I recently became engaged in the work of Siegel, Schmidt, Rogers, Macbeath regarding random lattices and geometry of numbers, e.g. Siegel proved that $$\int_{SL(n,\mathbb{R})/SL(n,\mathbb{Z})} \sum_{ 0 \neq u \in \Lambda} f(v) d \mu(\Lambda)= \int_{\mathbb{R}^n}f(x) dx$$ where $\mu$ is the normed Haar measure on $SL(n, \mathbb{R})$.

However, I believe that I've found some inconsistencies in the work of Rogers and Macbeath. I would be very grateful if someone who is familiar with this field could confirm these, or point out that I'm wrong.

  • In [1] I strongly suspect theorem 1, pg. 147 to be wrong. The theorem is a formula for the integral containing a sum over $l$ linear independent primitive points. In the case $l=2$ and after normalising the Haar measure in the theorem states $$\int_F \sum_{...} f(u,v)d \mu(\Lambda) = \frac{1}{\zeta(n)\zeta(n-1)} \int_{\mathbb{R}^{n^2}} f(x,y) d(x,y)$$ where the sum on the left ranges over all primitive points $u,v \in \Lambda$ with $\dim(u,v)=2$. $F$ is a fundamental domain for $SL(n,\mathbb{R})/SL(n,\mathbb{Z})$ here. However, this contradicts other results, e.g. by Rogers himself where the factor on the right hand side should be $\frac{1}{\zeta(n)^2}$ instead.
  • Also, in my opinion in [1] lemma 9, pg.150, should be wrong as well, because the desired representation should not be unique.
  • In [2], pg. 252, Rogers emphasises that theorem 5 works for $n \geq 2$. Although, in the proof of this theorem Rogers uses theorem 3 of the same paper which should not apply in the case $n=2$. This really surprised me, because Rogers explicitly highlights that the theorem 5 does apply for $n=2$ in the text above theorem 5.

[1] Siegel's mean value theorem in the geometry of numbers
A. M. Macbeath, C. A. Rogers http://www.ams.org/mathscinet-getitem?mr=0103183

[2] Mean values over the space of lattices C. A. Rogers http://www.ams.org/mathscinet-getitem?mr=0075243

  • $\begingroup$ Where did these papers appear? $\endgroup$ – paul garrett Feb 8 '14 at 23:14
  • $\begingroup$ Unfortunately, I couldn't find a free access via google(I looked beforehand and would had link both). If you have access to a mathscinet subscription you can find them here: ams.org/mathscinet-getitem?mr=0103183 and ams.org/mathscinet-getitem?mr=0075243 $\endgroup$ – Soeren Feb 8 '14 at 23:28
  • $\begingroup$ Ah, thanks. Even if something's behind a paywall, just telling what journal (if any) it appeared in helps specify it... $\endgroup$ – paul garrett Feb 8 '14 at 23:30
  • $\begingroup$ Yes, you are absolutely right! I added both links to my original post. $\endgroup$ – Soeren Feb 9 '14 at 2:09
  • $\begingroup$ In Theorem 1 of paper [1], Rogers makes a comment about some sets being measure zero in $SL(n,\mathbf{R})/SL(n, \mathbf{Z})$. I tried but couldn't prove it even for $n \ge 3$. Does anybody know how to do it? $\endgroup$ – Mahbub Alam Jan 23 '18 at 11:07

In this field it is a well known fact that Rogers had an error in his paper for the $n=2$ case. It was corrected by Schmidt in his paper A Metrical Theorem in the Geometry of numbers


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.