Let be a twice continuously differentiable function on an open domain stable under positive scaling, with . Apply the chain rule twice to the scalar function and set . The result is
In two variables, this reads . In matrix notation it is , where is the Hessian matrix. Equivalently, the radial differential operator obeys ; substituting explains the factor .
Second-order reaction 2026-10-14
A second-order reaction has overall order of reaction two: a rate law may depend on , or on with partial orders of reaction one and one. Doubling all relevant reactant concentrations multiplies the reaction rate by four at fixed conditions. This experimentally determined dependence does not establish molecularity or a unique reaction mechanism. For an elementary reaction obeying the law of mass action, molecularity and overall order of reaction agree; a net chemical equation alone cannot establish that the conversion is elementary.[ref]
For the single-concentration model, explicitly choose the disappearance rate convention
Here is the coefficient for loss of A, held fixed at the specified temperature and other conditions. Its subscript labels the chosen observed disappearance rate; it does not imply a pseudo-first-order reaction. Assume a uniformly mixed closed system, fixed volume, and no appreciable reverse or competing chemical reactions. The integrated rate law is
The reciprocal-concentration plot is a straight line with slope and intercept . A brief derivation uses the chain rule: , then applies the initial condition. This is a hyperbolic decay, rather than the exponential decay of a first-order reaction.[ref]
As an original prescribed example, choose and . At , the reciprocal-concentration calculation is , giving . The initial disappearance rate is ; after it is . Halving the concentration quarters the disappearance rate.
Dimensional analysis checks every formula: has units , has units , and is a dimensionless quantity. Changing concentration units changes the numerical rate constant: the same example is , , giving .
Figure 1.
Second-order reaction concentration decay and reciprocal-concentration lines for three prescribed starting concentrations
. Original model curves, not measured data. All use the disappearance rate coefficient . The reciprocal lines share a slope while their intercepts differ; marked points give each run's first half-life. Details.
For an elementary reaction , instead define the stoichiometrically normalized reaction rate . Two A entities disappear per P formed, so
Consequently the reciprocal-concentration slope is , not . In the prescribed example, the normalized rate constant would be ; initially while the disappearance rate is twice that. With , stoichiometry gives . Both rate constants have the same dimensions, so a units check alone cannot catch their factor-of-two difference. The symbol denotes the same normalization as in reaction rate.
The half-life, meaning the time to reach half a specified starting concentration, follows by setting :
For in , this becomes . Its units are seconds because has units . Doubling the starting concentration halves the first half-life; a quoted half-life needs its starting concentration. This differs from a first-order reaction, whose half-life is independent of starting concentration.[ref]
Within the prescribed run, reaches , , and at , , and . The successive halving intervals are , , and : each begins at a smaller concentration. More generally, reaching fraction of the initial concentration, with , requires , so consumption takes nine initial half-lives, rather than a fixed sequence of equal waiting times.
Figure 2.
Lengthening second-order reaction halving intervals and the dependence of the first half-life on initial concentration
. Original model predictions, not measurements. Left: the prescribed run. Right: at fixed , compared with a first-order reaction chosen to have . That comparison uses , with different units from . Details.
A straight reciprocal-concentration plot tests this particular single-concentration rate equation. It does not uniquely identify a reaction mechanism, nor does curvature rule out every second-order reaction: two independently changing reactant concentrations require their own integrated rate law. At low concentration, a small measurement error becomes approximately in the reciprocal, magnifying late-time errors. Use the line as a diagnostic; a nonlinear regression of the original concentrations can retain their measurement scale and explicit error model.
Sepp Hochreiter 2026-11-05
Sepp Hochreiter is a computer scientist working on machine learning and neural networks. His institutional profile identifies long short-term memory among his research topics. The1997paper with Jürgen Schmidhuber connects his earlier analysis of decaying learning signals to a recurrent architecture designed to preserve useful error flow.[ref] [ref]
A toy chain rule calculation explains the motivation. If a learning signal is multiplied by at each of twenty successive steps, its contribution is scaled by
That small contribution can make an early input difficult to associate with a much later training signal. The example illustrates the vanishing gradient problem; it is not a measurement of every recurrent neural network.
The long short-term memory design introduced a controlled internal path with constant error flow, while multiplicative gates regulate access. Its artificial-data experiments included successful learning across delays exceeding1000discrete steps. The reported capability belongs to those experiments and their architecture-specific learning rules, rather than implying that any trained instance will remember every input indefinitely.[ref]
Figure 1.
Sepp Hochreiter, identified on his JKU institutional profile
. External institutional portrait from Sepp Hochreiter profile by JKU.
For a positive harmonic function on the unit disc, choose a global harmonic conjugate and apply the Carathéodory derivative bound for the right half-plane to . This gives by the Cauchy-Riemann equations. A disk automorphism sending zero to has derivative modulus , so the chain rule gives the displayed bound. Composing a boundary-pole Poisson kernel with a disk automorphism attains equality, proving the coefficient is optimal at every point.
Sigmoid-softmax network gradients Created 2026-10-05 Updated 2026-10-14
For hidden activations and class probabilities , a one-hot label has log-likelihood . Put and . Then
These follow from the chain rule and give single-observation stochastic gradient descent updates for the negative log-likelihood.
If the zero set of a differentiable function is a submanifold, the chain rule always places its tangent space inside . Equality follows from the full-rank hypotheses of the regular level set theorem, but may fail for a singular defining function even when the zero set itself is smooth. For , , the zero set is the single point . Its tangent space is zero, while has kernel . One should therefore check the constraint qualification before using the kernel formula.
SIR first integral 2026-10-14
For , the SIR model equations have the first integral of an ordinary differential equation
Here is the natural logarithm. A short proof outline is to use the chain rule and substitute the SIR model equations:
The SIR first integral gives directly from , so it describes the curve in a phase portrait even before its time parametrization is known.
Specific kinetic energy 2026-10-25
Kinetic energy per unit mass. For an acceleration field , its change along an actual trajectory is , by the chain rule. This differs by a mass factor from the work integral of a physical force.
For , the chain rule gives . The radial term is killed by crossing with , giving the identity. Purely angular vector spherical harmonics are therefore unchanged as angular basis functions. This is distinct from pushing forward a physical vector field by the inversion Jacobian.
Squared-error backpropagation Created 2026-10-14 Updated 2026-10-25
For squared loss , let and . The chain rule gives and . A gradient descent step of size uses and , with every derivative evaluated at the same pre-update weights.
On an open set where , put . The chain rule gives , so the quadratic eikonal equation is equivalent to . For constant positive data on a smooth initial curve, the tangential derivative vanishes and the two allowed normal derivatives are . Locally the two solutions are , where is oriented distance along the normal rays; hence wherever the chosen stays positive and the normal-coordinate map remains invertible.
There are two separate obstructions to using this initial-data chart. A characteristic caustic occurs when normal rays focus and the projection Jacobian determinant vanishes. Alternatively can reach zero while the geometric rays remain separated: the square-root transformation then ceases to apply and the original Charpit's equations have a degenerate zero-gradient state. A continuation of beyond either obstruction needs separate justification; squaring an algebraic expression alone does not preserve a prescribed positive square-root branch.
Where , the product rule and chain rule give the displayed real derivative. Its stationary points therefore satisfy . The denominator uses the square of the real cube root, which stays positive for negative . Points with must be checked in the original function: a simple zero of with creates a vertical tangent, not a stationary point. If the prefactor also vanishes, a cancellation can change differentiability and must be analyzed separately.
Stellar adiabatic exponent Created 2026-09-28 Updated 2026-11-05
The stellar adiabatic exponents describe how pressure , temperature and mass density respond along an isentropic process at fixed chemical composition. With specific entropy ,
The chain rule gives . Thus the three names do not denote three independent thermodynamic responses. The adiabatic temperature gradient is , while determines the squared adiabatic sound speed .[ref]
For a fixed-composition monatomic ideal gas, all three exponents equal and . For equilibrium blackbody radiation, all three equal and . In a gas-radiation mixture they generally differ; the adiabatic exponents of a monatomic gas-radiation mixture keep that distinction explicit. Partial ionization can soften the compressional response because some work changes ionization rather than immediately raising the temperature. The relevant exponent in stellar convective stability therefore need not equal a constant ideal-gas specific-heat ratio.
Stochastic thermodynamics Created 2026-09-28 Updated 2026-11-05
Stochastic thermodynamics assigns work, heat and entropy production to individual fluctuating sample trajectories. It is useful for a small system, such as a trapped colloidal particle, whose exchanged energy is comparable to the thermal scale . The surroundings remain a thermal reservoir at temperature , while the observed coordinates follow a stochastic process.[ref]
For a one-dimensional optical trap with potential energy , moving its centre does incremental work on the particle. The coordinate change contributes , where the circle denotes the Stratonovich integral convention. The ordinary chain rule then gives the trajectory first law of thermodynamics, . A particle ahead of the moving trap centre can give a negative instantaneous work increment, even during a protocol with positive mean dissipated work.
In the optical-trap experiment of G. M. Wang and collaborators, a recorded colloidal particle trajectory allowed short-time entropy-consuming realizations to be compared with entropy-producing ones. Repeating the same driving protocol measures a probability distribution rather than only a mean response.[ref] The integral fluctuation theorem recovers nonnegative mean entropy production through Jensen's inequality, while allowing individual negative values.
A coordinate chart trivializes the tangent bundle by expressing in the coordinate basis. For , the chain rule gives the displayed transition. Smoothness and fibrewise linear invertibility of the Jacobian matrix construct the natural smooth vector bundle structure. The chain rule also supplies the transition cocycle.
Tangent space Created 2026-09-24 Updated 2026-10-14
At a fixed point of an -dimensional smooth manifold without boundary, the tangent space is the real vector space of all tangent vectors at . In manifold charts and , a tangent vector has components
The coordinate basis transforms inversely, so the tangent vector is unchanged. This invertible Jacobian matrix also preserves vector addition and scalar multiplication, giving a vector space of dimension independently of coordinates. [ref]
For example, at in , use the alternative manifold chart , . The Jacobian matrix sends to . Indeed, the smooth curve gives . Conversely, and , so . The numbers change; their action on every smooth function does not.
For an embedded submanifold , the tangent space can be identified with the vector subspace
A proof outline explains why every coordinate vector is possible: in a manifold chart , lift the short straight path using . Its tangent vector has components . For a regular local parametrization , the same chain rule gives , where . This is the tangent space from a local parametrization.
For the sphere , differentiating that equation gives . Conversely, for any nonzero with , the smooth curve
lies on the sphere and has ; use a constant smooth curve for . Thus , not merely a subset of that plane. At this vector subspace is , while the affine tangent plane drawn through is .
This also explains a mechanical constraint without requiring a force calculation. In the ideal mathematical model of a particle confined to a stationary sphere, its allowed velocities at are exactly : the radial component must vanish, but either independent tangential direction is available. For a general stationary holonomic constraint with independent equations, the regular level set theorem gives . These are instantaneous kinematic possibilities; they do not determine the actual subsequent motion. On a general smooth manifold, tangent spaces at different base points cannot be identified canonically; an affine connection supplies parallel transport when such comparisons are needed.
Tangent vector Created 2026-09-24 Updated 2026-10-18
A tangent vector at a point of a smooth manifold without boundary records an instantaneous velocity through . Its base point is part of the data; measuring its length additionally requires a Riemannian metric. Intrinsically, a smooth curve with defines its tangent vector by how every local smooth function changes along it:
This is a derivation at a point. Two smooth curves through represent the same tangent vector if these derivatives agree for every such smooth function; equivalently, their coordinate velocity vectors agree in one manifold chart. The chain rule makes that agreement independent of the chosen manifold chart. [ref]
For an embedded submanifold of Euclidean space, the intrinsic tangent vector is represented by the ordinary derivative , with its base point retained. A drawn arrow starting at represents that tangent vector; its endpoint need not lie on the embedded submanifold. Nor is a tangent vector required to be a unit tangent vector: a stationary smooth curve gives the zero vector, which belongs to the tangent space but has no direction to normalize.
Here are explicit calculations. In , and have different routes but the same tangent vector at . For , both give . For a circle of radius , has tangent vector at : changing changes the speed or reverses travel without changing the circle.
A reparametrization with gives, by the chain rule, . A regular change of curve parameter has . The singular choice instead gives
Both describe the same parabola, whose tangent line at the origin is still the horizontal axis. The zero velocity diagnoses this singular change of curve parameter, whose inverse is not differentiable at zero; it does not show that the geometric tangent line is absent. The regular curve supplies its nonzero direction.
Figure 1.
One parabola, different velocities: a singular parameter change can stop the arrow
. Details.
The diagram is an analytical illustration of these exact parametric curves, with equal parameter increments marked on each route. It separates geometry from traversal: the tangent line depends on the route, while a velocity vector also depends on how the curve parameter advances.
For a Planck function expressed as spectral radiance per frequency interval,
is the Planck constant, the speed of light and the Boltzmann constant. Differentiating at fixed frequency gives
The proof uses and the chain rule. At low frequency, the multiplier tends to one, recovering the Rayleigh-Jeans law. At high frequency it grows approximately as : the same fractional temperature change has a larger fractional brightness effect on the declining spectral tail.
For an original example choose and . At , , so the first-order fractional residual is . At , , it is . Both spectra are exact blackbodies: this residual is a temperature mismatch, not evidence of a nonthermal spectrum. Fitting temperature before assessing remaining spectral residuals therefore removes a physically distinct degree of freedom. The figure divides by a fixed peak brightness rather than by the rapidly vanishing high-frequency tail.
Suppose for positive times. The strictly increasing new time converts the scalar conservation law into , with . The chain rule proves the claim for classical solutions; a change of variables in the weak formulation proves it for weak solutions and preserves the time orientation of the Kruzhkov entropy inequality. An isolated zero of at the initial endpoint does not obstruct the positive-time transformation.
The Rankine-Hugoniot condition becomes , while characteristic curves obey . Thus a straight shock in the new time may accelerate in physical time. For , a constant-speed trajectory in is a parabola in . This separates genuine changes in wave structure from a common time-dependent speed factor.
Topological manifold Created 2026-09-24 Updated 2026-10-14
A topological -manifold without boundary is a topological space satisfying three conditions: it is Hausdorff, it is second countable, and every point lies in an open set with a homeomorphism onto an open set . Here is a fixed dimension of a manifold, and has the usual topology of Euclidean space. The pair is a coordinate chart. Unless explicitly qualified, “topological manifold” here means without boundary.[ref]
The Hausdorff condition requires disjoint open neighbourhoods for distinct points. Second countability requires a countable basis of a topology. Neither follows merely from the local homeomorphisms: non-Hausdorff manifolds fail the first condition, while an uncountable disjoint union of copies of the real line fails the second. Indeed, a basis of a topology must supply a different nonempty member inside each copy, so it cannot be countable.
The circle and the real line are both one-dimensional topological manifolds, with the same local model. Their global topologies differ: the circle is compact, while the real line is not. By contrast, the junction of a topological tripod already fails the local test. Thus local coordinate charts specify what kind of small pieces are allowed; global topological invariants distinguish how those pieces fit together.[ref]
The usefulness of a topological manifold is that local questions about continuity can be transported into Euclidean space, while compactness, connectedness, and the fundamental group still record global information. For calculus, further structure is needed: a smooth atlas requires smooth transition maps, letting the chain rule reconcile derivatives computed in different coordinate charts. A topological atlas by itself only guarantees homeomorphisms between coordinates.[ref]