problems arising from WLS

[econ11]

hi, i am currently tying to apply WLS method to solve heteroskedasticity in my data.

From the original model I got an output of Breusch Pagan LM Test below,
"Obs*R-squared 27.42129 Prob. Chi-Square(13) 0.0109"
and of White test
"Obs*R-squared 144.7361 Prob. Chi-Square(102) 0.0035"
which clearly show that i should reject the null hypothesis which says there is no heteroskedasticity.
btw, there is no serial correlation (this is cross-section data). The output of Breusch Godfrey LM Test is
"Obs*R-squared 1.206746 Prob. Chi-Square(2) 0.5470" (2lags included)
which means i cannot reject the null hypothesis saying there is no autocorrelation.
(i tried including 1lag, 2lags,and 5lags and all trials led to the same conclusion)

So i applied WLS method, using 1/sqr(Student) as the weight series. (variable "Student" was recommended by my professor)
As expected, heteroskedasticity was solved. The output of BP LM Test, and White test of the new model(WLS applied)
is "Obs*R-squared 20.19753 Prob. Chi-Square(13) 0.0904"
" Obs*R-squared 103.7419 Prob. Chi-Square(102) 0.4333"
so i concluded that the problem of heteroskedasticity has been solved.

However, a new problem of autocorrelation arised from this model.
the output of Serial Correlation test(Breusch Godfrey LM Test) is
"Obs*R-squared 98.37728 Prob. Chi-Square(2) 0.0000"

i don't know why serial correlation has occurred when i applied WLS method. Intuitively, this is a cross-section data so there should be no
autocorrelation. Can you please explain why this happened?