The naive approach:
int originalInput = 42;
int input = originalInput;
// Generate binary numbers
var binaryNumbers = Enumerable.Range(0, 31).Select(n => (int)Math.Pow(2, n)).ToArray();
// Largest first
Array.Reverse(binaryNumbers);
var result = new List<int>();
foreach (var bin in binaryNumbers)
{
if (input >= bin)
{
result.Add(bin);
input -= bin;
}
}
Console.WriteLine($"{originalInput} decomposed: " + string.Join(" ", result));
Generate a range of power-of-two numbers, ranging from 2^31 (1073741824) to 2^0 (1), then check whether the input is equal to or larger than those numbers, and if so, add that number to the result list and subtract it from the input.
Now that that's all written out, see how Sergey's answer greatly reduces the code required by some Linq and bitshifting magic.
A hybrid solution, inspired by combining both answers:
var input = 42;
var output = Enumerable.Range(0, 31)
.Select(n => (int)Math.Pow(2, n))
.Where(p => (p & input) > 0);
Console.WriteLine($"{input} decomposed: " + string.Join(" ", output));