Three-dimensional optical disk data storage via ... - Semantic Scholar
Three-dimensional optical disk data storage via the localized alteration of a format hologram R. R. McLeod,1,* A. J. Daiber,2 T. Honda,3 M. E. McDonald,4 T. L. Robertson,5 T. Slagle,6 S. L. Sochava,2 and L. Hesselink7 1
Department of Electrical and Computer Engineering, University of Colorado, Boulder, Colorado 80309-425, USA 2
Intel Corporation, Optical Platform Division, 8674 Thornton Avenue, Newark, California 94560, USA
3
Optical Technology Research Center, Canon Inc., 23-10, Kiyohara-Kogyodanchi, Utsunomiya-shi, Tochigi 321-3298, Japan 4 5
SolFocus, Inc., 510 Logue Avenue, Mountain View, California 94043, USA
Proteus Biomedical, 2600 Bridge Parkway, Suite 101, Redwood City, California 94065, USA 6
Foveon, Inc., 2880 Junction Avenue, San Jose, California 95134, USA
7
Department of Electrical Engineering, Stanford University, 161 Packard Building, 350 Serra Mall, Stanford, California 94305, USA *Corresponding author: [email protected] Received 3 January 2008; accepted 14 February 2008; posted 10 March 2008 (Doc. ID 91277); published 7 May 2008
Three-dimensional optical data storage is demonstrated in an initially homogenous volume by first recording a reflection grating in a holographic photopolymer. This causes the entire volume to be weakly reflecting to a confocal read/write head. Superposition of two or three such gratings with slightly different k-vectors creates a track and layer structure that specialized servo detection optics can use to lock the focus to these deeply-buried tracks. Writing is accomplished by locally modifying the reflectivity of the preexisting hologram. This modification can take the form of ablation, inelastic deformation via heating at the focus, or erasure via linear or two-photon continued polymerization in the previously unexposed fringes of the hologram. Storage by each method is demonstrated with up to eight data layers separated by as little as 12 microns. © 2008 Optical Society of America OCIS codes: 210.2860, 210.4590, 090.4220, 160.5470, 180.1790.
1. Introduction and Motivation
An attractive approach to significantly increase the capacity of optical disk data storage is to utilize many layers, effectively increasing the available storage area [1]. However, fabrication of laminated multilayer disks to submicron tolerances beyond roughly ten layers becomes impractical, motivating the use of a homogenous volume of storage material. Thick photopolymers developed for holography are an attractive candidate for this material because of their low cost, high sensitivity, ease of processing, and good 0003-6935/08/142696-12$15.00/0 © 2008 Optical Society of America 2696
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
optical qualities in large volumes [2,3]. However, unlike the reflective layers of a traditional disk, the smooth index perturbations created by a focused writing beam in these materials do not efficiently reflect light back to the head during readout [4]. This can be overcome by writing sharp features caused by nonlinear absorption at the focus of a Ti:sapphire laser [5]. However the laser’s small multiphoton absorption cross section restricts the transfer rate and its cost is prohibitive for most applications. In contrast, the high sensitivity of linear photopolymers enables a high transfer rate while using inexpensive diode lasers, but some modification of the drive is required in order to detect the bits.
One approach is to write smooth index features at the focus of a single objective lens and then detect their presence via refractive deviation of the transmitted beam using a pick-up head on the opposite side of the disk [6]. Alternatively, two opposing heads can be used during writing to record a microhologram at the focus of the counterpropagating beams [7,8]. This micron-scale Bragg grating can then be efficiently detected in reflection. Both methods suffer from the increased complexity of opposing heads which must be accurately coaligned in three dimensions. This alignment task is exacerbated by the lack of stamped servo features commonly employed in traditional single- or two-layer disks. The readout signal strength in multilayer data storage is also significantly smaller than with a traditional optical disk system. In any approach using linear recording, unintended writing by the portions of the beam out of the focus consume all but 1=M of the material dynamic range for M total layers, causing the read signal to drop as the number of layers squared [8]. This scaling strongly limits the total disk capacity. Two-photon writing methods can overcome this limitation although they typically require a nspulsed laser to reach the required photon flux [9]. Here we present a volume data storage method that uses only a single read/write head which is locked in radius and depth to a prefabricated volume servo pattern. In addition, the power in the writing focus does not depend on depth and the reflection efficiency of the bits is independent of the precise writing mechanism. This approach admits a large number of writing mechanisms; five are demonstrated in this paper. Several of these writing mechanisms do not suffer the 1=M 2 dynamic range penalty. Finally, by removing the holography from the drive, the requirements common to holographic approaches of subwavelength mechanical stability and laser coherence are removed. The foundation of the approach is the fabrication of the disk, which consists of a monolithic photopolymer between plastic or glass cover layers. The disk is exposed in the factory with a reflection grating such that it becomes an efficient, uniform reflector. As
shown in Fig. 1, the drive can therefore read the reflectivity at the focus of the single head with the addition of a confocal filter to restrict data and servo sensitivity to the focal depth. The grating can be uniform, providing a homogenous reflective volume, or can be efficiently patterned with radial tracks and layers in depth to guide the servo system. During writing, the reflectivity of this grating is locally modified by the single drive head and these modifications are detected by the same head during reading. Thus, only a single drive head is required, unlike previous methods. Since the function of the write head is now to destroy or perturb the preformatted precision reflection grating, there are many possible write mechanisms, expanding the potential array of recording approaches. In the following sections we describe a complete implementation of this architecture. Section 2 presents the design and demonstration of the holographic servo writer, while Section 3 describes the associated servo-tracking optics. Finally, in Section 4 we demonstrate multilayer writing via a number of different writing mechanisms. 2.
Holographic Servo Writer
The primary novelty of this storage architecture is the holographic formatting of the disk with a reflection grating in order to create a large volume that is an effective reflector at all locations. This holography can take place in the factory, removing the vibration isolation, coherent laser, and opposing heads from the drive. The drive, which now looks much like a standard platform, will write data by modulating the reflectivity of the existing reflection interference grating (see Section 4). This grating need not be completely uniform: via combination of two or three superimposed gratings, the amplitude of the reflection hologram can be periodically modulated in radius and depth to provide a track and layer structure. These increase the volume storage density by suppressing interlayer and intertrack sensitivity, thus reducing the recorded feature size while simultaneously providing a servo-tracking coordinate system. This section describes the selection of the proper
Fig. 1. (Color online) System architecture. The read/write head (dashed box) tracks the factory-written reflective servo pattern via feedback from confocal servo-tracking optics. Data is written by erasing or perturbing the reflective Bragg grating in micron volumes and read by detection of the change in reflectivity at the focus. 10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2697
grating structure, the design of a six-beam holographic servo writer to record this grating, and test results from formatted disks. A. Grating Combinations
A uniform reflection grating can be fabricated in a thick photopolymer disk by exposing it to counterpropagating plane waves. As shown in Fig. 2, this formatting can take place at a wavelength λF that is smaller than the drive read wavelength λR by tilting exposing plane waves according to
2
2πnR 2π 2πnF ¼ ð1 þ δÞð1 þ sÞ 2 ¼ ð1 þ δÞ cos θ; λR λF Λ ð1Þ
where kR ¼ 2 π nR =λR and kF ¼ 2 π nF =λF are the magnitude of the wave vectors at the reading and formatting wavelengths, K ¼ 2 π=Λ is the magnitude of the reflection grating wave vector, θ is the tilt of the formatting plane-wave propagation direction relative to the material surface inside the material, s is the fractional shrinkage of the material between formatting and reading, and δ is an intentional grating detuning parameter. This parameter, which can be expressed as δ ¼ Λ=ðλR =2nR Þ − 1, must be between 0 and 6% to fall within the coherent transfer function (CTF) of the confocal reflection head operating at numerical aperture 0.5 [10]. In Section 4 the choice of δ is shown to strongly influence the read/ write performance. Temperature changes of the disk will change δ via thermal expansion. Assuming a thermal expansion
Fig. 2. (Color online) Formatting via interference of two plane waves at λF (left) and reading via diffraction of a focused beam at λR (right) in both real (top) and Fourier (bottom) spaces. The two plane waves (top left) are tilted by an angle θ (bottom left) in order to record a grating with pitch Λ that, after shrinkage s, is intentionally larger than λR =2 by a factor of 1 þ δ. This detuning determines which angular cone of the incident beam is Bragg-matched, while the thickness of the material L controls the angular width of the diffraction (upper right). 2698
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
coefficient of 50 × 10−6 = °C which is typical for polymers, a 50 °C temperature swing will cause a change in the detuning by 0.25%. This is not enough to lose signal, but does change the strength and apparent size of the bits, as shown below. If the read/write wavelength λR is between 405 and 633 nm, this 1 to 1:5 nm wavelength shift can be compensated for via roughly 25 °C of laser diode temperature tuning. In contrast, the reflection grating arrangement is relatively insensitive to tilt beyond the typical limits imposed by aberrations. In these studies, cationic ring-opening holographic photopolymer (CROP) from Aprilis Inc. [2] was utilized with a postexposure shrinkage of s ≈ 0:2%. The material is sensitive at λF ¼ 532 nm and thus the read wavelengths λR are chosen between 532 and 670 nm to match potential laser sources for the drive. For most of the write methods discussed later, the disk is flood-cured after the holographic exposure, leaving it insensitive to ambient light. The hologram plane-wave reflection efficiency is set at roughly 50% in order to achieve maximum return signal through the confocal read head and simultaneously to avoid significant optical depletion near the bottom of the disk. As shown in Fig. 2, only the read signal suffers this depletion since the majority of the beam is not Bragg-matched and thus its attenuation is not governed by the strength of the reflection grating. Writing power is therefore nearly independent of depth. The uniform grating causes the entire material volume to reflect a portion of the read beam and thus erasure or distortion of this format grating would be detectable as a reduction in the uniform reflected signal. However, positioning the head within this uniform volume would be challenging without the benefit of traditional servo marks. These cues can be provided by structuring the grating to have regions of high and low reflectivity in the shape of layers and tracks. This patterning can be accomplished by superimposing several mutually incoherent gratings such that the holographic fringes beat to form moiré patterns. Referring to Fig. 3(b), two reflection holograms with grating vectors normal to the surface but differing in magnitude by the spatial beat frequency of δK Z ¼ π=δZ come into and out of phase to create a layered structure with layer spacing δZ. Between these layers the two gratings are out of phase, suppressing the reflected signal. Thus these layers both provide servo information and reduce the extent of recorded features, reducing interlayer crosstalk and enabling higher storage density. Similarly, two gratings tilted in the radial dimension such that the radial component of their wave vectors are δkR ¼ π =ð2 δRÞ create cylindrical tracks with spacing δR. In the case of layer-only formatting described in the previous paragraph, the two gratings can be created via tilt of plane waves according to Eq. (1). The recording optics are more complex in this radial track case because the R coordinate
Fig. 3. Possible holographic servo patterns including (a) uniform reflection grating, (b) layer patterning in depth, (c) track patterning in radius and (d) simultaneous layer and track patterning. The arrows represent the incoherent grating vectors needed to create each pattern.
vector changes around the disk. The solution is to restrict the recording beams to ∼100 μm slices in the RZ plane and then to rotate the disk through this thin recording plane. The holographic fringes are parallel to the disk surface and thus, if the motion is sufficiently stable, will be faithfully recorded in the moving material. Finally, as shown in Fig. 3(d), three mutually incoherent gratings with the amplitude of the central hologram twice that of the two tilted gratings will record a pattern with both layer and track structures. Assuming that all the gratings are of the same pitch but tilted, as shown, the track spacing δR and layer spacing δZ are related by sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2ffi π 2π 2π 2π : ð2Þ ¼ − − δZ Λ Λ δR This relationship does not always yield the desired layer spacing, so one can also use angle tuning [via Eq. (1)] to adjust the pitch of the individual gratings to achieve nearly any combination of track and layer spacing. Note that tracks in adjacent layers are interleaved, which further decreases interlayer crosstalk, although at the expense of slightly more complex interlayer seek algorithms. B.
Track and Layer Holographic Servo Writer
According to the previous discussion, the fabrication of depth and radial tracks in a cylindrical coordinate system requires an interference pattern confined to a thin RZ slice through which the disk is rotated during format recording. This in turn restricts any tilt of 10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2699
Fig. 4. (Color online) Creation of the layer and radial holographic tracking pattern. As shown in (a), three pairs of mutually incoherent beams record three reflection gratings in a narrow radial spoke. The k-space representation (b) shows how the six green plane waves create the three desired grating vectors, tuned for readout at a longer wavelength.
the beams to the RZ plane, as shown in Fig. 4(a). These three tilts, as shown in Fig. 4(b), are then selected to achieve the three pattern specifications of detuning (δ) at the read wavelength (λR ), the track pitch (δR), and the layer spacing (δZ). The diffraction of the focused read beam off of the resulting grating structure is three displaced rings, similar to the single ring shown in Fig. 2. Note that for a thick material (large L) only three narrow rings are Braggmatched out of the incident cone and reflected back to become the read signal. The disk is therefore nominally transparent to the remainder of the angular spectrum, enabling the absorption of the writing mechanism to be adjusted independently of the disk reflectivity used for reading and tracking.
The three pairs of waves must be mutually incoherent to avoid the recording of undesired grating crossterms. Since the disk must be moved during formatting, sequential recording of the gratings would require an impractical level of stability from the system spindle. Simultaneous recording with mutually incoherent beams could be accomplished with three independent lasers but alignment and precise relative power control would be challenging. This motivates the use of a single laser and an acousto-optic (AO) beam splitter. We thus employ a tangentiallymatched anisotropic AO deflector fabricated from TeO2 . The incident beam is ordinarily polarized, rather than the typical extraordinary polarization intended for the deflector [11]. When the crystal is oriented for symmetrical diffraction, the incident beam diffracts off of the moving acoustic grating into both the positive and negative orders, causing the diffracted beams to be Doppler up- and down-shifted by the acoustic drive frequency. The diffracted beams and the remaining undiffracted zero-order are thus mutually incoherent for recording times longer than one over the acoustic frequency, typically ∼100 MHz. The relative beam angles and powers are precisely controlled by the acoustic frequency and power. As shown in the inset of Fig. 5, weak prisms are used to tilt the diffracted beams back to the incident propagation direction, and a half-wave plate is used to rotate the zero-order output polarization to match the diffracted beams. The remainder of the servo writer design is shown in Fig. 5. The target Strehl ratio for the drive and disk is 0.9, which is equivalent to a wavefront variation of λ=20. The three separated beams are
Fig. 5. (Color online) Optical ray trace of the holographic servo writer. The acousto-optic (AO) deflector creates three mutually incoherent beams which are expanded in the plane of the drawing and split by the beam splitter (BS) into three pairs stacked in depth. Pickoff mirrors direct these to tilted plates, which bring the beams back to the same plane and titled cylindrical lenses, focussing the interference pattern into a thin sheet normal to the along-track (S) direction. The disk to be exposed is rotated through this interference pattern by a precision spindle. 2700
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
anamorphically expanded to cover the radial extent of the disk with a four-stage prism telescope since plano-convex cylindrical lenses at this stage would introduce unacceptable aberrations. The three anamorphic beams, stacked normal to the plane of the drawing, are then split with a beam splitter to generate the three pairs of mutually incoherent beams. Pick-off mirrors send these beams toward the disk, and tilted parallel plates are used to bring them back to the same plane. Cylindrical lenses (which are now acceptable due to their proximity to the end of the optical train) focus the beams into 100 μm thick RZ sheets. These lenses are tilted to compensate for the refraction at the surfaces of the coupling prisms, which are necessary to reach the required beam angles within the material. These prisms are oil-coupled to the 1 mm glass substrates of the polymer disk which is rotated on an air-bearing spindle to record the patterns around the entire disk. A RZ slice of a disk formatted for reading at 532 nm with this hardware is shown in Fig. 6. Thirteen layers with a 7:85 μm separation are recorded into the 100 μm polymer thickness. The reflectivity of the tracks, as measured by a 0.5 NA Geltech aspheric objective corrected for the appropriate spherical aberration, is 5 × 10−5 and the contrast between the on and off track is at least 10:1. The entire recording process requires approximately 1 min with a 400 mW diode-pumped solid-state (DPSS) laser at 532 nm. As shown in Fig. 7, the pattern recorded is quite accurate and could be made more so with the removal of a systematic error, probably due to a stray reflection. At the start and end point in angle, there is a radial offset of up to 1=2 track, limited by
Fig. 6. RZ confocal scanning reflection measurement made by the precision drive of a 13 layer formatted disk. The 180 nm fringes that provide the reflection efficiency of each track are not visible. The pattern is sampled at 50 nm radially and 300 nm in depth.
Fig. 7. Precision of the radial track position of the tracks shown in Fig. 6. The plot shows the measured deviation of the track center from its nominal position as a function of the radius, indicating a maximum track squeeze of 200 nm.
mechanical stability of the exposure system and shrinkage of the polymer during development. 3.
Servo Detection and Locking
The purpose of the structured reflectivity is to provide a radial and depth coordinate system, to which the read/write focus will be locked by closed-loop servo. In order to maintain this lock, the length of a mark which erases the servo track must be restricted; this limit is common in many modulation codes. Like the data detection, the servo detection optics must be sensitive to the layer near the focus and reject information from all other layers, suggesting the use of a confocal filter. This confocal filter can easily be integrated into the radial tracking servo optics. Light from the data-channel confocal filter is imaged onto a detector split in the radial direction to implement a confocally-filtered version of a pushpull radial servo. The image of the tracks on the confocal filter shifts as the head moves radially with respect to the disk, and the difference between the left and right halves of this image yields a directional error signal for radial tracking. Alternatively, the split detector itself can be used as the confocal filter if its diameter is equal to the diameter of the read focus times the magnification of the detection optics. However, this typically requires a long focal length lens in the detection arm or an inconveniently small detector. The servo optics for detection of focus and layer position error are conceptually the same as those used for radial tracking in that the goal is to subtract the reflected intensity of the layer structure at two locations displaced above and below the focus depth. This can be accomplished by splitting the detected light with a beam splitter into two confocal detection arms and axially displacing the pinhole in each arm relative to the focus of the detection lens [12], as shown in Fig. 8. The two pinholes are now conjugate to planes just above and below the focus of the read/write head. Experimental results from this system are shown in Fig. 9 for the layer-only format shown in Fig. 3(b). The upper plot shows how the displacement of the peak confocal sensitivity causes the two return 10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2701
Fig. 8. Structure of the combined radial and depth confocal servo detection optics. The beam splitter divides the focused beam which is sent through two pinholes placed before and after the focus for a collimated input. Split photo-diodes detect the radial tracking error. The data, radial error, and depth error signals are computed from A þ B þ C þ D, ðB þ DÞ − ðA þ CÞ, and ðA þ BÞ− ðC þ DÞ, respectively.
signals to shift in phase. The electronic difference of the two photocurrents, shown in the lower plot, yields a directional error signal with a negative-slope zero crossing at the peak of each track and a positiveslope zero crossing in between layers. 4. Read/Write Mechanisms
At this point, we have a simple, single-headed optical drive mechanism that can lock the head radially and in depth to an efficiently exposed tracking structure in an otherwise homogeneous photopolymer disk. Reflectivity in this volume is provided by a reflection grating structure recorded as approximately 200 nm pitch sinusoidal polymer density variations. The goal of the write mechanism is to locally modify this structure such that the amplitude of the Bragg-matched reflection signal is detectably changed. The second key innovation of this architecture is that this writing operation is not in itself holographic and thus can be done with a single focused beam, vastly simplify-
Table 1.
Fig. 9. Depth servo signal measured for a layer-only format grating. The top plot shows the signals detected from the two displaced confocal filters while the bottom is the electronic difference, indicating a layer peak at every negative-slope zero crossing. The dashed line indicates the depth of a representative layer crossing.
ing the drive layout. In this section, we demonstrate three primary writing strategies: (1) destruction of the grating via a short pulsed laser, (2) inelastic distortion of the grating via local heating at the writing focus, and (3) erasure of the grating via continued polymerization in the formerly unexposed fringes of the hologram. In the second scheme, the grating is tuned both into and out of the confocal CTF, while in the third scheme, both linear and two-photon polymerization processes are investigated. The experiments and conditions are summarized in Table 1 [13,14]. For each process, the erasure of a preformatted reflective structure has an inherent nonlinearity, even when the recording mechanism is linear and integrating. That is, one can not erase the reflectivity below zero. While not as strong as the threshold nonlinearities common to existing optical storage products, this saturating response serves to improve
Summary of Experimental Conditions for the Results in Figs. 10–14
Deformation Ablation Figure Detuning Laser Write λ Peak Power Write Time Energy/Bit NA Spot Radius Bit Spacing Layer Spacing # of Layers Read λ
2702
10.b ∼0% Ti:sapphire 800 nm 100 kW 120 fs 12 nJ 0.38 1:26 μm 3 μm NA 1 633 nm
Negative 10.c ∼3% Q-switched 532 nm 230 mW 20 μs 4:6 μJ 0.38 0:84 μm 2 μm NA 1 532 nm
Continued Polymerization
Positive
Linear
10.d ∼0% DPSS 532 nm 25 W 40 ns 1 μJ 0.5 0:64 μm 1:5 μm 12 μm 5 532 nm
10.e 3% Diode 670 nm 40 mW 2 ms 80 μJ 0.5 0:8 μm 2 μm 20 μm 5 670 nm
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
2-Photon 12 0:35–6:35% α − DFB 659 nm 50 mW 3s 150 mJ 0.5 0:79 μm 1:5 μm NA 1 659 nm
(10.f,13) 3:3% α − DFB 659 nm 50 mW 10 s 500 mJ 0.5 0:79 μm 1:5 μm (NA, 12 μm) (NA, 8) 659 nm
14 2:5% DPSS 532 nm 100 mW 6 ms 600 μJ 0.5 0:64 μm 2 μm 20 μm 6 532 nm
the signal-to-noise ratio (SNR) of the written “ones.” The SNR of the “zeros” is determined by the quality of the formatting and is discussed later in Subsection 4.D. A. Ablation with a Femtosecond Laser
The conceptually simplest form of writing is to locally destroy the polymer grating with an ultrafast laser pulse. This is similar to writing sharply-defined structures in glass [5], except that the strength of the read signal is determined by the grating reflectivity which is independent of the precise shape or mechanism of the laser/material interaction. The
uniform grating, exposed at 532 nm as previously described, is tuned for readout at 633 nm. The photopolymer is then completely cured via exposure to room lights for 24 hours. The confocal reflectivity of this grating at 633 nm is shown in Fig. 10(a). The sample is then placed at the 0:38 NA focus of a Ti:sapphire laser. A line of bits with 3 μm spacing is written at λ ¼ 800 nm using a single motorized stage. The sample was removed and placed on an XYZ motion system and a focused HeNe laser was used to read the data as shown in Fig. 10(b). The grating is nearly completely erased where writing occurred (marked by the symbols “1”). The profile of the
Fig. 10. (Color online) Results of different write mechanisms showing detected signal versus along-track distance (S). All horizontal scales save (a) are the same for ease of comparison of bit shapes. The graphics on the right indicate the writing mechanisms which are ablation, inelastic deformation, and one- or two-photon continued polymerization from top to bottom. Further details for each plot are given in Table 1.
10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2703
reflected signal fits well to a simple model in which the contrast of the index grating is reduced proportionally to pulse intensity, however the precise mechanism of erasure is unknown. Since the absorption method is presumably nonlinear, this indicates some broadening of the recording area. Note that the energy per bit stored, 12 nJ, is significantly lower than the roughly μJ pulse energies used for writing into glass [15], presumably due to the lower damage threshold of the organic material. B. Inelastic Deformation with a Q-Switched DPSS Laser
While the damage at the focus of an ultrafast laser demonstrates the feasibility of this storage mechanism, the resonant structure of the Bragg grating provides other writing mechanisms that require significantly less peak power. For example, as shown in Fig. 11, linear absorption at the focus will heat and expand the material, locally increasing the grating pitch. As the grating spatial frequency changes, it moves through the CTF of the confocal read head, tracing out the Bragg selectivity curve including the side-lobes. Writing with higher power expands the material beyond its elastic limit, permanently deforming the grating pitch at the focus. Thus it should be possible to permanently decrease the local reflectivity of the format grating by increasing the pitch and detuning the grating spatial frequency out of the passband of the confocal CTF. Unlike the previous case, where the femtosecond laser pulse erased the grating, independent of its detuning, here the recording process is very sensitive to the grating pitch. Incoherent optical curing of the format grating naturally causes a depth-dependent variation of the detuning parameter due to Beer–Lambert absorption and polymerization shrinkage. Therefore the format holograms in this case were thermally cured at 65 °C for 20 h. For this demonstration, a Q-switched DPSS laser at 532 nm was chosen as the writing laser source and, for convenience, also was used as the reading laser. A uniform format grating was exposed with a roughly 3% detuning. To increase the optical absorption and
proportionally decrease the energy requirement per bit, the material was exposed to 200 J=cm2 at 313 nm after the normal curing which, via an unknown photochromic reaction, darkened the material. Elastic deformation dynamics like those in Fig. 11 were observed until the incident power passed the inelastic threshold, at which point the recovery during cooling was not complete. Figure 10(c) shows the reflectivity of a track after exposure to 20 μs pulses with peak power of 230 mW spaced at 2 μm. To confirm that the decrease in signal was due to deformation, not some other damage mechanism, pumpprobe curves like Fig. 11 were taken while focused on a previously-written bit. The dynamical behavior was the same except that it started from a new point in the Bragg curve, as expected. As indicated by Fig. 11, deformation can either decrease or increase the format reflectivity depending on the initial grating detuning, δ, relative to the CTF of the confocal read channel. Since the permanent deformation is always an expansion, to write bits with a higher reflectivity requires an initial grating that has a finer pitch than is optimal for the confocal microscope or, in our definition, δ ≈ 0. To further confirm that the deformation was purely local, five layers of such “positive” bits separated by 12 μm were written in a uniform format grating, as shown in Fig. 10(d). These are narrower than those shown in Fig. 10(c) because of the larger NA used and the greater slope of the confocal CTF at its outer boundary. Also in this experiment, the dependence of the writing energy on the pulse duration was explored by reducing the pulse duration by a factor of 500. As expected for this linear absorption process, the writing energy is a very weak function of this duration, dropping roughly by a factor of 5. The additional energy required at longer pulse durations is probably due to thermal diffusion. Although it was not fully explored in this experiment, the inelastic deformation of the material is expected to be a highly nonlinear function of temperature. Thus, although optical absorption which creates the temperature rise is due to linear absorption, the deformation of the material should be confined to a region near the focus. This does not result in a smaller bit size (like the two-photon absorption discussed below) but does limit interlayer crosstalk and the integration of out-of-focus response, which is a concern for completely linear processes, discussed next. C. Erasure via Linear Photopolymerization with a DPSS Laser
Fig. 11. Elastic deformation via heating with a 25 μs pump pulse from a Q-switched DPSS laser of a uniform format grating with detuning δ < 0. As the grating expands locally, the reflected signal traces out the Bragg selectivity curve. 2704
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
To decrease the peak power to the range of 50 mW and reach the goal of using a diode laser for writing and reading, we investigated a third method of modifying the grating reflectivity. In this mechanism, the contrast of the grating is reduced by initiating polymerization in the regions of the hologram that were located at intensity nulls during formatting. This erases the sinusoidal polymer density
variation and, thus, the index grating that provides the reflected signal. A single focused beam can initiate polymerization in the unexposed fringes if the hologram recording process strongly decreases the material sensitivity in the exposed fringes. This can be engineered simply by adjusting the constituent chemical concentrations such that some critical but nondiffusing species is completely consumed in the bright fringes and, conversely, remains for later use in the dark fringes. Note that the format grating now serves a dual purpose of modulating the index of refraction to provide a reflected signal and modulating the recording sensitivity such that a single focused beam can write a sinusoidal index pattern that is precisely out-of-phase with the original hologram. This sinusoidal sensitivity pattern is created by significantly depleting some essential chemical species only in the bright fringes of the initial hologram recording. In this case, both the green-absorbing photosensitizer 5,12-bis(phenylethynyl)naphthacene (BPEN) and the photo-acid generator (PAG) are reduced to roughly 20% of the Aprilis commercial formulation [2]. A second, red-sensitive dye is added to the material. Like the green-sensitive BPEN, this dye does not directly initiate polymerization but instead chemically excites the intermediate PAG. The concentration of this PAG is spatially modulated by the recording of the hologram in the green such that the later recording with a focused beam in the red will preferentially write in the previously unexposed regions of the hologram, reducing the grating contrast. The green recording angles are adjusted to write a grating for read/write at 670 nm. Unlike all other cases presented, the material is not cured and is thus linearly photosensitive and must be shielded from ambient light before writing data. Five layers of bits separated by 2 μm transversely and 20 μm in depth are written into the red-sensitive material by a 40 mW diode laser, a single track of which is shown in Fig. 10(e). The red pulse energy of 80 μJ is significantly higher than counterpropagating microhologram storage demonstrated in this same material [8], which achieved roughly 1 nJ exposure energies in the green; however, continued optimization of the red sensitizer could presumably narrow this gap considerably.
initiation, and to understand their dependence on the grating detuning parameter, we formatted a series of polymer samples at 532 nm for readout by a 659 nm tapered distributed-feedback (α-DFB) laser diode. After formatting, the remaining green sensitizer BPEN was completely bleached. To ensure that the PAG was not completely consumed by this process, only 20% of the nominal BPEN concentration was used in the formulation, similar to the procedure described in Section 4.C. After the roughly s ¼ 0:2% shrinkage during the bleaching process, the detuning, δ, ranged from 0.17% to 6.01% in the five samples. After bleaching, the plane-wave diffraction efficiency of the samples was 55 10%. Single planes of bits were written into these materials according to the conditions of Table 1. Because of the small two-photon absorption cross section, the exposure energies are the largest of any method investigated. The bit size and 659 nm confocal reflectivity versus detuning are shown in Fig. 12. The fullwidth to half-maximum bit diameter varies from 1:4 μm down to 0:87 μm as the detuning is increased while the reflectivity of the format hologram drops by a factor of 14. As shown in the sketch of the CTF to the right of the plot, when the detuning is small the available, transverse spatial frequencies are restricted, resulting in larger observed bit size. As the detuning increases, the transverse passband of the confocal microscope is broadened, resulting in smaller observed bits but at the expense of lower total efficiency due to the roll-off of the CTF. This rolloff is caused by the Gaussian filling of the objective and thus could presumably be improved with a tophat apodization. Based on this study, a detuning of ∼3% is optimum for multilayer experiments. Note that this is the midpoint of the confocal CTF for the 0:5 NA objective used. In this experiment, the format grating is structured with a layer-only format to demonstrate the simultaneous use of a structured grating and data storage. Eight layers separated by 12 μm are formatted and cured. Two-photon initiation at 659 nm of the structured PAG is used to locally erase the format hologram in spots separated by 1:5 μm in the plane. The precision drive is then used as a 3D confocal microscope to produce images of the reflectivity with fine sampling resolution. A depth (Z) and
D. Erasure via Two-photon Polymerization
Iodonium PAGs have moderate two-photon cross sections [16]. In Section 4.C we showed that a PAG can be structurally patterned by its consumption in the formatting stage. Rather than initiate this PAG hologram with a red-sensitive linear dye, it can thus be directly initiated via two-photon absorption. This will start polymerization only in the unexposed fringes of the format grating, erasing the reflectivity in a region defined by the square of the intensity of the writing focus. To quantify the size of these bits, which are expected to be smaller than those formed by linear
Fig. 12. (Color online) Full-width to half-maximum transverse bit diameter (upper curve) written by two-photon continued polymerization and reflectivity of the uniform format grating (lower curve) versus grating detuning. The sketch of the CTF on the right qualitatively explains the observed behaviors. 10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2705
along-track (S) slice of these data is shown in Fig. 13 (right). The slight index mismatch between glass and cured polymer makes these surfaces visible to the confocal head, while the layered pattern of the format grating appears as horizontal stripes. The writing locations appear as dark holes in these formatted layers since the contrast of the photopolymer grating has been reduced at these locations by the point excitation of the PAG, which is out-of-phase with the holographic index pattern. Based on these experiments, the two-photon cross section of the PAG at 659 nm was estimated to be αΠ ¼ 1:5 10−5 cm4 s. The two-photon cross section of the PAG was found to be significantly larger at 532 nm, so as a final experiment, a uniform grating tuned for 532 nm reading was fabricated with 2.5% detuning. Using 100 mW of 532 nm light in 6 ms pulses, six layers separated by 20 μm were written. Figure 14 shows the resulting reflectivity measurements in each of the bit planes. The writing energy of 600 μJ is roughly an order of magnitude higher than previous studies of two-photon data storage in polymers [5], consistent with the fact that the PAG has not been optimized for a two-photon cross section. These images reveal that the reflectivity of the unwritten region is modulated by sinusoids at two different periods and orientations. These noise gratings are presumably due to unwanted reflections in the formatting hardware or within the sample itself and explain the large noise observed in the “zero” levels of Fig. 10. 5. Summary and Conclusions
Here we have addressed three problems associated with volumetric data storage in photopolymers. First, data storage in thick, homogeneous volumes traditionally requires heads on both sides of the media, complicating the mechanical packaging or highly nonlinear material response, requiring the use of expensive ultrafast lasers. Second, these volumes provide no servo-tracking marks, making the transition from instrumented lab bench to inexpensive servo-controlled product difficult. Finally, sensitive and inexpensive holographic photopolymers are a promising candidate for these bit-based 3D storage architectures, but the linear index re-
Fig. 13. (Color online) Storage via two-photon continued polymerization initiated by a DFB laser diode in a 100 μm thick photopolymer formatted with eight layers. 2706
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
Fig. 14. (Color online) Storage via two-photon continued polymerization initiated by a 532 nm DPSS laser in a uniform format grating. The 51 ðRÞ × 51 ðSÞ × 6 ðZÞ bits were separated by 2 ðRÞ× 2 ðSÞ × 20 ðZÞ μm. Note the three broad horizontal bands and finer stripes at 45° across the upper left image that indicate the presence of weak noise gratings.
sponse is not obviously the best candidate for the system due to the accumulated out-of-focus response. We have solved the difficulty of two opposing read/ write heads by separating the holography from the write mechanism so that a single-head optical drive can both read and write from one side of the disk. By fabricating the precision grating in the factory, the drive does not require vibration isolation or a highly coherent laser, both hallmarks of traditional holographic data storage. We addressed the problem of servo-tracking by patterning the reflective holographic format with a small number of gratings that beat to produce radial tracks, depth layers, or both. A servo writer capable of fabricating this track and layer structure was designed and implemented for 120 mm diameter photopolymer disks. We implemented servo and data detection optics with confocal filtering and demonstrated closed-loop radial and depth tracking of deeply-buried marks in a spinning disk. We expanded the range of bit-based read/write mechanisms for holographic photopolymers by demonstrating five different schemes including femtosecond ablation, thermally-initiated inelastic detuning for negative or positive bits, and continued polymerization driven by spatially structured linear or two-photon absorption. The final two cases demonstrated the complex response that can be engineered in such “homogenous” volumes: the focused write beam excited polymerization with a 220 nm period that was precisely out-of-phase with and the same magnitude as the previously-written hologram. The surface storage density D in storage locations per micron2 of this method should be similar to that derived for microholograms [8],
D≈
π NA4 L : 10 λ3R n
ð3Þ
This limit was derived for a uniform medium, thus in the case of layer and track formatting that reduce the recorded bit dimensions, this density should increase. This suggests the possibility of TByte capacity in submillimeter thickness disks of 120 mm diameter. When this work was performed, the authors were with Siros Technologies Incorporated, San Jose, California. References 1. K. A. Rubin, H. J. Rosen, W. W. Tang, W. Imaino, and T. C. Strand, “Multilevel volumetric optical disk storage,” Proc. SPIE 2338, 247–250 (1994). 2. D. A. Waldman, R. T. Ingwall, P. K. Dhal, M. G. Horner, E. S. Kolb, H.-Y. S. Li, R. A. Minns, and H. G. Schild, “Cationic ringopening photopolymerimization methods for volume hologram recording,” Proc. SPIE 2689, 127–141, (1996). 3. L. Dhar, A. Hale, H. E. Katz, M. L. Schilling, M. G. Schnoes, and F. C. Schilling, “Recording media that exhibit high dynamic range for digital holographic data storage,” Opt. Lett. 24, 487–489 (1999). 4. T. Wilson, Y. Kawata, and S. Kawata, “Readout of threedimensional optical memories,” Opt. Lett. 21, 1003–1005 (1996). 5. J. H. Strickler and W. W. Webb, “Three-dimensional optical data storage in refractive two photon point excitation,” Opt. Lett. 16, 1780–1783 (1991). 6. Y. Kawata, R. Juskaitis, T. Tanaka, T. Wilson, and S. Kawata, “Differential phase-contrast microscope with a split detector for the readout system of a multilayered optical memory,” Appl. Opt. 35, 2466–2470 (1996).
7. S. Orlic, S. Ulm, and H. J. U. Eichler, “3D bit-oriented optical storage in photopolymers,” J. Opt. A: Pure Appl. Opt. 3, 72– 81 (2001). 8. R. R. McLeod, A. J. Daiber, M. E. McDonald, T. L. Robertson, T. Slagle, S. L. Sochava, and L. Hesselink, “Micro-holographic multi-layer optical disk data storage,” Appl. Opt. 44, 3197– 3207 (2005). 9. E. Walker, A. Dvornikov, K. Coblentz, S. Esener, and P. Rentzepis, “Toward terabyte twophoton 3D disk,” Opt. Express 15, 12264–12276 (2007). 10. Min Gu, Principles of Three-Dimensional Imaging in Confocal Microscopes (World Scientific, 1996). 11. T. Weverka, K. Wagner, R. R. McLeod, and K. Wu, “Low-loss acousto-optic photonic switch,” in Acousto-Optic Signal Processing, N. J. Berg and J. M. Pellegrino, eds. (Marcel Dekker, 1994). 12. M. S. Wang and T. D. Milster, “Differential wax-wane focus servo,” Appl. Opt. 32, 4797–4807 (1993). 13. L. Hesselink, “Three dimensional recording (3DR) technology,” in Optical Data Storage 2000, IEEE Conference Digest (IEEE, 2000), pp. 149–151. 14. L. Hesselink, R. R. McLeod, and S. L. Sochava, “Optical data storage by selective localized alteration of a format hologram in a holographic storage disk,” US Patent 6,614,741 (Sept. 2, 2003). 15. E. N. Glezer, M. Milosavljevic, L. Huang, R. J. Finlay, T.-H. Her, J. P. Callan, and E. Mazur, “Three-dimensional optical storage inside transparent materials,” Opt. Lett. 21, 2023– 2025 (1996). 16. K. J. Schafer, J. M. Hales, M. Balu, K. D. Belfield, E. W. Van Stryland, and D. J. Hagan, “Two-photon absorption cross sections of common photoinitiators,” J. Photochemistry and Photobiology A: Chemistry 162, 497–502 (2004).
10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2707
Department of Electrical and Computer Engineering, University of Colorado, Boulder, Colorado 80309-425, USA 2
Intel Corporation, Optical Platform Division, 8674 Thornton Avenue, Newark, California 94560, USA
3
Optical Technology Research Center, Canon Inc., 23-10, Kiyohara-Kogyodanchi, Utsunomiya-shi, Tochigi 321-3298, Japan 4 5
SolFocus, Inc., 510 Logue Avenue, Mountain View, California 94043, USA
Proteus Biomedical, 2600 Bridge Parkway, Suite 101, Redwood City, California 94065, USA 6
Foveon, Inc., 2880 Junction Avenue, San Jose, California 95134, USA
7
Department of Electrical Engineering, Stanford University, 161 Packard Building, 350 Serra Mall, Stanford, California 94305, USA *Corresponding author: [email protected] Received 3 January 2008; accepted 14 February 2008; posted 10 March 2008 (Doc. ID 91277); published 7 May 2008
Three-dimensional optical data storage is demonstrated in an initially homogenous volume by first recording a reflection grating in a holographic photopolymer. This causes the entire volume to be weakly reflecting to a confocal read/write head. Superposition of two or three such gratings with slightly different k-vectors creates a track and layer structure that specialized servo detection optics can use to lock the focus to these deeply-buried tracks. Writing is accomplished by locally modifying the reflectivity of the preexisting hologram. This modification can take the form of ablation, inelastic deformation via heating at the focus, or erasure via linear or two-photon continued polymerization in the previously unexposed fringes of the hologram. Storage by each method is demonstrated with up to eight data layers separated by as little as 12 microns. © 2008 Optical Society of America OCIS codes: 210.2860, 210.4590, 090.4220, 160.5470, 180.1790.
1. Introduction and Motivation
An attractive approach to significantly increase the capacity of optical disk data storage is to utilize many layers, effectively increasing the available storage area [1]. However, fabrication of laminated multilayer disks to submicron tolerances beyond roughly ten layers becomes impractical, motivating the use of a homogenous volume of storage material. Thick photopolymers developed for holography are an attractive candidate for this material because of their low cost, high sensitivity, ease of processing, and good 0003-6935/08/142696-12$15.00/0 © 2008 Optical Society of America 2696
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
optical qualities in large volumes [2,3]. However, unlike the reflective layers of a traditional disk, the smooth index perturbations created by a focused writing beam in these materials do not efficiently reflect light back to the head during readout [4]. This can be overcome by writing sharp features caused by nonlinear absorption at the focus of a Ti:sapphire laser [5]. However the laser’s small multiphoton absorption cross section restricts the transfer rate and its cost is prohibitive for most applications. In contrast, the high sensitivity of linear photopolymers enables a high transfer rate while using inexpensive diode lasers, but some modification of the drive is required in order to detect the bits.
One approach is to write smooth index features at the focus of a single objective lens and then detect their presence via refractive deviation of the transmitted beam using a pick-up head on the opposite side of the disk [6]. Alternatively, two opposing heads can be used during writing to record a microhologram at the focus of the counterpropagating beams [7,8]. This micron-scale Bragg grating can then be efficiently detected in reflection. Both methods suffer from the increased complexity of opposing heads which must be accurately coaligned in three dimensions. This alignment task is exacerbated by the lack of stamped servo features commonly employed in traditional single- or two-layer disks. The readout signal strength in multilayer data storage is also significantly smaller than with a traditional optical disk system. In any approach using linear recording, unintended writing by the portions of the beam out of the focus consume all but 1=M of the material dynamic range for M total layers, causing the read signal to drop as the number of layers squared [8]. This scaling strongly limits the total disk capacity. Two-photon writing methods can overcome this limitation although they typically require a nspulsed laser to reach the required photon flux [9]. Here we present a volume data storage method that uses only a single read/write head which is locked in radius and depth to a prefabricated volume servo pattern. In addition, the power in the writing focus does not depend on depth and the reflection efficiency of the bits is independent of the precise writing mechanism. This approach admits a large number of writing mechanisms; five are demonstrated in this paper. Several of these writing mechanisms do not suffer the 1=M 2 dynamic range penalty. Finally, by removing the holography from the drive, the requirements common to holographic approaches of subwavelength mechanical stability and laser coherence are removed. The foundation of the approach is the fabrication of the disk, which consists of a monolithic photopolymer between plastic or glass cover layers. The disk is exposed in the factory with a reflection grating such that it becomes an efficient, uniform reflector. As
shown in Fig. 1, the drive can therefore read the reflectivity at the focus of the single head with the addition of a confocal filter to restrict data and servo sensitivity to the focal depth. The grating can be uniform, providing a homogenous reflective volume, or can be efficiently patterned with radial tracks and layers in depth to guide the servo system. During writing, the reflectivity of this grating is locally modified by the single drive head and these modifications are detected by the same head during reading. Thus, only a single drive head is required, unlike previous methods. Since the function of the write head is now to destroy or perturb the preformatted precision reflection grating, there are many possible write mechanisms, expanding the potential array of recording approaches. In the following sections we describe a complete implementation of this architecture. Section 2 presents the design and demonstration of the holographic servo writer, while Section 3 describes the associated servo-tracking optics. Finally, in Section 4 we demonstrate multilayer writing via a number of different writing mechanisms. 2.
Holographic Servo Writer
The primary novelty of this storage architecture is the holographic formatting of the disk with a reflection grating in order to create a large volume that is an effective reflector at all locations. This holography can take place in the factory, removing the vibration isolation, coherent laser, and opposing heads from the drive. The drive, which now looks much like a standard platform, will write data by modulating the reflectivity of the existing reflection interference grating (see Section 4). This grating need not be completely uniform: via combination of two or three superimposed gratings, the amplitude of the reflection hologram can be periodically modulated in radius and depth to provide a track and layer structure. These increase the volume storage density by suppressing interlayer and intertrack sensitivity, thus reducing the recorded feature size while simultaneously providing a servo-tracking coordinate system. This section describes the selection of the proper
Fig. 1. (Color online) System architecture. The read/write head (dashed box) tracks the factory-written reflective servo pattern via feedback from confocal servo-tracking optics. Data is written by erasing or perturbing the reflective Bragg grating in micron volumes and read by detection of the change in reflectivity at the focus. 10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2697
grating structure, the design of a six-beam holographic servo writer to record this grating, and test results from formatted disks. A. Grating Combinations
A uniform reflection grating can be fabricated in a thick photopolymer disk by exposing it to counterpropagating plane waves. As shown in Fig. 2, this formatting can take place at a wavelength λF that is smaller than the drive read wavelength λR by tilting exposing plane waves according to
2
2πnR 2π 2πnF ¼ ð1 þ δÞð1 þ sÞ 2 ¼ ð1 þ δÞ cos θ; λR λF Λ ð1Þ
where kR ¼ 2 π nR =λR and kF ¼ 2 π nF =λF are the magnitude of the wave vectors at the reading and formatting wavelengths, K ¼ 2 π=Λ is the magnitude of the reflection grating wave vector, θ is the tilt of the formatting plane-wave propagation direction relative to the material surface inside the material, s is the fractional shrinkage of the material between formatting and reading, and δ is an intentional grating detuning parameter. This parameter, which can be expressed as δ ¼ Λ=ðλR =2nR Þ − 1, must be between 0 and 6% to fall within the coherent transfer function (CTF) of the confocal reflection head operating at numerical aperture 0.5 [10]. In Section 4 the choice of δ is shown to strongly influence the read/ write performance. Temperature changes of the disk will change δ via thermal expansion. Assuming a thermal expansion
Fig. 2. (Color online) Formatting via interference of two plane waves at λF (left) and reading via diffraction of a focused beam at λR (right) in both real (top) and Fourier (bottom) spaces. The two plane waves (top left) are tilted by an angle θ (bottom left) in order to record a grating with pitch Λ that, after shrinkage s, is intentionally larger than λR =2 by a factor of 1 þ δ. This detuning determines which angular cone of the incident beam is Bragg-matched, while the thickness of the material L controls the angular width of the diffraction (upper right). 2698
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
coefficient of 50 × 10−6 = °C which is typical for polymers, a 50 °C temperature swing will cause a change in the detuning by 0.25%. This is not enough to lose signal, but does change the strength and apparent size of the bits, as shown below. If the read/write wavelength λR is between 405 and 633 nm, this 1 to 1:5 nm wavelength shift can be compensated for via roughly 25 °C of laser diode temperature tuning. In contrast, the reflection grating arrangement is relatively insensitive to tilt beyond the typical limits imposed by aberrations. In these studies, cationic ring-opening holographic photopolymer (CROP) from Aprilis Inc. [2] was utilized with a postexposure shrinkage of s ≈ 0:2%. The material is sensitive at λF ¼ 532 nm and thus the read wavelengths λR are chosen between 532 and 670 nm to match potential laser sources for the drive. For most of the write methods discussed later, the disk is flood-cured after the holographic exposure, leaving it insensitive to ambient light. The hologram plane-wave reflection efficiency is set at roughly 50% in order to achieve maximum return signal through the confocal read head and simultaneously to avoid significant optical depletion near the bottom of the disk. As shown in Fig. 2, only the read signal suffers this depletion since the majority of the beam is not Bragg-matched and thus its attenuation is not governed by the strength of the reflection grating. Writing power is therefore nearly independent of depth. The uniform grating causes the entire material volume to reflect a portion of the read beam and thus erasure or distortion of this format grating would be detectable as a reduction in the uniform reflected signal. However, positioning the head within this uniform volume would be challenging without the benefit of traditional servo marks. These cues can be provided by structuring the grating to have regions of high and low reflectivity in the shape of layers and tracks. This patterning can be accomplished by superimposing several mutually incoherent gratings such that the holographic fringes beat to form moiré patterns. Referring to Fig. 3(b), two reflection holograms with grating vectors normal to the surface but differing in magnitude by the spatial beat frequency of δK Z ¼ π=δZ come into and out of phase to create a layered structure with layer spacing δZ. Between these layers the two gratings are out of phase, suppressing the reflected signal. Thus these layers both provide servo information and reduce the extent of recorded features, reducing interlayer crosstalk and enabling higher storage density. Similarly, two gratings tilted in the radial dimension such that the radial component of their wave vectors are δkR ¼ π =ð2 δRÞ create cylindrical tracks with spacing δR. In the case of layer-only formatting described in the previous paragraph, the two gratings can be created via tilt of plane waves according to Eq. (1). The recording optics are more complex in this radial track case because the R coordinate
Fig. 3. Possible holographic servo patterns including (a) uniform reflection grating, (b) layer patterning in depth, (c) track patterning in radius and (d) simultaneous layer and track patterning. The arrows represent the incoherent grating vectors needed to create each pattern.
vector changes around the disk. The solution is to restrict the recording beams to ∼100 μm slices in the RZ plane and then to rotate the disk through this thin recording plane. The holographic fringes are parallel to the disk surface and thus, if the motion is sufficiently stable, will be faithfully recorded in the moving material. Finally, as shown in Fig. 3(d), three mutually incoherent gratings with the amplitude of the central hologram twice that of the two tilted gratings will record a pattern with both layer and track structures. Assuming that all the gratings are of the same pitch but tilted, as shown, the track spacing δR and layer spacing δZ are related by sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2ffi π 2π 2π 2π : ð2Þ ¼ − − δZ Λ Λ δR This relationship does not always yield the desired layer spacing, so one can also use angle tuning [via Eq. (1)] to adjust the pitch of the individual gratings to achieve nearly any combination of track and layer spacing. Note that tracks in adjacent layers are interleaved, which further decreases interlayer crosstalk, although at the expense of slightly more complex interlayer seek algorithms. B.
Track and Layer Holographic Servo Writer
According to the previous discussion, the fabrication of depth and radial tracks in a cylindrical coordinate system requires an interference pattern confined to a thin RZ slice through which the disk is rotated during format recording. This in turn restricts any tilt of 10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2699
Fig. 4. (Color online) Creation of the layer and radial holographic tracking pattern. As shown in (a), three pairs of mutually incoherent beams record three reflection gratings in a narrow radial spoke. The k-space representation (b) shows how the six green plane waves create the three desired grating vectors, tuned for readout at a longer wavelength.
the beams to the RZ plane, as shown in Fig. 4(a). These three tilts, as shown in Fig. 4(b), are then selected to achieve the three pattern specifications of detuning (δ) at the read wavelength (λR ), the track pitch (δR), and the layer spacing (δZ). The diffraction of the focused read beam off of the resulting grating structure is three displaced rings, similar to the single ring shown in Fig. 2. Note that for a thick material (large L) only three narrow rings are Braggmatched out of the incident cone and reflected back to become the read signal. The disk is therefore nominally transparent to the remainder of the angular spectrum, enabling the absorption of the writing mechanism to be adjusted independently of the disk reflectivity used for reading and tracking.
The three pairs of waves must be mutually incoherent to avoid the recording of undesired grating crossterms. Since the disk must be moved during formatting, sequential recording of the gratings would require an impractical level of stability from the system spindle. Simultaneous recording with mutually incoherent beams could be accomplished with three independent lasers but alignment and precise relative power control would be challenging. This motivates the use of a single laser and an acousto-optic (AO) beam splitter. We thus employ a tangentiallymatched anisotropic AO deflector fabricated from TeO2 . The incident beam is ordinarily polarized, rather than the typical extraordinary polarization intended for the deflector [11]. When the crystal is oriented for symmetrical diffraction, the incident beam diffracts off of the moving acoustic grating into both the positive and negative orders, causing the diffracted beams to be Doppler up- and down-shifted by the acoustic drive frequency. The diffracted beams and the remaining undiffracted zero-order are thus mutually incoherent for recording times longer than one over the acoustic frequency, typically ∼100 MHz. The relative beam angles and powers are precisely controlled by the acoustic frequency and power. As shown in the inset of Fig. 5, weak prisms are used to tilt the diffracted beams back to the incident propagation direction, and a half-wave plate is used to rotate the zero-order output polarization to match the diffracted beams. The remainder of the servo writer design is shown in Fig. 5. The target Strehl ratio for the drive and disk is 0.9, which is equivalent to a wavefront variation of λ=20. The three separated beams are
Fig. 5. (Color online) Optical ray trace of the holographic servo writer. The acousto-optic (AO) deflector creates three mutually incoherent beams which are expanded in the plane of the drawing and split by the beam splitter (BS) into three pairs stacked in depth. Pickoff mirrors direct these to tilted plates, which bring the beams back to the same plane and titled cylindrical lenses, focussing the interference pattern into a thin sheet normal to the along-track (S) direction. The disk to be exposed is rotated through this interference pattern by a precision spindle. 2700
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
anamorphically expanded to cover the radial extent of the disk with a four-stage prism telescope since plano-convex cylindrical lenses at this stage would introduce unacceptable aberrations. The three anamorphic beams, stacked normal to the plane of the drawing, are then split with a beam splitter to generate the three pairs of mutually incoherent beams. Pick-off mirrors send these beams toward the disk, and tilted parallel plates are used to bring them back to the same plane. Cylindrical lenses (which are now acceptable due to their proximity to the end of the optical train) focus the beams into 100 μm thick RZ sheets. These lenses are tilted to compensate for the refraction at the surfaces of the coupling prisms, which are necessary to reach the required beam angles within the material. These prisms are oil-coupled to the 1 mm glass substrates of the polymer disk which is rotated on an air-bearing spindle to record the patterns around the entire disk. A RZ slice of a disk formatted for reading at 532 nm with this hardware is shown in Fig. 6. Thirteen layers with a 7:85 μm separation are recorded into the 100 μm polymer thickness. The reflectivity of the tracks, as measured by a 0.5 NA Geltech aspheric objective corrected for the appropriate spherical aberration, is 5 × 10−5 and the contrast between the on and off track is at least 10:1. The entire recording process requires approximately 1 min with a 400 mW diode-pumped solid-state (DPSS) laser at 532 nm. As shown in Fig. 7, the pattern recorded is quite accurate and could be made more so with the removal of a systematic error, probably due to a stray reflection. At the start and end point in angle, there is a radial offset of up to 1=2 track, limited by
Fig. 6. RZ confocal scanning reflection measurement made by the precision drive of a 13 layer formatted disk. The 180 nm fringes that provide the reflection efficiency of each track are not visible. The pattern is sampled at 50 nm radially and 300 nm in depth.
Fig. 7. Precision of the radial track position of the tracks shown in Fig. 6. The plot shows the measured deviation of the track center from its nominal position as a function of the radius, indicating a maximum track squeeze of 200 nm.
mechanical stability of the exposure system and shrinkage of the polymer during development. 3.
Servo Detection and Locking
The purpose of the structured reflectivity is to provide a radial and depth coordinate system, to which the read/write focus will be locked by closed-loop servo. In order to maintain this lock, the length of a mark which erases the servo track must be restricted; this limit is common in many modulation codes. Like the data detection, the servo detection optics must be sensitive to the layer near the focus and reject information from all other layers, suggesting the use of a confocal filter. This confocal filter can easily be integrated into the radial tracking servo optics. Light from the data-channel confocal filter is imaged onto a detector split in the radial direction to implement a confocally-filtered version of a pushpull radial servo. The image of the tracks on the confocal filter shifts as the head moves radially with respect to the disk, and the difference between the left and right halves of this image yields a directional error signal for radial tracking. Alternatively, the split detector itself can be used as the confocal filter if its diameter is equal to the diameter of the read focus times the magnification of the detection optics. However, this typically requires a long focal length lens in the detection arm or an inconveniently small detector. The servo optics for detection of focus and layer position error are conceptually the same as those used for radial tracking in that the goal is to subtract the reflected intensity of the layer structure at two locations displaced above and below the focus depth. This can be accomplished by splitting the detected light with a beam splitter into two confocal detection arms and axially displacing the pinhole in each arm relative to the focus of the detection lens [12], as shown in Fig. 8. The two pinholes are now conjugate to planes just above and below the focus of the read/write head. Experimental results from this system are shown in Fig. 9 for the layer-only format shown in Fig. 3(b). The upper plot shows how the displacement of the peak confocal sensitivity causes the two return 10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2701
Fig. 8. Structure of the combined radial and depth confocal servo detection optics. The beam splitter divides the focused beam which is sent through two pinholes placed before and after the focus for a collimated input. Split photo-diodes detect the radial tracking error. The data, radial error, and depth error signals are computed from A þ B þ C þ D, ðB þ DÞ − ðA þ CÞ, and ðA þ BÞ− ðC þ DÞ, respectively.
signals to shift in phase. The electronic difference of the two photocurrents, shown in the lower plot, yields a directional error signal with a negative-slope zero crossing at the peak of each track and a positiveslope zero crossing in between layers. 4. Read/Write Mechanisms
At this point, we have a simple, single-headed optical drive mechanism that can lock the head radially and in depth to an efficiently exposed tracking structure in an otherwise homogeneous photopolymer disk. Reflectivity in this volume is provided by a reflection grating structure recorded as approximately 200 nm pitch sinusoidal polymer density variations. The goal of the write mechanism is to locally modify this structure such that the amplitude of the Bragg-matched reflection signal is detectably changed. The second key innovation of this architecture is that this writing operation is not in itself holographic and thus can be done with a single focused beam, vastly simplify-
Table 1.
Fig. 9. Depth servo signal measured for a layer-only format grating. The top plot shows the signals detected from the two displaced confocal filters while the bottom is the electronic difference, indicating a layer peak at every negative-slope zero crossing. The dashed line indicates the depth of a representative layer crossing.
ing the drive layout. In this section, we demonstrate three primary writing strategies: (1) destruction of the grating via a short pulsed laser, (2) inelastic distortion of the grating via local heating at the writing focus, and (3) erasure of the grating via continued polymerization in the formerly unexposed fringes of the hologram. In the second scheme, the grating is tuned both into and out of the confocal CTF, while in the third scheme, both linear and two-photon polymerization processes are investigated. The experiments and conditions are summarized in Table 1 [13,14]. For each process, the erasure of a preformatted reflective structure has an inherent nonlinearity, even when the recording mechanism is linear and integrating. That is, one can not erase the reflectivity below zero. While not as strong as the threshold nonlinearities common to existing optical storage products, this saturating response serves to improve
Summary of Experimental Conditions for the Results in Figs. 10–14
Deformation Ablation Figure Detuning Laser Write λ Peak Power Write Time Energy/Bit NA Spot Radius Bit Spacing Layer Spacing # of Layers Read λ
2702
10.b ∼0% Ti:sapphire 800 nm 100 kW 120 fs 12 nJ 0.38 1:26 μm 3 μm NA 1 633 nm
Negative 10.c ∼3% Q-switched 532 nm 230 mW 20 μs 4:6 μJ 0.38 0:84 μm 2 μm NA 1 532 nm
Continued Polymerization
Positive
Linear
10.d ∼0% DPSS 532 nm 25 W 40 ns 1 μJ 0.5 0:64 μm 1:5 μm 12 μm 5 532 nm
10.e 3% Diode 670 nm 40 mW 2 ms 80 μJ 0.5 0:8 μm 2 μm 20 μm 5 670 nm
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
2-Photon 12 0:35–6:35% α − DFB 659 nm 50 mW 3s 150 mJ 0.5 0:79 μm 1:5 μm NA 1 659 nm
(10.f,13) 3:3% α − DFB 659 nm 50 mW 10 s 500 mJ 0.5 0:79 μm 1:5 μm (NA, 12 μm) (NA, 8) 659 nm
14 2:5% DPSS 532 nm 100 mW 6 ms 600 μJ 0.5 0:64 μm 2 μm 20 μm 6 532 nm
the signal-to-noise ratio (SNR) of the written “ones.” The SNR of the “zeros” is determined by the quality of the formatting and is discussed later in Subsection 4.D. A. Ablation with a Femtosecond Laser
The conceptually simplest form of writing is to locally destroy the polymer grating with an ultrafast laser pulse. This is similar to writing sharply-defined structures in glass [5], except that the strength of the read signal is determined by the grating reflectivity which is independent of the precise shape or mechanism of the laser/material interaction. The
uniform grating, exposed at 532 nm as previously described, is tuned for readout at 633 nm. The photopolymer is then completely cured via exposure to room lights for 24 hours. The confocal reflectivity of this grating at 633 nm is shown in Fig. 10(a). The sample is then placed at the 0:38 NA focus of a Ti:sapphire laser. A line of bits with 3 μm spacing is written at λ ¼ 800 nm using a single motorized stage. The sample was removed and placed on an XYZ motion system and a focused HeNe laser was used to read the data as shown in Fig. 10(b). The grating is nearly completely erased where writing occurred (marked by the symbols “1”). The profile of the
Fig. 10. (Color online) Results of different write mechanisms showing detected signal versus along-track distance (S). All horizontal scales save (a) are the same for ease of comparison of bit shapes. The graphics on the right indicate the writing mechanisms which are ablation, inelastic deformation, and one- or two-photon continued polymerization from top to bottom. Further details for each plot are given in Table 1.
10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2703
reflected signal fits well to a simple model in which the contrast of the index grating is reduced proportionally to pulse intensity, however the precise mechanism of erasure is unknown. Since the absorption method is presumably nonlinear, this indicates some broadening of the recording area. Note that the energy per bit stored, 12 nJ, is significantly lower than the roughly μJ pulse energies used for writing into glass [15], presumably due to the lower damage threshold of the organic material. B. Inelastic Deformation with a Q-Switched DPSS Laser
While the damage at the focus of an ultrafast laser demonstrates the feasibility of this storage mechanism, the resonant structure of the Bragg grating provides other writing mechanisms that require significantly less peak power. For example, as shown in Fig. 11, linear absorption at the focus will heat and expand the material, locally increasing the grating pitch. As the grating spatial frequency changes, it moves through the CTF of the confocal read head, tracing out the Bragg selectivity curve including the side-lobes. Writing with higher power expands the material beyond its elastic limit, permanently deforming the grating pitch at the focus. Thus it should be possible to permanently decrease the local reflectivity of the format grating by increasing the pitch and detuning the grating spatial frequency out of the passband of the confocal CTF. Unlike the previous case, where the femtosecond laser pulse erased the grating, independent of its detuning, here the recording process is very sensitive to the grating pitch. Incoherent optical curing of the format grating naturally causes a depth-dependent variation of the detuning parameter due to Beer–Lambert absorption and polymerization shrinkage. Therefore the format holograms in this case were thermally cured at 65 °C for 20 h. For this demonstration, a Q-switched DPSS laser at 532 nm was chosen as the writing laser source and, for convenience, also was used as the reading laser. A uniform format grating was exposed with a roughly 3% detuning. To increase the optical absorption and
proportionally decrease the energy requirement per bit, the material was exposed to 200 J=cm2 at 313 nm after the normal curing which, via an unknown photochromic reaction, darkened the material. Elastic deformation dynamics like those in Fig. 11 were observed until the incident power passed the inelastic threshold, at which point the recovery during cooling was not complete. Figure 10(c) shows the reflectivity of a track after exposure to 20 μs pulses with peak power of 230 mW spaced at 2 μm. To confirm that the decrease in signal was due to deformation, not some other damage mechanism, pumpprobe curves like Fig. 11 were taken while focused on a previously-written bit. The dynamical behavior was the same except that it started from a new point in the Bragg curve, as expected. As indicated by Fig. 11, deformation can either decrease or increase the format reflectivity depending on the initial grating detuning, δ, relative to the CTF of the confocal read channel. Since the permanent deformation is always an expansion, to write bits with a higher reflectivity requires an initial grating that has a finer pitch than is optimal for the confocal microscope or, in our definition, δ ≈ 0. To further confirm that the deformation was purely local, five layers of such “positive” bits separated by 12 μm were written in a uniform format grating, as shown in Fig. 10(d). These are narrower than those shown in Fig. 10(c) because of the larger NA used and the greater slope of the confocal CTF at its outer boundary. Also in this experiment, the dependence of the writing energy on the pulse duration was explored by reducing the pulse duration by a factor of 500. As expected for this linear absorption process, the writing energy is a very weak function of this duration, dropping roughly by a factor of 5. The additional energy required at longer pulse durations is probably due to thermal diffusion. Although it was not fully explored in this experiment, the inelastic deformation of the material is expected to be a highly nonlinear function of temperature. Thus, although optical absorption which creates the temperature rise is due to linear absorption, the deformation of the material should be confined to a region near the focus. This does not result in a smaller bit size (like the two-photon absorption discussed below) but does limit interlayer crosstalk and the integration of out-of-focus response, which is a concern for completely linear processes, discussed next. C. Erasure via Linear Photopolymerization with a DPSS Laser
Fig. 11. Elastic deformation via heating with a 25 μs pump pulse from a Q-switched DPSS laser of a uniform format grating with detuning δ < 0. As the grating expands locally, the reflected signal traces out the Bragg selectivity curve. 2704
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
To decrease the peak power to the range of 50 mW and reach the goal of using a diode laser for writing and reading, we investigated a third method of modifying the grating reflectivity. In this mechanism, the contrast of the grating is reduced by initiating polymerization in the regions of the hologram that were located at intensity nulls during formatting. This erases the sinusoidal polymer density
variation and, thus, the index grating that provides the reflected signal. A single focused beam can initiate polymerization in the unexposed fringes if the hologram recording process strongly decreases the material sensitivity in the exposed fringes. This can be engineered simply by adjusting the constituent chemical concentrations such that some critical but nondiffusing species is completely consumed in the bright fringes and, conversely, remains for later use in the dark fringes. Note that the format grating now serves a dual purpose of modulating the index of refraction to provide a reflected signal and modulating the recording sensitivity such that a single focused beam can write a sinusoidal index pattern that is precisely out-of-phase with the original hologram. This sinusoidal sensitivity pattern is created by significantly depleting some essential chemical species only in the bright fringes of the initial hologram recording. In this case, both the green-absorbing photosensitizer 5,12-bis(phenylethynyl)naphthacene (BPEN) and the photo-acid generator (PAG) are reduced to roughly 20% of the Aprilis commercial formulation [2]. A second, red-sensitive dye is added to the material. Like the green-sensitive BPEN, this dye does not directly initiate polymerization but instead chemically excites the intermediate PAG. The concentration of this PAG is spatially modulated by the recording of the hologram in the green such that the later recording with a focused beam in the red will preferentially write in the previously unexposed regions of the hologram, reducing the grating contrast. The green recording angles are adjusted to write a grating for read/write at 670 nm. Unlike all other cases presented, the material is not cured and is thus linearly photosensitive and must be shielded from ambient light before writing data. Five layers of bits separated by 2 μm transversely and 20 μm in depth are written into the red-sensitive material by a 40 mW diode laser, a single track of which is shown in Fig. 10(e). The red pulse energy of 80 μJ is significantly higher than counterpropagating microhologram storage demonstrated in this same material [8], which achieved roughly 1 nJ exposure energies in the green; however, continued optimization of the red sensitizer could presumably narrow this gap considerably.
initiation, and to understand their dependence on the grating detuning parameter, we formatted a series of polymer samples at 532 nm for readout by a 659 nm tapered distributed-feedback (α-DFB) laser diode. After formatting, the remaining green sensitizer BPEN was completely bleached. To ensure that the PAG was not completely consumed by this process, only 20% of the nominal BPEN concentration was used in the formulation, similar to the procedure described in Section 4.C. After the roughly s ¼ 0:2% shrinkage during the bleaching process, the detuning, δ, ranged from 0.17% to 6.01% in the five samples. After bleaching, the plane-wave diffraction efficiency of the samples was 55 10%. Single planes of bits were written into these materials according to the conditions of Table 1. Because of the small two-photon absorption cross section, the exposure energies are the largest of any method investigated. The bit size and 659 nm confocal reflectivity versus detuning are shown in Fig. 12. The fullwidth to half-maximum bit diameter varies from 1:4 μm down to 0:87 μm as the detuning is increased while the reflectivity of the format hologram drops by a factor of 14. As shown in the sketch of the CTF to the right of the plot, when the detuning is small the available, transverse spatial frequencies are restricted, resulting in larger observed bit size. As the detuning increases, the transverse passband of the confocal microscope is broadened, resulting in smaller observed bits but at the expense of lower total efficiency due to the roll-off of the CTF. This rolloff is caused by the Gaussian filling of the objective and thus could presumably be improved with a tophat apodization. Based on this study, a detuning of ∼3% is optimum for multilayer experiments. Note that this is the midpoint of the confocal CTF for the 0:5 NA objective used. In this experiment, the format grating is structured with a layer-only format to demonstrate the simultaneous use of a structured grating and data storage. Eight layers separated by 12 μm are formatted and cured. Two-photon initiation at 659 nm of the structured PAG is used to locally erase the format hologram in spots separated by 1:5 μm in the plane. The precision drive is then used as a 3D confocal microscope to produce images of the reflectivity with fine sampling resolution. A depth (Z) and
D. Erasure via Two-photon Polymerization
Iodonium PAGs have moderate two-photon cross sections [16]. In Section 4.C we showed that a PAG can be structurally patterned by its consumption in the formatting stage. Rather than initiate this PAG hologram with a red-sensitive linear dye, it can thus be directly initiated via two-photon absorption. This will start polymerization only in the unexposed fringes of the format grating, erasing the reflectivity in a region defined by the square of the intensity of the writing focus. To quantify the size of these bits, which are expected to be smaller than those formed by linear
Fig. 12. (Color online) Full-width to half-maximum transverse bit diameter (upper curve) written by two-photon continued polymerization and reflectivity of the uniform format grating (lower curve) versus grating detuning. The sketch of the CTF on the right qualitatively explains the observed behaviors. 10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2705
along-track (S) slice of these data is shown in Fig. 13 (right). The slight index mismatch between glass and cured polymer makes these surfaces visible to the confocal head, while the layered pattern of the format grating appears as horizontal stripes. The writing locations appear as dark holes in these formatted layers since the contrast of the photopolymer grating has been reduced at these locations by the point excitation of the PAG, which is out-of-phase with the holographic index pattern. Based on these experiments, the two-photon cross section of the PAG at 659 nm was estimated to be αΠ ¼ 1:5 10−5 cm4 s. The two-photon cross section of the PAG was found to be significantly larger at 532 nm, so as a final experiment, a uniform grating tuned for 532 nm reading was fabricated with 2.5% detuning. Using 100 mW of 532 nm light in 6 ms pulses, six layers separated by 20 μm were written. Figure 14 shows the resulting reflectivity measurements in each of the bit planes. The writing energy of 600 μJ is roughly an order of magnitude higher than previous studies of two-photon data storage in polymers [5], consistent with the fact that the PAG has not been optimized for a two-photon cross section. These images reveal that the reflectivity of the unwritten region is modulated by sinusoids at two different periods and orientations. These noise gratings are presumably due to unwanted reflections in the formatting hardware or within the sample itself and explain the large noise observed in the “zero” levels of Fig. 10. 5. Summary and Conclusions
Here we have addressed three problems associated with volumetric data storage in photopolymers. First, data storage in thick, homogeneous volumes traditionally requires heads on both sides of the media, complicating the mechanical packaging or highly nonlinear material response, requiring the use of expensive ultrafast lasers. Second, these volumes provide no servo-tracking marks, making the transition from instrumented lab bench to inexpensive servo-controlled product difficult. Finally, sensitive and inexpensive holographic photopolymers are a promising candidate for these bit-based 3D storage architectures, but the linear index re-
Fig. 13. (Color online) Storage via two-photon continued polymerization initiated by a DFB laser diode in a 100 μm thick photopolymer formatted with eight layers. 2706
APPLIED OPTICS / Vol. 47, No. 14 / 10 May 2008
Fig. 14. (Color online) Storage via two-photon continued polymerization initiated by a 532 nm DPSS laser in a uniform format grating. The 51 ðRÞ × 51 ðSÞ × 6 ðZÞ bits were separated by 2 ðRÞ× 2 ðSÞ × 20 ðZÞ μm. Note the three broad horizontal bands and finer stripes at 45° across the upper left image that indicate the presence of weak noise gratings.
sponse is not obviously the best candidate for the system due to the accumulated out-of-focus response. We have solved the difficulty of two opposing read/ write heads by separating the holography from the write mechanism so that a single-head optical drive can both read and write from one side of the disk. By fabricating the precision grating in the factory, the drive does not require vibration isolation or a highly coherent laser, both hallmarks of traditional holographic data storage. We addressed the problem of servo-tracking by patterning the reflective holographic format with a small number of gratings that beat to produce radial tracks, depth layers, or both. A servo writer capable of fabricating this track and layer structure was designed and implemented for 120 mm diameter photopolymer disks. We implemented servo and data detection optics with confocal filtering and demonstrated closed-loop radial and depth tracking of deeply-buried marks in a spinning disk. We expanded the range of bit-based read/write mechanisms for holographic photopolymers by demonstrating five different schemes including femtosecond ablation, thermally-initiated inelastic detuning for negative or positive bits, and continued polymerization driven by spatially structured linear or two-photon absorption. The final two cases demonstrated the complex response that can be engineered in such “homogenous” volumes: the focused write beam excited polymerization with a 220 nm period that was precisely out-of-phase with and the same magnitude as the previously-written hologram. The surface storage density D in storage locations per micron2 of this method should be similar to that derived for microholograms [8],
D≈
π NA4 L : 10 λ3R n
ð3Þ
This limit was derived for a uniform medium, thus in the case of layer and track formatting that reduce the recorded bit dimensions, this density should increase. This suggests the possibility of TByte capacity in submillimeter thickness disks of 120 mm diameter. When this work was performed, the authors were with Siros Technologies Incorporated, San Jose, California. References 1. K. A. Rubin, H. J. Rosen, W. W. Tang, W. Imaino, and T. C. Strand, “Multilevel volumetric optical disk storage,” Proc. SPIE 2338, 247–250 (1994). 2. D. A. Waldman, R. T. Ingwall, P. K. Dhal, M. G. Horner, E. S. Kolb, H.-Y. S. Li, R. A. Minns, and H. G. Schild, “Cationic ringopening photopolymerimization methods for volume hologram recording,” Proc. SPIE 2689, 127–141, (1996). 3. L. Dhar, A. Hale, H. E. Katz, M. L. Schilling, M. G. Schnoes, and F. C. Schilling, “Recording media that exhibit high dynamic range for digital holographic data storage,” Opt. Lett. 24, 487–489 (1999). 4. T. Wilson, Y. Kawata, and S. Kawata, “Readout of threedimensional optical memories,” Opt. Lett. 21, 1003–1005 (1996). 5. J. H. Strickler and W. W. Webb, “Three-dimensional optical data storage in refractive two photon point excitation,” Opt. Lett. 16, 1780–1783 (1991). 6. Y. Kawata, R. Juskaitis, T. Tanaka, T. Wilson, and S. Kawata, “Differential phase-contrast microscope with a split detector for the readout system of a multilayered optical memory,” Appl. Opt. 35, 2466–2470 (1996).
7. S. Orlic, S. Ulm, and H. J. U. Eichler, “3D bit-oriented optical storage in photopolymers,” J. Opt. A: Pure Appl. Opt. 3, 72– 81 (2001). 8. R. R. McLeod, A. J. Daiber, M. E. McDonald, T. L. Robertson, T. Slagle, S. L. Sochava, and L. Hesselink, “Micro-holographic multi-layer optical disk data storage,” Appl. Opt. 44, 3197– 3207 (2005). 9. E. Walker, A. Dvornikov, K. Coblentz, S. Esener, and P. Rentzepis, “Toward terabyte twophoton 3D disk,” Opt. Express 15, 12264–12276 (2007). 10. Min Gu, Principles of Three-Dimensional Imaging in Confocal Microscopes (World Scientific, 1996). 11. T. Weverka, K. Wagner, R. R. McLeod, and K. Wu, “Low-loss acousto-optic photonic switch,” in Acousto-Optic Signal Processing, N. J. Berg and J. M. Pellegrino, eds. (Marcel Dekker, 1994). 12. M. S. Wang and T. D. Milster, “Differential wax-wane focus servo,” Appl. Opt. 32, 4797–4807 (1993). 13. L. Hesselink, “Three dimensional recording (3DR) technology,” in Optical Data Storage 2000, IEEE Conference Digest (IEEE, 2000), pp. 149–151. 14. L. Hesselink, R. R. McLeod, and S. L. Sochava, “Optical data storage by selective localized alteration of a format hologram in a holographic storage disk,” US Patent 6,614,741 (Sept. 2, 2003). 15. E. N. Glezer, M. Milosavljevic, L. Huang, R. J. Finlay, T.-H. Her, J. P. Callan, and E. Mazur, “Three-dimensional optical storage inside transparent materials,” Opt. Lett. 21, 2023– 2025 (1996). 16. K. J. Schafer, J. M. Hales, M. Balu, K. D. Belfield, E. W. Van Stryland, and D. J. Hagan, “Two-photon absorption cross sections of common photoinitiators,” J. Photochemistry and Photobiology A: Chemistry 162, 497–502 (2004).
10 May 2008 / Vol. 47, No. 14 / APPLIED OPTICS
2707
