The same integration by substitution gives the expected valueThe remaining integrand is an odd function. Its integral is absolutely convergent, since and is integrable. Thus cancellation under symmetry in integration is legitimate, rather than merely a principal-value calculation. ConsequentlyEquivalently, the logistic distribution is symmetric about , and its exponentially decaying tails ensure that its expected value exists.
For , integrate the probability density function to obtain the cumulative distribution functionIf , the requested probability is . Since the hyperbolic tangent is odd, this givesThe second case matters because the question places no sign restriction on : the event is empty when its upper endpoint is nonpositive. Using an oriented integral from to a negative would give a negative number, not a probability.
A probability density function must be nonnegative. The printed condition is therefore too weak: for the given prefactor makes the function negative everywhere. The location-scale form of the logistic distribution requires , which is assumed in the remaining parts.
Set , so . The normalization of a probability distribution follows from the antiderivative of the squared hyperbolic secant:For completeness, the same expression with integrates to , since the transformed limits reverse. If a signed scale were desired, replacing its prefactor by would give a valid density, but that is a correction to the printed formula.
The Maclaurin series give and . Inverting the latter using the geometric series givesMultiplying the two expansions yieldsThe remainder is because the product already generates an term from .
Use the Maclaurin series and . The terms through fourth degree giveIn particular the cubic coefficient cancels; it has not been omitted accidentally.
For , the local minimum at the origin has . Its full level set isThe connected component containing the minimum is just the isolated origin. The same level also has the outer branches for ; these are different connected components, not curves approaching the origin.
Because the exponential function is strictly increasing, the level sets are exactly those of . For they satisfyThere are no contours for , the level is the origin, and gives nested closed ovals symmetric about both axes. Their stronger fourth-power growth in makes them relatively narrower horizontally as the level rises.
For , the contour equation becomesThe central local minimum is surrounded by closed ovals for , accompanied by outer open branches. At , the level passes through the two saddle points and changes its connectivity. Levels have only outer branches, while levels above the saddle value have an upper and a lower branch. The sketch is restricted to the requested square; labels give , so they identify the corresponding contours of without exponentially large labels.
Put , so . Since and , its stationary points satisfyThe origin occurs for every real . There are two additional points only for , at .
At a stationary point, the Hessian matrix simplifies because :At the origin this is , so is a strict local minimum for every , with . If it is also the unique global minimum, since with equality only there. For , the two extra points have , so their Hessian has opposite-sign eigenvalues. ThusThere are no local maxima. When , the origin is not a global minimum: as , but never attains zero.
Write , , at the stationary point, and let . Under the usual second-order Taylor expansion hypothesis, the second-derivative test givesIndeed the quadratic term is . For , completing its square gives . The two coefficients have the same positive sign, the same negative sign, or opposite signs in the respective cases. A negative also forces the two Hessian eigenvalues to have opposite signs, including the case .
There is a qualification to the printed assumptions: existence of second partial derivatives only at the point does not ensure the needed quadratic approximation. Continuous second partial derivatives in a neighborhood are sufficient. The existence of second partial derivatives does not justify the Hessian test: for example, the function away from the origin, extended by , has Hessian at the origin but and . Thus the classification above uses the standard additional regularity assumption, which the smooth function in part (b) satisfies.
An interior stationary point has vanishing gradient:For a differentiable function, this is the necessary first-order condition for an interior local minimum or local maximum.
The nonzero factor can be cancelled in the local differential equation, leaving . On an interval with ,Hence the explicit general solution ison any interval that avoids the excluded origin. It also follows from fixing the local angle potential, with slope . As parameterized curves there are also vertical rays , which cannot be represented as single-valued functions . Geometrically the connected integral curves are rays, not lines continued through the undefined origin.
On a chart with , setThe chain rule verifies both components:A coordinate-independent description of the local potential is a continuous branch of , the complex argument of the scaled coordinates. Such branches can also cover points where , using a different angular chart. Continuing once around the missing origin changes that angle by a nonzero multiple of , so there is no global single-valued on the full stated domain.
There is a global-domain issue in the printed question. Write away from the origin. Direct partial differentiation givesThus the differential one-form is a closed differential form for every real . The Poincaré lemma makes it locally an exact differential form on small neighborhoods avoiding the origin. There is no additional condition such as .
However, the stated domain is the whole punctured plane, which is not simply connected. To test global exactness, use the closed ellipse , , . Here andThe line integral of a globally exact differential form around any closed curve would be zero. ConsequentlyThe distinction is precisely that of the elliptically scaled angular one-form. The remaining parts can be answered consistently using a local angle branch, as follows.
Differentiate the defining function with the other coordinates independent. Put for brevity. ThenThe implicit differentiation ratios therefore giveEach expression applies on the implicit surface wherever its own denominator is nonzero. The hyperbolic cosine appears because the derivative of with respect to either or is .
Assume is continuously differentiable near the point of interest on , and that the relevant denominators below are nonzero. The total differential on the surface isHolding fixed gives , so the implicit function theorem givesWhen , their product proves the reciprocity rule for constrained partial derivatives:Similarly, holding or fixed gives and . With , multiply the three ratios:This is the cyclic partial derivative identity. Nonzero denominators and local differentiable coordinate representations are substantive assumptions: at a singular coordinate representation, a formal cancellation of zero or infinite factors is not a valid proof.
Expand the defining complex exponential functions:Thus the real part and imaginary part of the hyperbolic cosine are
For , the complex logarithm is , where . ConsequentlyA specified branch chooses one continuous determination of the complex argument on an appropriate cut domain. For the usual principal-value convention use the principal argument . A formula using only would lose the quadrant information; it is a valid local angle only with the appropriate additive multiple of .
There are unlisted articles, also show them or only show them.
