ScalingStacks

6.5 Geometric interpretation [03M0]

Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.

Complete original source context · Original author HTML

6.5 Geometric interpretation

Suppose that u,vu,v are solutions of (32) near (0,0)(0,0) in ℝ2\mathbin{\mathbb{R}}^{2}, with an isolated singularity of order kk at (0,0)(0,0), and that u⁡(0,0)=v⁡(0,0)=0u(0,0)=v(0,0)=0. Let NN be the associated SL 3-fold in ℂ3\mathbin{\mathbb{C}}^{3}, defined by (31) with a=0a=0. Then NN has an isolated singular point at (0,0,0)(0,0,0). What can we say about it?

Well, the coordinates x,yx,y give a natural map from NN to ℝ2\mathbin{\mathbb{R}}^{2}. The fibre of this map over (0,0)(0,0) is a point, and the fibre of the map over other points (x,y)≠(0,0)(x,y)\neq(0,0) near (0,0)(0,0) in ℝ2\mathbin{\mathbb{R}}^{2} is a circle. Thus, NN near (0,0,0)(0,0,0) has the topology of 𝒮1×ℝ2{\mathcal{S}}^{1}\times\mathbin{\mathbb{R}}^{2} with 𝒮1×{(0,0)}{\mathcal{S}}^{1}\times\bigl\{(0,0)\bigr\} collapsed to a point. That is, topologically NN is a T2T^{2}-cone near (0,0,0)(0,0,0).

The author conjectures that when k=1k=1, to leading order uu and vv should agree with the functions u,vu,v of Proposition 6.7 or 6.8 with a=b=c=0a=b=c=0, at least in the generic case, and therefore that the tangent cone to NN at (0,0,0)(0,0,0) should be one of the T2T^{2}-cones L0±L_{0}^{\pm} from (16). This gives a good description of the local geometry of NN when k=1k=1.

When k>1k>1, the author conjectures that u,vu,v satisfy

u⁡(x,y)=o⁡(x2+|y|)andv⁡(x,y)=o⁡(|x|+|y|1/2)for small x,y.u(x,y)=o\bigl(x^{2}+|y|\bigr)\quad\text{and}\quad v(x,y)=o\bigl(|x|+|y|^{1/2}\bigr)\quad\text{for small $x,y$.} (48)

This is because u⁡(x,y)=O⁡(x2+|y|)u(x,y)=O(x^{2}+|y|) and v⁡(x,y)=O⁡(|x|+|y|1/2)v(x,y)=O(|x|+|y|^{1/2}) when k=1k=1, and by analogy with the zeros of holomorphic functions we expect zeros of higher order to decrease more quickly near (0,0)(0,0).

Let t>0t>0, and define t−1​N={t−1​𝐳:𝐳∈N}t^{-1}N=\{t^{-1}{\bf z}:{\bf z}\in N\}. Then t−1​Nt^{-1}N is also an SL 3-fold, and may be written in the form (30) with u,vu,v replaced by

ut​(x,y)=t−2​u​(t​x,t2​y)andvt​(x,y)=t−1​v​(t​x,t2​y).u^{t}(x,y)=t^{-2}u(tx,t^{2}y)\quad\text{and}\quad v^{t}(x,y)=t^{-1}v(tx,t^{2}y).

Now equation (48) implies that ut​(x,y)→0u^{t}(x,y)\rightarrow 0 and vt​(x,y)→0v^{t}(x,y)\rightarrow 0 as t→0t\rightarrow 0 for fixed x,yx,y. It follows that t−1​Nt^{-1}N converges to

N0={(z1,z2,z3)∈ℂ3:Re(z1z2)=0,Im(z3)=0,|z1|=|z2|}N_{0}=\bigl\{(z_{1},z_{2},z_{3})\in\mathbin{\mathbb{C}}^{3}:\mathop{\rm Re}(z_{1}z_{2})=0,\quad\mathop{\rm Im}(z_{3})=0,\quad|z_{1}|=|z_{2}|\bigr\}

as t→0t\rightarrow 0, and this is the tangent cone to NN at (0,0,0)(0,0,0).

But this is just the union of the two special Lagrangian 3-planes

Π+={(z,iz¯,x):z∈ℂ,x∈ℝ},Π−={(z,−iz¯,x):z∈ℂ,x∈ℝ},\Pi_{+}=\bigl\{(z,i\bar{z},x):z\in\mathbin{\mathbb{C}},\quad x\in\mathbin{\mathbb{R}}\bigr\},\quad\Pi_{-}=\bigl\{(z,-i\bar{z},x):z\in\mathbin{\mathbb{C}},\quad x\in\mathbin{\mathbb{R}}\bigr\},

which intersect in the real line {(0,0,x):x∈ℝ}\bigl\{(0,0,x):x\in\mathbin{\mathbb{R}}\bigr\}. Thus, when k>1k>1 we expect NN to resemble the union of two SL 3-planes Π±\Pi_{\pm} intersecting in a line, to leading order near (0,0,0)(0,0,0).

As this tangent cone is singular not just at (0,0,0)(0,0,0) but all along the line Π+∩Π−\Pi_{+}\cap\Pi_{-}, to have a good picture of NN near (0,0,0)(0,0,0) we need to include the next nonzero terms in uu and vv as well. Unfortunately, the author does not know what these terms are. But here is a rather crude approximation, which illustrates the kind of behaviour we expect.

For k>1k>1 even, suppose that u⁡(x,y)≈|x|αu(x,y)\approx|x|^{\alpha} for some α>2\alpha>2, and v⁡(x,y)≈0v(x,y)\approx 0 for small x,yx,y. Then we have

N≈{(z1,z2,z3)∈ℂ3:Re(z1​z2)=|Re(z3)|α,Im(z3)=0,|z1|=|z2|}.\begin{split}N\approx\bigl\{(z_{1},z_{2},z_{3})\in\mathbin{\mathbb{C}}^{3}:&\mathop{\rm Re}(z_{1}z_{2})=|\mathop{\rm Re}(z_{3})|^{\alpha},\\ &\mathop{\rm Im}(z_{3})=0,\quad|z_{1}|=|z_{2}|\bigr\}.\end{split} (49)

This may be written more nicely in different coordinates on ℂ3\mathbin{\mathbb{C}}^{3}. Define new coordinates (w1,w2,x1,x2)(w_{1},w_{2},x_{1},x_{2}) on ℂ3\mathbin{\mathbb{C}}^{3} by

w1=z1−i​z¯2,w2=z2+i​z¯1,x1=Re(z3),x2=Im(z3).w_{1}=z_{1}-i\bar{z}_{2},\quad w_{2}=z_{2}+i\bar{z}_{1},\quad x_{1}=\mathop{\rm Re}(z_{3}),\quad x_{2}=\mathop{\rm Im}(z_{3}).

Then w1​w2=2​Re(z1​z2)+i⁡(|z1|2−|z2|2)w_{1}w_{2}=2\mathop{\rm Re}(z_{1}z_{2})+i\bigl(|z_{1}|^{2}-|z_{2}|^{2}\bigr). Therefore, in these new coordinates, (49) becomes

N≈{(w1,w2,x1,x2)∈ℂ2×ℝ2:w1w2=2|x1|α,x2=0}.N\approx\bigl\{(w_{1},w_{2},x_{1},x_{2})\in\mathbin{\mathbb{C}}^{2}\times\mathbin{\mathbb{R}}^{2}:w_{1}w_{2}=2|x_{1}|^{\alpha},\quad x_{2}=0\bigr\}. (50)

So NN may be thought of as approximating a slowly varying 1-parameter family of complex quadratics w1​w2=cw_{1}w_{2}=c in ℂ2\mathbin{\mathbb{C}}^{2}, for c∈ℝc\in\mathbin{\mathbb{R}} varying with x1x_{1}. When x1=0x_{1}=0 the quadratic degenerates into w1​w2=0w_{1}w_{2}=0, the union of two complex lines in ℂ2\mathbin{\mathbb{C}}^{2}. For k>1k>1 odd, the appropriate approximation is u⁡(x,y)≈x​|x|α−1u(x,y)\approx x|x|^{\alpha-1} for some α>2\alpha>2 and v⁡(x,y)≈0v(x,y)\approx 0, and then

N≈{(w1,w2,x1,x2)∈ℂ2×ℝ2:w1w2=2x1|x1|α−1,x2=0}.N\approx\bigl\{(w_{1},w_{2},x_{1},x_{2})\in\mathbin{\mathbb{C}}^{2}\times\mathbin{\mathbb{R}}^{2}:w_{1}w_{2}=2x_{1}|x_{1}|^{\alpha-1},\quad x_{2}=0\bigr\}. (51)

We stress that (50) and (51) are just very approximate guesses, and the true behaviour of u,vu,v and NN will be different and more complicated than this.

Original mathematics by the credited authors. Source-backed reader collection; mathematical self-containment is not assessed.