Gdtyjr

I find that I have unpleasant spacing between the equals sign and (e.g.) the exponential function in this particular case, while using the align environment. I have a long expression which has to be split up in multiple rows. The alignment character & appears to gobble up all space when it comes after the equals sign. How could I remedy this while still preserving the plus sign alignment below?

Ideally, I would like to have the second row from the first equation, and the first row from the second equation.

documentclassarticle
usepackageamsmath
begindocument
beginalign % bad spacing at first row, but correctly placed second row
 2 cosh t =& e^t \
 &+ e^-t 
endalign
beginalign % good spacing at first row, but incorrectly placed second row
 2cosh t &= e^t \
 &+ e^-t 
endalign
enddocument

Put the ampersand before the equals sign. Then use quad to create the indentation in the second row.

documentclassarticle
usepackageamsmath
begindocument
beginalign 
 2 cosh t &= e^t \ 
 &quad+ e^-t 
endalign
enddocument

You can also use hspace if you want a different length for the indentation.

I wouldn't try aligning the plus with e^t, but if you insist, here's how.

documentclassarticle
usepackageamsmath
begindocument

noindent
The plus is flush with $e^t$ (I wouldn't recommend it):
beginalign
2cosh t =& e^t \
 & negmedspace+ e^-t
endalign
The plus is moved right (better):
beginalign
2cosh t &= e^t \
 &qquad+ e^-t
endalign
enddocument

With negmedspace we kill the space at the left of the binary operation symbol.

However, align is the wrong tool here:

documentclassarticle
usepackageamsmath
begindocument

beginequation
beginsplit
2cosh t &= e^t \
 &qquad + e^-t
endsplit
endequation

Suppose you wish to ensure that the two instances of e are aligned vertically, while respecting the fact that a binary operator (+) precedes the e in the second row. The most direct way to obtain this type of alignment is to use a pair of hphantom ("horizonal phantom") statements. The one in the first row mimics the + symbol (a binary operator) from the second row, and the hphantom statement in the second row mimics the = symbol (a relational operator) from the first row. The pairs are there to help TeX figure out which type of operator applies.

documentclassarticle
usepackageamsmath
setlengthtextwidth3in %% just for this example
begindocument
beginalign
 2cosh t &= phantom+ mathrme^t \
 &phantom= + mathrme^-t
endalign
enddocument 

You have placed the & wrong; it should go before the equal sign to get the correct spacing. Also, I've used hphantom to indent the expression in the second line to get the correct alignment. (Notice the before =.)

documentclassarticle

usepackageamsmath

begindocument

beginalign
 2cosh t
 &= e^t \
 &hphantom= + e^-t 
endalign

enddocument

Strictly speaking, only the first part of egreg's answer solves the problem as intended by Martin L. All other answers require space corrections of at least 1 or 2 mu. Here is an alternative.

LaTeX encloses relation symbols with thick spaces ;, and binary symbols with medium spaces :. The symbols = and + in our example are respectively of these kinds. Therefore, downgrading them to ordinary symbols, what we want is:

2 cosh t & ; mathord= ; e^t \
 & ; phantom= ; mathord+ : e^-t

A practical incarnation of the above is (see page 36 of l2kurz.pdf):

2 cosh t & = e^t \
 & mathrelphantom= negmedspace + e^-t

First, we redeem the relation status of = robbed by the phantom command. Second, we insert an empty group telling LaTeX to interpret + as a binary rather than a prefix symbol; but this creates a spurious medium space : that needs to be compensated.