Inspired by Rishabh Deep Singh

Calculus Level 5

0 100 x 0.5 d x \large \int _{ 0 }^{ 100 }{ \lfloor x-0.5 \rfloor \, d \lfloor x \rfloor }

Evaluate the Riemann-Stieltjes integral above.

Enter 666 if you come to the conclusion that this integral fails to exist.

Notation : \lfloor \cdot \rfloor denotes the floor function .


Inspiration .


The answer is 4950.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

2 solutions

Otto Bretscher
May 22, 2016

I will attempt to do this problem "from first principles", going back to the definition, with the notations used here .

Consider an approximating sum j = 0 n 1 f ( c j ) ( g ( x j + 1 g ( x j ) ) \sum_{j=0}^{n-1}f(c_j)\left(g(x_{j+1}-g(x_j)\right) with the integrand f ( x ) = x 0.5 f(x)=\lfloor x-0.5\rfloor and the integrator g ( x ) = x g(x)=\lfloor x\rfloor , with a mesh size of less than 0.5, meaning that x j + 1 x j < 0.5 x_{j+1}-x_j<0.5 for all j = 0 , . . , n 1 j=0,..,n-1 . Consider a positive integer k k with 1 k 100 1\leq k \leq 100 , arbitrary but fixed. If j j is the largest index such that x j < k x_{j}< k , then c j c_j will be on the interval ( k 0.5 , k + 0.5 ) (k-0.5,k+0.5) , so that f ( c j ) = k 1 f(c_j)=k-1 . Also, g ( x j ) = k 1 g(x_j)=k-1 and g ( x j + 1 ) = k g(x_{j+1})=k .Thus the integer k k makes a contribution of f ( c j ) ( g ( x j + 1 g ( x j ) ) = ( k 1 ) 1 = k 1 f(c_j)\left(g(x_{j+1}-g(x_j)\right)=(k-1)1=k-1 to the approximating sum. Adding up the contributions of all k k , we find that the approximating sum is k = 1 100 ( k 1 ) = 99 × 100 2 = 4950 \sum_{k=1}^{100}(k-1)=\frac{99\times 100}{2}=4950 .

Since all approximating sums with a mesh size of less than 0.5 are 4950, the Riemann-Stieltjes integral exists and has the same value. The answer is 4950 \boxed{4950} .

We can phrase this solution a little more generally as follows: If f ( x ) f(x) is continuous at the integers, then we have the Riemann-Stieltjes integral a b f ( x ) d x = f ( c ) \int_{a}^{b}f(x)d\lfloor x \rfloor =\sum f(c)

where the sum is taken over all integers c c with a < c b a<c\leq b . For example, 0 2 ( x 2 + 1 ) d x = f ( 1 ) + f ( 2 ) = 2 + 5 = 7 \int _{ 0 }^{ 2 }{ ( { x }^{ 2 }+1 ) \, d\lfloor x \rfloor } =f(1)+f(2)=2+5=7

in the inspirational problem.

wow,very good!!I hadn't seen this last...

Bonus: If f ( x ) f(x) is continuous at the integers, what is a b f ( x ) d x ? \int_a^b f(x) \space d\lceil x \rceil? , for example, what is 0 2 ( x 2 + 1 ) d x ? \int_0^2 {(x^2 + 1)} \space d\lceil x \rceil ?

And an award for me... : I almost have a counter-example here ( f = a r c t a n : R ( π 2 , π 2 ) f(x)f”(x) < 0, x R . . . { 0 } f = arc tan : \mathbb{R} \longrightarrow (-\frac{\pi}{2} , \frac{\pi}{2}) \Rightarrow \text{ f(x)f''(x) < 0, } \forall x \in \mathbb{R}... - \{0\} ). For this almost counter-example I got the First Premium ... I failed the question, "how a clever I am"!!... Good, tomorrow I'll have to work again, thanks God, this is the best...

Guillermo Templado - 5 years ago

Log in to reply

It is only a small generalization of my solution to show the following: If the integrator g ( x ) g(x) is a step function such that the integrand f ( x ) f(x) is continuous at all the steps, then the Riemann-Stieltjes integral is the sum of all the products of the value of f ( x ) f(x) at a step times the height of that step.

In your example, we have a b f ( x ) d x = f ( c ) \int_a^b f(x) \space d\lceil x \rceil=\sum f(c) , where the sum is taken over all integers c c with a c < b a\leq c< b .

Yes, that is certainly an impressive almost counterexample! Well done , compañero! You deserve a honorary membership in my "Circle of Comrades".

Otto Bretscher - 5 years ago

Log in to reply

Exactly(+1), Almost what I had in mind...

Guillermo Templado - 5 years ago

Log in to reply

@Guillermo Templado Compañero @Guillermo Templado : I must confess that I got your latest quiz wrong. Just taking a quick look, I would have thought that (c) was true. I look forward to your solution ;)

Otto Bretscher - 5 years ago

Log in to reply

@Otto Bretscher I must confess that I'm not sure about c).I think it's wrong... I'm sure the another two statements are false and I can prove it. If I don't get a proof of c) quickly, I'm probably going to delete it in two or 3 hours. Probably it's wrong.. I'm going to think about it...

Guillermo Templado - 5 years ago

Log in to reply

@Guillermo Templado Here is a rough outline of my thinking: Assume it's not Stieltje integrable over [ a , c ] [a,c] . Then there exists an ϵ 0 > 0 \epsilon_{0}>0 such that for every δ > 0 \delta >0 there exist two approximating sums S 1 S_1 and S 2 S_2 with mesh size < δ <\delta and S 1 S 2 ϵ 0 S_1-S_2\geq \epsilon_{0} . Combining this with an approximating sum S S with mesh size < δ <\delta on [ c , b ] [c,b] , we get two approximating sums S 1 + S S_1+S and S 2 + S S_2+S over [ a , b ] [a,b] that are at least ϵ 0 \epsilon_0 apart.

I must confess that I have heard about this Riemann-Stieltje integral only once in my life, in my introductory analysis class over 40 years ago, so, I feel a bit rusty on the subject. I might well be wrong; it's easy to miss some technicality.

Otto Bretscher - 5 years ago

Log in to reply

@Otto Bretscher I have been trying give you one reply twice , but I can't do it. I don't know why... Ok, give me some time, I have been researching... I think c) is false too. I have been trying prove it to you, but twice the computer has failed... I don't know what have happened... tomorrow, I'll try to post the solution, and we'll be able argue..if the computer lets me...

Guillermo Templado - 5 years ago

Log in to reply

@Guillermo Templado So, do you see a flaw in my reasoning, Compañero?

Otto Bretscher - 5 years ago

Log in to reply

@Otto Bretscher Yes and no, the condition if ϵ > 0 , δ > 0 \forall \epsilon > 0, \space \exists \delta > 0 such that for all partition Π = { x 0 , x 1 , . . . , x n } \Pi = \{x_0, x_1, ..., x_n\} with Δ x i < δ \Delta x_i < \delta and t i [ x i 1 , x i ] t_i \in [x_{i - 1} , x_i] , we have i = 1 n f ( t i ) ( g ( x i ) g ( x i 1 ) ) A < ϵ \displaystyle |\sum_{i = 1}^n f(t_i)(g(x_i) - g(x_{i -1})) - A| < \epsilon is a sufficient condition for f f to be R - S integrable respect to g g , but it's not a necessary condition...

Guillermo Templado - 5 years ago

Log in to reply

@Guillermo Templado What do you mean it is "not necessary." According to my (Soviet) sources, this is exactly the definition if Riemann-Stieltjes integrabilty. What definition are you using, and how does it differ from this one?

Even lowly Wikipedia defines it exactly this way.

Also, take a look at Theorem 6.12c in your trusted Rudin.

Otto Bretscher - 5 years ago

Log in to reply

@Otto Bretscher I'm using upper and lower R-S sums have to match...Tomorrow, I'll try to do a proof with all details,definitions... Ok, think now about this function f ( x ) = 0 f(x) = 0 if x c x \leq c and f ( x ) = 1 f(x) =1 if x > c x > c , and g ( x ) = 0 g(x) = 0 if (x < c) and g ( x ) = 1 g(x) = 1 if x c x \ge c . This function f is R-S integrable respect to g and it don't satisffies my reply above, I mean the necessary condition... and f is not R-S integrable respect to g in [a ,c] and in [c, b]...Ok, tomorrow, please... I don't have more time...

Guillermo Templado - 5 years ago

Log in to reply

@Guillermo Templado Take a look at Rudin's Theorem 6.12c, though, before you try too hard to find an (almost) counterexample ;)

I believe it's just the opposite: Your f f is not R-S integrable on the whole interval but it is R-S integrable on the subintervals.

Otto Bretscher - 5 years ago

Log in to reply

@Otto Bretscher ok, I'm going to do it. Maybe, I'm wrong as usual,I have been seeing Wikipedia and I'm wrong as usual, it's a necessary and sufficient condition... Anyway, tomorrow we'll know, I want a last word, but tomorrow,please...

Guillermo Templado - 5 years ago

@Otto Bretscher Lol. Rudin has become the bible in this arguement!

Log in to reply

@A Former Brilliant Member Yes, I'm wrong. My statement c) is true... Rudin and Otto are right... I'm goign to delete my problem...

Guillermo Templado - 5 years ago

Log in to reply

@Guillermo Templado Not so fast, Compañero @Guillermo Templado ! I believe that, as so often in mathematics, it depends on the definition you are using. Most people these days define the Riemann-Stieltjes integral in terms of the mesh size going to 0, as here , but there is an older definition out there according to which your statement may be false. I'll do some more research.

See the remarks on Pages 1 and 2 here . According to this more relaxed definition, this guy would be Riemann-Stieltjes integrable, for example, since there exists a partition (including 1 as a partition point) where all the approximating sums are the same.

Otto Bretscher - 5 years ago

Log in to reply

@Otto Bretscher I have got crazy searching a counterexample, but I I haven't got it, so I have decided to research propositions and I consider Rudin is right and its recyprocal is true too... Anyway, I'll continue researching...

Guillermo Templado - 5 years ago

@Otto Bretscher haha, it's curious that I believe the reciprocal of 6.12c) of Rudin, but I dont' believe 6.12c) of Rudin... How much do you bet I'm wrong?I would bet quite money,haha.. ok, tomorrow we'll see it...

Guillermo Templado - 5 years ago

Log in to reply

@Guillermo Templado Among Comrades, we usually bet a (1970 Lenin-) Ruble ;)

I'm with Rudin on this one! I believe the example you give in your post "I'm using upper and lower R-S sums have to match..." is Stieltjes integrable on the subintervals, but not Stieltjes integrable on the whole interval since both functions are then discontinuous at c c .

I'm waiting for your post that will convince us all; take your time!

Otto Bretscher - 5 years ago

Log in to reply

@Otto Bretscher I'm believing you and Rudin are right, I'm going to post a note in my problem for nobody answers it... Firstly, the condition above of mine is necessary and sufficient, so here, I'm wrong... Nevertheless, I have some doubts about my c) statement and I'm not going to delete it yet, because I still have some doubts... Tomorrow, if I don't see it clearly, it will be deleted...

Guillermo Templado - 5 years ago

Log in to reply

@Guillermo Templado I'm curious to see what conclusions you will reach. I cannot see a flaw in my outline of a proof that (c) is true.

Thinking about these Riemann-Stieltjes integrals is a a bit like walking in a labyrinth with funny mirrors: We need to forget (almost) all we learned about Riemann integrals all these decades and re-orient ourselves "from first principles."

Otto Bretscher - 5 years ago

Hey, this (and a slight) generalisation was given in this book by Walter Rudin!

Good to see that such problems are appreciated!

Log in to reply

@A Former Brilliant Member hm, Rudin, I believe that he is/was a very good teacher... I had a teacher in complex analysis who always quoted Rudin's book: Real and Complex Analysis

Guillermo Templado - 5 years ago

@A Former Brilliant Member Yes, indeed, these are well-known results of basic analysis (I learned them, with slightly different terminology, from old Soviet texts). It is fun to illustrate this theory with examples.

Otto Bretscher - 5 years ago

Yes, it makes sense. It uses simply the definition... and for this reason it's correct, because upper and lower sums and these sums match...

Guillermo Templado - 5 years ago

Log in to reply

Thanks for the feedback, compañero! We don't even have to mention upper and lower sums if we don't want to; the definition involves all approximating sums.

Otto Bretscher - 5 years ago

Log in to reply

Yes, that is compañero... As usual, a pleasure, thanks for you. I always learn something with you... See you soon...

Guillermo Templado - 5 years ago

Log in to reply

@Guillermo Templado The pleasure is mine. Your solution is interesting; I did not consider the possibility of breaking it up this way.

Otto Bretscher - 5 years ago

Wow Atleast someone is Inspired From Me. @Otto Bretscher

Rishabh Deep Singh - 4 years, 9 months ago

I'm going to use 3 propositions without proofs:

1.- Let's suppose that f f is a Riemann-Stieljes integrable function in [ a , c ] [a , c] and in [ c , b ] [c , b] ( a < c < b ) respect to function g g , then f f is a Riemann-Stieljes integrable function respect to g g in [ a , b ] [a , b] and a b f d g = a c f d g + c b f d g \int_a^b f \space dg = \int_a^c f \space dg + \int_c^b f \space dg .

This first proposition is only true if we are go to use (1) the generalizaded R-S integral definition here . It's not true if we are going to use the definition here , so the proof for this question will only be rigth if we are using (1) the generalized R-S integral . In other case is wrong...

For one proof of this proposition, take one partition P in [a,b] being the union of partitions P' in [a,c] and P'' in [c,b] given for ϵ 2 > 0 \frac{\epsilon}{2} > 0 and use the triangular inequality for the absolute value....Briefly and gruffly, A + B S ( P , f , g ) A S ( P , f , g ) + B S ( P , f , g ) ϵ 2 + ϵ 2 = ϵ |A + B - S(P,f,g)| \leq |A - S(P',f,g)| + |B - S(P'',f,g)| \leq \frac{\epsilon}{2} + \frac{\epsilon}{2} = \epsilon ...

2.- Let f f to be a real bounded function in [ a , b ] [a , b] right-continuous in a a and left-continuous in b b , and g ( x ) = k g(x) = k a constant function in ( a , b ) (a , b) such that g ( a ) k g ( b ) g(a) \leq k \leq g(b) with k k a real constant, then f f is a Riemann-Stieljes integrable function respect to g g and a b f d g = f ( a ) ( k g ( a ) ) + f ( b ) ( g ( b ) k ) \int_a^b f \space dg = f(a) \cdot ( k - g(a)) + f(b) \cdot (g(b) - k)

3.- If f f is a real bounded function and a Z a \in \mathbb{Z} then a a + 0.5 f d x = 0 \int_{a}^{a + 0.5} f \space d\lfloor x \rfloor = 0 . Proof: use upper and lower Riemann-Stieljes sums...

Solution:

0 100 x 0.5 d x = \int_0^{100} \lfloor x - 0.5 \rfloor \space d\lfloor x \rfloor = = 0 0.5 x 0.5 d x + 0.5 1 x 0.5 d x + 1 1.5 x 0.5 d x + . . . + 99 99.5 x 0.5 d x + 99.5 100 x 0.5 d x = = \int_0^{0.5} \lfloor x - 0.5 \rfloor \space d\lfloor x \rfloor + \int_{0.5}^{1} \lfloor x - 0.5 \rfloor \space d\lfloor x \rfloor + \int_1^{1.5} \lfloor x - 0.5 \rfloor \space d\lfloor x \rfloor + ... + \int_{99}^{99.5} \lfloor x - 0.5 \rfloor \space d\lfloor x \rfloor + \int_{99.5}^{100} \lfloor x - 0.5 \rfloor \space d\lfloor x \rfloor = 1.5 2 x 0.5 d x + 2.5 3 x 0.5 d x + . . . + 99.5 100 x 0.5 d x = \int_{1.5}^{2} \lfloor x - 0.5 \rfloor \space d\lfloor x \rfloor + \int_{2.5}^{3} \lfloor x - 0.5 \rfloor \space d\lfloor x \rfloor + ... + \int_{99.5}^{100} \lfloor x - 0.5 \rfloor \space d\lfloor x \rfloor = = 1 + 2 + 3 + . . . + 99 = ( 99 + 1 ) 99 2 = 4950 = 1 + 2 + 3 + ... + 99 = \frac{(99 + 1)99}{2} = 4950

How you evaluate your last Integrals? I didn't understand anything

Ciara Sean - 5 years ago

Log in to reply

I have edited the proof, if you don't understand something, tell me please. Firstly, you have to study elemental properties of Riemann-Stieljes integral and its sums... and later, think a little bit about the 3 propositions and the solution. I repeat, sorry, if you don't understand something tell me, please. For last integrals I use proposition 2.

Guillermo Templado - 5 years ago

Yes, compañero, very nicely done! Thanks! (+1)

In your second "proposition without proof", you want to state that k k is the value of g g on ( a , b ) (a,b) .

I used the definition to do this integral; I will write a solution when I get around to it.

I had not done a Riemann-Stieltjes integral since college, 40 years ago; the inspirational problem by @Rishabh Deep Singh was a nice refresher.

Otto Bretscher - 5 years ago

Log in to reply

Yes, g ( x ) = k g(x) = k in ( a , b ) (a , b) for my 2ª "proposition without proof". I can prove the propositions if you wish it... And notice, for example lim x 2. 5 + x 0.5 = 2 = lim x 3 x 0.5 \displaystyle \lim_{x \to 2.5^+} \lfloor x - 0.5 \rfloor = 2 = \displaystyle \lim_{x \to 3^-} \lfloor x - 0.5 \rfloor

Guillermo Templado - 5 years ago

Log in to reply

You need to state in your solution that g ( x ) = k g(x)=k on ( a , b ) (a,b) .

I understand your solution well; it's written up nicely. No further explanation is needed for me.

Please take a look at my solution "from first principles" and let me know whether it makes sense too.

Otto Bretscher - 5 years ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...