Solution
Solution 1
Note that 1+∑k=1n−1k⋅k!=1+∑k=1n−1((k+1)⋅k!−k!)=1+∑k=1n−1((k+1)!−k!)
=1+((2!−1!)+(3!−2!)+⋯+(n!−(n−1)!))=n!.
Thus for all m∈N,
(32m+16)!−(32m)!=(1+∑k=132m+15k⋅k!)−(1+∑k=132m−1k⋅k!)=∑k=32m32m+15k⋅k!.
So now,
16!−32!+48!−64!+⋯+1968!−1984!+2000!=16!+(48!−32!)+(80!−64!)⋯+(2000!−1984!)=16!+m=1∑62((32m+16)!−(32m)!)=16!+m=1∑62k=32m∑32m+15k⋅k!
Therefore we have f16=1, fk=k if 32m≤k≤32m+15 for some m=1,2,…,62, and fk=0 for all other k.
Therefore we have:
f1−f2+f3−f4+⋯+(−1)j+1fj=(−1)17⋅1+m=1∑62k=32m∑32m+15(−1)k+1k=−1+m=1∑62[j=16m∑16m+7(−1)2j+12j+j=16m∑16m+7(−1)2j+2(2j+1)]=−1+m=1∑62j=16m∑16m+7[(−1)2j+12j+(−1)2j+2(2j+1)]=−1+m=1∑62j=16m∑16m+7[−2j+(2j+1)]=−1+m=1∑62j=16m∑16m+71=−1+m=1∑628=−1+8⋅62=495.
Solution 2 (informal)
This is equivalent to Solution 1. I put up this solution merely for learners to see the intuition.
Let us consider a base n number system. It’s a well known fact that when we take the difference of two integral powers of n, (such as 1000010−10010) the result will be an integer in base n composed only of the digits n−1 and 0 (in this example, 9900). More specifically, the difference (nk)n−(nj)n, j,isanintegerkdigitslong(notethat(n^k)_nhask + 1digits).Thisintegerismadeupof(k-j)(n−1)’s followed byj0$’s.
It should make sense that this fact carries over to the factorial base, albeit with a modification. Whereas in the general base n, the largest digit value is n−1, in the factorial base, the largest digit value is the argument of the factorial in that place. (for example, 321! is a valid factorial base number, as is 3210!. However, 31! is not, as 3 is greater than the argument of the second place factorial, 2. 31! should be represented as 101!, and is 710.) Therefore, for example, 1000000!−10000! is not 990000!, but rather is 650000!. Thus, we may add or subtract factorials quite easily by converting each factorial to its factorial base expression, with a 1 in the argument of the factorial’s place and 0’s everywhere else, and then using a standard carry/borrow system accounting for the place value.
With general intuition about the factorial base system out of the way, we may tackle the problem. We use the associative property of addition to regroup the terms as follows: (2000!−1984!)+(1968!−1952!)+⋯+(48!−32!)+16! we now apply our intuition from paragraph 2. 2000!10 is equivalent to 1 followed by 1999 0’s in the factorial base, and 1984! is 1 followed by 1983 0’s, and so on. Therefore, 2000!−1984!=(1999)(1998)(1997)⋯(1984) followed by 1983 0’s in the factorial base. 1968!−1952!=(1967)(1966)⋯(1952) followed by 1951 0’s, and so on for the rest of the terms, except 16!, which will merely have a 1 in the 16! place followed by 0’s. To add these numbers, no carrying will be necessary, because there is only one non-zero value for each place value in the sum. Therefore, the factorial base place value fk is k for all 32m≤k≤32m+15 if 1≤m∈Z≤62, f16=1, and fk=0 for all other k.
Therefore, to answer, we notice that 1999−1998=1997−1996=1, and this will continue. Therefore, f1999−f1998+⋯−f1984=8. We have 62 sets that sum like this, and each contains 8 pairs of elements that sum to 1, so our answer is almost 8⋅62. However, we must subtract the 1 in the f16 place, and our answer is 8⋅62−1=495.
Solution 3 (less formality)
Let S=16!−32!+⋯−1984!+2000!. Note that since ∣S−2000!∣<<2000! (or ∣S−2000!∣=1984!+⋯ is significantly smaller than 2000!), it follows that 1999!<S<2000!. Hence f2000=0. Then 2000!=2000⋅1999!=1999⋅1999!+1999!, and as S−2000!<<1999!, it follows that 1999⋅1999!<S<2000⋅1999!. Hence f1999=1999, and we now need to find the factorial base expansion of
S2=S−1999⋅1999!=1999!−1984!+1962!−1946!+⋯+16!
Since ∣S2−1999!∣<<1999!, we can repeat the above argument recursively to yield f1998=1998, and so forth down to f1985=1985. Now S16=1985!−1984!+1962!+⋯=1984⋅1984!+1962!+⋯, so f1984=1984.
The remaining sum is now just 1962!−1946!+⋯+16!. We can repeatedly apply the argument from the previous two paragraphs to find that f16=1, and fk=k if 32m≤k≤32m+15 for some m=1,2,…,62, and fk=0 for all other k.
Now for each m, we have −f32m+f32m+1−f32m+2+⋯+f32m+31 =−32m+(32m+1)−(32m+2)+⋯−(32m−30)+(32m+31) =1+1+⋯+1+1 =8. Thus, our answer is −f16+8⋅62=495.
Solution 4 (my brain can't comprehend the other solutions)
First consider how we would express 48!−32!. We can't use any multiples of 48!, since we'll never be able to make the −32!. Instead, we have to start with using 47!s. Having 47⋅47! is the closest we can get to 48!, since 48!=48⋅47!. Thus, f47=47.
Now consider what we have left to express. We wanted to express 48!−32!, and now we have 47⋅47!. This gives us 47!−32! left. Clearly, we can't use anymore 47!s, so we have to use 46!s. The most 46!s we could use is 46 of them. Thus, f46=46. Since 47!=47⋅46!, we are left with 46!−32! to express. Clearly, each time, we have fn=n, leaving us with n!−32! to express.
Eventually, we will get down to needing to express 34!−32!. We will need 33 33!s, leaving 33!−32! to be expressed. Since 33!=33⋅32!, 33!−32!=32⋅32!. Thus, 48!−32! can be expressed as ∑i=3247i⋅(i!).
Returning to the problem, notice we can break the expression into groups of 2, having 16!+(48!−32!)+(80!−64!)+...(2000!−1984!). Each pair has an alternating sum of 8, and since there are 62 pairs, the sum is 496. Including the 16! at the start, which accounts for −1, we get 495.
-skibbysiggy
Solution 5 (Much simpler than the other solutions [besides 4])
Let's solve 3!−5! without numerically solving each. What do we know about number theory? Well, one way to rewrite a subtraction a from b given a>b is to borrow from a, add it to b, and subtract a as a result. More formally, if b−a:a>b→(b+a)−2a.
How does this apply to factorials? Well, say we borrow from the next highest factorial, that is, if we have b!−a! where a>b, then we can borrow once from (b+1)! to obtain b+1 copies of b! (which is exactly (b+1)! if you are wondering), and we subtract (b+1)! as a result. So, for our case 3!−5!, we can rewrite as 3!−(5)(4)3!, or 3!−20⋅3!, or −19⋅3!. We have the constraint however that K=f1(1!)+f2(2!)+...+fn(n!) where 0≤fn≤n. However, our case swaps the sign of K, giving us 0≥−fn≥−n. So, we borrow once from 4!, giving us 4 copies of 3!, and ultimately resulting in (−19+4)3!−1⋅4!. The result −19+4=−15 is still negative. To compensate, we take 1 copy of 5! to donate to 4!, which then we can donate the new copies of 4! to 3!. So, giving one copy of 5! to 4! gives us 5 copies of 4!, so we have (−15)3!+(−1+5)4!=(−15)3!+4(4!), in which 4⋅4! has 16 copies of 3! (as 4⋅3!=4!). Then, we have (−15+16)3!+4⋅4!−1⋅5!, with a final answer of 3!+4⋅4!−1⋅5!, where (f3,f4,f5)=(1,4,−1) before normalization (note that -1 is allowed for f5 as the expression itself is negative. For all positive K, the inequality must hold). This gives us an idea of what to do for the problem.
Consider n!−(n+16)!. We must borrow 1 copy from (n+1)!, which then results in borrowing 2 copies from (n+2)!, which then results in borrowing 3 copies from (n+3)!, and so on so forth until we borrow 15 copies from (n+15)!, and 16 copies from (n+16)!. Additionally, the number of copies we borrow from each factorial correspond to the fn power (We already proved that fn is determined by the explicit values of the remaining factorial expansion. However, as homework, try to see how each fn corresponds to the number of copies borrowed). Thus, we have f1−f2+f3,...,+f15−f16=1−2+3−4+...+15−16=−8. This means that each n!−(n+16)! pair contributes -8 to the count. So, it becomes clear that every (n+16)!−n! pair contributes 8 to the count. Thus, we can pair each term as 16!+(48!−32!)+(...)+(2000!−1984!). There are 62 of the (n+16)!−n! pairs, and thus we see that a total alternating sum of 496 arises. However, we have one last case, that is, 16!. We borrow 0 copies of n!, because 1 copy of n! is already present! Thus, the remaining (−1)n(fn) is nothing but (−1)125(1)=−1 (there are a total of 125 terms, and 16! is the 125th), and so our final sum is 496−1=495.
~Pinotation