Reflected Brownian Motions in the KPZ

Universality Class 1st Edition Thomas

Weiss
Visit to download the full and correct content document:
https://textbookfull.com/product/reflected-brownian-motions-in-the-kpz-universality-cla
ss-1st-edition-thomas-weiss/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

The Oxford Handbook on the United Nations Thomas G.

Weiss

https://textbookfull.com/product/the-oxford-handbook-on-the-
united-nations-thomas-g-weiss/

Motions of Ice Hydrometeors in the Atmosphere Numerical

Studies and Implications Pao K. Wang

https://textbookfull.com/product/motions-of-ice-hydrometeors-in-
the-atmosphere-numerical-studies-and-implications-pao-k-wang/

The United Nations and Changing World Politics Revised

and Updated with a New Introduction 8th Edition Thomas
G. Weiss

https://textbookfull.com/product/the-united-nations-and-changing-
world-politics-revised-and-updated-with-a-new-introduction-8th-
edition-thomas-g-weiss/

The new politics of class in Britain. The political

exclusion of the working class 1st Edition Evans

https://textbookfull.com/product/the-new-politics-of-class-in-
britain-the-political-exclusion-of-the-working-class-1st-edition-
evans/
America Inc Innovation and Enterprise in the National
Security State 1st Edition Linda Weiss

https://textbookfull.com/product/america-inc-innovation-and-
enterprise-in-the-national-security-state-1st-edition-linda-
weiss/

Sorting Out the New South City Race Class and Urban
Development in Charlotte 1875 1975 2nd Edition Thomas
W. Hanchett

https://textbookfull.com/product/sorting-out-the-new-south-city-
race-class-and-urban-development-in-charlotte-1875-1975-2nd-
edition-thomas-w-hanchett/

Brownian Motion Martingales and Stochastic Calculus 1st

Edition Jean-François Le Gall

https://textbookfull.com/product/brownian-motion-martingales-and-
stochastic-calculus-1st-edition-jean-francois-le-gall/

Brownian Motion Martingales and Stochastic Calculus 1st

Edition Jean-François Le Gall

https://textbookfull.com/product/brownian-motion-martingales-and-
stochastic-calculus-1st-edition-jean-francois-le-gall-2/

Class on Screen: The Global Working Class in

Contemporary Cinema Sarah Attfield

https://textbookfull.com/product/class-on-screen-the-global-
working-class-in-contemporary-cinema-sarah-attfield/
SPRINGER BRIEFS IN MATHEMATIC AL PHYSICS 18

Thomas Weiss
Patrik Ferrari
Herbert Spohn

Reflected
Brownian
Motions in the
KPZ Universality
Class

123
SpringerBriefs in Mathematical Physics

Volume 18

Series editors
Nathanaël Berestycki, Cambridge, UK
Mihalis Dafermos, Cambridge, UK
Tohru Eguchi, Tokyo, Japan
Atsuo Kuniba, Tokyo, Japan
Matilde Marcolli, Pasadena, USA
Bruno Nachtergaele, Davis, USA
SpringerBriefs are characterized in general by their size (50-125 pages) and fast
production time (2-3 months compared to 6 months for a monograph).

Briefs are available in print but are intended as a primarily electronic publication to
be included in Springer’s e-book package.

Typical works might include:

• An extended survey of a field

• A link between new research papers published in journal articles
• A presentation of core concepts that doctoral students must understand in order
• to make independent contributions
• Lecture notes making a specialist topic accessible for non-specialist readers.

SpringerBriefs in Mathematical Physics showcase, in a compact format, topics of

current relevance in the field of mathematical physics. Published titles will
encompass all areas of theoretical and mathematical physics. This series is intended
for mathematicians, physicists, and other scientists, as well as doctoral students in
related areas.

More information about this series at http://www.springer.com/series/11953

Thomas Weiss Patrik Ferrari
•

Herbert Spohn

Reflected Brownian Motions

in the KPZ Universality Class

123
Thomas Weiss Herbert Spohn
Zentrum Mathematik Zentrum Mathematik, M5
Technische Universität München Technische Universität München
Garching Munich
Germany Germany

Patrik Ferrari
Institut für Angewandte Mathematik
Universität Bonn
Bonn
Germany

ISSN 2197-1757 ISSN 2197-1765 (electronic)

SpringerBriefs in Mathematical Physics
ISBN 978-3-319-49498-2 ISBN 978-3-319-49499-9 (eBook)
DOI 10.1007/978-3-319-49499-9
Library of Congress Control Number: 2016958720

© The Author(s) 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

In our notes, we study a model of interacting one-dimensional Brownian motions in

considerable detail. At first sight, the dynamics look simple: the Brownian motions
are ordered, and Brownian motion with label j is reflected from the one with label
j 1. This model belongs to the Kardar–Parisi–Zhang universality class. Adding
to the results as presented in our notes a recent contribution by Kurt Johansson,
this model allows for the so far most complete asymptotic analysis. The limiting
expressions are likely to be valid for all models in the KPZ class.
The notes are based on two joint publications and the Ph.D. Thesis of Thomas
Weiss. Our research was partially supported by the DFG under the grant numbers
SP181/29-1 and SFB1060, project B04.

Bonn, Germany Thomas Weiss

Munich, Germany Patrik Ferrari
September 2016 Herbert Spohn

v
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 One-Sided Reflected Brownian Motions and Related Models . . . . . . . 9
3 Determinantal Point Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4 Airy Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5 Packed and Periodic Initial Conditions . . . . . . . . . . . . . . . . . . . . . . . . . 45
6 Stationary Initial Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7 More General Initial Conditions and Their Asymptotics . . . . . . . . . . . 97

vii
Chapter 1
Introduction

Back in 1931 Hans Bethe diagonalized the hamiltonian of the one-dimensional

Heisenberg spin chain through what is now called the “Bethe ansatz” (Bethe 1931).
At that time physicists were busy with other important developments and hardly real-
ized the monumental step: for the first time a strongly interacting many-body system
had been “solved exactly”. In the 1960s Lieb and Liniger (1963), Yang (1967), and
many more (Sutherland 2004) discovered that other quantum systems can be handled
via Bethe ansatz, which triggered a research area known as quantum integrability.
More details on the history of the Bethe ansatz can be found in Batchelor (2007).
Even with the Bethe ansatz at one’s disposal, it is a highly non-trivial task to arrive at
predictions of physical interest. This is why efforts in quantum integrability continue
even today, reenforced by the experimental realization of such chains through an
array of cold atoms (Simon et al. 2011).
On a mathematical level, quantum hamiltonians and generators of Markov
processes have a comparable structure. Thus one could imagine that the Bethe ansatz
is equally useful for interacting stochastic systems with many particles. The first
indication came somewhat indirectly from the Kardar-Parisi-Zhang equation (Kar-
dar et al. 1986), for short KPZ, in one dimension. We refer to books (Barabasi and
Stanley 1995; Meakin 1998), lecture notes (Johansson 2006; Spohn 2006; Quastel
2011; Borodin and Gorin 2012; Borodin and Petrov 2014; Spohn 2015), and sur-
vey articles (Halpin-Healy and Zhang 1995; Krug 1997; Sasamoto and Spohn 2011;
Ferrari and Spohn 2011; Corwin 2012; Takeuchi 2014; Quastel and Spohn 2015;
Halpin-Healy and Takeuchi 2015). The KPZ equation is a stochastic PDE governing
the time-evolution of a height function h(x, t) at spatial point x and time t, h ∈ R,
x ∈ R, t ≥ 0. The equation reads

∂t h = 21 (∂x h)2 + 21 ∂x2 h + W (1.1)

with W (x, t) normalized space-time white noise. We use here units of height, space,
and time such that all coupling parameters take a definite value. For a solution
of (1.1), the function x → h(x, t) is locally like a Brownian motion, which is
too singular for the nonlinearity to be well-defined as written. This difficulty was
© The Author(s) 2017 1
T. Weiss et al., Reflected Brownian Motions in the KPZ Universality Class,
SpringerBriefs in Mathematical Physics, DOI 10.1007/978-3-319-49499-9_1
2 1 Introduction

resolved through the regularity structures of Hairer (2013), see also Gubinelli and
Perkowski (2015) for a somewhat different approach. Kardar (1987) noted one link to
quantum integrability. He considered the moments of eh and established that they are
related to the δ-Bose gas with attractive interactions, which is an integrable quantum
many-body system (Lieb and Liniger 1963). More precisely, one defines


n 
E eh(yα ,t) = f t (y ) (1.2)
α=1

with y = (y1 , ..., yn ). Then

∂t f t = −Hn f t , (1.3)

where Hn is the n-particle Lieb-Liniger hamiltonian


n 
n
Hn = − 21 ∂ y2α − 1
2
δ(yα − yα ). (1.4)
α=1 α=α =1

Almost thirty years later the generator of the asymmetric simple exclusion process
(ASEP) was diagonalized through the Bethe ansatz. In case of N sites, the ASEP
configuration space is {0, 1} N signalling a similarity with quantum spin chains. In
fact, the ASEP generator can be viewed as the Heisenberg chain with an imagi-
nary XY-coupling. For the totally asymmetric limit (TASEP) and half filled lattice,
Gwa and Spohn (1992) established that the spectral gap of the generator is of order
N −3/2 . The same order is argued for the KPZ equation. This led to the strong belief
that, despite their very different set-up, both models have the same statistical prop-
erties on large space-time scales. In the usual jargon of statistical mechanics, both
models are expected to belong to the same universality class, baptized KPZ univer-
sality class according to its most prominent representative.
The KPZ equation is solved with particular initial conditions. Of interest are
(i) sharp wedge, h(x, 0) = −c0 |x| in the limit c0 → ∞, (ii) flat, h(x, 0) = 0,
and (iii) stationary, h(x, 0) = B(x) with B(x) two sided Brownian motion. The
quantity of prime interest is the distribution of h(0, t) for large t. More ambitiously,
but still feasible in some models, is the large time limit of the joint distribution of
{h(xα , t), α = 1, ..., n}.
In our notes we consider an integrable system of interacting diffusions, which is
governed by the coupled stochastic differential equations

dx j (t) = βe−β(x j (t)−x j−1 (t)) dt + dB j (t), (1.5)

where {B j (t)} a collection of independent standard Brownian motions. For the para-
meter β > 0 we will eventually consider only the limit β → ∞. But for the purpose
of our discussion we keep β finite for a while. In fact there is no choice, no other
system of this structure is known to be integrable. The index set depends on the prob-
lem, mostly we choose j ∈ Z. Note that x j interacts only with its left index neighbor
1 Introduction 3

x j−1 . The drift depends on the slope, as it should be for a proper height function.
But the exponential dependence on x j − x j−1 is very special, however familiar from
other integrable systems. The famous Toda chain (Toda 1967) is a classical integrable
system with exponential nearest neighbour interaction. Its quantized version is also
integrable (Sutherland 1978).
Interaction with only the left neighbor corresponds to the total asymmetric version.
Partial asymmetry would read
 
dx j (t) = pβe−β(x j −x j−1 ) − (1 − p)βe−β(x j+1 −x j ) dt + dB j (t) (1.6)

with 0 ≤ p ≤ 1. These are non-reversible diffusion processes. Only in the symmetric

case, p = 21 , the drift is the gradient of a potential and the diffusion process is
reversible. Then the model is no longer in the KPZ universality class and has very
distinct large scale properties, see Guo et al. (1988); Chang and Yau (1992), for
example.
Equations (1.5) and (1.6) should be viewed as a discretization of (1.1). The inde-
pendent Brownian motions are the natural spatial discretization of the white noise
W (x, t). For the drift one might have expected the form (x j − x j−1 )2 + x j+1 −
2x j + x j−1 , but integrability forces another dependence on the slope. For p = 1 the
similarity with the KPZ equation is even stronger when considering the exponential
moments
n 
E eβxmα (t) = f t (m),
 (1.7)
α=1

 ∈ Zn . Differentiating in t one obtains

m

d n 
n
β −2 f t (m)
 = ∂α f t (m)
 + 1
2
δ(m α − m α ) f t (m)
 − n f t (m),
 (1.8)
dt α=1 α,α =1

where δ is the Kronecker delta and ∂α f (m)  = f (..., m α , ...) − f (..., m α − 1, ...).
When comparing with (1.4), instead of ∂α one could have guessed the discrete Lapla-
cian α = −∂αT ∂α with T denoting the transpose. To obtain such a result, the drift
in (1.5) would have to be replaced by βe−β(x j −x j−1 ) + βe−β(x j −x j+1 ) . But the linear
equations for the exponential moments are no longer Bethe integrable. Note however
that the semigroups exp[∂α t] and exp[α t] differ on a large scale only by a uniform
translation proportional to t. With such a close correspondence one would expect
system (1.5) to be in the KPZ universality class for any β > 0, as has been verified to
some extent (Amir et al. 2011; Sasamoto and Spohn 2010; Borodin et al. 2015). Also
partial asymmetry, 21 < p < 1, should be in the KPZ universality class. In fact, the
true claim is by many orders more sweeping: the exponential in (1.6), p = 21 , can be
replaced by “any” function of x j − x j−1 , except for the linear one, and the system is
still in the KPZ universality class. There does not seem to be a promising idea around
of how to prove such a property. Current techniques heavily rely on integrability.
4 1 Introduction

This is a good opportunity to reflect another difficulty. Bethe ansatz is like a

first indication. But an interesting asymptotic analysis is yet another huge step. This
is well illustrated by the  KPZ equation. Solving the n-particle Eq. (1.2) yields the
exponential moment E enh(0,t) . However these moments grow rapidly as exp(n 3 ),
much too fast to determine the distribution of h(0, t). Because of the underlying
lattice, for system (1.5) the exponential moments grow only as exp(n 2 ), still too
fast. In replica computations one nevertheless continues formally, often with correct
results (Calabrese et al. 2010; Imamura and Sasamoto 2013; Dotsenko 2010). A proof
must exploit integrability, but cannot use exponential moments directly (Borodin and
Corwin 2013).
The system (1.5) simplifies substantially in the limit β → ∞. Then one arrives at
interacting Brownian motions, where the Brownian motions maintain their ordering
and Brownian motion with label j is reflected from its left neighboring Brownian
motion with label j − 1, see Sasamoto and Spohn (2015), Appendix B. These are
the reflected Brownian motions of the title. The proper definition of their dynamics
requires martingales involving local time, as will be discussed in Chap. 2. Our note
discusses exclusively this limit case. Thereby we arrive at a wealth of results on
universal statistical properties. Only for the TASEP a comparably detailed analysis
has been carried out (Ferrari and Spohn 2011), which does not come as a surprise,
since in the limit of low density, under diffusive rescaling of space-time and switching
to a moving frame of reference, the TASEP converges to system (1.5) (Karatzas et al.
2016). We will not exploit this limit. Our philosophy is to work in a framework which
uses only interacting diffusions.
The limit β → ∞ is meaningful also for p = 0, 1. Then the order of Brownian
particles is still preserved, but the reflection between neighbors is oblique. The sym-
metric version, p = 21 , corresponds to independent Brownian motions, maintaining
their order, a case which has been studied quite some time ago (Harris 1965). The
partially asymmetric version of the model is still Bethe integrable, but less tractable.
Only for the half-Poisson initial condition, an expression sufficiently compact for
asymptotic analysis has been obtained (Sasamoto and Spohn 2015).
The notion of integrability was left on purpose somewhat vague. In the β = ∞
limit for (1.5), integrability can be more concretely illustrated. For this case, let us set
j = 2, ..., n with x1 (t) a standard Brownian motion. Then the transition probability
from x to y at time t is given by
  
P x(t) ∈ d yx(0) = x = det{Φt
(i− j)
(y j − xi )}1≤i, j≤n d y, (1.9)

where
( j) 1 /2+ξw
w− j
2
Φt (ξ, t) = dw etw (1.10)
2πi iR+δ

with δ > 0 as first established by Sasamoto and Wadati (1998). There is a similar
formula for the TASEP (Schütz 1997). Such formuli nourish the hope to uncover
interesting features of the model.
1 Introduction 5

The three initial conditions of particular interest, wedge, flat, and stationary, are
easily transcribed to system (1.5) and become (i) packed, half-infinite system with
x j (0) = 0 for j = 1, 2, ..., (ii) periodic, x j (0) = j for j ∈ Z, (iii) Poisson,
{x j (0), j ∈ Z} is a Poisson process with constant density. According to our discus-
sion, in the latter case one might think that the quantity of prime interest is x1 (t). But
the reflection induces a propagation of statistical fluctuations, as can also be seen
from (1.7) together with (1.8). Their propagation speed is 1 and the correct quantity
is x t (t) with t denoting integer part. Along other space-time observation rays a
central limit type behavior would be observed.
As for other models in the KPZ universality class, our asymptotic analysis is lim-
ited to a single time and arbitrary number of spatial, resp. index points. Only recently
Johansson (2016) posted a result on the joint distribution of (x t (t), x αt (αt)),
α > 0, and identified its universal limit. Possibly such progress will lead eventually
to a complete understanding of the Airy sheet and the KPZ fixed point (Corwin et al.
2015).
Following the format of Springerbriefs, there is a short summary for each chapter.
But let us still explain of how our material is organized as a whole. The following
three chapters provide background material. In Chap. 2 we properly define the infinite
system of reflected Brownian motions as the solution of a martingale problem and
provide a variational formulation of this solution. Also the uniform Poisson process is
identified as stationary measure. In Chap. 3 we introduce the theory of determinantal
point processes and some related material on Fredholm determinants. At first sight
this looks unconnected. But to study the quantities of prime interest one first identifies
a “hidden” signed determinantal process which leads to an analytically more tractable
representation. In the long time limit we will arrive at a stochastic process which
describes the limiting spatial statistics. Such a process has been baptised Airy process,
in analogy to the Airy kernel and Airy operator which are one of the defining elements.
In fact there are several Airy processes depending on the initial conditions and on
the window of observation. The literature on Airy processes is somewhat dispersed.
Chapter 4 provides a streamlined account.
In Chap. 5 we investigate the two deterministic initial data, packed and periodic,
while in Chap. 6 we study random initial data as defined through a Poisson process.
These results will be used to discuss more general initial data, which should be
viewed as an open ended enterprise. One natural choice is to have in the left half
lattice either packed, periodic, or Poisson join up with either one of them in the right
half lattice. An example would be to have periodic to the left and Poisson to the
right. This then leads to distinct cross over processes. The mixed cases are studied
in Chap. 7. Slow decorrelation, referring to space-like paths more general than fixed
time, is a further topic. Each core chapter builds on a suitable asymptotic analysis.
In the early days the required techniques were developed ad hoc for the particular
model. Over the years a common strategy based on contour integrations has been
established, which will be also used here. Thus on the basis of a specific example
one can learn a technique applicable also to other models.
6 1 Introduction

References

H. Bethe. Zur Theorie der Metalle. I. Eigenwerte und Eigenfunktionen der linearen Atomkette.
Zeitschrift für Physik 71, 205–226 (1931)
E.H. Lieb, W. Liniger, Exact analysis of an interacting Bose gas. I. the general solution and the
ground state. Phys. Rev. 130, 1605–1616 (1963)
C.N. Yang, Some exact results for the many-body problem in one dimension with repulsive delta-
function interaction. Phys. Rev. Lett. 19, 1312–1315 (1967)
B. Sutherland, Beautiful Models. 70 Years of Exactly Solved Quantum Many-Body Problems (World
Scientific Publishing, New Jersey, 2004)
M.T. Batchelor, The Bethe ansatz after 75 years. Phys. Today 60, 36 (2007)
J. Simon, W.S. Bakr, R. Ma, M.E. Tai, P.M. Preiss, M. Greiner, Quantum simulation of antiferro-
magnetic spin chains in an optical lattice. Nat. 472, 307–312 (2011)
M. Kardar, G. Parisi, Y.Z. Zhang, Dynamic scaling of growing interfaces. Phys. Rev. Lett. 56,
889–892 (1986)
A.L. Barabasi, H.E. Stanley, Fractal Concepts in Surface Growth (Cambridge University Press,
Cambridge, 1995)
P. Meakin, Fractals Scaling and Growth Far from Equilibrium (Cambridge University Press, Cam-
bridge, 1998)
K. Johansson, Random matrices and determinantal processes. Math. Stat. Phys. Sess. LXXXIII:
Lect. Notes Les Houches Summer Sch. 2005, 1–56 (2006)
H. Spohn, Exact solutions for KPZ-type growth processes, random matrices, and equilibrium shapes
of crystals. Phys. A 369, 71–99 (2006)
J. Quastel, Introduction to KPZ. Curr. Dev. Math. 2011, 125–194 (2011)
A. Borodin and V. Gorin. Lectures on integrable probability, (2012). arXiv:1212.3351
A. Borodin, L. Petrov, Integrable probability: from representation theory to Macdonald processes.
Probab. Surv. 11, 1–58 (2014)
H. Spohn. The Kardar-Parisi-Zhang equation - a statistical physics perspective. in Stochastic
Processes and Random Matrices, École d’Été Physique, Les Houches (Oxford University Press,
2015). arXiv:1601.00499
T. Halpin-Healy, Y.-C. Zhang, Kinetic roughening phenomena, stochastic growth, directed polymers
and all that. Phys. Rep. 254, 215–414 (1995)
J. Krug, Origins of scale invariance in growth processes. Adv. Phys. 46, 139–282 (1997)
T. Sasamoto and H. Spohn, The 1+1-dimensional Kardar-Parisi-Zhang equation and its universality
class. J. Stat. Mech. P01031 (2011)
P.L. Ferrari and H. Spohn. Random growth models. in The Oxford Handbook of Random Matrix
Theory, ed. By J. Baik, G. Akemann, P. Di Francesco (2011)
I. Corwin, The Kardar-Parisi-Zhang equation and universality class. Random Matrices Theory Appl.
1, 1130001 (2012)
K. Takeuchi, Experimental approaches to universal out-of-equilibrium scaling laws: turbulent liquid
crystal and other developments. J. Stat. Mech. P01006 (2014)
J. Quastel, H. Spohn, The one-dimensional KPZ equation and its universality class. J. Stat. Phys.
160, 965–984 (2015)
T. Halpin-Healy, K. Takeuchi, A KPZ cocktail-shaken, not stirred: Toasting 30 years of kinetically
roughened surfaces. J. Stat. Phys. 160, 794–814 (2015)
M. Hairer, Solving the KPZ equation. Ann. Math. 178, 559–664 (2013)
M. Gubinelli and N. Perkowski. KPZ reloaded, (2015). arXiv:1508.03877
M. Kardar, Replica Bethe ansatz studies of two-dimensional interfaces with quenched random
impurities. Nucl. Phys. B 290, 582–602 (1987)
L.-H. Gwa, H. Spohn, Bethe solution for the dynamical-scaling exponent of the noisy Burgers
equation. Phys. Rev. A 46, 844–854 (1992)
M. Toda, Vibration of a chain with nonlinear interaction. J. Phys. Soc. Jpn. 22, 431 (1967)
References 7

B. Sutherland, A brief history of the quantum soliton with new results on the quantization of the
toda lattice. Rocky Mt. J. Math. 8, 413–430 (1978)
M.Z. Guo, G.C. Papanicolaou, S.R.S. Varadhan, Nonlinear diffusion limit for a system with nearest
neighbor interactions. Commun. Math. Phys. 118, 31–59 (1988)
C.-C. Chang, H.-T. Yau, Fluctuations of one dimensional Ginzburg-Landau models in nonequilib-
rium. Commun. Math. Phys. 145, 209–234 (1992)
G. Amir, I. Corwin, J. Quastel, Probability distribution of the free energy of the continuum directed
random polymer in 1+1 dimensions. Commun. Pure Appl. Math. 64, 466–537 (2011)
T. Sasamoto, H. Spohn, Exact height distributions for the KPZ equation with narrow wedge initial
condition. Nucl. Phys. B 834, 523–542 (2010)
A. Borodin, I. Corwin, P.L. Ferrari, B. Vető, Height fluctuations for the stationary KPZ equation.
Math. Phys. Anal. Geom. 18, 1–95 (2015)
P. Calabrese, P. Le Doussal, A. Rosso, Free-energy distribution of the directed polymer at high
temperature. EPL 90, 20002 (2010)
T. Imamura, T. Sasamoto, Stationary correlations for the 1D KPZ equation. J. Stat. Phys. 150,
908–939 (2013)
V. Dotsenko, Replica Bethe ansatz derivation of the Tracy-Widom distribution of the free energy
fluctuations in one-dimensional directed polymers. J. Stat. Mech. P07010, (2010)
A. Borodin and I. Corwin, Macdonald processes. Probab. Theory Relat. Fields (online first), (2013)
T. Sasamoto, H. Spohn, Point-interacting Brownian motions in the KPZ universality class. Electron.
J. Probab. 20(87), 28 (2015)
I. Karatzas, S. Pal, M. Shkolnikov, Systems of Brownian particles with asymmetric collisions. Ann.
Inst. H. Poincaré Probab. Stat. 52, 323–354 (2016)
T.E. Harris, Diffusion with collisions between particles. J. Appl. Probab. 2, 323–338 (1965)
T. Sasamoto, M. Wadati, Determinantal form solution for the derivative nonlinear Schrödinger type
model. J. Phys. Soc. Jpn. 67, 784–790 (1998)
G.M. Schütz, Exact solution of the master equation for the asymmetric exclusion process. J. Stat.
Phys. 88, 427–445 (1997)
K. Johansson. Two time distribution in Brownian directed percolation. Comm. Math. Phys., online
first:1–52, (2016)
I. Corwin, J. Quastel, D. Remenik, Renormalization fixed point of the KPZ universality class. J.
Stat. Phys. 160, 815–834 (2015)
Chapter 2
One-Sided Reflected Brownian Motions
and Related Models

For reflected Brownian motions with index set Z+ we properly define the dynamics
using an iterated Skorokhod construction. The full index set Z requires a proof
of well-posedness. In addition, we establish that the uniform Poisson process is
stationary under the dynamics and discuss previous results for the model.

2.1 Skorokhod Construction

Our definition of reflected Brownian motions is through the so-called Skorokhod

representation (Skorokhod 1961; Anderson and Orey 1976), which is a deterministic
function of the driving Brownian motions. This representation is the following: the
process x(t), driven by the Brownian motion B(t), starting from x(0) ∈ R and
being reflected (in the positive direction) at some continuous function f (t) with
f (0) ≤ x(0) is defined as:
 
x(t) = x(0) + B(t) − min 0, inf (x(0) + B(s) − f (s))
0≤s≤t
  (2.1.1)
= max x(0) + B(t), sup ( f (s) + B(t) − B(s)) .
0≤s≤t

Let Bn , n ∈ Z, be independent standard Brownian motions starting at 0. Through-

out this work, Bn always denotes these Brownian motions, allowing for coupling
arguments. The non-positive indices will be used from Sect. 2.3 on.

Definition 2.1 The half-infinite system of one-sided reflected Brownian motions

 ζn ≤ ζn+1 is defined recursively by
{xn (t), n ≥ 1} with initial condition x(0) = ζ,
x1 (t) = ζ1 + B1 (t) and, for n ≥ 2,
 
xn (t) = max ζn + Bn (t), sup (xn−1 (s) + Bn (t) − Bn (s)) . (2.1.2)
0≤s≤t

© The Author(s) 2017 9

T. Weiss et al., Reflected Brownian Motions in the KPZ Universality Class,
SpringerBriefs in Mathematical Physics, DOI 10.1007/978-3-319-49499-9_2
10 2 One-Sided Reflected Brownian …

Introducing the random variables


n
Yk,n (t) = sup (Bi (si ) − Bi (si−1 )) (2.1.3)
0≤sk ≤...≤sn−1 ≤t i=k

for k ≤ n, with the convention sk−1 = 0 and sn = t, allows for an equivalent explicit
expression:
xn (t) = max {Yk,n (t) + ζk }. (2.1.4)
k∈[1,n]

Although we are constructing an infinite system of particles, well-definedness is clear

in this case, as each process xn (t) is a deterministic function of only finitely many
Brownian motions Bk , 1 ≤ k ≤ n.
Adopting a stochastic analysis point of view, the system {xn (t), n ≥ 1} satisfies

xn (t) = ζn + Bn (t) + L n (t), for n ≥ 0, (2.1.5)

Here, L 1 (t) = 0, while L n , n ≥ 2, are continuous non-decreasing processes increas-

ing only when xn (t) = xn−1 (t). In fact, L n is twice the semimartingale local time at
zero of xn − xn−1 .

2.2 Packed Initial Conditions

A canonical and in fact the most studied initial condition for the system {xn (t), n ≥ 1}
is the one where all particles start at zero. We call this packed initial condition:

x(0) = ζpacked = 0. (2.2.1)

Using the monotonicity,


n
Yk−1,n (t) = sup (Bi (si ) − Bi (si−1 ))
0≤sk−1 ≤sk ≤...≤sn−1 ≤t i=k−1
(2.2.2)
n
≥ sup (Bi (si ) − Bi (si−1 )) = Yk,n (t),
0=sk−1 ≤sk ≤...≤sn−1 ≤t i=k

and inserting this initial condition into (2.1.4), leads to:


n
xn (t) = Y1,n (t) = sup (Bi (si ) − Bi (si−1 )), (2.2.3)
0≤s1 ≤...≤sn−1 ≤t i=1

again with the convention s0 = 0 and sn = t.

2.2 Packed Initial Conditions 11

2.2.1 Queues, Last Passage Percolation

and Directed Polymers

There are other interpretations for the quantity xn (t) than the system of reflected
Brownian motions focused on in this work, and each interpretation has inspired
different results over the last decades. One of these is seeing it as a sequence of
Brownian queues in series. A famous theorem of queueing theory, Burke’s theorem,
which states that the output of a stable, stationary M/M/1 queue is Poisson, can be
adapted to the Brownian setting (O’Connell and Yor 2001). This will be employed
in Sect. 2.3.4.
Furthermore, xn (t) can be viewed as a model of directed last-passage percolation
through a random medium, or equivalently a zero-temperature directed polymer in
a random environment. This model is constructed as follows:
Consider the space R+ × Z and assign white noise dBn as random background
weight on each line R+ × {n} for n ≥ 1. An up-right path is characterized by its
jumping points si and consists of line segments [sn−1 , sn ] × {n}, see Fig. 2.1. The set
of up-right paths going from (t1 , n 1 ) to (t2 , n 2 ) can then be parameterized by
 
Π (t1 , n 1 ; t2 , n 2 ) = s ∈ Rn 2 −n 1 +2 |t1 = sn 1 −1 ≤ sn 1 ≤ · · · ≤ sn 2 = t2 . (2.2.4)

The percolation time or weight of a path π ∈ Π is the integral over the background
weights along the path. Explicitly, we have:


n2
w(
π) = (Bi (si ) − Bi (si−1 )) . (2.2.5)
i=n 1

The last passage percolation time is given by the supremum over all such paths:

L (t1 ,n 1 )→(t2 ,n 2 ) := sup w(

π ). (2.2.6)
π ∈Π(t1 ,n 1 ;t2 ,n 2 )

Fig. 2.1 A path

π ∈ Π (0, 1; t, 5) (thick
black) and the random 5
background noise (gray)
4

1
s1 s2 s3 s4 t
12 2 One-Sided Reflected Brownian …

The supremum is almost surely attained by a unique path π ∗ , called the maximizer. It
exists because the supremum can be rewritten as a composition of a finite maximum
and a supremum of a continuous function over a compact set. Uniqueness follows
from elementary properties of the Brownian measure. Most importantly, from the
definition, we have
xn (t) = L (0,1)→(t,n) . (2.2.7)

This representation will be used repeatedly throughout this work as it nicely

visualizes coupling arguments, however, it also gives some connections to different
works. Our model can be seen as the semi-continuous limit of a more widely studied
discrete last passage percolation model (see for example Johansson (2000, 2003)).
This last passage percolation model is also the zero temperature limit of a directed
polymer model, which has been studied thoroughly in the recent past (Seppäläinen
and Valkó 2010; Borodin et al. 2015). In the directed polymer setting we have a
parameter β representing the inverse temperature, consider w( π ) as an energy and
assign a Gibbs measure on the set of paths according to the density eβw(π) , i.e. paths
with higher energy have a higher probability. The partition function of the polymer
is given by 
Z (t1 ,n 1 )→(t2 ,n 2 ) (β) = π eβw(π) ,
d (2.2.8)
Π(t1 ,n 1 ;t2 ,n 2 )

and satisfies the limit

1
lim log Z (t1 ,n 1 )→(t2 ,n 2 ) (β) = L (t1 ,n 1 )→(t2 ,n 2 ) . (2.2.9)
β→∞ β

Although apparently more difficult to handle, attention has been turning to these
positive temperature models recently, because, among other things, they allow for a
scaling limit to the KPZ equation by tuning the parameter β in the right way.

2.2.2 Previous Results

The behaviour of xn (t) under the packed initial condition is quite well under-
stood by now. Notice that by Brownian scaling we have the distributional identity
d √
{xn (t), n ≥√1} = { t xn (1), n ≥ 1}. The first result has been a law of large numbers,
i.e. xn (1)/ n converges to a constant (Glynn and Whitt 1991), therein it was already
conjectured it equals 2, which has been proven subsequently (Seppäläinen 1997).
Much of the results that followed exploited connections to a random matrix model,
so let us introduce it in full generality right away:
Let bi,i (t) for 1 ≤ i ≤ N and bi, j (t), bi, j (t) for 1 ≤ i < j ≤ N , be independent
Brownian motions. Define a stochastic process H (t), t ≥ 0, on the space of N × N
Hermitian matrices by
2.2 Packed Initial Conditions 13

Hi,i (t) = bi,i (t) for 1 ≤ i ≤ N

√  
Hi, j (t) = 2 bi, j (t) + ibi, j (t) for 1 ≤ i < j ≤ N (2.2.10)
√  
Hi, j (t) = 2 bi, j (t) − ibi, j (t) for 1 ≤ j < i ≤ N

Denote by λn1 (t) ≤ λn2 (t) ≤ · · · ≤ λnn (t) the ordered eigenvalues of the n × n
principal minor of the matrix H (t). The process {λnk (t), 1 ≤ k ≤ n ≤ N } is called
the Dyson Brownian minor process. It is a classical result, that the eigenvalues of
consecutive minors are interlaced, λkn+1 (t) ≤ λnk (t) ≤ λk+1
n+1
(t), so this process lives
in the Gelfand–Tsetlin cone:
 
GT N = xkn ∈ R, 1 ≤ k ≤ n ≤ N , xkn+1 ≤ xkn ≤ xk+1
n+1
. (2.2.11)

Restricted to one layer n, the process {λnk (t), 1 ≤ k ≤ n} is called Dyson’s

Brownian motion (Dyson 1962). It is a Markov process and satisfies the coupled
stochastic differential equation
 dt
dλnk = dBk + . (2.2.12)
i =k
λnk − λin

Its fixed time distribution is the eigenvalue distribution of the Gaussian unitary
ensemble (GUE):

  1  − yk2 
N
P λnk (t) ∈ dyk , 1 ≤ k ≤ n = e 2t (yi − yk )2 d y. (2.2.13)
Z n (t) k=1 k<i

Recognizing the second product as the square of the Vandermonde determinant,

one is able to describe this eigenvalue distribution as a determinantal point process
governed by the Hermite kernel. Asymptotic analysis of the distribution of the
largest eigenvalue in the appropriate edge scaling gives the GUE Tracy–Widom
distribution (Tracy and Widom 1994).
The first connections between our system of reflecting Brownian motions and the
matrix diffusion were found by Gravner et al. (2001), proving that for every n ≥ 1,
d d
xn (1) = λnn (1), and Baryshnikov (2001) generalizing this to {xn (1), n ≥ 1} =
{λn (1), n ≥ 1} by a combinatorial procedure originating from group representation
n

theory, called the Robinson–Schensted–Knuth (RSK) correspondence. Baryshnikov

(2001) also showed the remarkable fact that conditioned on the top layer {λkN (1),
1 ≤ k ≤ N } the distribution of {λnk (1), 1 ≤ k ≤ n < N } is uniform on
the compact set given by the Gelfand–Tsetlin interlacing inequality. Restricting
the Dyson Brownian minor process to a fixed time gives the GUE minor process
{λnk (1), 1 ≤ k ≤ n ≤ N }, whose full distribution has been found in Johansson and
Nordenstam (2006), again in the form of a determinantal point process.
There is a natural extension of the system of reflecting Brownian motions
{xn (t), n ≥ 1} to a process in the Gelfand–Tsetlin cone that is constructed in the
following way: Let B11 (t) be a Brownian motion. Let B12 (t) and B22 (t) be Brownian
14 2 One-Sided Reflected Brownian …

motions, which are reflected downwards resp. upwards from B11 (t). Iteratively con-
struct Bkn (t) as a Brownian motion reflected downwards from Bkn−1 (t) and upwards
n−1
from Bk−1 (t), with the peripheral processes Bkn (t) for k = 1 or k = n being
reflected from one process only. By construction, we have xn (t) = Bnn (t). The
process {Bkn (t), 1 ≤ k ≤ n ≤ N } is called Warren’s process, and has been intro-
duced and studied in Warren (2007). Restricted to one layer n, it is distributed as a
Dyson’s Brownian motion. Warren’s process shares the fixed time distribution with
the GUE minor process,

d
{λnk (1), 1 ≤ k ≤ n ≤ N } = {Bkn (1), 1 ≤ k ≤ n ≤ N }. (2.2.14)

There are also formulas for the transition density of the system along the edge, {Bnn (t),
1 ≤ n ≤ N } as well as for the system of two consecutive layers
{Bnk (t), 1 ≤ k ≤ n, N − 1 ≤ n ≤ N }.
The connection between Warren’s process and the Dyson Brownian minor process
does not, however, hold in full generality. The common dynamics of any amount
of consecutive layers of Warrens process, i.e. the process {Bnk (t), 1 ≤ k ≤ n,
N1 ≤ n ≤ N2 }, is simply given by Dyson’s SDE (2.2.12) for the layer n = N1
and the reflection SDE’s for the higher order layers. In the Dyson Brownian minor
process, on the other hand, the common dynamics of two consecutive layers is given
by a more complicated SDE (see (2.30) in Adler et al. (2014)) and the evolution of
three or more consecutive layers is not even a Markov process anymore. Interestingly,
both processes still show the same distribution along so-called space-like paths, i.e.
sequences of points (n i , ti ) satisfying ti ≤ ti+1 and n i ≥ n i+1 , in which case also
determinantal formulas exist (Adler et al. 2014; Ferrari and Frings 2010).
The determinantal formulas coming from the random matrix model are suitable
for asymptotic analysis to show multi-point scaling limits, where the Airy2 process
arises. It appears both for correlations of xn (t) along the n direction and the t direction,
as well as along general space-like paths. Reference (Johansson 2003) gives a sketch
of the proof for both directions, (Adler and Moerbeke 2005) prove the scaling limit
with correlations along t rigorously. A complete proof for correlations along n is
given in Sect. 5.1.

2.3 Infinite Particle Systems

The main focus of this work is showing determinantal formulas and scaling limits
for other initial conditions. Unfortunately the connection to random matrices breaks
down in this case. In fact, neither the Airy1 process nor the Airystat process have
ever been found in a scaling limit of a random matrix model. This has been rather
surprising, as the one-point distribution of the Airy1 process, the GOE Tracy–Widom
distribution, does arise in such a model, namely as the limiting distribution of the
largest eigenvalue of a Gaussian real symmetric matrix, the GOE ensemble (Borne-
mann et al. 2008).
2.3 Infinite Particle Systems 15

2.3.1 Definition

At first sight it might seem trivial to extend the definition of the half-infinite system
of one-sided reflected Brownian motions to an infinite number of particles {xn (t),
n ∈ Z}, by letting the index k run over (−∞, n] in (2.1.4). However, it has to be
shown that this maximum is finite, which is only the case for initial conditions which
are not too √closely spaced together. Roughly said, the growth rate of −ζ−k has to be
faster than k for large k. We call those initial conditions admissible.
Knowing the law of large numbers under packed initial conditions, it√is reasonable
to expect such a behaviour. It implies that Yk,n (t) grows roughly as 2 (n − k)t, so
in order for the maximum
√ in (2.3.2) being attained by a finite k for we need −ζ−k to
grow faster than k.

Definition 2.2 A random vector ζ ∈ RZ with ζn ≤ ζn+1 for all n ∈ Z is called an

admissible initial condition, if there exists a χ > 21 such that for any n ∈ Z the sum

P (ζn − ζ−M ≤ M χ ) (2.3.1)
M≥0

is finite.

Definition 2.3 Let ζ ∈ RZ be an admissible initial condition. The infinite system of

one-sided reflected Brownian motions {xn (t), n ∈ Z} with initial condition x(0) = ζ
is defined by
xn (t) = max{Yk,n (t) + ζk }. (2.3.2)
k≤n

By Proposition 2.4 below, which is proven in Sect. 2.3.2, this maximum exists and
is finite. More specifically, we will show that for ζ being any admissible initial con-
dition, the infinite system {xn (t), n ∈ Z} is the limit of certain half-infinite systems
{xn(M) (t), n ≥ −M} as M → ∞, where

xn(M) (t) = max {Yk,n (t) + ζk }, n ≥ −M. (2.3.3)

k∈[−M,n]

Notice that these processes indeed satisfy the Skorokhod equation,

 (M) 
xn(M) (t) = max ζn + Bn (t), sup (xn−1 (s) + Bn (t) − Bn (s)) , (2.3.4)
0≤s≤t

for n > −M, while the leftmost process is simply

(M)
x−M (t) = ζ−M + B−M (t). (2.3.5)

Thus as desired xn(M) (t) is a Brownian motion starting from ζn and reflected off by
(M)
xn−1 for n > −M.
16 2 One-Sided Reflected Brownian …

Proposition 2.4 For any t > 0, n ∈ Z there exists almost surely a k ≤ n maximizing
Yk,n (t) + ζk , i.e. the maximum in (2.3.2) exists. Furthermore, for any T > 0,

sup |xn (t)| < ∞, a.s., (2.3.6)

t∈[0,T ]

as well as
lim sup |xn(M) (t) − xn (t)| = 0, a.s.. (2.3.7)
M→∞ t∈[0,T ]

The convergence result allows for taking the limit in (2.3.4), implying that the system
{xn (t), n ∈ Z} satisfies the Skorokhod equation, too.
The initial conditions we are actually interested in are corresponding to the two
remaining fundamental geometries in the KPZ universality class. The first one is the
flat surface, which translates into periodic initial conditions x(0) = ζflat , defined by

ζnflat = n, for n ∈ Z, (2.3.8)

which is obviously admissible.

Finally we will study the case where the model starts in its random stationary
distribution, which is in our case a Poisson point process on the real line. However,
as already familiar from other models in the KPZ universality class (Borodin et al.
2015; Baik et al. 2010; Ferrari and Spohn 2006; Imamura and Sasamoto 2004),
Theorem 6.1 will be proven via a sequence of approximating initial conditions.
Let therefore be {Expn , n ∈ Z} be i.i.d. random variables with exponential distri-
bution with parameter 1. For parameters λ > 0 and ρ > 0 define the initial condition
x(0) = ζstat (λ, ρ) by

ζ0stat = 0,
λ−1 Expn , for n > 0, (2.3.9)
ζnstat − ζn−1
stat
=
ρ−1 Expn , for n ≤ 0.

Admissibility of this initial condition is also not hard to prove.

To recover the uniform Poisson process on the whole real line, we will set λ = 1
and carefully take the limit ρ → 1 in the determinantal formulas that hold in the case
ρ < λ. Finally, setting ζ0 = 0 will induce a difference of order one as compared to
the true Poisson process case. By Proposition 7.11 this difference will stay bounded
at all times, and consequently be irrelevant in the scaling limit. Thus it is enough to
prove Theorem 6.1 for the initial conditions x(0) = ζstat (1, 1).

2.3.2 Well-Definedness

For the proof of Proposition 2.4 we first need the following concentration inequality:
2.3 Infinite Particle Systems 17

Proposition 2.5 For each T > 0 there exists a constant C > 0 such that for all
k < m, δ > 0,

Yk,m (T )
≥ 2 + δ ≤ const · e−(m−k+1) δ .
2/3
P √ (2.3.10)
(m − k + 1)T

This proposition is proven in Sect. 5.1.2. Another necessary lemma, that will be
proven an intuitive way in Sect. 2.3.3, is:
Lemma 2.6 Consider 0 ≤ t1 ≤ t2 and m, Mt1 , Mt2 such that

(Mti ) (Mti )
xm (ti ) = xm (ti ) = xm (ti ), for i = 1, 2. (2.3.11)

Then
(Mt2 ) (Mt2 )
xm (t1 ) = xm (t1 ) = xm (t1 ). (2.3.12)

Proof (Proof of Proposition 2.4) Let us define an auxiliary system of processes,

which we will use later in proving Proposition 2.7, by
(M)
x−M (t) = ζ−M + B−M (t) + ρt, (2.3.13)

and
 (M) 
xn(M) (t) = max ζn + Bn (t), sup (xn−1 (s) + Bn (t) − Bn (s)) (2.3.14)
0≤s≤t

for n > −M. This system differs from xn(M) (t) just in the drift of the leftmost particle,
which of course influences all other particles as well (the choice of the extra drift
is because the system with infinite many particles in R− generates a drift ρ). This
system of particles satisfies
 
xn(M) (t) = max Y−M,n (t) + ζ−M , max {Yk,n (t) + ζk } , (2.3.15)
k∈[−M+1,n]

with

n 
Yk,n (t) = sup ρsk+1 + (Bi (si+1 ) − Bi (si )) . (2.3.16)
0≤sk+1 ≤...≤sm ≤t i=k

Also, we have the inequalities

Yk,n (t) ≤ Yk,n (t) ≤ Yk,n (t) + ρt. (2.3.17)

Consider the event


A M := {Y−M,n (T ) ≥ 3 (M + n + 1)T } ∪ {ζn − ζ−M ≤ M χ }
 (2.3.18)
∪ {Yn,n (T ) ≤ ρT + 3 (M + n + 1)T − M χ }.
18 2 One-Sided Reflected Brownian …


It is now straightforward to show ∞ M=0 P(A M ) < ∞. In fact, summability of the
probabilities of the first set in (2.3.18) is a consequence of Proposition 2.5, applied
with δ = 1, while the second set is covered by the definition of an admissible initial
condition. For the third set, notice that the left hand side is a Gaussian distribution
independent of M, while the right hand side is dominated by the M χ term for large
M. Finiteness of the sum allows applying Borel–Cantelli, i.e. A M occurs only finitely
many times almost surely. This means, that a.s. there exists a MT , such that for all
M ≥ MT the following three inequalities hold:

Y−M,n (T ) < 3 (M + n + 1)T
M χ < ζn − ζ−M (2.3.19)

− Yn,n (T ) < −ρT − 3 (M + n + 1)T + M χ

Adding up these, Y−M,n (T ) + ζ−M + ρT < Yn,n (T ) + ζn for all M ≥ MT and

dropping the term ρT shows us that the maximizing element in (2.3.2) cannot be a
k ≤ −MT , or
xn (T ) = xn(MT ) (T ). (2.3.20)

Moreover, applying (2.3.17), gives

Y−MT ,n (t) + ζ−MT ≤ Y−MT ,n (t) + ζ−MT + ρT < Yn,n (T ) + ζn , (2.3.21)

resulting in
xn(MT ) (T ) = xn(MT ) (T ). (2.3.22)

Repeating the same argument, we see that for every t ∈ [0, T ] there exists Mt
such that xn (t) = xn(Mt ) (t) = xn(Mt ) (t). Applying Lemma 2.6 then gives xn (t) =
xn(MT ) (t) = xn(MT ) (t) for every t ∈ [0, T ]. This settles the convergence and the
existence of a finite maximizing k in (2.3.2).
To see (2.3.6), which is equivalent to supt∈[0,T ] |xm(MT ) (t)| < ∞, we apply the
bound
n 
|Yk,n (t)| ≤ sup Bi (s) − inf Bi (s) < ∞. (2.3.23)
0≤s≤t 0≤s≤t
i=k

2.3.3 Last Passage Percolation

It is also possible to extend the last passage percolation interpretation to nontrivial

initial conditions. In order to do this, add non-negative Dirac background weights
ζk − ζk−1 on (0, k), k ∈ Z. The weight of a path is explicitly given by
2.3 Infinite Particle Systems 19


n2
 
w(
π) = Bi (si ) − Bi (si−1 ) + (ζi − ζi−1 )1si−1 =0 , (2.3.24)
i=n 1

and the percolation time L (0,n 1 )→(t,n 2 ) again as the supremum over the weight of all
paths. As t → 0 it is clear that any contribution from the Brownian background
weight will converge to 0, so the path tries to accumulate as much of the Dirac
weights as possible, i.e. we have the initial condition


n2
lim L (0,n 1 )→(t,n 2 ) = (ζi − ζi−1 ) = ζn 2 − ζn 1 −1 . (2.3.25)
t→0
i=n 1

By defining a normalized percolation time,


L (0,n 1 )→(t,n 2 ) = L (0,n 1 )→(t,n 2 ) + ζn 1 −1 , (2.3.26)

we recover the system

xn(M) (t) = 
L (0,−M)→(t,n) . (2.3.27)

For any M ≤ n, M = −∞ included, we can define an exit point of a path

π = (. . . , sn 2 −1 , sn 2 ) ∈ Π (0, −M; t, n) by

inf{k ∈ [n 1 , n 2 ], sk > 0}, (2.3.28)

which is of course the maximizing index k in (2.3.3).

We also can reproduce the system xn(M) (t) by adding a Lebesgue measure of
density ρ on the line {−M} × R+ .

Proof (Proof of Lemma 2.6) For t1 = t2 there is nothing to prove, so let t1 < t2 . For
(Mt )
each i, the equation xm (ti ) = xm i (ti ) implies that the maximizing paths of the LHS
and the RHS are equal on the restriction to (si , i ≥ Mi ), i.e. they have the same exit
point ei that satisfies ei ≥ Mti . Now if e1 ≥ e2 , then also e1 ≥ Mt2 , which means
that the path maximizing xm (t1 ) is contained in the set Π (0, −Mt2 ; t1 , m), resulting
(Mt )
in xm (t1 ) = xm 2 (t1 ).
If, however, e1 < e2 , then the maximizing path segments (0, e1 ) → (t1 , m)
and (0, e2 ) → (t2 , m) would need to have an intersection point (t ∗ , m ∗ ). We can
then construct a new maximizing path for xm (t1 ) by stringing together the seg-
ments (0, e1 ) → (0, e2 ) → (t ∗ , m ∗ ) → (t1 , m), where the middle segment is part
of the xm (t2 )-maximizing path and the last segment is part of the original xm (t1 )-
maximizing path. This contradicts the uniqueness of the maximizing path.
(Mt )
The equality xm (t1 ) = xm 2 (t1 ) is shown in the same way.
20 2 One-Sided Reflected Brownian …

2.3.4 Stationarity

We establish a useful property which will allow us to study our system of interacting
Brownian motions through a system with a left-most Brownian particle.
Proposition 2.7 Under the initial condition x(0) = ζ = ζstat (λ, ρ) defined in
(2.3.9), for each n ≤ 0 the process

xn (t) − ζn − ρt (2.3.29)

is a standard Brownian motion.

Remark 1 Proposition 2.7 allows us to restrict our attention to the half-infinite sys-
tem. In fact, conditioned on the path of x0 , the systems of particles {xn (t), n < 0}
and {xn (t), n > 0} are independent, as it is clear by the definition of the system. Then
(2.3.29) implies that the law of {xn (t), n > 0} is the same as the one obtained replac-
ing the infinitely many particles {xm (t), m ≤ 0} with a single Brownian motion
x0 (t) which has a drift ρ. This property will be used to derive our starting result,
Proposition 6.2.

Remark 2 From a stochastic analysis point of view, we find that the system
{xn (t), n ≥ 0} satisfies

xn (t) = ζn + Bn (t) + L n (t), for n ≥ 1,

(2.3.30)
x0 (t) = B0 (t) + ρt.

Here L n , n ≥ 2, are continuous non-decreasing processes increasing only when

xn (t) = xn−1 (t). In fact, L n is twice the semimartingale local time at zero of xn −xn−1 .
Notice that B0 (t) is a standard Brownian motion independent of {ζn , Bn (t), n ≥ 1},
but not equal to B0 (t).

Proof (Proof of Proposition 2.7) First notice that for any M,

(M)
x−M (t) − ζ−M − ρt, (2.3.31)

(M)
is a Brownian motion. Now assume xn−1 (t) − ζn−1 − ρt is a Brownian motion. By
definition,

xn(M) (t) − ζn−1

 (M) 
= max ζn − ζn−1 + Bn (t), sup (xn−1 (s) − ζn−1 + Bn (t) − Bn (s)) ,
0≤s≤t
(2.3.32)
which allows us to apply Proposition 2.8, i.e., we have that

xn(M) (t) − ζn−1 − (ζn − ζn−1 ) − ρt = xn(M) (t) − ζn − ρt (2.3.33)

2.3 Infinite Particle Systems 21

is a Brownian motion. Since xn(MT ) (t) = xn (t) the proof is completed.

It is clear, that in the case λ = ρ the process (2.3.29) is a Brownian motion for
n > 0, too, i.e., the system is stationary in n. We also have stationarity in t, in
the sense that for each t ≥ 0 the random variables {xn (t) − xn−1 (t), n ∈ Z} are
independent and distributed exponentially with parameter ρ. The following result is
a small modification of Theorem 2 in O’Connell and Yor (2001).
Proposition 2.8 (Burke’s theorem for Brownian motions) Fix ρ > 0 and let B(t),
C(t) be standard Brownian motions, as well as ζ ∼ exp(ρ), independent. Define the
process
 
D(t) = max ζ + C(t), sup (B(s) + ρs + C(t) − C(s)) . (2.3.34)
0≤s≤t

Then
D(t) − ζ − ρt (2.3.35)

is distributed as a standard Brownian motion.

Proof Extend the processes B(t), C(t) to two-sided Brownian motions indexed by
R. Defining

q(t) = sup {B(t) − B(s) + C(t) − C(s) − ρ(t − s)} (2.3.36)

−∞<s≤t

and
d(t) = B(t) + q(0) − q(t), (2.3.37)

we can apply Theorem 2 (O’Connell and Yor 2001), i.e., d(t) is a Brownian motion.
Now,

d √ d  ρ 
q(0) = sup{−B(s) − C(s) + ρs} = sup{ 2B(s) − ρs} = sup B(s) − s , (2.3.38)
s≤0 s≥0 s≥0 2

so by Lemma 2.9 it has exponential distribution with parameter ρ. As it is independent

of the processes {B(t), C(t), t ≥ 0} we can write q(0) = ζ. Dividing the supremum
into s < 0 and s ≥ 0 we arrive at:

−d(t) = q(t) − B(t) − q(0)

 
= max C(t) − ρt, sup {−B(s) + C(t) − C(s) − ρ(t − s)} − ζ ,
0≤s≤t
(2.3.39)
which is (2.3.35) up to a sign flip of B(s).
Lemma 2.9 Fix ρ > 0 and let B(t) be a standard Brownian motion. Then

sup(B(s) − ρs) ∼ exp(2ρ). (2.3.40)

s≥0
22 2 One-Sided Reflected Brownian …

Proof This formula can be found in Borodin and Salminen (2012), Part II, Sect. 2,
Eq. (1.1.4). We provide here two different proofs of the claim.
(a) The random variable
sup (B(s) − ρs) (2.3.41)
0≤s≤t

is distributed as B̃(t), where B̃ is a Brownian motion starting at 0, reflected (upwards)

at zero and drift −ρ. This follows from the Skorokhod construction. Indeed,

d
B̃(t) = sup (B(t) − B(s) − ρ(t − s))
0≤s≤t
(2.3.42)
d
= sup (B(t) − B(t − s) − ρs) = sup (B(s) − ρs),
0≤s≤t 0≤s≤t

where we changed s into t − s and used the fact that B(t) − B(t − s) has the same
distribution of B(s). As t → ∞, this converges to the stationary distribution of this
process, which is the exponential distribution with parameter 2ρ.
(b) For x ≥ 0, define the stopping time τx = inf{s ≥ 0 | B(s) − sρ/2 ≥ x}. Then,

P sup(B(s) − ρs) ≥ x = P(τx < ∞). (2.3.43)

s≥0

1
The process s → Ms = eρB(s)− 2 ρ s is a martingale (the geometric Brownian motion),
2

with M0 = 1. Thus by applying the optional sampling theorem (apply it first to τx ∧T

and then take T → ∞) one obtains

1 = E(Mτx ) = eρx P(τx < ∞) + 0P(τx = ∞), (2.3.44)

from which P(τx < ∞) = e−ρx as claimed (just replace ρ by 2ρ).

References

A.V. Skorokhod, Stochastic equations for diffusions in a bounded region. Theory Probab. Appl. 6,
264–274 (1961)
R.F. Anderson, S. Orey, Small random perturbation of dynamical systems with reflecting boundary.
Nagoya Math. J. 60, 189–216 (1976)
N. O’Connell, M. Yor, Brownian analogues of Burke’s theorem. Stoch. Proc. Appl. 96, 285–304
(2001)
K. Johansson, Shape fluctuations and random matrices. Comm. Math. Phys. 209, 437–476 (2000)
K. Johansson, Discrete polynuclear growth and determinantal processes. Comm. Math. Phys. 242,
277–329 (2003)
T. Seppäläinen, B. Valkó, Bounds for scaling exponents for a 1+1 dimensional directed polymer in
a brownian environment. ALEA 7, 451–476 (2010)
References 23

A. Borodin, I. Corwin, P.L. Ferrari, B. Vető, Height fluctuations for the stationary KPZ equation.
Math. Phys. Anal. Geom. 18, 1–95 (2015)
P.W. Glynn, W. Whitt, Departures from many queues in series. Ann. Appl. Probab. 1, 546–572
(1991)
T. Seppäläinen, A scaling limit for queues in series. Ann. Appl. Probab. 7, 855–872 (1997)
F.J. Dyson, A Brownian-motion model for the eigenvalues of a random matrix. J. Math. Phys. 3,
1191–1198 (1962)
C.A. Tracy, H. Widom, Level-spacing distributions and the Airy kernel. Comm. Math. Phys. 159,
151–174 (1994)
J. Gravner, C.A. Tracy, H. Widom, Limit theorems for height fluctuations in a class of discrete space
and time growth models. J. Stat. Phys. 102, 1085–1132 (2001)
Y. Baryshnikov, GUEs and queues. Probab. Theory Relat. Fields 119, 256–274 (2001)
K. Johansson, E. Nordenstam, Eigenvalues of GUE minors. Electron. J. Probab. 11, 1342–1371
(2006)
J. Warren, Dyson’s Brownian motions, intertwining and interlacing. Electron. J. Probab. 12, 573–
590 (2007)
M. Adler, E. Nordenstam, P. van Moerbeke, Consecutive minors for Dyson’s Brownian motions.
Stoch. Proc. Appl. 124, 2023–2051 (2014)
M. Adler, E. Nordenstam, P. van Moerbeke, The Dyson Brownian minor process. Annales de
l’Institut Fourier 64, 971–1009 (2014)
P.L. Ferrari, R. Frings, On the partial connection between random matrices and interacting particle
systems. J. Stat. Phys. 141, 613–637 (2010)
M. Adler, P. van Moerbeke, PDE’s for the joint distribution of the Dyson, Airy and Sine processes.
Ann. Probab. 33, 1326–1361 (2005)
F. Bornemann, P.L. Ferrari, M. Prähofer, The Airy1 process is not the limit of the largest eigenvalue
in GOE matrix diffusion. J. Stat. Phys. 133, 405–415 (2008)
J. Baik, P.L. Ferrari, S. Péché, Limit process of stationary TASEP near the characteristic line. Comm.
Pure Appl. Math. 63, 1017–1070 (2010)
P.L. Ferrari, H. Spohn, Scaling limit for the space-time covariance of the stationary totally asym-
metric simple exclusion process. Comm. Math. Phys. 265, 1–44 (2006)
T. Imamura, T. Sasamoto, Fluctuations of the one-dimensional polynuclear growth model with
external sources. Nucl. Phys. B 699, 503–544 (2004)
A.N. Borodin, P. Salminen, Handbook of Brownian Motion - Facts and Formulae (Birkhäuser,
Switzerland) (2012)
Another random document with
no related content on Scribd:
identify now Hamilton of Bangour, young David Malloch, a William
Crawford, and a William Walkinshaw,—contained about sixty songs
of Ramsay’s own composition. Similarly, among several mock-
antiques by modern hands inserted into The Evergreen, were two
by Ramsay himself, entitled The Vision and The Eagle and Robin
Redbreast.
The time had come for Ramsay’s finest and most characteristic
performance. More than once, in his miscellanies hitherto, he had
tried the pastoral form in Scotch, whether from a natural tendency to
that form or induced by recent attempts in the English pastoral by
Ambrose Philips, Pope, and Gay. Besides his pastoral elegy on the
death of Addison, and another on the death of Prior, he had written a
pastoral dialogue of real Scottish life in 162 lines, entitled Patie and
Roger, introduced by this description:—
“Beneath the south side of a craigy bield,
Where a clear spring did halesome water yield,
Twa youthfu’ shepherds on the gowans lay,
Tending their flocks ae bonny morn of May:
Poor Roger graned till hollow echoes rang,
While merry Patie hummed himsel a sang.”

This piece, and two smaller pastoral pieces in the same vein, called
Patie and Peggy and Jenny and Meggie, had been so much liked
that Ramsay had been urged by his friends to do something more
extensive in the shape of a pastoral story or drama. He had been
meditating such a thing through the year 1724, while busy with his
two editorial compilations; and in June 1725 the result was given to
the public in The Gentle Shepherd: A Scots Pastoral Comedy.
Here the three pastoral sketches already written were inwoven into a
simply-constructed drama of rustic Scottish life as it might be
imagined among the Pentland Hills, near Edinburgh, at that time,
still within the recollection of very old people then alive, when the
Protectorates of Cromwell and his son had come to an end and Monk
had restored King Charles. The poem was received with enthusiastic
admiration. There had been nothing like it before in Scottish
literature, or in any other; nothing so good of any kind that could be
voted as even similar; and this was at once the critical verdict. It is a
long while ago, and there are many spots in Edinburgh which
compete with one another in the interest of their literary
associations; but one can stand now with particular pleasure for a
few minutes any afternoon opposite that decayed house in the High
Street, visible as one is crossing from the South Bridge to the North
Bridge, where Allan Ramsay once had his shop, and whence the first
copies of The Gentle Shepherd were handed out, some day in June
1725, to eager Edinburgh purchasers.
The tenancy of this house by Ramsay lasted but a year longer.
He had resolved to add to his general business of bookselling and
publishing that of a circulating library, the first institution of the
kind in Edinburgh. For this purpose he had taken new premises, still
in the High Street, but in a position even more central and
conspicuous than that of “The Mercury” opposite Niddry’s Wynd.
They were, in fact, in the easternmost house of the Luckenbooths, or
lower part of that obstructive stack of buildings, already mentioned,
which once ran up the High Street alongside of St. Giles’s Church,
dividing the traffic into two narrow and overcrowded channels. It is
many years since the Luckenbooths and the whole obstruction of
which they formed a part were swept away; but from old prints we
can see that the last house of the Luckenbooths to the east was a tall
tenement of five storeys, with its main face looking straight down the
lower slope of the High Street towards the Canongate. The strange
thing was that, though thus in the very heart of the bustle of the town
as congregated round the Cross, the house commanded from its
higher windows a view beyond the town altogether, away to Aberlady
Bay and the farther reaches of sea and land in that direction. It was
into this house that Ramsay removed in 1726, when he was exactly
forty years of age. The part occupied by him was the flat immediately
above the basement floor, but perhaps with that floor in addition.
The sign he adopted for the new premises was one exhibiting the
heads or effigies of Ben Jonson and Drummond of Hawthornden.
Having introduced Ramsay into this, the last of his Edinburgh
shops, we have reached the point where our present interest in him
all but ends. In 1728, when he had been two years in the new
premises, he published a second volume of his collected poems,
under the title of Poems by Allan Ramsay, Volume II., in a
handsome quarto matching the previous volume of 1721, and
containing all the pieces he had written since the appearance of that
volume; and in 1730 he published A Collection of Thirty Fables.
These were his last substantive publications, and with them his
literary career may be said to have come to a close. Begun in the last
years of the reign of Queen Anne, and continued through the whole
of the reign of George I., it had just touched the beginning of that of
George II., when it suddenly ceased. Twice or thrice afterwards at
long intervals he did scribble a copy of verses; but, in the main, from
his forty-fifth year onwards, he rested on his laurels. Thenceforward
he contented himself with his bookselling, the management of his
circulating library, and the superintendence of the numerous
editions of his Collected Poems, his Gentle Shepherd, and his Tea
Table Miscellany that were required by the public demand, and the
proceeds of which formed a good part of his income. It would be a
great mistake, however, to suppose that, when Allan Ramsay’s time
of literary production ended, the story of his life in Edinburgh also
came to a close, or ceased to be important. For eight-and-twenty
years longer, or almost till George II. gave place to George III.,
Ramsay continued to be a living celebrity in the Scottish capital,
known by figure and physiognomy to all his fellow-citizens, and
Ramsay’s bookshop at the end of the Luckenbooths, just above the
Cross, continued to be one of the chief resorts of the well-to-do
residents, and of chance visitors of distinction. Now and then,
indeed, through the twenty-eight years, there are glimpses of him
still in special connections with the literary, as well as with the social,
history of Edinburgh. When the English poet Gay, a summer or two
before his death in 1732, came to Edinburgh on a visit, in the
company of his noble patrons, the Duke and Duchess of
Queensberry, and resided with them in their mansion of
Queensberry House in the Canongate,—now the gloomiest and
ugliest-looking house in that quarter of the old town, but then
reckoned of palatial grandeur,—whither did he tend daily, in his
saunterings up the Canongate, but to Allan Ramsay’s shop? One
hears of him as standing there with Allan at the window to have the
city notabilities and oddities pointed out to him in the piazza below,
or as taking lessons from Allan in the Scottish words and idioms of
the Gentle Shepherd, that he might explain them better to Mr. Pope
when he went back to London.
Some years later, when Ramsay had reached the age of fifty, and
he and his wife were enjoying the comforts of his ample success, and
rejoicing in the hopes and prospects of their children,—three
daughters, “no ae wally-draggle among them, all fine girls,” as
Ramsay informs us, and one son, a young man of three-and-twenty,
completing his education in Italy for the profession of a painter,—
there came upon the family what threatened to be a ruinous disaster.
Never formally an anti-Presbyterian, and indeed regularly to be seen
on Sundays in his pew in St. Giles’s High Kirk, but always and
systematically opposed to the unnecessary social rigours of the old
Presbyterian system, and of late under a good deal of censure from
clerical and other strict critics on account of the dangerous nature of
much of the literature put in circulation from his library, Ramsay
had ventured at last on a new commercial enterprise, which could
not but be offensive on similar grounds to many worthy people,
though it seems to have been acceptable enough to the Edinburgh
community generally. Edinburgh having been hitherto deficient in
theatrical accommodation, and but fitfully supplied with dramatic
entertainments, he had, in 1736, started a new theatre in Carrubber’s
Close, near to his former High Street shop. He was looking for great
profits from the proprietorship of this theatre and his partnership in
its management. Hardly had he begun operations, however, when
there came the extraordinary statute of 10 George II. (1737),
regulating theatres for the future all over Great Britain. As by this
statute there could be no performance of stage-plays out of London
and Westminster, save when the King chanced to be residing in some
other town, Ramsay’s speculation collapsed, and all the money he
had invested in it was lost. It was a heavy blow; and he was moved by
it to some verses of complaint to his friend Lord President Forbes
and the other judges of the Court of Session. While telling the story
of his own hardship in the case, he suggests that an indignity had
been done by the new Act to the capital of Scotland:—
“Shall London have its houses twa
And we doomed to have nane ava’?
Is our metropolis, ance the place
Where langsyne dwelt the royal race
Of Fergus, this gate dwindled down
To a level with ilk clachan town,
While thus she suffers the subversion
Of her maist rational diversion?”
However severe the loss to Ramsay at the time, it was soon tided
over. Within six years he is found again quite at ease in his worldly
fortunes. His son, for some years back from Italy, was in rapidly
rising repute as a portrait-painter, alternating between London and
Edinburgh in the practice of his profession, and a man of mark in
Edinburgh society on his own account; and, whether by a junction of
the son’s means with the father’s, or by the father’s means alone, it
was now that there reared itself in Edinburgh the edifice which at the
present day most distinctly preserves for the inhabitants the memory
of the Ramsay family in their Edinburgh connections. The
probability is that, since Allan had entered on his business premises
at the end of the Luckenbooths, his dwelling-house had been
somewhere else in the town or suburbs; but in 1743 he built himself a
new dwelling-house on the very choicest site that the venerable old
town afforded. It was that quaint octagon-shaped villa, with an
attached slope of green and pleasure-ground, on the north side of the
Castle Hill, which, as well from its form as from its situation, attracts
the eye as one walks along Princes Street, and which still retains the
name of Ramsay Lodge. The wags of the day, making fun of its
quaint shape, likened the construction to a goose-pie; and something
of that fancied resemblance may be traced even now in its extended
and improved proportions. But envy may have had a good deal to do
with the comparison. It is still a neat and comfortable dwelling
internally, while it commands from its elevation an extent of scenery
unsurpassed anywhere in Europe. The view from it ranges from the
sea-mouth of the Firth of Forth on the east to the first glimpses of the
Stirlingshire Highlands on the west, and again due north across the
levels of the New Town, and the flashing waters of the Firth below
them, to the bounding outline of the Fifeshire hills. When, in 1743,
before there was as yet any New Town at all, Allan Ramsay took up
his abode in this villa, he must have been considered a fortunate and
happy man. His entry into it was saddened, indeed, by the death of
his wife, which occurred just about that time; but for fourteen years
of widowerhood, with two of his daughters for his companions, he
lived in it serenely and hospitably. During the first nine years of
those fourteen he still went daily to his shop in the Luckenbooths,
attending to his various occupations, and especially to his circulating
library, which is said to have contained by this time about 30,000
volumes; but for the last five or six years he had entirely relinquished
business. There are authentic accounts of his habits and demeanour
in his last days, and they concur in representing him as one of the
most charming old gentlemen possible, vivacious and sprightly in
conversation, full of benevolence and good humour, and especially
fond of children and kindly in his ways for their amusement. He died
on the 7th of January 1758, in the seventy-second year of his age, and
was buried in Greyfriars Churchyard.
Ramsay had outlived nearly all the literary celebrities who had
been his contemporaries during his own career of active authorship,
ended nearly thirty years before. Swift and Pope were gone, after
Gay, Steele, Arbuthnot, and others of the London band, who had
died earlier. Of several Scotsmen, his juniors, who had stepped into
the career of literature after he had shown the way, and had attained
to more or less of poetic eminence under his own observation, three,
—Robert Blair, James Thomson, and Hamilton of Bangour,—had
predeceased him. Their finished lives, with all the great radiance of
Thomson’s, are wholly included in the life of Allan Ramsay. David
Malloch, who had been an Edinburgh protégé of Ramsay’s, but had
gone to London and Anglicised himself into “Mallet,” was about the
oldest of his literary survivors into another generation; but in that
generation, as Scotsmen of various ages, from sixty downwards to
one-and-twenty, living, within Scotland or out of it, at the date of
Ramsay’s death, we count Lord Kames, Armstrong, Reid, Hume,
Lord Monboddo, Hugh Blair, George Campbell, Smollett, Wilkie,
Blacklock, Robertson, John Home, Adam Smith, Adam Ferguson,
Lord Hailes, Falconer, Meikle, and Beattie. Such of these as were
residents in Edinburgh had known Allan Ramsay personally; others
of them had felt his influence indirectly; and all must have noted his
death as an event of some consequence.
The time is long past for any exaggeration of Allan Ramsay’s
merits. But, call him only a slipshod little Horace of Auld Reekie,
who wrote odes, epistles, satires, and other miscellanies in Scotch
through twenty years of the earlier part of the eighteenth century,
and was also, by a happy chance, the author of a unique and
delightful Scottish pastoral, it remains true that he was the most
considerable personality in Scottish literary history in order of time
after Drummond of Hawthornden, or, if we think only of the
vernacular, after Sir David Lindsay, and that he did more than any
other man to stir afresh a popular enthusiasm for literature in
Scotland after the Union with England. All in all, therefore, it is with
no small interest that, in one’s walks along the most classic
thoroughfare of the present Edinburgh, one gazes at the white stone
statue of Allan Ramsay, from the chisel of Sir John Steell, which
stands in the Gardens just below the famous “goose-pie villa.” It
looks as if the poet had just stepped down thence in his evening
habiliments to see things thereabouts in their strangely changed
condition. By the tact of the sculptor, he wears, one observes, not a
wig, but the true poetic night-cap or turban.
LADY WARDLAW AND THE BARONESS NAIRNE[6]

In 1719 there was published in Edinburgh, in a tract of twelve

folio pages, a small poem, 27 stanzas or 216 lines long, entitled
Hardyknute, a Fragment. It was printed in old spelling, to look like
a piece of old Scottish poetry that had somehow been recovered; and
it seems to have been accepted as such by those into whose hands the
copy had come, and who were concerned in having it published.
Among these were Duncan Forbes of Culloden, afterwards Lord
President of the Court of Session, and Sir Gilbert Elliott of Minto,
afterwards Lord Justice-Clerk; but there is something like proof that
it had come into their hands indirectly from Sir John Hope Bruce of
Kinross, baronet, who died as late as 1766 at a great age, in the rank
of lieutenant-general, and who, some time before 1719, had sent a
manuscript copy of it to Lord Binning, with a fantastic story to the
effect that the original, in a much defaced vellum, had been found, a
few weeks before, in a vault at Dunfermline.
The little thing, having become popular in its first published
form, was reproduced in 1724 by Allan Ramsay in his Evergreen,
which professed to be “a collection of Scots Poems wrote by the
ingenious before 1600”; but it there appeared with corrections and
some additional stanzas. In 1740 it had the honour of a new
appearance in London, under anonymous editorship, and with the
title “Hardyknute, a Fragment; being the first Canto of an Epick
Poem: with general remarks and notes.” The anonymous editor, still
treating it as a genuine old poem, of not later than the sixteenth
century, praises it very highly. “There is a grandeur, a majesty of
sentiment,” he says, “diffused through the whole: a true sublime,
which nothing can surpass.” It was but natural that a piece of which
this could be said should be included by Percy in his Reliques of
Ancient English Poetry, published in 1765. It appeared, accordingly,
in the first edition of that famous book, still as an old poem and in
antique spelling; and it was reprinted in the subsequent editions
issued by Percy himself in 1767, 1775, and 1794, though then with
some added explanations and queries.
It was through Percy’s collection that the poem first became
generally known and popular. Even there, though in very rich
company, it was singled out by competent critics for special
admiration. But, indeed, good judges, who had known it in its earlier
forms, had already made it a favourite. The poet Gray admired it
much; and Thomas Warton spoke of it as “a noble poem,” and
introduced an enthusiastic reference to it into one of his odes. Above
all, it is celebrated now as having fired the boyish genius of Sir
Walter Scott. “I was taught Hardyknute by heart before I could read
the ballad myself,” he tells us, informing us further that the book out
of which he was taught the ballad was Allan Ramsay’s Evergreen of
1724, and adding, “It was the first poem I ever learnt, the last I shall
ever forget.” In another place he tells us more particularly that it was
taught him out of the book by one of his aunts during that visit to his
grandfather’s farmhouse of Sandyknowe in Roxburghshire on which
he had been sent when only in his third year for country air and
exercise on account of his delicate health and lameness, and which
he remembered always as the source of his earliest impressions and
the time of his first consciousness of existence. He was accustomed
to go about the farmhouse shouting out the verses of the ballad
incessantly, so that the Rev. Dr. Duncan, the minister of the parish,
in his calls for a sober chat with the elder inmates, would complain of
the interruption and say, “One may as well speak in the mouth of a
cannon as where that child is.” Hardyknute, we may then say, was
the first thing in literature that took hold of the soul and imagination
of Scott; and who knows how far it may have helped to determine the
cast and direction of his own genius through all the future?
Afterwards, through his life in Edinburgh, Ashestiel, and Abbotsford,
he was never tired of repeating snatches of the strong old thing he
had learnt at Sandyknowe; and the very year before his death (1831)
we find him, when abroad at Malta in the vain hope of recruiting his
shattered frame, lamenting greatly, in a conversation about ballad-
poetry, that he had not been able to persuade his friend Mr. John
Hookham Frere to think so highly of the merits of Hardyknute as he
did himself.
 What is the piece of verse so celebrated? It must be familiar to
many; but we may look at it again. We shall take it in its later or
more complete form, as consisting of 42 stanzas or 336 lines; in
which form, though it is still only a fragment, the conception or story
is somewhat more complex, more filled out, than in the first
published form of 1719. The fragment opens thus:—
“Stately stept he east the wa’,
And stately stept he west;
Full seventy years he now had seen,
With scarce seven years of rest.
He lived when Britons’ breach of faith
Wrocht Scotland mickle wae;
And aye his sword tauld, to their cost,
He was their deadly fae.

High on a hill his castle stood,

With halls and towers a-hicht,
And guidly chambers fair to see,
Whare he lodged mony a knicht.
His dame, sae peerless ance and fair,
For chaste and beauty deemed,
Nae marrow had in a’ the land,
Save Eleanour the Queen.

Full thirteen sons to him she bare,

All men of valour stout;
In bluidy fecht with sword in hand
Nine lost their lives bot doubt:
Four yet remain; lang may they live
To stand by liege and land!
High was their fame, high was their micht,
And high was their command.

Great love they bare to Fairly fair,

Their sister saft and dear:
Her girdle shawed her middle jimp,
And gowden glist her hair.
What waefu’ wae her beauty bred,
Waefu’ to young and auld;
Waefu’, I trow, to kith and kin,
As story ever tauld!”
Here we see the old hero Hardyknute in peace in the midst of his
family, his fighting days supposed to be over, and his high castle on
the hill, where he and his lady dwell, with their four surviving sons
and their one daughter, Fairly Fair, one of the lordly boasts of a
smiling country. But suddenly there is an invasion. The King of
Norse, puffed up with power and might, lands in fair Scotland; and
the King of Scotland, hearing the tidings as he sits with his chiefs,
“drinking the blude-red wine,” sends out summonses in haste for all
his warriors to join him. Hardyknute receives a special message.
“Then red, red grew his dark-brown cheeks;
Sae did his dark-brown brow;
His looks grew keen, as they were wont
In dangers great to do.”

Old as he is, he will set out at once, taking his three eldest sons with
him, Robin, Thomas, and Malcolm, and telling his lady in his
farewell to her:—
“My youngest son sall here remain
To guaird these stately towers,
And shoot the silver bolt that keeps
Sae fast your painted bowers.”

And so we take leave of the high castle on the hill, with the lady,
her youngest son, and Fairly fair, in it, and follow the old lord and his
other three sons over the moors and through the glens as they ride to
the rendezvous. On their way they encounter a wounded knight,
lying on the ground and making a heavy moan:—
“‘Here maun I lie, here maun I die,
By treachery’s false guiles;
Witless I was that e’er gave faith
To wicked woman’s smiles.’”

Hardyknute, stopping, comforts him; says that, if he can but mount

his steed and manage to get to his castle on the hill, he will be tended
there by his lady and Fairly fair herself; and offers to detach some of
his men with him for convoy.
“With smileless look and visage wan
The wounded knicht replied:
‘Kind chieftain, your intent pursue,
 For here I maun abide.

‘To me nae after day nor nicht

Can e’er be sweet or fair;
But soon, beneath some drapping tree,
Cauld death sail end my care.’”

Farther pleading by Hardyknute avails nothing; and, as time presses,

he has to depart, leaving the wounded knight, so far as we can see, on
the ground as he had found him, still making his moan. Then, after
farther riding over a great region, called vaguely Lord Chattan’s land,
we have the arrival of Hardyknute and his three sons in the King of
Scotland’s camp, minstrels marching before them playing pibrochs.
Hardly have they been welcomed when the battle with the Norse
King and his host is begun. It is described at considerable length, and
with much power, though confusedly, so that one hardly knows who
is speaking or who is wounded amid the whirr of arrows, the
shouting, and the clash of armour. One sees, however, Hardyknute
and two of his sons fighting grandly in the pell-mell. At last it is all
over, and we know that the Norse King and his host have been
routed, and that Scotland has been saved.
“In thraws of death, with wallert cheek,
All panting on the plain,
The fainting corps of warriors lay,
Ne’er to arise again:
Ne’er to return to native land;
Nae mair wi’ blythesome sounds
To boist the glories of the day
And shaw their shinand wounds.

On Norway’s coast the widowed dame

May wash the rock with tears,
May lang look ower the shipless seas,
Before her mate appears.
Cease, Emma, cease to hope in vain:
Thy lord lies in the clay;
The valiant Scats nae reivers thole
To carry life away.

There, on a lea where stands a cross

Set up for monument,
Thousands full fierce, that summer’s day,
 Filled keen war’s black intent.
Let Scots, while Scots, praise Hardyknute;
Let Norse the name aye dread;
Aye how he foucht, aft how he spared,
Sall latest ages read.”

Here the story might seem to end, and here perhaps it was intended
at first that it should end; but in the completer copies there are three
more stanzas, taking us back to Hardyknute’s castle on the high hill.
We are to fancy Hardyknute and his sons returning joyfully thither
after the great victory:—
“Loud and chill blew the westlin wind,
Sair beat the heavy shower;
Mirk grew the nicht ere Hardyknute
Wan near his stately tower:
His tower, that used with torches’ bleeze
To shine sae far at nicht,
Seemed now as black as mourning weed:
Nae marvel sair he sich’d.

‘There’s nae licht in my lady’s bower;

There’s nae licht in my hall;
Nae blink shines round my Fairly fair,
Nor ward stands on my wall.
What bodes it? Robert, Thomas, say!’
Nae answer fits their dread.
‘Stand back, my sons! I’ll be your guide!’
But by they passed wi’ speed.

‘As fast I have sped ower Scotland’s faes.’

There ceased his brag of weir,
Sair shamed to mind oucht but his dame
And maiden Fairly fair.
Black fear he felt, but what to fear
He wist not yet with dread:
Sair shook his body, sair his limbs;
And all the warrior fled.”

And so the fragment really ends, making us aware of some dreadful

catastrophe, though what it is we know not. Something ghastly has
happened in the castle during Hardyknute’s absence, but it is left
untold. Only, by a kind of necessity of the imagination, we connect it
somehow with that wounded knight whom Hardyknute had met
lying on the ground as he was hurrying to the war, and whom he had
left making his moan. Was he a fiend, or what?
It is quite useless to call this a historical ballad. There was a
reference, perhaps, in the author’s mind, to the battle of Largs in
Ayrshire, fought by the Scots in 1263, in the reign of Alexander III.,
against the invading King Haco of Norway; and there is a Fairly
Castle on a hill near Largs which may have yielded a suggestion and a
name. But, in truth, any old Scottish reign, and any Norse invasion,
will do for time and basis, and the ballad is essentially of the
romantic kind, a story snatched from an ideal antique, and appealing
to the pure poetic imagination. A battle is flung in; but what rivets
our interest is the hero Hardyknute, a Scottish warrior with a Danish
name, and that stately castle of his, somewhere on the top of a hill, in
which he dwelt so splendidly with his lady, his four sons, and their
sister Fairly fair, till he was called once more to war, and in which
there was some ghastly desolation before his return. Such as it is, we
shall all agree, I think, with Gray, Warton, Scott, and the rest of the
best critics, in admiring the fragment. It has that something in it
which we call genius.
It seems strange now that any critic could ever have taken the
ballad for a really old one, to be dated from the sixteenth century or
earlier. Apart from the trick of old spelling, and affectation of the
antique in a word or two, the phraseology, the manner, the cadence,
the style of the Scotch employed, are all of about the date of the first
publication of the ballad, the first quarter of the eighteenth century.
The phrase “Let Scots, while Scots, praise Hardyknute,” and the
phrase “And all the warrior fled,” are decisive; and, while there might
be room for the supposition that some old legend suggested the
subject to the author, the general cast of the whole forbids the idea
that it is merely a version of some transmitted original.
Suspicions, indeed, of the modern authorship of Hardyknute
had arisen in various quarters long before any one person in
particular was publicly named as the author. That was first done by
Percy in 1767, in the second edition of his Reliques, when he gave his
reasons for thinking, from information transmitted to him from
Scotland by Sir David Dalrymple, Lord Hailes, that the ballad was
substantially the composition of a Scottish lady, who had died in
1727, eight years after it had first appeared in its less perfect form,
and three years after it had appeared with the improvements and the
additional stanzas. That lady was Elizabeth Halket, born in 1677, one
of the daughters of Sir Charles Halket of Pitfirran in Fifeshire,
baronet, but who had changed her name to Wardlaw in the year
1696, when she became the wife of Sir Henry Wardlaw of Pitreavie,
also a Fifeshire baronet. All subsequent evidence has confirmed the
belief that this Lady Wardlaw was the real author of Hardyknute,
though, to mystify people, it was first given out by her relatives as an
ancient fragment. This was the statement more especially of the
already-mentioned Sir John Hope Bruce of Kinross, who was one of
her brothers-in-law.
Of Lady Wardlaw herself we hear nothing more distinct than
that she was “a woman of elegant accomplishments, who wrote other
poems, and practised drawing and cutting paper with her scissors,
and who had much wit and humour, with great sweetness of
temper.” So we must be content to imagine her,—a bright-minded
and graceful lady, living in Fifeshire, or coming and going between
Fifeshire and Edinburgh, nearly two centuries ago, and who, while
attending to her family duties and the duties of her station, could
cherish in secret a poetic vein peculiarly her own, and produce at
least one fine ballad of an ideal Scottish antique. This in itself would
be much. For that was the age of Queen Anne and of the first of the
Georges, when poetry of an ideal or romantic kind was perhaps at its
lowest ebb throughout the British Islands, and the poetry most in
repute was that of the modern school of artificial wit and polish
represented by Addison and Pope.
But this is not all. In the year 1859 the late Mr. Robert Chambers
published a very ingenious and interesting essay entitled “The
Romantic Scottish Ballads: their Epoch and Authorship.” The ballads
to which he invited critical attention were the particular group which
includes Sir Patrick Spens, Gil Morrice, Edward Edward, The Jew’s
Daughter, Gilderoy, Young Waters, Edom o’ Gordon, Johnnie of
Braidislee, Mary Hamilton, The Gay Goss Hawk, Fause Foodrage,
The Lass of Lochryan, Young Huntin, The Douglas Tragedy, Clerk
Saunders, Sweet William’s Ghost, and several others. With but one
or two exceptions, these were first given to the world either in Percy’s
Reliques in 1765, or in the subsequent collections of Herd (1769),
Scott (1802), and Jamieson (1806); but, since they were published,
they have been favourites with all lovers of true poetry,—the “grand
ballad” of Sir Patrick Spens, as Coleridge called it, ranking perhaps
highest, on the whole, in general opinion. There is a certain common
character in all the ballads of the group, a character of genuine
ideality, of unconnectedness or but hazy connectedness with
particular time or place, of a tendency to the weirdly, and also of a
high-bred elegance and lightsome tact of expression, distinguishing
them from the properly historical Scottish Ballads, such as the Battle
of Otterbourne, or the Border Ballads proper, such as Kinmont
Willie, or the homely rustic ballads of local or family incident of
which so many have been collected. Hence the distinctive name of
“romantic,” usually applied to them.
Respecting these ballads the common theory was, and still is,
that they are very old indeed,—that they are the transmitted oral
versions of ballads that were in circulation among the Scottish
people before the Reformation. This theory Mr. Chambers
challenged, and by a great variety of arguments. Not only was it very
suspicious, he said, that there were no ancient manuscripts of them,
and that, save in one or two cases, they had never been heard of till
the eighteenth century; but the internal evidence, of conception,
sentiment, costume, and phraseology,—not in lines and passages
merely, where change from an original might be supposed, but
through and through, and back to the very core of any supposed
original,—all pointed, he maintained, to a date of composition not
farther back than the beginning of the century in which they first
came into print. He maintained farther that they all reveal the hand
of some person of superior breeding and refinement, with a
cultivated literary expertness and sense of the exquisite, and that,
just as the difference of age would be seen if one of them were placed
side by side with an authentic piece of old Scottish poetry of the
sixteenth century, so would this other difference of refined or
cultured execution be at once seen if one of them were placed side by
side with a genuine popular ballad of lowly origin, such as used to
please in sheets on street-stalls and in pedlars’ chap-books. Farther
still, in all or most of the ballads concerned, there are, he argued,
traces of feminine perception and feeling. And so, still pressing the
question, and noting the recurrence of phrases and ideas from ballad
to ballad of the group, not to be found in other ballads, but looking
like the acquired devices of one and the same writer’s fancy,—some
of the most remarkable of which recurring ideas and phrases he
chased up to the ballad of Hardyknute,—he arrived at the conclusion
that there was a “great likelihood” that all or most of the ballads he
was considering were either absolutely the inventions of Lady
Wardlaw of Pitreavie, or such complete recasts by her of traditional
fragments that she might be called the real author. He would not
advance the conclusion as more than a “great likelihood,” and he
allowed that it might be still controverted; but he cited in its favour
the fact that so high an authority as Mr. David Laing had previously
intimated his impression that Hardyknute and Sir Patrick Spens
were by the same hand.
Were Mr. Chambers’s conclusion to be verified, it would be a
sore wrench to the patriotic prejudices of many to have to abandon
the long-cherished fancy of the immemorial, or at least remote,
antiquity of so many fine Scottish favourites. But what a
compensation! For then that Lady Wardlaw whom we can already
station, for her Hardyknute alone, as undoubtedly one woman of
genius in the poverty-stricken Scotland of the beginning of the
eighteenth century, would shine out with greatly increased radiance
as the author of a whole cycle of the finest ballad-pieces in our
language, a figure of very high importance in Scottish literary
history, a precursor or sister of Burns and of Scott. For my own part,
I would willingly submit to the wrench for a compensation so
splendid. I am bound to report, however, that Mr. Chambers’s
speculation of 1859 was controverted strenuously at the time, has
been pronounced a heresy, and does not seem to have been
anywhere generally accepted. It was controverted especially, within a
year from its appearance, in a pamphlet of reply by Mr. Norval Clyne
of Aberdeen, entitled “The Scottish Romantic Ballads and the Lady
Wardlaw Heresy”; and I observe that Professor Child of America, in
his great Collection of English and Scottish Ballads, pays no respect
to it, treats it as exploded by Mr. Clyne’s reply, and expressly
dissociates Sir Patrick Spens and other ballads of the class from
Hardyknute. It may be enough, in these circumstances, merely to
intimate my opinion that the controversy is by no means closed.
There were shrewder and deeper suggestions, I think, in Mr.
Chambers’s paper of 1859 than Mr. Clyne was able to obviate; and,
having observed that most of the lore on the subject used by Mr.
Clyne in his reply, his adverse references and quotations included,
was derived from Mr. Chambers’s own Introduction and Notes to his
three-volume edition of Scottish Songs and Ballads in 1829, I cannot
but presume that Mr. Chambers had all that lore sufficiently in his
mind thirty years afterwards, and found nothing in it to impede or
disconcert him then in his new speculation. Apart, however, from the
special question of Lady Wardlaw’s concern in the matter, Mr.
Chambers seems to me to have moved a very proper and necessary
inquiry when he started his theory of the comparatively recent origin
of all or most of the Scottish Romantic Ballads. In what conception,
what kind of language, does the opposite theory couch itself? In the
conception that, besides the series of those literary products of past
Scottish generations, the work of learned or professional writers,
from the time of Barbour onwards, that have come down to us in
books, or in old manuscript collections like that of Bannatyne, there
was always a distinct literature of more lowly origin, consisting of
ballads and songs recited or sung in Scottish households in various
districts, and orally transmitted from age to age with no names
attached to them, and indeed requiring none, inasmuch as they were
nobody’s property in particular, but had “sprung from the heart of
the people.” Now, this phrase, “sprung from the heart of the people,”
I submit, is, if not nonsensical, at least hazy and misleading. Nothing
of fine literary quality ever came into existence, in any time or place,
except as the product of some individual person of genius and of
somewhat more than average culture. Instead of saying that such
things “spring from the heart of the people,” one ought rather to say
therefore that they “spring to the heart of the people.” They live after
their authors are forgotten, are repeated with local modifications,
and so become common property. It is, of course, not denied that
this process must have been at work in Scotland through many
centuries before the eighteenth. The proof exists in scraps of fine old
Scottish song still preserved, the earliest perhaps the famous verse
on the death of Alexander III., and in lists, such as that in The
Complaint of Scotland, of the titles of clusters of old Scottish songs
and tales that were popular throughout the country in the sixteenth
century, but have perished since. The very contention of Mr.
Chambers respecting Sir Patrick Spens and the other ballads in
question was that the fact that there is no mention of them in those
old lists is itself significant, and that they have a set of special
characteristics which came into fashion only with themselves.
 If Lady Wardlaw was the author of those ballads, or of some of
them, we have lost much by her secretiveness. We have been put in a
perplexity where perplexity there ought to have been none. The
cause, on her part, was perhaps less a desire for mystification than
an amiable shrinking from publicity, dislike of being talked of as a
literary lady. This was a feeling which the ungenerous mankind of
the last century,—husbands, brothers, uncles, and brothers-in-law,—
thought it proper to foster in any feminine person of whose literary
accomplishments they were privately proud. It affected the careers of
not a few later Scottish women of genius in the same century, and
even through part of our own. Passing over several such, and among
them Lady Anne Barnard, the authoress of Auld Robin Gray, let me
come to an instance so recent that it can be touched by the memories
of many that are still living.
In the year 1766, seven years after the birth of Burns, and five
before that of Scott, there was born, in the old house of Gask in
Strathearn, Perthshire, a certain Carolina Oliphant, the third child of
Laurence Oliphant the younger, who, by the death of his father the
next year, became the Laird of Gask and the representative of the old
family of the Oliphants.
They were a Jacobite family to the core. The Laird and his father
had been out in the Rebellion of 1745; they had suffered much in
consequence and been long in exile; and not till a year or two before
the birth of this little girl had they been permitted to return and
settle on their shattered estates. They were true to their Jacobitism
even then, acknowledging no King but the one “over the water,”
praying for him, corresponding with him, and keeping up the
recollection of him in their household as almost a religion. Carolina
was named Carolina because, had she been a boy, she was to have
been named Charles, and she used to say that her parents had never
forgiven her for having been born a girl. But two boys were born at
last, and there were sisters both older and younger; and so, among
Oliphants, and Robertsons of Struan, and Murrays, and other
relatives, all Jacobite, and all of the Scottish Episcopal persuasion,
Carolina grew up in the old house of Gask, hearing Jacobite stories
and Highland legends from her infancy, and educated with some
care. The mother having died when this, her third, child was but
eight years of age, the Laird was left with six young ones. “A poor
valetudinary person,” as he describes himself, he seems, however, to
have been a man of fine character and accomplishments, and to have
taken great pains with his children. King George III., hearing
somehow of his unswerving Jacobitism and the whimsicalities in
which it showed itself, is said to have sent him this message by the
member for Perthshire: “Give my compliments, not the compliments
of the King of England, but those of the Elector of Hanover, to Mr.
Oliphant, and tell him how much I respect him for the steadiness of
his principles.”
Somewhat stately and melancholic himself, and keeping up the
ceremonious distance between him and his children then thought
proper, the Laird of Gask had those liberal and anti-morose views of
education which belonged especially to Scottish nonjuring or
Episcopalian families. A wide range of reading was permitted to the
boys and the girls; dancing, especially reel-dancing, was incessant
among them,—at home, in the houses of neighbouring lairds, or at
county-balls; in music, especially in Scottish song, they were all
expert, so that the rumour of a coming visit of Neil Gow and his
violin to Strathearn, with the prospect it brought them of a week
extraordinary of combined music and reel-dancing, would set them
all madly astir; but the most musical of the family by far was
Carolina. She lived in music, in mirth, legend, Highland scenery, and
the dance, a beautiful girl to boot, and called “the Flower of
Strathearn,” of tall and graceful mien, with fine eyes, and fine
sensitive features, slightly proud and aquiline. And so to 1792, when
her father, the valetudinary laird, died, some of his children already
out in the world, but this one, at the age of twenty-six, still
unmarried.
For fourteen years more we hear of her as still living in the old
house of Gask with her brother Laurence, the new Laird, and with
the wife he brought into it in 1795,—the even tenor of her existence
broken only by some such incident as a visit to the north of England.
During this time it is that we become aware also of the beginnings in
her mind of a deep new seriousness, a pious devoutness, which,
without interfering with her passionate fondness for song and music,
or her liking for mirth and humour and every form of art, continued
to be thenceforth the dominant feeling of her life, bringing her into
closer and closer affinity with the “fervid” or “evangelical” in religion