Please see

https://www.researchgate.net/topic/Principal-Component-Analysis

https://en.wikipedia.org/wiki/Multivariate_normal_distribution

Hello dear colleague

thank you very much

Jolliffe I., 1986. Principal component analysis. Wiley Online Library. Springer serie. 518 p.

http://cda.psych.uiuc.edu/statistical_learning_course/Jolliffe%20I.%20Principal%20Component%20Analysis%20(2ed.,%20Springer,%202002)(518s)_MVsa_.pdf

PCA ia an exploratory technique, You can apply PCA on non normal multivariate data.

Thank u veru much!

PCA - I would like to confirm José Francisco Moreira Pessanha's answer.  

PCA (and SVD) are about the rotation and scaling of data, not about the probability distribution of a data set.

A good discussion appears in https://arxiv.org/ftp/physics/papers/0208/0208101.pdf

Gaussian - after you apply PCA (or SVD), you will have uncorrelated series (pca: principal component scores; svd: characteristic vectors).  You can then univariately review the pc score distributions using quantile-quantile (q-q) plots or other univariate techniques.  

Depending upon your application area, you may need to transform the variables to achieve something close to Gaussian.  If your data is "counts", try sqrt() if necessary.  If your data is strictly positive dimensional or weight measurements, try log().

Thank you very much.

Sincerly