Page 136 - Geometric Modeling and Algebraic Geometry
P. 136

7 Curve Parametrization Exploiting the Newton Polygon  137
                                     0= c 0 ,                 0= c 2 − c 1 ,
                                                                       1
                                                                  1
                                                                  4    4
                                     0= c 0 − c 1 + c 3 ,               5
                                                              0= −c 2 − c 1 +5c 3 + c 4
                                                                        2
                           Note that this step would in general yield linear constraints over a field extension, e.g.
                           over Q(γ 1 ) or Q(γ 2 ). They can always be written as a bigger number of constraints
                           over the ground field, here Q. This is the algorithmic side of the statement that the
                           divisor K [1,1] is Q-rational. Solving the system with respect to c 3 and setting c 3 =1,

                                                                                       (K [7,8] ):
                                           2      2  3 2 2
                           we get g 1 = xy + x y + xy − x y . We do the same again for g 2 ∈L   C
                                                     2
                                         g 2 := c 0 y + c 1 xy + c 2 x y + c 3 y + c 4 xy + c 5 x y
                                                                                  2 2
                                                                    2
                                                                           2
                                                             2
                                                           2                    2
                                     g 2,1,[7,8] = c 0 + c 1 u 1 + c 2 u + c 3 v 1 + c 4 u 1 v 1 + c 5 u v 1
                                                           1                    1
                                                                          yields the following con-
                                                       and ν 2 (g 2,1,[7,8] ) ≥ α ν 2
                              Requiring ν 1 (g 2,1,[7,8] ) ≥ α ν 1
                           straints:
                                    0= c 0 − c 1 + c 2 ,  0= c 4 − 2c 2 − c 5 − c 3 + c 1 ,
                                                                              1
                                                              1
                                                                         1
                                                              4          4    4
                                    0= c 0 − c 3 ,                  5
                                                          0= −c 4 − c 3 + c 1
                                                                    2
                           Solving the system with respect to c 3 and c 4 and setting c 3 =2 and c 4 = −1 (which
                           is a matter of choice), we get g 2 =2y +4xy +2x y +2y − xy − 3x y .
                                                                                     2 2
                                                                          2
                                                                   2
                                                                               2
                              The system for computing g 1 had got an 1-dimensional solution. This implies that
                           the divisor (g 1 ) [1,1] is equal to A. Hence its support is exactly the preimage of the

                           singular locus. On the other hand we were left with essentially one degree of freedom
                           when computing g 2 and therefore (g 2 ) [7,8] > A. Indeed we find that the support of

                           g 2 has an additional point corresponding to Q 3 ∈ U 1 , namely (u 1 ,v 1 )=(−1, −16).
                           Therefore we have to consider one more valuation ν 3 centered at Q 3 (which one
                           could get again from a Puiseux series solution at Q 3 ). All this is reflected in the
                           twisted orders of g 1 and g 2 :
                                  = ν 1,[1,1] (g 1 )=2,  β 1,ν 2  = ν 2,[1,1] (g 1 )=2,  β 1,ν 3  = ν 3,[1,1] (g 1 )=0,
                             β 1,ν 1
                                  = ν 1,[7,8] (g 2 )=2,  β 2,ν 1  = ν 2,[7,8] (g 2 )=2,  β 2,ν 3  = ν 3,[7,8] (g 2 )=1
                             β 2,ν 1
                                                                         =0.
                           Since Q 3 ∈ C is a smooth point, the adjoint order is α ν 3
                           7.7.4 Linear system of an anticanonical divisor
                                                                  (K [7,1] ):

                           Now we make an Ansatz for an element h ∈L   C
                                  h := c 1 + c 2 x + c 3 y + c 4 xy + c 5 x y + c 6 y + c 7 xy + c 8 x y
                                                                                   2 2
                                                                             2
                                                                      2
                                                               2
                              h 1,[7,1] := c 1 + c 2 u 1 + c 3 v 1 + c 4 u 1 v 1 + c 5 u v 1 + c 6 v + c 7 u 1 v + c 8 u v
                                                                          2
                                                                                        2 2
                                                                                  2
                                                                  2
                                                                  1       1       1     1 1
                                                                                             −
                                                                                      + β 2,ν 2
                                                             − α ν 1
                                                       + β 2,ν 1
                              Requiring ν 1 (h 1,[7,1] ) ≥ β 1,ν 1
                                                                  =2, ν 2 (h 1,[7,1] ) ≥ β 1,ν 2
                                                                  =1 (i.e. h 1,[7,1] also has to vanish
                           α ν 2
                               =2 and ν 3 (h 1,[7,1] ) ≥ β 1,ν 3
                                                             − α ν 3
                                                      + β 2,ν 3
                           on Q 3 ) yields the following constraints:
   131   132   133   134   135   136   137   138   139   140   141