21.14.05 · number-theory / sieve-methods-large-sieve

Kloosterman Sums and the Kuznetsov Spectral Formula

shipped3 tiersLean: none

Anchor (Master): Petersson 1932 *Math. Ann.* 117, 453-537 (the Petersson trace formula via Poincaré series and the Fourier coefficients of holomorphic cusp forms); Kuznetsov 1980 *Mat. Sb.* 111(153), 334-383 (*Petersson's conjecture for cusp forms of weight zero and Linnik's conjecture; sums of Kloosterman sums* — the original Kuznetsov formula); Bruggeman 1978 *Invent. Math.* 45, 1-18 (the independently discovered Bruggeman-Kuznetsov formula via spectral theory of the Laplacian); Selberg 1965 *Proc. Sympos. Pure Math.* VIII, 1-15 (*On the estimation of Fourier coefficients of modular forms* — the eigenvalue bound $\lambda_1 \ge 3/16$ and the conjecture $\lambda_1 \ge 1/4$); Deshouillers-Iwaniec 1982 *Invent. Math.* 70, 219-288 (the large sieve for sums of Kloosterman sums and applications); Iwaniec 2002 *Spectral Methods of Automorphic Forms* (GSM 53); Kim-Sarnak 2003 appendix to Kim *J. Amer. Math. Soc.* 16 ($\lambda_1 \ge 975/4096 \approx 0.238$, the current record toward Selberg); Iwaniec-Kowalski 2004 *Analytic Number Theory* (AMS Colloquium 53) Ch. 16

Intuition Beginner

A single Kloosterman sum is a tangle of arithmetic: you pair each residue with its multiplicative inverse modulo $c$ and add up the resulting points on the unit circle. The Weil bound controls one such sum, but number theory almost never needs one. It needs long sums of Kloosterman sums, added up as the modulus $c$ ranges over a long stretch. When you add many of them, fresh cancellation can appear that no single bound predicts. The question is how to measure that hidden cancellation, and the answer turns out to live in a completely different world: the world of modular forms.

The bridge is a trace formula. On one side you have the arithmetic: a weighted sum of Kloosterman sums, each multiplied by a smooth bump function of $mn / c$ . On the other side you have the spectrum: the natural frequencies of a drum. The drum here is a curved surface, the modular surface, built by folding up the upper half-plane by the modular group. Just as a drumhead has a discrete list of pure tones, this surface has a discrete list of vibration modes, and these modes are exactly the modular forms and their non-holomorphic cousins. The trace formula says that the arithmetic sum and the spectral list are two faces of the same coin.

The payoff is two-way. Knowing the Weil bound for each Kloosterman sum tells you something about the drum's tones, and knowing the tones tells you about the long sums of Kloosterman sums. The deepest tone has a special value: a famous conjecture of Selberg predicts the lowest interesting frequency can never dip below a precise floor, and that floor is exactly what would give the sharpest possible cancellation in the arithmetic sums.

Visual Beginner

The modular surface drawn as a curved triangle with one corner stretched off to infinity (the cusp), beside a list of its vibration frequencies. A bracket connects the surface to a second column: a long sum of small circle-diagrams, one per modulus $c$ , each the Kloosterman sum $S (m, n; c)$ , weighted by a smooth bump that switches on when $mn / c$ is moderate and off when $c$ is very large or very small. An arrow labelled "trace formula" runs between the two columns in both directions, showing the spectral list on the left equals the weighted arithmetic sum on the right.

The picture captures the slogan of the unit: a weighted sum of Kloosterman sums equals a sum over the natural frequencies of the modular surface. Cancellation in the arithmetic is the same fact as positivity of those frequencies.

Worked example Beginner

Take the simplest trace-formula identity in disguise: the average of $S (n, 1; p)$ over $n$ modulo a prime $p$ . This is the toy version of "the spectral side has a clean main term." Use $p = 5$ .

Step 1. Recall the inverses modulo $5$ : $\overset{ˉ}{1} = 1$ , $\overset{ˉ}{2} = 3$ , $\overset{ˉ}{3} = 2$ , $\overset{ˉ}{4} = 4$ . The Kloosterman sum $S (n, 1; 5)$ adds the four values $e ((n x + \overset{x}{ˉ}) /5)$ for $x = 1, 2, 3, 4$ , where $e (t) = cos (2 π t) + i sin (2 π t)$ .

Step 2. Add these over all $n$ from $0$ to $4$ . Swap the order of addition: first fix $x$ , then add over $n$ . For each fixed $x$ , the factor $e (\overset{x}{ˉ} /5)$ comes out, and what remains is the inner total $e (0) + e (x /5) + e (2 x /5) + e (3 x /5) + e (4 x /5)$ .

Step 3. That inner total walks once around the circle in five equal steps, so it adds to $0$ whenever $x$ is not a multiple of $5$ . Every $x$ from $1$ to $4$ is such a value, so the whole double total over $n = 0, 1, 2, 3, 4$ collapses to $0$ .

Step 4. Now remove the single term $n = 0$ . There $S (0, 1; 5)$ adds $e (\overset{x}{ˉ} /5)$ over $x = 1, 2, 3, 4$ , which is $e (1/5) + e (3/5) + e (2/5) + e (4/5)$ . The four fifth-roots other than $1$ add to $- 1$ . So the total of $S (n, 1; 5)$ over $n = 1, 2, 3, 4$ is $0 - (- 1) = 1$ .

What this tells us: averaging a Kloosterman sum over its first argument produces the clean value $1$ , independent of $p$ . That stubborn constant is the shadow of the "identity" main term $δ_{mn}$ that anchors the spectral side of the Petersson and Kuznetsov formulas. The arithmetic side always carries one rigid piece, and everything else is genuine cancellation.

Check your understanding Beginner

Exercise (easy, multiple choice).

The Kuznetsov and Petersson formulas relate a weighted sum of Kloosterman sums to a sum over which kind of object?

A. Prime numbers below $p$ . B. The natural vibration frequencies (spectrum) of the modular surface — its modular and Maass forms. C. The divisors of $c$ . D. The roots of unity modulo $p$ .

Hint

A trace formula has an arithmetic side and a spectral side. The "drum" picture is the spectral side.

Answer

B. The trace formula equates a weighted sum of Kloosterman sums (the arithmetic, or geometric, side) with a sum over the spectrum of the hyperbolic Laplacian on the modular surface (the spectral side): the Fourier coefficients of an orthonormal basis of cusp forms together with the Eisenstein continuous spectrum. This dictionary is what converts arithmetic cancellation into a statement about modular forms.

Formal definition Intermediate+

Throughout, $H = {z = x + i y : y > 0}$ is the upper half-plane with hyperbolic measure $d μ (z) = y^{- 2} d x d y$ and Laplacian $Δ = - y^{2} (\partial_{x}^{2} + \partial_{y}^{2})$ , $Γ = SL_{2} (Z)$ , and $e (t) = e^{2 π i t}$ . For integers $m, n \geq 1$ and a modulus $c \geq 1$ the Kloosterman sum is $S (m, n; c) = \sum_{x mod c, (x, c) = 1} e ((m x + n \overset{x}{ˉ}) / c)$ as in 21.15.04, where $\overset{x}{ˉ} x \equiv 1 (mod c)$ .

Definition (Maass cusp form). A Maass cusp form for $Γ$ is a smooth function $u : H \to C$ that is $Γ$ -invariant, $u (γ z) = u (z)$ for all $γ \in Γ$ , is an eigenfunction of the Laplacian, $Δ u = λ u$ , is square-integrable on $Γ\ H$ , and vanishes at the cusp in the sense that the constant term $\int_{0}^{1} u (x + i y) d x = 0$ . Writing the eigenvalue as $λ = \frac{1}{4} + t^{2}$ with spectral parameter $t$ , each such $u$ has a Fourier-Whittaker expansion $$ u(z) = \sum_{n \ne 0} \rho_u(n), \sqrt{y}, K_{it}(2\pi |n| y), e(nx), $$ where $K_{i t}$ is the $K$ -Bessel function and the $ρ_{u} (n)$ are the Fourier coefficients of $u$ .

Definition (Petersson normalisation, holomorphic case). For an even weight $k \geq 12$ let ${f}$ be an orthonormal basis of the cusp-form space $S_{k} (Γ)$ for the Petersson inner product of 21.04.01, with Fourier expansions $f (z) = \sum_{n \geq 1} a_{f} (n) n^{(k - 1) /2} e (n z)$ (Hecke normalisation). Define the harmonic weight $ω_{f} = Γ (k - 1) / ((4 π)^{k - 1} ∥ f ∥^{2})$ .

Definition (the Bessel test pair). A trace formula transports a test function $φ$ on the geometric side to a transform $\overset{φ}{^}$ on the spectral side. For the Kuznetsov formula the relevant transform of a smooth, compactly supported $φ : (0, \infty) \to C$ is the Bessel (Selberg-Harish-Chandra) transform $$ \hat\varphi(t) = \frac{\pi i}{\sinh(\pi t)} \int_0^\infty \bigl(J_{2it}(x) - J_{-2it}(x)\bigr), \varphi(x), \frac{dx}{x}, $$ with $J_{ν}$ the $J$ -Bessel function; the holomorphic Petersson formula uses instead the single Bessel kernel $J_{k - 1}$ evaluated at $4 π mn / c$ .

Definition (spectral decomposition). The space $L^{2} (Γ\ H)$ decomposes as the orthogonal sum of the constants, the discrete space spanned by Maass cusp forms ${u_{j}}$ with eigenvalues $λ_{j} = \frac{1}{4} + t_{j}^{2}$ , and the Eisenstein continuous spectrum carried by the Eisenstein series $E (z, \frac{1}{2} + i t)$ for $t \in R$ , whose Fourier coefficients $φ (n, t)$ involve the divisor function and the Riemann zeta function. The Selberg eigenvalue conjecture asserts $λ_{1} \geq \frac{1}{4}$ for congruence subgroups, equivalently that every cuspidal spectral parameter $t_{j}$ is real.

Counterexamples to common slips

A holomorphic cusp form of weight $k$ is not a Maass form: it solves $\overset{ˉ}{\partial} f = 0$ , whereas a Maass form solves the second-order equation $Δ u = λ u$ and is real-analytic but generally not holomorphic. The Petersson formula lives on the holomorphic side, the Kuznetsov formula on the Maass side; they are parallel, not identical.
The eigenvalue $λ = 0$ belongs only to the constant function, which is excluded from the cuspidal spectrum. The first cuspidal eigenvalue $λ_{1}$ is strictly positive; the content of Selberg's conjecture is the sharp lower floor $1/4$ , not mere positivity.
The continuous spectrum is genuinely present and must appear on the spectral side. Dropping the Eisenstein contribution turns the trace identity into a false statement: the Eisenstein term is exactly what supplies the divisor-function main terms in many applications.
The harmonic weights $ω_{f}$ are not all equal; they weight each form by the inverse square of its Petersson norm. A bound "on average over $f$ " with these weights is not the same as a bound for an individual $f$ until the weights are removed, which requires the spectral large sieve.

Key theorem with proof Intermediate+

Theorem (Petersson trace formula). Let $k \geq 12$ be even and ${f}$ an orthonormal Hecke-normalised basis of $S_{k} (Γ)$ with harmonic weights $ω_{f}$ . Then for all integers $m, n \geq 1$ , $$ \sum_{f} \omega_f, \overline{a_f(m)}, a_f(n) ;=; \delta_{mn} ;+; 2\pi, i^{-k} \sum_{c \ge 1} \frac{S(m,n;c)}{c}, J_{k-1}!\left(\frac{4\pi\sqrt{mn}}{c}\right). $$

Proof. The proof computes the Fourier coefficients of a Poincaré series in two ways. For $m \geq 1$ define the weight- $k$ Poincaré series $$ P_m(z) = \sum_{\gamma \in \Gamma_\infty \backslash \Gamma} \overline{j(\gamma, z)}^{-k}, e\bigl(m, \gamma z\bigr), \qquad \Gamma_\infty = \left{\pm\begin{pmatrix}1 & b \ 0 & 1\end{pmatrix}\right}, $$ where $j (γ, z) = cz + d$ is the automorphy factor of 21.04.01. The series converges absolutely for $k > 2$ , lies in $S_{k} (Γ)$ , and represents the $m$ -th coefficient functional: for any $g \in S_{k} (Γ)$ with $g (z) = \sum_{n} b_{g} (n) n^{(k - 1) /2} e (n z)$ , the unfolding computation $$ \langle g, P_m \rangle = \int_{\Gamma_\infty\backslash\mathbb{H}} g(z), \overline{e(mz)}, y^k, d\mu(z) = \frac{\Gamma(k-1)}{(4\pi m)^{k-1}}, b_g(m), m^{(k-1)/2} $$ shows $⟨ g, P_{m} ⟩$ is a fixed constant times $b_{g} (m)$ . Expanding $g = P_{n}$ gives the spectral evaluation: $⟨ P_{n}, P_{m} ⟩$ equals that constant times the $m$ -th coefficient of $P_{n}$ .

Now compute the $m$ -th Fourier coefficient of $P_{n}$ geometrically by splitting the coset sum on the lower-left entry $c$ of $γ$ . The $c = 0$ cosets contribute the identity, giving the diagonal $δ_{mn}$ after normalisation. For each $c \geq 1$ the cosets are indexed by pairs $(a, d)$ modulo $c$ with $a d \equiv 1 (mod c)$ , and applying the Fourier inversion to $e (nγ z)$ over the resulting completed sum produces, by the classical Bessel integral $\int_{- \infty}^{\infty} (t^{2} + 1)^{- k /2} e^{- 2 π i α t - π k a r g (\dots)} d t$ evaluating to a $J$ -Bessel value, the term $S (m, n; c) c^{- 1} J_{k - 1} (4 π mn / c)$ . Assembling, the $m$ -th coefficient of $P_{n}$ is $δ_{mn}$ plus the Kloosterman-Bessel series. Equating the spectral evaluation (an expansion of $P_{n}$ in the orthonormal basis ${f}$ , whose coefficients are $ω_{f} \overline{a_{f} (n)}$ ) with the geometric coefficient yields the stated identity. $□$

Bridge. The Petersson formula is the foundational reason that the arithmetic of Kloosterman sums and the harmonic analysis of cusp forms are the same subject: the Poincaré series is at once a sum over a coset space (producing Kloosterman sums) and an element of a Hilbert space (producing Fourier coefficients), and equating its two readings is exactly the trace identity. This builds toward the Kuznetsov formula, where the holomorphic basis is replaced by the Maass spectrum and the single Bessel kernel $J_{k - 1}$ is replaced by the Bessel test pair, and the same Kloosterman-Bessel series appears again in the geometric side. This is exactly the pattern that generalises: a kernel summed over $Γ_{\infty} \Γ/ Γ_{\infty}$ has a geometric expansion in Kloosterman sums and a spectral expansion over automorphic eigenfunctions. The central insight is that the diagonal $δ_{mn}$ is the rigid main term and the Kloosterman series is the fluctuation, so that putting these together converts a question about cancellation in $\sum_{c} S (m, n; c) / c$ into a question about the size and density of the spectral parameters $t_{j}$ . The bridge is the Poincaré series, dual to the coefficient functional on one side and to the spectral projection on the other.

Exercises Intermediate+

Exercise 3 (medium, symbolic).

Using only the Weil bound $∣ S (m, n; c) ∣ \leq τ (c) c g cd (m, n, c)^{1/2}$ (Weil-Estermann form, with $τ$ the divisor function) and the small-argument bound $J_{k - 1} (x) ≪ x^{k - 1}$ , show that the geometric side of the Petersson formula, $\sum_{c \geq 1} c^{- 1} S (m, n; c) J_{k - 1} (4 π mn / c)$ , converges absolutely for $k \geq 4$ .

Hint

For large $c$ the argument $4 π mn / c$ is small, so $J_{k - 1} (4 π mn / c) ≪ (mn / c)^{k - 1}$ . Combine with $∣ S (m, n; c) ∣/ c ≪ τ (c) c^{- 1/2}$ .

Answer

For each $c$ , $∣ S (m, n; c) ∣/ c ≪ τ (c) c^{- 1/2}$ by the Weil bound, and $J_{k - 1} (4 π mn / c) ≪_{m, n} c^{- (k - 1)}$ since the argument is $O (1/ c)$ . Hence the $c$ -th term is $≪_{m, n} τ (c) c^{- 1/2 - (k - 1)} = τ (c) c^{- (k - 1/2)}$ . The series $\sum_{c} τ (c) c^{- (k - 1/2)}$ converges whenever $k - 1/2 > 1$ , i.e. $k \geq 2$ ; in particular for $k \geq 4$ the geometric side is absolutely convergent, since $\sum_{c} τ (c) c^{- s} = ζ (s)^{2}$ converges for $ℜ s > 1$ . Rubric: full credit for the per-term bound $τ (c) c^{- (k - 1/2)}$ and the $ζ (s)^{2}$ convergence criterion.

Exercise 4 (medium, symbolic).

Show that the diagonal main term $δ_{mn}$ in the Petersson formula is forced by the $c = 0$ contribution, by computing $⟨ P_{n}, P_{m} ⟩$ 's identity-coset piece. Specifically, explain why the coset of the identity contributes $e (n z)$ whose $m$ -th Fourier coefficient is $δ_{mn}$ .

Hint

The cosets in $Γ_{\infty} \Γ$ with lower-left entry $c = 0$ are exactly $Γ_{\infty}$ itself, i.e. the identity coset, on which $γ z = z + b$ and $j (γ, z) = 1$ .

Answer

The cosets $Γ_{\infty} \Γ$ with $c = 0$ form the single identity coset (since $a d = 1$ with $c = 0$ forces $a = d = \pm 1$ , absorbed by $Γ_{\infty}$ ). On this coset $γ$ acts by $z \mapsto z$ up to translation and $j (γ, z) = 1$ , so the corresponding term of $P_{n} (z)$ is $e (n z)$ . Its $m$ -th Fourier coefficient (in the $e (m z)$ basis) is $1$ if $m = n$ and $0$ otherwise, namely $δ_{mn}$ . Every coset with $c \geq 1$ contributes to the Kloosterman-Bessel series instead, because there $j (γ, z) = cz + d$ is non-constant and the unfolding produces $S (m, n; c)$ . Rubric: full credit for identifying the $c = 0$ coset as the identity and extracting $δ_{mn}$ .

Exercise 5 (medium, numeric).

The Linnik-Selberg conjecture predicts $\sum_{c \leq X} S (m, n; c) / c ≪_{m, n, ε} X^{ε}$ . The Weil bound alone gives only $\sum_{c \leq X} ∣ S (m, n; c) ∣/ c ≪ \sum_{c \leq X} τ (c) c^{- 1/2}$ . Estimate this Weil-only bound's order of magnitude in $X$ (ignoring logarithms).

Hint

$\sum_{c \leq X} τ (c) c^{- 1/2}$ is a partial sum of a series whose terms are about $c^{- 1/2}$ ; the partial sum grows like $X^{1/2}$ up to a logarithmic factor.

Answer

$X^{1/2}$ up to a logarithmic power. By partial summation $\sum_{c \leq X} τ (c) c^{- 1/2} ≍ X^{1/2} lo g X$ , since $\sum_{c \leq X} τ (c) ≍ X lo g X$ and the weight $c^{- 1/2}$ shifts the growth to $X^{1/2}$ . So term-by-term the Weil bound yields $≪ X^{1/2 + ε}$ , whereas Linnik-Selberg predicts $X^{ε}$ — a saving of a full power of $X^{1/2}$ from the extra cancellation among the Kloosterman sums that only the spectral side exposes. Rubric: full credit for $X^{1/2}$ and the comparison with the conjectured $X^{ε}$ .

Exercise 6 (hard, short-answer).

State the Kuznetsov formula in the form "spectral side $=$ geometric side" and explain precisely where an exceptional eigenvalue $λ_{j} = \frac{1}{4} - σ_{j}^{2} < \frac{1}{4}$ would enter, and why proving the Selberg conjecture $λ_{1} \geq 1/4$ removes those terms.

Hint

On the spectral side each cusp form contributes $\overset{φ}{^} (t_{j}) \overline{ρ_{j} (m)} ρ_{j} (n)$ . An exceptional eigenvalue makes $t_{j} = i σ_{j}$ imaginary, and the Bessel transform $\overset{φ}{^} (i σ_{j})$ can grow.

Answer

The Kuznetsov formula reads $$ \sum_j \hat\varphi(t_j), \overline{\rho_j(m)}\rho_j(n) + (\text{Eisenstein integral}) = \frac{\delta_{mn}}{\pi}!\int_0^\infty ! J\text{-terms} + \sum_{c\ge 1}\frac{S(m,n;c)}{c},\Phi!\left(\frac{4\pi\sqrt{mn}}{c}\right), $$ where $Φ$ is the Bessel transform of the test function and $\overset{φ}{^}$ its spectral image. For a real spectral parameter $t_{j}$ the factor $\overset{φ}{^} (t_{j})$ is bounded. An exceptional eigenvalue $λ_{j} < \frac{1}{4}$ forces $t_{j} = i σ_{j}$ with $0 < σ_{j} \leq 1/2$ , and then $\overset{φ}{^} (i σ_{j})$ can be exponentially large in the relevant ranges, so its term dominates and obstructs the cancellation one wants on the geometric side. Selberg's conjecture $λ_{1} \geq 1/4$ asserts no such $t_{j} = i σ_{j}$ occurs for congruence groups, removing every exceptional term and leaving only bounded oscillatory contributions; this is exactly the input that would upgrade the Linnik-Selberg bound to the conjectured $X^{ε}$ . Rubric: full credit for the spectral/geometric split, the location of the exceptional term, and the role of $λ_{1} \geq 1/4$ .

Exercise 7 (hard, short-answer).

Sketch how the equidistribution of Kloosterman fractions ${\overset{x}{ˉ} / c : 0 < x < c, (x, c) = 1}$ in $[0, 1]$ follows from cancellation in sums of Kloosterman sums, via Weyl's criterion.

Hint

Weyl's criterion: a sequence equidistributes iff every non-zero frequency exponential sum $\sum e (m \cdot (point))$ is $o (count)$ . The frequency- $m$ exponential sum over $\overset{x}{ˉ} / c$ is essentially $\sum_{c} S (m, 0; c)$ -type, a Ramanujan-Kloosterman average.

Answer

Fix $m \neq = 0$ . Weyl's criterion requires $\sum_{c \leq X} \sum_{(x, c) = 1} e (m \overset{x}{ˉ} / c) = o (\sum_{c \leq X} φ (c))$ . The inner sum $\sum_{(x, c) = 1} e (m \overset{x}{ˉ} / c)$ is the Kloosterman sum $S (m, 0; c)$ (a Ramanujan sum once one argument vanishes), so the criterion is a statement of cancellation in $\sum_{c \leq X} S (m, 0; c)$ . The Kuznetsov formula, applied with a test function localising $c \leq X$ , bounds this average by the spectral side, where the bound $λ_{1} > 0$ already gives a power saving over the unconditional size $\sum_{c} φ (c) ≍ X^{2}$ . Hence each non-zero Weyl sum is $o (X^{2})$ , and the Kloosterman fractions equidistribute. The quality of equidistribution (the discrepancy) sharpens as $λ_{1}$ approaches the Selberg bound $1/4$ . Rubric: full credit for the Weyl-criterion reduction, the identification of the Weyl sum with a Kloosterman average, and the spectral input.

Exercise 8 (hard, short-answer).

Explain why the Petersson formula (holomorphic weight $k$ ) does not by itself prove the Selberg eigenvalue conjecture, even though it is "the same" trace identity as Kuznetsov's.

Hint

The Petersson formula sees only holomorphic cusp forms, whose "eigenvalue" is the weight $k$ , fixed and large. The Maass spectrum with its small eigenvalues $λ_{j}$ is invisible to it.

Answer

The Petersson formula is an identity over the holomorphic cusp forms of a fixed weight $k$ . These correspond, in spectral language, to discrete-series representations whose Laplace parameter is pinned by $k$ and is large; they carry no information about the small Laplace eigenvalues $λ_{j}$ of Maass (non-holomorphic) cusp forms. The Selberg conjecture is a statement about precisely those Maass eigenvalues near $1/4$ , which appear only on the spectral side of the Kuznetsov formula for weight-zero automorphic forms. So although both formulas arise from the same Poincaré-series mechanism and share the Kloosterman-Bessel geometric side, only the Kuznetsov formula's spectral side contains the eigenvalues the conjecture constrains; the Petersson formula is silent about them. Rubric: full credit for distinguishing holomorphic discrete series (fixed weight) from the Maass spectrum and locating the conjecture on the latter.

Advanced results Master

Each trace formula is an instance of one mechanism: a kernel on $Γ\ H$ , summed over the double coset space, has a geometric expansion in Kloosterman sums and a spectral expansion over automorphic eigenfunctions. The named results below are the holomorphic, the Maass, and the inverted forms, together with the eigenvalue inputs that calibrate them.

Theorem (Kuznetsov-Bruggeman formula). Let $φ : (0, \infty) \to C$ be smooth and compactly supported, $\overset{φ}{^}$ its Bessel transform, and $m, n \geq 1$ . With ${u_{j}}$ an orthonormal basis of Maass cusp forms with spectral parameters $t_{j}$ and Fourier coefficients $ρ_{j} (\cdot)$ , and the Eisenstein contribution written as a $t$ -integral against $∣ φ (n, t) ∣^{2}$ -type coefficients, $$ \sum_j \frac{\hat\varphi(t_j),\overline{\rho_j(m)}\rho_j(n)}{\cosh(\pi t_j)} + \frac{1}{\pi}\int_{-\infty}^{\infty} \hat\varphi(t), \sigma_{m,n}(t), dt ;=; \frac{\delta_{mn}}{\pi^2}\int_0^\infty \varphi(x)\frac{dx}{x} ;+; \sum_{c\ge 1}\frac{S(m,n;c)}{c},\varphi!\left(\frac{4\pi\sqrt{mn}}{c}\right). $$ The transform $\overset{φ}{^}$ extends holomorphically in $t$ through the strip $∣ℑ t ∣ \leq 1/2$ that detects exceptional eigenvalues.

The formula is reversible: choosing $φ$ to localise $c$ near a scale $C$ extracts $\sum_{c ≍ C} S (m, n; c) / c$ as a spectral average, while choosing $φ$ to localise $t$ near a parameter $T$ extracts a count of Maass forms. This duality between the arithmetic scale $c$ and the spectral scale $t$ is the engine of all applications.

Theorem (Selberg, $λ_{1} \geq 3/16$ ). For every congruence subgroup $Γ_{0} (N)$ , the smallest positive Laplace eigenvalue satisfies $λ_{1} \geq 3/16$ , so every exceptional spectral parameter has $∣ℑ t_{j} ∣ \leq 1/4$ .

Selberg's argument bounds the Fourier coefficients of Maass forms by relating them, through the formula, to Kloosterman sums controlled by the Weil bound; the Weil exponent $\frac{1}{2}$ converts into the eigenvalue floor $3/16$ . The conjectured floor $1/4$ would follow from the Ramanujan-Petersson conjecture for Maass forms, equivalently from the temperedness of the corresponding automorphic representations of $GL_{2}$ .

Theorem (Kuznetsov, Linnik-Selberg toward). For fixed $m, n \geq 1$ , $$ \sum_{c \le X} \frac{S(m,n;c)}{c} ;\ll_{m,n,\varepsilon}; X^{1/6 + \varepsilon}. $$ The Linnik-Selberg conjecture predicts the exponent $ε$ in place of $1/6$ .

Kuznetsov's exponent $1/6$ already beats the Weil-only exponent $1/2$ of Exercise 5 by exposing the cancellation among different moduli through the spectral side. The conjectural $X^{ε}$ would follow from sufficiently strong spectral large-sieve inequalities together with the Selberg bound, but remains open.

Theorem (Deshouillers-Iwaniec spectral large sieve). For complex coefficients $(a_{n})$ and $T, N \geq 1$ , $$ \sum_j \frac{1}{\cosh(\pi t_j)} \Bigl|\sum_{n \le N} a_n, \rho_j(n)\Bigr|^2 ,\mathbf 1_{|t_j| \le T} ;\ll; (T^2 + N)\sum_{n\le N}|a_n|^2, $$ with analogous bounds for the holomorphic and Eisenstein spectra.

This is the spectral analogue of the classical large-sieve inequality 21.14.04: it removes the harmonic weights $ω_{f}$ and controls Fourier coefficients of cusp forms on average over the spectrum, and it is the workhorse behind bounds for moments of $L$ -functions, the linear and bilinear Kloosterman-sum estimates, and the Bombieri-Vinogradov theorem beyond the $1/2$ barrier.

Theorem (subconvexity input). The Kuznetsov formula converts a moment of $L$ -functions $\sum_{j} L (u_{j}, \frac{1}{2})^{2} \overset{φ}{^} (t_{j})$ into a sum of Kloosterman sums whose Weil-bound cancellation yields a subconvex bound $L (u_{j}, \frac{1}{2}) ≪ t_{j}^{1/3 + ε}$ in the spectral aspect, beating the convexity exponent $1/2$ .

The amplification method inserts a Dirichlet polynomial to isolate a single form, and the Kloosterman-sum geometric side, bounded by Weil, supplies the saving. This is the prototype for the entire subconvexity industry, where breaking $1/2$ in some aspect of the analytic conductor is the goal.

Synthesis. The unifying principle is that the Poincaré-series mechanism produces one identity with two expansions, and every theorem of the subject is read off by choosing which side to control. The foundational reason the Petersson, Kuznetsov, and Bruggeman formulas coincide in structure is that each computes the Fourier coefficients of a kernel summed over $Γ_{\infty} \Γ/ Γ_{\infty}$ : the geometric reading is the Kloosterman-Bessel series, the spectral reading is the eigenfunction expansion, and the diagonal $δ_{mn}$ is the rigid main term in both. This is exactly the dictionary in which the Weil bound for individual Kloosterman sums is dual to a bound on individual Fourier coefficients of cusp forms, and the spectral large sieve is dual to cancellation in long sums of Kloosterman sums. The central insight is the eigenvalue calibration: the Weil exponent $1/2$ feeds the formula and emerges as Selberg's $λ_{1} \geq 3/16$ , while the conjectural Ramanujan exponent $0$ would emerge as Selberg's $λ_{1} \geq 1/4$ and as the Linnik-Selberg cancellation $X^{ε}$ — three faces of temperedness.

Putting these together places the analytic theory of automorphic forms, the arithmetic of exponential sums, and the spectral geometry of the modular surface into a single frame, generalising from $SL_{2}$ to higher rank where the Kuznetsov formula becomes the relative trace formula of the Langlands program. The bridge is the Bessel transform pair, dual on one side to the test function localising the modulus $c$ and on the other to the spectral measure localising the parameter $t$ ; once that pair is in hand, equidistribution of Kloosterman fractions, subconvexity for $L$ -functions, and the large-sieve inequalities are corollaries of the single identity.

Full proof set Master

Theorem (Petersson trace formula), proof. Given in the Intermediate section: compute the inner product $⟨ P_{n}, P_{m} ⟩$ of weight- $k$ Poincaré series two ways. The spectral reading expands $P_{n}$ in an orthonormal basis ${f}$ of $S_{k} (Γ)$ , giving $\sum_{f} ω_{f} \overline{a_{f} (m)} a_{f} (n)$ . The geometric reading splits the coset sum on the lower-left entry $c$ : the $c = 0$ coset gives $δ_{mn}$ , and each $c \geq 1$ gives $S (m, n; c) c^{- 1} J_{k - 1} (4 π mn / c)$ via the Bessel integral. Equating the two yields the identity. $□$

Proposition (orthogonality of the Petersson formula recovers dimension). Setting $m = n$ and summing the diagonal recovers $dim S_{k} (Γ)$ asymptotically: $\sum_{f} ω_{f} ∣ a_{f} (n) ∣^{2} = 1 + O_{n} (1)$ termwise, and the harmonic-weighted count $\sum_{f} ω_{f}$ is $\frac{k - 1}{12} + O (1)$ .

Proof. Take $m = n$ in the Petersson formula: the diagonal is $1$ and the Kloosterman-Bessel series is $O_{n} (1)$ by Exercise 3's absolute convergence, so $\sum_{f} ω_{f} ∣ a_{f} (n) ∣^{2} = 1 + O_{n} (1)$ . To recover the dimension, note $ω_{f} = Γ (k - 1) / ((4 π)^{k - 1} ∥ f ∥^{2})$ and that $\sum_{f} ω_{f}$ equals the harmonic-weighted cardinality of an orthonormal basis. Comparing with the Petersson formula at $m = n = 1$ and using the crude Kloosterman estimate gives $\sum_{f} ω_{f} = dim S_{k} (Γ) \cdot \overline{ω} = \frac{k - 1}{12} + O (1)$ , since $dim S_{k} (Γ) = \frac{k}{12} + O (1)$ from the valence formula of 21.04.01 and the average harmonic weight is $≍ 1/ k$ . $□$

Proposition (Weil bound forces $λ_{1} \geq 3/16$ ). If every Kloosterman sum satisfies $∣ S (m, n; c) ∣ \leq τ (c) (m, n, c)^{1/2} c^{1/2}$ , then no cuspidal eigenvalue lies in $(0, 3/16)$ .

Proof. Insert a positive test function $φ$ into the Kuznetsov formula so that $\overset{φ}{^} (t) \geq 0$ for real $t$ and $\overset{φ}{^} (iσ)$ is large and positive for an exceptional parameter $t = iσ$ , $0 < σ \leq 1/2$ . The geometric side is then bounded, using the Weil bound and the small-argument Bessel estimate, by $δ_{mn} / π^{2} \cdot (const) + O (\sum_{c} τ (c) c^{- 1/2} φ (4 π mn / c))$ . Tracking the dependence of this bound on the localisation scale and comparing with the lower bound furnished by the single exceptional term $\overset{φ}{^} (iσ) \overline{ρ_{j} (m)} ρ_{j} (n) / cosh (π iσ)$ forces $σ \leq 1/4$ , equivalently $λ_{1} = \frac{1}{4} - σ^{2} \geq \frac{1}{4} - \frac{1}{16} = \frac{3}{16}$ . The exponent $1/4$ on $σ$ is the image of the Weil exponent $1/2$ on $c$ under the Bessel transform; improving the Kloosterman-sum bound to Ramanujan strength would push $σ \to 0$ and yield the Selberg floor $1/4$ . $□$

Proposition (real-valuedness of the spectral side). The cuspidal spectral sum $\sum_{j} \overset{φ}{^} (t_{j}) \overline{ρ_{j} (m)} ρ_{j} (n) / cosh (π t_{j})$ is real when $m = n$ and $φ$ is real.

Proof. For $m = n$ each summand is $\overset{φ}{^} (t_{j}) ∣ ρ_{j} (n) ∣^{2} / cosh (π t_{j})$ . The Bessel transform of a real $φ$ satisfies $\overset{φ}{^} (t_{j}) \in R$ for real $t_{j}$ (the kernel $J_{2 i t} - J_{- 2 i t}$ over $i sinh (π t)$ is real for real $t$ ), and $∣ ρ_{j} (n) ∣^{2} \geq 0$ , $cosh (π t_{j}) > 0$ . For an exceptional $t_{j} = i σ_{j}$ the same kernel remains real by the symmetry $J_{2 i t} \leftrightarrow J_{- 2 i t}$ . Hence every summand is real, and so is the sum. This reality is the structural counterpart of the reality of Kloosterman sums proved in 21.15.04. $□$

Theorem (Kuznetsov-Bruggeman formula), proof sketch via the resolvent. Let $G_{s} (z, w)$ be the resolvent kernel of $Δ$ on $Γ\ H$ , i.e. the Green's function for $Δ - s (1 - s)$ . Its spectral expansion is a sum over ${u_{j}}$ plus the Eisenstein integral; its geometric expansion is a sum over $γ \in Γ$ of an automorphic kernel $k_{s} (z, γ w)$ . Take the $(m, n)$ -th Fourier coefficient in both $z$ and $w$ : the spectral side produces $\sum_{j} ρ_{j} (m) \overline{ρ_{j} (n)}$ weighted by the resolvent eigenfactor $(s (1 - s) - λ_{j})^{- 1}$ , and the geometric side, after summing over the cosets with lower-left entry $c$ , produces $\sum_{c} S (m, n; c) / c$ times a Bessel-Whittaker integral. Inverting the resolvent eigenfactor by a contour integral against a test function $φ$ — equivalently applying the Selberg-Harish-Chandra transform — converts the eigenfactor into $\overset{φ}{^} (t_{j})$ and the Bessel-Whittaker integral into $φ (4 π mn / c)$ , yielding the stated formula. Bruggeman 1978 carried out the resolvent route; Kuznetsov 1980 used Poincaré series directly, paralleling the Petersson proof above with $K$ -Bessel in place of the holomorphic kernel. $□$

Connections Master

Gauss, Jacobi, Kloosterman, Salié sums and the Weil bound 21.15.04. This unit takes the Weil bound $∣ S (m, n; p) ∣ \leq 2 p$ as a black-box arithmetic input and feeds it into the geometric side of the trace formula; the reality and twisted multiplicativity of $S (m, n; c)$ proved there are exactly the properties that make the Kloosterman-Bessel series well-defined and the spectral side real. The eigenvalue floor $3/16$ is the trace-formula image of the Weil exponent $1/2$ .
Modular forms on $SL_{2} (Z)$ 21.04.01. The Petersson inner product, the cusp-form spaces $S_{k} (Γ)$ , the valence and dimension formulas, and the Fourier expansion of cusp forms are the holomorphic spectral side; the Poincaré series used in the proof is built from the automorphy factor introduced there, and the harmonic-weighted dimension count reuses the valence formula directly.
Hecke operators and the Hecke algebra 21.04.02. The orthonormal basis ${f}$ diagonalising the Petersson side is the Hecke-eigenform basis; multiplicativity of the coefficients $a_{f} (n)$ and the Euler product of the attached $L$ -function are what let the trace formula address moments of $L$ -functions, and the Ramanujan-Petersson conjecture for these eigenforms is the holomorphic counterpart of the Selberg eigenvalue conjecture for Maass forms.
The large sieve and bilinear forms 21.14.04. The Deshouillers-Iwaniec spectral large sieve is the automorphic-spectrum analogue of the classical large-sieve inequality of this chapter; it bounds Fourier coefficients of cusp forms on average and supplies the off-diagonal control behind Bombieri-Vinogradov-type theorems and bilinear Kloosterman-sum estimates.
Dirichlet $L$ -functions and the approximate functional equation 21.13.05. The subconvexity application of the Kuznetsov formula breaks the convexity exponent $1/4$ of the analytic conductor introduced there; the amplified second moment of $L$ -functions is opened by the approximate functional equation and closed by the Kloosterman-sum cancellation on the geometric side.

Historical & philosophical context Master

Petersson introduced his trace formula in 1932 ^{[Petersson 1932]} while computing the Fourier coefficients of holomorphic cusp forms through Poincaré series; the appearance of Kloosterman sums and Bessel functions in the coefficient formula was the first sign that the arithmetic of 21.15.04 and the analysis of modular forms were the same subject. Selberg, in his 1956 development of the trace formula for the Laplacian on $Γ\ H$ and his 1965 American Mathematical Society address ^{[Selberg 1965]}, proved the eigenvalue bound $λ_{1} \geq 3/16$ for congruence subgroups by routing the estimation of Fourier coefficients of Maass forms through the Weil bound for Kloosterman sums, and stated the conjecture $λ_{1} \geq 1/4$ that bears his name.

The non-holomorphic analogue — the formula relating sums of Kloosterman sums to the Maass spectrum — was found independently by Bruggeman in 1978 ^{[Bruggeman 1978]}, via the resolvent of the hyperbolic Laplacian, and by Kuznetsov in 1980 ^{[Kuznetsov 1980]}, via Poincaré series; Kuznetsov inverted the identity to prove the bound $\sum_{c \leq X} S (m, n; c) / c ≪ X^{1/6 + ε}$ , the first unconditional progress on Linnik's 1962 conjecture that this sum is $X^{ε}$ . Deshouillers and Iwaniec in 1982 ^{[Deshouillers-Iwaniec 1982]} turned the formula into the spectral large sieve, opening the systematic application to analytic number theory, including bounds beyond the $1/2$ -level in the Bombieri-Vinogradov range. The Selberg conjecture remains open; the current record $λ_{1} \geq 975/4096 \approx 0.238$ is due to Kim and Sarnak in 2003 ^{[Kim 2003]}, obtained from the symmetric-fourth functorial lift for $GL_{2}$ , which places the eigenvalue question inside the Langlands functoriality program where temperedness of automorphic representations is the governing principle.

Bibliography Master

@article{Petersson1932,
  author  = {Petersson, Hans},
  title   = {Über die Entwicklungskoeffizienten der automorphen Formen},
  journal = {Acta Mathematica},
  volume  = {58},
  pages   = {169--215},
  year    = {1932},
  doi     = {10.1007/BF02547776}
}

@article{Selberg1965,
  author    = {Selberg, Atle},
  title     = {On the estimation of {F}ourier coefficients of modular forms},
  journal   = {Proceedings of Symposia in Pure Mathematics},
  publisher = {American Mathematical Society},
  volume    = {VIII},
  pages     = {1--15},
  year      = {1965}
}

@article{Bruggeman1978,
  author  = {Bruggeman, Roelof W.},
  title   = {Fourier coefficients of cusp forms},
  journal = {Inventiones Mathematicae},
  volume  = {45},
  number  = {1},
  pages   = {1--18},
  year    = {1978},
  doi     = {10.1007/BF01406220}
}

@article{Kuznetsov1980,
  author  = {Kuznetsov, Nikolai V.},
  title   = {Petersson's conjecture for cusp forms of weight zero and {L}innik's conjecture; sums of {K}loosterman sums},
  journal = {Matematicheskii Sbornik},
  volume  = {111(153)},
  number  = {3},
  pages   = {334--383},
  year    = {1980}
}

@article{DeshouillersIwaniec1982,
  author  = {Deshouillers, Jean-Marc and Iwaniec, Henryk},
  title   = {Kloosterman sums and {F}ourier coefficients of cusp forms},
  journal = {Inventiones Mathematicae},
  volume  = {70},
  number  = {2},
  pages   = {219--288},
  year    = {1982},
  doi     = {10.1007/BF01390728}
}

@article{KimSarnak2003,
  author  = {Kim, Henry H.},
  title   = {Functoriality for the exterior square of {$\mathrm{GL}_4$} and the symmetric fourth of {$\mathrm{GL}_2$}},
  journal = {Journal of the American Mathematical Society},
  volume  = {16},
  number  = {1},
  pages   = {139--183},
  year    = {2003},
  note    = {With appendices by D. Ramakrishnan and by H. Kim and P. Sarnak},
  doi     = {10.1090/S0894-0347-02-00410-1}
}

@book{Iwaniec2002,
  author    = {Iwaniec, Henryk},
  title     = {Spectral Methods of Automorphic Forms},
  edition   = {2nd},
  series    = {Graduate Studies in Mathematics},
  volume    = {53},
  publisher = {American Mathematical Society},
  year      = {2002}
}

@book{IwaniecKowalski2004,
  author    = {Iwaniec, Henryk and Kowalski, Emmanuel},
  title     = {Analytic Number Theory},
  series    = {American Mathematical Society Colloquium Publications},
  volume    = {53},
  publisher = {American Mathematical Society},
  year      = {2004}
}

Prerequisites

21.15.04
21.04.01
21.04.02

Tier anchors

beginner: Iwaniec-Kowalski 2004 *Analytic Number Theory* (AMS Colloquium Publications 53) Ch. 16 opening; Iwaniec 2002 *Spectral Methods of Automorphic Forms* (Graduate Studies in Mathematics 53, 2nd ed.) Ch. 9 informal motivation for trace formulas; Sarnak's 1990 Stanford lectures *Some Applications of Modular Forms* (Cambridge Tracts 99) Ch. 1 on Kloosterman sums and equidistribution
intermediate: Iwaniec-Kowalski 2004 *Analytic Number Theory* (AMS Colloquium 53) §16.1-§16.5 (Petersson formula, Kuznetsov formula, sums of Kloosterman sums); Iwaniec 2002 *Spectral Methods of Automorphic Forms* (GSM 53) Ch. 3 (Maass forms and the spectral decomposition), Ch. 9 (the Kuznetsov-Bruggeman formula); Cogdell-Piatetski-Shapiro 1990 *The Arithmetic and Spectral Analysis of Poincaré Series* (Academic Press) Ch. 1-3
master: Petersson 1932 *Math. Ann.* 117, 453-537 (the Petersson trace formula via Poincaré series and the Fourier coefficients of holomorphic cusp forms); Kuznetsov 1980 *Mat. Sb.* 111(153), 334-383 (*Petersson's conjecture for cusp forms of weight zero and Linnik's conjecture; sums of Kloosterman sums* — the original Kuznetsov formula); Bruggeman 1978 *Invent. Math.* 45, 1-18 (the independently discovered Bruggeman-Kuznetsov formula via spectral theory of the Laplacian); Selberg 1965 *Proc. Sympos. Pure Math.* VIII, 1-15 (*On the estimation of Fourier coefficients of modular forms* — the eigenvalue bound $\lambda_1 \ge 3/16$ and the conjecture $\lambda_1 \ge 1/4$); Deshouillers-Iwaniec 1982 *Invent. Math.* 70, 219-288 (the large sieve for sums of Kloosterman sums and applications); Iwaniec 2002 *Spectral Methods of Automorphic Forms* (GSM 53); Kim-Sarnak 2003 appendix to Kim *J. Amer. Math. Soc.* 16 ($\lambda_1 \ge 975/4096 \approx 0.238$, the current record toward Selberg); Iwaniec-Kowalski 2004 *Analytic Number Theory* (AMS Colloquium 53) Ch. 16

References

Petersson, H. — Über die Entwicklungskoeffizienten der automorphen Formen · Acta Mathematica 58 (1932), 169-215; and Mathematische Annalen 117 (1932). Introduces the Poincaré series $P_m$ of weight $k$, computes their Fourier coefficients as a main term plus a Kloosterman-Bessel sum, and obtains the trace formula expressing $\sum_f \overline{a_f(m)}a_f(n)/\|f\|^2$ over an orthonormal basis of holomorphic cusp forms as $\delta_{mn}$ plus $\sum_c S(m,n;c)/c \cdot J_{k-1}(4\pi\sqrt{mn}/c)$.
Kuznetsov, N. V. — Petersson's conjecture for cusp forms of weight zero and Linnik's conjecture; sums of Kloosterman sums · Matematicheskii Sbornik 111(153) (1980), 334-383 (English: Math. USSR-Sb. 39 (1981), 299-342). Derives the trace formula for Maass cusp forms of weight zero, inverts it to bound sums of Kloosterman sums $\sum_{c \le X} S(m,n;c)/c \ll X^{1/6+\varepsilon}$, and applies the eigenvalue bound $\lambda_1 \ge 3/16$.
Bruggeman, R. W. — Fourier coefficients of cusp forms · Inventiones Mathematicae 45 (1978), 1-18. Independently establishes the spectral sum formula relating Kloosterman sums to the Laplace eigenvalues and Fourier coefficients of Maass forms on $\mathrm{SL}_2(\mathbb{Z})$, via the resolvent of the hyperbolic Laplacian and the spectral decomposition of $L^2(\Gamma\backslash\mathbb{H})$.
Selberg, A. — On the estimation of Fourier coefficients of modular forms · Proceedings of Symposia in Pure Mathematics VIII (American Mathematical Society, 1965), 1-15. Proves $\lambda_1(\Gamma_0(N)) \ge 3/16$ for congruence subgroups and conjectures the sharp bound $\lambda_1 \ge 1/4$ (the Selberg eigenvalue conjecture), equivalent to the absence of exceptional eigenvalues below the continuous-spectrum edge.
Deshouillers, J.-M. & Iwaniec, H. — Kloosterman sums and Fourier coefficients of cusp forms · Inventiones Mathematicae 70 (1982), 219-288. Develops the large sieve for Fourier coefficients of cusp forms via the Kuznetsov formula, obtaining mean-value and individual bounds on bilinear sums of Kloosterman sums and the spectral large sieve inequality used throughout analytic number theory.
Iwaniec, H. — Spectral Methods of Automorphic Forms · Graduate Studies in Mathematics 53 (American Mathematical Society, 2nd ed. 2002). Ch. 3 (Maass forms, spectral decomposition of $L^2(\Gamma\backslash\mathbb{H})$ into cusp forms, residual spectrum, and Eisenstein continuous spectrum); Ch. 9 (the Kuznetsov-Bruggeman trace formula, the Bessel transform pair, applications to Kloosterman-sum cancellation). The standard modern reference for the spectral side.
Iwaniec, H. & Kowalski, E. — Analytic Number Theory · American Mathematical Society Colloquium Publications 53 (2004). Ch. 16 develops the Petersson and Kuznetsov formulas, the spectral decomposition, the Bessel-Selberg-Harish-Chandra transform, the Linnik-Selberg conjecture, and the applications to equidistribution of Kloosterman fractions and subconvexity.
Kim, H. — Functoriality for the exterior square of $\mathrm{GL}_4$ and the symmetric fourth of $\mathrm{GL}_2$ · Journal of the American Mathematical Society 16 (2003), 139-183, with appendices by D. Ramakrishnan and by Kim-Sarnak. The Kim-Sarnak appendix deduces $\lambda_1 \ge 975/4096 \approx 0.2380$ for congruence subgroups, the current record toward Selberg's $\lambda_1 \ge 1/4$, from the symmetric-fourth functorial lift.

Estimated time

beginner: 18m
intermediate: 46m
master: 88m