Write $t=1-\rho \in (0,1)$. The target Poisson distribution and proposal geometric distribution have probability mass functions $p_k=e^{-\lambda }{\lambda }^k/k!$ and $q_k=(1-t)t^k$. Their ratio isThe successive ratio $R_{k+1}/R_k=\lambda /[t(k+1)]$ shows that a maximizing integer is $m=\lfloor \lambda /t\rfloor$, with a tie at $m-1$ when $\lambda /t$ is an integer. Thus the tight envelope isFor any fixed $t$, draw $Y\sim \mathrm{Geom}(1-t)$ using (ii)'s inversion formula with this new parameter, and draw an independent $U$ from the uniform distribution on $(0,1)$. Accept whenotherwise restart. The acceptance probability is $1/M(t)$ and the expected number of proposals is $M(t)$.

To minimize that expected count, set $K=\lceil \lambda \rceil$. The exact answer isHere is a proof that includes the changes of maximizing integer. Each $\log R_k(t)$ is a convex function of $t$, since its second derivative is $(1-t{)}^{-2}+k/t^2>0$. Their pointwise maximum $\log M(t)$ is also a convex function. When $m$ is fixed its derivative isIf $K-1<\lambda \le K^2/(K+1)$, take $t_{\ast }=\lambda /K$, where $m$ changes from $K$ on the left to $K-1$ on the right. The left derivative is nonpositive because $\lambda \le K^2/(K+1)$; the right derivative is nonnegative because $\lambda \ge K-1$. This is a minimum. If $K^2/(K+1)<\lambda \le K$, take $t_{\ast }=K/(K+1)$. There $K<\lambda /t_{\ast }\le K+1$ and the active branch has $m=K$ except for the endpoint tie; its derivative vanishes. Convexity again proves the global minimum. The formulas agree at the boundary.

For a fully specified optimal algorithm, compute $K,t_{\ast },{\rho }_{\ast }$ and $m_{\ast }=\lfloor \lambda /t_{\ast }\rfloor$ once, then use the preceding acceptance test with $t=t_{\ast }$ and $m=m_{\ast }$. At an integer tie either maximizing integer gives the same test. In code, factorials can be evaluated through the logarithm of the gamma function:and acceptance is $\log U\le \min (0,\log a_Y)$. This avoids overflowing factorials.

The optimum minimizes expected trials, the usual idealization of time when each proposal and test has constant cost. Actual runtime also depends on the implementation of discrete inversion and factorial evaluation. Simply minimizing an upper bound for $M(t)$ would not necessarily give the exact optimum above; see optimal geometric envelope for a Poisson distribution.

Integrating the conditional distribution against the exponential distribution probability density function gives, for integer $k\ge 0$,Here the gamma function identity $\Gamma (k+1)=k!$ evaluates the integral. Henceon $\{0,1,2,\dots \}$. Put $r=1/(1+\lambda )$ and $\rho =1-r$. The geometric series gives $F(k)=1-r^{k+1}$. Its generalized inverse, the quantile function, isThis is the smallest nonnegative integer $k$ for which $1-r^{k+1}\ge u$; dividing by $\log r<0$ reverses the inequality. Draw $U$ from the uniform distribution on $(0,1)$ and use this formula. The commonly used expression $\lfloor \log (1-U)/\log r\rfloor$ has the same probability distribution, but differs from the exact generalized inverse when $U$ is precisely a jump value of the cumulative distribution function. Such values have probability zero for an ideal continuous uniform draw.

For the exponential distribution with rate $\lambda$, the cumulative distribution function is $F(x)=1-e^{-\lambda x}$ for $x\ge 0$, and zero for $x<0$. Its quantile function isFor inverse transform sampling, draw $U$ from the uniform distribution on $(0,1)$ and return $-\log (1-U)/\lambda$. Equivalently return $-\log U/\lambda$, since $1-U$ has the same uniform distribution. Indeed, for $x\ge 0$, $\mathbb{P}(-\log U/\lambda >x)=\mathbb{P}(U<e^{-\lambda x})=e^{-\lambda x}$.

The integrand is positive, so the minimum-variance importance distribution is its target-weighted probability density function, normalized by $\theta =\sqrt{e}$:By completing the square,Thus $q^{\star }$ is the multivariate normal distribution $N(-e_1,I_d)$, where $e_1=(1,0,\dots ,0)$ is the first standard basis vector. Its importance weight and weighted integrand areA variance cannot be negative, and this proposal makes every summand constant, proving optimality directly. It is admissible under the arbitrary-proposal condition even though its mean vector is not zero. Although optimal proposals usually involve an unknown normalizing integral, here both that integral and the resulting normal distribution are explicit.

At equal sample count, rejection sampling eventually has smaller variance for fixed ${\sigma }^2>1$; at equal expected proposal cost, importance sampling eventually wins. The distinction matters because a sample of the rejection sampling estimator means an accepted draw.

For the importance sampling estimator, all coordinates contribute to the importance weight although only one enters the integrand. The same normal distribution transform as in (c), now with linear coefficient $-2$ for coordinate one, givesBoth estimators are unbiased estimators of $\sqrt{e}$, andFor fixed $s>1$, the first variance grows exponentially while the second stays constant. This is the benefit of returning exact target draws in rejection sampling. However, obtaining those draws costs $s^{d/2}$ proposals each on average.

To compare computational work, multiply each variance by its expected proposal count. For importance sampling that count is $n$; for rejection sampling it is $n{s}^{d/2}$. The ratio of the resulting cost-normalized variances isbecause $A/\sqrt{s}=\sqrt{s/(2s-1)}<1$. Thus importance sampling eventually gives better precision for the same expected number of proposals, assuming comparable work per proposal. This is an asymptotic cost comparison, not a guarantee that a fixed small computing budget produces any accepted draws. Both methods still deteriorate exponentially in this comparison. For $s=1$ they coincide and every proposal is accepted.

A practical improvement is to exploit the integrand's one-coordinate structure directly: average $e^{-G_1}$ for standard normal distribution draws $G_1$, without generating unused coordinates. The optimal change of proposal distribution in (iii) improves further.

For the new integrand, $\theta ={\mathbb{E}}_p[e^{-X_1}]=e^{1/2}$ and ${\mathrm{Var}}_p(e^{-X_1})=e^2-e$, independently of dimension. Chebyshev's inequality therefore yieldsA sufficient positive integer sample count isThe ceiling function ensures that the bound is at most $\delta$, so the strict-error event in the question has probability at least $1-\delta$. The required number of accepted draws is independent of $d$; their expected proposal count is $n{\sigma }^d$.

For $\alpha >1$, $\theta =0$ is an unstable stationary point and the interior minimum satisfies $u={\alpha }^{-1/2}$. HenceThe potential then diverges at $\pi$. For $ϵ={\omega }_0/C$, the turning point is specified exactly by the smaller quadratic rootFor small positive ${\omega }_0$, this approaches $\boxed{{\theta }_{\max }\simeq 2\arccos {\alpha }^{-1/2}}$, a finite excursion even when the initial impulse is arbitrarily small. An excursion from the north pole to the outer turning angle and back has periodNear the unstable pole, $V\simeq -{\Omega }_u^2{\theta }^2$ with ${\Omega }_u=C\sqrt{\alpha -1}/2$. The near-pole contribution gives $T_{\theta }\sim (2/{\Omega }_u)\log (\mathrm{constant}\times C/{\omega }_0)$, so the period diverges logarithmically as ${\omega }_0\to 0$. The exactly unperturbed top can remain at the unstable vertical orientation; the limiting excursions spend ever longer close to it. The direction also precesses according to part c.

Put $u={\cos }^2(\theta /2)$. Then $V/C^2=u^{-1}-1-\alpha +\alpha u$. For $\alpha <1$ it increases from zero at the north pole to infinity as $\theta \to \pi$ and has no interior minimum. The vertical orientation is stable. Near zero,For ${\omega }_0/C\ll 1$, away from the marginal case $\alpha =1$, this givesThe locally signed tilt oscillates at angular frequency $\Omega$, while the azimuth precesses approximately at $C/2$. The nonnegative polar coordinate is $\theta \simeq {\theta }_{\max }|\sin \Omega t|$, so its successive pole-to-pole excursions have period $\pi /\Omega$ and angular repetition frequency $2\Omega$. Stating this convention avoids confusing a signed small displacement with the polar-angle magnitude. The full axis direction combines that excursion with precession.

With $p=-P$, the allowed range is $|x|\le X_2$, where $X_2=\sqrt{2P/a}$. But an orbit starting with $x\ne 0$ never crosses the midplane in finite time, becauseOne positive-side separatrix, with $u=\omega (t-t_c)$ and $\omega =\sqrt{2Pa}/m$, isThe negative-side orbit is its reflection in $x=0$. For the positive branch, $x$ rises from zero asymptotically, reaches $X_2$ at $u=0$, then returns asymptotically to zero. Thus the infimum is zero but it is not attained in finite time on this nontrivial branch. Since $\dot{z}=(P/m)(2{\mathrm{sech}}^{2}u-1)$, the two axial-velocity zeros occur at $x=\sqrt{P/a}$. Motion is toward decreasing $z$ near either asymptotic midplane tail and toward increasing $z$ near the outer excursion. These directions are marked in part i.

The exactly midplane initial state $x=0$, $\dot{x}=0$ is a separate straight trajectory with $\dot{z}=-P/m$; it must not be described as a finite-time crossing of the separatrix.

Dividing the two velocities givesHere $\dot{z}=a{x}^2/m\ge 0$: the orbit always advances toward increasing $z$, except for instantaneous zeros at each midplane crossing $x=0$. The transverse motion oscillates between $-X$ and $X$. At these outer turning points $\dot{x}=0$ and the trajectory is tangent to the $z$ direction; at the midplane $\dot{x}\ne 0$ while $\dot{z}=0$, so the tangent is along $x$. The sign of the initial $\dot{x}$ only chooses the phase or reflected trace.

Fully ionized helium supplies three particles per four hydrogen masses, so ${\mu }_{\mathrm{He}}\simeq 4/3$. The same small $Z$, but negligible hydrogen fraction, gives ${\kappa }_{0,\mathrm{He}}\propto Z$, half the hydrogen-star coefficient. The relative luminosity isConsequentlyThe very strong molecular-weight dependence outweighs the reduction in mass. Treating the two opacities as identical would miss a factor two, and using neutral-gas molecular weights would be inconsistent with the stated hot ionized stars.

For fully ionized hydrogen with a small metal fraction, there is approximately one proton and one electron per hydrogen mass, so ${\mu }_H\simeq 1/2$. The Kramers coefficient is proportional to $Z(1+X)\simeq 2Z$. Denote this star's luminosity by $L_H$; the supplied proportionality law alone does not fix its absolute normalization.

Relative to the helium-dominated comparison star, the calculation in part ii givesIf one uses a roughly solar luminosity as an illustrative calibration for this roughly solar-mass, solar-central-temperature reference, the helium star would be of order a hundred solar luminosities. That calibration is only approximate because a mostly hydrogen composition with $\mu \simeq 1/2$ is not the actual solar composition.

The zero trace and unit eigenvalues imply equal numbers of $+1$ and $-1$ eigenvalues, so $N$ is even. A two-dimensional representation is impossible: all four matrices ${\alpha }_1,{\alpha }_2,{\alpha }_3,\beta$ would be traceless $2\times 2$ matrices. Anticommutation gives $\mathrm{Tr}(AB)=0$ for distinct matrices, whereas $\mathrm{Tr}(A^2)=2$. They would therefore be four linearly independent members of the three-dimensional complex space of traceless $2\times 2$ matrices, a contradiction.The usual irreducible $3+1$ dimensional Dirac representation has $N=4$. Larger representations can contain multiple copies; the requested lower bound is not itself a proof of irreducibility.

For any one of the four anticommuting matrices $A$, choose another $B$. Since $B^{-1}=B$ and $BAB=-A$, cyclic invariance of the trace givesThereforeEquivalently, the paired positive and negative unit eigenvalues cancel. This argument does not depend on a particular matrix representation.

If $Av=\lambda v$ for $A={\alpha }_i$ or $\beta$, then $A^2=I$ implies ${\lambda }^2=1$. The polynomial $x^2-1$ has distinct roots, so such a matrix is diagonalizable andAnticommutation with another invertible Dirac matrix maps a $+1$ eigenspace to a $-1$ eigenspace, so both signs occur with equal multiplicities in a nontrivial representation. A Hermitian operator is appropriate for the physical Hamiltonian and additionally guarantees an orthonormal eigenbasis.

For a free particle the Dirac matrices are constant and the momenta commute. Squaring the Hamiltonian cancels all mixed terms by their anticommutators:Applying another time derivative to the Dirac equation therefore givesfor every component. This is the Klein-Gordon equation. Its componentwise satisfaction is a necessary consequence of the first-order Dirac equation, not a converse selecting all spinor solutions of the squared equation.

For a nonzero everywhere attractive weak potential, $\gamma <0$, the pole has $k=i\kappa$ with $\kappa =m|\gamma |/{\hslash }^2>0$. The exterior waves decay as $e^{-\kappa |x|}$, giving one shallow bound state to leading weak-coupling order:Its size ${\kappa }^{-1}$ is much larger than the potential range because $\kappa a=O(\lambda )\ll 1$. For an everywhere repulsive potential there is no negative-energy normalizable bound state: the kinetic and potential quadratic forms are both nonnegative. Its lower-half-plane pole is a virtual state, not a stable bound state. More complicated sign-changing potentials with a vanishing integral require higher-order analysis and are not covered by this simple leading result.

Write $\nu =m\gamma /{\hslash }^2$, so $t(k)\simeq k/(k+i\nu )$. Its nontrivial pole isFor negative $\gamma$ this is on the positive imaginary $k$ axis, the physical bound-state sheet. For positive $\gamma$ it lies on the negative imaginary axis and is a virtual-state pole reached by analytic continuation through the continuum cut. These different sheets can have the same negative numerical energy; specifying only $E_{\ast }$ would conceal the distinction. As a function of $E$, the momentum square root also introduces the continuum threshold branch point at zero. If $\gamma =0$, numerator and denominator cancel at this order and no pole is predicted by this leading approximation.

Replace the slowly varying interior wave by $C$ and $e^{ik|x-x^{\prime }|}$ by one when both arguments lie within the range. The integral equation givesHenceThe integrated strength has dimensions of energy times length. For $U=-\alpha \delta (x)$, $\gamma =-\alpha$ and this is exactly the result in part c. For a regular weak potential the approximation sums the important ratio $m\gamma /(k{\hslash }^2)$ without requiring it to be small, which is why it remains useful beyond the naive first Born expansion at small momentum.

Let $\lambda =m{U}_0{a}^2/{\hslash }^2\ll 1$ and $ka\ll 1$. Inside a region of size $a$, the Schrödinger equation estimates the curvature as $a^2{\psi }^{\prime \prime }=O(\lambda +(ka{)}^2)\psi$. Matching to the incoming and reflected waves constrains the slope as well. To leading order the interior wave has nearly constant value $C$, while its small accumulated derivative change encodes the potential's integrated strength.

This is not a Born assumption that $C$ is close to one. At very low $k$ the wave can be strongly reflected even by a weak potential, and $C$ can be small. The matching solution in the next part gives $C\sim k/(k+im\gamma /{\hslash }^2)$. It implies a fractional variation across the range of order $ka+m|\gamma |a/{\hslash }^2$, which is small under the two specified conditions. ThusThe approximation requires a finite short-range potential with a controlled width and integral; it does not describe arbitrary long-range tails.