# When exactly does error tend to zero in calculus?

I've come across many instances where sometimes the error tends to zero but other times it does not. Let me give you a few examples.

1. When I calculate the area of a sphere summing up discs of height $dy$ from -R to +R, the error in volume tends to zero as $\Delta y->0$ but when I'm calculating the surface area using rings of height $dy$, this error does not tend to zero. Same with a hollow cone vs a solid cone.

2. $ds$ is the arc length. We know $ds/dt$ = speed = $|dr/dt|$ since the error tends to zero.

However, in another case, particularly from Irodov's problems:

3. I'm trying to find the work done by the spring as the block moves from one end to the other(Yes' I'm aware there are easier ways of going about it)

Here, where I think I'm going wrong is assuming the spring force to be constant in the interval $dy$ when it can only remain constant during an infinitesimal displacement along the spring.

However, here we assume pressure to be constant in the interval $Rd\theta$ when it really only is constant in the interval $dH$ since it's a function of $h$ And here, we assume the potential energy of the chain to be constant in the interval $Rd\theta$ when it should be only constant in $dh$ yet in these two cases the error tends to zero but it doesn't in the first case.

• In your first one the path it travels on is the arc, so ds will obviously be the same portion of the arc that dr is. As JMLCarter mentioned in his answer; it isn't error. It's usually a physical significance. – JMac Feb 14 '17 at 11:38
• -1. Unclear. There is too much confusing content in your question, and too much variety in styles. It would be far better if you focussed on one problem - the first is probably good enough. – sammy gerbil Feb 14 '17 at 18:01
• Your first example is addressed in the Math SE questions Why does the same limit work in one case but fail in another? and Why is arc length not in the formula for the volume of a solid of revolution? – sammy gerbil Feb 14 '17 at 18:29
• @sammygerbil I've given all these examples so that the answerer has a more clear idea about where I'm going wrong. – xasthor Feb 15 '17 at 1:48
• I think the question is good. Preferably, write down whatever's in the picture using latex for the sake of clarity. – User2956 Feb 17 '17 at 12:10

I think there is no general rule for deciding when the error tends to zero. There are an infinite variety of cases - as you illustrate. You must examine the appropriateness of the approximation being made in each case.

Take your first example, which I will simplify to the 2D problems of finding the area and perimeter of a circle.

Approximate the area of the circle with horizontal strips which do not extend beyond the perimeter. As the strips are made narrower the total area of the strips increases hence the error decreases. In the limit of infinitesimally thin strips we get the exact value of the area.

Approximate the perimeter of the circle using horizontal and vertical straight lines, as done in the paradox in the Math SE question Is $\pi = 4$?

Start by approximating the circle with a square of perimeter 4. Then make indents by removing corners - the perimeter is still 4. As more and more corners are removed, the resulting staircase curve approximates more closely to the circle, yet the perimeter remains the same. The error never gets any smaller.

The correct way of doing this approximation is to use the hypotenuse of the triangles as the element of perimeter. While the length of the staircase curve never gets any smaller, the total length of the hypotenuses decreases towards the circumference of the circle.

In the Math SE answers it is pointed out that the staircase curve does not approach the circle smoothly. Smoothness is related to derivatives, so another way of putting this is that while two curves can be arbitrarily close at some point, their derivatives at that point can be significantly different.

The key to ensuring that the approximation gives the correct result when you integrate, is to examine the smoothness of the approximation. You can see that the staircase curve never gets any smoother - it is always jagged if you look at it on a small enough scale. So it can never provide an accurate approximation to the circle.

In practical terms examining smoothness of an approximation is doing the same as what you are asking a shortcut for - ie checking that the error tends to zero. I think there is no shortcut, no easy alternative to checking in each case that the approximation does in fact get closer to what it is approximating as the step size decreases. In many cases you will know from experience that it works. Otherwise you need to convince yourself that the error does decrease towards zero as the step size is reduced.

1. While $ds=r\delta \theta$ works when the centre is the origin O (and $s=r\theta$ is in fact exact even for finite angles), it does not work when a point on the circumference is the origin. When OA is a diameter the relation $ds=r\delta\theta$ is good as $\delta\theta \to 0$, but as A moves closer to O then OA shrinks to zero and the approximation gets increasingly bad - as your hand-drawn diagram beings to show. Your guess ("Probably mistake : $r\delta\theta \ne ds$") is correct.
2. (i) Spring and block. You have expressed the spring force in terms of $y$, so your mistake is not failing to do so. Your mistake is failing to relate $dx, dy$ correctly : $dy=\sin\theta dx$. Then work done is $kxdx=kxdy/\sin\theta$. A diagram linking $dx$ and $dy$ will help you avoid this error.
(ii) PE of Chain. The height of the CM is $h=R(1-\cos\theta)$ so $dh=R\sin\theta d\theta$ not $R d\theta$. Your mistake is again that you have not drawn a diagram to relate $dh$ to $Rd\theta$.
(iii) Pressure in bowl. Not enough explanation of what you are doing. You seem to have ignored $l=2\pi R\cos\theta$.
• In the spring and block system, I corrected x to $x = l/cos\theta -l$ which gave me the correct answer. For the chain, what I was trying to ask is why I can treat the potential energy to be constant throughout the element of $Rd\theta$ when potential energy which is a function of $h$ should only be constant in the interval $dh$, and not $Rd\theta$ I think my question wasn't clear, but what I was trying to ask is why I can treat a force $F(x)$ which I thought is only constant in the interval $[x,x+dx)$, to be constant in unrelated infinitesimal intervals such as $dy, Rd\theta$ etc. – xasthor Feb 20 '17 at 1:29