I'm reading Qing Liu. I feel confused about the proof of theorem 4.3.12 and corollary 3.14.
Theorem 4.3.12. Let $f$ be a morphism of locally Noetherian schemes. Let $x\in X$ and $y= f(x)$. Then $\dim (\mathcal{O}_{X_y,x})\geq \dim (\mathcal{O}_{X,x})-\dim (\mathcal{O}_{Y,y})$. If, moreover, $f$ is flat, then we have equality.
My question:
1.(red sentences) How do we know $t$ is not invertible in $B=\mathcal O_{X,x}$?
2.(green sentence) After base change with respect to $Y'$, $X'\rightarrow X$ is a closed immersion. But why $x$ is in $X' = X\times_YY'$?
Corollary 4.3.14. Let $f:X\rightarrow Y$ be a flat subjective morphism of algebraic varieties. We suppose that $Y$ is irreducible and that $X$ is equidimensional. Then for every $y\in Y$, the fiber $X_y$ is equidimensional, and we have $\dim(X_y) = \dim(X)-\dim(Y)$
My Question: Qing Liu defines algebraic variety as Noetherian scheme of finite type over $k$. That means $X_i$ may not be reduced. But Proposition 2.5.23(a) only works for domains. In this case how to get the highlight equality?
Theorem 2.5.15 Let $(A,\frak m)$ be a Noetherian local ring, $f\in \frak m$. Then we have $\dim(A/fA)\geq \dim(A)-1$. Moreover, equality holds if $f$ is not contained in any minimal prime ideal of $A$.
Theorem 2.5.23(a) Let $A$ be a finitely generated integral domain over $k$. Let $p$ be a prime ideal of $A$, we have $\operatorname{ht}(p) + \dim(A/p) = \dim(A)$.


$A\to \mathcal{O}_{X,x}$ is a local homomorphism of local rings, so it sends $\mathfrak{m}_A$ to $\mathfrak{m}_x$. $t$ is in the first by assumption, so it's in the second, and the maximal ideal of a local ring is precisely the non-invertible elements.
Since $Y'\to Y$ hits $y$ by construction, there are maps from $\operatorname{Spec} k(x)$ to $Y'$ and $X$ which agree after composing with the map to $Y$. So there's a map $\operatorname{Spec} k(x)\to X'$ which after composing with $X'\to X$ agrees with the map $\operatorname{Spec} k(x)\to X$ which picks out $x$. As $X'\to X$ is a closed immerison, this means $x\in X'$.
Dimension is a purely topological invariant insensitive to the scheme structure. So you can replace your scheme by its reduction.