how to expand variables in tensorflow to a higher dimension?

Question

I have defined a weight matrix w,and the matrix is in shape [1, 500], but I want to copy w to new variable w1 in shape [100, 1, 500]. in w1, the 100 matrix in shape [1, 500] are all value-equal and they can change its value at the same time when running Back-propagation algorithm, how can I do with that?