Let $I $ denote the ideal in $Q [x,y,z]$ generated by $[x^2+y^2,xz−y,z^3−zy^3,xy+zy^2]$
Compute a generating set for $ I∩Q[y].$
Compute a generating set for $ I∩(y). $
Compute a generating set for $ (I:y) $
For $ I∩(y). $
I have set $(p1,p2,p3,p4)$ and the $ [x,0,z] $
Then I have obtained $ P1x^2 + P2xz + P3z^3 = 0 $
Is that the right step, and what do I do next?
I don't actually know where to start for the first generating set.
Thanks :)
Keeping in mind that it is an exam question, I propose the following solution:
Let's name the binomials:$$f_1=x^2+y^2 \\ f_2=xz-y \\ f_3=z^3-zy^3 \\ f_4=xy+zy^2 \\ I=\langle f_1, f_2, f_3, f_4\rangle$$
I. A generating set for $I \cap \mathbb{Q}[y]$
It means that we have to eliminate $x$ and $z$ from the equations $f_i=0, i=1,..,4$. This can be done by considering the $lex$ order $x>z>y$ and simply finding the Gröbner basis: I used Buchberger's algorithm by finding the S-polynomial ($S(f_i,f_j)$) of each pair in $F=(f_1,f_2,f_3,f_4)$ and its remainder on division by F (denoted by $\overline{S(f_i,f_j)}^F, i\neq j$). If the remainder is not zero, it is added to $F$ until all $\overline{S(f_i,f_j)}^F=0$.
$$S(f_1,f_2) = xy+zy^2 =f_4 \in I \implies \overline{S(f_1,f_2)}^F=0 \\ S(f_1,f_3) = x^2zy^3+z^3y^2 =zy^3f_1 +y^2f_3 \in I \implies \overline{S(f_1,f_3)}^F=0\\ S(f_1,f_4) = xzy^2-y^3 = y^2f_2 \in I \implies \overline{S(f_1,f_4)}^F=0\\ S(f_2,f_3) = xzy^3-z^2y = y^3f_2 + (y^4-z^2y) \notin I \implies \overline{S(f_2,f_3)}^F=-z^2y+y^4 \\ $$ Thus, we must add $f_5=-z^2y+y^4$ to our ideal $I$ and call it $I'$. Now, $F' = (f_1,f_2,f_3,f_4,f_5)$ $$S(f_1,f_5)=x^2y^4+y^3z^2y^4f_1-y^2f_5 \in I' \implies \overline{S(f_1,f_5)}^{F'}=0\\ S(f_2,f_4)=-z^2-y^2=yf_5+(-y^5-y^2) \notin I' \implies \overline{S(f_1,f_5)}^{F'}=-y^5-y^2 $$ Thus, we must add $f_6=-y^5-y^2$ to our ideal $I'$. Hence $F''=(f_1,...,f_6)$ It turns out that the remaining S-polynomials yield $\overline{S(f_1,f_6)}^{F''}=\overline{S(f_2,f_5)}^{F''}=\overline{S(f_2,f_6)}^{F''}=\overline{S(f_3,f_4)}^{F''}=\overline{S(f_3,f_5)}^{F''}=\overline{S(f_3,f_6)}^{F''}=\overline{S(f_4,f_5)}^{F''}=\overline{S(f_4,f_6)}^{F''}=\overline{S(f_5,f_6)}^{F''}=0$.
Therefore the Gröbner basis (It is easy to verify that it is also the reduced Gröbner basis by observing the leading terms) of I is $$G=\langle x^2+y^2, xz-y, z^3-zy^3, xy+z^2, z^2y-y^4, y^5+y^2\rangle$$ resulting in $I \cap \mathbb{Q}[y] =\langle y^5+y^2 \rangle$.
II. A generating set for $I \cap \langle y \rangle$
Like you started, we are looking for a polynomial $f \in I$ and $f \in \langle y \rangle$ at the same time. Let $$f=p_1f_1+p_2f_2+p_3f_3+p_4f_4 = q_1y \tag{1}$$ where $p_1,p_2,p_3,p_4$ and $q_1$ are functions of $x,y$ and $z$. Our aim is to find $p_i,i=1,..,4$. $\{(y,0,0,0),(0,y,0,0),(0,0,y,0),(0,0,0,1)\}$ are trivial solutions. When $y=0$, equation $(1)$ becomes $$r_1x^2+r_2xz+r_3z^3=0 \tag{2}$$ where $r_i$ are functions of $x,z$. The generating set of all solutions of equation $(2)$ is $\{(z,-x,0),(0,z^2,-x),(z^3,0,-x^2)\}$.
Substituting these 7 solutions in equation $(1)$ gives us the following polynomials $$g_1=x^2y+y^3\\ g_2=xyz-y^2\\ g_3=z^3y-zy^4\\ g_4=xy+zy^2\\ g_5=xy+zy^2=g_4\\ g_6=xzy^3-yz^2\\ g_7=x^2zy^3+z^3y^2=zy^2g_1+yg_3\\$$
Hence, the generating set for the given ideal is $\langle x^2y+y^3, xyz-y^2,z^3y-zy^4, xy+zy^2, xzy^3-yz^2 \rangle$. Note that this is not a Gröbner basis and finding it shouldn't be hard by following the same procedure for the first question. It turns out to be$$I \cap \langle y \rangle = \langle y^5+y^2, z^2y-y^4, xy+zy^2 \rangle$$
III. A generating set for $I:y$
There is a theorem that states if $\{h_1,...,h_p\}$ is a basis of the ideal $I \cap \langle g \rangle$, then $\{\dfrac{h_1}{g},...,\dfrac{h_p}{g}\}$ is a basis of $\langle I:g \rangle$. Therefore, $$I:y = \langle \frac{y^5+y^2}{y}, \frac{z^2y-y^4}{y}, \frac{xy+zy^2}{y} \rangle\\ = \langle y^4+y, z^2-y^3, x+zy \rangle$$
I have only briefly read the Cox-Little-O'shea book and I have come up with this solution based on my understanding. Nonetheless, there might be a simpler solution or a way to shorten these computations (probably by using improvements of the Buchberger's algorithm). I will be happy to know if someone can suggest improvements or shortcuts to these approaches.