional source of optical energy.

For an original numerical example, $\rho =0.96$ and $f=0.04$ give $q=0.9216$, soAlthough only four percent of the spherical area is open, almost half the initial photon ensemble eventually exits: surviving photons get repeated opportunities. After $N$ completed encounters, the escaped fraction is $P_{\mathrm{esc}}(1-q^N)$, the absorbed fraction is $P_{\mathrm{abs}}(1-q^N)$, and the fraction still inside is $q^N$. The three fractions sum to one for every $N$.

The stopping encounter has a geometric distribution with termination probability $1-q$. Its expected count is $1/(1-q)$, about 12.76 in this example. This counts wall encounters; it does not specify elapsed time without a model of path lengths and propagation.

Boundary cases clarify the assumptions. With $\rho <1$ and $f=0$, every photon is eventually absorbed. With $\rho =1$ and $f>0$, every photon eventually escapes. If both $\rho =1$ and $f=0$, the ideal photon never terminates and the formula's $0/0$ is inapplicable. Starting after a successful first reflection instead would also give a different conditional escape probability; the starting point must accompany the result.

A measurement can vary a known aperture while injecting a stable radiant flux that first strikes the wall. A monitored detector then tests the apparatus response against a model including its viewing geometry. A single detector signal is not automatically the total energy escaping through all ports. Real coatings, reflective port fixtures and direct paths require corrections; baffles prevent the receiver from directly viewing an unusually bright first-illumination patch.^[ref] The probability above is therefore a useful ideal-model prediction, not a universal detector efficiency.