--- title: Exponential Pulsation Ratio (EPR) Overview url: https://www.emergentmind.com/topics/exponential-pulsation-ratio-epr type: topic --- # Exponential Pulsation Ratio (EPR) Overview Exponential Pulsation Ratio (EPR) is not used uniformly across the cited literature. In the stochastic pulse-superposition model of uncorrelated exponential pulses, EPR denotes the intermittency parameter $$ \gamma=\frac{\langle \tau\rangle}{\langle w\rangle}, $$ with \(\langle \tau\rangle\) the average pulse duration and \(\langle w\rangle\) the average waiting time [1702.00105]. In a pulsar polarization model with orthogonal polarization modes, the paper does **not explicitly use the term "Exponential Pulsation Ratio (EPR)"** as a separate parameter; instead, the central dimensionless quantity is the mean intensity ratio $$ M=\frac{\mu_1}{\mu_2}, $$ which parameterizes the normalized observables [2209.00743]. A further, terminologically distinct usage appears in work on an **EPR Hamiltonian**, where EPR names the optimization problem rather than a ratio [2512.09896]. ## 1. Terminological scope The cited works attach the label EPR to different mathematical objects. One paper identifies EPR with an intermittency parameter in a filtered-Poisson or shot-noise model of exponential pulses. Another discusses exponential statistics in pulsar polarization without adopting EPR as formal notation, while a third uses EPR as the name of a Hamiltonian optimization problem [1702.00105] [2209.00743] [2512.09896]. | Context | Quantity | Role | |---|---|---| | Super-position of uncorrelated exponential pulses | \(\gamma=\langle\tau\rangle/\langle w\rangle\) | Intermittency parameter or EPR | | Pulsar orthogonal polarization modes | \(M=\mu_1/\mu_2\) | Central parameter; the paper does not explicitly use EPR | | EPR Hamiltonian problem | \(H(G)\) | EPR names the Hamiltonian problem | A common source of confusion is therefore terminological rather than mathematical. The literature summarized here supports the narrower statement that EPR is a standard symbol only within the pulse-overlap model, whereas the pulsar and Hamiltonian papers use related or distinct notation. ## 2. EPR as an intermittency parameter in exponential pulse superposition In the pulse-superposition framework, the stochastic process is $$ \Phi_K(t)=\sum_{k=1}^{K(T)} A_k\,\phi\!\left(\frac{t-t_k}{\tau_k}\right), $$ where \(A_k\) are pulse amplitudes, \(\tau_k\) pulse durations, \(t_k\) arrival times, \(K(T)\) the number of pulses in an interval \(T\), and \(\phi\) a normalized pulse shape satisfying $$ \int_{-\infty}^{\infty}\phi(\theta)\,d\theta=1. $$ All random variables are assumed independent and uncorrelated across pulses [1702.00105]. Within this model, the mean value and variance are $$ \Phi=\frac{\langle\tau\rangle}{\langle w\rangle}\,I_1\langle A\rangle, \qquad \sigma^2=\gamma I_2\langle A^2\rangle, $$ where \(I_1=\int\phi(\theta)\,d\theta\), \(I_2=\int[\phi(\theta)]^2\,d\theta\), and $$ \gamma=\frac{\langle\tau\rangle}{\langle w\rangle}. $$ The relative fluctuation level is $$ \frac{\sigma^2}{\Phi^2} = \frac{1}{\gamma}\frac{I_2}{I_1^2}\frac{\langle A^2\rangle}{\langle A\rangle^2}. $$ Small \(\gamma\) yields strongly intermittent, burst-dominated fluctuations, while large \(\gamma\) yields quasi-Gaussian behavior [1702.00105]. In this usage, EPR measures pulse overlap rather than spectral shape. The parameter controls intermittency statistics and fluctuation amplitude, but the same paper states that the auto-correlation function and the frequency power spectral density are independent of the degree of pulse overlap and thereby the intermittency of the stochastic process. This sharply separates overlap statistics from spectral asymptotics. ## 3. Auto-correlation and spectral structure under exponential pulses For the normalized variable, the auto-correlation function and frequency power spectral density are $$ R_{\widetilde{\Phi}}(r) = \frac{1}{\langle\tau\rangle} \int_0^\infty \tau P_\tau(\tau)\, \rho_\phi\!\left(\frac{r}{\tau}\right)d\tau, $$ and $$ \Omega_{\widetilde{\Phi}}(\omega) = \frac{1}{\langle\tau\rangle} \int_0^\infty \tau^2 P_\tau(\tau)\, \varrho_\phi(\tau\omega)\,d\tau, $$ with $$ \rho_\phi(\theta)=\frac{1}{I_2}\int \phi(\chi)\phi(\chi+\theta)\,d\chi, \qquad \varrho_\phi(\vartheta)=\frac{1}{I_2}\lvert\varphi(\vartheta)\rvert^2. $$ For constant pulse duration, these reduce to $$ R_{\widetilde{\Phi}}(r)=\rho_\phi\!\left(\frac{r}{\langle\tau\rangle}\right), \qquad \Omega_{\widetilde{\Phi}}(\omega)=\varrho_\phi(\langle\tau\rangle\omega) $$ [1702.00105]. For a one-sided exponential pulse, $$ \phi(\theta)=\Theta(\theta)e^{-\theta}, $$ the normalized auto-correlation is $$ \rho_\phi(\theta)=e^{-|\theta|}, $$ and the spectrum is Lorentzian, $$ \varrho_\phi(\vartheta)=\frac{2}{1+\vartheta^2}. $$ For constant duration, $$ \frac{1}{2}\Omega_{\widetilde{\Phi}}(\omega)=\frac{1}{1+(\langle\tau\rangle\omega)^2}, $$ which is flat at low frequencies and scales as \(1/\omega^2\) at high frequencies. The algebraic tail is demonstrated to result from the discontinuity of the pulse at its starting point [1702.00105]. For the two-sided exponential pulse $$ \phi(\theta;\lambda) = \Theta(-\theta)e^{\theta/\lambda} + \Theta(\theta)e^{-\theta/(1-\lambda)}, \qquad 0<\lambda<1, $$ the spectrum is $$ \varrho_\phi(\vartheta;\lambda) = \frac{2}{[1+(1-\lambda)^2\vartheta^2][1+\lambda^2\vartheta^2]}. $$ The symmetric case \(\lambda=1/2\) gives the square of a Lorentzian and a \(1/\omega^4\) tail, while strongly asymmetric pulses give a broken power law with two scaling regions. Random pulse durations lengthen effective correlation times and increase low-frequency power, but the high-frequency power-law tail remains unchanged. Additive white noise leads to a flat spectrum at high frequencies [1702.00105]. ## 4. Exponential mode statistics in pulsar polarization A separate use of exponential statistics appears in the polarization of pulsar radio emission. The model assumes the **incoherent superposition of two orthogonally polarized modes**, each fully polarized and independent in its intensity fluctuations, with mode intensities \(X_1\) and \(X_2\) and mean values \(\mu_1\) and \(\mu_2\). In the linear model, $$ {\rm I}=X_1+X_2, \qquad {\rm Q}=X_1-X_2, $$ while \(U\) and \(V\) vanish; elliptical generalizations are also treated [2209.00743]. To account for heavy modulation and observed asymmetries, the mode intensities are taken to be exponential random variables, $$ f_j(x)=\frac{1}{\mu_j}\exp\!\left(-\frac{x}{\mu_j}\right), \qquad x\ge 0, $$ for \(j=1,2\). For an exponential distribution, the mean and standard deviation are equal, \(\mathrm{E}[X]=\mathrm{Std}[X]=\mu\), and the modulation index is \(\beta=1\). The central parameter is $$ M=\frac{\mu_1}{\mu_2}. $$ The paper states that all normalized observable parameters depend only on \(M\), not on the absolute intensities [2209.00743]. The same source explicitly notes that it does **not** introduce "Exponential Pulsation Ratio" as a separate parameter. Instead, the ratio \(M\) plays the organizing role. This suggests that any identification of EPR with \(M\) is interpretive rather than terminological. ## 5. Distributional consequences and observable parameters in the pulsar model Under the exponential-mode assumption, the total-intensity distribution for unequal means is $$ f_{\rm I}(x) = \frac{1}{\mu_1-\mu_2} \left[ \exp\!\left(-\frac{x}{\mu_1}\right) - \exp\!\left(-\frac{x}{\mu_2}\right) \right], \qquad x\ge 0, $$ and for \(\mu_1=\mu_2=\mu\), $$ f_{\rm I}(x)=\frac{x}{\mu^2}\exp\!\left(-\frac{x}{\mu}\right), \qquad x\ge 0, $$ which is a Gamma distribution. The Stokes-\(Q\) distribution is an asymmetric two-sided exponential, $$ f_{\rm Q}(x)= \begin{cases} \frac{1}{\mu_1+\mu_2}\exp\!\left(-\frac{x}{\mu_1}\right), & x\ge 0,\\[4pt] \frac{1}{\mu_1+\mu_2}\exp\!\left(\frac{x}{\mu_2}\right), & x<0. \end{cases} $$ The linear polarization amplitude satisfies $$ f_{\rm L}(x) = \frac{1}{\mu_1+\mu_2} \left[ \exp\!\left(-\frac{x}{\mu_1}\right) + \exp\!\left(-\frac{x}{\mu_2}\right) \right], \qquad x\ge 0 $$ [2209.00743]. With $$ m=\frac{\mu_1-\mu_2}{\mu_1+\mu_2}=\frac{M-1}{M+1}, $$ the fractional linear polarization \(z=L/I\) has density $$ f_{\rm ml}(z) = \frac{(1-m^2)(1+m^2z^2)}{(1-m^2z^2)^2}, \qquad 0\le z\le 1, $$ which reduces to a uniform distribution when \(M=1\). Additional normalized observables are $$ \nu_1=\frac{M}{M+1}, \qquad \nu_2=\frac{1}{M+1}, $$ for the occurrence frequencies of the primary and secondary modes, $$ \bar{\rm L}=\frac{M^2+1}{(M+1)^2}, \qquad m=\frac{M-1}{M+1}, $$ for normalized mean linear polarization and normalized mean Stokes \(Q\), $$ \beta=\frac{\sqrt{M^2+1}}{M+1}, \qquad r_{IQ}=\frac{M^2-1}{M^2+1}, $$ for the total-intensity modulation index and the intensity-correlation coefficient, and $$ ml=1+\frac{1-m^2}{2m^2}\ln(1-m^2), \qquad 0