IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 35, NO.

2, FEBRUARY 1999 129

Analytical Design of Double-Chirped Mirrors with

Custom-Tailored Dispersion Characteristics
Nicolai Matuschek, Franz X. Kärtner, and Ursula Keller, Member, IEEE

Abstract— We present a theory for the analytical design of do not answer important questions such as “Is it possible to
double-chirped mirrors with special dispersion characteristics. design chirped mirrors with some arbitrarily chosen dispersion
A simple analytical equation takes an arbitrarily desired group characteristics?” or “What is a good starting structure for a
delay dispersion (that also includes possible higher order disper-
sion) as an input function and gives the chirp law as an output. chirped mirror that can be used for computer optimization,
The chirp law determines the local Bragg wavelengths in the given the desired reflectance and GDD of the mirror?”
mirror. It allows the calculation of the thicknesses of the high- Recently, we developed a theory for chirped mirror analysis
and low-index layers if the double chirp of the layers in the front that resulted in the double-chirped mirror (DCM) design
part of the mirror is taken into account. We use this method to technique [7], [8]. The theory, which is based on an exact
design a highly dispersive double-chirped semiconductor Bragg
mirror and a double-chirped TiO2 –SiO2 mirror for higher or- coupled-mode analysis [9], allows for an easy and still pre-
der dispersion compensation in optical parametric oscillators cise calculation of the group delay and GDD by using the
operating in the visible spectral range. The design formulas are WKB approximation. In the case of weakly index-modulated
applicable to general chirped Bragg gratings and provide insight nonuniform grating structures (e.g., chirped fiber gratings), the
into the reasons why certain dispersion characteristics might be usefulness of the WKB approach has been already demon-
impossible to achieve.
strated by Poladian and Sipe et al. [10], [11]. In this paper, we
Index Terms—Chirped mirrors, coatings, coupled-mode analy- use our analytical expressions and derive the chirp law for the
sis, dielectric films, dispersion control, thin-film devices, ultrafast design of mirrors with an arbitrary dispersion characteristic.
optics, WKB analysis.
This paper widely uses the notation and results derived
in [8] and is organized as follows. In Section II, the main
I. INTRODUCTION results of [8] are summarized in a self-contained manner. In
Section III, a differential equation for the chirp law, given
R APID advances in ultrashort pulse generation have re-
sulted in sub-10-fs pulses from Ti:sapphire lasers [1]–[4].
It became evident that the main limitation to short pulse
a desired GDD, is derived within certain approximations.
Subsequently, in Section IV, we use the analytical solution
generation is given by higher order dispersion. Chirped mirrors of this differential equation and explicitly determine the chirp
[5] provide a powerful and compact technique for disper- law for some analytically and some numerically solvable cases.
sion compensation. Additionally, they exhibit a broader high- Finally, in Section V, the applicability and use of the derived
reflectance range than standard quarter-wave Bragg mirrors. formulas is demonstrated by investigation of two examples,
A chirped mirror introduces a controlled negative group delay a semiconductor Bragg mirror with a constant negative GDD
dispersion (GDD) by increasing the local Bragg period of the and a dielectric mirror for dispersion compensation in optical
mirror along the grating. parametric oscillators (OPO’s). We discuss the final design of
However, so far there were no analytical laws for the the dispersion compensating OPO mirrors obtained from the
chirping of the layer thicknesses for mirrors with a given theoretical starting design after numerical refinement. These
dispersion characteristic. Therefore, the mirrors are often de- DCM’s have the potential to support 6-fs pulses from an OPO
signed by time-consuming numerical optimization. The reason in the visible [12].
for this lack in appropriate design formulas is that the standard
design methods for optical interference coatings were mostly II. WKB DESCRIPTION OF A DOUBLE-CHIRPED MIRROR
focused on their reflection and transmission properties but According to Fig. 1, a DCM is an optical interference
not on the phase properties of the coatings. There exist only coating that in general consists of four multilayer subsections
a few papers concerning the phase properties of multilayer deposited on a substrate. For dispersion-compensating broad-
coatings by analytical means (see, for example, [6]). Most of band high reflectors, the layer materials are dielectrics such
these results are only of a qualitative nature and give little as SiO and TiO . The first section of a DCM is a broad-
analytical insight into the design problem. Additionally, they band antireflection (AR) coating, which typically consists of
8–14 layers and matches the subsequent optical coating to the
Manuscript received June 5, 1998; revised October 20, 1998. This work ambient medium [7], [8]. Here, we do not consider the problem
was supported by the Swiss National Science Foundation. of finding a proper high-quality AR coating (see, for example,
The authors are with the Swiss Federal Institute of Technology (ETH), [13]). We start to design the phase and amplitude properties
Institute of Quantum Electronics, Ultrafast Laser Physics Laboratory, ETH
Hönggerberg-HPT E9, CH-8093 Zurich, Switzerland. of the mirror under the assumption of perfect matching to
Publisher Item Identifier S 0018-9197(99)00857-X. the ambient medium. Therefore, we count the layers starting
0018–9197/99$10.00  1999 IEEE
130 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 35, NO. 2, FEBRUARY 1999

Fig. 1. Schematic drawing of a DCM, which is composed of four multilayer

sections. The different task of each section is described in the text. The
position in the mirror is determined by the variable m; where m = 0 defines
the beginning of the theoretical DCM structure directly after the AR coating.
Light, with wavenumber k; is reflected at the wavenumber-dependent classical
turning point mt (k ).
Fig. 2. Refractive-index profile of a DCM composed of two different
layer materials with the indices nh and n1 . The physical thickness of the
from the first layer following the AR-coating section and we symmetrically defined index steps is given by 3(zm ) = dh;m + d1;m ;
calculate all quantities with respect to this layer, which we where dh=1;m denotes the physical thickness of the high-/low-index layer
assume to be made of the low-index material. 0
at the ( m)th index step.
The double-chirp section is necessary for the avoidance of
Gires–Tournois-like oscillations in the group delay. Double connection to the coordinate is given via
chirping means that the local Bragg wavelength of the index where is the physical thickness of the
grating and the local coupling of the incident wave to the individual index step at position . Thus, in (1), the thickness
reflected wave are chirped simultaneously. This corresponds of a complete index step is normalized to . Note that
to an adiabatic matching of the impedance of the Bragg we place the mirror on the negative side of the axis, i.e.,
grating to the low-index layer at the front, as discussed in where defines the beginning of the mirror
[7], [8]. In the subsequent simple-chirp section (Fig. 1), only (see Fig. 1). As we have recently shown, a chirped mirror
the Bragg wavelength is chirped. This means that the high- is exactly described by the coupled-mode equations (1) even
and low-index layers are quarter-wave layers for maximum for arbitrarily high index differences of the layer materials,
reflection of waves with a wavelength close to the local Bragg if the exact normalized coupling coefficient and detuning
wavelength. The fourth section is a quarter-wave Bragg mirror coefficient are used [8], [9].
with fixed Bragg wavelength. The third and fourth sections are In [8], we used the WKB approximation to derive the
optional and depend on the special design problem considered. following expression for the phase of the wave reflected
However, they are necessary to obtain high reflectances over by a DCM with an ideally matched front part of the mirror,
a wavelength range as broad as possible. which is achieved by the double-chirp section:
The following theoretical derivation of the chirp law, as
presented in Sections III and IV, deals only with the double-
chirp and simple-chirp sections of the DCM and does not (2)
include the quarter-wave section. The reason is that wave-
lengths around the fixed Bragg wavelength of the quarter-wave
with
section are reflected approximately at the same position in
the mirror. This leads to an almost vanishing dispersion and
the one-to-one mapping of the wavelength with its point (3)
of reflection (classical turning point) is lost. However, the
quarter-wave stack at the end of the mirror can be considered Here, is the vacuum wavenumber. The propagation constant
as a continuation of the simple-chirp section in the case of in the grating is the square root of the negative scattering
a desired small amount of negative dispersion for the long potential of the Schrödinger equation, which is equivalent
wavelengths, as will be explained in more detail in Section to the coupled-mode equations (1), and is the right turning
V-C. point of the corresponding classical motion. The classical
The Bragg grating is decomposed into symmetrically de- turning points are defined by the condition
fined index steps [7], [8], and the discrete variable
counts the index steps of the chirped Bragg grating (see Fig. (4)
2). In [7] and [8], we found an exact description of a chirped
Bragg grating by exact coupled-mode equations of the form In the following, we consider only generic potentials with one
or two classical turning points [8, Fig. 6], where the right
(1) turning point corresponds to the positive sign in (4). In Fig.
3, the negative potential and the effective wavenumber
where and are the slowly varying amplitudes of the are shown for a typical DCM for a wavelength in the high-
forward and backward propagating waves, respectively. Here, reflectance region. The area in the figure corresponds to the
is considered to be a continuous variable, where the integral that has to be calculated when evaluating (2).
MATSUCHEK et al.: DOUBLE-CHIRPED MIRRORS WITH CUSTOM TAILORED DISPERSION CHARACTERISTICS 131

which agree with the coefficients from standard coupled-mode

theory [14]. Here, we have introduced the Fresnel reflectivity

(8)

where and are the refractive indices of the high- and

low-index layer materials, respectively. In the following, we
denote

(9)

as the local Bragg wavenumber and

as the corresponding local Bragg wavelength, where
and are the physical layer thicknesses of the high-
and low-index layers, respectively, at the th index step
( )
Fig. 3. The effective propagation constant q m in the grating is the square
0 ( )
root of the negative scattering potential U0 m . The phase of the reflected
(see Fig. 2).
( )
light is essentially given by twice the area A under the q m -curve, according
( = 0)
to (2). The area has to be integrated from the beginning of the mirror m B. Chirp Laws for the Bragg Wavenumber
to the right classical turning point mt . In the zeroth-order approximation, one
j ( )j
has to integrate over the absolute value of the detuning coefficient  m . Using the approximations (5)–(7), we derive the group delay
The error in the phase due to this approximation is indicated by the additional from (2)
1
area A. The curves in this figure are obtained for a typical wavelength in
the high-reflectance range of a dielectric DCM.

III. ANALYTICAL DERIVATION OF THE CHIRP LAW

In this section, we derive approximate expressions for the
group delay and the GDD, based on (2). The relation for the
GDD can also be interpreted as a differential equation for the
classical turning point of the Bragg grating as a function of
wavenumber, if a certain GDD is desired. (10)

A. Approximation of the WKB Formulas where is the velocity of light in vacuum with the
angular frequency . Note that the second equality holds
Generally, it is not possible to derive analytical expressions because the integrand (3) vanishes at the classical turning point
for the GDD due to the complicated integrand according to (4). Thus, we first built the derivative of (2) with
in (2). Therefore, we expand the square root in a respect to wavenumber, and afterwards we applied the zeroth-
Taylor series according to order approximation (5). The approximate group delay (10) has
a simple physical interpretation. It is the optical path length
for a wave traveling from the mirror surface to the classical
(5) turning point and back. From (10), we obtain the GDD

and take only the zeroth-order term for an analytical

estimation of the phase properties. The expansion (5) is always (11)
possible since in the interval . Fig. 3 also shows
the error in the integral (2) in zeroth-order approximation. The meaning of this equation is twofold. The equation
Note that, in the impedance-matched case considered here, the allows one to calculate the dispersion if the Bragg wavenumber
propagation constant and its approximation have the same is given as a function of the index step which
tangent at the beginning of the mirror, i.e., . also specifies the classical turning point as a function of
In principle, one has to use the exact coupling and detuning wavenumber. The equation can also be interpreted as a first-
coefficients for the calculation of the phase integral. However, order differential equation for the classical turning point if a
for an analytical treatment of the DCM dispersion, further certain dispersion is desired. Here, we are interested
approximations are necessary. Therefore, in this paper, we use in the second interpretation, since it gives us the Bragg
the expressions wavenumber for the index steps, and therefore the layer
thicknesses, for a desired dispersion characteristic. Thus, we
(6) write

(7) (12)
132 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 35, NO. 2, FEBRUARY 1999

where we have used the equation for the classical turning point to (9), the optical thickness as a function of the index step. At
first, (17) might look surprising, since it contains the product
(13) and not only because that would
relate to the change of the group delay over a distance of the
index steps, if each step would have equal optical length. But
which follows from condition (4) with the approximations (6) this is not the case, and the optical thickness of the th step
and (7). Equation (13) shows that, for negligible coupling is proportional to the inverse of .
coefficients, the incident wave is reflected at the point where
the wavenumber equals the Bragg wavenumber. This happens IV. ANALYTICAL EXAMPLES
where the approximate detuning coefficient vanishes according
In this section, we analytically discuss examples for using
to (7).
(17).
The general solution of the differential equation (12) is given
by
A. Analytically Invertible Examples

(14) For some simple cases of the desired GDD, the chirp law
can be derived explicitly. This is the case for a GDD that
obeys a power law of the form
where is an arbitrary integration constant. This equation
determines the classical turning point for a given wavenumber
(18)
. By using (13), which connects the Bragg wavenumber at
the classical turning point with the wavenumber the classical
Here, is the absolute value of the desired GDD at
turning point can be written as a function of the corresponding
wavenumber . For such a GDD, the third-order dispersion
Bragg wavenumber. Thus, we get the following inverse law
(TOD) is given by
for chirping the Bragg wavenumber with position in the
mirror:
(19)

One can see that the TOD is positive for negative for
(15) and vanishes for . If we substitute (13) into (18),
the calculation of the integral in (17) results in
Note that the index at the variable is no longer necessary,
since we consider as an independent variable. The inte-
gration constant is determined by a boundary condition—for
example, the Bragg wavenumber at the beginning of the
mirror. Hence, we have a condition of the form

(16)

where denotes the maximum Bragg wavenumber of a

chirped mirror that produces a negative GDD. In that case,
the Bragg wavenumber is a monotonically decreasing function
(20)
with respect to the negative axis.
Using (16), we obtain the following closed-form solution The inverse relationship leads to (21), shown at the bottom
for the turning point: of the page, where we have used . Note that the
chirp law for the case follows directly from the case
(17) in the limit . As an example, to obtain a
constant negative GDD , one has to chirp the Bragg
This equation allows us to calculate the position of the index wavenumber according to the following square root law:
step at which the Bragg wavenumber has the value in
order to achieve the given GDD. The inverse of this equation (22)
determines the Bragg wavenumber and, therefore, according

(21)
MATSUCHEK et al.: DOUBLE-CHIRPED MIRRORS WITH CUSTOM TAILORED DISPERSION CHARACTERISTICS 133

which is a rather unexpected result. An example using this

analytical chirp law will be discussed in detail in Section V-
A in the context of the design of a chirped semiconductor
mirror. Reference [15] shows a comparison of group delays
with different TOD’s obtained when using (21) for different
values of the chirp exponent .
Since we will want to compare the exact calculated group
delay with the group delay obtained from the WKB solution,
we give the group delay for the chirp law (21) according to
(10) by using (21) and (13) for the evaluation of the integral
and determination of the classical turning point. The result,
Fig. 4. Chirp law, which gives the Bragg wavenumber or Bragg wavelength
(23), is shown at the bottom of the page. In the same way as m
as a function of the normalized position j j in the mirror, for a highly
mentioned above, the case follows from the dispersive DCSM. In the case of a desired constant GDD, the chirp law with
case in the limit . respect to Bragg wavenumber is a square root function according to (22). For
the example discussed in the text, the chirp law has been used only over the
first 60 index steps.
B. Taylor Expansion of the Desired Dispersion Characteristic
In general, the relationship (17) cannot be inverted. How- V. NUMERICAL EXAMPLES
ever, we can always compute the index step as a
function of Bragg wavelength if the dispersion is given as A. Double-Chirped Semiconductor Mirror
a Taylor expansion around a center wavenumber
Here, we discuss the design of a double-chirped semi-
conductor mirror (DCSM) with a constant negative GDD.
This example illustrates the method and demonstrates the
applicability of the theory not only to the design of chirped
dielectric mirrors but also to the design of chirped Bragg
(24) gratings in general. However, we do not discuss the problem of
matching the DCSM to air and the improvement of the design
by computer optimization. A complete design including these
Here, we have introduced the dispersion coefficients
issues is described in Section V-B.
Here, we use the constant refractive indices and
(25) which are close to the refractive indices of AlAs and
GaAs, respectively, at around 1050 nm. For these indices, the
approximate coupling coefficient is according
Substitution of (13) into (24) and integrating the GDD ac- to (6). We desire a constant GDD. Thus, (22) is used as the
cording to (17) finally leads to chirp law for the Bragg wavenumber. For the desired GDD,
a typical absolute value of fs is assumed. A
dispersion value at this order of magnitude is desired in many
applications where large amounts of dispersion have to be
compensated for within a single bounce on the mirror. The
(26) region of high reflectance and constant dispersion should be
centered at around 1050 nm. The value for the minimum Bragg
wavelength is set to nm, which corresponds to
Once this expression has been obtained and plotted, the a maximum Bragg wavenumber of nm
inversion can always be obtained by inspection. In Section m .
V-B, we discuss the complete design of a DCM starting from Fig. 4 shows the Bragg wavenumber and Bragg wavelength
the analytical equation (26). This will show the usefulness with respect to the normalized distance in the mirror, obtained
of this approach for the analytical determination of an initial for the values above. Obviously, the Bragg wavenumber
design that is already very close to the desired dispersion decreases to zero at . Thus, with the chirp law
characteristic. used here, one can only chirp the Bragg wavenumber over a

(23)
134 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 35, NO. 2, FEBRUARY 1999

TABLE I
CENTER WAVELENGTH AND DISPERSION COEFFICIENTS, ACCORDING TO (24),
FOR THE DESIRED GDD OF A DIELECTRIC DCM THAT CAN BE USED FOR
DISPERSION COMPENSATION IN AN OPO, AS DESCRIBED IN THE TEXT

increase with a higher power law, e.g., with an exponent 2,

would lead to unrealistically thin high-index layers in the front
part of the mirror. Additionally, a double chirp which is too
strong would superimpose an additional dispersion effect that
is undesired for the demonstration of the method described
(a)
here.
As can be seen from Fig. 5(a), the high-reflectance range
has a bandwidth of more than 100 nm. However, due to the
long matching section, a part of the high-reflectance range
has been lost on the short wavelength side. This is why the
minimum Bragg wavelength was chosen as short as 980 nm.
Fig. 5(b) shows the group delay and GDD. Obviously, the
theoretical design is very close to the desired group delay
and GDD, respectively. The designed GDD oscillates around
the desired constant value. The average value is in excellent
agreement with the desired value. The oscillations are due to
imperfections in the double-chirp section and could be further
reduced by methods such as numerical optimization. One has
to note that the desired group delay is only determined up to
a constant offset value. Therefore, we adjust the desired group
delay for one wavelength. The dash–dotted line in Fig. 5(b)
(b) is the group delay obtained from the WKB solution (23) for
Fig. 5. Reflectance, group delay, and GDD of a double-chirped semiconduc- the analytical chirp law (22) . The constant difference
tor Bragg mirror using the chirp law shown in Fig. 4. The thickness of the
high-index layer was chirped as described in the text. (a) The bandwidth of
between the adjusted group delay and the WKB solution is
the highly reflective region is about 100 nm. (b) The designed GDD (solid about 50 fs. This difference is explained by the assumption
line) oscillates around the desired constant GDD (dashed line). The difference of a constant coupling coefficient, according to (6). As we
between the designed group delay (solid line) and the group delay obtained
from the WKB solution (dash–dotted line) is mostly due to the assumed
discussed in [8], that approximation leads to an almost constant
constant coupling coefficient. difference between the exact computed group delay and the
WKB solution. However, this is irrelevant for the physically
maximum distance of about 280 index steps. However, due to most important quantity, which is the GDD.
practical limitations arising from semiconductor growth (e.g.,
with MOCVD), the number of index steps cannot be much
larger than 100. Thus, for a demonstration, we use 60 index B. Design of a Double-Chirped Dielectric Mirror
steps. For the computation of the mirror properties, the Bragg Here, we discuss the complete design of a dielectric DCM
wavenumber has to be taken at 60 discrete points defined with a desired dispersion characteristic, given as a Taylor
by with . The reflectance and expansion (24) up to sixth order. The center wavelength
phase properties upon reflection are exactly calculated with and dispersion coefficients are summarized in Table I. Six
the transfer matrix formalism. bounces on this mirror should compensate for the material
Fig. 5 shows the mirror reflectance, group delay, and dispersion introduced by a 2-mm-long BBO crystal and 1 mm
GDD, where all properties are calculated without consid- of fused silica [12]. It is of considerable interest to have broad-
ering the index jump to an ambient medium. In order to band dispersion compensating mirrors at around 600 nm for
avoid undesired oscillations, the impedance was matched very short-pulse OPO’s [16], [17] and optical parametric amplifiers
slowly over the first 50 index steps by an appropriate slow (OPA’s) [18], [19] in that wavelength range.
tapering of the coupling coefficient [7], [8]. Thus, the design For the numerical evaluation of (26), the maximum Bragg
presented here consists of only two parts, the double-chirp wavenumber was chosen to be nm
section (50 index steps) and a simple-chirp section (10 index m . For the coupling coefficient, we used the constant
steps). Here, the coupling coefficient was increased almost value which is obtained from (6) for the indices
linearly via the thickness of the high-index layers according and . These indices are close to the
to . The reason is that an indices of SiO and TiO at wavelengths around nm.
MATSUCHEK et al.: DOUBLE-CHIRPED MIRRORS WITH CUSTOM TAILORED DISPERSION CHARACTERISTICS 135

Fig. 6. Chirp law for a dielectric DCM with a desired GDD, as summarized
in Table I. The minimum Bragg wavelength was chosen to be 500 nm. At
m
j j 24; the inverse chirp law is not invertible due to the increase of the
desired GDD for longer wavelengths [see Fig. 7(b)]. Also shown are the 23
discrete points at which the chirp law was evaluated in order to generate the
theoretical designs shown in Fig. 7. The dashed vertical line symbolizes the (a)
position in the mirror where all wavelengths around the Bragg wavelength are
reflected, when a quarter-wave section follows after the simple-chirp section.

Fig. 6 shows the Bragg wavenumber and Bragg wavelength

again as a function of the position in the mirror. One can see
that, for this example, (26) is not a one-to-one function. This
means that, at , the inverse chirp law (24) is locally
not invertible. The occurrence of this phenomenon is discussed
in detail in Section V-C. Thus, using this law, one can chirp the
Bragg wavenumber only over a maximum number of 24 index
steps. In order to generate a mirror design with not too many
layers, the design presented here was chirped over 23 index
steps. The 23 discrete points for the Bragg wavenumbers,
where the chirp law is evaluated, are also shown in Fig. 6.
Fig. 7 shows the mirror properties for a final design that (b)
can be fabricated. This design is achieved by starting from Fig. 7. Broad-band design of a DCM that is useful for dispersion com-
the analytical design obtained with the discrete points for the pensation in an OPO. The designed reflectance, group delay, and GDD are
Bragg wavenumber from Fig. 6. Additionally, the problem obtained using the chirp law shown in Fig. 6. The thickness of the high-index
layer was chirped as described in the text. The mirror properties are shown
of matching to air is solved by putting a broad-band AR after adding an AR coating on top of the mirror structure and a subsequent
coating on top of the mirror [7], [8]. Finally, the design computer refinement. (a) The mirror is highly reflective over a bandwidth of
performance is improved by computer optimization with a about 250 nm. The designed group delay (solid line) is extremely smooth and
is very close to the desired group delay (dashed line) over a bandwidth of
standard gradient algorithm from [20]. It is important to note about 180 nm. (b) The designed GDD (solid line, upper plot) oscillates around
that, for the refinement procedure and calculation of the final the desired GDD (dashed line, upper plot) with an amplitude of about 6–7
mirror properties, the wavelength dependence of the refractive fs2 in that range. The lower plot shows the difference between the designed
and desired GDD.
indices as well as the absorption and scattering losses have
to be taken into account. In contrast, this is not necessary for
the determination of the discrete Bragg wavenumbers from Fig. 8 shows the physical layer thicknesses of the starting
(26), because these effects are considered to be small. For the design and of the final design after computer optimization.
generation of the analytical initial design, the thickness of the The total design consists of 54 layers, in which 46 layers
high-index layer was almost linearly chirped over the first 10 have been used for the theoretical DCM structure and 8 layers
index steps according to . for the AR coating. The layer thicknesses of the AR coating
Fig. 7(a) shows the reflectance and group delay of the final are only shown after optimization. The reason is that we
design. The reflectance is high over a bandwidth of 250 nm started from a 12-layer AR coating, but during the optimization
(500–750 nm), and the group delay is very smooth over a procedure it was possible to eliminate the two thinnest layers
bandwidth of about 180 nm (540–720 nm). In this range, of the AR coating leading to an effective reduction of 4
the deviation from the desired group delay is only 0.14 fs layers. The figure illustrates two facts. The first observation
(rms). As is demonstrated by Fig. 7(b), the designed GDD is that the analytical initial design of the DCM does not
oscillates around the desired GDD in this region, with an change very much during optimization. Most corrections to
oscillation amplitude of 6–7 fs . Although the starting design the starting design occur in the AR coating in the front part of
is computer optimized, the oscillations in the GDD cannot be the total mirror. The other observation is that the different
removed completely. The reason is that the AR coating does sections of the DCM can still be clearly identified. In the
not allow the generation of a perfect matching to the ambient AR coating, the layer thicknesses seem to change randomly,
medium (air) over a bandwidth of 180 nm. whereas the double-chirp section starts with very thin high-
136 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 35, NO. 2, FEBRUARY 1999

is left, are clearly indicated by plotting and analyzing the

chirp law, derived from (21) or (26). As an example, the
noninvertible region in Fig. 6 at a finite Bragg wavelength
is a consequence of the increase of the desired GDD with
respect to wavelength [see Fig. 7(b)]. In fact, the critical Bragg
wavenumber at which this happens corresponds exactly to the
wavenumber where the desired GDD becomes positive. Using
the values for the design discussed in Section V-B, the critical
Bragg wavenumber is m nm)
and the wavenumber for vanishing dispersion is found to be
m nm). These wavenumbers
are connected via (13). Now, it is quite easy to understand what
happens. As long as the desired GDD is negative, the Bragg
Fig. 8. Layer thicknesses of the DCM design corresponding to the mirror wavenumber is a monotonically decreasing function. When
properties shown in Fig. 7. The thicknesses of the high- and low-index layers
are shown before and after computer optimization. The layer thicknesses of the the desired GDD becomes positive, the direction of the chirp
AR coating are only shown after optimization because the two thinnest layers has to be changed, which means that the Bragg wavenumber
of the AR coating are removed during the refinement procedure. Obviously, should now increase. Of course, the formulas derived from
the different sections of a DCM can still be clearly identified even after
computer optimization. WKB solutions cannot cover this case. Additionally, the fast
chirp for the longer wavelengths also explains the problem of
keeping the reflectance at a high level, as is demonstrated by
index layers with increasing thicknesses along the DCM. The the decrease of the reflectance with respect to wavelength [see
thickness of the low-index layers decreases in the double- Fig. 7(a)]).
chirp section. In the simple-chirp section, the thicknesses A quarter-wave section following the simple-chirp section
of the high- and low-index layers increase simultaneously. offers a possible solution over a limited wavelength range
The stability of the starting design against changes due to when the noninvertible region is reached. The quarter-wave
the optimization procedure is understood by the fact that a stack solves both problems: it increases the reflectance and
sufficiently good AR coating can be designed over a bandwidth leads to an almost vanishing dispersion for wavelengths around
of about 180 nm at wavelengths around 600 nm. If one tries the corresponding Bragg wavelength. Thus, in such a case, the
to extend the bandwidth with smooth dispersion properties Bragg mirror can be considered as the “natural” continuation
beyond the bandwidth of the AR coating, the optimization of the simple-chirp section, symbolized by the dashed vertical
will significantly change the layer thicknesses of the DCM line in Fig. 6. The Bragg wavelength of the quarter-wave
structure. Additionally, the amplitude of the oscillation in the section is a free parameter which can be adjusted to achieve
GDD will increase with the required bandwidth. maximum performance. It should be chosen much smaller
The mirror design presented here will be manufactured in than and close to the last value used in the simple-chirp
the near future. If the properties of the fabricated mirrors will section. However, this method works only over a restricted
be close to the designed properties, ultrashort laser pulses from wavelength range.
an OPO with a duration of about 5–7 fs can be expected [12]. The conclusion is that, if the inverse chirp law is not invert-
ible in the region of interest, as described above, the design
C. Unfavorable Choice of the Desired method proposed in this paper will not lead to an appropriate
Dispersion Characteristic theoretical starting design for later computer refinement. A
The analytical discussion of the dispersion behavior of possible solution is to change the desired dispersion char-
DCM’s gives insight into reasons why certain dispersion acteristic or to add a quarter-wave stack after the simple-
characteristics might be difficult, if not impossible, to achieve. chirp section. On the other hand, the failure of producing a
For example, such a situation arises if the absolute value of reasonable initial design indicates that an optimum design for
the desired second-order dispersion coefficient is chosen a high-reflectance coating with the desired dispersion might
too low or too high, or if the ratio between the different not exist.
dispersion coefficients is chosen in an unfavorable
way. Generally, one has to chirp relatively slowly to obtain a
high amount of negative dispersion and relatively quickly for VI. CONCLUSION
a small amount. As a consequence, in the first case, one needs We have presented an analytical design method for DCM’s
many index steps to achieve a broad enough high-reflectance or general chirped Bragg gratings with a given dispersion char-
bandwidth. In the second case, it is possible that the reflectance acteristic. Problems in the mirror design that may arise due to
is reduced due to the fast chirp over a small number of index an unfavorable choice of the desired GDD are indicated by the
steps. Thus, for given materials, there is always an optimum corresponding chirp law. The chirp law also directly indicates
range for the dispersion, over which a high reflectance and how many layers have to be used for the theoretical design
smooth GDD can be achieved simultaneously. in order to achieve the desired dispersion characteristic over a
The great advantage of the design method described in given bandwidth. The applicability of our design method has
this paper is that situations, in which the optimum range been demonstrated by two examples, a semiconductor and a
MATSUCHEK et al.: DOUBLE-CHIRPED MIRRORS WITH CUSTOM TAILORED DISPERSION CHARACTERISTICS 137

dielectric DCM. As has been shown, the analytical designs Nicolai Matuschek was born in Biberach an der
achieved by this method are very close to the desired design Riß, Germany, in 1968. He received the diploma
degree in physics from the University of Ulm,
goals and can be used as excellent starting designs for a Ulm, Germany in 1995. He is currently working
subsequent computer optimization. toward the Ph.D. degree at the Institute of Quantum
Electronics, Swiss Federal Institute of Technology
(ETH), Zurich, Switzerland.
REFERENCES After finishing his diploma degree in the field of
theoretical polymer physics, he changed his research
[1] J. Zhou, G. Taft, C.-P. Huang, M. M. Murnane, and H. C. Kapteyn,
interests to laser physics, where he is currently
“Pulse evolution in a broad-bandwidth Ti:sapphire laser,” Opt. Lett.,
focused on the theory of mode-locked solid-state
vol. 19, pp. 1149–1151, 1994.
lasers.
[2] A. Stingl, M. Lenzner, Ch. Spielmann, F. Krausz, and R. Szipöcs, “Sub-
10 fs mirror-dispersion-controlled Ti:sapphire laser,” Opt. Lett., vol. 20,
pp. 602–604, 1995.
[3] I. D. Jung, F. X. Kärtner, N. Matuschek, D. H. Sutter, F. Morier-Genoud,
G. Zhang, U. Keller, V. Scheuer, M. Tilsch, and T. Tschudi, “Self-
starting 6.5 fs pulses from a Ti:sapphire laser,” Opt. Lett., vol. 22, pp.
1009–1011, 1997.
[4] D. H. Sutter, I. D. Jung, F. X. Kärtner, N. Matuschek, F. Morier-Genoud, Franz X. Kärtner was born in Cham, Germany, in
V. Scheuer, M. Tilsch, T. Tschudi, and U. Keller, “Self-starting 6.5-fs 1961. He received the diploma degree in electrical
pulses from a Ti:sapphire laser using a semiconductor saturable absorber engineering from the Technical University Munich,
and double-chirped mirrors,” IEEE J. Select. Topics Quantum Electron., Munich, Germany, in 1986 and the Ph.D. degree
vol. 4, pp. 169–178, 1998. for work on noise in microwave circuits and optical
[5] R. Szipöcs, K. Ferencz, Ch. Spielmann, and F. Krausz, “Chirped devices in 1989.
multilayer coatings for broadband dispersion control in femtosecond From 1991 to 1993, he was a Feodor-Lynen
lasers,” Opt. Lett., vol. 19, pp. 201–203, 1994. Research Fellow of the Humboldt Foundation at
[6] A. V. Tikhonravov, P. W. Baumeister, and K. V. Popov, “Phase Massachusetts Institute of Technology (MIT), Cam-
properties of multilayer coatings,” Appl. Opt., vol. 36, pp. 4382–4392, bridge, MA, where he carried out research on the
1997. generation of squeezed states in fibers and quantum
[7] F. X. Kärtner, N. Matuschek, T. Schibli, U. Keller, H. Haus, C. Heine, nondemolition measurements. From 1993 to 1997, he was with the Institute of
R. Morf, V. Scheuer, M. Tilsch, and T. Tschudi, “Design and fabrication Quantum Electronics, Swiss Federal Institute of Technology (ETH), Zurich,
of double-chirped mirrors,” Opt. Lett., vol. 22, pp. 831–833, 1997. Switzerland. He received his Habilitation in 1997 with work on ultrashort
[8] N. Matuschek, F. X. Kärtner, and U. Keller, “Theory of double-chirped pulse generation. Since 1998, he has been a Visiting Assistant Professor at
mirrors,” IEEE J. Select. Topics Quantum Electron., vol. 4, pp. 197–208, MIT.
1998. Dr. Kärtner is a member of the German Physical Society, the Optical
[9] , “Exact coupled-mode theories for multilayer interference coat- Society of America, and the IEEE.
ings with arbitrary strong index modulations,” IEEE J. Quantum Elec-
tron., vol. 33, pp. 295–302, 1997.
[10] L. Poladian, “Graphical and WKB analysis of nonuniform Bragg grat-
ings,” Phys. Rev. E, vol. 48, pp. 4758–4767, 1993.
[11] J. E. Sipe, L. Poladian, and C. M. de Sterke, “Propagation through
nonuniform grating structures,” J. Opt. Soc. Amer. A, vol. 11, pp.
1307–1320, 1994. Ursula Keller (M’89) was born in Zug, Switzer-
[12] G. M. Gale, private communication 1998.
land, in June 1959. She received the “Diplom” in
[13] J. A. Dobrowolski, A. V. Tikhonravov, M. K. Trubetskov, B. T. Sullivan,
physics from the Swiss Federal Institute of Tech-
and P. G. Verly, “Optimal single-band normal-incidence antireflection
nology (ETH), Zurich, Switzerland, in 1984 and
coatings,” Appl. Opt., vol. 35, pp. 644–658, 1996.
[14] H. A. Haus, Waves and Fields in Optoelectronics. Englewood Cliffs, the M.S. and Ph.D. degrees in applied physics
NJ: Prentice-Hall, 1998. from Stanford University, Stanford, CA, in 1987
[15] N. Matuschek, F. X. Kärtner, and U. Keller, “Double-chirped mirror and 1989, respectively. Her Ph.D. research was
designs with adjustable higher order dispersion,” in Proc. Conf. Lasers in optical probing of charge and voltage in GaAs
and Electrooptics (CLEO ’98), San Francisco, CA, May 3–8, 1998, paper integrated circuits and in low-noise ultrafast laser
CThC6. systems.
[16] G. M. Gale, M. Cavallari, and F. Hache, “Femtosecond visible optical From late 1984 to 1985, she worked on optical
parametric oscillator,” J. Opt. Soc. Amer. B, vol. 15, pp. 702–714, 1998. bistability at Heriot-Watt University, Edinburgh, Scotland, U.K. In 1989, she
[17] G. M. Gale, F. Hache, and M. Cavallari, “Broad-bandwidth parametric joined AT&T Bell Laboratories, Holmdel, NJ, as a Member of Technical
amplification in the visible: Femtosecond experiments and simulations,” Staff, where she conducted research on photonic switching, ultrafast laser
IEEE J. Select. Topics Quantum Electron., vol. 4, pp. 224–229, 1998. systems, and semiconductor spectroscopy. Since March 1993, she has been
[18] G. Cerullo, M. Nisoli, and S. De Silvestri, “Generation of 11 fs pulses an Associate Professor and since October 1997 a Full Professor in the Physics
tunable across the visible by optical parametric amplification,” Appl. Department at the ETH, Zurich. Her current research interests are in ultrafast
Phys. Lett., vol. 71, pp. 3616–3618, 1997. lasers, spectroscopy, local probes, and novel devices for applications in optical
[19] A. Shirakawa, I. Sakane, and T. Kobayashi, “Pulse-front-matched optical information processing, communication and medicine. She has published more
parametric amplification for pulse generation tunable in the visible and than 80 journal papers and one book chapter and holds five patents.
near-infrared reaching to 6.5 fs,” in Proc. Conf. Lasers and Electrooptics Prof. Keller is a member of the Optical Society of America, the European
(CLEO ’98), San Francisco, CA, May 3–8, 1998, paper CPD11-2. Physical Society, the Swiss Physical Society, and the Swiss Academy of
[20] W. H. Press, S. V. Teukolsky, W. T. Vetterling, and B. P. Flannery, Technical Sciences. During 1985–1986, she was a Fulbright Fellow, in
Numerical Recipes in Fortran, 2nd ed. Cambridge, U.K.: Cambridge 1987–1988 she received an IBM Predoctoral Fellowship, and in 1998 she
Univ., 1994. received the Carl Zeiss Research Award.