Sunteți pe pagina 1din 86

High-Resolution Radius of Curvature Measurements for FEL Cavity Mirrors Agee Springer Senior Thesis Research Advisor: Dr.

Ying Wu

Abstract In this project, research into optimizing a radius of curvature measurement system for FEL cavity mirrors was undertaken. The system is characterized by significant diffraction effects due to the use of a He-Ne laser beam as a light source. Optimization of image formation through diffraction control was studied in both computer simulations and the physical system. Radius of curvature measurements of many FEL cavity mirrors were conducted. The results show the method to be theoretically sound and capable of producing measurements of long ROC mirros accurate to within 0.5%.

Table of Contents
Abstract...............................................................................................................1 Table of Contents ................................................................................................2 1. Introduction.....................................................................................................3 2. Optical Theory 2.1 Ray Optics.......................................................................................7 2.2 Diffraction Theory...........................................................................13 2.3 Gaussian Optics...............................................................................24 3. Simulation Method 3.1 Description of OPC Code ................................................................30 3.2 Simulation Examples .......................................................................31 4. ROC Measurements of FEL Mirrors 4.1 Experimental Setup..........................................................................39 4.2 Measurement Results.......................................................................42 4.3 Simulation Results...........................................................................53 4.4 Discussion .......................................................................................71 5. Conclusion and Future Research .....................................................................74 6. Acknowledgements .........................................................................................76 7. Works Cited ....................................................................................................77 Appendices A.1: OPC Code Used for Plane Wave Simulation ...............................................78 A.2: OPC Code Used for Gaussian Beam Simulation..........................................82

1. Introduction
The Duke Free-Electron Laser Laboratory (DFELL) is a world leader in developing storage ring based ultraviolet (UV) and vacuum-ultraviolet (VUV) free-electron lasers (FELs). The Duke FEL is also a driver for the world's most powerful high-energy Compton gamma-ray facility, the High Intensity Gamma-ray Source (HIGS), a premier light source facility codeveloped by the DFELL and Triangle Universities Nuclear Laboratory (TUNL). The UVVUV FEL and HIGS have been used for frontier research in a number of scientific areas, from subfields of physics such as nuclear physics and surface physics, to chemistry, biology, and medical sciences. The Duke FEL employs world's longest active laser cavity with two concave mirrors separated by 54 meters. The performance of the FEL and HIGS gamma-ray source critically depends on the proper choice of FEL mirror parameters. More specifically, the correct choice of radius of curvature (ROC) of the FEL mirrors is essential for stable and high power operation of the FEL and HIGS. If the ROC is smaller than half the length of the cavity (for a symmetric design), the cavity will be unstable and lasing will not occur. However, if the ROC is too large, the large optical beam will not be matched to the small size of the FELs electron beam. This mismatch will lead to poor interaction between the optical and electron beams, resulting in nonoptimal operation of both the FEL and HIGS. Therefore, it is imperative that a high-precision ROC measurement technique is developed to ensure the optimal and successful operation of the FEL. Historically, the long ROC of FEL cavity mirrors has made ROC measurements extremely difficult. While there has not been significant work done developing measurement techniques of this type, interferometry has been traditionally been employed to obtain

measurements for mirrors with long ROC. One method is to use Newtons rings by placing an optical flat over the surface of a concave mirror. This distance between the mirror and the optical flat varies as dictated by the mirrors ROC (see figure 1).

Figure 1: An example setup for generating Newtons rings.

When this setup is illuminated from above by coherent light, the variation in distance leads to an interference pattern, Newtons rings. While these rings can be used to measure a mirrors ROC, it is not possible to obtain the high-precision measurements of long ROC mirrors required for successful FEL operation from this system. For longer ROC measurements, researchers at the Korea Institute of Standards and Science have developed a measurement apparatus using a half-aperture bidirectional shearing interferometer (see figure 2).1

Yun Woo Lee, et. al., Noncollimated bidirectional shearing interferometer for measuring a long radius of curvature, Applied Optics, 36.22 (1997): 5317. 4

Figure 2: An interferometric measurement system.

In this system, the beam is split into two components, with one being reflected off a plane mirror, and the other off the mirror under study. The beams are reconstituted and the interference fringes of the two branches are compared. The point source is then displaced until the interference fringes from the two branches are identical. The ROC of the mirror can then be determined from Newtons lens equation. While the above system is capable of giving results accurate to within 1%, it suffers from several limitations. In addition to being very complex, the system also depends on a perfectly flat plane mirror for measurements. For FEL cavity mirror ROC measurements, we require an accuracy of

"R ~ 0.5% , R
which corresponds to 50 nm surface flatness for the plane mirror. This means that anything less

! than a plane mirror of near-perfect flatness will introduce significant systematic error into ROC

measurements. Because of this limitation, this system has only been used to test mirrors with a ROC of 10 m or less, and is not optimal for measuring the cavity mirrors at DFELL.2 A different system employing the geometric imaging principle has been used for many years at DFELL for ROC measurements of FEL cavity mirrors.3 The principle states that the image of an object viewed through a lens (or spherical mirror) can be found according to
1 1 1 + = , s s" f

where s is the distance from the object being viewed to the lens, s is the distance from the lens
! to the image, and f is the focal length of the lens. In the measurement system employed at

DFELL, a light source is used to illuminate a target object. The target object is located approximately 28 m from the cavity mirror under study. The light is reflected off the mirror and a CCD camera is used to locate the image. Once the location of the image has been determined, the ROC of the mirror can be determined from the above formula, using f = -R/2 as the focal length for a spherical mirror. The original incarnation of this measurement setup utilized a white light source to illuminate the target object. This design choice has an important advantage over the use of a coherent source, such as a laser. Because the light emitted by an incandescent source has neither spatial nor temporal coherence, diffraction effects are essentially eliminated from the system. While this method proved useful for mirrors with a high enough reflectivity in the visible spectrum, it was not useful for measurement of the ROC of UV mirrors. These mirrors reflect

2 3

Lee, et. al., 5317-5320. J. Li, Y. K. Wu, C. Sun, Improved Long Radius of Curvature Measurement System for FEL Mirrors. Proc. of 2005 Particle Accelerator Conf., 16 May-20 May 2005, Knoxville, Tennessee: 1787-1789.

very little visible light, and the white light source was simply not powerful enough to allow for the formation of a reflected image. In order to facilitate the measurement of UV mirrors, an upgrade was begun several years ago to replace the white light source with a coherent source in the form of a He-Ne laser. The high intensity of the laser beam allows for the formation of an image despite the low reflectivity of the mirror. However, the diffraction effect due to the use of a highly coherent light source causes degradation of the image, which, in many cases, prevents accurate ROC measurements. Thus, we arrive at the physics problem at hand - optimizing the image formation of an object illuminated by a coherent light source by reducing diffraction effects. This project explores this problem in detail through the use of computer simulations and other theoretical tools. The physical ROC measurement technique is also optimized with the goal of creating a system capable of measuring a wide range of mirrors, ranging from the IR to the UV regimes, to an accuracy of 0.5%.

2. Optical Theory
2.1 Ray Optics Although usually covered in the most basic introductory physics course, the concepts of geometric optics play an important role in this project and merit a brief review. Light is fundamentally a wave phenomenon. However, thinking of light in terms of the propagation of wavefronts often presents unnecessary complexities in dealing with simple problems in optics. To deal with this, one utilizes optical rays to visualize the system. A ray is a line drawn perpendicular to a wavefront and indicates the direction of propagation. It is important to note that a ray is a mathematical construct. The electromagnetic wave is the physical entity.4 The ray

Eugene Hecht, Optics (San Francisco, Pearson Education, 2002, 4th ed.) 98. 7

formulation allows for understanding of the important geometric optical effects, such as reflection and refraction, image formation, and misalignment effects, without the constant need to consider the wave nature of light. While Fermats principle dictates that light travels in straight lines in homogeneous media, the laws that govern a beams propagation through various optical instruments are complex and in general not linear. To illustrate, consider an optical beam that makes an angle with the optical axis, (see figure 3).

Figure 3: An example optical ray.

We know that the ray can be described by the parameters, r, the distance from the optical axis (which will henceforth be referred to as the z-axis by convention), and the slope of the line,

r" = tan(# ) .
While the second parameter is simple enough in free space, difficulties are encountered when we consider the effects of an optical ! instrument, such as a lens, on the propagation of the ray. However, if is small, we can make the paraxial approximation and linearize the tangent.5 This gives

r" # $ ,

Dieter Meschede, Optics, Light and Lasers (Mrlenbach, Germany, Verlag-GmbH & Co., ! 2007) 20. 8

r and we can describe the ray with a vector r = ( r," ) . This approximation might at first seem to
be of little utility. After all, most light sources emit radiation over a wide range of angles.

! However, the paraxial approximation is remarkably accurate when describing the beam
generated by a laser, such as the one utilized in the ROC measurement setup, and will be sufficient here. Using this framework, it is a simple matter to describe the propagation of an optical

r beam. If we have a ray characterized by the parameters r1 = ( r1,"1 ) and propagate it a distance d
(see figure 4), then the ray is characterized by
r2 =! + "1d , r1 " 2 = "1. ! !

Figure 4: Ray propagation in the paraxial approximation.

This linear relation can easily be written down in the form of a matrix,

"1 d% MD = $ ', #0 1&


where D indicates that propagation over a drift space is being described. This gives

r r r2 = M D r1 .

While a matrix might seem unnecessary for this simple example, its usefulness becomes clear
! when systems of optical instruments is considered.

There are two optical devices that are especially relevant to this experiment, the converging lens and the concave spherical mirror. Most generally, a lens is a refracting device...that reconfigures a transmitted energy distribution.6 More practically, a converging lens is a device that takes the divergent light from a point source and focuses it back down to a point.7 Mathematically, if we have an incoming ray with parameters r1 and 1, the effect of a lens is found according to8
r2 = r1, #1 " 2 = r1 + "1 , f ! where f is the focal length of the lens or curved mirror being studied. From these formulas, it is

! clear that we can express the effects of the lens in the form of a matrix, much as we did with the

drift space propagation of a ray. This matrix is


# 1 0& M L = %"1 ( . $ f 1'

Like a lens, a spherical mirror is a device that converges incoming light rays to a point.
! However, a mirror is a primarily reflective device, while a lens is primarily refractive. A

spherical mirror has an associated focal length, given by

f =

R , 2

where R is the mirrors radius of curvature, and is defined as positive for a concave mirror. The

! transformation matrix associated with a mirror is then


#1 M M = %"2 $R
6 7

0& (. 1'

Hecht, 150. ! Meschede, 16. 8 Anthony Siegman, Lasers (Mill Valley, CA, University Science Books, 1986) 582. 10

It is important to note that, while an ideal mirror reflects all incident light, this is not necessarily the case for a physical mirror. Indeed, a real mirror will reflect the majority of incident rays, but some will be transmitted through the medium of the mirror. The portion of the incident beam that is reflected may also depend on wavelength, making a mirror only useful in a certain region of the electromagnetic spectrum. This is a real problem when trying to obtain ROC measurements. FEL cavity mirrors designed for UV lasing are designed with a high reflectivity in the UV regime, but reflect very little visible light, making ROC measurements difficult. Our discussion of common optical elements and their associated transformation matrices illustrates the utility of thinking of optics in this way. Consider a system of lenses (any optical elements will do, but we will consider lenses for ease of illustration here) separated by drift
r spaces. If we have an incoming optical ray, r1 (see figure 5), and we wish to know the effect of

these lenses, we must propagate the ray to the first lens, then apply the lens transform for the first lens, then propagate to the second ! and so on. lens

Figure 5: An example lens system.

However, if we view this system in the form of a series of linear transformation matrices, it is a fairly simple matter to find the rays parameters at d5 using matrix multiplication.
r r r r5 = M D 5 M L 4 M D 4 M L 3 M D 3 M L 2 M D 2 M L1 M D1r1 = Mr1 .

11

While this result was derived for a system of lenses, it is valid for any optical system, since each optical element has a transformation matrix that is linear and of the correct dimension. Thus, we have in general:
r r rn = Mr1 ,

where
!M = M M M ...M M . n n"1 n"2 2 1

And we see that any system of optical elements can be reduced to a single transformation matrix
! acting on the incident ray. As the measurement system is composed of lenses and mirrors, this

method of analysis is extremely useful. One result that is especially relevant to this experiment and merits further discussion is the geometric magnification of an image viewed through a lens. Recall that the location of the image of an object viewed through a lens (or a curved mirror) can be found according to
1 1 1 + = , s s" f

where s is the distance from the object being viewed to the lens, s is the distance from the lens
! to the image, and f is the focal length of the lens. It is clear from this expression that the image

will be in focus if a screen is placed a distance s away from the lens. However, the size of the image will vary geometrically as the screen is moved away from the focus in either direction (see figure 6). This effect is relevant to the ROC measurement system, as it is comprised mainly of lenses and mirrors.

12

Figure 6: Geometric magnification effect due to a thin converging lens.

2.2 Diffraction Theory Although we have thus far largely ignored the wave nature of light, we must take it into account if we are to understand the diffraction effect found in the ROC measurement system. The most commonly cited example of diffraction is that of a narrow slit under the illumination of monochromatic light. If a screen is placed a large distance away from the slit, a pattern of light and dark spots is observed (see figure 7). This effect is clearly not geometric, as a purely geometric analysis would predict that the image formed on the screen would simply be that of the slit.

13

Figure 7: Diffraction pattern produced by a single slit illuminated by monochromatic plane waves.9

There are two distinct diffraction regimes, each with a different mathematical treatment. The pattern depicted in figure 7 arises due to Fraunhofer, or far-field, diffraction and bears little resemblance to the aperture being imaged. The other type of diffraction is known as Fresnel, or near-field, diffraction. As is noted in Hechts text on optics, as long as the incoming and outgoing waves approach being planar (differing by only a small fraction of a wavelength) over the extent of the diffracting aperture (or obstacles), Fraunhofer diffraction obtains.10 Otherwise, one must apply the more mathematically complex Fresnel diffraction theory. Since the measurement system utilized in this experiment satisfies the Fraunhoffer condition, we will focus on that treatment.

10

Image from: http://upload.wikimedia.org/wikipedia/commons/e/e1/Diffraction1.png Hecht, 448. 14

In our measurement setup, a thin filament is used as a target object. However, that diffraction pattern generated when the filament is illuminated by a monochromatic plane wave is nearly identical to the conceptually simpler single slit, with the only difference being in the zeroth order mode. Therefore, let us consider a long, narrow slit in the x-y plane, of width b in the y direction (see figure 8).

Figure 8: Single slit diffraction.

Since the slit is long and narrow, we can treat the problem as one dimensional in the y-direction. We wish to know the optical field a distance R>>b away from this line source. To find this, we apply the Huygens-Frensel Principle, which states that any unobstructed point on a wavefront can be thought of as a secondary source of spherical wavelets. The amplitude of the field at a farther point can then be thought of as the superposition of the waves emanating from the point sources on the original wavefront.11 For the purposes of our example, we can think of the line

11

Hecht, 444. 15

source as being made up of N point sources, each emitting a spherical wave. Conceptually then, it is not surprising that a diffraction pattern like the one in figure 5 results, as these spherical waves will interfere with each other. Based on this analysis, the field due to a differential section of the line source is

dE =
where

"L sin(#t $ kr) dy , R

"L =

1 lim ("0 N ) , b N #$

is the source strength per unit length. We can then expand r to make it an explicit function of y.

!
r = R " y sin # +

y2 cos2 # + ... . 2R

We can ignore the higher order terms, even for large y, as long as R is sufficiently big so as to make
! y2 " 0. 2R

This is an explicit statement of the Fraunhofer condition, which holds whenever r can be
! considered to be linear in y.12 Using this, the field due to one differential segment of our line

source can be written

dE =

"L sin(#t $ k ( R $ y sin % )) dy . R

And we can now integrate to find the total field

E=

"L R

&

b /2 $b / 2

sin(#t $ k ( R $ y sin % )) dy ,

which gives

!
12

Hecht, 452. 16

E=

"L b sin[( kb /2) sin # ] sin($t % kR) . R ( kb /2) sin #

If we introduce a new term


!

"=

kb sin # , 2

we can express the field as

E=

"L b sin # sin($t % kR) . R #

While this result is correct, it is not immediately useful. We are more interested in the
! irradiance, as that is the quantity that is most easily measured. Recall that

I " E2 , T
which is the time average of the square of the field. This operation is defined as13

f (t )

"

1 T

t +T t

f ( t #) dt # .

Calculating the irradiance, we have

$ sin # ' 2 I (" ) = I (0)& ) , % # (

where we recall that is a function of . A graph of this result is plotted in figure 9. The
! similarity to the diffraction pattern depicted in figure 7 is immediately apparent.

13

Hecht, 387. 17

Figure 9: Relative intensity as a function of .

We are often interested in knowing the locations of the subsidiary minima in a diffraction pattern. These are easily found by taking the derivative of I with respect to ,

2sin " (" cos " # sin " ) dI = I (0) . d" "3
This expression will be zero when sin " = 0 , which means

!
and when
! !

" = m# , m = 1,2,3...,

" cos " # sin " = 0 , tan " = " .

The solutions to the transcendental equation give the local maxima of I, while the condition on ! ! sin " gives the minima. If we recall the definition of , we can rewrite the minima condition as

18

" = m# =

kb 2#b sin $ m = sin $ m , 2 2%

which gives the commonly known formal for locating the diffraction minima,

bsin " m = m# .

This result can also be found by considering the phase differences arising from the relative path
! length differences (i.e. interference effects) of the modeled point sources on the y-axis.

Another elegant framework for considering Fraunhofer diffraction is that of the Fourier transform. Recall that the Fourier transform of a function in one dimension is given by

F (k) =

# "#

f ( x )e ikx dx ,

from which we can recover the original function using the inverse transform

f ( x) =

1 2"

$ #$

F ( k )e#ikx dk .

The Fourier transform and its inverse can easily be expanded to two dimensions according to

F ( kx ,k y ) =

f ( x, y ) =

1 2 (2" )

$ $ f ( x, y )e ( % % F (k ,k )e
"# "#

i kx x +ky y )

dxdy ,

#i( kx x +ky y )

#$

#$

dk x dk y .

! If we now consider a monochromatic plane wave incident on an arbitrary diffracting aperture in

! the x-y plane, the optical disturbance at a point on a screen a distance z away is given by14
E ( X,Y ) =

"A e i(#t$kR ) ik( Xx +Yx ) / R e dxdy , %% R Aperture

where A is the source strength per unit area, and X and Y are the coordinates of the point of observation on the! screen (see figure 10).

14

Hecht, 539. 19

Figure 10: Coordinate system used in the Fourier treatment of diffraction due to an arbitrary aperture.

This expression arises by considering each unobstructed point on the aperture as a point source of spherical waves and then integrating to find the total field. The leading 1/R term corresponds to the drop off of the optical field with distance from the aperture. However, if we only consider a small region in on the screen, R is essentially constant and we can redefine the first term inside the integral as
A( x, y ) = A0 ( x, y )e i" ( x,y ) .

This expression is known as the aperture function. The A0 term describes the amplitude of the
! field at a given point (x,y) on the aperture. For a monochromatic plane wave in a uniform

medium, such as air, this value will be a constant. The term exp[i(x,y)] describes the phase of

20

the wave at a given point on the aperture. So, we can rewrite the expression for the field at the screen as

E ( X,Y ) =

$ $
"#

# "#

A( x, y )e ik( Xx +Yy ) / R dxdy ,

where the limits of integration have been added because the function A is only nonzero over the

! dimensions of the physical aperture. If we now define the spatial frequencies of the wave as
kx " kX , R ky " kY , R

the diffracted field can be expressed as


!

E ( kx ,k y ) =

$ $
"#

A( x, y )e ( "#

i kx x +ky y )

dxdy .

When written in this form, it is easy to see that the optical field at a screen a distance away from

! an aperture under the Fraunhofer condition is merely the Fourier transform of the field
distribution across the aperture, which is expressed in the aperture function. As an illustration of this property, let us revisit the example of the single slit, of width b, illuminated by a monochromatic plane wave. Since the slit is infinitely wide in the y direction, we can treat this as a 1-dimensional problem. Assuming a perfect plane wave, with no phase or amplitude variations across the slit, we can represent the aperture function (see figure 11a) as

# A for b /2 " z " b /2 A( y ) = $ 0 . % 0 otherwise


Taking the Fourier transform of this aperture function, we can easily compute the field at a point on the screen.

!
E ( ky ) = A0 #
b /2 "b / 2

e ikY y dy .

This gives

21

" sin( k b /2) % y ' E ( ky ) = A0b$ $ k b /2 ' . y # &

From our definition of kY we can show that


!
ky = k sin " ,

where is defined in figures 8 and 10. Substituting this value into the expression for the field, we have
!

# sin( kb sin " ) & # sin ) & E (" ) = A0b% kb 2 ( = A0b% (. $ ) ' $ 2 sin " '

Finally, computing the irradiance, we arrive at the result


!
$ sin # ' 2 $ sin # ' 2 I (" ) = A b & ) = I (0)& ) , % # ( % # (
2 2 0

which is the exact result obtained using the Huygens-Fresnel Principle.


!

x Figure 11a: Aperture function for the single slit.

22

Figure 11b: Field distribution resulting from the Fourier transform of the aperture function.

This discussion of diffraction as a Fourier transform is especially useful in understanding an important feature of the ROC measurement system. It is clear from the above analysis that the optical field distribution of the diffracted aperture is the spatial-frequency spectrum of the aperture function. Moreover, the original aperture function can be reconstructed simply by taking the inverse transform of the optical field. This can be done physically by a lens, which also acts as a Fourier transformer, or by a spherical mirror.15 Since the inverse transform is taken over all values of k in each dimension, all the information of the original aperture function (which is distributed throughout the field) is captured. However, a problem arises when the optical instrument used as a Fourier transformer (the FEL cavity mirror in our case), is not capable of capturing all the diffraction modes. In this case, there will be a spatial intensity variation in the re-formed image associated with the spatial-frequency content not captured by

15

Hecht, 524. 23

the mirror (see figure 12). This effect is of real practical significance in our system, as the FEL cavity mirrors have a finite diameter.

Figure 12: Intensity profile of re-formed aperture function with 3 diffraction modes captured. The original aperture function is superimposed in red.

2.3 Gaussian Optics In our analysis up to this point we have made an important assumption, namely that light is propagated as a monochromatic plane wave. However, this neglects the fact that beam generated by the He-Ne laser beam in the measurement system has a Gaussian profile. Beams of this type are known as Gaussian beams, and have special properties that must be taken into

24

account if the physics of the ROC measurement system are to be fully understood. It is important to note that what we refer to as a Gaussian beam is really the lowest, or principle, mode in a family of infinite solutions to the paraxial wave equation.16 However, since most lasers operate on the principle mode, we will focus on this as we explore the physical characteristics of Gaussian optics. The lowest order Gaussian beam can be described by two parameters, its width, w, and the radius of the wavefront, R. A beam will always have a location where the width of the beam is minimized, known as the waist, w0. At this point, the wavefront is a plane wave, with R = , and the beam diverges in either direction along the optical axis from this waist, which we conventionally define to be z = 0 (see figure 13).17

Figure 13: A typical Gaussian beam. The red lines represent the width of the beam as a function of z.

If the waist size is known (and we have defined its location to be z = 0), then the normalized field at any other plane along the z-axis will be given by

E ( x, y,z) =
16 17

* $ 2 q0 x 2 + y 2 'exp,#ik& z + )/ , " w 0q( z) 2q( z) (/ , % + .

Siegman, 642. Siegman, 663. ! 25

where q( z) is the complex radius of curvature.18 This quantity is related to the beam width and
radius of curvature by

1 1 # = "i 2 , q( z) R( z) $w ( z)
and obeys the propagation law

q( z) = q0 + z = z + iz R ,

in free space. Here, we have introduced the quantity

!
zR =

2 "w 0 , #

which is known as the Rayleigh range. The Rayleigh range is the distance from the waist where
! the Gaussian wavefront exhibits the greatest variation.19 Referring to figure 13, the confocal

parameter of the beam, which is the region where the beam is considered to be collimated, is given by
b = 2zR .

With these definitions, we can describe the width and radius of a Gaussian beam according to
!

w ( z) = w 0

" z %2 1+ $ ' # zR &

( " z %2+ R( z) = z*1+ $ R ' ) # z&,


with

!
w0 =

"zR . #

!
18 19

Siegman, 664. Meschede, 48. 26

It is clear from this result that the properties of a Gaussian beam at any location along the optical axis are completely determined by a single parameter w0 (or equally well by zR) and the wavelenth of the light. The above equations describe how to propagate a Gaussian beam through free space. The next step is to examine the effect of a lens on the beam characteristics. Recalling from geometric optics that an ideal thin lens will alter the direction of propagation of an incoming ray while leaving its position in the lens plane unaffected, it is unsurprising that a lens will have a similar effect on a Gaussian beam. Immediately after passing through a thin lens, the beam width remains unchanged. The radius of curvature is transformed according to
1 1 1 = " , R2 R1 f

where R1 is the radius before the lens, R2 is the radius after the lens, and f is the focal length.
! The end result of this transformation is an altered beam with a new associated Rayleigh range,

which must be calculated following passage through a lens. Consider as an example a Gaussian beam passed through a single lens, as shown in figure 14.

Figure 14: Example lens transformation of Gaussian beam.

27

If the incident beam has wavelength and waist w01 (with associated Rayleigh range zR1), then the width and radius of curvature at the lens interface can easily be found according to
% "z1 ( 2 #zR1 w1 = w ("z1 ) = 1+ ' * , $ & zR1 ) ) # &2, z R1 = R("z1 ) = "z1+1+ % R1 ( . . + $ "z1 ' . * -

Applying the lens transform


!
w1 = w 2 , fR1 R2 = , f " R1 !

we have

! % $z1 ( 2 % $z2 ( 2 "zR1 "zR 2 1+ ' * = 1+ ' * , # # & zR1 ) & zR 2 )

and
! ) # &2, z f"z1+1+ % R1 ( . 2, ) # + . zR 2 & . * $ "z1 ' = "z 2 +1+ % ( . ) # &2, + $ "z 2 ' . zR1 . * f / "z1+1+ % ( + $ "z1 ' . * -

Solving these two equations for the new Rayleigh range, zR2, and the location of the new waist
! plane, z2, we can describe the properties of the transformed beam and propagate it further in

free space. This process can be repeated for multiple lenses, a process which lends itself well to the matrix method detailed in the geometric optics section. However, to describe a Gaussian beam, a more complex matrix is required than in the simple geometric case, and the discussion of this method is beyond the scope of this paper.

28

While it is a fairly simple matter to propagate a Gaussian beam in free space and through a lens, the effect of an aperture merits further examination. About 86.5% of the power of a Gaussian beam is contained within the width of the beam, w, and the intensity falls off rapidly with r beyond that value.20 However, an aperture of size w will still clip a significant portion of the beams edges. This is of real practical importance when creating an optical system, as lenses and other elements of finite size can cause clipping in much the same manner as an aperture. To investigate this effect, recall the total power in an optical beam is given by
Ptotal " ## u
2

dA ,

all space

where we are integrating over the cross sectional area of the beam. Using this formula, if we
! have an circular aperture of radius b, centered along the optical axis of a Gaussian beam of width

w, the power transmitted through the aperture will be


Ptransmitted 2 = Ptotal "w 2

b 0

2"re#2r

/w2

dr = 1# e#2b

/w2

From this formula, we can see that an aperture of size w will only transmit approximately 86% of the power in the! incident beam, a significant loss. However, an aperture with radius

d=

" w # 1.571w 2

will transmit over 99% of the power in the beam.21

! In addition to power reduction, an aperture can also produce significant diffraction


effects, even if only a small portion of the incident wave is cut off. However, in the case of a single slit, the resultant diffraction pattern differs from that of a uniform plane wave in that the

20

Rdiger Paschotta, Gaussian beams, laser beams, fundamental transverse modes, Encyclopedia of Laser Physics and Technology, 9 May 2009 <http://www.rpphotonics.com/gaussian_beams.html> 21 Siegman, 666. 29

intensity drops off when going off-axis. This is because the effective aperture function is not uniform, but rather varies as a Gaussian function. This effect is explored to a great extent to benefit the ROC measurement system. Decreased intensity of the higher order diffraction modes means less information will be lost when only the lower orders are collected and reflected by the mirror. The result is a reduction in the diffraction effects in the re-formed image when compared to the plane wave example.

3. Simulation Method
3.1 Description of OPC Simulation Code In order to explore the physics of the ROC measurement system and to confirm the theory of our experimental method, a series of computer simulations were created using the Optical Propagation Code (OPC) written by the Laser Physics and Non-Linear Optics group at the university of Twente.22 The code is designed to simulate the behavior of both plane waves and Gaussian beams in a user designed optical system. The optical field is simulated on a grid of finite size, allowing for the adjustment of the spacing between grid points to determine the resolution of the simulation. The user is able to completely define the initial properties of the field by specifying such parameters as wavelength and beam waist size (in the case of a Gaussian beam). The simulated optical beam can be propagated in free space using either Fourier transforms or the Fresnel integral. The code can also simulate the effect a variety of optical instruments, including lenses, mirrors, and apertures on the beam. Finally, the code can generate outputs with the optical field values at any point in the simulated system. These outputs can be read and analyzed using Matlab. Because both free space propagation methods fully treat the

22

The latest version of the OPC code can be downloaded at http://lpno.tnw.utwente.nl/opc.html 30

wave properties of the field, the code accurately reproduces the diffraction effects seen in physical systems. 3.2 Simulation Examples The first simulation is a simple example to confirm the codes ability to accurately reproduce the effects of a Gaussian beam. A simple, 2-lens system is set up as shown in figure 15.

Figure 15: Gaussian beam propagation through 2-lens system.

A Gaussian beam with w 01 = 1.25 "10#4 m and " = 1.064 #10$6 m is generated at z = 0 as shown. The beam is propagated a distance z1 = 0.2 m where a lens (f = 0.1 m) transform is applied.
! ! The beam is then propagated another 0.52 m before a second lens (f = 0.25 m) transform is

applied. Finally, the beam is propagated to the theoretically calculated waist location after the second lens. The goal of the simulation is to show that the waist sizes and locations are as expected after each lens. With the given initial parameters of the beam

w 01 = 1.25 "10#4 m , " = 1.064 #10$6 m ,


we can easily calculate the beam characteristics at the interface of the first lens ! !

31

" 0.2m % 2 )4 w (0.2m) = w 01 1+ $ ' = 5.561(10 m , zR1 & # " " z %2 % R(0.2m) = (0.2m)$1+ $ R1 ' ' = 0.21064 m . $ # 0.2m & ' # & ! Applying the lens transform
!
w 2 = w1, 1 1 1 = " , R2 R1 f

! we have the new waist location and associated zR

"z2 = 0.18245 m , zR 2 = 0.03839 m ,

which gives a new waist size of ! !

w 02 = 1.135 "10#4 m .

These values completely determine the properties of the transformed beam. If we now propagate

! the beam to the second lens and repeat the transformation, we arrive at the final beam parameters
"z4 = 0.8486 m , zR 3 = 0.2581 m ,

and a final waist size of

! !

w 03 = 2.957 "10#4 m .

A sample of the graphical results appears below in figures 16a-16c. For each cross section, and intensity profile! generated and fit with a Gaussian function using Matlab, giving was a result for the beam with. These results appear in Table 1.

32

Figure 16a: Incident Gaussian beam generated by OPC code fitted using Matlab.

33

Figure 16b: Beam profile after passing after second lens transformation.

34

Figure 16c: Beam profile at the theoretical location of the final waist plane.

Position z=0 before lens 1 after lens 1 z2 z2-zR2 z2+zR2 before lens 2 after lens 2 z4 z4-zR3 z4-zR3

Predicted Beam Width (mm) 0.125 0.5561 0.5561 0.1135 0.1605 0.1605 1.0149 1.0149 0.2957 0.4182 0.4182

Measured Beam Width (mm) 0.125 0.55605 0.55605 0.11355 0.16032 0.16056 1.0144 1.0144 0.29588 0.41745 0.41852

Difference (%) 0 0.0090 0.0090 0.044 0.11 0.037 0.049 0.049 0.061 0.18 0.077

Table 1: Simulation results for the two-lens system compared with theoretical expectations.

35

These results have several important features. The simulated beam widths are well within one percent of the theoretically predicted values. The width of the beam immediately before and after a simulated lens is unchanged. Finally, the spot size one zR on either side of each waist is essentially the same, and equal to
w ( z = zR ) = 2w 0 ,

indicating that the beam waists are at the expected locations. These features demonstrate the
! ability of the OPC code to accurately reproduce the essential physics of Gaussian beams.

For the second simulation, the OPC code was used to construct an asymmetrical laser cavity. A laser cavity consists of two spherical mirrors separated by a distance d as shown in figure 17.

Figure 17: An asymmetrical laser cavity.

A Gaussian beam is injected into the cavity and reflected continually between the two mirrors. In a symmetric cavity, the mirrors have the same radius of curvature, and the beam waist is located at d/2. In an asymmetric cavity, the mirrors have different radii and the beam waist is located at a different location away from the center of the cavity. In order for a stable system to arise, the Gaussian beam parameters must be matched to physical of properties the cavity. The

36

goal of this simulation is to show that a beam matched to the cavity parameters will produce a stable waist at the desired location. Consider the cavity shown in figure 17. Suppose we require z1 = 0.3 m and z2 = 0.7 m, for a total cavity length of 1 m. We inject a Gaussian beam with w 0 = 2.85 "10#4 m and

" = 6.33 #10$7 m at z = 0 and propagate it 0.3 m to the first mirror. Since we require the waist of
! the reflected beam to be at z = 0, the mirror transformation equations dictate
!
w 2 = w1, R2 = "R1.

Recalling that the beam radius transforms as ! ! 1


R2

1 2 " , R1 RM

where RM is the radius of curvature of the mirror. Applying the above requirements, we can
! easily see that we need a mirror that satisfies at z = 0.3 m
RM = R1 = "R2 .

Plugging in for w0 and , we have


! RM 1 = 0.8416 m .

To find r2, the same analysis is used, with z = 0.7 m. This gives
! RM 2 = 0.9321 m .

To test this analysis, the simulation propagates the beam between the two mirrors, taking data
! each time the expected waist plane is reached. Each image is then analyzed and fitted using

Matlab. A sample of the graphical results appears in figure 18, and the tabulated results appear in Table 2.

37

Figure 18: Example simulation output fitted with Gaussian curve.

Iteration 1 2 3 4 5 6 7 8

Waist Size (mm) 0.28498 0.28501 0.28499 0.28502 0.28498 0.28503 0.28497 0.28505

Difference (%) 0.0070 0.0035 0.0035 0.0070 0.0070 0.011 0.011 0.018

Table 2: Simulation results for stable laser cavity simulation. The expected waist size is 0.285mm.

The result of this simulation demonstrates the stability of this laser cavity and once again demonstrates the ability of the OPC code to accurately simulate the physics of Gaussian beams.

38

4. ROC Measurements of FEL Cavity Mirrors


4.1 Experimental Setup The ROC measurement system is based on the geometric imaging principle for a spherical mirror. As discussed in the theoretical framework section of this paper, the location of the image of an object placed a distance s from a mirror of radius R can be found using the thin lens equation

1 1 2 + =# , s s" R
where s is the distance from the mirror to the image plane. If both s and s are known, then it is

! a simple matter to calculate the radius of curvature of the mirror from this formula. Our setup is
shown in figure 19 below.

Figure 19: ROC measurement system. Distances not to scale.

In this system, a He-Ne laser beam is used as a light source. It generates a highly coherent, nearly monochromatic beam with a Gaussian profile. The beam is then collimated using a pair of lenses, the second of which is focused on the mirror. A thin filament is used as a target object and is mounted on a movable stage, along with a high quality CCD camera. The collimated beam strikes the filament before traveling approximately 28 meters to the mirror (this

39

distance varies slightly depending on the position of the stage). The beam is then reflected back to the CCD camera, where an image is formed. Several important physical properties of this system merit further explanation. Because the beam is essentially monochromatic and has a high degree of spatial coherence, the collimated beam approaches a plane wave across the extent of target object. Coupled with the fact that the size of the target is significantly smaller than the distance to the mirror, and it is clear that there will be non-negligible diffraction effects in the system. This poses a practical limitation, as the intensity of the incident beam is distributed over a large spatial region at the mirror. A typical cavity mirror has a diameter of ~5 cm, which is large enough to capture only ~ 3 diffraction orders. Therefore, there will be information missing when the image is re-formed on the camera and diffraction effects will persist as was demonstrated in the theoretical background section of this paper. It should be noted that these diffraction effects will be mitigated somewhat by the fact that the beam has a Gaussian profile. From this analysis, it is clear that there are three principle physical effects at work in the system. The first is the geometric imaging principle, which dictates that the image size, or more importantly the imaged width of the target, will increase linearly with distance from the mirror. The second is the diffraction effect introduced to the system by the use of a coherent light source. Because only a finite number of diffraction orders are captured by the mirror, the re-formed image will have internal intensity variations due to missing spatial frequency information. While these variations reduce the quality of the image, a sharp image is expected to be formed at the location of the geometric image plane if enough diffraction orders are captured. Finally, the use of a Gaussian beam reduces the number of diffraction orders that must be captured in order to form a sharp image by decreasing the amplitudes of the higher order diffraction modes. The

40

combined effects of these factors lead us to expect the image of the target filament will be sharpness at the image plane. Measurement of the location of the best focused image is facilitated by the CCD camera. The stage is moved to its farthest position from the mirror and an image is recorded and stored for analysis. The stage is moved forward and another image is taken and stored. This procedure is repeated until the camera has scanned a 40 cm distance. Matlab is used to analyze the image taken for each camera location. This is accomplished by summing the CCD pixel values to create an intensity profile of the image. However, due to the diffraction effects in the system, it is not immediately clear how to measure the width of the image to determine the location of the image plane. In order to deal with this difficulty, the derivative of the plot is taken in order to measure the sharpness of the edge of the image. The sharpness of each image is then plotted as a function of camera position and a quadratic curve fit is applied. The location of the maximum of this curve is taken to be the location of the sharpest image, and therefore the location of the image plane. This position is then used to determine the mirror ROC according to the mirror equation derived from the geometric imaging principle. Much work has been done to determine the optimal target object for use with this method. In the original version of the measurement system utilizing a white light source, a wire mesh was used. This proved to be an effective choice, since diffraction effects were negligible. However, the use of a coherent source requires a simpler target to simplify the diffraction problem. Initially, a single slit was used. This proved to be non-optimal, as a large portion of the beam was blocked. The slit was replaced by a thin filament. However, this presented problems as well, as the filament was initially too narrow. The resulting diffraction pattern was so spread out that most mirrors were only able to capture 1-2 orders. A wider filament ( 3 mm)

41

also proved problematic, as the image position on the CCD camera shifts as the camera is moved. A large filament limits the range the camera is able to scan before a portion of the image lies off the CCD. A target filament with a width of 1.6 mm is a compromise between these two extremes. The filament is large enough to allow for the capture of 3 diffraction orders, while still small enough to allow the camera to scan a significant distance. 4.2 Measurement Results Since the upgrade began several years ago, the ROC measurement system has been used to measure the ROC of numerous FEL cavity mirrors. These measurements can be divided up into two classes, those with a wider filament and those with a more narrow target. Measurements of the first class were conducted during the initial phases of the upgrade, and represent the closest correlation to the physics discussed previously and depicted in the simulations described in the next section. Due to the sheer number of measurements taken, it is not practical to reproduce them all here. A sample set of measurement images from this class, taken from a mirror substrate with a 4 inch diameter, as well as analyzed intensity profiles are given in figures 21a-21b, and 22.

42

Figure 21a: Sample CCD capture, intensity profile and derivative located before the image plane.

43

Figure 21b: Sample CCD capture, intensity profile and derivative located at approximately the image plane. The maximal values of the derivative show significant sharpening of the image when compared with figure 21a.

44

Figure 21c: Sample CCD capture, intensity profile and derivative located after the image plane. The maximal values of the derivative are closer to those of figure 21a, indicating that the image is not as sharp as the one in figure 21b.

In figure 22, the intensity profiles of images from several camera locations, as well as their derivatives, are plotted for comparison.

45

Figure 22: Intensity profiles and computed derivatives for several camera locations plotted together for comparison.

46

From these results, we can clearly see how image sharpness varies with distance. The intensity plot suggests that the image at z = 28 cm (which corresponds to a distance of 28.09 0.28 = 27.81 m from the mirror) has the greatest steepness. This observation is born out when we examine the derivative, which offers a more quantitative measure of sharpness. From the graph, it is clear that the z = 0.28 m position is considerably sharper than either the z = 0.20 m or the z = 0.36 m positions. In order to more exactly find the location of the image plane from this data, the values for steepness found from the derivatives are plotted and a second order polynomial curve fit is applied. This technique can be seen in the second example data run below. As mentioned above, a second class of measurements, including the ones taken during the 2008-2009 academic year, were undertaken using a smaller filament. We know from observation that the resulting diffraction pattern in the mirror plane is characterized by greater spacing between diffraction minima. As the result of changing to a narrower filament, the mirror is now only capable of capturing up to the first diffraction order. We expect, then, that the system will not be able to form an image on the CCD. However, experimental observation shows that the image plane can still be located through the aforementioned measurement technique. A sample of the measurements taken using the narrower target is reproduced here in table 3. The raw CCD images are shown in figures 23a-23c, and the graphed results for the selected data runs are then shown in figures 24a-24f23. Two measurements are taken for each mirror, one with the mirror initially set in a default orientation using a marked line on the edge of the mirror (R1), and the second with the mirror rotated 90 about the optical axis (R2). These measurements give values for the ROC in two perpendicular planes, thereby providing some
23

Jingyi Li, Radius of Curvature and Transmission Measurements of 780 nm Mirrors (Durham, NC: DFELL, 2008), 2-4. 47

indication of the uniformity of the mirrors curvature. For each mirror, the ROC specified by the

#+0.4 & manufacturer is 27.46 % ( m, and the control wavelength is 780 nm. $ "0.2'
Mirror R1 (m) 28.320.09 28.440.17 27.730.10 27.660.10 27.550.13 27.900.08 R2 (m) 28.380.11 28.330.14 27.660.10 27.690.10 27.640.10 27.970.10 Within Manufacturer Specification? No No Yes Yes Yes No

! CCV014 CCV054 CCV077 CCV078 CCV080 CCV081

Table 3: Measured Radii of selected 780 nm mirrors.

Figure 23a: Sample CCD capture before the image plane.

48

Figure 23b: Sample CCD capture at approximately the image plane. The imaged filament is sharper than in the first image and is closer to the center of the CCD.

49

Figure 23c: Sample CCD capture after the image plane. The image has shifted further on the CCD and is not as sharp as the second image.

Figure 24a: ROC measurements for CCV014.

50

Figure 24b: ROC measurements for CCV054.

Figure 24c: ROC measurements for CCV077.

51

Figure 24d: ROC measurements for CCV078.

Figure 24e: ROC measurements for CCV080.

52

Figure 24f: ROC measurements for CCV081.

It is important to note that the technique for analyzing the data generated from the smaller filament setup is different from that employed in the earlier measurements. Because less than three diffraction orders have been captured, it is not possible to determine the location of the image plane from the sharpness information contained in the derivative of the intensity profile. Instead the location of the first order diffraction maxima of the reformed image, which is believed to be narrowest at the image plane, is measured and plotted. A quadratic fit is applied, and the minimum taken to be the location of the image plane. While the physics behind this method is not completely understood, the results indicate the system is still capable of determining the location of the image plane to the desired precision. A more detailed discussion follows in section 4.4. 4.3 Simulation Results In order to study the physics behind the measurement system, a pair simulations were created. The first is designed to examine only the diffraction effects in the system, while the second explores the effect of a Gaussian beam on the diffraction problem. Since the OPC code is

53

not capable of modeling a thin filament, a narrow slit is used instead. The first simulation is setup as shown in figure 25.

Figure 25: Plane wave simulation setup.

A monochromatic plane wave is generated and then propagated 0.28 m where a slit of width 2 mm is applied. The beam is then propagated 6 m, where it strikes a mirror with a radius of curvature of 6 m. At the mirror interface, a circular aperture is applied to simulate a mirror of finite size. The beam is reflected back and cross sections of the field are taken every 3 cm in the range of 5.94 - 6.06 m from the mirror. This simulates the camera movement in the physical measurement system and allows us to view the effect of diffraction on the imaged slit width. As was discussed in the theoretical framework section of this paper, geometric optics predicts that the measured width of the reflected image will increase linearly with distance from the mirror. However, because the mirror is of finite size, information will be missing when the reflected image is formed and significant diffraction effects will still be present, overwhelming the geometric effect and distorting the image. To get a good measure of the location of the image plane, the simulation examines the sharpness of the intensity profile at each camera location. We expect the most sharply focused image of the slit to be a distance of 6 m away from the mirror. It is important to note, that the above analysis is only useful if enough

54

information is captured at the mirror to reform the image of the slit. If a low number of diffraction orders are captured, it becomes impossible to form a discernable image. For this simulation, three data runs were undertaken. In the first run, the aperture was sized to allow the mirror capture up to the fourth diffraction order. In the second, the first 8 diffraction orders were captured. Finally, in the last simulation, 12 diffraction orders were reflected back to the camera. In each case, the simulations outputs were analyzed using Matlab and an intensity profile, drawn from the data from all five camera positions, was compiled. These results appear in figures 26a-26d.

Figure 26a: Example simulation output.

55

Figure 26b: Zoomed in view of the right corner 4-diffraction orders run. Little differentiation can be seen as insufficient information has been captured.

56

Figure 26c: Zoomed in view of the right corner for 8-diffraction orders run. The slit width at 6m appears to be the narrowest, though it is still difficult to differentiate.

57

Figure 26d: Zoomed in view of the 12-diffraction orders run. The slit width at 6 m is clearly the narrowest.

From these results, it is apparent that a mirror must be large enough to capture at least 6-8 diffraction orders if an image is to be formed when the slit is illuminated by a monochromatic plane wave. A more quantitative analysis follows the results of the next simulation. In the second simulation, the plane wave is replaced by a Gaussian beam, and a lens is added to focus the beam on the mirror, a feature present in the physical measurement system. The system is set up as shown in figure 27.

58

Figure 27: Gaussian beam simulation setup.

A Gaussian distribution ( " = 6.33 #10$7 m, w 0 = 0.01 m ) is generated and then propagated 0.28 m where lens with focal length f = 8 m is applied. This has the effect of placing the waist of the

! beam at the mirror interface, a distance 8 m away. The beam is transmitted a distance of 2 m
berfore a slit of width 2 mm is applied. After being propagated another 6 m, the beam strikes the mirror with a radius of curvature of 6 m. At the mirror interface, a circular aperture is applied to simulate a mirror of finite size. The beam is reflected back and cross sections of the field are taken every 6 cm in the range of 5.88 - 6.12 m from the mirror. This simulates the camera movement in the physical measurement system and allows us to view the effect of diffraction, combined with the properties of the Gaussian beam, on the imaged slit width. A series of simulations are conducted for Gaussian case, with 3, 4, 5, and 6 diffraction orders being captured. The outputs at z = 6 m away from the mirror were compiled using Matlab and compared with the same outputs for the plane wave. The result can be seen in figures 28a and 28b.

59

Figure 28a: Compiled results for 3-6 diffraction orders captured for both Gaussian and plane wave illumination.

60

Figure 28b: Detailed view of the right edge of figure 21a. Noticeable sharpening can be seen in the Gaussian outputs when compared with the plane wave outputs of the same order.

These results show the image formed to be qualitatively sharper when a Gaussian beam is used in lieu of plane wave illumination. However, a more quantitative analysis is required if these simulations are to confirm the validity of the ROC measurement technique. To get a numerical value for steepness, the simulation outputs for both the plane wave and Gaussian beam simulations are fitted with exponential curves on both the left and right edges of the image. The curve has the form

f ( x ) = a3e

" ( x"a1 ) a2

+ a4 ,

where ai denotes a fitting constant. Example fitted edges are shown in figures 29a-29d. These

! results are taken from the outputs of both simulations with 6 diffraction orders captured.

61

Figure 29a: Left edge fits for the plane wave simulation with 6 diffraction orders captured.

62

Figure 29b: Right edge fits for the plane wave simulation with 6 diffraction orders captured.

63

Figure 29c: Left edge fits for the Gaussian simulation with 6 diffraction orders captured.

64

Figure 29c: Right edge fits for the Gaussian simulation with 6 diffraction orders captured.

65

For these curve fits, the sharpness value is determined from the constant a2. These values are plotted as a function of camera position for the plane wave case in figure 30a, and for the Gaussian simulation in figure 30b.

66

Figure 30a: Steepness values for the plane wave simulation.

67

Figure 30b: Steepness values for the Gaussian simulation. The fit is sharper at all positions and there is a clear maximum at 6 m.

68

While these fits show a sharpening trend when a Gaussian beam is used instead of a plane wave, and that the image sharpness is greatest at the image plane, there is a second method that can be used for evaluating the simulation outputs. This is accomplished by examining the derivative of the intensity profiles to observe the sharpening through the slope directly. An example of this method is shown is figure 31, where the intensity distributions and their derivatives are shown. The figure is from a simulation using a Gaussian beam and a mirror capturing 6 diffraction orders.

69

Figure 31: Intensity profiles and computed derivatives for several simulated camera locations plotted together for comparison.

70

Like the curve fits, the derivatives show that the sharpness is greatest at the image plane. However, to offer a more quantitative description, the peaks of the derivative graph would need to be fit to determine the location of the maxima and minima. It is important to note that these simulations serve to illustrate an important limitation of the OPC code. In theory, if the aperture in front of the mirror were removed from the simulation, the mirror would have infinite extent and would capture all diffraction orders. The result would be the complete elimination of the diffraction effects from the reflected image, allowing us to view only the geometric properties. This would certainly be true if the simulation utilized an infinite plane to model the optical beam. However, in reality, the code simulates the optical field using a grid of finite size and spacing. To simulate an infinite plane, the code imposes periodic boundary conditions at the edges of the grid. This can introduce disruptive diffraction-like effects if the field has a value significantly greater than zero near the boundaries. However, even if the field is close to zero, the size of the infinitely large mirror is limited by the size of the grid. Thus, it is impossible to completely eliminate the diffraction effects from this simulation, as an infinitely large grid would be required. However, for the purposes of studying the effects of capturing only a low number of diffraction maxima, the finite grid is sufficient. 4.4 Discussion The above simulations give important physical insight into the measurement results of section 4.2. Firstly, it is clear that the FEL cavity mirror must capture a certain number of diffraction orders in order to form a sharp image on the CCD camera at the image plane of the geometric imaging. For a slit illuminated by a monochromatic plane wave, the simulations have shown that between 6 and 8 orders must be captured. However, the FEL cavity mirrors used in the physical measurement system are only capable of capturing 3 diffraction orders, making

71

them seemingly incapable of forming a sharp image on the CCD camera. The results of the second simulation show a definite sharpening trend when an incident beam with a Gaussian profile is used. The laser used in for ROC measurements generates a beam with a Gaussian distribution. Therefore, based on these simulations, we must conclude that the use of a laser beam as a source in the ROC measurement system reduces the diffraction effects. What is not immediately clear however, is why we are able to form a sharp image with only 3 diffraction orders in our measurement system when the simulations predict that 4 or more are needed. The goal of the simulations was to recreate the ROC measurement system on a smaller scale to avoid reflections arising from the finite grid size as described in the previous section. This being the case, we expect the number of captured diffraction orders required to form a sharp image to be the same in both the physical and simulated measurement systems. However, the results do not support our hypothesis. This result might be explained by the fact that the simulation does not exactly reproduce the physical system. Aside from the disparities between the dimensions of the physical and simulated systems, the primary difference between the two is that the simulation uses a slit in lieu of a thin filament. While this would have no effect under plane wave illumination, the use of a Gaussian beam might have an effect on the results. A more accurate simulation, which accurately reproduces the ROC measurement system, is needed in order to better understand this discrepancy. One aspect of the physical measurement results that merits further discussion is error. The reported error in reported in table 3 and in figures 21a-21f is purely statistical in origin. We believe the primary source of systematic error was the measurement of the dimensions of the experimental setup, more importantly the distances from the target to the mirror and from the mirror to the camera. Another source of systematic error arises from pressure waves due to

72

airflow in the linear accelerator tunnel, where the measurement system is housed. However, this was minimized by deactivating the ventilation fan before measurements and allowing the air in the tunnel to become still. A third source of systematic error is the possible non-uniformity of the index of refraction of the medium (air) between the slit and the mirror due to temperature gradients. This is difficult to quantify, and we believe housing the measurement setup in the tunnel, which is well-insulated, minimized any temperature variations. Of these three sources, we believe the error in distance measurement to be the most important. Yet, this source produces error on the order of millimeters at most. The statistical error given in section 4.2 is on the order of ten centimeters, two orders of magnitude greater than the systematic error. It is for this reason that the systematic error has been omitted from our experimental results. Another feature that warrants further discussion is the second class of measurements described in section 4.2. In this setup, a thinner filament was used, resulting in a diffraction pattern with greater spacing between maxima. In this case, the mirror is only able to capture the first diffraction order. In spite of our predictions, we are still able to form an image on the camera and determine the location of the image plane. However, this is done by examining the spacing between the first order diffraction maxima as a function of camera position, rather than by measuring the sharpness of the image. While any explanation for this result is mostly conjecture at this point, we believe that this technique is actually made possible by the fact that so few diffraction orders are captured by the mirror. This amplifies the spatial intensity variation of the image on the camera, allowing for the clear determination of the spacing between the first order maxima in the recorded images. It is also likely the lenses used in the optical system also reduce the diffraction effects in the system by modifying the beams coherence. This allows us to reform the image with fewer orders. The effect may also be present in the first class of

73

measurements as well. Further research into this feature is necessary if it is to be completely understood. In spite of this uncertainty, the ROC measurements have several other important characteristics. The first is reproducibility. For each measurement direction of each mirror, two data runs were performed. These show that the system gives consistent results when measurements are repeated. The results also show that this setup yields high-precision results. The average uncertainty on the measurements in section 4.2 is 0.32%. This achieves the goal of a system capable of measuring ROC to within 0.5%.

5. Conclusion and Future Research


The ROC measurement system has been used to measure numerous mirrors with control wavelengths ranging from the IR to the UV regimes. At this point, we believe the physics behind the systems operation is well understood. It is clear that the use of a Gaussian beam reduces the diffraction effect associated with a coherent, monochromatic source. The experimental results also show that the capture of the first 3 diffraction orders is sufficient to form a sharp image on the camera at the focus location. The measurements of FEL mirrors show that the system is accurate to within 10 cm on average. This corresponds to a less than 0.5% uncertainty, demonstrating the success of the new ROC measurement system in meeting design goals. As was discussed somewhat in the results section, there is a discrepancy between the number of captured diffraction orders needed to form an image in the simulation and the number needed in the ROC measurement system, which cannot be easily explained. Work is needed to develop a new simulation that more accurately depicts the actual measurement setup. The greatest obstacle to accomplishing this is grid size, which becomes a significant issue when the

74

drift space between the target and the mirror is expanded to the 27.5 meters of the physical system. Another challenge is to simulate a filament as a target. The OPC code cannot simulate this by default, and an addition to the code would need to be made. However, development of this simulation is essential if the experimental method is to be accurately modeled. Improvements can also be made to the ROC measurement system itself. Currently the system is capable of measuring the ROC of a mirror in one direction at a time. However, one goal of the measurements is to determine the uniformity of the mirrors in multiple directions. Currently this is accomplished by rotating the mirror 90 after the first data run to get a measurement for the perpendicular direction. This method gives a rough idea of the uniformity of the mirrors surface. However, the system could be improved by the addition of a mirror mount that could be easily rotated by a prescribed amount, possibly by a motor. Another option would be to implement a target object that would allow for the measurement of a mirrors ROC in multiple directions simultaneously. The challenge presented by this choice is that the diffraction problem becomes significantly more complicated to understand. However, either improvement would give a much more complete picture of the uniformity of a mirrors surface. As it stands now, however, the ROC measurement system is well qualified to measure a wide range of FEL cavity mirrors. The method has been shown to work consistently and to be capable of measuring a mirrors ROC to within 0.5%. In addition, the fundamental physics of the system is well understood. While detailed measurements of the uniformity of a mirrors surface remain difficult with the current apparatus, this problem could be dealt with in a future upgrade. The project goal of understanding the optimization of image formation through the control of diffraction effects has been successfully achieved, and the subsequent application of

75

this knowledge to improve the ROC measurement apparatus will benefit operations at DFELL for years to come.

6. Acknowledgments
I would like to thank Jingyi Li for his earlier work with the ROC measurement system. His assistance with the measurement and data analysis portion of this project has been invaluable and I am extremely grateful to him for all that he has done. I would also like to thank Professor Ying Wu for serving as my mentor and research advisor throughout the completion of this project. Without his help and support, I would never have been able to accomplish what I did. Finally, I would like to thank my defense committee for their role in helping me complete and revise this thesis.

76

7. Works Cited
Hecht, Eugene. Optics. San Francisco: Pearson Education, 2002. Lee, Yun Woo, et. al. Noncollimated bidirectional shearing interferometer for measuring a long radius of curvature. Applied Optics, 36.22 (1997): 5317-5320. Li, J., Y. K. Wu, C. Sun, Improved Long Radius of Curvature Measurement System for FEL Mirrors. Proc. of 2005 Particle Accelerator Conf., 16 May-20 May 2005, Knoxville, Tennessee: 1787-1789. Meschede, Dieter. Optics, Light and Lasers. Mrlenbach, Germany: Verlag-GmbH & Co., 2007. Paschotta, Rdiger. Gaussian beams, laser beams, fundamental transverse modes. Encyclopedia of Laser Physics and Technology. 9 May 2009 <http://www.rpphotonics.com/gaussian_beams.html>. Siegman, Anthony. Lasers. Mill Valley, CA: University Science Books, 1986.

77

A.1: OPC Code Used For Plane Wave Simulation


#!/usr/bin/perl # This script is created by # Agee Springer (ags10@duke.edu) # based on script written by Botao Jia (botaojia@fel.duke.edu) # April, 2009 #The script simulates the diffraction effect of a single slit illuminated by a monochromatic plane #wave and the image formation when a the diffracted light is reflected by a spherical mirror of #finite size. #The optics components set up is as the following: # 1. Generate a plane wave z=0; # 2. Apply a single slit aperture; # 3. Propagate the beam z=6 (m) to the mirror; # 4. Apply a mirror transform with R=6 (m); # 5. Propagate the beam an additional ~6 (m) back to the camera; #Screens are set up every 0.3m from z=5.88 m to z=6.12 m after the mirror to observe the image profile. use lib '/home/research1/appl/lib4k'; #if use public domain #use lib './lib'; #if use personal machine use Physics::OPC; #$VERBOSE = 1; $field = field init => 'plane', I => 1, lambda => 6.33e-7, npoints => 2048, mesh => 2e-5, nslices => 1; #first screen after slit $optics4=optics" slit d=0.0020 # slit size can be adjusted by user # rectangle dx=0.01 dy=0.0005 "; unlink 'after_slit.cross.txt'; unlink 'after_slit.dfl'; run $optics4; cross $field '>after_slit.cross.txt';

78

move $field => 'after_slit.dfl'; #propagate to mirror $optics5=optics" fresnel z=6 "; unlink 'before_mirror_cross.txt'; unlink 'before_mirror.dfl'; run $optics5; cross $field '>before_mirror.cross.txt'; move $field => 'before_mirror.dfl'; #screen after mirror $optics6=optics" diaphragm r=0.012 mirror r=6 ";

#can be used to define size of mirror

unlink 'after_mirror.cross.txt'; unlink 'after_mirror.dfl'; run $optics6; cross $field '>after_mirror.cross.txt'; move $field => 'after_mirror.dfl'; #propagate another ~6 m $optics7=optics" fresnel z=5.88 "; unlink 'at5.88m.txt'; unlink 'at5.88m.dfl'; run $optics7; cross $field '>at5.88m.txt'; move $field => 'at5.88m.dfl'; $optics8=optics" fresnel z=0.03 "; unlink 'at5.91m.txt'; unlink 'at5.91m.dfl'; run $optics8; cross $field '>at5.91m.txt'; move $field => 'at5.91m.dfl';

79

$optics9=optics" fresnel z=0.03 "; unlink 'at5.94m.txt'; unlink 'at5.94m.dfl'; run $optics9; cross $field '>at5.94m.txt'; move $field => 'at5.94m.dfl'; $optics10=optics" fresnel z=0.03 "; unlink 'at5.97m.txt'; unlink 'at5.97m.dfl'; run $optics10; cross $field '>at5.97m.txt'; move $field => 'at5.97m.dfl'; $optics11=optics" fresnel z=0.03 "; unlink 'at6.00m.txt'; unlink 'at6.00m.dfl'; run $optics11; cross $field '> at6.00m.txt'; move $field => 'at6.00m.dfl'; $optics12=optics" fresnel z=0.03 "; unlink 'at6.03m.txt'; unlink 'at6.03m.dfl'; run $optics12; cross $field '> at6.03m.txt'; move $field => 'at6.03m.dfl'; $optics13=optics" fresnel z=0.03 "; unlink 'at6.06m.txt'; unlink 'at6.06m.dfl'; run $optics13; cross $field '> at6.06m.txt'; move $field => 'at6.06m.dfl';

80

$optics14=optics" fresnel z=0.03 "; unlink 'at6.09m.txt'; unlink 'at6.09m.dfl'; run $optics14; cross $field '>at6.09m.txt'; move $field => 'at6.09m.dfl'; $optics15=optics" fresnel z=0.03 "; unlink 'at6.12m.txt'; unlink 'at6.12m.dfl'; run $optics15; cross $field '>at6.12m.txt'; move $field => 'at6.12m.dfl';

81

A.2: OPC Code Used For Gaussian Beam Simulation


#!/usr/bin/perl # This script is created by # Agee Springer (ags10@duke.edu) # based on script written by Botao Jia (botaojia@fel.duke.edu) # April, 2009 #The script simulates the diffraction effect of a single slit illuminated by a monochromatic plane #wave and the image formation when a the diffracted light is reflected by a spherical mirror of #finite size. #The optics components set up is as the following: # 1. Generate a Gaussian beam z=0 with user defined parameters; # 2. Propagate the beam through a lens focused on the mirror. # 3. Propagate and apply a single slit aperture; # 4. Propagate the beam z=6 (m) to the mirror; # 5. Apply a mirror transform with R=6 (m); # 6. Propagate the beam an additional ~6 (m) back to the camera; #Screens are set up every 0.3m from z=5.88 m to z=6.12 m after the mirror to observe the image profile. use lib '/home/research1/appl/lib4k'; #if use public domain #use lib './lib'; #if use personal machine use Physics::OPC; #$VERBOSE = 1; $field = field init => 'gaussian', I => 1, w0 => 0.01, lambda => 6.33e-7, npoints =>2048, mesh =>2e-5, nslices => 1; #$field = field #init => 'plane', #I => 1, #lambda => 6.33e-7, #npoints => 2048, #mesh => 2e-5, #nslices => 1; #first screen immediate before lens

82

$optics1=optics" fresnel z=0.28 "; unlink 'before_lense.cross.txt'; unlink 'before_lense.dfl'; run $optics1; cross $field '>before_lense.cross.txt'; move $field => 'before_lense.dfl'; #second screen immediate after lens $optics2=optics" diaphragm r=0.0125 lens f=8 "; unlink 'after_lense.cross.txt'; unlink 'after_lense.dfl'; run $optics2; cross $field '>after_lense.cross.txt'; move $field => 'after_lense.dfl'; #third screen immediate before slit $optics3=optics" fresnel z=2 "; unlink 'before_slit.cross.txt'; unlink 'before_slit.dfl'; run $optics3; cross $field '>before_slit.cross.txt'; move $field => 'before_slit.dfl'; #fourth screen after slit $optics4=optics" slit d=0.0020 # rectangle dx=0.1 dy=0.0005 "; unlink 'after_slit.cross.txt'; unlink 'after_slit.dfl'; run $optics4;

83

cross $field '>after_slit.cross.txt'; move $field => 'after_slit.dfl'; #propagate to mirror $optics5=optics" fresnel z=6 "; unlink 'before_mirror_cross.txt'; unlink 'before_mirror.dfl'; run $optics5; cross $field '>before_mirror.cross.txt'; move $field => 'before_mirror.dfl'; #screen after mirror $optics6=optics" diaphragm r=0.012 mirror r=6 "; unlink 'after_mirror.cross.txt'; unlink 'after_mirror.dfl'; run $optics6; cross $field '>after_mirror.cross.txt'; move $field => 'after_mirror.dfl'; #propagate another ~6 m $optics7=optics" fresnel z=5.88 "; unlink 'at5.88m.txt'; unlink 'at5.88m.dfl'; run $optics7; cross $field '>at5.88m.txt'; move $field => 'at5.88m.dfl'; $optics8=optics" fresnel z=0.03 "; unlink 'at5.91m.txt'; unlink 'at5.91m.dfl'; run $optics8; cross $field '>at5.91m.txt'; move $field => 'at5.91m.dfl';

84

$optics9=optics" fresnel z=0.03 "; unlink 'at5.94m.txt'; unlink 'at5.94m.dfl'; run $optics9; cross $field '>at5.94m.txt'; move $field => 'at5.94m.dfl'; $optics10=optics" fresnel z=0.03 "; unlink 'at5.97m.txt'; unlink 'at5.97m.dfl'; run $optics10; cross $field '>at5.97m.txt'; move $field => 'at5.97m.dfl'; $optics11=optics" fresnel z=0.03 "; unlink 'at6.00m.txt'; unlink 'at6.00m.dfl'; run $optics11; cross $field '> at6.00m.txt'; move $field => 'at6.00m.dfl'; $optics12=optics" fresnel z=0.03 "; unlink 'at6.03m.txt'; unlink 'at6.03m.dfl'; run $optics12; cross $field '> at6.03m.txt'; move $field => 'at6.03m.dfl'; $optics13=optics" fresnel z=0.03 "; unlink 'at6.06m.txt'; unlink 'at6.06m.dfl'; run $optics13; cross $field '> at6.06m.txt'; move $field => 'at6.06m.dfl';

85

$optics14=optics" fresnel z=0.03 "; unlink 'at6.09m.txt'; unlink 'at6.09m.dfl'; run $optics14; cross $field '>at6.09m.txt'; move $field => 'at6.09m.dfl'; $optics15=optics" fresnel z=0.03 "; unlink 'at6.12m.txt'; unlink 'at6.12m.dfl'; run $optics15; cross $field '>at6.12m.txt'; move $field => 'at6.12m.dfl';

86

S-ar putea să vă placă și