The original problem is on Binomial distribution with n = 100 trials and  p = 1/3 for individual 
success  (p= 1/3 to answer correctly and q = 1-p = 2/3 to answer incorrectly).


So, formally there are standard formulas for binomial distribution, but the Math does not 
recommend to use it, because the formulas include high degree exponents and too great binomial
coefficients.  Instead, there is another, very effective and standard approach on solving such problems.


This approach is to use an appropriate normal distribution as an approximation to the Binomial distribution.


The relevant normal distribution has the mean  m = n*p =  = 33.33333
and the standard deviation  S =  =  = 4.714.



This normal curve covers all 100 questions on the horizontal axis, and we should take the area under
this normal curve on the right of the threshold value 40.


A convenient tool to find the area of interest is to use the standard function  normcdf 
(which means Cumulative Distrifution Functionfor normal distribution).  
This function is in any regular calculator like TI-83/84.


So, you write/print in your calculator

                 z1     z2      mean     SD     <<<---===  formatting pattern.
    P = normcdf(39.5,  9999,  33.33333, 4.714)


Here z1 = 39.5  stands for 40 and presents the continuing correction;  z2 stand for infinity on the horizontal axis,
mean is the mean and SD is the standard deviation.


Then the function will return the probability, which you are looking for: P = 0.0954.


It is the ANSWER to the problem's question  P = 0.0954, obtained with the normal distribution approximation.