OctoInformBeta

OpenAI preprint, September 2026 · §3. Shape inequalities for finite step functions

Lemma 3.4: Under shape conditions, seven certificate quantities are nonnegative for x>0x>0, and (rρF2)′=ρD̃F2(r\rho F_2)'=\rho\,\widetilde{D}F_2 where r≠0r\neq0 and ρ′=(r−s)ρ\rho'=(r-s)\rho

Certificate signs on the half-line and the flux derivative.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Let γ\gamma be an order parameter that is a finite step function, and fix a time tt inside one of its steps, where γ=c\gamma = c. On that step, let Φγ\Phi_\gamma be the value function of γ\gamma, let uγ=Φγ,xu_\gamma = \Phi_{\gamma,x}, and let XX be the diffusion dX=γ(t)uγ(t,X)dt+dBdX = \gamma(t)\,u_\gamma(t,X)\,dt + dB, X0=0X_0 = 0, driven by the Brownian motion BB; its drift r=cuγr = c\,u_\gamma solves the backward Burgers equation ∂tr+12rxx+rrx=0\partial_t r + \tfrac12 r_{xx} + r\,r_x = 0.

Lemma 3.4 concerns rr at the time tt, as a function of xx, together with a second function ss and the certificate variables built from their derivatives. The argument needs two things of these variables: their signs at every x>0x > 0, under shape conditions on rr and ss, and an integration by parts against a weight ρ\rho, which turns on the derivative of a flux built from a polynomial F2F_2 in the variables. The library states both for arbitrary smooth real functions rr and ss on ℝ\mathbb{R}.

The Backward Shape

The Burgers rate of a function rr of time tt and space xx is the function ℛ=12rxx+rrx,\mathcal{R} = \tfrac12\, r_{xx} + r\,r_x ,(1)✦ so that a solution of the backward Burgers equation satisfies ∂tr=−ℛ\partial_t r = -\mathcal{R}. The backward Wronskian of a function rr of time tt and space xx is the function rℛx−rxℛr\,\mathcal{R}_x - r_x\,\mathcal{R} of tt and xx, the Wronskian in space of rr and its Burgers rate ℛ\mathcal{R}.✦

A function rr on ℝ\mathbb{R} has the backward shape if it is odd with r′>0r' > 0 everywhere and if, for every x≥0x \ge 0, its second derivative, its Burgers rate and its backward Wronskian obey r″(x)≤0,ℛ(x)≤0,r(x)ℛ′(x)−r′(x)ℛ(x)≥0.r''(x) \le 0, \qquad \mathcal{R}(x) \le 0, \qquad r(x)\,\mathcal{R}'(x) - r'(x)\,\mathcal{R}(x) \ge 0 .(2)✦

Certificate Variables

The certificate variables of real functions rr and ss on ℝ\mathbb{R} are the functions R=r2,K=rs,v=r′,w=r″,z=r‴,b=−wr,j=s′−srv,L=−rs″\begin{gathered} R = r^2, \qquad K = rs, \qquad v = r', \qquad w = r'', \qquad z = r''', \\ b = -\frac{w}{r}, \qquad j = s' - \frac{s}{r}\,v, \qquad L = -r\,s'' \end{gathered}(3) on ℝ\mathbb{R}; moreover SS is the function z+bv−2Rbz + bv - 2Rb on ℝ\mathbb{R}. For the drift of the setting, vv, ww and zz are cΦγ,xxc\,\Phi_{\gamma,xx}, cΦγ,xxxc\,\Phi_{\gamma,xxx} and cΦγ,xxxxc\,\Phi_{\gamma,xxxx} at the time tt.

The Polynomial F2F_2

The polynomials F1F_1 and F2F_2 defined below are built from the rational constants A=113,C=5324,T0=7348,g0=233110000,h=872500,k=1916,m=−1051400,n0=−1073500,x0=46635000,y=685310000.\begin{gathered} A = \tfrac{1}{13}, \quad C = \tfrac{53}{24}, \quad T_0 = \tfrac{73}{48}, \quad g_0 = \tfrac{2331}{10000}, \quad h = \tfrac{87}{2500}, \\ k = \tfrac{19}{16}, \quad m = -\tfrac{1051}{400}, \quad n_0 = -\tfrac{1073}{500}, \quad x_0 = \tfrac{4663}{5000}, \quad y = \tfrac{6853}{10000}. \end{gathered}(4)

The polynomials F1F_1 and F2F_2 in the real variables RR, KK, vv, bb and jj are the cubic forms F1=b2(2kR−Av)−bv(2kK+g0R+2T0R+2Cv+hj),F_1 = b^2\,(2kR - Av) - bv\,(2kK + g_0R + 2T_0R + 2Cv + hj),(5)✦ F2=v2(x0j+yv+mK+n0R)−F1.F_2 = v^2\,(x_0 j + y v + mK + n_0 R) - F_1 .(6)✦

The quadratic forms QQ and TT in RR, KK, vv, bb and jj below satisfy ∂bF2=2Q+T\partial_b F_2 = 2Q + T. Q=Avb+Cv2+kKv−2kRb+T0Rv,T=v(g0R+hj).Q = Avb + Cv^2 + kKv - 2kRb + T_0Rv, \qquad T = v\,(g_0R + hj).(7)

The flow derivative DF2DF_2 at real numbers RR, KK, vv, bb, jj, zz and LL is the derivative of F2F_2 along the vector field (2Rv,2vK+Rj,−Rb,−z−bv,Kb−jv−L)(2Rv,\ 2vK + Rj,\ -Rb,\ -z - bv,\ Kb - jv - L) in the variables (R,K,v,b,j)(R, K, v, b, j), DF2=2Rv∂RF2+(2vK+Rj)∂KF2−Rb∂vF2−(z+bv)∂bF2+(Kb−jv−L)∂jF2,\begin{aligned} DF_2 = {}& 2Rv\,\partial_R F_2 + (2vK + Rj)\,\partial_K F_2 - Rb\,\partial_v F_2 \\ & - (z + bv)\,\partial_b F_2 + (Kb - jv - L)\,\partial_j F_2 , \end{aligned}(8)✦ each partial taken at (R,K,v,b,j)(R, K, v, b, j), and the weighted derivative of F2F_2 is D̃F2=DF2+(v+R−K)F2.\widetilde{D}F_2 = DF_2 + (v + R - K)\,F_2 .(9)✦

The certificate flux Ψ\Psi of real functions rr, ss and ρ\rho on ℝ\mathbb{R} and their flux divergence Θ\Theta are the functions Ψ=rρF2(R,K,v,b,j),Θ=ρD̃F2(R,K,v,b,j,z,L)\Psi = r\rho\,F_2(R, K, v, b, j), \qquad \Theta = \rho\,\widetilde{D}F_2(R, K, v, b, j, z, L)(10) on ℝ\mathbb{R}, with RR, KK, vv, bb, jj, zz and LL the certificate variables of rr and ss.

The Claim

Theorem 1. Let rr and ss be smooth real functions on ℝ\mathbb{R}, with certificate variables as in (3).

(i) Suppose that rr has the backward shape, that ss is odd with s′≥0s' \ge 0 everywhere, and that s″≤0s'' \le 0 and the Wronskian rs′−sr′rs' - sr' is nonnegative on [0,∞)[0, \infty). Then at every x>0x > 0, R(x),K(x),v(x),j(x),L(x)≥0,b(x)−2v(x)≥0,S(x)≥0.R(x),\ K(x),\ v(x),\ j(x),\ L(x) \ge 0, \qquad b(x) - 2v(x) \ge 0, \qquad S(x) \ge 0 .(11)✦

(ii) Let xx be a point with r(x)≠0r(x) \neq 0, and let ρ\rho be a real function on ℝ\mathbb{R} that is differentiable at xx with ρ′(x)=(r(x)−s(x))ρ(x)\rho'(x) = \bigl(r(x) - s(x)\bigr)\rho(x). Then the certificate flux of rr, ss and ρ\rho is differentiable at xx, and its derivative there is the flux divergence: Ψ′(x)=Θ(x).\Psi'(x) = \Theta(x) .(12)✦

In the paper, the result reads: for finite step data, at a fixed positive time inside one step, the variables RR, KK, vv, bb, jj, LL, b−2vb - 2v (the paper’s BB) and SS are even and extend smoothly across zero, with RR, KK, vv, jj, LL, b−2vb - 2v, S≥0S \ge 0; with 𝒟=r∂x\mathcal{D} = r\,\partial_x they obey 𝒟(R,K,v,b,j)=(2Rv,2vK+Rj,−Rb,−z−bv,−jv+Kb−L)\mathcal{D}(R, K, v, b, j) = (2Rv,\ 2vK + Rj,\ -Rb,\ -z - bv,\ -jv + Kb - L); and, with ff the density at that time of the diffusion dX=rdt+dBdX = r\,dt + dB, X0=0X_0 = 0, every polynomial FF in RR, KK, vv, bb, jj satisfies 𝔼f[𝒟F+(v+R−K)F]=0\mathbb{E}_f[\mathcal{D}F + (v + R - K)F] = 0, every term absolutely integrable.

What Is Checked

Both statements concern smooth real functions rr and ss on ℝ\mathbb{R} and their certificate variables (3): the first is stated over the backward shape, with its Burgers rate (1) and backward Wronskian, and the second over the certificate flux and flux divergence (10), built on the polynomial F2F_2 of (6) with its constants (4), its flow derivative (8) and its weighted derivative (9).