akaike criterion: ar() vs arima() in R

akaike criterion: ar() vs arima() in R

I've ve got drastic difference between AIC for ar(1) and arima(1,0,0):
a<-ar(rn, lags=1)
a$aic 0 1 2 3 4 5 6 7 ...
6.0215169 1.2184962 2.0020937 1.1786418 0.9002231 0.0000000 1.1728207 ...
b<-arima(rn, order=c(1,0,0)) b$aic [1] -6840.676
The regression coeff are rather close though: -0.068 for ar and -0.077 for
arima. Will greatly appreciate any comments. Alec