# Aligning cases conditions from multiple nesting levels

\begin{dmath*}
y_{ic}[n] = \begin{dcases}
y_0 + x_0 - x_0 (1 - k_2)^n & k_1 = 0 \\
k_5 + \left( y_0 - k_5 \right) (1 - k_1)^n + \begin{dcases}
x_0 k_1 n (1 - k_1)^{n-1} & k_1 \neq 0, k_1 = k_2 \\
x_0 \frac{k_2}{k_1 - k_2} \left[ \begin{split}&(1 - k_2)^n \\ & - (1 - k_1)^n \end{split} \right] & k_1 \neq 0, k_1 \neq k_2
\end{dcases}
\end{dcases}
\end{dmath*}


Is there any way I can get the conditions, which come from multiple dcases nesting levels, to appear in a single vertical column? (This would not only look nicer but save a lot of horizontal space, allowing the LHS of the entire equation to fit on the same line instead of wrapping.)

Followup: The solution provided by Harish worked very nicely... and then I noticed that I didn't need a separate case for $k_1 = 0$ (The analysis was different but led to the same result as the others -- I guess that's fairly common with partial fraction expansions).

• I believe it would be clearer with a single three item cases environment. – egreg Dec 31 '12 at 17:11
• Related: How to align nested cases? – Werner Dec 31 '12 at 17:48
• @Werner: I already found and tried to apply that approach... but it didn't work correctly since my third row is taller than the minimum. – Ben Voigt Dec 31 '12 at 18:27
• IMHO using Iversonians or Characteristic Functions instead of case-based definitions would make this expression clearer. – user10274 Jan 1 '13 at 9:02
• @BenVoigt -- yes, i did, but obviously not carefully enough. you have undone the intended correction, which was to double the backslash intended to separate the lines in dcases, a problem which was caused by a system update affecting well over 7000 items. (please look at the meta explanation of the mess. we've been left by the system managers to fix it on our own; we're trying desperately to do that before we get more questions about code that doesn't work.) i apologize for the mess, but ask you to please make that necessary correction. – barbara beeton Jun 12 '17 at 0:45

If I were to do this, I would have listened to egreg's comments above. But the following works though coding may be ugly:

\documentclass{article}
\usepackage{mathtools,breqn}
\begin{document}
\begin{dmath*}
y_{ic}[n] = \begin{dcases}
y_0 + x_0 - x_0 (1 - k_2)^n & k_1 = 0 \\
k_5 + \left( y_0 - k_5 \right) (1 - k_1)^n + \begin{dcases}
x_0 k_1 n (1 - k_1)^{n-1}  &  \makebox[-11pt][l]{$k_1 \neq 0, k_1 = k_2$}\\
x_0 \frac{k_2}{k_1 - k_2} \left[ \begin{aligned}(1 - k_2)^n \\  - (1 - k_1)^n \end{aligned} \right] & \makebox[-11pt][l]{$k_1 \neq 0, k_1 \neq k_2$}
\end{dcases}
\end{dcases}
\end{dmath*}
\end{document}


I think I would do it in the way that egreg hinted at, and just use one dcases envrionment

Note that I have replaced your split environment with an array and used @{}c@{} to remove column space (check your original .log file- you should have got a warning that you weren't allowed to use split in that context).

MWE

\documentclass{article}
\usepackage{mathtools}

\begin{document}
$y_{ic}[n] = \begin{dcases} y_0 + x_0 - x_0 (1 - k_2)^n & k_1 = 0 \\ k_5 + \left( y_0 - k_5 \right) (1 - k_1)^n + x_0 k_1 n (1 - k_1)^{n-1} & k_1 \neq 0, k_1 = k_2 \\ k_5 + \left( y_0 - k_5 \right) (1 - k_1)^n+ x_0 \frac{k_2}{k_1 - k_2} \left[ \begin{array}{@{}c@{}}(1 - k_2)^n \\ - (1 - k_1)^n \end{array} \right] & k_1 \neq 0, k_1 \neq k_2 \end{dcases}$
\end{document}


Here are some more options that you could explore. Note that cases and friends can be easily emulated using an ordinary array:

\documentclass{article}
\begin{document}
$\setbox1=\hbox{k_1 \neq 0, k_1 = k_2} y_{ic}[n] = \left\{\begin{array}{@{}l@{\quad}l} y_0 + x_0 - x_0 (1 - k_2)^n, & \hspace*{\dimexpr-1em-\wd1-\nulldelimiterspace}k_1 = 0 \\[2\jot] k_5 + (y_0 - k_5) (1 - k_1)^n + \left\{\begin{array}{@{}l@{\quad}l@{}} x_0 k_1 n (1 - k_1)^{n-1}, & k_1 \neq 0, k_1 = k_2 \\[\jot] \frac{x_0 k_2}{k_1 - k_2} [(1 - k_2)^n - (1 - k_1)^n], & k_1 \neq 0, k_1 \neq k_2 \end{array}\right. \end{array}\right.$

\bigskip

$\setbox1=\hbox{k_1 \neq 0, k_1 = k_2} y_{ic}[n] = \left\{\begin{array}{@{}l@{\quad}l} y_0 + x_0 - x_0 (1 - k_2)^n, & \hspace*{\dimexpr-1em-\wd1-\nulldelimiterspace}k_1 = 0 \\[2\jot] k_5 + (y_0 - k_5) (1 - k_1)^n \\ \qquad {} + \left\{\begin{array}{@{}l@{\quad}l@{}} x_0 k_1 n (1 - k_1)^{n-1}, & k_1 \neq 0, k_1 = k_2 \\[\jot] \frac{x_0 k_2}{k_1 - k_2} [(1 - k_2)^n - (1 - k_1)^n ], & k_1 \neq 0, k_1 \neq k_2 \end{array}\right. \end{array}\right.$
\end{document}


The space-correction in the first row consists of 3 parts:

1. An em skip, stemming from the forced @{\quad} in the separation between the two columns;
2. \wd1 (or the width of \box1) equal to the width of the condition(s) in row 2 and 3;
3. \nulldelimiterspace, since using \right. actually inserts a small gap.
• What you've done is make both rows in the subordinate case block the same height... an inline fraction makes that possible but ugly. I'd already tried this with a display-mode fraction, but the braces and conditions didn't line up then. Thanks for the ideas though. – Ben Voigt Dec 31 '12 at 18:56