Sharedwww / tables / Notes / padictwist.texOpen in CoCalc
Author: William A. Stein
1
% padictwist.tex
2
\documentclass{article}
3
\include{macros}
4
5
\title{Approximating twists by finite slope newforms}
6
\author{Robert Coleman\footnote{Coleman hasn't read this version yet,
7
so any mistakes are Stein's.} \and William A.~Stein}
8
\begin{document}
9
\maketitle
10
11
\section{Introduction}
12
We have experimentally investigated the question of approximation
13
of twists of finite slope forms.
14
15
Let~$p$ be a prime and let~$N$ be a positive
16
integer not divisible by~$p$.
17
Consider a classical newform~$f$ in
18
$S_k(\Gamma_0(N),\Zp) := S_k(\Gamma_0(N),\Z)\tensor\Zp.$
19
The \defn{slope} of~$f$ is $\ord_p(a_p(f))$.
20
By~\cite[Prop.~3.6]{shimura:intro}, the twist of~$f$ by a Dirichlet character
21
$\eps:(\Z/p\Z)^*\ra \C^*$
22
is an eigenform on $\Gamma_1(Np^{2r})$; the slope of this
23
twist is infinite.
24
Assume that~$f$ has finite slope.
25
Is it possible to approximate $f^{\eps}$ by forms of finite
26
slope? One way to make ``approximate'' precise and falsifiable,
27
is to ask the following: let $n\geq 1$ be an integer;
28
does there exist a newform
29
$g \in S_{k + \vphi(p^n)}(\Gamma_0(Np);\Z_p)$
30
such that $g \con f^{\eps}\pmod{p^n}$?
31
(By this congruence, we mean that
32
$\ord_p(g -f^\eps) = \min_{i\geq 1}\{\ord_p(a_i(g)-a_i(f^\eps))\} \geq n.$)
33
34
The question of approximation of twists by finite slope forms
35
is motivated by our attempt to complete the
36
eigencurve~\cite{eigencurve}, and by questions
37
of N.~Jochnowitz.
38
If the above question has an affirmative
39
answer for $r=1$, then we obtain a sequence $\{g_n\}$ of
40
finite slope classical overconvergent forms that converges
41
to the infinite slope form $f^{\eps}$.
42
An unusual feature of the \nobreak{computations} described in this
43
paper is that we approximate~$f$ by {\em newforms} on $\Gamma_0(Np)$, so the
44
slopes increase quickly.
45
46
\section{A false conjecture}
47
In this section we state a very general, {\em and very false}, conjecture.
48
Let~$p$ be a prime and let~$N$ be a positive
49
integer not divisible by~$p$.
50
\begin{conjecture}[False conjecture]\label{conj:false}
51
Let $f \in S_k(\Gamma_0(N),\Zp)$ be a finite slope newform,
52
and~$\eps$ a character mod~$p$ such that $\eps^2=1$.
53
Then for all $n\geq 1$ there
54
exists an eigenform
55
$g \in S_{k + \vphi(p^n)}(\Gamma_0(Np);\Z_p)$
56
such that $g \con f^{\eps}\pmod{p^n}$.
57
\end{conjecture}
58
59
60
\section{Numerical experiments}
61
\label{sec:plan}
62
In the remainder of this paper we report on numerical experiments
63
designed to gather evidence about Conjecture~\ref{conj:false}.
64
These computations were carried out in \magma{} (cf.~\cite{magma})
65
using algorithms described in~\cite{stein:phd}.
66
67
We perform the following experiments:
68
\begin{dashlist}
69
\item Section~\ref{exp:delta}.
70
Let $f=\Delta$ be the unique cusp form of weight~$12$.
71
For each $p=2,3,5,7,11$ consider the twist $f^{\eps}$ by either the
72
trivial or quadratic mod~$p$ character.
73
We compute $\ord_p(g- f^{\eps})$ for each rational
74
newform $g\in S_{k+\vphi(p^n)}(\Gamma_0(Np);\Z_p)$,
75
for the first few values of~$n$.
76
In fact, $\ord_p(g-f^{\eps})$ was computed using only~$17$ terms of the
77
$q$-expansion, which is probably enough, but we have {\em not}
78
proved this.
79
\item Section~\ref{exp:11}.
80
Let~$f$ be the weight-$2$ newform
81
corresponding to the elliptic curve $X_0(11)$.
82
For $p=2,3,5$ we perform the same computation as for~$\Delta$.
83
\end{dashlist}
84
\vspace{1ex}
85
86
87
88
Since, in this experiment, we only twist $f$ by the trivial
89
or quadratic character, we adopt the notation $f^{(1)}$ and
90
$f^{(-1)}$ for these twists, respectively.
91
Thus $f^{(1)}$ is the $p$-deprivation
92
of~$f$, and $f^{(-1)}$ is the quadratic twist.
93
94
\begin{remark}
95
The newforms at level~$Np$ all have slope $(k-2)/2$
96
(cf.~\cite[\S6]{diamond-im}).
97
\end{remark}
98
99
\begin{remark}
100
Some of the tables are missing entries; these are being computed...
101
\end{remark}
102
103
\subsection{Twisting~$\Delta$}\label{exp:delta}
104
Consider the weight-12 cuspform $\Delta$; the $q$-expansion of $\Delta$ is
105
\begin{eqnarray*}
106
\Delta&=&
107
\prod(1-q^n)^{24} \\
108
&=&q - 24q^2 + 252q^3 - 1472q^4 + 4830q^5 - 6048q^6 - 16744q^7 +\cdots.
109
\end{eqnarray*}
110
Let~$p$ be a prime and let $\eps:(\Z/p\Z)^*\ra\{\pm 1\}$ be a character;
111
then $\eps^2=1$, so $\Delta^{\eps}\in S_{12}(\Gamma_0(9))$.
112
As outlined in Section~\ref{sec:plan}, we compute each of the
113
rational newforms $g \in S_{12+\vphi(p^n)}(\Gamma_0(p),\Z_p),$
114
and then $\ord_p(\Delta^{\eps}-g)$.
115
116
The computations are given in the following tables.
117
For example, the 4th line of the first table below
118
says that when $n=4$ there are two
119
$\Z_2$-rational newforms
120
$g_1$ and $g_2$; and that $\Delta^{(1)} \con g_1 \pmod{2^5}$
121
and $\Delta^{(1)} \con g_2 \pmod{2^5}$.
122
The non-$\Z_p$-rational newforms in $S_{12+\vphi(p^n)}(\Gamma_0(p),\Zbar_p)$
123
are not listed.
124
125
\begin{center}
126
\vspace{1ex}
127
Eigenforms on $\Gamma_0(2)$ congruent mod $2$ to the
128
$2$-deprivation of $\Delta$.\vspace{1ex}\\
129
\begin{tabular}{l|l}
130
$n\quad$ & $\quad\ord_2(\Delta^{(1)}-g_i)$\\\hline
131
$1$ & none\\
132
$2$ & 3, 3\\
133
$3$ & 4\\
134
$4$ & 5, 5\\
135
$5$ & 6, 6\\
136
$6$ & 7, 7, 7, 7
137
\end{tabular}
138
\end{center}
139
140
\newpage
141
\begin{center}
142
\vspace{1ex}
143
Eigenforms on $\Gamma_0(3)$ congruent mod $3$ to
144
twists of $\Delta$\vspace{1ex}\\
145
\begin{tabular}{l|l|l}
146
$n\quad$ & $\quad\ord_3(\Delta^{(1)}-g_i)$ & $\quad\ord_3(\Delta^{(-1)}-g_i)$
147
\\\hline
148
$1$ & 1, 1, 1 & 1, 1, 1\\
149
$2$ & 1, 2, 1 & 2, 1, 2 \\
150
$3$ & 3, 1, 3, 3, 1 &1, 3, 1, 1, 3\\
151
$4$ & 4, 4, 1, 1, 1, 3, 4, 4, 1, 1, 1
152
& 1, 1, 4, 4, 3, 1, 1, 1, 4, 4, 3
153
\end{tabular}
154
\end{center}
155
156
\begin{center}
157
\vspace{1ex}
158
Eigenforms on $\Gamma_0(5)$ congruent mod $5$ to
159
twists of $\Delta$\vspace{1ex}\\
160
\begin{tabular}{l|l|l}
161
$n\quad$ & $\quad\ord_5(\Delta^{(1)}-g_i)$ & $\quad\ord_5(\Delta^{(-1)}-g_i)$
162
\\\hline
163
$1$ & 1, 0, 1, 0, 0 & 0, 1, 0, 1, 1\\
164
$2$ & 2, 1, 1, 0, 0, 2, 1, 1, 0, 0, 0& 0, 0, 0, 1, 2, 0, 0, 0, 1, 1, 2\\
165
\end{tabular}
166
\end{center}
167
168
\begin{center}
169
\vspace{1ex}
170
Eigenforms on $\Gamma_0(7)$ congruent mod $7$ to twists of $\Delta$
171
\vspace{1ex}\\
172
\begin{tabular}{l|l}
173
$n\quad$ & $\quad\ord_7(\Delta^{(1)}-g_i)$, $\quad\ord_7(\Delta^{(-1)}-g_i)$
174
\\\hline
175
$1$ & [1, 1, 0, 0, 1, 1, 0, 0, 0], [1, 1, 0, 0, 1, 1, 0, 0, 0] \\
176
$2$ & [1, 2, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 2, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0], \\
177
& [1, 1, 2, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, 2, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0]
178
\end{tabular}
179
\end{center}
180
181
\begin{center}
182
\vspace{1ex}
183
Eigenforms on $\Gamma_0(11)$ congruent mod $11$ to twists of $\Delta$
184
\vspace{1ex}\\
185
\begin{tabular}{l|l}
186
$n\quad$ & $\quad\ord_{11}(\Delta^{(1)}-g_i)$,
187
$\quad\ord_{11}(\Delta^{(-1)}-g_i)$
188
\\\hline
189
$1$ & [ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ],\\
190
& [ 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 ] \\
191
$2$ & \\
192
\end{tabular}
193
\end{center}
194
195
\comment{
196
\begin{remark}
197
There seems to be a form~$F$ of weight~$30$ congruent to $\Delta \pmod{27}$
198
and a form~$G$ congruent to $\Delta$ $3$-deprived, $\Delta(3)$,
199
modulo~$27$ of slope~$14$. But the evil twin
200
of~$F$ is also congruent to $\Delta(3) \pmod{27}$
201
and it has slope~$27$! The evil twin is an old eigenform in
202
the space $S_{30}(\Gamma_0(3),\Z_3)$.
203
\end{remark}}
204
205
206
\subsection{Twisting the weight-$2$ newform on $X_0(11)$}\label{exp:11}
207
Let~$f$ be the form corresponding to the elliptic curve of conductor~$11$.
208
Fix a prime~$p$, and let $\eps$ be the trivial character modulo~$p$.
209
We list the congruences between $f^{\eps}$ and the $\Z_2$-rational
210
newforms in $S_{2+\vphi(p^n)}(\Gamma_0(p\cdot 11))$.
211
212
For example, the 2nd line of the table below
213
says that when $n=2$ there are three
214
$\Z_2$-rational newforms
215
$g_1,g_2,g_3$; and that $f^{(1)} \con g_1 \pmod{3}$,
216
$f^{(1)} \not\con g_2 \pmod{3}$,
217
and $f^{(1)} \con g_3 \pmod{3}$.
218
The 3rd line says there are five $\Z_2$-rational newforms
219
in $S_{2+\vphi(2^3)}(\Gamma_0(22))$,
220
and the first, $g$, satisfies $f^{(1)}\con g\pmod{3^2}$.
221
222
\begin{center}
223
\vspace{1ex}
224
Eigenforms on $\Gamma_0(22)$ congruent mod~$2$ to
225
the $2$-deprivation of~$f$\vspace{1ex}\\
226
\begin{tabular}{l|l}
227
$n\quad$ & $\quad\ord_2(f^{(1)}-g_i)$\\\hline
228
$1$ & none\\
229
$2$ & 1, 0, 1\\
230
$3$ & 2, 1, 0, 1, 2 \\
231
$4$ & 1, 1, 3, 3, 0 \\
232
$5$ & 4, 1, 0, 4, 0, 1, 0
233
\end{tabular}
234
\end{center}
235
236
\newpage
237
\begin{center}
238
\vspace{1ex}
239
Eigenforms on $\Gamma_0(33)$ congruent mod $3$ to the
240
$3$-deprivation of~$f$\vspace{1ex}\\
241
\begin{tabular}{l|l}
242
$n\quad$ & $\quad\ord_3(f^{(1)}-g_i)$\\\hline
243
$1$ & 0, 1, 1, 0 \\
244
$2$ & 1, 0, 0, 1, 0, 0 \\
245
$3$ & 0, 0, 0, 0, 1, 1, 1, 0, 1, 1, 1, 0, 0, 0, 0, 0
246
\end{tabular}
247
\end{center}
248
249
\begin{center}
250
\vspace{1ex}
251
Eigenforms on $\Gamma_0(33)$ congruent mod $3$ to the
252
quadratic twist of~$f$\vspace{1ex}\\
253
\begin{tabular}{l|l}
254
$n\quad$ & $\quad\ord_3(f^{(-1)}-g_i)$\\\hline
255
$1$ & 1, 0, 0, 0\\
256
$2$ & 0, 2, 1, 0, 1, 0\\
257
$3$ & 3, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 1, 1, 1
258
\end{tabular}
259
\end{center}
260
261
\begin{center}
262
\vspace{1ex}
263
Eigenforms on $\Gamma_0(55)$ congruent mod $5$
264
to twists of~$f$\vspace{1ex}\\
265
\begin{tabular}{l|l|l}
266
$n\quad$ & $\quad\ord_5(f^{(-1)}-g_i)$ & $\quad\ord_5(f^{(-1)}-g_i)$ \\\hline
267
$1$ & 1, 0 & 0, 0\\
268
$2$ & 0, 2, 1, 0, 1, 0 & \\
269
\end{tabular}
270
\end{center}
271
272
\subsection{Twisting $X_0(14)$}\label{exp:14}
273
\begin{center}
274
\vspace{1ex}
275
Eigenforms on $\Gamma_0(3\cdot 14)$ congruent mod $3$
276
to twists of $h\in S_2(\Gamma_0(14))$\vspace{1ex}\\
277
\begin{tabular}{l|l|l}
278
$n\quad$ & $\quad\ord_3(h^{(1)}-g_i)$& $\quad\ord_3(h^{(-1)}-g_i)$
279
\\\hline
280
$1$ & 0, 1 & 0, 0\\
281
$2$ & 0, 1, 0, 0, 0, 0 & 0, 0, 2, 0, 2, 0
282
\end{tabular}
283
\end{center}
284
285
\bibliographystyle{amsalpha}
286
\bibliography{biblio}
287
288
\end{document}
289
290
291
Dear Robert,
292
293
I did a few more computations, up to weight 12+phi(3^4),
294
and things look perfect. Can you prove Conjecture 1.1 in
295
the attached file? Shall we plan a big systematic computation
296
to test a ``twisting conjecture''?
297
298
Best,
299
William
300
301
302
303
304
5 Oct 1999:
305
306
Dear Robert,
307
308
I THINK ORD_3(A_3(\DELTA))=2, IF I READ YOUR TABLES CORRECTLY.
309
310
Yes, this is correct.
311
312
313
314
315
Maybe we can fortmulate a conjecture for infinite slope forms which would
316
implie the GM conjecturer but is not implied by it. Here is a
317
false conjecture
318
319
\proclaim False Conjecture. Let F(q) be the q-expansion of a weight k
320
tame level N infinitre slope overconvergent p-adic eigenform. Then
321
for intgers r and n\ge 1 (r,p)=1, there exists an overconvergent
322
eigenform of weight k+r\phi(p^n) of slope n and tamre level N such that
323
G(q) is congruent to F(q) modulo p^n.
324
325
Remark. We know if n < the weight of G minus one, G must be
326
classical.
327
328
GM says: if d(k,\alp)=1 ("obvious" notation) and F is as form of weight
329
k and slope \alp, then for every integer of k'\ge k congruent to k modulo
330
\phi(p^n), n\ge\alp , there exists G of weight k' and
331
slope \alp such that
332
333
G(q)\con F(q) modulo p^n
334
335
When p=3 and F=\Delta, this implies there exists a form of weight 2*9
336
+12 of slope 2 congruernt to \Delta modulo 27.
337
338
I just checked that this is true.
339
340
You found one of slope \ge 3 congruent to the twist of
341
Delta by \omega of this weight.
342
What was the slope of the form you found?
343
344
Over Z_3 there are five distinct eigenforms at weight 30.
345
Two of them, g and h, are congruent modulo 27 to the twist of
346
Delta by omega. The slope of g is 14 (i.e., ord_3(a_3(g))=14.),
347
and the slope of h is also FOURTEEN.
348
349
This is interesting: let's see what the slopes are for the
350
other examples:
351
slopes
352
--------
353
k = 12: 5
354
k = 14: 6
355
k = 30: 14, 14
356
k = 66: 32,32, 32,32
357
In general, by Atkin-Lehner theory, because we are considering
358
newforms on Gamma_0(3), the slope is (k-2)/2. [See e.g.,
359
page 64 of Diamond-Im].
360
361
Is there a form of this weight congruent
362
to the twist of \Delta by \omega^2, i.e., thre
363
3-deprived version of \Delta, \Delta(3)?
364
365
Let depriv := Delta x omega^2.
366
First, at weight 12 there is a form congruent to deprive mod 3, but not
367
modulo 3^2.
368
Next, weight 14: There is a form congruent mod 3, but NOT modulo 3^2.
369
Next, weight 30: There are three forms cong. mod 3^3 to deprive.
370
Next, weight 66: There are forms cong. mod 3^4 to deprive.
371
372
Is there a form of weight 6 congruent to \Delta(3) modulo 9?
373
374
YES, there is a weight 6 form congruent to \Delta(3) modulo 9, but not
375
modulo 27.
376
377
We might be heading to another onjectural generalization of
378
Hida theory.
379
380
Great!
381
382
William
383
384
Buffers Files Tools Edit Search Mule Help
385
There seems to be a form F of weight 30 congruent to \Delta mod 27
386
and a form G congruent to \Delta 3-deprived, \Delta(3), mod 27 of
387
slope 14. But the evil twin of F is also congruent to \Delta(3) mod
388
27 and it has slope 27!
389
390
Amazing!
391
392
Let f be the form corresponding to the ellptic curve of conductor 11.
393
Let f(2) be the 2-deprivation of f. I wrote a program to list the
394
Z_2-rational newforms in S_{2+Phi(2^n)}(Gamma_0(2*11)). The third
395
line of the following table says that when n=3 there are 3 Z_2 rational
396
newforms h1,h2,h3; and that f(2) = h1 (mod 3), f(2) =/= h2 (mod 3),
397
and f(2) = h3 (mod 3).
398
399
p = 2, f in S_2(Gamma_0(11)).
400
n | congruences
401
-----------------------------
402
0 | none
403
1 | none
404
2 | [ 1, 0, 1 ]
405
3 | [ 2, 1, 0, 1, 2 ]
406
4 | [ 1, 1, 3, 3, 0 ]
407
5 | [ 4, 1, 0, 4, 0, 1, 0 ]
408
409
Amazingly there are forms in S_{2+Phi(2^5)}(Gamma_0(22)) that
410
are congruent mod 2^4 to f(2)!
411
412
Note: Because I am only considering newforms, the slopes are (k-2)/2.
413
414
p = 3, f in S_2(Gamma_0(11)).
415
416
n | congruences
417
-----------------------------
418
0 | none
419
1 | [ 0, 1, 1, 0 ]
420
2 | [ 1, 0, 0, 1, 0, 0 ]
421
3 |
422
423
William
424
425
426
427