Abstract Generated abstract
This note studies how planar area measure changes under a quasiconformal mapping whose dilatation is close to one. For a normalized mapping between rectangles, the argument estimates a boundary integral by the Jacobian and derives an inequality involving the square root of the Jacobian, then applies Bunyakovsky’s inequality to subsets of the domain. The resulting estimate bounds the change in the area of a measurable set, giving, with \(1 - 1/q = \varepsilon\), \(|\Delta \sigma| \leq \varepsilon |1 - 2\sigma| + 2\sqrt{\sigma(1-\sigma)}\sqrt{\varepsilon+\varepsilon^2}\), and hence \(|\Delta \sigma| \leq \sqrt{\varepsilon}+O(\varepsilon)\) for small \(\varepsilon\). The same method also yields another proof that quasiconformal mappings in the relevant closure class send sets of measure zero to sets of measure zero.
Full Text
MATHEMATICS
P. P. BELINSKII
ON THE AREA MEASURE UNDER A QUASICONFORMAL MAPPING
(Presented by Academician M. A. Lavrent'ev on 6 III 1958)
It is known that under a \(q\)-quasiconformal mapping with small \(1 - 1/q = \varepsilon\), the mapping is close to a conformal one, and, under the corresponding normalization, close to the identity mapping \((^1,^2)\). However, up to now estimates of the change of area under such a mapping have been unknown. In the present note an estimate is given for the change of area of the form \(|\Delta \sigma| \leq \sqrt{\varepsilon} + O(\varepsilon)\) for \(q\)-quasiconformal mappings. Along the way, the fact is once again proved that under quasiconformal mappings, sets of measure zero go into sets of measure zero \((^3)\).
Let the rectangle \(K: 0 < x < \lambda,\ 0 < y < 1/\lambda\) of the plane \(z = x + iy\) be mapped \(q\)-quasiconformally onto the rectangle \(0 < u < \mu,\ 0 < v < 1/\mu\) of the plane \(w = u + iv\), in such a way that the vertices go into the corresponding vertices. Let us compute \(\operatorname{Im} \int_{\Gamma} w(z)\,dz\), where \(\Gamma\) is the boundary of the rectangle \(K\):
\[ \operatorname{Im}\int_{\Gamma} w(z)\,dz = \int_{\Gamma} u\,dy + v\,dx = \int_0^{1/\lambda} \mu\,dy - \int_0^\lambda \frac{1}{\mu}\,dx = \frac{\mu}{\lambda} - \frac{\lambda}{\mu}. \]
On the other hand:
\[ \left|\int_{\Gamma} w(z)\,dz\right| \leq \iint_K |-(u_y + v_x) + i(u_x - v_y)|\,dx\,dy \leq \left(\sqrt{q} - \frac{1}{\sqrt{q}}\right) \iint_K \sqrt{\frac{D(u,v)}{D(x,y)}}\,dx\,dy . \]
Comparing with the preceding, we obtain
\[ \left|\frac{\mu}{\lambda} - \frac{\lambda}{\mu}\right| \leq \left(\sqrt{q} - \frac{1}{\sqrt{q}}\right) \iint_K \sqrt{\frac{D(u,v)}{D(x,y)}}\,dx\,dy . \tag{1} \]
Now let, for definiteness, \(\mu \geq \lambda\). Subject the plane \(w\) to a compression in the direction of the \(v\)-axis by \(t \geq 1\) times and to a stretching along the \(u\)-axis by the same factor (in the case \(\mu < \lambda\) we interchange \(u\) and \(v\)). In this case there will be obtained a mapping of the rectangle \(K\) onto a rectangle with sides \(t\mu\) and \(1/t\mu\). The characteristic \(p(z)\), obviously, will not exceed \(t^2 q\), and the Jacobian will not change. Then, according to (1):
\[ \left|\frac{t\mu}{\lambda} - \frac{\lambda}{t\mu}\right| \leq \left(t\sqrt{q} - \frac{1}{t\sqrt{q}}\right) \iint_K \sqrt{\frac{D(u,v)}{D(x,y)}}\,dx\,dy . \]
Dividing both sides of the obtained inequality by \(t\), passing to the limit as \(t \to \infty\), and taking into account that \(\mu \geq \lambda\), we obtain
\[ \frac{1}{\sqrt{q}} \leq \iint_K \sqrt{\frac{D(u,v)}{D(x,y)}}\,dx\,dy . \tag{2} \]
We divide the rectangle \(K\) into two subsets \(S_1\) and \(S_2\). Applying Bunyakovsky’s inequality to (2), we obtain
\[ \frac{1}{\sqrt{q}} \leq \left[ \operatorname{mes} S_1 \cdot \iint_{S_1} \frac{D(u,v)}{D(x,y)}\, dx\, dy \right]^{1/2} + \left[ \operatorname{mes} S_2 \cdot \iint_{S_2} \frac{D(u,v)}{D(x,y)}\, dx\, dy \right]^{1/2}. \tag{3} \]
From the inequality obtained one can derive yet another proof of the fact that a mapping from the closure of the class of continuously differentiable \(q\)-quasiconformal mappings carries a set of measure zero into a set of measure zero.
To prove this, let us first note that inequality (3) remains valid, since it is known that Green’s formula is legitimately applicable in the case of a mapping belonging to the closure of the class of \(q\)-quasiconformal mappings.
Denote the images of the sets \(S_1\) and \(S_2\) by \(S'_1\) and \(S'_2\). It is clear that
\[ \operatorname{mes} S'_2 \geq \iint_{S_2} \frac{D(u,v)}{D(x,y)}\, dx\, dy . \]
If \(\operatorname{mes} S_1 = 0\), then from (3) it follows that
\[ \operatorname{mes} S'_1 = 1-\operatorname{mes} S'_2 \leq 1-\iint_{S_2} \frac{D(u,v)}{D(x,y)}\, dx\, dy \leq 1-\frac{1}{q}. \tag{4} \]
If we assume that \(\operatorname{mes} S'_1 > 0\), then, considering the mapping in a neighborhood of some point of density of \(S'_1\), it is possible, by means of auxiliary conformal transformations, to obtain a mapping carrying a set of measure zero into a set whose measure is arbitrarily close to one, which contradicts (4).
Denote \(\operatorname{mes} S_1\) by \(\sigma\), and \(\operatorname{mes} S'_1\) by \(\sigma+\Delta\sigma\). Then (3) gives
\[ \frac{1}{\sqrt{q}} \leq \sqrt{\sigma(\sigma+\Delta\sigma)} + \sqrt{(1-\sigma)(1-\sigma-\Delta\sigma)} . \tag{3'} \]
Solving the last inequality with respect to \(\Delta\sigma\) and denoting \(1-1/q=\varepsilon\), we obtain
\[ |\Delta\sigma| \leq \varepsilon |1-2\sigma| + 2\sqrt{\sigma(1-\sigma)}\sqrt{\varepsilon+\varepsilon^2}, \tag{5} \]
or, for small \(\varepsilon\),
\[ |\Delta\sigma| \leq \sqrt{\varepsilon}+O(\varepsilon). \tag{5'} \]
Inequality (5′) is easily transferred to the case of a mapping of the disk onto itself with the normalization \(w(0)=0\).
Mathematical Institute
Siberian Branch of the Academy of Sciences of the USSR
Received
15 II 1958
CITED LITERATURE
\(^{1}\) M. A. Lavrent’ev, Matem. sborn., 42, issue 2 (1935).
\(^{2}\) P. P. Belinskii, DAN, 91, No. 2 (1953).
\(^{3}\) I. P. Pesin, DAN, 102, No. 2 (1955).