[Tex/LaTex] Using “&” symbol to split and align multiple equations

Question

I would like to use the "&" symbol:

1. to align multi-line equations as well

2. to use it as an index to mark the placement while splitting an equation.

However, when I do so, I don't want the split equation to be in alignment with rest other equation lines. Is there a better way to do this? Thank you for your suggestions.

A minimal working example:

\documentclass[12pt]{article}
\usepackage{amsmath,amssymb,amsbsy}
\begin{eqnarray*}
\begin{split}
f(t_1,t_2|X_1,X_2) &=
\frac{\partial^2}{\partial{t_{1}}\partial{t_{2}}}\mathscr{L}\left[\Lambda_{01}(t_{1})e^{\beta^\top_{1}X_{1}}+\Lambda_{02}(t_{2})e^{\beta^\top_{2}X_{2}}\right] \\
&= \mathscr{L}^{\prime\prime}\left[\Lambda_{01}(t_1)e^{\beta_1^\top X_1}+\Lambda_{02}(t_2)e^{\beta_2^\top X_2}\right] \\
&= \frac{k+1}{k}\left(\nu_1\varphi_1{t_1}^{\varphi_1-1}e^{\beta_1^\top X_1}\right)\left(\nu_2\varphi_2{t_2}^{\varphi_2-1}e^{\beta_2^\top X_2}\right) \times\\ 
&\quad\left[\frac{k+\nu_1{t_1}^{\varphi_1}{e^{\beta_1^\top X_1}}+\nu_2{t_2}^{\varphi_2}{e^{\beta_2^\top X_2}}}{k}\right]^{-(k+2)}
\end{split}
\end{eqnarray*}
\end{document}

Answer 1

I'm not sure if I fully understand your objectives. At any rate, instead of using both an eqnarray* and a split environments, it's better to use just an align* environment.

When splitting a long expression across two (or more) lines, it's customary (but not an absolute requirement) to place the operator of consequence (here: \times) on the continuation line. And, when using | to denote conditioning on some events or variables, it's a good idea to use the \mid command to gain some well-chosen horizontal spacing around this operator. I'd also recommend giving special treatment to "e" as the base of the natural exponential function, say by declaring it as a math operator -- as is done in the modified form of your MWE.

\documentclass{article}
\usepackage{amsmath,amssymb}
\DeclareMathOperator{\e}{e} % upright-roman font, thin space ahead of "e"
\begin{document}
\begin{align*}
f(t_1,t_2\mid X_1,X_2) &=
\frac{\partial^2}{\partial{t_1}\partial{t_2}} 
\mathcal{L}\left[\Lambda_{01}(t_1) \e^{\beta_1^\top \!  X_1}+
                 \Lambda_{02}(t_2)\e^{\beta_2^\top \! X_2}\right] \\
&= \mathcal{L}^{\prime\prime}
\left[ \Lambda_{01}(t_1) \e^{\beta_1^\top \!  X_1}+
       \Lambda_{02}(t_2) \e^{\beta_2^\top \!  X_2}\right] \\
&= \frac{k+1}{k}
\left(\nu_1\varphi_1{t_1}^{\varphi_1-1}\e^{\beta_1^\top \!  X_1}\right)
\left(\nu_2\varphi_2{t_2}^{\varphi_2-1}\e^{\beta_2^\top \!  X_2}\right) \\ 
&\qquad\quad \times
\left[ \frac{k+\nu_1{t_1}^{\varphi_1}{\e^{\beta_1^\top \!  X_1}}
              +\nu_2{t_2}^{\varphi_2}{\e^{\beta_2^\top \!  X_2}}}{k} 
\right]^{-(k+2)}
\end{align*}
\end{document}