The pp-adic Number System and the Artin-Hasse Exponential

The pp-adic Number System and the Artin-Hasse Exponential

Achyut Bharadwaj, Lex Harie Pisco, Krittika Garg, Swayam Chaulagain, Counsellor: Sanskar Agrawal, Mentor: Nischay Reddy
June 2023
pp-adic numbers, Artin-Hasse exponential, radius of convergence, power series, integral power series, discs, leftist numbers

Introduction #

In this paper, we explore the pp-adic system, by defining it in multiple ways: as an extension of the pp-adic integers, as well as an extension of the rationals. We then proceed to perform analysis in the pp-adics, by defining convergence, continuity and discs. We then describe exponentiation and logarithmic functions over the pp-adics as functions derived from power series. We explore the radius of convergence and other properties of these functions. We then explore the Artin-Hasse exponential, which, though seemingly random, turns out to be an integral power series.

Formal Power Series #

The Ring R[[x]]R[[x]] #

Prove that R[[x]]R[[x]] is a ring under the natural operations of addition and multiplication.

Addition : Let f,gR[[x]]f,g \in R[[x]] such that f(x)=i=0aixif(x)= \sum_{i=0}^\infty a_i x^i and g(x)=j=0bjxjg(x)= \sum_{j=0}^\infty b^j x^j. Then f(x)+g(x)=i=0aixi+j=0bjxj=i,j=0(ai+bj)xif(x) + g(x) = \sum_{i=0}^\infty a_i x^i + \sum_{j=0}^\infty b^j x^j = \sum_{i,j=0}^\infty (a_i + b_j) x^i. Since ai+bj{a_i + b_j } is closed under addition as aia_iand bjRb_j \in R we can say, R[[x]]R[[x]] is a ring under addition.

Multiplication : For proving the multiplication, we need to prove the commutative property in a formal power series. Let f,gR[[x]]f,g \in R[[x]] such that f(x)=i=0aixif(x)= \sum_{i=0}^\infty a_i x^i and g(x)=j=0bjxjg(x)= \sum_{j=0}^\infty b^j x^j. Then f(x)g(x)=i=0aixi×j=0bjxj=k=0(i=0kaibki)xkf(x)g(x) = \sum_{i=0}^\infty a_i x^i \times \sum_{j=0}^\infty b^j x^j = \sum_{k=0}^\infty\left(\sum_{i = 0}^ka_ib_{k-i}\right)x^k

Since the sum i=0kaibki\sum_{i=0}^ka_ib_{k-i} is a sum of products of elements of RR, the sum itself is in RR since RR is closed under addition and multiplication. Therefore, the above sum is a power series in RR, that is the products fgR[[x]]fg \in R[[x]].
 
Now, since the ring is commutative over both ++ and \cdot, it follows that R[[x]]R[[x]] is also commutative. Moreover, it is clearly associative over ++ since the ring RR is itself associative over ++. Next, the ring has a zero element, i.e. 0R[[x]]0 \in R[[x]] since for any fR[[x]]f\in R[[x]], we have f+0=ff + 0 = f. Similarly, it also has an identity element, i.e. 11. Also, the additive inverse of ff is the power series with the coefficients each being the additive inverse of the coefficients of ff. Thus, every element of R[[x]]R[[x]] has an inverse element. Distributivity can also be easily seen from the fact that RR is itself distributive.
 
We will now prove associativity over multiplication. Let a=i=0aixi,b=i=0bixi,c=i=0cixia = \sum_{i=0}^\infty a_ix^i, b = \sum_{i=0}^\infty b_ix^i, c = \sum_{i=0}^\infty c_ix_i Now, ab=k=0(i=0kaibki)xk=i=0pixiab = \sum_{k=0}^\infty\left(\sum_{i=0}^ka_ib_{k-i}\right)x^k = \sum_{i=0}^\infty p_ix^i So, (ab)c=k=0(i=0kpicki)xk=k=0(i=0k(j=0iajbij)cki)xk(ab)c = \sum_{k=0}^\infty \left(\sum_{i=0}^k p_ic_{k-i}\right)x^k = \sum_{k=0}^\infty\left(\sum_{i=0}^k \left(\sum_{j=0}^ia_jb_{i-j}\right)c_{k-i}\right)x^k The above can be expanded as k=0(c0a0b0+c1(a0b1+a1b0)+ck(a0bk+akb0)xk\sum_{k=0}^\infty (c_0a_0b_0 + c_1(a_0b_1 + a_1b_0) + \cdots c_k(a_0b_k + \cdots a_kb_0)x^k Now, a(bc)=(bc)aa(bc) = (bc)a by commutativity. Therefore, we may use a similar approach as above to show that a(bc)=k=0(a0b0c0+a1(b0c1+b1c0)++ak(b0ck+bkc0))xka(bc) = \sum_{k=0}^\infty (a_0b_0c_0 + a_1(b_0c_1 + b_1c_0) + \cdots + a_k(b_0c_k + \cdots b_kc_0))x^k By rearranging the terms above, we get a(bc)=k=0(c0a0b0+c0(a0b1+a1b0)++ck(a0bk+akb0))xk=(ab)ca(bc) = \sum_{k=0}^\infty (c_0a_0b_0 + c_0(a_0b_1 + a_1b_0) + \cdots + c_k(a_0b_k + \cdots a_kb_0))x^k = (ab)c This proves associativity over multiplication.

Units in R[[x]]R[[x]] #

Definition 1. f(x)R[[x]]f(x) \in R[[x]] then f(0)=a0f(0) = a_0

Definition 2. Units in R[[x]]R[[x]] are all ff and gg in R[[x]]R[[x]] such that [fg](x)=1[f \cdot g](x) = 1.

Examples of these units are:

  1. f(x)=(x1)f(x) = -(x-1)

  2. f(x)=i=0xif(x) = \sum_{i=0}^{\infty} x^i

  3. f(x)=1+i=1naixif(x) = 1 + \sum_{i=1}^n a_ix^i for aiRa_i \in R

  4. f(x)=(±(x1)i=0xi)nf(x) = (\pm (x-1) \cdot \sum_{i=0}^{\infty} x^i)^n for nNn \in \mathbb{N}

Proposition 1. If f(x)R[[x]]f(x) \in R[[x]] and f(0)=1f(0) = 1, then f(x)f(x) is a unit in R[[x]]R[[x]].

Proof. Let f(x),g(x)f(x), g(x) be in R[[x]]R[[x]] such that fg=1f \cdot g = 1. We need to prove that such a power series, gg, exists if and only if we have f(0)=1f(0) = 1. Let f(x)=a0+a1x+f(x) = a_0 + a_1x + \cdots and g(x)=b0+b1x+.g(x) = b_0 + b_1x + \cdots. Thus, proving the existence of gg is equivalent to proving that there exists a sequence {bi}\lbrace b_i \rbrace such that (a0+a1x+)(b0+b1x+)=1.(a_0 + a_1x + \cdots)(b_0 + b_1x + \cdots) = 1. Expanding this product, and combining the terms of equal degree, we get: a0b0+(a0b1+a1b0)x++(a0bn++anb0)xn+=1a_0b_0 + (a_0b_1 + a_1b_0)x + \cdots + (a_0b_n + \cdots + a_nb_0)x^n + \cdots = 1 Since the RHS is simply 11, we need all terms (a0bn+anb0)=0(a_0b_n + \cdots a_nb_0) = 0 for all n1n\geq1 and a0b0=1a_0b_0 = 1.
 
Now, if f(0)=a0f(0) = a_0 is not a unit, then we cannot find a b0b_0 in RR such that a0b0=1a_0b_0 = 1. This shows that we cannot find the desired sequence {bi}\lbrace b_i \rbrace. Thus, if f(x)f(x) is a unit in R[x]R[x], it follows that f(0)f(0) is a unit.
 
We now prove the converse. That is, if f(0)f(0) is a unit in RR, then there exists some sequence {bi}\lbrace b_i\rbrace so that g(x)=i0bixig(x)=\sum_{i\geq0}b_ix^i and f(x)g(x)=1f(x)g(x) = 1. We use induction in order to do so. We will first prove that there exists some sequence {bi}i=0\lbrace b_i\rbrace_{i=0} (i.e. a single term b0b_0) so that a0b0=1a_0b_0 = 1. This finishes our base case. Now, assume that there exists some sequence {bi}i=0n\lbrace b_i\rbrace_{i=0}^n so that a0b0=1a_0b_0 = 1 and for all 0<kn0 < k \leq n we have (a0bk+akb0)=0.(a_0b_k + \cdots a_kb_0) = 0. We prove that there exists a sequence {bi}i=0n+1\lbrace b_i\rbrace_{i=0}^{n+1} so that the same holds for k=n+1k = n+1 as well. We have an+1b0+a0bn+1=0.a_{n+1}b_0 + \cdots a_0b_{n+1} = 0. Therefore, we have bn+1=an+1b0++a1bna0b_{n+1} = -\frac{a_{n+1}b_0 + \cdots + a_1b_n}{a_0} Since we know bib_i exist for all ii less than or equal to nn, we have that bn+1b_{n+1} also exists, completing our inductive step. Thus, if f(0)f(0) is a unit, then f(x)f(x) has an inverse. ◻

Compositions of formal power series #

We will now generalize as to when f(g(x))f(g(x)) where g(x)R[[x]]g(x) \in R[[x]] is an element of R[[x]]R[[x]] itself.

Proposition 2. Let f,gR[[x]]f, g \in R[[x]]. Then, f(g(x))R[[x]]f(g(x)) \in R[[x]] if and only if g(0)=0g(0) = 0.

Proof. Suppose g(0)=0g(0) = 0. Then, we can write g(x)=xk+h(x)g(x) = x^k + h(x) where kk is the smallest power of xx in g(x)g(x), and k0k \neq 0 since g(0)=0g(0)=0. Moreover, the smallest power of xx in h(x)h(x) is more than kk. Also, let f(x)=i=0aixif(x) = \sum_{i=0}^\infty a_ix^i Therefore, f(g(x))=a0+a1g(x)+a2g(x)2+=a0+a1(xk+h(x))+a2(xk+h(x))2+f(g(x)) = a_0 + a_1g(x) + a_2g(x)^2 + \cdots = a_0 + a_1(x^k + h(x)) + a_2(x^k + h(x))^2 + \cdots Also let f(g(x))=c0+c1x+c2x2+f(g(x)) = c_0 + c_1x + c_2x^2 + \cdots In order to prove that f(g(x))R[[x]]f(g(x)) \in R[[x]] it suffices to prove that cnRc_n \in R for all nn. So, we will work on finding the coefficient of the xnx^n term, i.e. cnc_n.
 
Consider the term an+1(xk+h(x))n+1a_{n+1}(x^k + h(x))^{n+1} in the expansion for f(g(x))f(g(x)). The term with the lowest power in this expansion is x(n+1)kx^{(n+1)k}. Since k0k \neq 0, it follows that (n+1)k>n(n+1)k > n. Therefore, the coefficient cnc_n of xnx^n is independent of an+1a_{n+1}. By a similar argument, cnc_n is independent of aja_j for all j>nj > n. In other words, it follows that cn=i=0naitic_n = \sum_{i = 0}^n a_it_i where tiRt_i \in R. Note that there will be no powers of aia_i in the expansion since the coefficients aia_i are not raised to a power in the expansion of f(g(x))f(g(x)). Since ai,tiRa_i, t_i \in R, it follows that aitiRa_it_i \in R. Thus, i=0naitiR\sum_{i=0}^na_it_i \in R. Therefore, for all nn, we have cnRc_n \in R, where f(g(x))=c0+c1x+f(g(x)) = c_0 + c_1x + \cdots Therefore, by definition of a formal power series, f(g(x))R[[x]]f(g(x)) \in R[[x]].
 
Now, suppose g(0)=c0g(0) = c \neq 0. Let g(x)=c+h(x)g(x) = c + h(x). Then, f(g(x))=a0+a1(c+h(x))+a2(c+h(x))2+=c0+c1x+f(g(x)) = a_0 + a_1(c + h(x)) + a_2(c+h(x))^2 + \cdots = c_0 + c_1x + \cdots In the nnth term of the above expansion, we have a constant cnc^n. Since there are infinitely many terms in the expansion, c0c_0 is an infinite sum, which is not defined in RR. Therefore, c0Rc_0 \notin R. It follows that f(g(x))R[[x]].f(g(x)) \notin R[[x]]. ◻

Multivariate Power Series #

We can define the system R[[x,y]]R[[x,y]] to (R[[x]])[[y]](R[[x]])[[y]]. Inductively, we may define R[[x1,,xk]]=(R[[x1,,xk1]])[[xk]]R[[x_1, \dots, x_k]] = (R[[x_1, \dots, x_{k-1}]])[[x_k]] Since RR is a ring implies R[[x]]R[[x]] is a ring (base case) Then, we assume that R[[x1,x2,x3,xk]]R[[x_1,x_2, x_3, … x_k]] is a ring Now, R[[x1,x2,X3,xk+1]]R[[x_1, x_2, X_3, … x_k+1]] is (R[[x1,x2,x3,xk]])[[xk+1]](R[[x_1,x_2, x_3, … x_k]])[[x_k+1]] which is a ring (inductive case). Thus, by induction we can say that R[[x1,x2,X3,xn+1]]R[[x_1, x_2, X_3, … x_n+1]] is a ring.

We will now generalize. For which power series gR[[x,y]]g \in R[[x,y]] can we define f(g(x,y))f(g(x,y)) for any fR[[t]]f \in R[[t]]?

Proposition 3. For f,gR[[x,y]]f, g \in \mathbb{R}[[x, y]], we have that f(g(x,y))f(g(x, y)) is defined in R[[x,y]]R[[x, y]] if and only if g(0,0)0g(0, 0) \neq 0.

Proof. Suppose g(0,0)=0g(0, 0) = 0. Thus, the smallest power of xx and yy can be represented as bmnxmynb_{mn}x^my^n where both mm and nn are not 00 at the same time. So, we let g(x,y)=bmnxmyn+h(x,y)g(x, y) = b_{mn}x^my^n + h(x, y). Thus, f(g(x,y))=f(bmnxmyn+h(x,y))=a0+a1(bmnxmyn+h(x,y))+a2(bmnxmyn+h(x,y))2+=c00+c10x+c01y+ \begin{aligned} f(g(x, y)) &= f(b_{mn}x^my^n + h(x, y)) \\ &= a_0 + a_1(b_{mn}x^my^n + h(x, y))+a_2(b_{mn}x^my^n + h(x, y))^2+\cdots \\ &= c_{00} + c_{10}x + c_{01}y + \cdots \end{aligned} Now, consider the coefficient, cpqc_{pq}. We know that the term ap+q+1(bmnxm+yn+h(x,y)p+q+1)a_{p+q+1}(b_{mn}x^m+y^n + h(x, y)^{p+q+1}) has smallest power xm(p+q+1)yn(p+q+1)x^{m(p+q+1)}y^{n(p+q+1)} which is clearly greater than the power of xpyqx^py^q. Hence, cpqc_{pq} is independent of aka_k for all k>p+qk> p+q. Therefore, we have that cpq=i=0p+qaidic_{pq} = \sum_{i=0}^{p+q}a_id_i which is clearly an element of RR. Therefore, f(g(x,y))f(g(x, y)) makes sense in R[[x,y]]R[[x, y]].
 
Now, if g(0,0)=cg(0, 0) = c which is a constant, then let g(x,y)=c+h(x,y)g(x, y) = c + h(x, y). Therefore, we have f(g(x,y))=a0+a1(c+h(x,y))+a2(c+h(x,y))2+f(g(x, y)) = a_0 + a_1(c+h(x, y)) + a_2(c+h(x, y))^2 + \cdots The constant term of the above expansion will be a0+a1c+a2c2+a_0 + a_1c + a_2c^2 + \cdots which is an infinite sum. This does not make sense in RR since we can only compute finite sums in RR. Therefore, if c0c\neq 0, then there is no way to make sense of f(g(x,y))f(g(x, y)) as an element of R[[x,y]]R[[x, y]]. ◻

Polynomial Fields and Rational Functions #

For some field kk, we write k[x]k[x] to denote the set of polynomials with coefficients in kk. We write k(x)k(x) to denote the set {g(x)h(x)g(x),h(x)R[[x]]}\left\lbrace \left .\frac{g(x)}{h(x)} \right \rvert g(x), h(x) \in R[[x]]\right \rbrace We will now explore the relation between k(x),k[x]k(x), k[x] and k[[x]]k[[x]]. Firstly, note that k[x]k(x)k[x] \in k(x) by letting g(x)=1g(x) = 1. Also, clearly k[x]k[[x]]k[x] \in k[[x]]. This is because a polynomial is a formal power series with the coefficients equal to 00 for all xnx^n where nn is greater than the degree of the polynomial. We now check when an element of k(x)k(x) can be written as a formal power series. Consider an example of the power series expansion of a rational function g(x)/h(x)g(x)/h(x), t(x)t(x). Let

g(x)=2x+1g(x) = 2x+1 and h(x)=x1+1h(x) = x^1 + 1. Then, g(x)h(x)=2x+1x2+1=1+x+2xx32x4+x5+2x6\frac{g(x)}{h(x)} = \frac{2x+1}{x^2 + 1} = 1 + x + 2x -x^3 -2x^4 + x^5 + 2x^6 - \cdots Notice how the terms of the power series are recursive. The nnth term of t(x)t(x) can be found in terms of the previous terms of t(x)t(x). In fact, this holds in general too!

Proposition 4. We claim that a rational function g(x)/h(x)g(x)/h(x), i.e. an element of k(x)k(x) with h(0)0h(0) \neq 0 can be written as a power series t(x)t(x).

Proof. If h(0)0h(0) \neq 0, then we have already proved that h(x)h(x) is a unit. Thus, 1/h(x)R[[x]]1/h(x) \in R[[x]]. So, since formal power series are closed under multiplication, g(x).(1/h(x))R[[x]]g(x). (1/h(x)) \in R[[x]]. Thus, g(x)/h(x)g(x)/h(x) can be written as a power series t(x)t(x). If h(0)=0h(0)=0, then h(x)h(x) is not a unit which implies 1/h(x)R[[x]]1/h(x) \notin R[[x]] and thus, g(x)/h(x)g(x)/h(x) cant be expressed as a formal power series. ◻

Proposition 5. Let t(x)=g(x)/h(x)t(x) = g(x)/h(x) be the power series expansion of a rational function. Then, the terms ana_n of t(x)t(x) satisfy a linear recursion. Formally, this means that there exists a number m1m\geq 1 and constants c1,,cmkc_1, \dots, c_m \in k such that for all nn that is sufficiently large, an=i=1mciania_n = \sum_{i=1}^mc_ia_{n-i}

Proof. Firstly, let the degree of h(x)h(x) be mm and the degree of g(x)g(x) be kk. Let h(x)=b0+b1x+bmxmh(x) = b_0 + b_1x + \cdots b_mx^m and t(x)=a0+a1x+t(x) = a_0 + a_1x + \cdots. Now, since t(x)=g(x)/h(x)t(x) = g(x)/h(x), we may write g(x)=t(x)h(x)g(x) = t(x)h(x).
 
Since the degree of g(x)g(x) is kk, it follows that the coefficients of xnx^n in the expansion of the product on the RHS is 00 for all nn that is greater than kk.
 
Now, consider some n>max(m,k)n > \max(m, k). Clearly, n>kn>k. Therefore, the coefficient of xnx^n in the expansion will be 00. Let us find the coefficient using the fact that t(x)=a0+a1x+t(x) = a_0 + a_1x + \cdots and h(x)=b0+b1x++bmxmh(x) = b_0 + b_1x + \cdots + b_mx^m. This will be anb0+an1b1++anmbma_nb_0 + a_{n-1}b_1 + \cdots + a_{n-m}b_m Since n>kn>k, the above must be 00. Rearranging the terms, we get anb0=(an1b1+anmbm)a_nb_0 = -(a_{n-1}b_1 + \cdots a_{n-m}b_m) Since kk is a field and b0=h(0)0b_0 = h(0) \neq 0, we have an=b01(an1b1+anmbm)=i=1mciania_n = -b_0^{-1}\left(a_{n-1}b_1 + \cdots a_{n-m}b_m\right) = \sum_{i=1}^mc_ia_{n-i} where ci=b01bic_i = -b_0^{-1}b_i. This proves that the coefficients of the power series t(x)t(x) satisfy a linear recursion. ◻

Proposition 6. The converse of the previous lemma also holds true. That is, if t(x)=nanxnk[[x]]t(x) = \sum_{n\geq}a_nx^n\in k[[x]] is such that the coefficients ana_n satisfy a linear recursion, then there exists some rational function g(x)/h(x)g(x)/h(x) which has a power series expansion of t(x)t(x).

Proof. We know that for t(x)t(x), we have an=c1an1+c2an2+cmanma_n = c_1a_{n-1} + c_2a_{n-2} + \cdots c_ma_{n-m} for nn that is sufficiently large enough, since we are given that the coefficients of tt satisfy a linear recursion. Therefore, we have anc1an1cmanm=0a_n - c_1a_{n-1} - \cdots - c_ma_{n-m} = 0 Let b0=1b_0 = 1 and bi=cib_i = -c_i when i1i\geq1. Therefore, we may write anb0+an1b1+anmbm=0a_nb_0 + a_{n-1}b_1 + \cdots a_{n-m}b_m = 0 Now, consider h(x)=b0+b1x+bmxmh(x) = b_0 + b_1x + \cdots b_mx^m. Consider the product t(x)h(x)t(x)h(x). The coefficient of xnx^n in this product will be b0an+b1an1+bmanmb_0a_n + b_1a_{n-1} + \cdots b_ma_{n-m} which is 00 by our assumption. Therefore, all terms of t(x)h(x)t(x)h(x) have a coefficient of 00 for nn sufficiently large. Suppose for all n>kn>k we have that the coefficient of xkx^k is 00 (and not 00 for nn not more than kk). Therefore, t(x)h(x)=g(x)t(x)h(x) = g(x) is a polynomial of degree kk. Therefore, g(x)/h(x)g(x)/h(x) is a rational function that is represented by t(x)t(x). Therefore, we have found some rational function that is represented by t(x)t(x) given that the coefficients of t(x)t(x) satisfy a linear recursion. This completes our proof. ◻

pp-adic numbers #

The pp-adic numbers are defined as follows:

Definition 3. The set of pp-adic numbers Zp\mathbb{Z_p} is defined as:

Zp={(a1,a2,)aiZ/piZ and ai+1ai(modpi)\mathbb{Z_p} = \lbrace (a_1, a_2, \dots) \mid a_i \in \mathbb{Z}/p^i \mathbb{Z} \text{ and } a_{i+1} \equiv a_i \pmod{p^i}

One can prove that the set of pp-adic numbers, Zp\mathbb{Z_p} forms a ring under term-wise addition and multiplication.

Proposition 7. The ring Zp\mathbb{Z_p} is an integral domain, i.e., if for some a,bZpa, b \in \mathbb{Z_p} we have ab=0ab = 0, then either a=0a=0 or b=0b=0.

Proof. In Zp\mathbb{Z_p}, we have 0=(0,0,)0 = (0, 0, \dots). Thus, we need to prove that either a=(0,0,)a = (0, 0, \dots) or b=(0,0,)b = (0, 0, \dots) given that ab=(0,0,)ab = (0, 0, \dots). First, let a=(a1,a2,)a = (a_1, a_2, \dots) and b=(b1,b2,)b = (b_1, b_2, \dots). Then, ab=(a1b1,a2b2,)=(0,0,)ab = (a_1b_1, a_2b_2, \dots) = (0, 0, \dots) This implies that we must have aibi0(modpi)a_ib_i \equiv 0 \pmod{p^i} for all ii. Clearly, we must also have a1b10(modp)a_1b_1 \equiv 0 \pmod{p}. Since Zp\mathbb{Z_p} is an integral domain, we must have a10(modp)a_1 \equiv 0 \pmod p or b10(modp)b_1 \equiv 0 \pmod{p}. Assume without loss of generality that a10(modp)a_1 \equiv 0 \pmod{p}. We will prove that either ai=0a_i = 0 for all ii or bi=0b_i = 0 for all ii.
 
We will prove this by contradiction. Suppose that ai0a_i \neq 0 and bi0b_i \neq 0. Then, there exists some aka_k such that ak0(modpk)a_k \equiv 0 \pmod {p^k}. Also assume without loss of generality that bi0b_i \neq 0 for some iki \leq k (if the smallest such ii is greater than kk, then we just exchange aa and bb). If ak0(modpk)a_k \equiv 0 \pmod{p^k}, then by the definition of a pp-adic number, we must have ak+1ak≢0(modpk)a_{k+1} \equiv a_k \not \equiv 0 \pmod{p^k}. Thus, ak+10a_{k+1} \neq 0. It follows that for any iki \geq k, we must have ai0a_i \neq 0 as well as bi0b_i \neq 0.
 
Now, since ab=0ab = 0, we must have aibi0(modpi)a_ib_i \equiv 0 \pmod{p^i} for all iki \geq k. Now, we have ak10(modpk1)a_{k-1} \equiv 0 \pmod{p^{k-1}} but ak0(modpk)a_k \equiv 0 \pmod{p^k}. Moreover, by definition, we have akak1(modpk1)a_k \equiv a_{k-1} \pmod{p^{k-1}}. Therefore, it follows that akpk1m1(modpk)a_k \equiv p^{k-1}m_1 \pmod{p^k} where 0<m1<p0 < m_1 < p. Next, we have ak+1pkm0+pk1m1(modpk+1)a_{k+1} \equiv p^km_0 + p^{k-1}m_1 \pmod{p^{k+1}} where 0<m0<p0 < m_0 < p. In general, we therefore have ak+ipk+i1mi1++pkm0(modpk+i)a_{k+i} \equiv p^{k+i - 1}m_{i-1} + \cdots + p^km_0 \pmod{p^{k+i}} Let us say the smallest value of ii for which bi0b_i \neq 0 is i=ri = r. We assumed earlier (WLOG) that r<kr < k. Now, just as before, we may write: br+jpr+j1nj1++prn0(modpr+j)b_{r+j} \equiv p^{r+j-1}n_{j-1} + \cdots + p^rn_0 \pmod{p^{r+j}} Since k>rk > r, let k=r+lk = r + l where ll is some natural number. Also, let i=rk+ji = r - k + j. Therefore, br+j=bk+ipk+i1nj1++prn0(modpk+i)b_{r+j} = b_{k+i} \equiv p^{k+i - 1}n_{j-1} + \cdots + p^rn_0 \pmod{p^{k+i}} Now, vp(ak+ibk+i)=vp(ak+i)+vp(bk+i)v_p(a_{k+i}b_{k+i}) = v_p(a_{k+i}) + v_p(b_{k+i}). Since the smallest power of pp in the expansion of ak+ia_{k+i} is kk. Therefore, we have vp(ak+i)=kv_p(a_{k+i}) = k. Since the smallest power of pp in the expansion of bk+ib_{k+i} is rr, we have that vp(bk+i)=r<kv_p(b_{k+i}) = r < k. Thus, vp(ak+ibk+i)=krk2v_p(a_{k+i}b_{k+i}) = kr \leq k^2. Now, consider the case when i=k2k+1i = k^2 - k + 1. We therefore have vp(ak+1bk2+1)k2v_p(a_{k^+1}b_{k^2+1}) \leq k^2. This implies that ak2+1bk2+1≢0(modpk2+1)a_{k^2+1}b_{k^2 +1} \not \equiv 0 \pmod{p^{k^2+1}} However, since ab=0ab = 0, we must have ak2+1bk2+10(modpk2+1)a_{k^2+1}b_{k^2+1} \equiv 0 \pmod{p^{k^2+1}}. A contradiction. Thus, we cannot have both a0a \neq 0 as well as b0b \neq 0. ◻

Let us explore what the units in Zp\mathbb{Z_p} are. We first take an example. Consider Z3\mathbb{Z_3}. Consider a=(1,4,13,40,)Zpa = (1, 4, 13, 40, \dots)\in \mathbb{Z_p}. Also, let b=(1,7,25,79,)b = (1, 7, 25, 79, \dots). Clearly, bb also is an element of Zp\mathbb{Z_p}. Moreover, ab=(11,47,1325,4079,)=(1,1,1,1,)ab = (1\cdot1, 4\cdot7, 13\cdot25, 40\cdot 79, \dots) = (1, 1, 1, 1, \dots) Thus, bb is the inverse of aa in Zp\mathbb{Z_p}. Since aa has an inverse in Zp\mathbb{Z_p}, it follows that aa is a unit in Zp\mathbb{Z_p}. Note that however, a=(0,3,12,)a = (0, 3, 12, \dots) is not a unit in Zp\mathbb{Z_p}. It seems that in general, aZpa\in \mathbb{Z_p} is a unit if and only if the first term of aa is non-zero. In other words, we claim that aZpa \in \mathbb{Z_p} is a unit if and only if a0(modp)a \equiv 0 \pmod{p}.

Proposition 8. In Zp\mathbb{Z_p}, a pp-adic number uu is a unit if and only if u1≢0(modp)u_1 \not \equiv 0 \pmod{p}

Proof. Let u=(u1,u2,)u = (u_1, u_2, \dots). Suppose u1=0u_1 = 0. Then, there is clearly no inverse for uu since there is no b1Z/pZb_1 \in \mathbb{Z}/p\mathbb{Z} such that u1b11(modp)u_1b_1 \equiv 1 \pmod{p} (00 is not a unit modulo pp). Hence, if uu is a unit, then it follows that u1≢0u_1 \not \equiv 0 or equivalently, u0(modp)u \equiv 0 \pmod{p}.
 
We now prove the other direction. Suppose u≢0(modp)u \not \equiv 0 \pmod{p}. Then, we must prove that uu is a unit in Zp\mathbb{Z_p}. Since pp is prime, if u0(modp)u \neq 0 \pmod{p}, it means that u1≢0(modp)u_1 \not \equiv 0 \pmod{p}. Since pp is a prime, it follows that (u1,p)=1(u_1, p) = 1. Thus, u1u_1 is a unit in Z/pZ\mathbb{Z}/p\mathbb{Z}.
 
Now, suppose that (uk,p)=1(u_k, p) = 1. Now, by definition of a pp-adic number, we have uk+1uk(modpk)u_{k+1} \equiv u_k \pmod{p^k}. This in turn implies that uk+1uk(modp)u_{k+1} \equiv u_k \pmod{p}. Thus, (uk+1,p)=1(u_{k+1}, p) = 1. Hence, (uk+1,pk+1)=1(u_{k+1}, p^{k+1}) = 1. Thus, uk+1u_{k+1} is a unit in Z/pk+1Z\mathbb{Z}/p^{k+1}\mathbb{Z}.
 
By induction, this implies that for all kk, we have uku_k is a unit in Z/pkZ\mathbb{Z}/p^k\mathbb{Z}. Since the multiplication operation is termwise in Zp\mathbb{Z_p}, it follows that uu is a unit in Zp\mathbb{Z_p}. ◻

This leads to another important result:

Proposition 9. Let a0a\neq 0 be a pp-adic number. Then, there exists a unique pair (u,k)(u, k) such that a=pkua = p^ku and uu is a unit.

Proof. We will first prove existence. Let a=(a1,a2,)a = (a_1, a_2, \dots). If a10a_1 \neq 0, then by the previous proposition, we have that aa itself is a unit. Thus, a=p0ua = p^0u where u=au=a.
 
Now, suppose that ak=0a_k = 0 for some kk (this automatically implies that a1,a2,,ak=0a_1, a_2, \dots, a_k = 0 by the construction of pp-adic numbers) such that ak+10a_{k+1} \neq 0. Now, consider the pp-adic number b=(ak+1/pk,ak+2/pk,)b = (a_{k+1}/p^k, a_{k+2}/p^k, \dots) We must first prove that bb is indeed a pp-adic integer.
 
For any i>ki>k, by the definition of pp-adics, we know that aiak0(modpk)a_i \equiv a_k \equiv 0 \pmod{p^k}. Thus, ai/pka_i/p^k is an integer for all aia_i.
 
Now, for some ii, we have ak+i+1ak+i(modpk+i)a_{k+i+1} \equiv a_{k+i} \pmod{p^{k+i}} by the definition of pp-adics. Thus, ak+i+1=pin+ak+ia_{k+i+1} = p^in + a_{k+i} where 0<n<p0<n<p. Dividing both sides by pkp^k, we get ak+i+1pk=pin+ak+ipk\frac{a_{k+i+1}}{p^k} = p^{i}n + \frac{a_{k+i}}{p^k} Hence, taking mod pip^i on both sides, we get ak+i+1pkak+ipk(modpi)\frac{a_{k+i+1}}{p^k} \equiv \frac{a_{k+i}}{p^k} \pmod{p^i} Hence, we conclude that bZpb\in \mathbb{Z_p}. Now, consider the pp-adic integer pkbp^kb. We get pkb=pk(ak+1/pk,ak+2/pk,,ak+k/pk,ak+k+1/pk,)=(ak+1,ak+2,,ak+k,ak+k+1,)\begin{aligned} p^kb &= p^k(a_{k+1}/p^k, a_{k+2}/p^k, \dots, a_{k+k}/p^k, a_{k+k+1}/p^k, \dots) \\ &= (a_{k+1}, a_{k+2}, \dots, a_{k+k}, a_{k+k+1}, \dots) \end{aligned} Now, ai+1ak(modpi)a_{i+1} \equiv a_k \pmod{p^i}, or in general, ak+iai(modpi)a_{k+i} \equiv a_i \pmod{p^i} by induction on kk. Thus, pkb=(a1,a2,)=ap^kb = (a_1, a_2, \dots) = a Now, clearly, ak+1/pk≢0(modp)a_{k+1}/p^k \not \equiv 0 \pmod{p} since ak+1a_{k+1} is nonzero. Thus, bb is a unit. Thus, for arbitrary aa, we have found a specific unit bb and power kk such that a=pkba = p^kb, which completes our proof for existence.
 
Now, we prove uniqueness. To do so, suppose that a=pk1u1=pk2u2a = p^{k_1}u_1 = p^{k_2}u_2. Assume WLOG that k1k2k_1 \geq k_2. Then, pk1u1pk2u2=pk2(pk1k2u1u2)=0p^{k_1}u_1 - p^{k_2}u_2 = p^{k_2}(p^{k_1-k_2}u_1 - u_2) = 0 Since Zp\mathbb{Z_p} is an integral domain, it follows that either pk2=0p^{k_2} = 0 or pk1k2u1u2=0p^{k_1-k_2}u_1 - u_2 = 0. Since pk1p^{k_1} is clearly not 00, it follows that pk1k2u1=u2p^{k_1-k_2}u_1 = u_2 If k1>k2k_1>k_2, then the first term of u2u_2 will be 00, which is a contradiction since u2u_2 is a unit. Therefore, k1=k2k_1 = k_2. Thus, u2=p0u1=u1u_2 = p^0u_1 = u_1. Hence, the pair (u,k)(u, k) is unique. ◻

Now, we will explore some properties related to Zp[x]\mathbb{Z_p}[x], i.e. polynomials with coefficients in Zp\mathbb{Z_p}.

Theorem 1. Let f(x)Zp[x]f(x) \in \mathbb{Z_p}[x]. Consider some a1a_1 (if there exists one) such that f(a1)0(modp)f(a_1) \equiv 0 \pmod{p} such that f(a1)≢0(modp)f’(a_1) \not \equiv 0 \pmod{p}. Then, there exists a unique aZpa\in \mathbb{Z_p} such that f(a)=0f(a) = 0 and aa1(modp)a \equiv a_1 \pmod{p}.

Proof. If we prove that if there exists a root, aka_k to f(x)f(x) in Z/pkZ\mathbb{Z}/p^k\mathbb{Z} with f(ak)0f’(a_k) \neq 0, then there exists a unique ak+1a_{k+1} such that f(ak+1)=0f(a_{k+1}) = 0 in Z/pk+1Z\mathbb{Z}/p^{k+1}\mathbb{Z}, it will by induction imply that there exists some unique pp-adic number a=(a1,a2,)a=(a_1, a_2, \dots) such that f(a)=0f(a) = 0 in Zp\mathbb{Z_p}. However, note the condition that ak+1ak(modpk)a_{k+1} \equiv a_k \pmod{p^k}. Let us first prove this result for k=1k=1.
 
Suppose f(a1)0(modp)f(a_1) \equiv 0 \pmod{p} and f(a1)≢0(modp)f’(a_1) \not \equiv 0 \pmod{p}. We will try to find some a2a_2 such that f(a2)0(modp2)f(a_2) \equiv 0 \pmod{p^2} and a2a1(modp)a_2 \equiv a_1 \pmod{p}. f(x)=c0+c1x++cnxnf(x) = c_0 + c_1x + \cdots + c_nx^n in Z/pZ\mathbb{Z}/p\mathbb{Z}. Then, we have f(a1)=c0+c1a1++cna1nf(a_1) = c_0 + c_1a_1 + \cdots + c_na_1^n. Consider a2=a1+mpa_2 = a_1 + mp for some integer 0<m<p0<m<p. Then, we have f(a2)=f(a1+mp)=c0+c1(a1+mp)++cn(a1+mp)n=c0+c1(a1+mp)++cn(a1n+na1n1mp++(mp)n)\begin{aligned} f(a_2) &= f(a_1 + mp) = c_0 + c_1(a_1+mp) + \cdots + c_n(a_1 + mp)^n \\ &= c_0 + c_1(a_1 + mp) + \cdots + c_n(a_1^n + na_1^{n-1}mp + \cdots + (mp)^n) \end{aligned} Now, if we take mod p2p^2 on both sides of the above, all terms with powers of pp over 22 get cancelled to 00. Thus, we get: f(a2)c0+c1(a1+mp)+c2(a12+2a1mp)++cn(a1n+na1n1mp)(c0+c1a1+cnan)+(c1(mp)+2c2(mp)a1+ncn(mp)a1n1)f(a1)+mpf(a1)(modp2)\begin{aligned} f(a_2) &\equiv c_0 + c_1(a_1 + mp) + c_2(a_1^2 + 2a_1mp) + \cdots + c_n(a_1^n + na_1^{n-1}mp) \\ &\equiv (c_0 + c_1a_1 + \cdots c_na_n) + (c_1(mp) + 2c_2(mp)a_1 + \cdots nc_n(mp)a_1^{n-1}) \\ & \equiv f(a_1) + mpf’(a_1) \pmod{p^2} \end{aligned} Now, we know that a2a2(modp)a_2 \equiv a_2 \pmod{p} since a2=mp+a1a_2 = mp + a_1. If a2a_2 is a root modulo p2p^2, it implies that f(a2)0(modp2)f(a_2) \equiv 0 \pmod{p^2}. Therefore, f(a1)+mpf(a1)0(modp2)f(a_1) + mpf’(a_1) \equiv 0 \pmod{p^2}. Since f(a1)0(modp)f(a_1) \equiv 0 \pmod{p}, let f(a1)=kpf(a_1) = kp. So, we must have kp+mpf(a1)0(modp)kp + mpf’(a_1)\equiv 0 \pmod{p}. So, mk(f(a1))1(modp)m \equiv -k(f’(a_1))^{-1} \pmod{p}. We can do this step since f(a1)≢0(modp)f’(a_1) \not \equiv 0 \pmod{p}. Thus, we have found a unique mm such that f(a1+mp)0(modp2)f(a_1 + mp) \equiv 0 \pmod{p^2}. In other words, given a value a1a_1, we have proved that there exists a unique a2a_2 such that f(a2)0(modp2)f(a_2) \equiv 0 \pmod{p^2}.
 
Now, suppose we are given some aka_k such that f(ak)0(modpk)f(a_k) \equiv 0 \pmod{p^k}. We will prove that there exists a unique ak+1a_{k+1} such that f(ak+1)0(modpk+1)f(a_{k+1}) \equiv 0 \pmod{p^{k+1}} and ak+1ak(modpk)a_{k+1} \equiv a_k \pmod{p^k}.
 
Let f(x)=c0++cnxnf(x) = c_0 + \cdots + c_nx^n. Then, f(ak)c0+c1ak+cnann0(modpk)f(a_k) \equiv c_0 + c_1a_k + \cdots c_na_n^n \equiv0 \pmod{p^k} Now, since ak+1ak(modpk)a_{k+1} \equiv a_k \pmod{p^k}, we have ak+1pkm+aka_{k+1} \equiv p^km + a_k where 0<m<p0<m<p. Therefore, f(ak+1)=c0+c1ak+1++cnak+1n=c0+c1(ak+pkm)++cn(ak+pkm)n=c0+c1(ak+pkm)++cn(akn+nakn1mp+(mp)n)\begin{aligned} f(a_{k+1}) &= c_0 + c_1a_{k+1} + \cdots + c_na_{k+1}^{n} \\ &= c_0 + c_1(a_k + p^km) + \cdots + c_n(a_k + p^km)^n \\ &= c_0 + c_1(a_k + p^km) + \cdots + c_n(a_k^n + na_k^{n-1}mp + \cdots (mp)^n) \end{aligned} If we take mod pk+1p^{k+1} on both sides, all terms apart from the pkp^k and constant terms cancel out since 2k>k+12k > \geq k+1 for all positive kk. Thus, f(ak+1)c0+c1(ak+pkm)+c2(ak2+2akpkm)++cn(akn+nakn1pkm)(c0+c1ak+cnakn)+pkm(c1+2c2ak+ncnakn1)f(ak)+pkmf(ak)0(modpk+1)\begin{aligned} f(a_{k+1}) &\equiv c_0 + c_1(a_k + p^km) + c_2(a_k^2 + 2a_kp^km) + \cdots + c_n(a_k^{n} + na_k^{n-1}p^km) \\ &\equiv (c_0 + c_1a_k + \cdots c_na_k^n) + p^km(c_1 + 2c_2a_k + \cdots nc_na_k^{n-1}) \\ &\equiv f(a_k) + p^kmf’(a_k) \equiv 0 \pmod{p^{k+1}} \end{aligned} Since f(ak)0(modpk)f(a_k) \equiv 0 \pmod{p^k}, we have f(ak)=pktf(a_k) = p^kt. So, pkt+pkmf(ak)0(modpk+1)p^kt + p^kmf’(a_k) \equiv 0 \pmod{p^{k+1}}. Hence, t+mf(ak)0(modp)-t + mf’(a_k) \equiv 0 \pmod{p} Hence, mt(f(ak))1(modp)m \equiv t(f’(a_k))^{-1} \pmod{p}. Since f(ak)≢0(modpk)f’(a_k) \not \equiv 0 \pmod{p^k}, its inverse exists and is unique. Thus, there exists a unique mm, given as above, such that ak+1=mpk+aka_{k+1} = mp^k + a_k is a root of f(x)f(x) in Z/pk+1Z\mathbb{Z}/p^{k+1}\mathbb{Z} and ak+1ak(modpk)a_{k+1} \equiv a_k \pmod{p^{k}}.
 
We are given the value a1a_1 that is a root of f(x)f(x) in Z/pZ\mathbb{Z}/p\mathbb{Z}. Thus, we can find the corresponding value of a2a_2 and hence a3a_3 and so on. Thus, by induction, there exists a unique sequence a=(a1,a2,)a=(a_1, a_2, \dots) that satisfies f(ak)0(modpk)f(a_k) \equiv 0 \pmod{p^k} and ak+1ak(modpk)a_{k+1} \equiv a_k \pmod{p^k}. By definition of a pp-adic number, aZpa\in \mathbb{Z_p}. Moreover, since f(ak)0(modpk)f(a_k) \equiv 0 \pmod{p^k} it follows that f(a)=0f(a) = 0 in Zp\mathbb{Z_p}. Thus, there exists a unique pp-adic number aZpa\in \mathbb{Z_p} such that f(a)=0f(a) = 0 given a1a_1. ◻

The pp-adic Numbers Qp\mathbb{Q_p} #

As an Extension of Zp\mathbb{Z_p} #

Definition 4. We define Qp=Zp[1p]\mathbb{Q_p} = \mathbb{Z_p}[\frac{1}{p}].

For example, (1,4,13,)1+(3,3,3,)3+(2,5,14,)9Q3\frac{(1, 4, 13, \dots)}{1} + \frac{(3, 3, 3, \dots)}{3} + \frac{(2, 5, 14, \dots)}{9}\in\mathbb{Q_3} Note that the above is NOT equal to (1,4,13,)+(1,1,1,)+(29,59,139,)\left(1, 4, 13, \dots\right) + \left(1, 1, 1, \dots\right) + \left(\frac{2}{9}, \frac{5}{9}, \frac{13}{9}, \dots\right) The divided by sybmol is merely used as a notation and does not translate to the above. More generally, any element of Qp\mathbb{Q_p} is: a0+a1p++akpka_0 + \frac{a_1}{p} + \cdots + \frac{a_k}{p^k} By taking pkp^k as the common denominator, the above can be rewritten as: a0pk+a1pk1++akpk=apk\frac{a_0p^k + a_1p^{k-1} + \cdots + a_k}{p^k} = \frac{a}{p^k} where ai,aZpa_i, a \in \mathbb{Z_p}. Moreover, we have that two pp-adic numbers apk\frac{a}{p^k} and bpm\frac{b}{p^m} are equal if and only if apm=bpkap^m = bp^k.

Proposition 10. Qp\mathbb{Q_p} is a field. Moreover, QQp\mathbb{Q}\in \mathbb{Q_p}.

Proof. Consider some pp-adic number α=apk\alpha = \frac{a}{p^k}. We will prove that there exists some pp-adic number β=bpm\beta = \frac{b}{p^m} such that αβ=1\alpha\beta = 1 for every αQp\alpha \in \mathbb{Q_p} that is non-zero, i.e. a0a\neq 0.
 
Firstly, we know that since aZpa\in \mathbb{Z_p}, we have that there exists some aa’ that is a unit in Zp\mathbb{Z_p} and n1n_1 a non-negative integer so that a=apn1a = a’p^{n_1}. Let βQp\beta\in \mathbb{Q_p} be such that β=b1=(a)1pkn11\beta = \frac{b}{1} = \frac{(a’)^{-1}p^{k-n_1}}{1} The inverse of aa’ exists since aa’ is a unit as defined. Then, αβ=apn1pk(a)1pkn11=pkpk\alpha\beta = \frac{a’p^{n_1}}{p^k}\cdot \frac{(a’)^{-1}p^{k-n_1}}{1} = \frac{p^k}{p^k} Now since a/pk=b/pm    apm=bpka/p^k = b/p^m \iff ap^m = bp^k, it follows that the above equals 11 if and only if pk=pkp^k=p^k. Thus, αβ=1\alpha\beta = 1. Thus, we have found a βQp\beta\in \mathbb{Q_p} given αQp{0}\alpha \in \mathbb{Q_p}\setminus\lbrace 0 \rbrace such that αβ=1\alpha\beta = 1. Hence, Qp\mathbb{Q_p} is a field.
 
We will now prove that QQp\mathbb{Q}\in \mathbb{Q_p}. Consider an element of Q\mathbb{Q}, say a/ba/b. Let a=pk1aa = p^{k_1}a’ and b=pk2bb = p^{k_2}b’ where aa’ and bb’ are co-prime with pp. Thus, ab=pk1abpk2\frac{a}{b} = p^{k_1}\frac{\frac{a’}{b’}}{p^{{k_2}}} where αZp\alpha \in \mathbb{Z_p}. Thus, any rational number is also a pp-adic number. Hence Qp\mathbb{Q_p} contains Q\mathbb{Q}. ◻

Qp\mathbb{Q_p} as Completion of Q\mathbb{Q} #

We will first define metric spaces.

Definition 5. We call a set XX along with a function (metric) d ⁣:X×XRd\colon X\times X \to \mathbb{R}, (X,d)(X, d), a metric space if the following properties are satisfied for any x,y,zXx, y, z \in X:

  1. d(x,x)=0d(x, x) = 0

  2. If xyx \neq y then d(x,y)>0d(x, y) > 0

  3. d(x,z)d(x,y)+d(y,z)d(x, z) \leq d(x, y) + d(y, z)

  4. d(x,y)=d(y,x)d(x, y) = d(y, x)

For example, (Q,)(\mathbb{Q}, |\cdot|) is a metric space.

Definition 6. For a metric space (X,d)(X, d), we define a Cauchy sequence of XX wrt dd as follows: A sequence (xn)n0(x_n)_{n\geq0} where xnXx_n \in X is called a Cauchy sequence if for all ϵR+\epsilon\in \mathbb{R}^+, there exists some NϵNN_\epsilon \in \mathbb{N} such that for all integers m,nNm, n\geq N, we have d(xn,xm)<ϵd(x_n, x_m) < \epsilon.

In simple terms, any sequence that should converge under the given metric is called a Cauchy sequence. We now define what a completion with respect to a metric is.

Definition 7. Let S(X,d)S_{(X, d)} be the set of Cauchy sequences in XX with respect to the metric dd. Then, given two elements xn,ynx_n, y_n of S(X,d)S_{(X, d)}, we say that xnx_n is equivalent to yny_n iff for every real number ϵ\epsilon, there exists a natural number NϵN_\epsilon such that for all integers nNn\geq N, we have d(xn,yn)<ϵd(x_n, y_n)<\epsilon. We denote this equivalence by xnynx_n \sim y_n.

Definition 8. We define the completion of XX with respect to dd as the set S(X,d)/S_{(X, d)}/\sim.

In more intuitive terms, the completion of a metric space XX with respect to dd is the set of limits of the Cauchy sequences in XX with respect to dd.
 
As an example, R\mathbb{R} is a completion of Q\mathbb{Q} with respect to the usual metric, the absolute value.
 
Now, we will define Qp\mathbb{Q_p} as a completion of Q\mathbb{Q}. In order to do so, we must first define the metric on Q\mathbb{Q} that yields Qp\mathbb{Q_p}.

Definition 9. Let pp be a prime and aa be an integer. Then, we define vp(a)v_p(a) to be the largest nn such that pnap^n|a when a0a\neq 0 and to be \infty when a=0a=0. Now, we extend this definition to Q\mathbb{Q}. Let q=a/bQq = a/b\in Q where aa and bb are integers where b0b\neq 0. Then, we define vp(q)=vp(a)vp(b)v_p(q) = v_p(a)-v_p(b).

Lemma 1. The above definition of vp()v_p() is well-defined. That is, if a/b=c/d=qa/b = c/d = q, then we have vp(a/b)=vp(c/d)v_p(a/b) = v_p(c/d).

Proof. Suppose a/b=c/da/b = c/d such that vp(a/b)=vp(c/d)+mv_p(a/b) = v_p(c/d)+m. Then, vp(a)vp(b)=vp(c)vp(d)+mv_p(a) -v_p(b) = v_p(c) - v_p(d)+m. Let a=pk1a,b=pk2b,c=pk3c,d=pk4da = p^{k_1}a’, b = p^{k_2}b’, c = p^{k_3}c’, d = p^{k_4}d’ with each of a,b,c,da’, b’, c’, d’ relatively prime to pp. Therefore, k1k2k3k4+mk_1 - k_2 k_3 - k_4+m. Moreover, we have a/b=pk1k2a/ba/b = p^{k_1-k_2}a’/b’ and c/d=pk3k4c/dc/d = p^{k_3-k_4}c’/d’. Since a/b=c/da/b = c/d, we have pk1k2ab=pk3k4cdp^{k_1-k_2}\frac{a’}{b’} = p^{k_3-k_4}\frac{c’}{d’} We can rewrite this as pk1k2ad=pk3k4bcp^{k_1-k_2}a’d’ = p^{k_3-k_4}b’c’ Now, we assumed that k1k2=k3k4+mk_1-k_2 = k_3-k_4+m. Let k3k4=nk_3-k_4 = n. So, pn+mad=pnbcp^{n+m}a’d’ = p^nb’c’ which implies that pmad=bcp^ma’d’ = b’c’. Taking the vpv_p of both sides, we have m=0m = 0. Thus ◻

Definition 10. We define the pp-adic absolute value to be as follows. Let qQq\in \mathbb{Q}. Then, we define qp=pvp(q)\lvert q \rvert_p = p^{-v_p(q)} when q0q\neq0 and 00 when q=0q=0.

We will first explore some properties of vp(a)v_p(a) which will help us find a metric on Q\mathbb{Q} that gives Qp\mathbb{Q_p}.

Proposition 11. For any a,bZa, b \in \mathbb{Z}, the following are true:

  1. vp(ab)=vp(a)+vp(b)v_p(ab) = v_p(a) + v_p(b)

  2. vp(a+b)min(vp(a),vp(b))v_p(a+b) \geq \min(v_p(a), v_p(b))

  3. If vp(a)vp(b)v_p(a) \neq v_p(b), then vp(a+b)=min(vp(a),vp(b))v_p(a+b) = \min(v_p(a), v_p(b))

Proof. Let vp(a)=k1v_p(a) = k_1 and vp(b)=k2v_p(b) = k_2. WLOG, assume that k1k2k_1 \geq k_2. Then, clearly a=pk1aa = p^{k_1}a’ and b=pk2bb = p^{k_2}b’ where (a,p)=1(a’, p) = 1 and (b,p)=1(b’, p) = 1. Then:

  1. We have ab=abpk1+k2ab = a’b’p^{k_1+k_2} where (ab,p)=1(a’b’, p) = 1 since each of aa’ and bb’ is co-prime to pp. Hence, vp(ab)=k1+k2=vp(a)+vp(b)v_p(ab) = k_1 + k_2 = v_p(a) + v_p(b).

  2. We have a+b=apk1+bpk2a+b = a’p^{k_1} + b’p^{k_2}. Since k1k2k_1 \geq k_2, we can write a+b=pk2(apk1k2+b)a+b = p^{k_2}(a’p^{k_1-k_2} + b’) Hence, vp(a+b)=k2+vp(apk1k2+b)v_p(a+b) = k_2 + v_p(a’p^{k_1-k_2} + b’) Clearly, the above is at least k2k_2 since vp(apk1k2+b)0v_p(a’p^{k_1-k_2} + b’) \geq 0. Hence, vp(a+b)k2=min(vp(a),vp(b))v_p(a+b)\geq k_2 = \min(v_p(a), v_p(b)) since we assumed that vp(a)vp(b)v_p(a) \leq v_p(b).

  3. Since we have that vp(a)vp(b)v_p(a) \neq v_p(b), we know k1>k2k_1 > k_2. Thus, apk1k2+b0+bb(modp)a’p^{k_1-k_2} + b’ \equiv 0 + b’ \equiv b’ \pmod{p}. Since bb’ is co-prime with pp, we have apk1k2+b≢0(modp)a’p^{k_1-k_2} + b’ \not \equiv 0 \pmod{p}. Hence, p(pk1k2a+b)p\nmid(p^{k_1-k_2}a’ + b’). Thus, vp(apk1k2+b)=0v_p(a’p^{k_1-k_2} + b’) = 0. Hence, vp(a+b)=k2=vp(b)=min(vp(a),vp(b))v_p(a+b) = k_2 = v_p(b) = \min(v_p(a), v_p(b)) since we assume that vp(a)vp(b)v_p(a) \leq v_p(b).

 ◻

Proposition 12. The above properties all hold for any r,sQr, s \in \mathbb{Q}.

Proof. Let r=a1/b1r = a_1/b_1 and s=a2/b2s = a_2/b_2 where ai,bia_i, b_i are integers with bi0b_i \neq 0. We know by definition that vp(r)=vp(a1)vp(b1)v_p(r) = v_p(a_1)-v_p(b_1) and similarly vp(s)=vp(a2)vp(b2)v_p(s) = v_p(a_2) - v_p(b_2). Hence:

  1. We have vp(rs)=vp(a1a2b1b2)=vp(a1a2)vp(b1b2)v_p(rs) = v_p\left(\frac{a_1a_2}{b_1b_2}\right) = v_p(a_1a_2)-v_p(b_1b_2) Since aiZa_i\in \mathbb{Z} we have vp(a1a2)=vp(a1)+vp(a2)v_p(a_1a_2) = v_p(a_1) + v_p(a_2) and similarly vp(b1b2)=vp(b1)+vp(b2)v_p(b_1b_2) = v_p(b_1) + v_p(b_2). Thus, vp(rs)=(vp(a1)vp(b1))+(vp(a2)vp(b2))=vp(r)+vp(s)v_p(rs) = (v_p(a_1)-v_p(b_1)) + (v_p(a_2)-v_p(b_2)) = v_p(r)+v_p(s)

  2. We can write vp(r+s)v_p(r+s) as vp(a1b1+a2b2)=vp(a1b2+a2b1b1b2)=vp(a1b2+a2b1)vp(b1b2)v_p\left(\frac{a_1}{b_1}+\frac{a_2}{b_2}\right)= v_p\left(\frac{a_1b_2+a_2b_1}{b_1b_2}\right)=v_p(a_1b_2+a_2b_1)-v_p(b_1b_2) Now, a1b2a_1b_2 and a2b1a_2b_1 are integers. So, vp(a1b2+a2b1)min(vp(a1b2),vp(a2b1))v_p\left(a_1b_2 + a_2b_1\right)\leq \min(v_p(a_1b_2), v_p(a_2b_1)) Thus, vp(r+s)min(vp(a1)+vp(b2),vp(a2)+vp(b1))(vp(b1)+vp(b2))=min(vp(a1)vp(b1),vp(a2)vp(b2))=min(vp(r),vp(s)) \begin{aligned} v_p(r+s) &\geq \min(v_p(a_1)+v_p(b_2), v_p(a_2)+v_p(b_1)) - (v_p(b_1)+v_p(b_2)) \\ &=\min(v_p(a_1)-v_p(b_1), v_p(a_2)-v_p(b_2)) = \min(v_p(r), v_p(s)) \end{aligned} We can legally perform the above step since vp(b1b2)v_p(b_1b_2) is a constant.

  3. If vp(a1b2)=vp(a2b1)v_p(a_1b_2) = v_p(a_2b_1), then we have vp(a1)+vp(b2)=vp(a2)vp(b1).v_p(a_1) + v_p(b_2) = v_p(a_2) v_p(b_1). Thus, by rearranging the terms, we have vp(r)=vp(s)v_p(r) = v_p(s). Thus, if vp(r)vp(s)v_p(r) \neq v_p(s), it follows that vp(a1b2)vp(a2b1).v_p(a_1b_2) \neq v_p(a_2b_1). Hence, vp(a1b2+a2b1)=min(vp(a1b2),vp(a2b1)).v_p(a_1b_2 + a_2b_1) = \min(v_p(a_1b_2), v_p(a_2b_1)). It follows directly that vp(r+s)=min(vp(r),vp(s)).v_p(r+s) = \min(v_p(r), v_p(s)).

We have thus extended all the properties that hold for integers under vpv_p to rationals. ◻

We will now use these properties to prove some properties regarding p\left\lvert \cdot \right\rvert_p, which will in turn prove that (Q,p)(\mathbb{Q}, \left\lvert \cdot \right\rvert_p) is a metric space.

Proposition 13. The following properties are true regarding p\left\lvert \cdot \right\rvert_p:

  1. qp0\left\lvert q \right\rvert_p \geq 0 and equality holds iff q=0q=0

  2. q+rpmax(qp,rp)\left\lvert q+r \right\rvert_p \leq \max(\left\lvert q \right\rvert_p, \left\lvert r \right\rvert_p)

  3. qrp=qprp\left\lvert qr \right\rvert_p = \left\lvert q \right\rvert_p \left\lvert r \right\rvert_p

Proof.

  1. We have qp=pvp(q)\left\lvert q \right\rvert_p = p^{-v_p(q)}. Clearly this is at least 00 for all vp(q)v_p(q) since vp(q)Qv_p(q) \in \mathbb{Q}. We have pvp(q)=0p^{-v_p(q)} = 0 iff vp(q)=v_p(q) = \infty which happens only when q=0q=0.

  2. We have q+rp=pvp(q+r)pmin(vp(q),vp(r))=pmax(vp(q)),vp(r)\left\lvert q+r \right\rvert_p = p^{-v_p(q+r)} \leq p^{-\min(v_p(q), v_p(r))} = p^{\max(-v_p(q)), -v_p(r)} since vp(q+r)min(vp(q),vp(r)))v_p(q+r) \geq \min(v_p(q), v_p(r))). Thus, q+rpmax(qp,rp)\left\lvert q+r \right\rvert_p \leq \max(\left\lvert q \right\rvert_p, \left\lvert r \right\rvert_p) Clearly, equality holds when qprp\left\lvert q \right\rvert_p \neq \left\lvert r \right\rvert_p, which follows from the property of vpv_p.

  3. We have qrp=pvp(qr)=pvpqvp(r)=pvp(q)pvp(r)=qprp\left\lvert qr \right\rvert_p = p^{-v_p(qr)} = p^{-v_p{q} - v_p(r)} = p^{-v_p(q)}p^{-v_p(r)} = \left\lvert q \right\rvert_p\left\lvert r \right\rvert_p

 ◻

With these properties, we may now define a new metric on Q\mathbb{Q}. Consider the metric dpd_p which is defined as follows:

Definition 11. We define the function dp ⁣:Q×QRd_p \colon Q\times \mathbb{Q}\to \mathbb{R} as follows: dp(x,y)=xypd_p(x, y) = \left\lvert x-y \right\rvert_p.

Proposition 14. Q\mathbb{Q} is a metric space over dpd_p.

Proof. In order to prove this, we must prove all properties listed in the definition of a metric space.

  1. dp(x,x)=xxp=0p=0d_p(x, x) = \left\lvert x-x \right\rvert_p = \left\lvert 0 \right\rvert_p = 0

  2. dp(x,y)=xyp>0d_p(x, y) = \left\lvert x-y \right\rvert_p > 0 when xyx\neq y by the previous proposition.

  3. dp(x,z)=xzp=(xy)+(yz)pd_p(x, z) = \left\lvert x-z \right\rvert_p = \left\lvert (x-y)+(y-z) \right\rvert_p. By property 22 above, (xy)+(yz)pmax(xyp,yzp)xyp+yzp\left\lvert (x-y)+(y-z) \right\rvert_p \leq \max(\left\lvert x-y \right\rvert_p, \left\lvert y-z \right\rvert_p) \leq \left\lvert x-y \right\rvert_p + \left\lvert y-z \right\rvert_p since qp0\left\lvert q \right\rvert_p \geq 0. Hence, dp(x,z)dp(x,y)+dp(y,z)d_p(x, z) \leq d_p(x, y) + d_p(y, z).

  4. dp(x,y)=xyp=yxp=dp(y,x)d_p(x, y) = \left\lvert x-y \right\rvert_p = \left\lvert y-x \right\rvert_p = d_p(y, x)

Thus, Q\mathbb{Q} is a metric space over dpd_p. ◻

We can thus finally define Qp\mathbb{Q_p} in terms of Q\mathbb{Q} as follows:

Definition 12. We define Qp\mathbb{Q_p} to be the completion of Q\mathbb{Q} with respect to dpd_p. Thus, Qp={[an]anSp}\mathbb{Q_p} = \lbrace [a_n]\mid a_n \in S_p \rbrace where SpS_p denotes the set of Cauchy sequences in Q\mathbb{Q} with respect to dpd_p.

Operations in Qp\mathbb{Q_p} #

Now, we know that Q\mathbb{Q} is a field with operations (+,)(+, \cdot). We will now try to find a pair of binary operations (+,)(+, \cdot) on Qp\mathbb{Q_p} such that (Qp,+,)(\mathbb{Q_p}, +, \cdot) is a field.
 
Let αQp\alpha \in \mathbb{Q_p}. Therefore, we can write α=[a]\alpha = [a] for some aSpa\in S_p by definition of QpQ_p. Similarly, let β=[b]Qp\beta = [b] \in \mathbb{Q_p}. We define the addition of two elements in the p-adics as α+β=[a]+[b]=[a+b]\alpha + \beta = [a] + [b] = [a+b] and their product as αβ=[a][b]=[ab]\alpha\cdot \beta = [a]\cdot[b] = [a\cdot b] However, we have not yet defined what addition and multiplication are in SpS_p (aa and bb are elements of SpS_p). We define the sum of two sequences a=(ai)a=(a_i) and b=(bi)b = (b_i) to be a+b=(ai+bi)a+b = (a_i+b_i), i.e. the termwise sum of the terms of the sequence. Similarly, we define their product to be ab=(aibi)a\cdot b = (a_ib_i). Since both aa and bb converge, it follows that both a+ba+b as well as abab converge. Thus, SpS_p is closed under multiplication.

Definition 13. Let SS be the set of all Cauchy sequences of Q\mathbb{Q}, we define a set of equivalence classes in SS such that S/={[a]aS}S/\sim \hspace{0.5 mm}= \lbrace [a] | a \in S \rbrace [a]={(ai)y(yi)Slimiaiyip=0}[a] = \lbrace (a_i) \sim y | \exists (y_i) \in S \lim\limits_{i \to \infty}{|a_i - y_i|_p} = 0 \rbrace

Lemma 2. The set S/S/\sim has a well-defined addition and multiplication [a]+[b][a+b][a][b][ab] \begin{aligned} [a] + [b] \sim [a + b] \\ [a][b] \sim [ab] \end{aligned}

Proof. We need to show that, (ai+bi)(ai+bi)p=0|(a_i + b_i) - (a^{’}_i + b^{’}_i)|_p = 0. To prove this, we can observe the following,

(ai+bi)(ai+bi)p=aiai+bibip|(a_i + b_i) - (a^{’}_i + b^{’}_i)|_p = |a_i - a^{’}_i + b_i - b^{’}_i|_p

It follows that, aiai+bibipaiaip+bibip|a_i - a^{’}_i + b_i - b^{’}_i|_p \le |a_i - a^{’}_i|_p + |b_i - b^{’}_i|_p the RHS converges to 00 as ii \rightarrow \infty. For multiplication, we have, aibiaibi=aibiaibi+aibi+aibiai(bibi)+bi(aiai)pai(bibi)p+bi(aiai)pai(bibi)p+bi(aiai)p=aip(bibi)p+bip(aiai)p\begin{aligned} |a_ib_i - a^{’}_ib^{’}_i| = |a_ib_i - a_ib^{’}_i + a_ib^{’}_i + a^{’}_ib^{’}_i| \\ |a_i(b_i - b^{’}_i) + b^{’}_i(a_i - a^{’}_i)|_p \le |a_i(b_i - b^{’}_i)|_p + |b^{’}_i(a_i - a^{’}_i)|_p \\ |a_i(b_i - b^{’}_i)|_p + |b^{’}_i(a_i - a^{’}_i)|_p = |a_i|_p|(b_i - b^{’}_i)|_p + |b^{’}_i|_p|(a_i - a^{’}_i)|_p \end{aligned} Since (ai)(a_i) and (bi)(b^{’}_i) are bounded, then we can proceed: aiplimi(bibi)p+biplimi(aiai)p=ϵ0+ϵ0=0|a_i|_p\lim\limits_{i \to \infty}|(b_i - b^{’}_i)|_p + |b^{’}_i|_p|\lim\limits_{i \to \infty}(a_i - a^{’}_i)|_p = \epsilon \cdot 0 + \epsilon \cdot 0 = 0 ◻

Now, we will prove that Qp\mathbb{Q_p} is a field.

Definition 14. If λ\lambda is an element of Qp\mathbb{Q_p} and (xn)S/(x_n) \in S/\sim is any Cauchy sequence representing λ\lambda, we define λp=limnxnp|\lambda|_p = \lim\limits_{n \to \infty}|x_n|_p

Proposition 15. Qp=S/\mathbb{Q_p} = S/\sim is a field

Proof. Let (1/xi)(1/x_i) denote the multiplicative inverse of (xi)(x_i) in Qp\mathbb{Q_p}. To know that 1/xn1/x_n is Cauchy, observe the following 1xi1xjp=xjxixjxip\left |\frac{1}{x_i} - \frac{1}{x_j} \right|_p = \left|\frac{x_j - x_i}{x_jx_i}\right|_p Since (xn)(x_n) is Cauchy, we have that (xjxi)(x_j - x_i) is bounded by ϵ\epsilon. It follows that for a fix large NN and for every i,j>Ni, j > N, we have that for every ϵ>0\epsilon > 0, the following holds 1xi1xjp<ϵ\left |\frac{1}{x_i} - \frac{1}{x_j} \right|_p < \epsilon ◻

Sequences and Series in Qp\mathbb{Q_p} #

Till now, we only talked about sequences in Q\mathbb{Q}, and used them to define the system Qp\mathbb{Q_p}. We will now talk about sequences in Qp\mathbb{Q_p} itself. They can be thought of as sequences of sequences, since each element of Qp\mathbb{Q_p} is itself a sequence of elements in Q\mathbb{Q}. We will now extend the pp-adic absolute value to even pp-adic numbers. There are two things we need to take care of here. Firstly, we only know what qp\left\lvert q \right\rvert_p where qq is a rational number is, not a sequence of rational numbers. So, we need to define the notion of pp-adic absolute value for Cauchy sequences. Next, [a][a] is not a single sequence. It is a set of sequences. Therefore, we need to show that for any sequence in [a][a], the result of [a]p\left\lvert [a] \right\rvert_p is the same. In other words, if xyx\sim y are two sequences, then we must show that xp=yp\left\lvert x \right\rvert_p = \left\lvert y \right\rvert_p.

Definition 15. Let (xn)(x_n) be a Cauchy sequence. Then, the sequence (xnp)(\left\lvert x_n \right\rvert_p) converges to some real number, say yRy \in \mathbb{R}. Then, we define (xn)p\left\lvert (x_n) \right\rvert_p to be yy. In other words, (xn)p=limnxnp\left\lvert (x_n) \right\rvert_p = \lim_{n\to\infty}\left\lvert x_n \right\rvert_p.

Proof. We will prove that if (xn)(x_n) is a Cauchy sequence, then (xnp)(\left\lvert x_n \right\rvert_p) converges in R\mathbb{R}. So, we must prove that for all ϵR+\epsilon \in \mathbb{R}^+, we have that there exists some NϵNN_\epsilon \in \mathbb{N} such that for all m,nNϵm,n\geq N_\epsilon we have xnpxmp<ϵ\lvert \left\lvert x_n \right\rvert_p-\left\lvert x_m \right\rvert_p \rvert<\epsilon By the definition of a Cauchy sequence with respect to p\left\lvert \cdot \right\rvert_p, we have that for every ϵR+\epsilon \in \mathbb{R}^+, there exists some natural number NϵN_\epsilon such that for all m,nNϵm,n\geq N_\epsilon, we have xnxmp<ϵ\left\lvert x_n - x_m \right\rvert_p < \epsilon. Let us now keep ϵ\epsilon and NϵN_\epsilon fixed. Now, consider some integers m,nNϵm, n \geq N_\epsilon such that xnpxmp\left\lvert x_n \right\rvert_p \neq \left\lvert x_m \right\rvert_p. Then, we have xnxmp=max(xnp,xmp)<ϵ\left\lvert x_n - x_m \right\rvert_p = \max(\left\lvert x_n \right\rvert_p, \left\lvert x_m \right\rvert_p)< \epsilon Therefore, we have 0<xmp<xnp<ϵ0<\left\lvert x_m \right\rvert_p<\left\lvert x_n \right\rvert_p<\epsilon (WLOG). Thus, xnpxmp<ϵ\left\lvert x_n \right\rvert_p-\left\lvert x_m \right\rvert_p<\epsilon. In general, xnpxmp<ϵ\lvert \left\lvert x_n \right\rvert_p-\left\lvert x_m \right\rvert_p \rvert < \epsilon.
 
Now, if xnp=xmp\left\lvert x_n \right\rvert_p = \left\lvert x_m \right\rvert_p, this implies that xnpxmp=0<ϵ\lvert \left\lvert x_n \right\rvert_p-\left\lvert x_m \right\rvert_p \rvert = 0 < \epsilon since ϵR+\epsilon \in \mathbb{R}^+. Thus, for all m,nNϵm,n \geq N_\epsilon we have xmpxnp<ϵ\lvert \left\lvert x_m \right\rvert_p-\left\lvert x_n \right\rvert_p \rvert<\epsilon. This holds true for every value of ϵ\epsilon that is a positive real and its corresponding NϵN_\epsilon. Therefore, the sequence (xnp)(\left\lvert x_n \right\rvert_p) converges. ◻

We will now show that [a]p\left\lvert [a] \right\rvert_p is well defined where [a][a] is the equivalence class of a Cauchy sequence aa of Q\mathbb{Q}. Recall that this is equivalent to proving that if xnynx_n \sim y_n then xnp=ynp\left\lvert x_n \right\rvert_p = \left\lvert y_n \right\rvert_p.

Proposition 16. If (xn),(yn)Sp(x_n), (y_n) \in S_p are equivalent (where equivalence is as defined before), then (xn)p=(yn)p\left\lvert (x_n) \right\rvert_p = \left\lvert (y_n) \right\rvert_p. In other words, limnxnp=limnynp\lim_{n\to\infty}\left\lvert x_n \right\rvert_p = \lim_{n\to \infty}\left\lvert y_n \right\rvert_p

Proof. Given that (xn)(yn)(x_n)\sim (y_n), by definition, we have that for all ϵR+\epsilon \in \mathbb{R}^+, there exists some NϵNN_\epsilon \in \mathbb{N} such that for all nNϵn\geq N_\epsilon, we have xnynp<ϵ\left\lvert x_n-y_n \right\rvert_p<\epsilon. In other words, if (xn)(yn)(x_n) \sim (y_n), then limnxnynp=0\lim_{n\to\infty}\left\lvert x_n-y_n \right\rvert_p = 0 Let us assume for the sake of contradiction that limnxnplimnynp\lim_{n\to\infty}{\left\lvert x_n \right\rvert_p} \neq \lim_{n\to\infty}\left\lvert y_n \right\rvert_p. Also, let us assume WLOG that limnxnp>limnynp\lim_{n\to\infty}{\left\lvert x_n \right\rvert_p} > \lim_{n\to\infty}\left\lvert y_n \right\rvert_p. Since xnynp=max(xnp,ynp)\left\lvert x_n - y_n \right\rvert_p = \max(\left\lvert x_n \right\rvert_p, \left\lvert y_n \right\rvert_p) when xnpynp\left\lvert x_n \right\rvert_p\neq \left\lvert y_n \right\rvert_p, it follows that limnxnynp=limnxnp=0\lim_{n\to\infty}\left\lvert x_n-y_n \right\rvert_p = \lim_{n\to\infty}\left\lvert x_n \right\rvert_p = 0 However, we assumed that limnynp<limnxnp=0\lim_{n\to \infty}\left\lvert y_n \right\rvert_p<\lim_{n\to\infty}\left\lvert x_n \right\rvert_p = 0. A contradiction since the absolute value is always positive. It follows that limnxnp=limnynp\lim_{n\to\infty}\left\lvert x_n \right\rvert_p = \lim_{n\to \infty}\left\lvert y_n \right\rvert_p. Hence, (xn)p=(yn)p\left\lvert (x_n) \right\rvert_p = \left\lvert (y_n) \right\rvert_p. ◻

From this, one may conclude that αp\left\lvert \alpha \right\rvert_p is well defined for all αQp\alpha \in \mathbb{Q_p}. Moreover, the same properties hold for αp\left\lvert \alpha \right\rvert_p for αQp\alpha \in \mathbb{Q_p} as in Q\mathbb{Q}, which can be easily seen from the fact that (xn)p=limnxnp\left\lvert (x_n) \right\rvert_p = \lim_{n\to\infty}\left\lvert x_n \right\rvert_p.
 
With the new extension of the pp-adic absolute value in the pp-adic numbers, we may now extend the notion of convergence to Qp\mathbb{Q_p} as well.

Definition 16. Consider a sequence (xn)(x_n) where xnQpx_n \in \mathbb{Q_p}. We say that (xn)(x_n) converges to a limit LQpL\in \mathbb{Q_p}, iff for all ϵR+\epsilon \in\mathbb{R}^+, there is some natural number NϵN_\epsilon such that for all nNϵn \geq N_\epsilon we have xnLp<ϵ\left\lvert x_n- L \right\rvert_p<\epsilon. Note that here, each xkx_k is itself a pp-adic number. So, (xn)(x_n) is essentially a sequence of sequences of rationals.

We now give the condition of convergence for a sequence in Qp\mathbb{Q_p}.

Definition 17. A sequence (an)(a_n) converges iff there exists an element of Qp\mathbb{Q_p} that it converges to.

Proposition 17. In Qp\mathbb{Q_p}, a sequence (xn)(x_n) converges then (xnp)(\left\lvert x_n \right\rvert_p) converges in R\mathbb{R}. Moreover, limnxnp=limnxnp\lim_{n\to\infty}{\left\lvert x_n \right\rvert_p} = \left\lvert \lim_{n\to\infty}x_n \right\rvert_p

Proof. Suppose (xn)(x_n) converges to LQpL\in \mathbb{Q_p}. Then, by definition, we have that ϵR+NϵN st nNϵ,xnLp<ϵ\forall\epsilon \in \mathbb{R}^+\exists N_\epsilon \in \mathbb{N}\text{ st } \forall n \geq N_\epsilon, \left\lvert x_n-L \right\rvert_p<\epsilon Now, xnLp=max(xnp,Lp)<ϵ\left\lvert x_n-L \right\rvert_p = \max(\left\lvert x_n \right\rvert_p, \left\lvert L \right\rvert_p)<\epsilon if xnpLp\left\lvert x_n \right\rvert_p\neq \left\lvert L \right\rvert_p. Therefore, we have 0xnp<ϵ0 \leq \left\lvert x_n \right\rvert_p<\epsilon and 0Lp<ϵ0 \leq \left\lvert L \right\rvert_p < \epsilon for all real numbers ϵ>0\epsilon > 0. Thus, we have xnpLp<ϵ\lvert \left\lvert x_n \right\rvert_p-\left\lvert L \right\rvert_p \rvert<\epsilon. If xnp=Lp\left\lvert x_n \right\rvert_p= \left\lvert L \right\rvert_p the previous statement still holds. Therefore, limnxnp=Lp=limnxnp\lim_{n\to\infty}\left\lvert x_n \right\rvert_p = \left\lvert L \right\rvert_p=\left\lvert \lim_{n\to\infty}x_n \right\rvert_p which proves the desired result. ◻

Power series in Qp\mathbb{Q_p} #

Definition 18. A series i0ai\sum_{i\geq0}a_i is said to converge if and only if the sequence (Sn)(S_n) given by Sn=i=0naiS_n = \sum_{i=0}^na_i converges.

We will now give the condition of convergence for a series in Qp\mathbb{Q_p}. The condition turns out to be much simpler than that in R\mathbb{R}, in which there are multiple convergence tests for series.

Proposition 18. The series n0an\sum_{n\geq0}a_n converges in Qp\mathbb{Q_p}     \iff the sequence (an)n0(a_n)_{n\geq0} converges to 00 in Qp\mathbb{Q_p}.

Proof. Consider the sequence (Sn)(S_n) given by Sn=i=0naiS_n = \sum_{i=0}^na_i. We need to show that (Sn)(S_n) converges if and only if (an)(a_n) converges.
 
Suppose (Sn)(S_n) converges. We know that, by definition, (Sn)(S_n) converges if and only if for all ϵR+\epsilon \in \mathbb{R}^+ we have that there exists some natural number NϵN_\epsilon such that for all m,nNϵm,n\geq N_\epsilon we have SmSnp<ϵ\left\lvert S_m-S_n \right\rvert_p < \epsilon. Consider some nNϵn\geq N_\epsilon. Then, we clearly have that Sn+1Snp<ϵ\left\lvert S_{n+1}-{S_n} \right\rvert_p < \epsilon Now, by definition of SnS_n, we have Sn+1Sn=an+1S_{n+1}-S_n = a_{n+1}. Thus, an+1p<ϵ\left\lvert a_{n+1} \right\rvert_p<\epsilon for all nNϵn\geq N_\epsilon. Therefore, in general, we have that for every ϵR+\epsilon \in \mathbb{R}^+, there is some Mϵ=Nϵ+1M_\epsilon = N_\epsilon + 1 such that for all nMϵn \geq M_\epsilon we have anp<ϵ\left\lvert a_n \right\rvert_p<\epsilon. It follows that (an)(a_n) converges to 00, by definition. Note that this direction of the proof holds in R\mathbb{R} as well. The opposite direction is what makes convergence in R\mathbb{R} more difficult.
 
Now, we will prove that if (an)(a_n) converges to 00, then (Sn)(S_n) converges. By definition, we have that for every ϵR+\epsilon \in \mathbb{R}^+, there exists some natural number NϵN_\epsilon such that for all nNϵn\geq N_\epsilon we have anp<ϵ\left\lvert a_n \right\rvert_p < \epsilon. Now, consider some integers m,nm, n such that m>nNϵm>n\geq N_\epsilon. We have SmSn=an+1++amS_m - S_n = a_{n+1} + \cdots + a_m Thus, SmSnp=an+1++ampmax(an+1,,am)\begin{aligned} \left\lvert S_m - S_n \right\rvert_p = \left\lvert a_{n+1} + \cdots + a_m \right\rvert_p \leq \max(a_{n+1}, \dots, a_m) \end{aligned} Since for all nNϵn\geq N_\epsilon, we have that an<ϵa_n < \epsilon, it follows that max(an+1,,am)<ϵ\max(a_{n+1}, \dots, a_m)<\epsilon. Hence SmSnp<ϵ\left\lvert S_m - S_n \right\rvert_p < \epsilon for all m,n>Nϵm, n > N_\epsilon.
 
Thus, in general, for all ϵR+\epsilon \in \mathbb{R}^+, we have that there exists some NϵNN_\epsilon \in \mathbb{N} such that for all m,nNϵm, n\geq N_\epsilon, SmSnp\left\lvert S_m - S_n \right\rvert_p. Hence, (Sn)(S_n) converges. ◻

Note that the above proof holds entirely because of the fact that a+bpmax(ap,bp)\left\lvert a+b \right\rvert_p\leq \max(\left\lvert a \right\rvert_p, \left\lvert b \right\rvert_p).

Radius of Convergence #

We define the radius of convergence of a power series n0anxn\sum_{n\geq0}a_nx^n to be the value rr so that the sequence anpcn\left\lvert a_n \right\rvert_pc^n converges to 00 for all c<rc<r and does not converge for c>rc>r. The following result is fundamental:

Proposition 19. The radius of convergence of n0anxn\sum_{n\geq0}a_nx^n is given by r=(limsupanp1/n)1r = \left(\lim\sup\left\lvert a_n \right\rvert_p^{1/n}\right)^{-1}

Discs #

Definition 19. For any aQpa\in \mathbb{Q_p} and rR+r\in \mathbb{R}^+, we define a closed disc of radius rr, centered at aa to be the set D(a;r):={zQp ⁣:zapr}D(a;r) \vcentcolon=\lbrace z\in \mathbb{Q_p}\colon \left\lvert z-a \right\rvert_p\leq r \rbrace and an open disc of radius rr, centered at aa to be the set D(a;r):={zQp ⁣:zap<r}D(a; r^-)\vcentcolon=\lbrace z\in \mathbb{Q_p} \colon \left\lvert z-a \right\rvert_p<r \rbrace.

Now, consider f(x)=n0anxnQp[[x]]f(x) = \sum_{n\geq 0}a_nx^n \in \mathbb{Q_p}[[x]], a power series, and suppose that its radius of convergence is rr. Therefore, we can define a function f ⁣:D(0;r)Qpf\colon D(0; r^-) \to \mathbb{Q_p} so that for any tD(0;r)t\in D(0; r^-), we have f(t)=limn(k=0naktk)f(t) = \lim_{n\to\infty}\left(\sum_{k=0}^n a_kt^k\right) Since tD(0;r)t\in D(0; r^-), the above sum indeed converges, and therefore, ff is well defined.
 
We now defined continuity in Qp\mathbb{Q_p}.

Definition 20. We say that a function f ⁣:SQpf\colon S\to \mathbb{Q_p} is continuous at a point xSx \in S if for all ϵR+\epsilon \in \mathbb{R}^+, there exists some positive real δ\delta such that xyp<δ\left\lvert x-y \right\rvert_p<\delta implies f(x)f(y)p<ϵ\left\lvert f(x)-f(y) \right\rvert_p<\epsilon.

Definition 21. We say that a function f ⁣:SQpf\colon S\to\mathbb{Q_p} is continuous, if it is continuous at every point in SS.

Proposition 20. Every function f ⁣:D(0;r)Qpf\colon D(0; r^-) \to \mathbb{Q_p} such that f(x)=limnk=0nakxkf(x) = \lim_{n\to\infty}\sum_{k=0}^n a_kx^k for all xD(0;r)x \in D(0; r^-) is continuous in Qp\mathbb{Q_p}.

Proof. We first prove a lemma:

Lemma 3. Let Sn=k=0nakS_n = \sum_{k=0}^na_k and let S=limnSnS = \lim_{n\to\infty}S_n. Then, Splimnmax(a1p,,anp)\left\lvert S \right\rvert_p \leq \lim_{n\to\infty}\max(\left\lvert a_1 \right\rvert_p, \dots, \left\lvert a_n \right\rvert_p). This can also be written as max(a1p,a2p,)\max(\left\lvert a_1 \right\rvert_p, \left\lvert a_2 \right\rvert_p, \dots). Notationally, we shall express this as max(akp)n0\max(\left\lvert a_k \right\rvert_p)_{n\geq0}.

Proof. We know that S=limnSnS = \lim_{n\to \infty}S_n Thus, Sp=limnSnp\left\lvert S \right\rvert_p = \left\lvert \lim_{n\to\infty}S_n \right\rvert_p As we saw earlier, the above is equal to limnSnp\lim_{n\to\infty}\left\lvert S_n \right\rvert_p, which is at most limnmax(a1p,,anp)\lim_{n\to\infty}\max(\left\lvert a_1 \right\rvert_p, \dots, \left\lvert a_n \right\rvert_p). ◻

Let xD(0;r)x\in D(0; r^-) be any point in D(0;r)D(0; r^-). Let yy be another point in D(0;r)D(0; r^-). Therefore, we have xp<r\left\lvert x \right\rvert_p<r and yp<r\left\lvert y \right\rvert_p < r. Consider some positive real ϵ\epsilon. Now, suppose that there is some δR+\delta \in \mathbb{R}^+ such that xyp<δ\left\lvert x-y \right\rvert_p<\delta. Now, we have f(x)f(y)p=limn(k=0nakxk)limn(k=0nakyk)p=limn(k=0nak(xkyk))p=limn((xy)k=0nak(xk1+xk2y++xyk2+yk1))p=xyplimn(k=0nak(xk1+xk2y++xyk2+yk1))p\begin{aligned} \left\lvert f(x)-f(y) \right\rvert_p &= \left\lvert \lim_{n\to\infty}\left(\sum_{k=0}^n a_kx^k\right)-\lim_{n\to\infty}\left(\sum_{k=0}^n a_ky^k\right) \right\rvert_p = \left\lvert \lim_{n\to\infty}\left(\sum_{k=0}^na_k(x^k-y^k)\right) \right\rvert_p \\ &= \left\lvert \lim_{n\to\infty}\left((x-y)\sum_{k=0}^na_k(x^{k-1}+x^{k-2}y+\cdots+ xy^{k-2}+y^{k-1})\right) \right\rvert_p \\ &= \left\lvert x-y \right\rvert_p\left\lvert \lim_{n\to\infty}\left(\sum_{k=0}^na_k(x^{k-1}+x^{k-2}y + \cdots + xy^{k-2}+ y^{k-1})\right) \right\rvert_p \end{aligned} By the Lemma, we may simplify the above to get: f(x)f(y)pxypmax(an(xn1+xn2y++xyn2+yn1)p)n0\left\lvert f(x) - f(y) \right\rvert_p\leq \left\lvert x-y \right\rvert_p\max\left(\left\lvert a_n(x^{n-1}+x^{n-2}y + \cdots + xy^{n-2}+y^{n-1}) \right\rvert_p\right)_{n\geq0} Now, xn1+xn2y+xyn2+yn1pmax(xnkyk1p)k=1n1\left\lvert x^{n-1}+x^{n-2}y + \cdots xy^{n-2} + y^{n-1} \right\rvert_p\leq \max\left(\left\lvert x^{n-k}y^{k-1} \right\rvert_p\right)_{k=1}^{n-1} Since xp<r\left\lvert x \right\rvert_p<r and yp<r\left\lvert y \right\rvert_p<r, it follows that xnkyk1p<rnkrk1=rn1\left\lvert x^{n-k}y^{k-1} \right\rvert_p<r^{n-k}r^{k-1} = r^{n-1}. Therefore, xn1+xn2y+xyn2+yn1p<rn1\left\lvert x^{n-1}+x^{n-2}y + \cdots xy^{n-2} + y^{n-1} \right\rvert_p<r^{n-1} Thus, f(x)f(y)p<xypmax(anprn1)n0\left\lvert f(x)-f(y) \right\rvert_p< \left\lvert x-y \right\rvert_p\max\left(\left\lvert a_n \right\rvert_pr^{n-1}\right)_{n\geq0} Now, by the definition of continuity, we must prove that for all ϵR+\epsilon \in \mathbb{R}^+, there exists some δR+\delta \in \mathbb{R}^+ such that xyp<δ\left\lvert x-y \right\rvert_p<\delta implies f(x)f(y)p<ϵ\left\lvert f(x)-f(y) \right\rvert_p<\epsilon. We can let δ=ϵmax(anprn1)n0\delta = \frac{\epsilon}{\max\left(\left\lvert a_n \right\rvert_pr^{n-1}\right)}_{n\geq0} for any given positive real ϵ\epsilon. Then, clearly xyp<δ\left\lvert x-y \right\rvert_p<\delta implies f(x)f(y)p\left\lvert f(x)-f(y) \right\rvert_p. Therefore, there always exists such a real number δ\delta and hence f(x)f(x) is continuous. ◻

For an alternative proof, we will use the notion of continuity between the mapping of two topological spaces.

Definition 22. Let XX and YY be topological spaces. The map f ⁣:XYf\colon X \to Y is continuous     \iff the preimage of the open set is open.

In other words, if you have a function ff mapping from one topological space XX to another topological space YY, and for any open set UYU \subset Y, the set of all points in XX that map to points in UU (i.e., the preimage of UU) is open in XX, then ff is a continuous map.

Proof. To begin, let’s prove first the following lemma:

Lemma 4. Let a,bQpa, b \in \mathbb{Q_p} and r,sR+r, s \in \mathbb{R}^+, we have the following properties of D(a;r)D(a;r^{-}):

  1. If bD(a;r)b \in D(a; r^{-}), then D(a;r)=D(b;r)D(a; r^{-}) = D(b; r^{-}).

  2. The open disc D(a;r)D(a; r^{-}) is also a closed set.

  3. D(a;r)D(b;s)    D(a;r)D(b;s)D(a;r^{-}) \cap D(b; s^{-}) \neq \emptyset \iff D(a; r^{-}) \subset D(b; s^{-}) or D(a;r)D(b;s)D(a; r^{-}) \supset D(b; s^{-})

Proof.

  1. Observe the following, we can rewrite xD(a;r)x \in D(a; r^{-}) as, xap<rxap=(xb+bap)max(xbp,bap)<r\begin{aligned} |x - a |_p < r \\ |x - a |_p = (|x - b + b - a|_p) \le \max(|x-b|_p, |b-a|_p) < r \end{aligned} We have that max(xbp,bap)\max(|x-b|_p, |b-a|_p) is contained D(b;r)D(b; r^{-}), but then xapmax(xbp,bap)|x-a|_p \le \max(|x-b|_p, |b-a|_p), thus D(a;r)D(b;r)D(a; r^{-}) \subset D(b; r^{-}), and since it is given that bD(a;r)b \in D(a; r^{-}) we also have D(b;r)D(a;r)D(b; r^{-}) \supset D(a; r^{-}). Hence, D(a;r)=D(b;r)D(a; r^{-}) = D(b; r^{-}) as claimed.

  2. By definition, D(a;r)D(a; r^{-}) is an open set. We will show that it is also a closed set. Pick a boundary point in D(a;r)D(a; r), and call it xx, and also choose srs \le r. Since xx is a boundary point, we have D(a;r)D(x;s)D(a; r) \cap D(x; s) \neq \emptyset, then yD(a;r)D(x,s)\exists y \in D(a; r) \cap D(x, s), this means that ya<r|y - a| < r and yx<sr|y - x| < s \le r. Using the non-archimedean inequality, we have: xa<max(xy,ya)<max(s,r)r|x - a| < \max(|x - y|, |y - a|) < \max(s, r) \le r thus xD(a;r)x \in D(a; r) such that D(a;r)D(a; r) contains each of its boundary points, making D(a;r)D(a; r) a closed set by definition.

  3. Assume W.L.O.G that rsr \le s. If the intersection is non-empty then there exists a cc in D(a;r)D(b;s)D(a; r) \cap D(b; s). Then we know, from (i)(i), that D(a;r)=D(c;r)D(a; r) = D(c; r) and D(b;s)=D(c;s)D(b; s) = D(c; s). Hence D(a;r)=D(c;r)D(c;s)=D(b;s)D(a; r) = D(c; r) \subset D(c; s) = D(b; s)

 ◻

Now to begin the proof, we define the preimage of D(y;s)D(y; s^{-}) under ff as, f1(D(y;s))={aD(0;r)f(a)D(y;s)}f^{-1}(D(y; s^{-})) = \lbrace a \in D(0; r^{-}) | f(a) \in D(y; s^{-}) \rbrace For a sketch-proof, when these set of points in D(0;r)D(0; r^{-}) that is in the preimage of ff are open then ff is continuous. Now, fix an element of the preimage of D(y;s)D(y; s^{-}) under ff, and call it tt such that tp<r|t|_p < r. By definition, f(t)f(t) converges in D(y;s)D(y; s^{-}). By Proposition 16, it follows that antn|a_nt^n| converges to 00 in D(y;s)D(y; s^{-}). We have that antnp|a_nt^n|_p is a Cauchy sequence. Since anQpa_n \in \mathbb{Q_p}, it is Cauchy, then we have antnp=anptnp<ϵtnp|a_nt^n|_p = |a_n|_p|t^n|_p < \epsilon\cdot |t^n|_p. For which it follows that (tn)(t^n) is Cauchy since we have amtmp<ϵ|a_mt^m|_p < \epsilon

for every m>Mm > M (for a fixed large MM). Then it follows that tt converges in D(y;s)D(y; s^{-}). Next, we have that tD(y;s)t \in D(y; s^{-}), and also tD(0;r)t \in D(0; r^{-}) then D(y;s)D(0;r)={t}D(y; s^{-}) \cap D(0; r^{-}) = \lbrace t \rbrace. By Lemma 4, we know that D(0;r)=D(t;r)D(t;s)=D(y;s)D(0; r^{-}) = D(t; r^{-}) \subset D(t; s^{-}) = D(y; s^{-}). Then we have a union of open disks which are the preimage of D(y;s)D(y; s^{-}) under ff, f1(D(y;s))=tp<rD(t;r)f^{-1}(D(y; s^{-})) = \bigcup_{|t|_p < r} D(t; r^{-}) Hence it follows that ff is continuous as claimed. ◻

Remarks 1. The characterization all power series f(x)Qp[[x]]f(x) \in \mathbb{Q_p}[[x]] such that f(x)f(x) converges at every point of the closed unit disk D(0;1)D(0; 1) are as follows:

  • The function f:D(0;r)D(0;1)f: D(0; r^{-}) \to D(0; 1^{-}) must be continuous.

  • Given tD(0;r)t \in D(0; r^{-}), the sequence of coefficient of ff, given by (an)(a_n) satisfy that antnp|a_nt^n|_p converges to 0 in D(0;1)D(0; 1).

  • For a power series to converge to 11, it needs f(0)=1f(0) = 1. Since there exists gQp[[x]]g \in \mathbb{Q_p}[[x]] such that fg=1f \cdot g = 1 for which multiplication is closed in Qp[[x]]\mathbb{Q_p}[[x]]. In other words, f(x)1+xQp[[x]]f(x) \in 1 + x\mathbb{Q_p}[[x]].

Exponentiation in pp-adics #

We define the exponentiation of xx in the pp-adics to be the following element of Qp[[x]]\mathbb{Q_p}[[x]]: exp(x)=n0xnn!\exp(x) = \sum_{n\geq0}\frac{x^n}{n!} We see that the exp\exp function has the following properties:

Proposition 21. We have exp(x+y)=exp(x)exp(y)\exp(x+y) = \exp(x)\cdot\exp(y) in Qp[[x,y]]\mathbb{Q_p}[[x, y]].

Proof. We have exp(x)=1+x+x22!+x33!+\exp(x) = 1 + x + \frac{x^2}{2!} + \frac{x^3}{3!}+\cdots and exp(y)=1+x+y22!+y33!+\exp(y) = 1 + x + \frac{y^2}{2!} + \frac{y^3}{3!}+\cdots Thus, exp(x)exp(y)=(1+x+x22!+)(1+y+y22!+)\exp(x)\cdot\exp(y) = \left(1+x+\frac{x^2}{2!}+\cdots\right)\left(1+y+\frac{y^2}{2!}+\cdots\right) Expanding the above product and combining all terms with the same degree, we get: exp(x)exp(y)=1+(x+y)+(x22!+xy+y22!)++k=0n(xkynkk!(nk)!)+\exp(x)\cdot\exp(y) = 1+(x+y) + \left(\frac{x^2}{2!}+ xy + \frac{y^2}{2!}\right) + \cdots + \sum_{k=0}^n\left(\frac{x^ky^{n-k}}{k!(n-k)!}\right)+\cdots Writing the above as a summation with respect to nn, we get: exp(x)exp(y)=n0(k=0nxkynkk!(nk)!)=n0(k=0n(nk)xkynkn!)=n0(x+y)nn!=exp(x+y)\exp(x)\cdot\exp(y) = \sum_{n\geq0}\left(\sum_{k=0}^n\frac{x^ky^{n-k}}{k!(n-k)!}\right) = \sum_{n\geq0}\left(\frac{\sum_{k=0}^n\binom{n}{k}x^ky^{n-k}}{n!}\right) = \sum_{n\geq0}\frac{(x+y)^n}{n!}=\exp(x+y) by the binomial theorem, and we are done. Hence, we have a homomorphism ϕ:Qp2Qp\phi : \mathbb{Q_p}^2 \to \mathbb{Q_p} such that exp(x)\exp(x) satisfies ϕ(xy)=ϕ(x)ϕ(y)\phi(x \circ y) = \phi(x) \circ \phi(y) ◻

Proposition 22. The radius of convergence of exp(x)\exp(x) is p1p1p^{-\frac{1}{p-1}}

Proof. We know that the radius, rr, of convergence of any power series in the pp-adics is given by r=(limsupanp1/n)1r=(\lim\sup\left\lvert a_n \right\rvert_p^{1/n})^{-1}. Therefore, we know that the radius of convergence, rr of exp(x)\exp(x) is given by (limsup1n!p)1(\lim\sup\left\lvert \frac{1}{n!} \right\rvert_p)^{-1}. We need to therefore prove that (limsup1n!p)1=p1/(p1)\left(\lim\sup\left\lvert \frac{1}{n!} \right\rvert_p\right)^{-1} = p^{-1/(p-1)} In order to do so, we need the following lemma

Lemma 5. vp(n!)=nSp(n)p1v_p(n!) = \frac{n - S_p(n)}{p-1} (Sp(n)S_p(n) is the sum of all digits of nn over base pp)

Proof. Notice that vp(n!)=vp(n)+vp(n1)+=lnvp(l)v_p(n!) = v_p(n) + v_p(n-1) + \cdots = \sum_{l \le n} v_p(l). We take vp(l)v_p(l) (lnl \le n), and expand ll over base pp. Take l=lmpm++lrprl = l_mp^m + \cdots + l_rp^r (mrm \le r, lm0l_m \neq 0), where we have vp(l)=mv_p(l) = m. Using telescoping techniques, observe that: 1=(p1)+(p1)p+(p1)p2++(p1)pmpml1=(p1)+(p1)p+(p1)p2++(p1)pm1+(lm1)pm++lrpr\begin{aligned} - 1 = (p - 1) + (p - 1)p + (p - 1)p^2 + \cdots + (p - 1)p^m - p^m \\ l - 1 = (p - 1) + (p - 1)p + (p - 1)p^2 + \cdots + (p - 1)p^{m-1} + (l^m - 1) p^m + \cdots + l_rp^r \end{aligned} We have that the sum of the digits (over base pp) of l1l - 1 is, Sp(l1)=m(p1)+Sp(l)1S_p(l - 1) = m(p - 1) + S_p(l) - 1 The reason why there is 1-1 on the RHS since we have lm1l_m - 1 from the previous equation, thus it follows we have Sp(l)1S_p(l) - 1. We know that vp(l)=mv_p(l) = m, then solving for mm, we have m=1p1[Sp(l1)Sp(l)+1]m = \frac{1}{p - 1}\left[ S_p(l - 1) - S_p(l) + 1 \right] Then we have, vp(n!)=lnvp(l)=1p1ln[Sp(l1)Sp(l)+1]v_p(n!) = \sum_{l \le n} v_p(l) = \frac{1}{p-1}\sum_{l \le n} [S_p(l - 1) - S_p(l) + 1] Since this is a telescoping series, we have that: vp(n!)=1p1(Sp(n)+n)=nSp(n)p1v_p(n!) = \frac{1}{p-1}(-S_p(n) + n) = \frac{n - S_p(n)}{p - 1} ◻

Lemma 6. limsup((vp(n!)/n)\lim\sup(\left(v_p(n!)/n\right) converges to 1/(p1)1/(p-1).

Proof. Consider some non-negative integer kk. Consider the sequence of all reals vp(n!)/nv_p(n!)/n where nn is so that pkn<pk+1p^k \leq n < p^{k+1}. The maximum value of this sequence occurs when n=pkn = p^k. Thus, the supremum of the sequence vp(n!)/nv_p(n!)/n when pkn<pk+1p^k \leq n < p^{k+1} is vp((pk)!)/pkv_p((p^k)!)/p^k. By Legendre’s formula vp(n!)=np+np2+v_p(n!) = \left\lfloor\frac{n}{p}\right\rfloor + \left\lfloor\frac{n}{p^2}\right\rfloor+\cdots When n=pkn=p^k, this simplifies to: vp(n!)=pk1++1=pk1p1v_p(n!) = p^{k-1}+\cdots + 1 = \frac{p^k-1}{p-1} Thus, vp(n!)n=pk1pk(p1)=1p1pk1pk\frac{v_p(n!)}{n} = \frac{p^{k}-1}{p^k(p-1)} = \frac{1}{p-1}\cdot\frac{p^k-1}{p^k} The above gives the maximum value of vp(n!)v_p(n!) for nn between pkp^k and pk+1p^{k+1}. Thus, one may conclude that limsup(vp(n!)/n)=limk(1p1pk1pk)\lim\sup(v_p(n!)/n) = \lim_{k\to\infty}\left(\frac{1}{p-1}\cdot\frac{p^k-1}{p_k}\right) The 1/(p1)1/(p-1) term is constant. Moreover, the sequence px1(p1)px\frac{p^x-1}{(p-1)p^x} where xx is a real number converges to the same real number as (pk1)/(pk(p1))(p^k-1)/(p^k(p-1)) where kk is an integer since ZR\mathbb{Z}\in \mathbb{R}. Therefore limk1p1pk1pk=limx1p1px1px=1p1\lim_{k\to\infty}\frac{1}{p-1}\cdot\frac{p^k-1}{p^k} = \lim_{x\to\infty}\frac{1}{p-1}\cdot\frac{p^x-1}{p^x} = \frac{1}{p-1} Hence limsup(vp(n!)/n)=1/(p1)\lim\sup(v_p(n!)/n) = 1/(p-1). ◻

Now, our goal is to find limsup\lim\sup of the sequence 1/n!p1/n\left\lvert 1/n! \right\rvert_p^{1/n}. The sequence (1/n!p1/n)(\left\lvert 1/n! \right\rvert_p^{1/n}) can be rewritten as (pvp(1/n!)/n)=(pvp(n!)/n)\left(p^{-v_p(1/n!)/n}\right) = \left(p^{v_p(n!)/n}\right) Now, limsup(pvp(n!)/n)=plimsup(vp(n!)/n)\lim\sup\left(p^{v_p(n!)/n}\right) = p^{\lim\sup(v_p(n!)/n}) since pkp^k is a strictly increasing function with respect to kk. By our lemma, we therefore have limsup(1/n!p1/n)=p1/(p1)\lim\sup(\left\lvert 1/n! \right\rvert_p^{1/n}) = p^{1/(p-1)} Thus, r=p1/(p1)r = p^{-1/(p-1)} and we are done. ◻

Proposition 23. For all a,bD(0;(p1/(p1)))a, b \in D(0; \left(p^{-1/(p-1)}\right)^-), we have a+bD(0;(p1/(p1)))a+b \in D(0; \left(p^{-1/(p-1)}\right)^-). Therefore, we have exp(a+b)=exp(a)exp(b)\exp(a+b) = \exp(a)\cdot\exp(b).

Proof. By definition, from a,bD(0;(p1/(p1)))a,b\in D\left(0;\left(p^{-1/(p-1)}\right)^-\right) we have ap,bp<p1/(p1)\left\lvert a \right\rvert_p, \left\lvert b \right\rvert_p<p^{-1/(p-1)}. Thus, we have max(ap,bp)<p1/(p1)\max(\left\lvert a \right\rvert_p, \left\lvert b \right\rvert_p)<p^{-1/(p-1)}. Hence, a+bpmaxap,bp<p1/(p1)\left\lvert a+b \right\rvert_p\leq \max{\left\lvert a \right\rvert_p, \left\lvert b \right\rvert_p}<p^{-1/(p-1)}. Thus, we have a+bD(0;(p1/(p1)))a+b\in D\left(0;\left(p^{-1/(p-1)}\right)^-\right).
 
Now, we know exp(a)=limn(k=0nakk!)\exp(a) = \lim_{n\to\infty}\left(\sum_{k=0}^n\frac{a^k}{k!}\right) and exp(b)=limn(k=0nbkk!)\exp(b) = \lim_{n\to\infty}\left(\sum_{k=0}^n\frac{b^k}{k!}\right) Thus, exp(a)exp(b)=(limn(k=0nakk!))(limn(k=0nbkk!))=limn(k=0nakk!k=0nbkk!)\exp(a)\cdot\exp(b) = \left(\lim_{n\to\infty}\left(\sum_{k=0}^n\frac{a^k}{k!}\right)\right)\cdot\left(\lim_{n\to\infty}\left(\sum_{k=0}^n\frac{b^k}{k!}\right)\right) = \lim_{n\to\infty}\left(\sum_{k=0}^n\frac{a^k}{k!}\cdot\sum_{k=0}^n\frac{b^k}{k!}\right) We can expand the product of the summations and combine the terms of the same degree to get: k=0nakk!k=0nbkk!=k=0n(a+b)kk!+fn(a,b)\sum_{k=0}^n\frac{a^k}{k!}\cdot\sum_{k=0}^n\frac{b^k}{k!} = \sum_{k=0}^n\frac{(a+b)^k}{k!}+f_n(a, b) where fn(x,y)f_n(x, y) is some polynomial in Qp[x,y]\mathbb{Q_p}[x,y] such that the smallest degree of the terms is n+1n+1. Therefore, limn(k=0nakk!k=0nbkk!)=limn=(k=0n(a+b)kk!)+limnfn(a,b)=exp(a+b)+limnfn(a,b)\lim_{n\to\infty}\left(\sum_{k=0}^n\frac{a^k}{k!}\cdot\sum_{k=0}^n\frac{b^k}{k!}\right) = \lim_{n\to=\infty}\left(\sum_{k=0}^n\frac{(a+b)^k}{k!}\right)+\lim_{n\to\infty}f_n(a, b)=\exp(a+b)+\lim_{n\to\infty}f_n(a, b) by definition of exp(a+b)\exp(a+b). We know that the degree of the term with lowest degree in f(x,y)f(x, y) is n+1n+1. Moreover, we have xp,yp<p1/(p1)\left\lvert x \right\rvert_p, \left\lvert y \right\rvert_p<p^{-1/(p-1)}. Assume WLOG that xpyp<p1/(p1)\left\lvert x \right\rvert_p\leq \left\lvert y \right\rvert_p<p^{-1/(p-1)}. Hence, xkyn+1kp<yn+1p<pn/(p1)\left\lvert x^ky^{n+1-k} \right\rvert_p<\left\lvert y^{n+1} \right\rvert_p<p^{-n/(p-1)} for any kk. Moreover, for any term in f(x,y)f(x, y) with the power of xx and yy being s,ts, t respectively, we must have xsytp<pn/(p1)\left\lvert x^sy^t \right\rvert_p<p^{-n/(p-1)} since s+tn+1s+t\geq n+1 for f(x,y)f(x,y). But, each term of f(x,y)f(x, y) also has coefficients. We know that the coefficients of f(x)f(x) are at least (1/n!)2(1/n!)^2. Thus, the pp-adic absolute value of a term of f(x,y)f(x, y) is at most p2vp(n!)pn/(p1)p^{2v_p(n!)}p^{-n/(p-1)} Now, we know that vp(n!)1/(p1)v_p(n!)\leq 1/(p-1) as nn goes to infinity as we saw earlier. Thus, the maximum value of the pp-adic absolute value of the individual terms of f(x,y)f(x, y) is p2np1p^{\frac{2-n}{p-1}} as nn approaches \infty. As nn approaches infinity, the above approaches 00. Thus, limnfn(a,b)p=0\lim_{n\to\infty}\left\lvert f_n(a, b) \right\rvert_p = 0. Hence, fn(a,b)f_n(a, b) itself approaches 00 in Qp\mathbb{Q_p} as we have seen earlier. Hence, exp(a+b)=exp(a)exp(b)\exp(a+b) = \exp(a)\exp(b). ◻

Corollary 1. exp(na)=exp(an)\exp(na) = \exp(a^n) for all integers nn and aD(0;(p1/(p1)))a \in D(0; \left(p^{-1/(p-1)}\right)^-).

Proposition 24. We have exp(x)exp(y)p=xyp\left\lvert \exp(x)-\exp(y) \right\rvert_p = \left\lvert x-y \right\rvert_p.

Proof. We first prove that the statement is true when y=0y=0. When y=0y=0, we have exp(y)=0\exp(y) = 0. Therefore, we need to prove that exp(x)1p=xp\left\lvert \exp(x) - 1 \right\rvert_p = \left\lvert x \right\rvert_p. We have exp(x)1=n1xnn!\exp(x)-1 = \sum_{n\geq1}\frac{x^n}{n!} As we saw earlier, we have exp(x)1pmax(xnn!p)n1\left\lvert \exp(x) - 1 \right\rvert_p \leq \max\left(\left\lvert \frac{x^n}{n!} \right\rvert_p\right)_{n\geq1} We claim that the maximum absolute value is xp\left\lvert x \right\rvert_p. In order to do so, let us suppose for the sake of contradiction that there is some term xn/n!x^n/n! with an absolute value that is more than or equal to the absolute value of xx. So, we have xn/n!pxp\left\lvert x^n/n! \right\rvert_p \geq \left\lvert x \right\rvert_p, which happens if and only if xpn11/n!p1\left\lvert x \right\rvert_p^{n-1}\left\lvert 1/n! \right\rvert_p\geq1 or xpn1n!p\left\lvert x \right\rvert_p^{n-1} \geq \left\lvert n! \right\rvert_p. By the definition of the pp-adic absolute value, this happens if and only if p(n1)vp(x)pvp(n!)    (n1)vp(x)vp(n!)    (n1)vp(x)vp(n!)    vp(n!)/(n1)vp(x) \begin{aligned} p^{-(n-1)v_p(x)} \geq p^{-v_p(n!)} &\iff -(n-1)v_p(x) \geq -v_p(n!) \\ &\iff (n-1)v_p(x)\leq v_p(n!) \\ &\iff v_p(n!)/(n-1) \geq v_p(x) \end{aligned}

Now, xp<p1/(p1)\left\lvert x \right\rvert_p < p^{-1/(p-1)}. Therefore, vp(x)>1/(p1)v_p(x) > 1/(p-1). Thus, we finally get vp(n!)/(n1)>1/(p1)v_p(n!)/(n-1) > 1/(p-1) which is a contradiction. Therefore, we must have that xp\left\lvert x \right\rvert_p is the unique maximum value in the sequence xn/n!p\left\lvert x^n/n! \right\rvert_p. Since the maximum value is unique, we have n1xnn!p=max(xnn!p)n1=xp\left\lvert \sum_{n\geq 1}\frac{x^n}{n!} \right\rvert_p = \max\left(\left\lvert \frac{x^n}{n!} \right\rvert_p\right)_{n\geq1} = \left\lvert x \right\rvert_p This proves that exp(x)1p=xp\left\lvert \exp(x) - 1 \right\rvert_p = \left\lvert x \right\rvert_p.
 
Now, consider any yy in D(0;(p1/(p1)))D\left(0;\left(p^{-1/(p-1)}\right)^-\right). Therefore, we have exp(x)exp(y)p=(exp(x)1)(exp(y)1)pmax(exp(x)1p,exp(y)1p)\left\lvert \exp(x) - \exp(y) \right\rvert_p = \left\lvert (\exp(x) - 1) - (\exp(y) - 1) \right\rvert_p\leq \max(\left\lvert \exp(x) -1 \right\rvert_p, \left\lvert \exp(y)-1 \right\rvert_p) Since exp(x)1p=xp\left\lvert \exp(x) - 1 \right\rvert_p = \left\lvert x \right\rvert_p, we have the above may be written as max(xp,yp)\max(\left\lvert x \right\rvert_p, \left\lvert y \right\rvert_p). If xpyp\left\lvert x \right\rvert_p \neq \left\lvert y \right\rvert_p, we have exp(x)1pexp(y)1p\left\lvert \exp(x) - 1 \right\rvert_p \neq \left\lvert \exp(y) - 1 \right\rvert_p. Therefore, we have exp(x)exp(y)p=max(xp,yp)=xyp\left\lvert \exp(x) - \exp(y) \right\rvert_p = \max(\left\lvert x \right\rvert_p, \left\lvert y \right\rvert_p) = \left\lvert x-y \right\rvert_p. ◻

Artin-Hasse Exponential Function #

Theorem 2. E(x)Zp[[x]]E(x) \in \mathbb{Z_p}[[x]]

To prove this theorem, we need to prove the following Lemmas first:

Lemma 7. Let f(x)1+xQp[[x]]f(x) \in 1 + x\mathbb{Q_p}[[x]] be a power series with pp-adic rational coefficients. Then f(x)1+xZp[[x]]    f(xp)f(x)p1+pxZp[[x]]f(x) \in 1 + x\mathbb{Z_p}[[x]] \iff \frac{f(x^p)}{f(x)^p} \in 1 +px\mathbb{Z_p}[[x]]

Proof. We will begin with the assumption that f(x)1+xZp[[x]]f(x) \in 1 + x\mathbb{Z_p}[[x]]. We can see that the constant term is given by, F(0)=f(0)pf(0p)=(1+i1ai0i)p(1+i1ai0pi)=1p1=0F(0) = f(0)^p - f(0^p) = \left( 1 + \sum_{i \ge 1}a_i0^i\right)^p - \left(1 + \sum_{i \ge 1} a_i0^pi\right) = 1^p - 1 = 0 Thus we have that f(0)p=f(0p)=1f(0)^p = f(0^p) = 1, then it follows that f(x)pf(x)^p and f(xp)f(x^p) are both invertible formal power series. Then there exists t(x)1+pxZp[[x]]t(x) \in 1 + px\mathbb{Z_p}[[x]] such that f(xp)f(x)p=t(x)\frac{f(x^p)}{f(x)^p} = t(x). The reason why t(x)1+pxZp[[x]]t(x) \in 1 + px\mathbb{Z_p}[[x]] is because that the coefficients of t(x)t(x) satisfy a linear recursion that is derived from the product f(x)pt(x)=f(xp)f(x)^p \cdot t(x) = f(x^p). By construction, the coefficients of t(x)t(x) are an=i=1mcian1a_n = \sum_{i = 1}^{m} c_ia_{n-1} (m0\exists m \ge 0, n>0n > 0), and (ci)(c_i) are the coefficients of f(x)pf(x)^p. By the multinomial theorem, the coefficients (ci)(c_i) is given by p!/(r1!r2!r3!())p!/(r_1!r_2!r_3!(\cdots)) such that i0ri=p\sum_{i \ge 0} r_i = p. From this, we know that the coefficients (except the constant term) of f(x)pf(x)^p is a multiple of pp. Hence it follows that t(x)1+pxZp[[x]]t(x) \in 1 + px\mathbb{Z_p}[[x]] exists. Now, supposed that f(xp)=f(x)pg(x)f(x^p) = f(x)^p\cdot g(x) with g(x)1+pxZp[[x]]g(x) \in 1 + px\mathbb{Z_p}[[x]]. Let f(x)=n0anxnf(x) = \sum_{n \ge 0} a_nx^n, g(x)=n0bnxng(x) = \sum_{n \ge 0} b_nx^n. By assumption, a0=1a_0 = 1, and suppose that we have the required integrality for ana_n with 0n<N10 \le n < N -1. We will show that the NNth coefficient of f(x)pg(x)f(x)^p\cdot g(x) is equal to the NNth coefficient of (nNanxn)p+nNbnxn(\sum_{n \le N} a_nx^n)^p + \sum_{n \le N} b_nx^n for which this sum is just a redefinition of f(xp)f(x^p) if we take NN \to \infty. The proof will proceed by induction. To begin, let’s expand these power series to get a sense of the terms, f(x)p=(i=0N1aixi+aNxN+kN+1akxk)p=t=0p(pt)(i=0N1aixi+kN+1akxk)pt(aNxN)tf(x)^p = \left( \sum_{i = 0}^{N-1} a_ix^i + a_Nx^N + \sum_{k \ge N + 1}a_kx^k\right)^p = \sum_{t=0}^{p} \binom{p}{t}\left(\sum_{i = 0}^{N-1} a_ix^i + \sum_{k \ge N + 1}a_kx^k\right)^{p - t}(a_Nx^N)^t (nNanxn)p=i=0p(pi)(nN1anxn)pi(aNxN)i\left(\sum_{n \le N} a_nx^n\right)^p =\sum_{i = 0}^{p} \binom{p}{i} \left(\sum_{n \le N - 1}a_nx^n \right)^{p-i}(a_Nx^N)^i

We have two cases, p∤Np \not\mid N and pNp \mid N. Suppose p∤Np \not\mid N, then there is no mZm \in \mathbb{Z} such that N=pmN = pm (that would give us another NNth coefficient which is aNppa_{\frac{N}{p}}^p). Thus we have that the NNth coefficient of (nNanxn)p+nNbnxn(\sum_{n \le N} a_nx^n)^p + \sum_{n \le N} b_nx^n is (p1)aNa0+bN=p(1)(aN)+bN=paN+bN\binom{p}{1}a_Na_0 + b_N = p(1)(a_N) + b_N = pa_N + b_N, while we have that the NNth coefficient of f(x)pg(x)f(x)^p\cdot g(x) is g(0)(p1)aN+a0bNg(0)\binom{p}{1}a_N + a_0b_N. Since we have assumed that a0=1a_0 = 1, then by the linear recursion definition of the coefficient of g(x)g(x), we have that a0b0=1a_0b_0 = 1 then b0=1b_0 = 1. It follows that the NNth term of f(x)pg(x)f(x)^p\cdot g(x) is (1)paN+(1)bN=paN+bN(1)pa_N + (1)b_N = pa_N + b_N. The intuition behind the multiplication of f(x)pg(x)f(x)^p\cdot g(x) is just termwise multiplication, thus we can find the pairs such that they’re the NNth term. Since by definition, g(x)1+pZp[[x]]g(x) \in 1 + p\mathbb{Z_p}[[x]], it follows that bNpZpb_N \in p\mathbb{Z_p}. Now, it follows that we can always construct the NNth coefficient using the linear-recursion definition of the coefficient of g(x)g(x) then the construction is followed by induction. Hence both of them have the same NNth coefficient such that p∤Np \not\mid N. Also, we can conclude that aNZpa_N \in \mathbb{Z_p}, since the NNth coefficient is not divisible by p2p^2 but by pp only, thus it follows that the NNth coefficient is in pZpp\mathbb{Z_p} and aNZpa_N \in \mathbb{Z_p}. Now for our second case, suppose that pNp \mid N, then there exists mZm \in \mathbb{Z} such that N=pmN = pm. We have that the NNth coefficient of (nNanxn)p+nNbnxn(\sum_{n \le N} a_nx^n)^p + \sum_{n \le N} b_nx^n is aNpp+(p1)aN+bN=aNpp+(1)aN+bN=aNpp+aN+bNa_{\frac{N}{p}}^p + \binom{p}{1}a_N + b_N = a_{\frac{N}{p}}^p + (1)a_N + b_N = a_{\frac{N}{p}}^p + a_N + b_N. To find aNppa_{\frac{N}{p}}^p in (nNanxn)p(\sum_{n \le N} a_nx^n)^p, consider the ppth term and set i=0i = 0, observe the following:

(nN1anxn)p=(nNp1anxn+aNpxNp+n=Np1N1anxn)p\left(\sum_{n \le N - 1}a_nx^n \right)^{p} = \left(\sum_{n \le \frac{N}{p} - 1}a_nx^n + a_{\frac{N}{p}}x^{\frac{N}{p}} + \sum_{n = \frac{N}{p} - 1}^{N - 1}a_nx^n \right)^{p}

(nNp1anxn+aNpxNp+n=Np1N1anxn)p=((nNp1anxn+n=Np1N1anxn)+aNpxNp)p\left(\sum_{n \le \frac{N}{p} - 1}a_nx^n + a_{\frac{N}{p}}x^{\frac{N}{p}} + \sum_{n = \frac{N}{p} - 1}^{N - 1}a_nx^n \right)^{p} = \left(\left( \sum_{n \le \frac{N}{p} - 1}a_nx^n + \sum_{n = \frac{N}{p} - 1}^{N - 1}a_nx^n \right) + a_{\frac{N}{p}}x^{\frac{N}{p}} \right)^{p}

Then by the binomial theorem, it follows that we have aNppa_{\frac{N}{p}}^p as the additional term for the NNth coefficient of (nNanxn)p+nNbnxn(\sum_{n \le N} a_nx^n)^p + \sum_{n \le N} b_nx^n. While for the NNth coefficient of f(x)pg(x)f(x)^p\cdot g(x), we have aNpp+g(0)(p1)aN+a0bNa_{\frac{N}{p}}^p + g(0)\binom{p}{1}a_N + a_0b_N. To find the value of aNppa_{\frac{N}{p}}^p, consider the ppth term and set t=0t = 0, observe the following:

(i=0N1aixi+kN+1akxk)p=(i=0(N/p)1aixi+aNpxNp+kNp+1akxk)p(i=0(N/p)1aixi+aNpxNp+kNp+1akxk)p=((i=0(N/p)1aixi+kNp+1akxk)+aNpxNp)p\begin{aligned} \left(\sum_{i = 0}^{N-1} a_ix^i + \sum_{k \ge N + 1}a_kx^k \right)^{p} = \left(\sum_{i = 0}^{(N/p) - 1} a_ix^i + a_{\frac{N}{p}}x^{\frac{N}{p}} + \sum_{k \ge \frac{N}{p} + 1}a_kx^k \right)^{p}\\ \left(\sum_{i = 0}^{(N/p) - 1} a_ix^i + a_{\frac{N}{p}}x^{\frac{N}{p}} + \sum_{k \ge \frac{N}{p} + 1}a_kx^k \right)^{p} = \left( \left( \sum_{i = 0}^{(N/p) - 1} a_ix^i + \sum_{k \ge \frac{N}{p} + 1}a_kx^k \right) + a_{\frac{N}{p}}x^{\frac{N}{p}} \right)^{p} \end{aligned} Then by the binomial theorem, it follows that we have aNppa_{\frac{N}{p}}^p as the additional term for the NNth coefficient of f(x)pg(x)f(x)^p\cdot g(x). Now, it follows that we can always construct the NNth coefficient using the linear-recursion definition of the coefficient of g(x)g(x) then the construction is followed by induction. Hence both of them have the same NNth coefficient such that pNp \mid N. Also in our second case, we can conclude that aNZpa_N \in \mathbb{Z_p}, since the NNth coefficient is not divisible by ppp^p but by pp only, thus it follows that the NNth coefficient is in pZpp\mathbb{Z_p} and aNZpa_N \in \mathbb{Z_p}. Since we have successfully concluded that aNZpa_N \in \mathbb{Z_p}, it follows that f(x)1+xZp[[x]]f(x) \in 1 + x\mathbb{Z_p}[[x]]. ◻

Lemma 8. exp(px)1+pZp[[x]]\exp(-px) \in 1 + p\mathbb{Z_p}[[x]]

Lemma 9. E(xp)E(x)p=exp(px)\frac{E(x^p)}{E(x)^p} = \exp(-px)

Now we are ready to prove Theorem 2.

Proof. It follows that E(x)1+xZp[[x]]E(x) \in 1 + x\mathbb{Z_p}[[x]] by Lemma 7 due to Lemma 9, thus it follows that the coefficients of E(x)E(x) is in Zp\mathbb{Z_p} by Lemma 7. ◻