Ttyjfyk

I am using an array environment to get aligned portions of a series of equations to center (instead of left-justify), as shown below:

usepackagearray,amsmath
[
beginarray>displaystylec @= >displaystylec @+ >displaystylec @+ >displaystylec
sumlimits_r=0^n+1 binomn+1r & binomn+10 & binomn+11 + ldots + binomn+1n & binomn+1n+1 \
& 1 & sumlimits_r=1^n binomn+1r & 1 \
endarray
]

The array environment (I believe) is necessary here to get each of the columns to center instead of justifying left.

Now my problem is that these two lines are part of a greater series of equations, where the others do not follow this pattern to be aligned. However, I need the equals signs to line up across all lines.

My current approach is follow the array with a normal align environment, having one equation line mirroring the longest line above but enclosed in phantom to get the align spacing right. But this leaves a single empty line with an equals in it.

...

beginalign*
&= 2 + sum_r=1^nleft[binomnr + binomnr-1right] \
phantomsumlimits_r=0^n+1 binomn+1r &= phantom binomn+10 + binomn+11 + ldots + binomn+1n + binomn+1n+1
endalign*

How can I get this result, but without the extraneous equals line at the end? Preferable a more elegant one, as this idea relies on several iffy factors such as none of the following equations exceeding the size of the one governing the special alignment.

try

documentclassarticle
usepackagearray,amsmath
begindocument
[
beginarray>displaystylec @= >displaystylec @+ >displaystylec @+ >displaystylec
sum_r=0^n+1 binomn+1r 
 & binomn+10 & binomn+11 + ldots + binomn+1n & binomn+1n+1 \
 & 1 & sumlimits_r=1^n binomn+1r & 1 \
 & multicolumn3>displaystylel
 2 + sum_r=1^nleft[binomnr + binomnr-1right]
 
endarray
]
enddocument

Use the [t] option. Then you do not need to use multicolumn many times if you have many subsequent lines.

documentclassarticle
usepackagearray,amsmath
begindocument
beginalign*
sumlimits_r=0^n+1 binomn+1r
&beginarray[t]@>displaystylec @=@>displaystylec @+ >displaystylec @+ >displaystylec
& binomn+10 & binomn+11 + ldots + binomn+1n & binomn+1n+1 \
& 1 & sumlimits_r=1^n binomn+1r & 1 \
endarray\
&=2 + sum_r=1^nleft[binomnr + binomnr-1right]
endalign*
enddocument

eqparbox allows you to store the lengths of boxes via a <tag>. Boxes with the same <tag> are set with the maximum width across all content. Below I use this approach with a newly-defined eqmathbox[<tag>][<align>] (default for <align> is to centre the content) to add content to three different <tag>ged boxes:

documentclassarticle

usepackageeqparbox,xparse,amsmath

% https://tex.stackexchange.com/a/34412/5764
makeatletter
NewDocumentCommandeqmathboxo Oc m%
 IfValueTF#1
 defeqmathbox@##1##2eqmakebox[#1][#2]$##1##2$
 defeqmathbox@##1##2eqmakebox$##1##2$
 mathpaletteeqmathbox@#3

makeatother

begindocument

beginalign*
sum_r = 0^n + 1 binomn + 1r 
 &= eqmathbox[LEFT]binomn + 10 + eqmathbox[CENTRE]binomn + 11 + dots + binomn + 1n + eqmathbox[RIGHT]binomn + 1n + 1 \
 &= eqmathbox[LEFT]1 + eqmathbox[CENTRE]sum_r = 1^n binomn + 1r + eqmathbox[RIGHT]1 \
 &= 2 + sum_r = 1^n biggl[ binomnr + binomnr - 1 biggr]
endalign*

enddocument

Since eqparbox uses TeX's label-ref system, you need to compile twice for every change in the content of the maximum width.