The KPZ fixed point and Brownian motion share the same null sets Pantelis Tassopoulos, Sourav Sarkar Abstract . We show that the increments of the KPZ fixed point started from arbitrary initial data are mutually absolutely continuous with respect to Brownian motion with diffusion parameter 2 2 on compacts, extending the one-sided Brownian absolute continuity relation of the KPZ fixed point established in [ 29 ] . We also show that additive Brownian motion is absolutely continuous with respect to the centred Airy sheet on compacts, but it is not mutually absolutely continuous globally. As applications, we show that with probability strictly between zero and one, there exist record times of the KPZ fixed point away from any reference point, obtain a characterisation for the hitting probabilities of the graph of the KPZ fixed point to be positive in terms of a certain thermal capacity in the sense of [ 33 , 34 ] and compute essential suprema of Hausdorff dimensions of these random intersections. Finally, we compute essential suprema of Hausdorff dimensions of images of subsets in the plane under the Airy sheet and give a condition for the positivity of their Lebesgue measure in terms of Bessel-Riesz capacity. 2010 Mathematics Subject Classification 82B23 , 82C22 and 60H15 Figure 1 . Left : the centred KPZ fixed point started from flat initial data. Right : Brownian motion with diffusion parameter 2 . 1. Introduction In 1986, Kardar, Parisi and Zhang [ 18 ] predicted universal scaling behaviour for many planar random growth processes. Models in the KPZ universality class have a height function t\mapsto h_{t} , t\geq 0 which is conjectured to converge at large time and small length scales under the KPZ 1:2:3 scaling to a universal object called the KPZ fixed point, t\mapsto\mathfrak{h}_{t} , t\geq 0 . In [ 22 ] , Matetski-Quastel-Remenik constructed the KPZ fixed point as a Markov process in t , and they showed that it is a limit of the height function evolution of the totally asymmetric simple exclusion process with arbitrary initial condition. The natural domain of initial data for the KPZ fixed point is the space of upper semicontinuous functions satisfying a certain sub-parabolic growth condition. The KPZ fixed point at time t started from an admissible initial data h_{0} can also be expressed in terms of a variational formula with respect to a random metric on space-time \mathbb{R}^{4}_{\uparrow}=\{(x,s;y,t)\in\mathbb{R}^{4}:s0 , -\infty1 for the Radon-Nikodym derivative, a problem which remains open. By different means, the authors in [ 29 ] proved absolute continuity of the KPZ fixed point with respect to Brownian motion on compacts for general initial conditions, [ 29 , Theorem 1.2] . In [ 31 , 30 ] , we strengthened the above comparison by obtaining an explicit functional relationship between the law of the increments of the KPZ fixed point started from arbitrary initial data and Brownian motion on compacts. Whether the KPZ fixed point is mutually absolutely continuous with respect to Brownian motion on compacts, still remained open. Our main result in Theorem 3.7 of this paper settles the question of mutual absolute continuity of the KPZ fixed point started from arbitrary initial data against Brownian motion on compacts, by answering it in the affirmative. It crucially leverages the fact that the directed landscape at unit time, the Airy sheet, can be fully recovered as a deterministic function of the Airy line ensemble, [ 8 , Theorem 1.21] . In particular, we use a new coupling between the Airy sheet and the Airy line ensemble which extends the coupling in [ 6 ] . In Section 4 , we apply the techniques developed in Section 3 to the Airy sheet. In particular, we prove a non-disjointness result for the geodesics in the directed landscape in Proposition 4.1 . Moreover, we prove that the additive Brownian motion is absolutely continuous with respect to the Airy sheet (up to centering) on compact subsets of \mathbb{R}^{2} , Theorem 4.2 . We then discuss applications of the tools developed in Sections 3 and 4 . We revisit a certain notion of thermal capacity from [ 33 , 34 ] , [ 19 ] and use Theorem 3.7 to characterise when certain hitting probabilities of the KPZ fixed point are positive. Finally, using Theorem 4.2 , we compute essential suprema of their Hausdorff dimensions of images of compact subsets in the plane under the Airy sheet and give a condition for them to have positive Lebesgue measure using potential theory for additive Brownian motion, see for example [ 20 ] . \NoHyper Mutual absolute continuity of the KPZ fixed point started from finitary initial data Theorem 3.7 Mutual absolute continuity of the KPZ fixed point started from initial data with support bounded in at least one direction Mutual absolute continuity of the KPZ fixed point started from compactly supported initial data Brownian regularity of \mathcal{A}_{1} Ergodic properties of Airy sheet Geodesic geometry of the parabolic Airy line ensemble Theorem 3.6 and coupling ( 2.5 ) Theorem 3.5 Coupling 2.5 Thm 2.10 , coupling 2.5 and ( 2.5 ) Prop 3.1 , Lemma 3.4 \endNoHyper Figure 2 . Flowchart of main steps in the proof of Theorem 3.7 . 1.2. Organization of the paper First, in Section 1.2 we provide necessary background material including properties of last passage percolation, the Pitman transform and facts about Brownian bridges, ergodic properties of the Airy line ensemble, the Brownian regularity of the parabolic Airy 2 process, the symmetries of the Airy sheet as well as couplings thereof to the Airy line ensemble, ending the section with some background and setup for the KPZ fixed point. In Section 3 , we prove that for arbitrary initial data one obtains mutual absolute continuity of the laws of the spatial increments of the KPZ fixed point against rate two Brownian motion on compacts; the arguments leverage the construction of the entire Airy sheet as a deterministic function of the Airy line ensemble. Figure 2 shows the key steps of the proof of the main result, Theorem 3.7 . In Section 4 , we prove a non-disjointness result for the geodesics in the Airy sheet, Proposition 4.1 and that the additive Brownian motion is absolutely continuous with respect to the (centred) Airy sheet on compacts, Theorem 4.2 , but is not mutually absolutely continuous, Proposition 4.3 . We then discuss applications of the mutual absolute continuity result, Theorem 3.7 , and the absolute continuity result, Theorem 4.2 , in Section 5 . These pertain to record times of the KPZ fixed point, certain hitting probabilities for the graph of the KPZ fixed point in terms of a certain parabolic capacity and computations of L^{\infty} -norms of Hausdorff dimension of these (random) intersections. Finally, we use Theorem 4.2 and tools from potential theory to study geometric properties of the images of subsets of the plane under the Airy sheet, Corollary 4.4 , including computing the essential suprema of the Hausdorff dimensions of these random sets. Finally, in the Appendix, Section 6 , we prove a non-degeneracy result for geodesic jump times in the Airy line ensemble which we use in Section 3 . 1.3. Notation Let \mathscr{C}([a,b];\mathbb{R}^{d}) denote the space of all continuous functions f:[a,b]\mapsto\mathbb{R}^{d} , d\geq 1 and \mathscr{C}_{0}([a,b];\mathbb{R}^{d}) denote the space of all f\in\mathscr{C}([a,b];\mathbb{R}^{d}) with f(a)=0 . We say that a Brownian motion or a Brownian bridge has rate (or diffusion parameter) v if its quadratic variation in an interval [s,t] is equal to v(t-s) . From now on, all Brownian motions/bridges are rate two unless stated otherwise. Moreover, we denote by \mu the Wiener measure associated to a rate two Brownian motion starting from the origin on \mathscr{C}([0,\infty);\mathbb{R}) (or in a slight abuse of notation starting from any other point on the line). Finally, for a random variable Y on some probability space (\Omega,\mathscr{F},\mathbb{P}) , we will sometimes denote a version of the regular conditional distribution of \mathbb{P} given Y (whenever the latter exists, see [ 17 , Theorem 8.5] for sufficient conditions which will suffice in the present case), by \mathbb{P}_{Y}(\cdot)\equiv\mathbb{P}(\cdot|\sigma(Y)) . For sigma algebras \mathcal{A},\mathcal{B} on some set \Omega , we denote the minimal sigma algebra containing both \mathcal{A} and \mathcal{B} by \mathcal{A}\lor\mathcal{B} . 2. Preliminaries We first recall the definition of absolute continuity of measures on a measurable space (\Omega,\Sigma) . Definition 2.1 (Absolute continuity) . Let \mu,\nu denote measures on (\Omega,\Sigma) . Then, we say \mu is absolutely continuous with respect to \nu , written as \mu\ll\nu , if for all A\in\Sigma such that \nu(A)=0 , \mu(A)=0 . We say two measures \mu and \nu are mutually absolutely continuous if both \mu\ll\nu and \nu\ll\mu are satisfied. In what follows, a random line ensemble is a random variable taking values in an indexed (at most countably infinite) family of continuous paths defined on a common subset of \mathbb{R} . 2.1. Last passage percolation We begin with the collection of some preliminary facts regarding last passage percolation (LPP). For more details, see [ 6 , Section 2] . Formally, let I\subset\mathbb{Z} be a possibly finite index set and define the space \mathscr{C}^{I} of sequences of continuous functions with real domains, that is, the space of maps f:\mathbb{R}\times I\to\mathbb{R}\quad(x,i)\mapsto f_{i}(x) . Definition 2.2 (Path) . Let x\leq y\in\mathbb{R} , and m\leq\ell\in\mathbb{Z} respectively. A path , from (x,\ell) to (y,m) is a non-increasing function \pi:[x,y]\to\mathbb{N} which is cadlag on (x,y) and takes the values \pi(x)=\ell and \pi(y)=m . This also leads one to naturally define a derived quantity, namely the last passage value . Definition 2.3 (Length) . Let x\leq y\in\mathbb{R} and m0\,. Remark . In conjunction with Lemma 2.6 , Lemma 2.7 yields the corresponding result for Brownian motion. We now state informally a decomposition result for Brownian bridges, which is similar in spirit to the Lévy-Ciesielski construction of Brownian motion. Lemma 2.8 (Lemma 2.8 in [ 4 ] ) . Fix j\in\mathbb{N} , T>0 and consider a sequence of times 0=t_{0}} where \leq is the coordinate-wise partial order, and let g_{1},g_{2} be two bounded Borel measurable functions from [s,t]\to\mathbb{R}\cup\{-\infty\} such that g_{1}(x)\leq g_{2}(x) for all x\in[s,t] . For i=1,2 , let B^{i} be a k -tuple of Brownian bridges from (s,\underline{x}^{i}) to (t,\underline{y}^{i}) , conditioned to avoid each other and g_{i} . Then there exists a coupling such that B^{1}_{j}(r)\leq B^{2}_{j}(r) for all r\in[s,t],j\in\llbracket 1,k\rrbracket . For a sketch of a proof, see the proof of Lemmas 2.6 and 2.7 in [ 4 ] . For a more complete argument, see the proof of Lemma 2.15 in [ 10 ] . 2.4. Airy line ensemble and the Brownian Gibbs property The Airy line ensemble is a non-intersecting random sequence of continuous functions \mathcal{A}=(\mathcal{A}_{1},\mathcal{A}_{2},\dots) (see Theorem 2.1 in [ 6 ] ), such that \mathcal{A}_{1}>\mathcal{A}_{2}>\cdots . It was introduced by Prähofer and Spohn [ 26 ] in the version (\mathcal{A}^{\mathrm{stat}}_{i})_{i\in\mathbb{N}}\stackrel{{\scriptstyle\mathrm{def}}}{{=}}(\mathcal{A}_{i}(\cdot)+(\cdot)^{2})_{i\in\mathbb{N}} , which is stationary in time, see also [ 4 ] and [ 5 ] . We will thus call it the stationary Airy line ensemble . The top line \mathcal{A}_{1} is known as the parabolic \text{Airy}_{2} process that appears as the limiting spatial fluctuation of random growth models starting from a single point. We now recall the Brownian Gibbs resampling property enjoyed by the Airy line ensemble, first established in [ 4 ] . Informally, it states that for a0 . We collect some important properties of the Airy line ensemble and the Airy sheet, which will prove useful later. Recall that \mathcal{A} is the parabolic Airy line ensemble and that \mathcal{S} is the Airy sheet. In [ 8 ] , the Airy sheet was constructed as a deterministic function of the Airy sheet on the entire plane extending the coupling on the half-plane used in [ 6 ] . As a by-product, the Airy sheet \mathcal{S}(\cdot,\cdot) can be coupled with the (parabolic) Airy line ensemble \mathcal{A} so that \mathcal{S}(0,\cdot)=\mathcal{A}_{1}(\cdot) and almost surely for all (x,y,z)\in\mathbb{Q}\setminus\{0\}\times\mathbb{Q}^{2} , there exists a random integer K_{x,y,z} such that for all k\geq K_{x,y,z} , (recalling the definition for last passage values, Definition 2.4 ) \displaystyle\mathcal{S}(x,z)-\mathcal{S}(x,y) \displaystyle=\mathcal{A}[x_{k}\to(z,1)]-\mathcal{A}[x_{k}\to(y,1)]\,,\quad x>0 (2.3) \displaystyle\mathcal{S}(x,z)-\mathcal{S}(x,y) \displaystyle=\tilde{\mathcal{A}}[(-x)_{k}\to(-z,1)]-\tilde{\mathcal{A}}[(-x)_{k}\to(-y,1)]\,,\quad x<0\,, where x_{k}=(-\sqrt{k/2x},k) , x>0 and \tilde{\mathcal{A}}(\cdot)=\mathcal{A}(-\cdot) . Aside from this coupling, we will also need some global shape estimates for the Airy sheet. In particular, the Airy sheets satisfy almost sure pointwise bounds (2.4) |\mathcal{S}(x,y)+(x-y)^{2}|\leq\mathfrak{C}+c\log^{2/3}(2+|x|+|y|)\,,\qquad\text{ for all }x,y\in\mathbb{R} for some universal constant c>0 and some \mathfrak{C} satisfying \mathbb{E}[a^{\mathfrak{C}^{3/2}}]<\infty for some a>1 , [ 7 ] . Finally, we recall some properties of the Airy sheet, from Section 9 in [ 6 ] and Section 14 in [ 8 ] . More precisely, we have almost surely, that as a random continuous function in \mathbb{R}^{2} , the Airy sheet is translation invariant, that is, for any c\in\mathbb{R} , \mathcal{S}(\cdot+c,\cdot+c)\stackrel{{\scriptstyle\mathrm{d}}}{{=}}\mathcal{S}(\cdot,\cdot) and (2.5) \mathcal{S}(x,y)\stackrel{{\scriptstyle\mathrm{d}}}{{=}}\mathcal{S}(-y,-x)\,,\quad\mathcal{S}(x,y)\stackrel{{\scriptstyle\mathrm{d}}}{{=}}\mathcal{S}(-x,-y)\quad\mathrm{and}\quad\mathcal{S}(x,y)\stackrel{{\scriptstyle\mathrm{d}}}{{=}}\mathcal{S}(x,y+c)+2c(y-x)+c^{2}. Moreover, almost surely for all x\leq x^{\prime},\;y\leq y^{\prime} (2.6) \mathcal{S}(x,y)+\mathcal{S}(x^{\prime},y^{\prime})\geq\mathcal{S}(x,y^{\prime})+\mathcal{S}(x^{\prime},y). Since \mathcal{S}(x,y+\cdot)\stackrel{{\scriptstyle\mathrm{d}}}{{=}}\mathcal{S}(0,y-x+\cdot)=\mathcal{A}_{1}(y-x+\cdot) by the translation invariance and the coupling above, we now have, by Theorem 2.10 and the pointwise ergodic theorem for all x,y\in\mathbb{R}^{2} almost surely, \displaystyle\mathcal{S}(x,y)+(x-y)^{2} \displaystyle=\mathbb{E}[\mathcal{A}_{1}(0)]+\lim_{m\to\infty}\frac{1}{m}\sum_{m=1}^{\infty}(\mathcal{S}(x,y)-\mathcal{S}(x,y+m)+(x-y)^{2}-(x-y-m)^{2}) (2.7) \displaystyle=\mathbb{E}[\mathcal{A}_{1}(0)]+\lim_{m\to\infty}\frac{1}{m}\sum_{m=1}^{\infty}(\mathcal{S}(x,y)-\mathcal{S}(x,y-m)+(x-y)^{2}-(x-y+m)^{2})\,. Now we introduce some geodesic geometry on the Airy line ensemble. For x\leq y\in\mathbb{R} and \ell\in\mathbb{N} , we shall denote the rightmost geodesic between (x,\ell) and (y,1) in the Airy line ensemble \mathcal{A} by \pi[(x,\ell)\to y] (see Section 2 of [ 29 ] ). Next we recall the definition of infinite geodesics. Definition 2.12 . For any x\in\mathbb{R}^{+} and y\in\mathbb{R} with x_{k}=(-\sqrt{k/2x},k) , we define the geodesic \pi[x\to y] as the almost sure pointwise limit of \pi[x_{k}\to y] as k\to\infty , whenever the limit exists. We define the length of the geodesic \pi[x\to y] as \mathcal{S}(x,y) . Remark . The fact that these limits exist almost surely for all x,y in a countable dense set of \mathbb{R}^{+}\times\mathbb{R}^{2} is the content of [ 29 , Lemma 3.4] . More generally, the directed landscape \mathcal{L}:\mathbb{R}^{4}_{\uparrow}=\{(x,s;y,t)\in\mathbb{R}^{4}:s0 , recall the definition of t -finitary initial data. Definition 2.15 . ( t -finitary initial data) For t>0 , we denote by \mathscr{I}_{t} the set of locally bounded upper semicontinuous h_{0}\in\mathrm{UC} satisfying the sub-parabolic growth condition \displaystyle\lim_{|x|\to\infty}\frac{h_{0}(x)-x^{2}/t}{|x|}=-\infty\,. This condition on the initial data (for any t>0 fixed) is both necessary and sufficient to guarantee that the KPZ fixed point (at time t>0 ) does not explode, see [ 29 , Proposition 6.1] . Starting from admissible data h_{0}\in\mathscr{I}_{t} , t>0 the KPZ fixed point at time t>0 , \mathfrak{h}_{t}(\cdot,h_{0}) (or \mathfrak{h}(\cdot) when t,h_{0} are clear from the context), can be expressed in terms of a variational formula involving the directed landscape \mathcal{L} : (2.10) \mathfrak{h}_{t}(y,h_{0})=\max_{x\in\mathbb{R}}(h_{0}(x)+\mathcal{L}(x,0;y,t))\,,\quad y\in\mathbb{R}\,. We denote the law of \mathfrak{h}_{t}(y,h_{0})-\mathfrak{h}_{t}(a,h_{0})\,,y\in[a,b] for a0 (see Definition 2.15 ), one obtains mutual absolute continuity of the laws of the spatial increments of the KPZ fixed point against rate two Brownian motion on compacts. We crucially leverage the ‘full-space’ coupling between the Airy sheet and Airy line ensemble on \mathbb{R}^{2} , ( 2.5 ). Figure 2 shows the key steps of the proof of the main mutual absolute continuity result, Theorem 3.7 . We begin with a proposition regarding the monotonicity of some functionals of the Airy line ensemble and compactly supported initial data (appearing as boundary data in the variational characterisation of the KPZ fixed point; see the discussion in [ 30 , Section 5] ). First, recall from [ 29 , Theorem 3.7] the notation the semi-infinite last passage values for x>0 , \ell\geq 1 (3.1) \mathcal{A}[x\to(0,\ell)]\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\lim_{k\to\infty}(\mathcal{A}[x_{k}\to(0,\ell)]-\mathcal{A}[x_{k}\to(0,1)])+\mathcal{S}(x,0)\,, for x_{k}=(-\sqrt{k/2x},k) , k\geq 1 . By [ 29 , Lemma 3.8] , they are non-decreasing in \ell (hence finite) and by [ 29 , Lemma 3.9] \mathscr{F}_{-}\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\sigma(\{\mathcal{A}_{i}(x):x\leq 0,i=1,2,\cdots\}) -measurable. Proposition 3.1 . Fix finitary initial data h_{0}\in\mathrm{UC}_{c} with ‘max-plus’ support \mathrm{supp}_{-\infty}(h_{0})\subseteq(0,\infty) and denote the random ‘boundary data’ as G_{\ell}=\max_{x\in\mathbb{R}}(h_{0}(x)+\mathcal{A}[x\to(0,\ell)])\,. Then, almost surely, G_{m}>G_{m+1} for m\geq 1 . Moreover, if the initial data is compactly-supported such that \mathrm{supp}_{-\infty}(h_{0})\subseteq(0,\beta) for some \beta>0 , we obtain the almost sure uniform lower bounds for m>m^{\prime} , \displaystyle G_{m}-G_{m^{\prime}}\geq\mathcal{A}[(\varepsilon^{\infty}_{\beta,m^{\prime}},m^{\prime})\to(0,m)]-\mathcal{A}[(\varepsilon^{\infty}_{\beta,m^{\prime}},m^{\prime})\to(0,m^{\prime})]>0\,, where \varepsilon^{\infty}_{\beta,m^{\prime}} is the first time the semi-infinite geodesic \pi[\beta\to(0,m^{\prime})] reaches level m^{\prime} . Proof. We show the strict monotonicity for m=1 . The other cases (and corresponding lower bounds) follow analogously. We can express almost surely from [ 29 , Proposition 5.1] , G_{1}=\max_{x\in\mathbb{R}}(h_{0}(x)+\mathcal{S}(x,0)) , and G_{2}=\max_{x\in\mathbb{R}}(h_{0}(x)+\lim_{k\to\infty}(\mathcal{A}[x_{k}\to(0,2)]-\mathcal{A}[x_{k}\to(0,1)]+\mathcal{S}(x,0))\,, where x_{k}=(-\sqrt{k/(2x)},k) , x>0 , k\geq 1 . By monotonicity of geodesics, we have for any fixed 00 , 2\leq\ell\leq k , \varepsilon^{k}_{x,\ell} is the first time the geodesic from x_{k} to (0,2) reaches level \ell . Note for any fixed x>0 , as k\to\infty , \varepsilon^{k}_{x,\ell} eventually stabilises to the respective jump time of the semi-infinite geodesic on the Airy line ensemble. Now, by the metric composition law for last passage percolation, we have almost surely eventually in k\geq 1 , for any x\in\mathrm{supp}_{-\infty}(h_{0})\subseteq(0,\beta) (where we assume the support of h_{0} is countable and \beta=\sup\mathrm{supp}_{-\infty}(h_{0})<\infty ), \displaystyle\mathcal{A}[x_{k}\to(0,2)]-\mathcal{A}[x_{k}\to(0,1)] \displaystyle\leq\mathcal{A}[x_{k}\to(0,2)]-\mathcal{A}[x_{k}\to(\varepsilon^{k}_{\beta,2},2)]-\mathcal{A}[(\varepsilon^{k}_{\beta,2},2)\to(0,1)] \displaystyle=\mathcal{A}[x_{k}\to(\varepsilon^{k}_{\beta,2},2)]+\mathcal{A}[(\varepsilon^{k}_{\beta,2},2)\to(0,2)]\qquad\hfill(\varepsilon^{k}_{x,2}\leq\varepsilon^{k}_{\beta,2}) \displaystyle-\mathcal{A}[x_{k}\to(\varepsilon^{k}_{\beta,2},2)]-\mathcal{A}[(\varepsilon^{k}_{\beta,2},2)\to(0,1)]\,, where \varepsilon^{k}_{\beta,2}<0 almost surely, which follows from from the locally Brownian nature of the Airy line ensemble, see Lemma 6.1 in the Appendix. We thus have \displaystyle\mathcal{A}[x_{k}\to(0,2)]-\mathcal{A}[x_{k}\to(0,1)] \displaystyle\leq\mathcal{A}[(\varepsilon^{k}_{\beta,2},2)\to(0,2)]-\mathcal{A}[(\varepsilon^{k}_{\beta,2},2)\to(0,1)]\,. Taking k\to\infty we obtain the almost sure bounds \displaystyle G_{2} \displaystyle=\max_{x\in\mathbb{R}}(h_{0}(x)+\lim_{k\to\infty}(\mathcal{A}[x_{k}\to(0,2)]-\mathcal{A}[x_{k}\to(0,1)]+\mathcal{S}(x,0)) \displaystyle\leq G_{1}+\lim_{k\to\infty}(\mathcal{A}[(\varepsilon^{k}_{\beta,2},2)\to(0,2)]-\mathcal{A}[(\varepsilon^{k}_{\beta,2},2)\to(0,1)]) \displaystyle=G_{1}+\mathcal{A}[(\varepsilon^{\infty}_{\beta,2},2)\to(0,2)]-\mathcal{A}[(\varepsilon^{\infty}_{\beta,2},2)\to(0,1)]-\infty . Then, for every y\in\mathbb{R} , there exists some random open interval J_{y}\subseteq[y,\infty) (in the subspace topology of [y,\infty) ) such that, \mathfrak{h}_{t}(\cdot,h_{0})|_{J_{y}}-\mathfrak{h}_{t}(y,h_{0})\stackrel{{\scriptstyle\mathrm{law}}}{{=}}t^{\frac{1}{3}}(A(t^{-\frac{2}{3}}\;\cdot)|_{J^{\prime}_{y}}-A(t^{-\frac{2}{3}}y)\,, where A(\cdot) is the parabolic \mathrm{Airy}_{2} process and y\in J^{\prime}_{y}\subseteq[y,\infty) is another random open interval depending on the law of the Airy line ensemble, which contains y in its interior. Proof. By 1:2:3 scaling and translation symmetries of the Airy sheet, it suffices to take t=1 , \inf\mathrm{supp}_{-\infty}(h_{0})>0 and y>0 . We can now represent using the coupling ( 2.5 ) (cf. [ 29 , Proposition 5.1] ) \mathfrak{h}(s)=\mathfrak{h}_{1}(s,h_{0})=\max_{\ell\geq 1}(G_{\ell}+\mathcal{A}[(y,\ell)\to(s,1)])\,,\qquad s\geq y\,, with G_{\ell}\equiv\max_{x\in\mathbb{R}}\bigg(h_{0}(x)+\lim_{k\to\infty}(\mathcal{A}[x_{k}\to(y,\ell)]-\mathcal{A}[x_{k}\to(y,1)])+\mathcal{S}(x,y)\bigg)\,, where x_{k}=(-\sqrt{k/(2x)},k) , x>0 , k\geq 1 . Now, by the shape estimates of the Airy sheet, ( 2.4 ) (cf. [ 29 , Proposition 6.1] ), there exists a random m^{*}\in\mathbb{N} such that almost surely, we can represent the KPZ fixed point \mathfrak{h} on [y,y+1] as \mathfrak{h}(s)=\max_{\ell\geq 1}(G^{m^{*}}_{\ell}+\mathcal{A}[(y,\ell)\to(s,1)])\,,\qquad s\in[y,y+1]\,, where G^{m^{*}}_{\ell}=\max_{x\in[-m^{*},m^{*}]}(h_{0}(x)+\mathcal{A}[x\to(y,\ell)])\,,\quad\ell\geq 1\,. By [ 29 , Proposition 5.1] , there exists a random L_{0} such that we can represent the KPZ fixed point \mathfrak{h} on [y,y+1] as \displaystyle\mathfrak{h}(s) \displaystyle=\max_{1\leq\ell\leq L_{0}}(G^{m^{*}}_{\ell}+\mathcal{A}[(y,\ell)\to(s,1)])=G^{m^{*}}_{1}+\mathcal{A}_{1}(s)-\mathcal{A}_{1}(y) \displaystyle+\max_{y\leq r\leq s}\bigg(\max_{2\leq\ell\leq L_{0}}(G^{m^{*}}_{\ell}+\mathcal{A}[(y,\ell)\to(r,2)])-G^{m^{*}}_{1}-\mathcal{A}_{1}(r)+\mathcal{A}_{1}(y)\bigg)\lor 0\,,\quad s\in[y,y+1]\,. By Proposition 3.1 , we have G^{m^{*}}_{1}>G^{m^{*}}_{2} almost surely (by sectioning on events \{m^{*}=m\} , m\geq 1 ). Thus, since L_{0},m^{*} are uniform for s\in[y,y+1] , by continuity of last passage values and the monotonicity of the boundary data G^{m^{*}}_{\ell} (cf. Proposition 3.1 ), we have \lim_{s\to y}\max_{2\leq\ell\leq L_{0}}(G^{m^{*}}_{\ell}+\mathcal{A}[(y,\ell)\to(s,2)])=G^{m^{*}}_{2}0 and \alpha,y\in\mathbb{R} . Then, there exists a coupling of (\mathfrak{h}_{t}(\cdot,h_{0})\,:h_{0}\in\mathscr{I}_{t}\,,\inf\mathrm{supp}_{-\infty}(h_{0})>-\alpha) such that almost surely, for any two h_{0},h^{\prime}_{0} as above, there exists some random \tau_{y}=\tau_{y}(h_{0},h^{\prime}_{0})>y with \mathfrak{h}_{t}(\cdot,h_{0})|_{[y,\tau_{y}]}-\mathfrak{h}_{t}(y,h_{0})=\mathfrak{h}_{t}(\cdot,h^{\prime}_{0})|_{[y,\tau_{y}]}-\mathfrak{h}_{t}(y,h^{\prime}_{0})\,, that is the increments of the KPZ fixed points eventually coalesce. Figure 3 . Illustration of the coupling between the KPZ fixed points on [0,2] started from compactly supported flat 0 initial data on [0,1] , that is 0\cdot\delta_{[0,1]} (recall ( 2.9 )) ( green ) and the superposition of two narrow wedges at 0,1 , that is 0\cdot\delta_{\{0,1\}} ( blue ). We now prove a continuity result for the infinite last passage representation of the KPZ fixed point at the origin. This is the content of the following lemma. Lemma 3.4 . Fix finitary initial data with bounded ‘max-plus’ support \mathrm{supp}_{-\infty}(h_{0})\subseteq(0,\infty) and denote the random ‘boundary data’ as G_{\ell}=\max_{x\in\mathbb{R}}(h_{0}(x)+\mathcal{A}[x\to(0,\ell)]) , \ell\geq 1 (cf. ( 3.1 )). Then, for any k\geq 1 and t\in K , K\subseteq[0,\infty) compact, there exists a random almost surely finite L^{k}_{0} such that \max_{\begin{subarray}{c}k\leq\ell\end{subarray}}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,k)])=\max_{\begin{subarray}{c}k\leq\ell\leq L^{k}_{0}\end{subarray}}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,k)])\,,\quad\text{ for all }0\leq s\leq t\,. In particular, we have the continuity near the origin \lim_{t\searrow 0}\max_{\begin{subarray}{c}k\leq\ell\end{subarray}}(G_{\ell}+\mathcal{A}[(0,\ell)\to(t,k)])=G_{k}\,,\qquad k\geq 1\,. Proof. We first show that for any k\geq 1 and t\in K , K\subseteq[0,\infty) compact, there exists a random almost surely finite L^{k}_{0} such that \max_{\begin{subarray}{c}k\leq\ell\end{subarray}}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,k)])=\max_{\begin{subarray}{c}k\leq\ell\leq L^{k}_{0}\end{subarray}}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,k)])\,,\quad\text{ for all }0\leq s\leq t\,. The rest of the proof would follow by the continuity of last passage values (over the continuous environment). Indeed, fix k\geq 1,t\geq 0,K\subseteq\mathbb{R} as above. Also set x_{0}=\sup\mathrm{supp}_{-\infty}(h_{0})>0 and for x>0 , n\geq 1 , x^{n}_{0}=(-\sqrt{n/2x},n) . Now, by [ 29 , Lemma 3.6] , there exists a random Y\in\mathbb{N} such that the (rightmost) semi-infinite geodesic starting from (Y,1) , \pi[x,Y] satisfies \pi[x,Y](\sup K)\geq k . Moreover, for any fixed \ell\geq k , there exists a random N_{x,\ell}\in\mathbb{N} , such that the rightmost geodesics \pi[x,Y],\pi[x^{n}_{0},(Y,1)],\pi[x,(0,1)],\pi[x^{n}_{0},(0,1)] , pass through a common random point (T,d(T)) with T\leq 0 for all n\geq N_{x,\ell} . Now, by the ordering of geodesics, [ 29 , Proposition 2.8] , we have for any fixed x\in\mathrm{supp}_{-\infty}(h_{0}) , almost surely, \pi[x^{n},1]\leq\pi[x^{n},(\sup K,\ell)]\leq\pi[x^{n},Y]\leq\pi[x,Y] for all n\geq N_{x,\ell} . Hence, we have for all t\in K by the ordering of geodesics, almost surely, uniformly in t\in K \mathcal{A}[x^{n}\to(t,k)]=\max_{k\leq\ell\leq L^{k}_{0}}\mathcal{A}[x^{n}\to(0,\ell)]+\mathcal{A}[(0,\ell)\to(t,k)]\,,\qquad n\geq N_{x,\ell}\,, with L^{k}_{0}=\pi[x,Y](0)\leq\pi[x_{0},Y](0) (by [ 29 , Lemma 3.5] ). Thus, we have by [ 29 , Theorem 3.7] , for \ell\geq k , x\in\mathrm{supp}_{-\infty}(h_{0}) , \displaystyle\max_{k\leq\ell}(\mathcal{A}[x\to(0,\ell)]+\mathcal{A}[(0,\ell)\to(t,k)]) \displaystyle=\lim_{n\to\infty}\mathcal{A}[x^{n}\to(t,k)]-\mathcal{A}[x^{n}\to(t,1)] \displaystyle=\max_{k\leq\ell\leq L^{k}_{0}}\mathcal{A}[x\to(0,\ell)]+\mathcal{A}[(0,\ell)\to(t,k)]\,. Since the L^{k}_{0} is uniform in x , and the support of the initial data can always be taken to be countable, we conclude almost surely, for all t\in K , \max_{k\leq\ell}(G_{\ell}+\mathcal{A}[(0,\ell)\to(t,k)])=\max_{k\leq\ell\leq L^{k}_{0}}(G_{\ell}+\mathcal{A}[(0,\ell)\to(t,k)])\,, which concludes the proof. ∎ Remark . The above proof shows that one can make sense of the ‘infinite’ Skorokhod reflections \max_{\ell\geq m}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,m)])\,,\quad s\geq 0\,,m\geq 1 as random continuous functions on the entire path space \mathscr{C}(\mathbb{R};\mathbb{R}) , since on compacts by geodesic geometry, there is an almost surely finite random maximiser. This means we can represent the KPZ fixed point at unit time started from h_{0} , \mathfrak{h}_{1}(s,h_{0})=\max_{\ell\geq 1}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,1)])\,,\quad s\geq 0\,,m\geq 1\,. For more details, see [ 29 , Proposition 5.1] and [ 30 , Section 5] . We are now in a position to prove that the increments of the KPZ fixed point started from compact initial data are mutually absolutely continuous with respect to Brownian motion (see Figure 4 for an illustration). Figure 4 . Illustration of the Brownian Gibbs resampling for the top line of the Airy line ensemble appearing in the variational expression for the KPZ fixed point on the compact interval [0,r] . In short, it can be expressed (up to mutual absolute continuity) as a concatenation of a Brownian bridge W and Brownian motion B (conditionally independent given the Airy line ensemble) at the point (\varepsilon,\mathcal{A}_{1}(\varepsilon-\mathcal{A}_{1}(0)+G_{1}) , conditioned to not hit \mathcal{A}_{2} . The increments of the KPZ fixed point in the interior interval [\varepsilon,r] on the event \mathcal{A}_{1}(\cdot)-\mathcal{A}_{1}(0)+G_{1} avoids \max_{\ell\geq 2}(G_{\ell}+\mathcal{A}[(0,\ell)\to(\cdot,2)]) are simply the increments of the Brownian motion B . By standard monotonicity results for Brownian bridges, conditionally on B|_{[\varepsilon,r]} , the above non-intersection condition can always be ensured to occur with positive probability. Theorem 3.5 . Fix t>0 and let h_{0}\in\mathrm{UC}_{c} , that is, its max-plus support (cf. Definition 2.13 ) is bounded. Then, for any y,\in\mathbb{R} and r>0 , we have the (mutual) absolute continuity relation \nu^{h_{0}}\ll\mu\ll\nu^{h_{0}} , on paths on \mathscr{C}([y,y+r];\mathbb{R}) where \nu^{h_{0}} denotes the law of the increments of the KPZ fixed point started from h_{0} and \mu the appropriate restriction of the rate two Wiener measure. Remark . The above mutual absolute continuity result can be partially extended to the full Airy sheet, see Theorem 4.2 . Proof. By the local Brownianness of the KPZ fixed point, (cf. [ 29 ] ) it suffices to show \mu\ll\nu^{h_{0}} . By 1:2:3 -scaling, we can without loss of generality set t=1 . By the translation symmetries of the Airy sheet, we can also set y=0 , and fix the support of the initial data to lie in [0,\infty) . Now, suppose A\subseteq\mathscr{C}([\varepsilon,r];\mathbb{R}) , 0\leq\varepsilon\mathcal{A}_{2}(s) , for all s\in[\varepsilon,r] ( B has the law of the rate two Wiener measure \mu ) and \nu is the conditional law of the increments of \mathcal{A}_{1}(y)-\mathcal{A}_{1}(\varepsilon)\,,y\in[\varepsilon,r] . Hence, we have \displaystyle 0 \displaystyle=\mathbb{P}\bigg(B(\cdot)\in A\,,\max_{\ell\geq 2}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])\leq G_{1}+\mathcal{A}_{1}(s)-\mathcal{A}_{1}(0)\,,s\in[0,\varepsilon]\,, \displaystyle\max_{\ell\geq 2}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])\leq G_{1}+B(s)+\mathcal{A}_{1}(\varepsilon)-\mathcal{A}_{1}(0)\,,s\in[\varepsilon,r]\,, \displaystyle\mathcal{A}_{1}(\varepsilon)+B(s)>\mathcal{A}_{2}(s)\,,s\in[\varepsilon,r]\bigg)\,. Now, conditioning on the sigma algebra \mathscr{F}^{\prime}\equiv\sigma(\{\mathcal{A}_{i}(x):x\not\in(0,\varepsilon)\times\{1\}\} , and using the Brownian Gibbs property, we have as before \displaystyle 0 \displaystyle=\mathbb{P}\bigg(B(\cdot)\in A\,,\max_{\ell\geq 2}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])\leq G_{1}+W(s)\,,s\in[0,\varepsilon]\,, \displaystyle\mathcal{A}_{1}(0)+W(s)>\mathcal{A}_{2}(s)\,,s\in[0,\varepsilon]\,,\mathcal{A}_{1}(\varepsilon)+B(s)>\mathcal{A}_{2}(s)\,,s\in[\varepsilon,r]\,, \displaystyle\max_{\ell\geq 2}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])\leq G_{1}+B(s)+\mathcal{A}_{1}(\varepsilon)-\mathcal{A}_{1}(0)\,,s\in[\varepsilon,r]\bigg) \displaystyle=\mathbb{P}(B(\cdot)\in A\,,E)\,, where \displaystyle E \displaystyle\equiv\bigg\{\max_{\ell\geq 2}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])\leq G_{1}+W(s)\,,s\in[0,\varepsilon]\,, \displaystyle\mathcal{A}_{1}(0)+W(s)>\mathcal{A}_{2}(s)\,,s\in[0,\varepsilon]\,,\mathcal{A}_{1}(\varepsilon)+B(s)>\mathcal{A}_{2}(s)\,,s\in[\varepsilon,r]\,, \displaystyle\max_{\ell\geq 2}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])\leq G_{1}+B(s)+\mathcal{A}_{1}(\varepsilon)-\mathcal{A}_{1}(0)\,,s\in[\varepsilon,r]\bigg\}\,, W is a rate two Brownian bridge starting at (0,0) and ending at (\varepsilon,\mathcal{A}_{1}(\varepsilon)-\mathcal{A}_{1}(0)) (independent from B , see Figure 4 ). Now, by the Brownian Gibbs property and the fact that conditionally on \mathscr{F}^{\prime\prime}\equiv\sigma(\{\mathcal{A}_{i}(x):x\not\in(\varepsilon/2,r)\times\{1\}\}) , the stochastic domination holds (cf. Lemma 2.9 ): \mathcal{A}_{1}(\varepsilon)\geq_{d}\tilde{W}_{1}(\varepsilon) , where for two random variables X,Y , X\leq_{d}Y denotes the stochastic domination of X by Y and \tilde{W}_{1}(\cdot) has the law of a rate two Brownian bridge starting from (\varepsilon/2,\mathcal{A}_{1}(\varepsilon/2)) and ending at (r,\mathcal{A}_{1}(r)) . Moreover, we can express W as W_{0}+L , where W_{0} is a Brownian bridge vanishing at both endpoints and L an affine function with endpoints (0,0) and (\varepsilon,\mathcal{A}_{1}(\varepsilon)-\mathcal{A}_{1}(0)) . Now, conditionally on \sigma(B|_{[\varepsilon,r]})\lor\mathscr{F}^{\prime\prime} , we have by Lemma 2.7 and the above for any a\in\mathbb{R},\eta>0 , \displaystyle\mathbb{P}(\mathcal{A}_{1}(\varepsilon)>a\,,\left\lVert W_{0}\right\rVert_{L^{\infty}}\leq\eta|\sigma(B|_{[\varepsilon,r]})\lor\mathscr{F}^{\prime\prime}) \displaystyle=\mathbb{P}(\mathcal{A}_{1}(\varepsilon)>a|\sigma(B|_{[\varepsilon,r]})\lor\mathscr{F}^{\prime\prime})\cdot\mathbb{P}(\left\lVert W_{0}\right\rVert_{L^{\infty}}\leq\eta) \displaystyle\geq\mathbb{P}(\tilde{W}_{1}(\varepsilon)>a|\sigma(B|_{[\varepsilon,r])})\cdot\mathbb{P}(\left\lVert W_{0}\right\rVert_{L^{\infty}}\leq\eta)>0 almost surely. Moreover, conditioning on \sigma(B|_{[\varepsilon,r]})\lor\mathscr{F}^{\prime\prime} , we can treat the data (3.2) \bigg(B,\max_{\ell\geq 2}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)]),\mathcal{A}_{2},\mathcal{A}_{1}(0),G_{1}\bigg)\in\mathscr{C}([\varepsilon,r];\mathbb{R})\times\mathscr{C}^{2}([0,r];\mathbb{R})\times\mathbb{R}^{2} as fixed. Taking \mathcal{A}_{1}(\varepsilon) sufficiently large and \varepsilon>0 sufficiently small, which will depend on the data ( 3.2 ), we obtain \mathbb{P}(E\,|\sigma(B|_{[\varepsilon,r]})\lor\mathscr{F}^{\prime\prime})>0 almost surely. Since 0=\mathbb{P}(B(\cdot)\in A\,,E)=\mathbb{E}[\mathbf{1}(B(\cdot)\in A)\cdot\mathbb{P}(E\,|\sigma(B|_{[\varepsilon,r]})\lor\mathscr{F}^{\prime\prime})] , and \mathbb{P}(E\,|\sigma(B|_{[\varepsilon,r]})\lor\mathscr{F}^{\prime\prime})>0 , we finally obtain \mathbb{P}(B(\cdot)\in A)=0 , concluding the proof. ∎ Remark . Observe that the same argument in conjunction with Proposition 3.1 gives for any m\geq 1 , the laws of the increments of the ‘infinite’ Skorokhod reflections \max_{\ell\geq m}(G_{\ell}+\mathcal{A}[(0,\ell)\to(s,m)])\,,\quad s\geq 0 are mutually absolutely continuous with respect to the law of a rate two Brownian motion on paths on \mathscr{C}([\varepsilon,t];\mathbb{R}) , for any 0<\varepsilon0 we have the (mutual) absolute continuity relation \mu\ll\nu^{h_{0}}\ll\mu , on paths on \mathscr{C}_{0}(K;\mathbb{R}) where \nu^{h_{0}} denotes the law of the increments of the KPZ fixed point started from h_{0} at time t>0 in K . Proof. By the local Brownianness of the KPZ fixed point, (cf. [ 29 ] ) it suffices to show \mu\ll\nu^{h_{0}} . By 1:2:3 -scaling, we can without loss of generality set t=1 . Also, the translation and reflection symmetries of the Airy sheet allow us to set y>0 (by enlarging K if necessary), and by taking the supports to be bounded from the left, fix the support of the initial data to lie in (0,\infty) (since it is compact). Recall the notation for the sigma algebra \mathscr{F}_{-}=\sigma(\{\mathcal{A}_{i}(x):i\geq 1\,,x\leq 0\}) . Now, by the coupling 2.5 and the metric composition law ( 2.1 ), we can express the KPZ fixed point as (recall the notation for the Pitman transform ( 2.2 )), \mathfrak{h}(s)=\max_{\ell\geq 1}(G^{\infty}_{\ell}+\mathcal{A}[(0,\ell)\to(s,1)])=WF_{1}(s)\,,\quad s\in[y,y+r] with F=\bigg(G^{\infty}_{1}+\mathcal{A}_{1}(\cdot)-\mathcal{A}_{1}(0)\,,\max_{2\leq\ell}(G^{\infty}_{\ell}+\mathcal{A}[(0,\ell)\to(\cdot,2)])\bigg)\,, where G^{\infty}_{\ell}\equiv\max_{x\in\mathbb{R}}(h_{0}(x)+\mathcal{A}[x\to(0,\ell)]) and G^{m}_{\ell}\equiv\max_{x\in(m-1,m]}(h_{0}(x)+\mathcal{A}[x\to(0,\ell)])\,,\quad m\geq 1\,. Moreover, we have \max_{2\leq\ell}(G^{\infty}_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])=\max_{m\geq 1}\max_{2\leq\ell\leq L^{m}_{0}}(G^{m}_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])\,,\quad s\in[y,y+r] where L_{0}^{m} is the semi-infinite geodesic intercept \pi[m\to y+r](0) (cf. ( 2.12 )). We now show this maximum is attained for an almost surely finite m^{*}\geq 1 uniformly in s\in[y,y+r] . Moreover, m^{*} is measurable with respect to \mathscr{F}_{-} . Indeed, we have as m\to\infty , coupling the Airy line ensemble to an Airy sheet \mathcal{S}(\cdot,\cdot) using ( 2.5 ), \displaystyle\max_{2\leq\ell\leq L^{m}_{0}}(G^{m}_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)]) \displaystyle\leq\max_{1\leq\ell\leq L^{m}_{0}}(G^{m}_{\ell}+\mathcal{A}[(0,\ell)\to(s,1)]) \displaystyle\leq\max_{x\in(m-1,m]}(h_{0}(x)-x^{2})+\max_{s\in[0,y+r]}\max_{x\in(m-1,m]}(\mathcal{S}(x,s)+x^{2}) \displaystyle\leq(m-1)\cdot\max_{x\in(m-1,m]}\frac{h_{0}(x)-x^{2}}{|x|}+\mathfrak{C}+2m(y+r) \displaystyle+c\log^{\frac{2}{3}}(2+|m|+|y+r|)\,,\quad s\in[0,y+r] for some absolute c>0 and almost surely finite \mathscr{G} -measurable \mathfrak{C} where the last bound follows from the Airy sheet shape estimates ( 2.4 ). Since h_{0} is finitary (see Definition 2.15 ), we have \lim_{m\to\infty}\max_{x\in(m-1,m]}\frac{h_{0}(x)-x^{2}}{|x|}=-\infty\,, and so for all m\geq 1 sufficiently large, \displaystyle\max_{2\leq\ell\leq L^{m}_{0}}(G^{m}_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)]) \displaystyle\leq\max_{2\leq\ell\leq L^{1}_{0}}(G^{1}_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])\,,\quad s\in[0,y+r]\,. Hence, we have almost surely, (3.3) \max_{\begin{subarray}{c}2\leq\ell\leq L^{m}_{0}\\ m\geq 1\end{subarray}}(G^{m}_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])=\max_{\begin{subarray}{c}2\leq\ell\leq L^{m}_{0}\\ 1\leq m\leq m^{*}\end{subarray}}(G^{m}_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])\in\mathscr{C}([0,y+r];\mathbb{R})\,, where the continuity follows from Lemma 3.4 . We now conclude \lim_{s\to 0}\max_{2\leq\ell}(G^{\infty}_{\ell}+\mathcal{A}[(0,\ell)\to(s,2)])=G_{2}^{\infty}0 and h_{0}\in\mathscr{I}_{t} . Then, for any compact K\subset\mathbb{R} , have the (mutual) absolute continuity relation \mu\ll\nu^{h_{0}}\ll\mu on paths on \mathscr{C}_{0}(K;\mathbb{R}) , where \nu^{h_{0}} denotes the law of the increments of the KPZ fixed point started from h_{0} at time t>0 in K . Proof. From [ 29 , Theorem 1.2] , we have \nu^{h_{0}}\ll\mu . It thus remains to prove that \mu\ll\nu^{h_{0}} . As before, we lose no generality in assuming that t=1 and h_{0}\in\mathscr{C}(\mathbb{R};\mathbb{R}) . This is because of the metric composition law for the directed landscape ( 2.8 ) (cf. the proof of [ 29 , Theorem 1.2] ). It thus suffices to prove the mutual absolute continuity of the law of the increments of \mathfrak{h}_{1}(y,h_{0})=\max_{x\in\mathbb{Q}\setminus\{0\}}(h_{0}(x)+\mathcal{S}(x,y))\,,\quad y\in\mathbb{R}\,, against the rate two Wiener measure on some compact K . By translation symmetries of the Airy sheet, we can also without loss of generality take K\subset(\varepsilon,y_{0}-\varepsilon) for some y_{0}>0 and \varepsilon\in(0,y_{0}) . More precisely, let \mathcal{A}:\mathbb{N}\times\mathbb{R}\to\mathbb{R} be an Airy line ensemble and set \tilde{\mathcal{A}}_{1}(x)=\mathcal{A}_{1}(-x),x\in\mathbb{R} . Then, by the coupling between the Airy sheet and Airy line ensemble ( 2.5 ), we can express for all y\in\mathbb{Q}_{+} , \displaystyle\max_{x\in\mathbb{R}}\left(h_{0}(x)+\mathcal{S}(x,y)\right) \displaystyle=\bigg(\max_{x\in(0,\infty)}\max_{\ell\geq 1}\left(h_{0}(x)+\mathcal{A}[x\to(0,\ell)]\right)+\mathcal{A}[(0,\ell)\to(y,1)]\bigg) \displaystyle\vee\bigg(\max_{x\in(-\infty,0)}\max_{\ell\geq 1}\left(h_{0}(x)+\tilde{\mathcal{A}}[x\to(-y_{0},\ell)]+\tilde{\mathcal{A}}[(-y_{0},\ell)\to(-y,1)]\right)\bigg)\,. We thus have the alternative semi-discrete variational characterisation for the KPZ fixed point \displaystyle\mathfrak{h}_{1}(y,h_{0}) \displaystyle=\left(\max_{\ell\geq 1}G_{\ell}^{h_{0}}+\mathcal{A}[(0,\ell)\to(y,1)]\right) \displaystyle\vee\left(\max_{\ell\geq 1}\tilde{G}_{\ell}^{h_{0}}+\tilde{\mathcal{A}}[(-y_{0},\ell)\to(-y,1)]\right)\,,\quad y\in\mathbb{Q}\cap[0,y_{0}]\,, where for \ell\geq 1 , G_{\ell}^{h_{0}}=\max_{x\in(0,\infty)}\left(h_{0}(x)+\lim_{k\to\infty}\mathcal{A}[x_{k}\to(0,\ell)]-\mathcal{A}[x_{k}\to(0,1)]+\mathcal{S}(x,0)\right) and \tilde{G}_{\ell}^{h_{0}}=\max_{x\in(-\infty,0)}\left(h_{0}(x)+\lim_{k\to\infty}\tilde{\mathcal{A}}[(-x)_{k}\to(-y_{0},\ell)]-\tilde{\mathcal{A}}[(-x)_{k}\to(-y_{0},1)]+\mathcal{S}(x,y_{0})\right)\,, where x_{k}=(-\sqrt{k/(2x)},k) for x>0 . Note by the ergodic properties of the Airy line ensemble and the above coupling with the Airy sheet, in particular, we have from ( 2.5 ) that for all \ell\geq 1 , G_{\ell}^{h_{0}} are \mathscr{F}_{-}=\sigma(\{\mathcal{A}_{i}(x):x\leq 0\}) measurable and \tilde{G}_{\ell}^{h_{0}} are \mathscr{F}_{(y_{0},\infty)}\equiv\sigma(\{\mathcal{A}_{i}(x):x\geq y_{0}\}) measurable since \mathcal{S}(x,0) , x>0 and \mathcal{S}(x,y_{0}) , x<0 are \mathscr{F}_{-} and \mathscr{F}_{(y_{0},\infty)} -measurable respectively (cf. ( 2.5 )). We can now express using the metric composition law, Lemma ( 2.5 ), almost surely (recall the notation for the top line of a Pitman transform, ( 2.2 )), (3.4) \displaystyle\mathfrak{h}_{1}(y,h_{0}) \displaystyle=WF_{1}(y)\lor WG_{1}(y)\,,\quad y\in\mathbb{Q}\cap[0,y_{0}]\,, with the environments F=(F_{1},F_{1})=\bigg(G^{h_{0}}_{1}+\mathcal{A}_{1}(\cdot)-\mathcal{A}_{1}(0),\max_{\ell\geq 2}(G_{\ell}^{h_{0}}+\mathcal{A}[(0,\ell)\to(\cdot,2)]\bigg)\,,\quad\mbox{ and} G=(G_{1},G_{1})=\bigg(\tilde{G}^{h_{0}}_{1}+\mathcal{A}_{1}(\cdot)-\mathcal{A}_{1}(y_{0}),\max_{\ell\geq 2}(\tilde{G}_{\ell}^{h_{0}}+\tilde{\mathcal{A}}[(-y_{0},\ell)\to(-\cdot,2)])\bigg)\,. One can show that these two environments are actually continuous. Indeed, by the shape estimates for the Airy sheet, ( 2.4 ) one can show by arguing similarly as in the proof of Theorem 3.6 leading up to ( 3.3 ), that there exists m^{*}\in\mathbb{N} almost surely finite such that (3.5) \max_{\ell\geq 2}(G_{\ell}^{h_{0}}+\mathcal{A}[(0,\ell)\to(\cdot,2)]=\max_{\ell\geq 2}(G_{\ell}^{h_{0},m^{*}}+\mathcal{A}[(0,\ell)\to(\cdot,2)])\,, and (3.6) \max_{\ell\geq 2}(\tilde{G}_{\ell}^{h_{0}}+\tilde{\mathcal{A}}[(-y_{0},\ell)\to(-\cdot,2)])=\max_{\ell\geq 2}(\tilde{G}_{\ell}^{h_{0},m^{*}}+\tilde{\mathcal{A}}[(-y_{0},\ell)\to(-\cdot,2)]) where for m\geq 1 , G_{\ell}^{h_{0},m}=\max_{x\in(0,m)}\left(h_{0}(x)+\lim_{k\to\infty}\mathcal{A}[x_{k}\to(0,\ell)]-\mathcal{A}[x_{k}\to(0,1)]+\mathcal{S}(x,0)\right) and \tilde{G}_{\ell}^{h_{0},m}=\max_{x\in(-m,0)}\left(h_{0}(x)+\lim_{k\to\infty}\tilde{\mathcal{A}}[(-x)_{k}\to(-y_{0},\ell)]-\tilde{\mathcal{A}}[(-x)_{k}\to(-y_{0},1)]+\mathcal{S}(x,y_{0})\right)\,, then continuity follows from Lemma 3.4 . Moreover, the boundary data ( 3.5 ), ( 3.6 ) are both (3.7) \mathscr{G}\equiv\sigma(\{\mathcal{A}_{i}(x):i\geq 1\,,x\leq 0\text{ or }x\geq y_{0}\})\, measurable (by simply taking the minimal m=m^{*}\in\mathbb{N} so that the maxima are attained). Observe that for all \ell\geq 1 we have that the boundary data G^{h_{0},m}_{\ell},\tilde{G}^{h_{0},m}_{\ell} are strictly monotone by Proposition 3.1 and the flip symmetry of the Airy line ensemble. For ease of notation, we set K=[a,b] with 00,a\in\mathbb{R},K\subset\mathbb{R} with a<\inf K bounded and h_{0}\in\mathscr{I}_{t} , we have that there exists some probability measure on \mathbb{R} , \eta , such that \eta\otimes\mu\ll\mathrm{Law\,of}\;(\mathfrak{h}_{t}(a,h_{0}),\mathfrak{h}_{t}(\cdot,h_{0})-\mathfrak{h}_{t}(a,h_{0}))\ll\mathrm{Leb}\otimes\mu\,, where \mu denotes the law of a rate two Brownian motion starting at (a,0) , restricted to K and \mathrm{Leb} the Lebesgue measure on \mathbb{R} . Proof. First, by translation symmetries of the Airy sheet ( 2.5 ), we can take both a,\inf K>0 , with B a rate two Brownian motion starting at (a,0) (and independent from \mathfrak{h}(a) ). By the metric composition law for the directed landscape, ( 2.8 ) and independence, we can without loss of generality take h_{0} to be continuous. Inspecting the proof of Theorem 3.7 with [a,\sup K] in place of K and conditioning on \mathscr{G}\equiv\sigma(\{\mathcal{A}_{i}(x):(i,x)\not\in\{2,\ldots\}\times\mathbb{R}\cup\{1\}\times(0,\sup K+1)\}) instead of ( 3.7 ) we obtain with \mathfrak{h}_{t}(\cdot)\equiv\mathfrak{h}_{t}(\cdot,h_{0}) , \eta\otimes\mu\ll\mathrm{Law\,of}\;(\mathfrak{h}_{t}(a),\mathfrak{h}_{t}(\cdot)-\mathfrak{h}_{t}(a)) , for some probability measure on \mathbb{R} , \eta . For the converse absolute continuity relation, observe that by a localisation argument for the support of h_{0} (using the shape bounds for the Airy sheet ( 2.4 )), for any Borel A , \mathbb{P}((\mathfrak{h}_{t}(a),\mathfrak{h}_{t}(\cdot)-\mathfrak{h}_{t}(a))\in A)\leq\sum_{n=1}^{\infty}\mathbb{P}((\mathfrak{h}^{n}_{t}(a),\mathfrak{h}^{n}_{t}(\cdot)-\mathfrak{h}^{n}_{t}(a))\in A)\,, where \mathfrak{h}^{n}_{t} denotes the KPZ fixed point started from initial data h_{0}\cdot\delta_{[-n,n]} (recall ( 2.9 )). Now, for any fixed n , by the translation symmetries of the Airy sheet ( 2.5 ), it suffices to prove that \mathrm{Law\,of}\;(\mathfrak{h}^{n}_{t}(a),\mathfrak{h}^{n}_{t}(\cdot)-\mathfrak{h}^{n}_{t}(a))\ll\mathrm{Leb}\otimes\mu\,, for a=0 , K\subseteq(0,\infty) compact. In this case, the coupling 2.5 between the Airy sheet and Airy line ensemble gives \mathfrak{h}_{t}^{n}(y)=\max_{\begin{subarray}{c}1\leq\ell\leq L_{0}\end{subarray}}(G_{\ell}+\mathcal{A}[(0,\ell)\to(y,1)])\,,\quad y\in[0,\sup K+1]\,, for some almost surely finite random L_{0} . It thus suffices to show that the laws of (3.9) (G_{1},\max_{\begin{subarray}{c}1\leq\ell\leq m\end{subarray}}(G_{\ell}+\mathcal{A}[(0,\ell)\to(\cdot,1)])\ll\mathrm{Leb}\otimes\mu on \mathscr{C}(K;\mathbb{R}) for any m\geq 1 , by a similar localisation argument as above, sectioning on events \{L_{0}\leq m\} . Clearly G_{1}=\max(h_{0}(x)+\mathcal{S}(x,0))=\mathfrak{h}_{1}(0)\ll\mathrm{Leb} . Now ( 3.9 ) follows from the Brownian Gibbs property, 2.4 conditioning on \mathscr{F}_{-}=\sigma(\{\mathcal{A}_{i}(x):x\leq 0\}) , using the fact that G_{\ell} are \mathscr{F}_{-} measurable and then applying [ 31 , Theorem 7.9] to the iterated Skorokhod reflections (cf. 1.2 ) with boundary data (G_{\ell})_{1\leq\ell\leq m} on K . ∎ The mutual absolute continuity of the increments of the KPZ fixed point against Brownian motion on compacts gives the former has ‘full topological support’. This is the content of the following corollary. Corollary 3.9 . Let K\subseteq\mathbb{R} be a bounded interval, t>0 , h_{0}\in\mathscr{I}_{t} and f\in\mathscr{C}_{0}(K;\mathbb{R}) . Then for any \varepsilon>0 , we have \mathbb{P}(\left\lVert f-\mathfrak{h}_{t}(\cdot,h_{0})+\mathfrak{h}_{t}(\inf K,h_{0})\right\rVert_{L^{\infty}(K)}<\varepsilon)>0 . Proof. The proof of both parts follows from the mutual absolute continuity relation in Theorem 3.7 and Lemmas 2.6 and Lemma 2.7 . ∎ 4. The Airy Sheet and additive Brownian motion In this section, we prove mutual absolute continuity of additive Brownian motion to the Airy sheet on compacts. Figure 6 . Illustrations of: left : Additive Brownian motion, centre-left : \mathcal{S}(\cdot,\cdot^{\prime}) , centre-right : \mathcal{S}(\cdot,0)+\mathcal{S}(0,\cdot^{\prime}) and right : \mathcal{S}(\cdot,\cdot^{\prime})-\mathcal{S}(\cdot,0)-\mathcal{S}(0,\cdot^{\prime})+\mathcal{S}(0,0) on [0,1]^{2} . Observe the quadrangle inequality 2.6 \mathcal{S}(x,y)+\mathcal{S}(x^{\prime},y^{\prime})\geq\mathcal{S}(x,y^{\prime})+\mathcal{S}(x^{\prime},y)\,,\quad x\leq x^{\prime},\;y\leq y^{\prime} gives for any x_{0},y_{0} , the random continuous function \mathcal{S}(x,y)-\mathcal{S}(x,y_{0})-\mathcal{S}(x_{0},y)+\mathcal{S}(x_{0},y_{0})\,,\quad x\geq x_{0}\,,y\geq y_{0} is monotone in x,y and is thus the cumulative distribution function of a random Borel measure \mu_{x_{0},y_{0}} on [x_{0},\infty)\times[y_{0},\infty) (cf. right in Figure 6 ). We can thus write (4.1) \mu_{x_{0},y_{0}}([x_{0},x]\times[y_{0},y])=\mathcal{S}(x,y)-\mathcal{S}(x,y_{0})-\mathcal{S}(x_{0},y)+\mathcal{S}(x_{0},y_{0})\,. We start with a lemma that states with probability strictly between zero and one, the quadrangle inequality ( 2.6 ), becomes an equality on compacts (cf. centre-right in Figure 6 ), or in other words the measures \mu_{x_{0},y_{0}} are not degenerate but can vanish on compacts with positive probability. Proposition 4.1 . Fix M>0 , then we have (4.2) \mathbb{P}(\mathcal{S}(x,y)=\mathcal{S}(x,-M)+\mathcal{S}(-M,y)-\mathcal{S}(-M,-M)\,,\text{ for all }|x|+|y|\leq M)>0\,. Moreover, (4.3) \lim_{m\to\infty}\mathbb{P}(\mathcal{S}(x,y)=\mathcal{S}(x,-M)+\mathcal{S}(-M,y)-\mathcal{S}(-M,-M)\,,\text{ for all }|x|+|y|\leq M)=0\,. Remark . This result also has a geometric interpretation in terms of geodesics in the directed landscape (cf. [ 6 , Section 12] for a definition). In particular, by [ 11 , Lemma 3.15] , the above event is equivalent to the fact that no two directed geodesics with endpoints (x_{1},0),(y_{1},1) and (x_{2},0),(y_{2},1) for -M\leq x_{1}0 with probability tending to one as N\to\infty , concluding the proof. ∎ We arrive at the following absolute continuity result. It states that the law of the Airy sheet (up to centering) majorises the law of additive Brownian motion , that is the sum of two independent rate two Brownian motions, on compacts. This is the content of the following theorem. Theorem 4.2 . Fix K\subset\mathbb{R}^{2} compact. Then, with B,B^{\prime} two independent Brownian motions starting on (\inf K,0) , the law of B(\cdot)+B^{\prime}(\cdot^{\prime}) on \mathscr{C}_{0}(K;\mathbb{R}) is absolutely continuous with respect to the law of the Airy sheet \mathcal{S}(\cdot,\cdot^{\prime})-\mathcal{S}(\inf K,\inf K) on \mathscr{C}_{0}(K;\mathbb{R}) . Proof. By translation symmetries of the Airy sheet, we can replace K with [1,2\mathrm{diam}K+1]^{2} . Let A\subseteq\mathscr{C}_{0}(K;\mathbb{R}) be Borel measurable. Using the coupling of the Airy sheet with the Airy line ensemble, 2.5 , we have with M=\mathrm{diam}K on the event (recall \eta_{M} from ( 4.4 ) in Proposition 4.1 ) A_{M}\equiv\{\mathcal{A}_{1}(y)-\mathcal{A}_{1}(0)\geq\eta_{M}+\mathcal{A}[(0,L_{M})\to(y,2)]\,,\quad y\in[0,2M+1]\}\,, \mathcal{S}(x,y)=\mathcal{S}(0,y)-\mathcal{S}(0,0)+\mathcal{S}(x,0)=\mathcal{A}_{1}(y)-\mathcal{A}_{1}(0)+\mathcal{S}(x,0)\,. Note by the coupling ( 2.5 ) and ergodic properties of the Airy sheet ( 2.5 ), \mathcal{S}(\cdot,0) , \eta_{M} , L_{M} are \mathscr{F}_{(-\infty,0)\times\mathbb{N}\cup\mathbb{R}\times\{2,\ldots\}}\equiv\sigma(\{\mathcal{A}_{i}(x):(i,x)\in\mathbb{N}\times(-\infty,0)\cup\{2,\ldots\}\times\mathbb{R}\}) measurable. We can thus estimate \displaystyle\mathbb{P}(\mathcal{S}(\cdot,\cdot^{\prime}) \displaystyle-\mathcal{S}(\inf K,\inf K)\in A)\geq\mathbb{P}(\mathcal{A}_{1}(\cdot)-\mathcal{A}_{1}(1)+\mathcal{S}(\cdot^{\prime},0)-\mathcal{S}(1,0)\in A,A_{M}) \displaystyle=\mathbb{P}\big(\mathcal{A}_{1}(\cdot)-\mathcal{A}_{1}(1)+\mathcal{S}(\cdot^{\prime},0)-\mathcal{S}(1,0)\in A, \displaystyle\mathcal{A}_{1}(y)-\mathcal{A}_{1}(1)+\mathcal{A}_{1}(1)-\mathcal{A}_{1}(0)\geq\eta_{M}+\mathcal{A}[(0,L_{M})\to(y,2)])\,,y\in[0,2M+1]\big)\,. Now, by the Brownian Gibbs property, (cf. 2.4 ), one can resample the Airy line ensemble on \{1\}\times[1,2M+2] to have the law of the affine shift L(\cdot)+W(\cdot) of an independent Brownian bridge W(\cdot) starting from (1,0) and ending at (2M+2,0) , with L(0)=\mathcal{A}_{1}(1) and L(2M+2)=\mathcal{A}_{1}(2M+2) conditioned to avoid \mathcal{A}|_{[1,2M+2]\times\{2\}} . In particular, by Lemma 2.6 , we have the law of L(\cdot)+W(\cdot) restricted to [1,2M+1] is mutually absolutely continuous with respect to the law of \mathcal{A}_{1}(1)+B(\cdot) , where B is a Brownian motion starting at (1,0) . Thus, we have \displaystyle\mathbb{P}\big(L(\cdot)+W(\cdot)-\mathcal{A}_{1}(1)+\mathcal{S}(\cdot^{\prime},0)-\mathcal{S}(1,0)\in A, \displaystyle\mathcal{A}_{1}(y)-\mathcal{A}_{1}(0)\geq\eta_{M}+\mathcal{A}[(0,L_{M})\to(y,2)])\,,y\in[0,1]\,, \displaystyle W(\cdot)+L(\cdot)-\mathcal{A}_{1}(1)+\mathcal{A}_{1}(1)-\mathcal{A}_{1}(0)\geq\eta_{M}+\mathcal{A}[(0,L_{M})\to(y,2)])\,,y\in[1,2M+1]\,, \displaystyle W(y)+L(y)\geq\mathcal{A}_{2}(y)\,,y\in[1,2M+1]\big)=0 if and only if \displaystyle\mathbb{P}\big(B(\cdot)+\mathcal{S}(\cdot^{\prime},0)-\mathcal{S}(1,0)\in A,\mathcal{A}_{1}(y)-\mathcal{A}_{1}(0)\geq\eta_{M}+\mathcal{A}[(0,L_{M})\to(y,2)])\,,y\in[0,1]\,, \displaystyle B(\cdot)+\mathcal{A}_{1}(1)-\mathcal{A}_{1}(0)\geq\eta_{M}+\mathcal{A}[(0,L_{M})\to(y,2)])\,,y\in[1,2M+1]\,, \displaystyle\mathcal{A}_{1}(1)+B(\cdot)\geq\mathcal{A}_{2}(y)\,,y\in[1,2M+1]\big)=0\,. Now, conditioning on \mathscr{F}^{\mathrm{ext}}_{[1,2M+1]\times\{1\}}\lor\sigma(B_{t}:t\in[1,2M+1]) where \mathscr{F}^{\mathrm{ext}}_{[1,2M+1]\times\{1\}}\equiv\sigma(\{\mathcal{A}_{i}(x):(i,x)\not\in\{1\}\times[1,2M+1]) by the Brownian bridge property 2.4 , stochastic domination arguments using Lemma 2.9 and 2.7 (recall \mathcal{S}(\cdot,0) , \eta_{M} , L_{M} are \mathscr{F}_{(-\infty,0)\times\mathbb{N}\cup\mathbb{R}\times\{2,\ldots\}} -measurable), \displaystyle\mathbb{P}\bigg(\mathcal{A}_{1}(y)-\mathcal{A}_{1}(0)\geq\eta_{M}+\mathcal{A}[(0,L_{M})\to(y,2)])\,,y\in[0,1]\,, \displaystyle B(\cdot)+\mathcal{A}_{1}(1)-\mathcal{A}_{1}(0)\geq\eta_{M}+\mathcal{A}[(0,L_{M})\to(y,2)])\,,y\in[1,2M+1]\,, \displaystyle\mathcal{A}_{1}(1)+B(\cdot)\geq\mathcal{A}_{2}(y)\,,y\in[1,2M+1]\bigg|\mathscr{F}^{\mathrm{ext}}_{[1,2M+1]\times\{1\}}\lor\sigma(B_{t}:t\in[0,1,2M+1])\bigg)>0 almost surely. We thus deduce that the law of B(\cdot)+\mathcal{S}(\cdot^{\prime},0) (for B and \mathcal{S} independent) on \mathscr{C}_{0}(K;\mathbb{R}) is absolutely continuous with respect to the law of the Airy sheet \mathcal{S}(\cdot,\cdot^{\prime})-\mathcal{S}(\inf K,\inf K) on \mathscr{C}_{0}(K;\mathbb{R}) . To conclude the proof observe that the law of \mathcal{S}(\cdot,0)-\mathcal{S}(\inf K,0) is mutually absolutely continuous with respect to a rate two Brownian motion on K , by the coupling of the Airy sheet with the Airy line ensemble ( 2.5 ) and Theorem 2.11 . ∎ This means the law of the Airy sheet has full topological support in the space of certain separable continuous functions in \mathbb{R}^{2} . In particular, generalising the quadrangle equality in the statement of Proposition 4.1 , for K\subset\mathbb{R}^{2} compact, we denote by (4.6) \bm{\mathrm{Rect}}_{K}\equiv\{h\in\mathscr{C}^{2}(K;\mathbb{R}):h(x_{1},y_{1})+h(x_{2},y_{2})-h(x_{1},y_{2})-h(x_{2},y_{1})=0\,,\text{ for all }x_{1,2},y_{1,2}\in K\}\,, the set of all continuous functions on K satisfying the so-called rectangle property above. Theorem 4.2 cannot be strengthened to mutual absolute continuity between the Airy sheet and additive Brownian motion. This is indeed the case and is the content of the following proposition. Proposition 4.3 . Fix a,b\in\mathbb{R} and B,B^{\prime} two independent Brownian motions starting from (a,0) and (b,0) respectively. Then, the law of the centred Airy sheet \mathcal{S}(\cdot,\cdot^{\prime})-\mathcal{S}(a,b) on \mathscr{C}_{0}([a,\infty)\times[b,\infty);\mathbb{R}) is not absolutely continuous with respect to the law of the additive Brownian motion B(\cdot)+B^{\prime}(\cdot^{\prime}) on \mathscr{C}_{0}([a,\infty)\times[b,\infty);\mathbb{R}) . Proof. Suppose for a contradiction that the law of the Airy sheet \mathcal{S}(\cdot,\cdot^{\prime})-\mathcal{S}(a,b) on \mathscr{C}_{0}([a,\infty)\times[b,\infty);\mathbb{R}) is absolutely continuous with respect to the law of the additive Brownian motion B(\cdot)+B^{\prime}(\cdot^{\prime}) on \mathscr{C}_{0}([a,\infty)\times[b,\infty);\mathbb{R}) . Consider for any M>1 , the continuous (with respect to the uniform topology) functional on \mathscr{C}_{0}([a,\infty)\times[b,\infty);\mathbb{R}) , f\mapsto\Delta^{M}(f)\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\max_{(x,y)\in[a,a+M]\times[b,b+M]}(f(x,y)-f(x,b)-f(a,y)+f(a,b))\,. By Proposition 4.1 , the Airy sheet satisfies \Delta^{M}(\mathcal{S})=\max_{(x,y)\in[a,a+M]\times[b,b+M]}\mu_{a,b}([a,x]\times[b,y])>0 with positive probability for all M>1 sufficiently large, whereas the restriction of \Delta^{M} to \bm{\mathrm{Rect}}_{[a,a+M]\times[b,b+M]^{2}} is identically zero for all M . This is a contradiction as additive Brownian motion is supported on \bm{\mathrm{Rect}}_{[a,a+M]\times[b,b+M]^{2}} . ∎ We end this subsection with the following corollary regarding the topological support of the Airy sheet. Corollary 4.4 . Fix K\subset\mathbb{R}^{2} compact, h\in\bm{\mathrm{Rect}}_{K} (cf. ( 4.6 )) and \varepsilon>0 . Then, for any compact K\subseteq\mathbb{R}^{2} , \mathbb{P}(||\mathcal{S}(\cdot,\cdot^{\prime})-\mathcal{S}(\inf K,\inf K)-(h(\cdot,\cdot^{\prime})-h(\inf K,\inf K))||_{L^{\infty}(K)}<\varepsilon)>0\,. Proof. The proof is a direct application of Theorem 4.2 , Lemma 2.7 and the fact that any h\in\bm{\mathrm{Rect}}_{K} satisfies h(\cdot,\cdot^{\prime})=h(0,\cdot)-h(0,0)+h(\cdot^{\prime},0)-h(0,0) . ∎ 5. Applications In this section, we discuss some applications of the mutual absolute continuity result of the KPZ fixed point established in Theorem 3.7 and the absolute continuity result of the Airy sheet proved in Theorem 4.2 . 5.1. Record times for the KPZ fixed point We obtain as a direct consequence of Theorem 3.7 a result involving the ‘topological support’ of record times for the KPZ fixed point, that is the times the KPZ fixed point attains its running maximum (relative to some starting point). Define the random closed subset of record times for the KPZ fixed point \mathfrak{h}_{t}(\cdot,h_{0}) at time t>0 started from initial data h_{0}\in\mathscr{I}_{t} , \mathscr{R}(a,t,h_{0})\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\{y\geq a:\mathfrak{h}_{t}(y,h_{0})=\max_{a\leq s\leq y}\mathfrak{h}_{t}(s,h_{0})\}\,,\quad a\in\mathbb{R}\,. In the following corollary, we obtain that in any interval [b,c] , a0 , a0 and let h_{0}\in\mathscr{I}_{t} and recall the notation for the KPZ fixed point started from h_{0} at time t , \mathfrak{h} (suppressing time dependence). We are interested in giving necessary and sufficient conditions for the positivity of probabilities of the form (5.1) \mathbb{P}(\mathfrak{h}(E)\cap F\neq\emptyset)=\mathbb{P}(\mathrm{Gr}(\mathfrak{h})\text{ hits }E\times F)>0\,, where the graph of the KPZ fixed point is denoted by \mathrm{Gr}(\mathfrak{h})\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\{(t,\mathfrak{h}(t)):t\in\mathbb{R}\}\subseteq\mathbb{R}^{2} . In other words, we are interested in the probability the KPZ fixed point on some compact E hits another compact F , see Figure 7 below. Figure 7 . Illustration of the graph of the rescaled increments of the KPZ fixed point in blue and the set E\times F in red . The probability \mathfrak{h}(E)\cap F\neq\emptyset if and only if the graph of the KPZ fixed point, \mathrm{Gr}(\mathfrak{h}) hits the red region with positive probability. We will in fact be able to obtain a characterisation for ( 5.1 ) by comparing ( 5.1 ) to Brownian hitting probabilities using the mutual absolute continuity of increments of the KPZ fixed point, Corollary 3.8 . It is a well-known folklore fact that for a Brownian motion W , W(E) intersects F with positive probability if and only if E\times F has positive thermal capacity in the sense of Watson [ 33 , 34 ] . Now, by Corollary 3.8 and analytic properties of tail probabilities of the KPZ fixed point, see [ 22 , Section 4] (which give a positive density against the Lebesgue measure on \mathbb{R} ) we obtain the absolute continuity relations \mathrm{Law\,of}\;(Y,B(\cdot))\ll\mathrm{Law\,of}\;(\mathfrak{h}(a),\mathfrak{h}(\cdot)-\mathfrak{h}(a))\ll\mathrm{Leb}\otimes\mu\,, on \mathbb{R}\times\mathscr{C}(K;\mathbb{R}) , for any K\subseteq\mathbb{R} compact and a<\inf K , with B a rate two Brownian motion starting at (a,0) (and independent from the random variable Y ), the above statements transfer verbatim to the KPZ fixed point and E,F as above. We first formulate the statement of the characterisation in the case where the Lebesgue measure of F is zero, since otherwise the intersection probability \mathbb{P}(\mathfrak{h}(E)\cap F\neq\emptyset) is strictly positive for every non-empty Borel set E\subset\mathbb{R} . Definition 5.2 . Let E\subset\mathbb{R} and F\subset\mathbb{R} be compact sets. For \gamma\geq 0 , the \gamma -thermal capacity of the set E\times F is defined by (5.2) C_{\gamma}(E\times F)=\frac{1}{\inf\{\mathscr{E}_{\gamma}(\mu):\mu\in\mathscr{P}(E\times F)\}}, where \mathscr{P}(E\times F) denotes the set of all Borel probability measures \mu on E\times F such that \mu(\{t\}\times F)=0 for all t\in E and \mathscr{E}_{\gamma}(\mu) is the \gamma -thermal energy of \mu defined by (5.3) \mathscr{E}_{\gamma}(\mu)\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\int_{E\times F}\int_{E\times F}\frac{\mathrm{e}^{-|x-y|^{2}/(4|t-s|)}}{|t-s|^{1/2}\cdot|x-y|^{\gamma}}\mu(\operatorname{d}\!s\,\operatorname{d}\!x)\mu(\operatorname{d}\!t\,\operatorname{d}\!y)\,. Corollary 5.3 . Suppose F\subset\mathbb{R} is compact and has Lebesgue measure 0. Then \mathbb{P}(\mathfrak{h}(E)\cap F\neq\emptyset)>0 if and only if \mathscr{C}_{0}(E\times F)>0 . Proof. First assume \mathscr{C}_{0}(E\times F)>0 . Then, observe for any a<\inf E , \mathbb{P}(\mathfrak{h}(E)\cap F\neq\emptyset)=\mathbb{P}((\mathfrak{h}(E)-\mathfrak{h}(a))\cap(F-\mathfrak{h}(a))\neq\emptyset)=\mathbb{P}(1/\sqrt{2}(\mathfrak{h}(E)-\mathfrak{h}(a))\cap 1/\sqrt{2}(F-\mathfrak{h}(a))\neq\emptyset)\,. By Corollary, 3.8 , we have for some probability measure \eta on \mathbb{R} , \eta\otimes\mu\ll\mathrm{Law\,of}(\mathfrak{h}(a),\mathfrak{h}(\cdot)-\mathfrak{h}(a)) on \mathbb{R}\times\mathscr{C}(E;\mathbb{R}) . Hence, there exist a random variable Y with law \eta and a standard Brownian motion W starting from (a,0) (mutually independent) such that (Y,W(\cdot)) is absolutely continuous with respect to (\mathfrak{h}(a),1/\sqrt{2}(\mathfrak{h}(\cdot)-\mathfrak{h}(a))) on \mathbb{R}\times\mathscr{C}(E;\mathbb{R}) . Thus, to prove \mathbb{P}(\mathfrak{h}(E)\cap F\neq\emptyset)>0 , it suffices to prove \mathbb{P}(W(E)\cap 1/\sqrt{2}(F-Y)\neq\emptyset)>0 . Now, conditionally on Y , by [ 19 , Proposition 1.4] , this is true if and only if there exists a probability measure \sigma on (E-a)\times 1/\sqrt{2}(F-Y) such that \int_{(E-a)\times 1/\sqrt{2}(F-Y)}\int_{(E-a)\times 1/\sqrt{2}(F-Y)}\frac{\mathrm{e}^{-|x-y|^{2}/(2|t-s|)}}{|t-s|^{1/2}}\sigma(\operatorname{d}\!s\,\operatorname{d}\!x)\sigma(\operatorname{d}\!t\,\operatorname{d}\!y)>0\,. Now, by translation and scaling this is true if and only if there exists some probability measure \tilde{\sigma} on E\times F such that \int_{E\times F}\int_{E\times F}\frac{\mathrm{e}^{-|x-y|^{2}/(4|t-s|)}}{|t-s|^{1/2}}\tilde{\sigma}(\operatorname{d}\!s\,\operatorname{d}\!x)\tilde{\sigma}(\operatorname{d}\!t\,\operatorname{d}\!y)>0\,, or equivalently, if and only if \mathscr{C}_{0}(E\times F)>0 . For the converse implication, suppose that \mathscr{C}_{0}(E\times F)=0 , then use the absolute continuity relation from Corollary 3.8 , \mathrm{Law\,of}\;(\mathfrak{h}(a),\mathfrak{h}(\cdot)-\mathfrak{h}(a))\ll\mathrm{Leb}\otimes\mu for a<\inf E , on \mathbb{R}\times\mathscr{C}(E;\mathbb{R}) and argue entirely analogously, using the characterisation in [ 19 , Proposition 1.4] . ∎ The condition in Corollary 5.3 can also be recast in terms of a geometric condition on the set E\times F involving a certain kind of Hausdorff dimension, which we turn to now. Let us define \varrho to be the parabolic metric on \mathbb{R}^{2} , that is, \varrho\bigl((s,x);(t,y)\bigr)\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\max\bigl(|t-s|^{1/2},|x-y|\bigr). On the metric space \mathbf{S}=(\mathbb{R}^{2},\varrho) , also called space-time as we distinguish the spatial and temporal variables, we can define a notion of Hausdorff dimension associated to it. More precisely, for any \alpha\geq 0 and A\subset\mathbb{R}\times\mathbb{R} , the \alpha -dimensional parabolic Hausdorff measure of A is defined by \mathscr{H}^{\alpha}(A;\varrho)=\lim_{\delta\to 0}\inf\left\{\sum_{n=1}^{\infty}(\mathrm{diam}_{\varrho}\,U_{n})^{\alpha}:\text{open cover }(U_{n})_{n=1}^{\infty}\text{ of }A\,,\,\sup_{n\geq 1}\,(\mathrm{diam}_{\varrho}\,U_{n})\leq\delta\right\}\,. The parabolic Hausdorff dimension of A is defined by \mathrm{dim}_{\mathscr{H}}(A;\varrho)\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\inf\{\alpha\geq 0:\mathscr{H}^{\alpha}(A;\varrho)=0\}. Corollary 5.4 (Intersection probabilities) . If \mathrm{dim}_{\mathscr{H}}(E\times F;\varrho)>1 then \mathbb{P}(\mathfrak{h}(E)\cap F\neq\emptyset)>0 . If \mathrm{dim}_{\mathscr{H}}(E\times F;\varrho)<1 then \mathbb{P}(\mathfrak{h}(E)\cap F\neq\emptyset)=0 . Proof. Proceed as in the proof of Corollary 5.3 and use Corollary 3.8 to argue as in page 407 of [ 19 ] applying the characterisation of the parabolic Hausdorff dimension in terms of thermal capacity. ∎ In the case where the above probability does not vanish, we are interested in describing the Hausdorff dimension \mathrm{dim}_{\mathscr{H}}(\mathfrak{h}(E)\cap F) of the random intersection set \mathfrak{h}(E)\cap F . In particular, we seek only to compute the L^{\infty}(\mathbb{P}) -norm of that Hausdorff dimension, since in general it is not constant. Below for F\subseteq\mathbb{R} compact, we denote its Lebesgue measure by |F| . We also denote by \mathrm{dim}_{\mathscr{H}} , the Hausdorff dimension on \mathbb{R} with respect to the Euclidean metric (in any dimension). Corollary 5.5 . If F\subset\mathbb{R} is compact and |F|>0 , then (5.4) \left\lVert\mathrm{dim}_{\mathscr{H}}\bigl(\mathfrak{h}(E)\cap F\bigr)\right\rVert_{L^{\infty}(\mathbb{P})}=2\mathrm{dim}_{\mathscr{H}}(E)\land 1. If in addition, \mathrm{dim}_{\mathscr{H}}(E)>1/2 , then \mathbb{P}(|\mathfrak{h}(E)\cap F|>0)>0 . Proof. Use [ 19 , Proposition 1.2] to obtain that for a standard Brownian motion starting from (a,0) , a<\inf E , with positive probability for any \varepsilon>0 , \mathbb{P}(\mathrm{dim}_{\mathscr{H}}\bigl(W(E)\cap F\bigr)\geq 2\mathrm{dim}_{\mathscr{H}}(E)\land 1-\varepsilon)>0 and almost surely, \mathrm{dim}_{\mathscr{H}}\bigl(W(E)\cap F\bigr)\leq 2\mathrm{dim}_{\mathscr{H}}(E)\land 1 . Then, argue as in Corollary 5.3 and note that the rescaling therein is a bi-Lipschitz homeomorphism, thereby preserving the Hausdorff dimension almost surely. ∎ The remaining case, and arguably most interesting case, is when F has Lebesgue measure 0, that is |F|=0 . The following result gives a suitable (though quite complicated) formula that also generalises for Brownian motion in higher dimensions. The proof is entirely analogous to that of Corollary 5.5 , hence omitted. Corollary 5.6 . If F\subset\mathbb{R} is compact and |F|=0 , then (5.5) \left\lVert\mathrm{dim}_{\mathscr{H}}\bigl(\mathfrak{h}(E)\cap F\bigr)\right\rVert_{L^{\infty}(\mathbb{P})}=\sup\Bigl\{\gamma>0:C_{\gamma}(E\times F)>0\Bigr\}\,. 5.3. Geometric properties of the Airy sheet images Having established the absolute continuity of additive Brownian motion against the Airy sheet in Theorem 4.2 we compute essential suprema of Hausdorff dimensions of images of compact sets under the Airy sheet and give conditions for the positivity of their Lebesgue measure in terms of the one–dimensional Bessel–Riesz capacity, which we now define. Definition 5.7 (Bessel-Riesz capacity) . Let E\subset\mathbb{R}^{2} be compact. The (one-dimensional) Bessel - Riesz capacity of the set E\subset\mathbb{R}^{2} is defined by (5.6) C_{\mathrm{BR}}(E)=\frac{1}{\inf\{\mathscr{E}_{\mathrm{BR}}(\mu):\mu\in\mathscr{P}(E)\}}\in[0,+\infty], where \mathscr{P}(E) denotes the set of all Borel probability measures \mu on E and \mathscr{E}_{\mathrm{BR}}(\mu) is the ‘energy’ of \mu defined by \mathscr{E}_{\mathrm{BR}}(\mu)\stackrel{{\scriptstyle\mathrm{def}}}{{=}}\int_{E}\int_{E}\frac{1}{|x-y|^{1/2}}\mu(\operatorname{d}\!x)\mu(\operatorname{d}\!y)\,. We now compute the essential supremum of the Euclidean Hausdorff dimension of Airy sheet images using the absolute continuity result in Theorem 4.2 and the spatial regularity of the Airy sheet, [ 6 , Proposition 10.5] . This is the content of the following corollary. Corollary 5.8 . Let E\subset\mathbb{R}^{2} be bounded Borel and \mathcal{S} an Airy sheet. Then we have \left\lVert\mathrm{dim}_{\mathscr{H}}(\mathcal{S}(E))\right\rVert_{L^{\infty}(\mathbb{P})}=1\land 2\mathrm{dim}_{\mathscr{H}}(E)\,. Proof. By translation symmetries of the Airy sheet, ( 2.5 ), we can assume without loss of generality E\subset(0,\infty)^{2} . The almost sure upper bound on the random variable \mathrm{dim}_{\mathscr{H}}(\mathcal{S}(E)) can be readily established from the Hölder 1/2- regularity of the Airy sheet, [ 6 , Proposition 10.5] , which means the dimension of the image can at most double. For the lower bound, [ 35 , Theorem 6.11, page 181] gives almost surely, \mathrm{dim}_{\mathscr{H}}(X(E))=1\land 2\mathrm{dim}_{\mathscr{H}}(E) where X denotes an additive Brownian motion. Thus, by Theorem 4.2 and translation invariance of Hausdorff dimension, we have with positive probability \mathrm{dim}_{\mathscr{H}}(\mathcal{S}(E))=1\land 2\mathrm{dim}_{\mathscr{H}}(E) , and so we deduce \left\lVert\mathrm{dim}_{\mathscr{H}}(\mathcal{S}(E))\right\rVert_{L^{\infty}(\mathbb{P})}\geq 1\land 2\mathrm{dim}_{\mathscr{H}}(E) , which yields the result. ∎ We now state a condition for when the above random sets have positive Lebesgue measure. Corollary 5.9 . Let E\subset\mathbb{R}^{2} be compact and \mathcal{S} an Airy sheet. If C_{\mathrm{BR}}(E)>0 , then \mathbb{P}(\big|\mathcal{S}(E)\big|>0)>0 . Remark . By Frostman’s characterisation of Hausdorff dimension, if \mathrm{dim}_{\mathscr{H}}(E)>1/2 , \mathbb{P}(\big|\mathcal{S}(E)\big|>0)>0 and by Corollary 5.8 if \mathrm{dim}_{\mathscr{H}}(E)<1/2 , then \big|\mathcal{S}(E)\big|=0 almost surely. Proof. This, is a direct application of the translation invariance of the Lebesgue measure, the absolute continuity of the Airy sheet (up to centering) against additive Brownian motion, Theorem 4.2 and [ 20 , Theorem 1.1] (which is the analogous result of 5.9 for additive Brownian motion). ∎ 6. Appendix Next result shows infinite geodesics in the Airy line ensemble do not ‘jump instantaneously’. Lemma 6.1 . Fix x<0 , 1\leq k<\ell . Then, with \varepsilon^{\infty}_{\ell,k} the last jump time of the almost-surely unique geodesic on the Airy line ensemble from (x,\ell) to (0,k) , we have \varepsilon^{\infty}_{\ell,k}<0 almost surely. Proof. We show \mathbb{P}(\varepsilon^{\infty}_{\ell,k}=0)=0 . Expressing last passage values in terms of the Pitman transform see subsection 2.2 , we have \varepsilon^{\infty}_{\ell,k}=\operatorname*{argmax}_{z\in[x,0]}(\mathcal{A}[x\to(z,k+1)]+\mathcal{A}[(z,k)\to(0,k)])\,. Now, by the mutual absolute continuity of the centred of the Airy line ensemble with respect to independent Brownian motions, [ 9 , Theorem 1.1] , \varepsilon^{\infty}_{\ell,k} is mutually absolutely continuous with respect to \operatorname*{argmax}_{z\in[x,0]}(B[(x,\ell)\to(z,k+1)]+B[(z,k)\to(0,k)])=\operatorname*{argmax}_{z\in[x,0]}(B[(x,\ell)\to(z,k+1)]-B_{k}(z))\,, where B is a family of independent Brownian motions. By [ 31 , Proposition 4.1] and [ 31 , Theorem 7.1] , the laws of B[(x,\ell)\to(\cdot,k+1)] restricted to [x/2,0] is absolutely continuous with respect to that of a standard Brownian motion starting from 0 restricted to [x,0] . Thus, by independence of B[(x,\ell)\to(\cdot,k+1)] with B_{k} , time-reversal and flip symmetry of Brownian motion, \displaystyle\tilde{\varepsilon}^{\infty}_{\ell,k} \displaystyle:=\operatorname*{argmax}_{z\in[x/2,0]}(B[(x,\ell)\to(z,k+1)]-B_{k}(z)) \displaystyle=\operatorname*{argmax}_{z\in[x/2,0]}(B[(x,\ell)\to(z,k+1)]-B[(x,\ell)\to(0,k+1)]-B_{k}(z)) has the law of the argmax of a Brownian motion on [x/2,0] . Now, by Lévy’s arcsine law, \tilde{\varepsilon}^{\infty}_{\ell,k} has a density with respect to the Lebesgue measure, which concludes the proof. ∎