On sum of squares certificates of non-negativity on a strip (Q1996082)

Every polynomials \(f\in\mathbb{R}[X,Y]\) that is non-negative on \([0,1]\times \mathbb{R}\) can be written as \[ \sigma_0+\sigma_1X(1-X) \] where \(\sigma_0,\sigma_1\in\mathbb{R}[X,Y]\) are sums of squares. In the paper, the authors give two kind of degree bounds for \(\sigma_0\) and \(\sigma_1\). First, under the assumption that \(\deg_Yf\leq 2\), they showed that \[ \deg(\sigma_0),\deg(\sigma_1X(1-X))\leq \deg_Xf+3. \] Second, under the assumptions that 1) \(f\) is fully \(m\)-ic on \([0,1]\) -- i.e., that for each \(x\in[0,1]\), \(f(x,Y)\) has exactly degree \(m\) -- , 2) \(f\) is positive on \([0,1]\times\mathbb{R}\), and 3) \(\deg_Xf\geq 2\) (although this hypothesis can be ommitted), they show that \[ \deg(\sigma_0),\deg(\sigma_1X(1-X))\leq d^3(m+1)\frac{\inf\{f(x,y/z)z^m\mid x\in[0,1],\,y^2+z^2=1,\,z\neq 0\}}{\|f\|_\infty} \] where \(\|f\|_\infty\) is the maximum absolute value of a coefficient of \(f\). They later show that the hypothesis 2) might be weakened, if restrictions on the zeros of \(f\) are imposed.