We describe some periodic structure in the cohomology of the moduli stack of 1-dimensional formal group laws, also known as the E_2-page of the classical Adams–Novikov spectral sequence. This structure is distinct from the familiar v_n-periodicities, and it displays interesting number-theoretic properties. Our techniques involve the \mathbb{C}-motivic Adams spectral sequence, and we obtain analogous periodic structure in \mathbb{C}-motivic stable homotopy.
(see Theorem 4.7 and Theorem 4.11) For all n \geq 1, there is a non-zero element in \mathbb{C}-motivic stable homotopy such that:
the coweight of the element is 4n-1.
the element is not divisible by \eta.
when n = 2^j, the element equals (the \mathbb{C}-motivic version of) the Mahowald element \eta_{j+3} in stem 2^{j+3}.
for n \geq 2, the nth term of sequence (?) gives the smallest number N such that \eta^N annihilates the element.
Theorem 1.2 can be compared to the following well-known claims about classical v_1-periodicity. For all n \geq 1, there is a non-zero element in classical stable homotopy such that:
the stem of the element is 4n-1.
the element is not divisible by 2.
the nth term of the sequence
gives the smallest number N such that 2^N annihilates the element.
See ? below for a further description of this analogous classical v_1-periodic structure.
As an aside, we point out that the classical Adams spectral sequence for \mathit{tmf} can be used to detect additional infinite w_1-periodic families of non-zero differentials. For example, the element d_0 g^4 in the 94-stem is the value of a classical Adams d_2 differential [IWX22b]. Comparison to the Adams spectral sequence for \mathit{tmf} shows that d_0 g^k is non-zero in the Adams E_2-page for all k. Therefore, the elements d_0 g^k, for k \geq 4, are the targets of a w_1-periodic family of classical differentials.
In the \mathbb{C}-motivic context, there is a corresponding family of Adams d_2 differentials whose targets are \tau^6 d_0 g^k for k \geq 4.
(see Corollary 8.1 and Corollary 8.3) For all n \geq 1, there is a non-zero element in the cohomology of the moduli stack of 1-dimensional formal group laws such that:
the difference between the stem and the cohomological filtration of the element is 8n-2.
the element is not divisible by \alpha_1.
When n = 2^j, the element equals the \beta-family element that detects Mahowald’s \eta_{j+3} in stem 2^{j+3}.
For n \geq 2, the nth term of sequence (?) gives the smallest number N such that \alpha_1^N annihilates the element.
Determine the first unknown value in the sequence (?). In other words, compute the \eta-exponent of the element in coweight 95 = 4 \cdot 24 - 1.
What is the \eta-exponent in coweight 3 \cdot 2^n - 1?
Prove that the values of \eta-exponents in coweight 4k-1 depend only on the 2-adic valuation of k, with the exception of coweights of the form 2^n-1.
Give a number-theoretic description of the sequence (?).
Compute more \mathbb{C}-motivic w_1-periodic homotopy.
Consider the element H_1 in degree (62,5,33). Determine the w_1-periodic behavior of this element.
Is there a w_1-periodic version of the classical image of J spectrum?
Determine the differentials and hidden extensions in the Burklund–Xu spectral sequence in Chow degree one in a range, thereby determining the Adams E_2-page in Chow degree one in that range.
[Isa19, Theorem 2.19] There are highly structured isomorphisms between H^{**} \mathcal{A}^{\mathrm{cl}}, H^{***} \mathcal{B}, and the part of H^{***} \mathcal{A} that lies in Chow degree zero.
One consequence of Remark 5.3 is that elements of the form 1^{BX} \cdot x are permanent cycles in the Burklund–Xu spectral sequence [Benson24, Proposition 15]. Therefore, the spectral sequence collapses in Chow degree zero.
Moreover, the \mathbb{C}-motivic Cartan–Eilenberg spectral sequence in Chow degree zero also collapses. There are two main points to this observation. First, \mathbb{C}-motivic Cartan–Eilenberg differentials preserve Chow degree, so we need only consider differentials whose sources and targets are both in Chow degree zero. Second, \mathbb{C}-motivic Cartan–Eilenberg differentials take values in multiples of \tau [BX23, Theorem 2.8], and there are no multiples of \tau in Chow degree zero.
Hence, the \mathbb{C}-motivic Burklund–Xu E_1-page in Chow degree zero is isomorphic to the \mathbb{C}-motivic Cartan–Eilenberg E_2-page in Chow degree zero, which is isomorphic to the part of H^{***} \mathcal{A} in Chow degree zero.
Every non-zero element of H^{***}\mathcal{A} has a non-negative Chow degree.
By inspection of degrees, the cobar complex for \mathcal{A} is concentrated in non-negative Chow degrees.
In ?, we study h_0-submodules of the classical Adams E_2-page. Here we are working in an \mathbb{F}_2-algebra, so the characteristic of every h_0-submodule is 2.
In ?, we study h_1-submodules of the \mathbb{C}-motivic Adams E_2-page. Again we are working in an \mathbb{F}_2-algebra, so the characteristic of every h_1-submodule is 2.
In ?, we study \eta-submodules of the \mathbb{C}-motivic stable homotopy ring \pi_{*,*}. Now we are not working in an \mathbb{F}_2-algebra, so the characteristic can be larger than 2. In practice, some of the \eta-submodules that we study do have characteristics larger than 2, but we will not study characteristics because we do not have techniques to analyze them. Rather, we are more interested in \eta-exponents. Note that 2 \eta = 0, so our simpler classification of \eta-submodules applies.
In ?, we study \alpha_1-submodules of the cohomology of the moduli stack of 1-dimensional formal group laws, also known as the classical Adams–Novikov E_2-page. Again, we are not working in an \mathbb{F}_2-algebra, so the characteristic can be larger than 2, but we will not study characteristics. Rather, we are more interested in \alpha_1-exponents. Note that 2 \alpha_1 = 0, so our simpler classification of \alpha_1-submodules applies.
The elements \frac{\overline{v_1^{4k}}}{h_0} are closely related to each other via the usual Massey products that express v_1-periodicity. Specifically, we have
with no indeterminacy for all j and k.
More generally, assume that n \leq m. For each choice of \frac{\overline{v_1^{4j}}}{h_0^m} and each choice of \frac{\overline{v_1^{4k}}}{h_0^n}, the Massey product
has no indeterminacy and equals an element that is a possible choice for \frac{\overline{v_1^{4j+4k}}}{h_0^n}. Beware that the unique element in the Massey product may depend on the choices of \frac{\overline{v_1^{4j}}}{h_0^m} and of \frac{\overline{v_1^{4k}}}{h_0^n}.
For a fixed value of n, consider the elements of the form \frac{\overline{v_1^{4k}}}{h_0^n} in H^{**}\mathcal{A}^{\mathrm{cl}}. Such an element exists only when the h_0-coexponent of \frac{\overline{v_1^{4k}}}{h_0} is at least n. This family of existing elements forms a v_1-periodic family, but the degree of the periodicity depends on n.
For all j \geq 2, the h_0-coexponent of \frac{\overline{v_1^{2^j}}}{h_0} is 2^{j-1}.
This follows immediately from the fact that
For all k \geq 0, the h_0-coexponent of:
\frac{\overline{v_1^{8k+12}}}{h_0} is 6.
\frac{\overline{v_1^{16k+24}}}{h_0} is 12.
\frac{\overline{v_1^{32k+48}}}{h_0} is 28.
In all three cases, the proof follows a similar logic. Determine the h_0-coexponents in low dimensions by explicit computation. Then Adams periodicity [May-never] [Li20] determines the h_0-coexponents in higher dimensions.
For part (1), the h_0-coexponents of \frac{\overline{v_1^{12}}}{h_0}, \frac{\overline{v_1^{20}}}{h_0}, and \frac{\overline{v_1^{28}}}{h_0} are 6 by explicit computation [IWX22b]. For k \geq 3, we use that the Adams periodicity operator P^4 of degree (32,16) induces an isomorphism on H^{**}\mathcal{A}^\mathrm{cl} in the relevant range.
For part (2), the h_0-coexponent of \frac{\overline{v_1^{24}}}{h_0} is 12 by explicit computation [IWX22b]. For k \geq 1, we use that the Adams periodicity operator P^4 of degree (32,16) induces an isomorphism on H^{**}\mathcal{A}^\mathrm{cl} in the relevant range.
For part (3), the h_0-coexponents of \frac{\overline{v_1^{48}}}{h_0} and \frac{\overline{v_1^{80}}}{h_0} are 28 by explicit computation [IWX22b] [BFCC23]. For k \geq 2, we use that the Adams periodicity operator P^8 of degree (64,32) induces an isomorphism on H^{**}\mathcal{A}^\mathrm{cl} in the relevant range.
The h_0-coexponent of \frac{\overline{v_1^{96}}}{h_0} is 58.
This specific case is explicitly computed in [BFCC23].
Unlike Lemma 3.3, Lemma 3.4 only considers a single coexponent rather than a periodic family. The problem is that the generator \frac{\overline{v_1^{96}}}{h_0^{58}} lies slightly outside of the range in which (some power of) Adams periodicity is known to be an isomorphism. If we could verify that the h_0-coexponent of \frac{\overline{v_1^{64+96}}}{h_0} is 58 in the 319-stem, then we could use P^{16} periodicity to establish that the h_0-coexponent of \frac{\overline{v_1^{64k+96}}}{h_0} is 58 for all k \geq 0.
Analogous to Remark 3.1, the elements \frac{\overline{w_1^{4k}}}{h_1^n} are related to each other via the Massey products that express w_1-periodicity. Specifically, we have
with no indeterminacy for all j and k.
More generally, assume that n \leq m. For each choice of \frac{\overline{w_1^{4j}}}{h_1^m} and each choice of \frac{\overline{w_1^{4k}}}{h_1^n}, the Massey product
has no indeterminacy and equals an element that is a possible choice for \frac{\overline{w_1^{4j+4k}}}{h_1^n}. Beware that the unique element in the Massey product may depend on the choices of \frac{\overline{w_1^{4j}}}{h_1^m} and of \frac{\overline{w_1^{4k}}}{h_1^n}.
For each fixed value of j \geq 3, the h_0-submodules cogenerated by \displaystyle \frac{\overline{v_1^{3 \cdot 2^j + i \cdot 2^{j+1}}}}{h_0} have the same h_0-coexponent for all i.
For j \geq 3, what is the h_0-coexponent of \frac{\overline{v_1^{3 \cdot 2^j}}}{h_0}?
? describes some cyclic h_1-submodules of H^{***}\mathcal{A} in Chow degree zero.
We use the isomorphism of Theorem 2.1 from H^{**} \mathcal{A}^{\mathrm{cl}} to the Chow degree zero part of H^{***} \mathcal{A}. Apply this isomorphism to Lemma 3.2, Lemma 3.3, and Lemma 3.4, or directly to ?.
No element of the form \frac{\overline{w_1^{4k}}}{h_1^n} or \frac{\overline{w_1^{4k+2}}}{h_1^n} is hit by an Adams differential.
All of these elements have Chow degree zero. The Adams d_r differential increases the Chow degree by r-1. The \mathbb{C}-motivic Adams E_2-page is zero in negative Chow degrees by Proposition 2.3, so all differentials entering Chow degree zero are null.
No element of the form \frac{\overline{w_1^{4k}}}{h_1^n} or \frac{\overline{w_1^{4k+2}}}{h_1^n} is the target of a hidden \eta-extension.
Each element under consideration has Chow degree zero, so it has a degree of the form \left( s,f,\frac{s+f}{2} \right) for some s and f. If that element were the target of a hidden \eta extension, then the source of the extension would have degree \left( s-1, f', \frac{s+f}{2} -1 \right), where f' < f-1. Then the Chow degree of this source would be f' - f + 1, which is negative. The \mathbb{C}-motivic Adams E_2-page is zero in negative Chow degrees by Proposition 2.3.
No element of the form \frac{\overline{w_1^{4k}}}{h_1^n} or \frac{\overline{w_1^{4k+2}}}{h_1^n} is the target of a (hidden or non-hidden) \tau-extension.
Each element under consideration has Chow degree zero, so it has a degree of the form \left( s,f,\frac{s+f}{2} \right) for some s and f. If that element were the target of a \tau extension, then the source of the extension would have degree \left( s, f', \frac{s+f}{2} +1 \right), where f' \leq f. (Note that f' = f precisely when the \tau-extension is not hidden.) Then the Chow degree of this source would be f' - f - 2, which is negative. The \mathbb{C}-motivic Adams E_2-page is zero in negative Chow degrees by Proposition 2.3.
The element h_1 h_n is a permanent cycle in the \mathbb{C}-motivic Adams spectral sequence.
Using [Mahowald77], we know that h_1 h_n is a permanent cycle in the \tau-localization of the \mathbb{C}-motivic Adams spectral sequence. It is possible that \tau^m h_1 h_n supports differentials for small values of m, but there exists a smallest N such that \tau^N h_1 h_n is a permanent cycle. Let \alpha be the element in \pi_{2^n, 2^{n-1}+1-N} that is detected by \tau^N h_1 h_n.
The framework of [Isa19, Chapter 6] tells us that if an element of \pi_{s,w} is not divisible by \tau, then it is detected in filtration 2w-s in the Adams–Novikov spectral sequence. In particular, \alpha is detected in filtration 2-2N in the Adams–Novikov spectral sequence. The only possibility is that N = 0.
Proposition 4.5 is not an independent proof of the existence of Mahowald’s \eta_n family. Rather, we use Mahowald’s classical result to extend to the \mathbb{C}-motivic context.
For all n \geq 1, there are non-zero elements \gamma_n in \pi_{20n-2,12n-1} such that:
\gamma_n is detected by \frac{\overline{w_1^{4n}}}{h_1} in filtration 4n in the \mathbb{C}-motivic Adams spectral sequence.
\gamma_n is annihilated by \eta.
the elements form a w_1-periodic family in the sense that \gamma_{n+1} is contained in the Toda bracket \langle \gamma_n, \eta, \gamma_1 \rangle for all n \geq 1.
\gamma_{2^j} equals \eta^{2^{j+2}-2} \cdot \eta_{j+4}, where \eta_{j+4} is the Mahowald element in the 2^{j+4}-stem.
the \eta-coexponent of \gamma_n is given in ? for some values of n.
The proofs of (1), (2), and (3) proceed by induction. Let \gamma_1 be an element of \pi_{18,11} that is detected by \frac{\overline{w_1^4}}{h_1}. We already know that \gamma_1 is annihilated by \eta [IWX20].
Inductively define \gamma_{n+1} to be an element in the Toda bracket
The bracket is well-defined by the induction hypothesis. The Moss convergence theorem [Mos70] then tells us that \gamma_{n+1} is detected by
Beware that the Moss convergence theorem requires that there are no crossing Adams differentials. In this case, there are no crossing differentials because the sources of such differentials would have negative Chow degree, and there are no non-zero elements in negative Chow degree by Proposition 2.3.
To show that \gamma_{n+1} is annihilated by \eta, consider the shuffle
This last expression equals \nu \gamma_n \gamma_1 by [Toda62, Theorem 3.6], which equals zero because \nu \gamma_1 is zero [IWX20].
For part (4), we use Proposition 4.5 together with the observation that \frac{\overline{w_1^{2^{j+2}}}}{h_1} equals h_1^{2^{j+2}-2} \cdot h_1 h_{j+4}. This last claim is a translation along Theorem 2.1 of the observation that \frac{\overline{v_1^{2^{j+2}}}}{h_0} equals h_0^{2^{j+2}-2} \cdot h_0 h_{j+3} in the classical Adams E_2-page.
Finally, we must study the \eta-coexponent of \gamma_n. Proposition 4.5 shows that h_1 h_{j+1} is a permanent cycle. Therefore, \frac{\overline{w_1^{2^{j-1}}}}{h_1} has h_1-coexponent equal to 2^{j-1} - 1 in the Adams E_\infty-page. Lemma 4.3 rules out hidden \eta extensions that could increase this coexponent in homotopy.
Similarly, Proposition 7.8 and Proposition 7.26, to be proved later, show that \frac{\overline{w_1^{8k+12}}}{h_1} and \frac{\overline{w_1^{16k+24}}}{h_1} have h_1-coexponents equal to 6 and 12 respectively in the Adams E_\infty-page. Lemma 4.3 rules out hidden \eta extensions that could increase this coexponent in homotopy.
The proofs of parts (1), (2), (3), and (4) of Theorem 4.7 are self-contained. Only the proof of part (5) depends on results that appear later in this manuscript. We emphasize this point to clarify the absence of circular arguments. Also, parts (1), (2), and (3) were previously proved by Andrews [And].
Beware that the Toda brackets in part (3) of Theorem 4.7 may have indeterminacies.
In part (5) of Theorem 4.7, the \eta-coexponent of \gamma_n is unknown for many values of n. The smallest such value is n = 12. There are two potential sources for this uncertainty. First, we have incomplete knowledge about h_1-coexponents in the \mathbb{C}-motivic Adams E_2-page, as discussed in ?. Second, even if we did know more about the Adams E_2-page, it is possible that the presence of non-zero of Adams differentials could affect the \eta-coexponent of \gamma_n.
For all n \geq 1, there are non-zero elements \delta_n in \pi_{20n+9,12n+6} such that:
\delta_n is detected by \overline{w_1^{4n+2}} in filtration 4n+3 in the \mathbb{C}-motivic Adams spectral sequence.
\delta_n is annihilated by \eta.
the elements form a w_1-periodic family in the sense that \delta_{n+1} is contained in the Toda bracket \langle \delta_n, \eta, \gamma_1 \rangle for all n.
the \eta-coexponent of \delta_n is 2.
Identically to the proof of Theorem 4.7, the proofs of (1), (2), and (3) proceed by induction. Let \delta_1 be an element of \pi_{29,18} that is detected by \overline{w_1^6}. We already know that \delta_1 is annihilated by \eta [IWX20].
Inductively define \delta_{n+1} to be an element in the Toda bracket
The bracket is well-defined by the inductive hypothesis. The Moss convergence theorem [Mos70] then tells us that \delta_{n+1} is detected by
Beware that the Moss convergence theorem requires that there are no crossing Adams differentials. In this case, there are no crossing differentials because the source of such differentials would have a negative Chow degree.
To show that \delta_{n+1} is annihilated by \eta, consider the shuffle
This last expression equals \nu \delta_n \gamma_1 by [Toda62, Theorem 3.6], which equals zero because \nu \gamma_1 is zero [IWX20].
Finally, we must study the \eta-coexponent of \delta_n. Proposition 7.3 and Proposition 7.5 show that h_1 h_3 g^k is a permanent cycle, but h_3 g^k is not. Therefore, \overline{w_1^{4n+2}} has h_1-coexponent equal to 2 in the Adams E_\infty-page. Lemma 4.3 rules out hidden \eta extensions that could increase this coexponent in homotopy.
The proofs of parts (1), (2), and (3) of Theorem 4.11 are self-contained. Only the proof of part (4) depends on results that appear later in this manuscript. We emphasize this point to clarify the absence of circular arguments. Also, parts (1), (2), and (3) were previously proved by Andrews [And].
In Theorem 4.11, one can define \delta_0 to be an element in \pi_{9,6} that is detected by \overline{w_1^2}, which is more commonly known as h_1^2 h_3. However, this element is qualitatively different. Because of the presence of h_1 c_0 in higher filtration, there is a hidden \eta extension, and \delta_0 is not annihilated by any power of \eta. Rather, in coweight 3, the element \sigma of \pi_{7,4} generates the \eta-submodule \mathbb{Z}[\eta]/(16, 2 \eta), and \delta_0 equals \eta^2 \sigma. In a sense, the \eta-coexponent of \delta_0 equals 3 because it can be divided twice by \eta. However, the product \eta \delta_0 is not zero, so \delta_0 is not a cogenerator.
Also, \delta_1 cannot be constructed as the Toda bracket \left\langle \delta_0, \eta, \gamma_1 \right\rangle because the bracket is not well-defined.
Note that \mathcal{B} is similar to the construction \mathcal{P}^{\mathrm{mot}} of [BX23, Construction 2.7]. The difference is that we do not include \tau in \mathcal{B}.
Recall from [Benson24, Proposition 15] that the Burklund–Xu differentials are H^{***}\mathcal{B}-linear, in the sense that
for all y in the E_r-page and all x in H^{***}\mathcal{B}.
There is an isomorphism from the \mathbb{C}-motivic Cartan–Eilenberg E_2-page in Chow degree one to the part of H^{***} \mathcal{A} in Chow degree one.
As in the proof of Theorem 2.1 just above, \mathbb{C}-motivic Cartan–Eilenberg differentials take values in multiples of \tau. Therefore, the values of such differentials have Chow degree of at least 2.
The element q_2 \cdot h_0^2 has degree (3,3) = (3,1) + (0,2). This degree converts back to (8,3,5) in the traditional motivic grading, and in fact the element q_2 \cdot h_0^2 represents c_0.
There is also a formula involving threefold Massey products for the d_2 differential [Benson24, Proposition 14] [BX23, Lemma 4.10]. For i \geq 2 we have
where the Massey product occurs in H^{**}\mathcal{A}^{\mathrm{cl}}. The well-definedness of the bracket is assured by the assumption that d_1(q_i \cdot x) = 0, and the indeterminacy is zero in the E_2-page because it lies in the image of the d_1 differential.
Similarly, there are formulas involving higher order Massey products for the higher differentials. Specifically, for i \geq r we have
See [BX23, Lemma 4.10] for an explicit statement of the formula when r = 3. The earlier differentials take care of the well-definedness and indeterminacy of the bracket. These formulas are algebraic analogues of the homotopical Toda bracket formulas of [CF17]. We do not discuss the proofs of these higher differential formulas because we do not need them.
The element q_n of the Burklund–Xu spectral sequence has degree (2^n-1,1), so its v_1-intercept is 2^n-3.
The v_1-intercept is multiplicative in the sense that the v_1-intercept of xy is equal to the sum of the v_1-intercepts of x and y.
[Ada66] The cohomology of the classical Steenrod algebra is concentrated in degrees with v_1-intercept greater than or equal to -3, with the exception of the elements h_0^k in degree (0,k) for k \geq 2.
? lists all elements in the cohomology of the classical Steenrod algebra whose v_1-intercepts equal -2, 0, or 1.
The degrees under consideration lie within the region in which the Adams periodicity operator P of degree (8,4) is an isomorphism [Ada66] [Li20]. The result follows from explicit low-dimensional computations in degrees up to (12,6).
? lists all elements of the Burklund–Xu E_1-page in Chow degree one with v_1-intercept equal to -1.
We are searching for expressions of the form q_n \cdot x such that x belongs to H^{**} \mathcal{A}^{\mathrm{cl}}, and q_n \cdot x has v_1-intercept -1. Since q_n has v_1-intercept 2^n-3, our desired x must have v_1-intercept 2 - 2^n.
If n \geq 3, then 2 - 2^n \leq -6. By Proposition 6.3, x must be zero or of the form h_0^j. More specifically, the elements q_n \cdot h_0^{2^{n-1} - 1} are the only elements of the form q_n \cdot x such that n \geq 3, the v_1-intercept is -1, and the Chow degree is 1.
Next, consider elements of the form q_2 \cdot x. Then x has v_1-intercept -2 since q_2 has v_1-intercept 1. By Lemma 6.4, x must be h_0 or of the form P^k h_1^2 for some k \geq 0.
Now consider elements of the form q_1 \cdot x. Then x has v_1-intercept 0 since q_1 has v_1-intercept -1. By Lemma 6.4, x must be 1.
Finally, consider elements of the form q_0 \cdot x. Then x has v_1-intercept 1 since q_0 has v_1-intercept -2. By Lemma 6.4, x must be of the form P^k h_2, P^k h_0^2 h_3, or P^k h_1 c_0 for some k \geq 0.
Consider the Burklund–Xu E_1-page in Chow degree one and v_1-intercept -7. Under these restrictions, the only elements are q_n \cdot h_0^{2^{n-1} + 2} in degree (2^n - 1, 2^{n-1} + 3) for n \geq 1.
We are searching for expressions of the form q_n \cdot x with x in H^{**}\mathcal{A}^{\mathrm{cl}}, such that q_n \cdot x has v_1-intercept -7. Since q_n has v_1-intercept 2^n-3, our desired x must have v_1-intercept -4 - 2^n.
If n \geq 1, then -4-2^n \leq -6. By Proposition 6.3, x can only be a power of h_0.
For k \geq 1, the elements q_0 \cdot P^k h_1^2 in degree (8k+2, 4k+3) are non-zero permanent cycles in the Burklund–Xu spectral sequence.
The elements q_0 \cdot P^k h_1^2 are in the lowest possible Burklund–Xu filtration, and the Burklund–Xu d_r differential decreases the filtration by r. Therefore, these elements are permanent cycles.
To prove that q_0 \cdot P^k h_1^2 cannot be the target of a differential, we need to rule out all possible differentials that might hit those elements. The degree of q_0 \cdot P^k h_1^2 is (8k+2, 4k+3), and the sources of possible differentials have degree (8k+3, 4k+2). The v_1-intercept of such elements is -1. Lemma 6.5 classifies elements with v_1-intercept -1. Of those elements, only the elements q_0 \cdot P^k h_2 lie in a degree of the form (8k+3,4k+2) with k \geq 1. The Burklund–Xu filtration of q_0 \cdot P^k h_2 is 0, so it cannot support a differential.
The condition k \geq 1 in Lemma 6.7 is necessary. If k = 0, then there is an element q_2 \cdot h_0 in degree (3,2) = (8k+3, 4k+2). In fact, there is a differential
As in Remark 5.10, this differential is equivalent to the Massey product
Therefore, q_0 \cdot h_1^2 is zero in the E_\infty-page.
For all k \geq 1, the product h_0 h_2 \cdot h_2 g^k in degree (20k + 6, 4k + 3) is non-zero in H^{***} \mathcal{A}.
The condition k \geq 1 in Proposition 6.9 is necessary. When k = 0, the product h_0 h_2 \cdot h_2 is in fact zero. This corresponds to the differential discussed in Remark 6.8.
? shows some infinite families of non-zero Massey products in the cohomology of the \mathbb{C}-motivic Steenrod algebra. All of the indeterminacies are zero.
For the first five families, all of the relevant elements have Chow degree zero. The Massey products are analogous to the classical non-zero Massey products
under the isomorphism of Theorem 2.1 from H^{**} \mathcal{A}^{\mathrm{cl}} to the Chow degree zero part of H^{***} \mathcal{A}.
For the last family, we analyze the indeterminacy first. The indeterminacy of the bracket consists of linear combinations of expressions of the form h_0 h_2 \cdot h_2 g^k \cdot x, where x has degree (60,12,36); and expressions of the form y \cdot h_1^3 i_1, where y has degree (20k+10, 4k+5, 12k+7). There are no non-zero elements in degree (60,12,36).
We now consider elements of degree (20k+10, 4k+5, 12k+7). Note that the Chow degree of such elements is 1, so our earlier analysis of the Burklund–Xu spectral sequence in ? is relevant. After converting to the degrees that we used in that spectral sequence, we are looking for elements of degree (8k+3, 4k+5). The v_1-intercept of such elements is -7. Lemma 6.6 classifies elements of the Burklund–Xu E_1-page with v_1-intercept -7, but none of those elements has a degree of the form (8k+3, 4k+5) with k \geq 1. Therefore, there are no non-zero elements of degree (20k+10, 4k+5, 12k+7), and the Massey product has no indeterminacy.
Finally, we must determine the unique element that is contained in the Massey product. We have
where the first equality holds because there is no indeterminacy, and the last equality is the second family of Massey products in ?. Proposition 6.9 shows that the Massey product is non-zero.
The k=0 case of the last family in ? behaves differently. In that case, we have the Massey product \langle 0, h_1^3, h_1^3 i_1 \rangle in degree (66,15,40) because h_0 h_2^2 = 0. In that degree, there is a relation
So the non-zero element h_0 h_2 \cdot h_2 g^3 does belong to the bracket, but the indeterminacy makes this fact not useful.
For all k \geq 1, there is a non-zero \mathbb{C}-motivic Adams d_2 differential
The proof is by induction. The base cases are 1 \leq k \leq 3. These cases follow from explicit low-dimensional computation [IWX20].
There is a “higher Leibniz rule” [Mos70, Theorem 1.1] that describes how the Adams d_2 differential interacts with Massey products in the Adams E_2-page. Applied to the first family of Massey products in ?, the rule says that d_2(h_3 g^{k+3}) belongs to
The latter two terms consist of the multiples of h_3 g^k and of h_1^3 i_1 in the appropriate degree, and we must determine these multiples.
First, we look for expressions of the form h_3 g^k \cdot x, where x has degree (59,14,36). By inspection, x must be zero.
Second, we look for expressions of the form y \cdot h_1^3 i_1, where y has degree (20k+10, 4k+5, 12k+7). We already considered this situation in the proof of Proposition 7.1, and we found that y must be zero.
We have now determined that d_2(h_3 g^{k+3}) belongs to \langle d_2(h_3 g^k), h_1^3, h_1^3 i_1 \rangle. By induction, this bracket equals \langle h_0 h_2 \cdot h_2 g^k, h_1^3, h_1^3 i_1 \rangle, which equals h_0 h_2 \cdot h_2 g^{k+3} as shown in ?.
The use of g^3-periodicity, rather than g-periodicity, is a curious feature of the proof of Proposition 7.3. The key point is that there are no elements in H^{***}\mathcal{A} in degree (59,14,36). Note that the expression h_1^2 e_0 g^2 would have degree (59,14,36) if it existed, but it does not exist in H^{***}\mathcal{A}. See [KILRZ-e0gk] [Tha21] for information about the non-existence of this element.
If we had used g-periodicity, then the indeterminacy in our computations would have involved elements in degree (19,6,12). In this case, the element h_1^2 e_0 is present. That would result in indeterminacy in the brackets, and the argument would fail. In fact, the same problem occurs for g^{2^k}-periodicity, in which the presence of h_1^2 e_0 g^{2^k-1} creates indeterminacy problems. We chose to use g^3-periodicity because 3 is the smallest number that is not a power of 2.
The element h_1 h_3 g^k is a non-zero permanent cycle in the \mathbb{C}-motivic Adams spectral sequence for all k \geq 0.
Lemma 4.2 shows that h_1 h_3 g^k is non-zero if it does not support any differentials.
The proof is by induction. The case k= 0 is exceptional, but we know from explicit computation that h_1 h_3 is a permanent cycle. The base case k=1 says that h_1 h_3 g is a permanent cycle, which we also know from explicit computation [IWX20]. Let \alpha_1 be an element of \pi_{28,17} that is detected by h_1 h_3 g. Note that \eta^3 \alpha_1 is zero for degree reasons [IWX22b].
We inductively define \alpha_k to be an element of the Toda bracket \langle \alpha_{k-1}, \eta^3, \eta_4 \rangle. Part (2) of Theorem 4.11 ensures that \alpha_{k-1} \cdot \eta^2 is zero, so the bracket is well-defined.
The induction hypothesis is that \alpha_{k-1} is detected by h_1 h_3 g^{k-1}. The Moss convergence theorem [BK21, Theorem 1.1] implies that the Toda bracket \langle \alpha_{k-1}, \eta^3, \eta_4 \rangle is detected by the Massey product \langle h_1 h_3 g^{k-1}, h_1^3, h_1 h_4 \rangle, which equals h_1 h_3 g^k as shown in ?. The Moss convergence theorem requires that there be no “crossing differentials”. In our case, such crossing differentials would have sources with negative Chow degree, but there are no such elements by Proposition 2.3. Therefore, \alpha_k is detected by h_1 h_3 g^k.
Extending the notation of the proof, we may define \alpha_0 to be an element of \pi_{8,5} that is detected by h_1 h_3. However, \eta^3 \alpha_0 is non-zero, and it is detected by h_1^3 c_0. Consequently, the base case for the induction is \alpha_1, which is annihilated by \eta^3.
The elements h_1^2 h_3 g^k are also non-zero permanent cycles in the \mathbb{C}-motivic Adams spectral sequence. This follows immediately from Proposition 7.5 by multiplication with h_1. This family of elements was previously studied by Andrews [And].
The element i_1 g^{2k} is a non-zero permanent cycle in the \mathbb{C}-motivic Adams spectral sequence for all k \geq 0.
Lemma 4.2 shows that i_1 g^{2k} is non-zero if it does not support any differentials.
The proof proceeds by induction. The base case k=0 says that i_1 is a permanent cycle, which we know from explicit computation [IWX20]. Let \alpha_0 be an element of \pi_{53,30} that is detected by i_1. Note that \eta^7 \alpha_0 is zero for degree reasons [IWX22b].
We inductively define \alpha_k to be an element of the Toda bracket \langle \alpha_{k-1}, \eta^7, \eta_5 \rangle. Part (2) of Theorem 4.7 ensures that \alpha_{k-1} \cdot \eta^6 is zero, so the bracket is well-defined.
The induction hypothesis is that \alpha_{k-1} is detected by i_1 g^{2k-2}. The Moss convergence theorem [BK21, Theorem 1.1] implies that the Toda bracket \langle \alpha_{k-1}, \eta^7, \eta_5 \rangle is detected by the Massey product \langle i_1 g^{2k-2}, h_1^7, h_1 h_5 \rangle, which equals i_1 g^{2k} as observed in ?. The Moss convergence theorem requires that there be no “crossing differentials”. In our case, such crossing differentials would have sources with negative Chow degree, but there are no such elements by Proposition 2.3. Therefore, \alpha_k is detected by i_1 g^{2k}.
The elements h_1^m i_1 g^{2k} are also non-zero permanent cycles for 0 \leq m \leq 5. This follows from Proposition 7.8 by multiplication with powers of h_1.
(107, 13, 60) d_2 \left( \emph{\texttt{\{13-300\}}} \right) = 0.
By inspection, the only possible non-zero value for d_2 \left( \texttt{\{13-300\}} \right) is \texttt{\{15-313\}}, also known as h_1^6 h_6 e_0 g in degree (106,15,60). If there were a differential
then this differential would be detected by inclusion S \rightarrow S/\tau of the bottom cell of S/\tau, since \texttt{\{15-313\}} is not divisible by \tau. However, according to [IWX22], no such differential occurs in the Adams spectral sequence for S/\tau.
(107, 13, 60) d_3 \left( \emph{\texttt{\{13-300\}}} \right) = 0.
By inspection, the only possible non-zero value for d_3 \left( \texttt{\{13-300\}} \right) is \texttt{\{16-344\}}, also known as h_2 g^2 C' in degree (106,16,60). We will show that \tau \texttt{\{16-344\}} is non-zero on the E_3-page. This will rule out the possible differential since \texttt{\{13-300\}} is annihilated by \tau.
We know that \tau \texttt{\{16-344\}} is non-zero in the E_2-page, so we just need to show that it cannot be hit by a d_2 differential. The only possible source of such a differential is \texttt{\{14-306\}} in degree (107,14,59). Note that \texttt{\{14-306\}} equals h_1 \texttt{\{13-293\}}, but \tau \texttt{\{16-344\}} is not a multiple of h_1. Therefore, d_2 \left( \texttt{\{14-306\}} \right) cannot equal \tau \texttt{\{16-344\}}.
(107, 13, 60) d_6 \left( \emph{\texttt{\{13-300\}}} \right) does not equal \emph{\texttt{\{19-447\}}} nor \emph{\texttt{\{19-447\}}} + \emph{\texttt{\{19-449\}}}.
Inclusion S \rightarrow S/\tau of the bottom cell takes the element \texttt{\{19-447\}} to h_1^5 x_{101,14} on E_2-pages. The latter element is still non-zero on the E_6-page of the Adams spectral sequence for S/\tau, and it is not hit by a d_6 differential. Therefore, \texttt{\{19-447\}} cannot be hit by a d_6 differential in the Adams spectral sequence for S.
Similarly, inclusion of the bottom cell takes \texttt{\{19-447\}} + \texttt{\{19-449\}} to h_1^5 x_{101,14} + \tau M h_1 g^3. The latter term is hit by a d_2 differential in the Adams spectral sequence for S/\tau, so inclusion of the bottom cell takes \texttt{\{19-447\}} + \texttt{\{19-449\}} to h_1^5 x_{101,14} on Adams E_3-pages. The rest of the argument is identical to the previous paragraph.
(107, 17, 59) d_2 \left( \emph{\texttt{\{17-398\}}} \right) = 0.
The only possible non-zero value is \tau \texttt{\{19-449\}}. However, \texttt{\{17-398\}} is a multiple of h_3, while \tau \texttt{\{19-449\}} is not.
(100, 17, 56) d_2(\emph{\texttt{\{17-339\}}}) = \emph{\texttt{\{19-393\}}}.
This follows by comparison along the inclusion S \rightarrow S/\tau of the bottom cell of S/\tau. The d_2 differential is present in the Adams spectral sequence for S/\tau [IWX22].
(101, 14, 56) The element \emph{\texttt{\{14-268\}}} survives to the E_4-page.
There are no possible non-zero values for d_2( \texttt{\{14-268\}} ). There are also no possible non-zero values for d_3( \texttt{\{14-268\}} ) because \texttt{\{17-339\}} does not survive to the E_3-page by Lemma 7.14.
(107, 16, 60) The element \emph{\texttt{\{16-356\}}} survives to the E_4-page.
Note that \texttt{\{16-356\}} equals h_2^2 \texttt{\{14-268\}}, so the result follows from Lemma 7.15.
(105, 13, 57) The element \emph{\texttt{\{13-288\}}} survives to the E_5-page.
The only possible non-zero value for d_2( \texttt{\{13-288\}}) is \tau \texttt{\{15-293\}}. However, h_2 \cdot \tau \texttt{\{15-293\}} is non-zero, while h_2 \texttt{\{13-288\}} is zero.
There are no possible non-zero values for d_3( \texttt{\{13-288\}}) or d_4( \texttt{\{13-288\}}).
(107, 15, 59) The element \emph{\texttt{\{15-323\}}} survives to the E_5-page.
This follows from Lemma 7.17 and the relation \texttt{\{15-323\}} = h_1^2 \texttt{\{13-288\}}.
(107, 14, 59) d_2 \left( \emph{\texttt{\{14-306\}}} \right) = \emph{\texttt{\{16-352\}}}.
This follows by comparison along the inclusion S \rightarrow S/\tau of the bottom cell of S/\tau. The d_2 differential is present in the Adams spectral sequence for S/\tau [IWX22].
This shows that d_2 \left( \texttt{\{14-306\}} \right) equals either \texttt{\{16-352\}} or \texttt{\{16-352\}} + \tau \texttt{\{16-344\}}. The element \texttt{\{14-306\}} is a multiple of h_1, but \texttt{\{16-352\}} + \tau \texttt{\{16-344\}} is not a multiple of h_1. This rules out the latter possibility.
(106, 20, 60) d_2 \left( \emph{\texttt{\{20-485\}}} \right) = \emph{\texttt{\{22-548\}}}.
This follows by comparison along the inclusion S \rightarrow S/\tau of the bottom cell. The differential is present in the Adams spectral sequence for S/\tau [IWX22].
(107, 17, 61) d_2 \left( \emph{\texttt{\{17-391\}}} \right) = 0, and d_3 \left( \emph{\texttt{\{17-391\}}} \right) = \emph{\texttt{\{20-480\}}}.
Both differentials are detected by comparison along the inclusion S \rightarrow S/\tau of the bottom cell. The only possible non-zero value for d_2 \left( \texttt{\{17-391\}} \right) is \texttt{\{19-448\}}. If such a differential occurred, then it would be detected by a d_2 differential in the Adams spectral sequence for S/\tau. However, there is no such differential for S/\tau.
The only possible non-zero value for d_3 \left( \texttt{\{17-391\}} \right) is \texttt{\{20-480\}}. In this case, there is a corresponding d_3 differential in the Adams spectral sequence for S/\tau.
(106,19,59) The element \tau \emph{\texttt{\{19-449\}}} is non-zero in the Adams E_6-page.
By inspection of the E_2-page, there are several possible differentials that might kill \tau \texttt{\{19-449\}}. The table shows each possibility and gives the reason why that differential does not in fact occur.
(107, 13, 60) d_6 \left( \emph{\texttt{\{13-300\}}} \right) = 0.
Because of Lemma 7.12, it only remains to show that d_6 \left( \texttt{\{13-300\}} \right) cannot equal \texttt{\{19-449\}}. Lemma 7.22 shows that \tau \texttt{\{19-449\}} is non-zero in the E_6-page, while \tau \texttt{\{13-300\}} is zero.
(107, 13, 60) d_7 \left( \emph{\texttt{\{13-300\}}} \right) = 0.
In the E_2-page, there are two possible non-zero values \texttt{\{20-485\}} and \tau \texttt{\{20-480\}} for the differential. However, Lemma 7.20 and Lemma 7.21 show that these elements do not occur in the E_7-page.
(107, 13, 60) The element \emph{\texttt{\{13-300\}}} is a permanent cycle.
The previous lemmas have shown that \texttt{\{13-300\}} survives to the E_8-page. For higher differentials, all possible non-zero values disappear already in the E_5-page.
The element \Delta_1 h_1^2 i_1 g^{4k} is a permanent cycle in the \mathbb{C}-motivic Adams spectral sequence for all k \geq 0.
Lemma 4.2 shows that \Delta_1 h_1^2 i_1 g^{4k} is non-zero if it does not support any differentials.
The proof proceeds by induction. The base case k=0 says that \Delta_1 h_1^2 i_1 is a permanent cycle, which we know from Proposition 7.25. Let \alpha_0 be an element of \pi_{107,60} that is detected by \Delta_1 h_1^2 i_1. Note that \eta^{15} \alpha_0 is zero by part (2) of Theorem 4.7.
We inductively define \alpha_k to be an element of the Toda bracket \langle \alpha_{k-1}, \eta^{15}, \eta_6 \rangle. Part (2) of Theorem 4.7 ensures that \alpha_{k-1} \cdot \eta^{12} is zero, so the bracket is well-defined.
The induction hypothesis is that \alpha_{k-1} is detected by \Delta_1 h_1^2 i_1 g^{4k-4}. The Moss convergence theorem [BK21, Theorem 1.1] implies that the Toda bracket \langle \alpha_{k-1}, \eta^{15}, \eta_6 \rangle is detected by the Massey product \langle \Delta_1 h_1^2 i_1 g^{4k-4}, h_1^{15}, h_1 h_6 \rangle, which equals \Delta_1 h_1^2 i_1 g^{4k} as observed in ?. The Moss convergence theorem requires that there be no “crossing differentials”. In our case, such crossing differentials would have sources with negative Chow degree, but there are no such elements by Proposition 2.3. Therefore, \alpha_k is detected by \Delta_1 h_1^2 i_1 g^{4k}.
For all n \geq 1, there are non-zero elements a_n in degree (20n-2, 4n) in the Adams–Novikov E_2-page such that:
a_n is annihilated by \alpha_1.
the elements form a w_1-periodic family in the sense that a_{n+1} is contained in the Massey product \langle a_n, \alpha_1, a_1 \rangle for all n.
the element a_{2^j} equals \alpha_1^{2^{j+2} - 2} \cdot \beta_{2^{j+2}/2^{j+2} -1}.
the \alpha_1-coexponent of a_n is given in ? for some values of n.
According to Lemma 4.4, the elements of the form \frac{\overline{w_1^{4k}}}{h_1^n} in the Adams E_\infty-page detect elements in \mathbb{C}-motivic homotopy that are not divisible by \tau. Therefore, they map under inclusion of the bottom cell to non-zero elements in the homotopy of the cofiber of \tau. The homotopy of the cofiber of \tau is isomorphic to the Adams–Novikov E_2-page [Isa19, Theorem 6.12]. Parts (1), (2), and (4) are translations of Theorem 4.7 along this isomorphism. Part (3) follows from the fact that both the Adams E_\infty-page element h_1 h_{j+4} and Adams–Novikov E_\infty-page element \beta_{2^{j+2}/2^{j+2} -1} detect Mahowald’s \eta_{j+4}.
In part (4) of Corollary 8.1, the \alpha_1-coexponent of a_n is unknown for many values of n. The smallest such value is n = 12. See Remark 4.10 for further discussion of these uncertainties.
For all n \geq 1, there are non-zero elements b_n in degree (20n+9, 4n+3) in the Adams–Novikov E_2-page such that:
b_n is annihilated by \alpha_1.
the elements form a w_1-periodic family in the sense that b_{n+1} is contained in the Massey product \langle b_n, \alpha_1, a_1 \rangle for all n.
the \alpha_1-coexponent of b_n is 2.
Similarly to the proof of Corollary 8.1, these results are translations of Theorem 4.11 along the isomorphism between the Adams–Novikov E_2-page and the homotopy of the cofiber of \tau.
The \mathbb{C}-motivic homotopy elements \gamma_n and \delta_n of Theorem 4.7 and Theorem 4.11 are detected respectively by the elements a_n and b_n from Corollary 8.1 and Corollary 8.3 in the \mathbb{C}-motivic Adams–Novikov E_\infty-page.
When restricted to Chow degree zero, the cohomology of \mathbb{C}-motivic \mathcal{A}(2) is
The dual \mathcal{A}(2)_* of \mathcal{A}(2) is
Define \mathcal{B}(2)_* to be the subalgebra
of \mathcal{A}(2)_*, and note that the cobar complex of \mathcal{B}(2) is equal to the part of the cobar complex of \mathcal{A}(2) in Chow degree zero. Therefore, H^{***}\mathcal{B}(2) is isomorphic to the part of H^{***} \mathcal{A}(2) in Chow degree zero.
It remains to compute H^{***}\mathcal{B}(2). Observe that \mathcal{B}(2)_* is isomorphic to classical \mathcal{A}^{\mathrm{cl}}(1)_*, with shifted degrees. Consequently, we thoroughly know the cohomology of \mathcal{B}(2) by comparison to the well-known cohomology
of classical \mathcal{A}^{\mathrm{cl}}(1).
The E_\infty-page of the Burklund–Xu spectral sequence for the cohomology of \mathbb{C}-motivic A(2) in Chow degree one is displayed in ?.
Because \mathcal{A}(2)_* contains \tau_0, \tau_1, and \tau_2, but no \tau_i for i \geq 3, the E_1-page of the \mathbb{C}-motivic Burklund–Xu spectral sequence in Chow degree one takes the form
with differentials d_1(q_1) = q_0 \cdot h_0 and d_1(q_2) = q_1 \cdot h_1. Beware of the notation that we are using for the coefficients on q_n, as discussed carefully in ?.
The E_2-page now takes the form
where the kernels and images are taken within H^{**}\mathcal{A}^{\mathrm{cl}}(1). More explicitly, the E_2-page is
Note the curious fact that \frac{\ker(h_0)}{\operatorname{im}(h_1)} equals 0 in H^{**}\mathcal{A}^{\mathrm{cl}}(1), so the q_1-summand contributes nothing to the E_2-page. ? illustrates the E_2-page that we have just described symbolically.
Next, we must analyze d_2 differentials, which can only take the form d_2(q_2 \cdot x) = q_0 \cdot y. For degree reasons, there are a few but not many possible differentials. Namely, we have
As in Remark 5.10, these formulas correspond to Massey product computations
There are no further differentials, and we directly obtain the E_3-page, which equals the E_\infty-page.
In the \mathbb{C}-motivic Burklund–Xu spectral sequence for the cohomology of A(2) in Chow degree one, there are no hidden h_0 extensions. The only hidden h_1 extensions are from q_2 \cdot P^k a to q_0 \cdot P^{k+1}.
For degree reasons, there are no possible hidden h_0 extensions, and the only possible hidden h_1 extensions are from q_2 \cdot P^k a to q_0 \cdot P^{k+1}.
Recall the Massey product \langle a, h_1, h_1^2 \rangle = P h_0 in H^{**} \mathcal{A}^{\mathrm{cl}}(1). Then there is a similar Massey product \langle q_2 \cdot a, h_1, h_1^2 \rangle = q_2 \cdot P h_0 in the Burklund–Xu E_1-page. If this Massey product were well-defined in H^{***} \mathcal{A}(2), then the May convergence theorem [May69] would imply that q_2 \cdot P h_0 is a permanent cycle in the Burklund–Xu spectral sequence. But we already saw in the proof of Proposition 9.2 that q_2 \cdot P h_0 is not a permanent cycle. Therefore, the Massey product is not well-defined in H^{***} \mathcal{A}(2). The only possible problem is that there is a hidden h_1 extension on q_2 \cdot a, and q_0 \cdot P is the only possible target for this hidden extension.
The rest of the hidden extensions follow from the first one by P^k multiplication.
Converting from our Burklund–Xu spectral sequence notation to standard \mathbb{C}-motivic notation for H^{***} \mathcal{A}(2), the E_\infty-elements q_2 \cdot a and q_0 \cdot P detect e_0 and h_0 g respectively. The hidden h_1 extension from q_2 \cdot a to q_0 \cdot P corresponds to the relation h_2 \cdot e_0 = h_0 g. This relation is elementary from the perspective of the \mathbb{C}-motivic May spectral sequence [Isa09].
The text on this page is from arXiv:2504.14383v2 by Daniel C. Isaksen, Hana Jia Kong, Guchuan Li, Yangyang Ruan and Heyi Zhu, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.
The record of each item is in analysis/objects/2504/2504.14383.jsonl.