Holographic data storage is often thought as a possi-ble replacement technology for the high-capacity op-tical disks, CDs, DVDs, and BRs. Very impressiverealizations are already close to a complete achieve-ment [1]. Nevertheless, various holographic arrange-ments exist and many are still on the race. Some ofthem are based on a bit-oriented approach: bits ofdata are stored by recording microholograms inthe focus of two counterpropagating beams [2,3].Although large capacities are expected with thesebit-oriented systems [4,5], we are interested in page-oriented approaches for which parallelism is thoughtto provide very high data rates [6]. In these architec-tures, data pages are imprinted on an image beamthat is subsequently recorded by holography in thevolume of the recording material by interfering itwith a reference beam. Thanks to Bragg selectivity,several pages of data can be recorded in the samevolume using a multiplexing procedure. The mostcommonly used is angular multiplexing and its deri-
vatives, such as shift or phase-coded multiplexings[1,7–10]. Nevertheless, wavelength multiplexing, inwhich a new page is recorded at a different wave-length, has also been investigated [11]. The mostcommon configuration for holographic wavelengthmultiplexing corresponds to a configuration in whichthe image beam and the reference beam enter the ho-lographic recording medium by two opposite faces.This arrangement is known as Denisyuk holography,a technique also often employed for color reproduc-tions of artworks in museography [12].
Besides these holographic approaches, data sto-rage architectures inspired from the principle ofLippmann interference color photography were pro-posed a long time ago. Incidentally Denisyuk holo-graphy was inspired by Lippmann photography[13]. In these Lippmann architectures, the incidentimage beam goes through the recording mediumand is reflected back by a mirror set in contact withthis recording medium, the image plane coincideswith the mirror plane. The incident and reflectedbeams interfere and record a complex Bragg struc-ture in the recording material. After this recordingstep, the mirror is removed and the material is pro-cessed. Illuminating the Bragg structure with white
light diffracts the recorded image with its whole ori-ginal wavelength spectrum. For color photographies,all spectral components are recorded at once, whilefor data storage, images at different wavelengths areusually recorded sequentially and correspond to dif-ferent sets of data.
Lippmann photography and Denisyuk holographyshare several common features. They both rely onthe recording of interferences between counterpropa-gating waves and exploit the wavelength Braggselectivity to record multiple data at the same loca-tion. The main difference originates from the refer-ence beam. In Lippmann architectures, the imagebeam itself acts as the reference beam. There is thusno external phase reference and this is the reasonwhy phase information cannot be recorded. Althoughrecording phase information can slightly increase thecapacity of holographic memories [14,15], most oftenthis information is useless and the reconstructedpages are grabbed with an intensity detector. There-fore, the intensity-coded data pages recorded in aLippmann system should be adequate for datastorage.
The first report on data storage inspired fromLippmann’s ideas dates from 1965 [16]. From then,several variants have been investigated [17–20]. Inall these demonstrations light corresponding to theadjacent data pixels does not interfere, either be-cause these pixels are sequentially recorded orbecause the material thickness is smaller than thedepth of field of the image. These architecturescan thus be said as being bit-oriented. It has beenshown that the capacity of these variants is rela-tively limited and does not benefit from the materialthickness [21]. To get higher capacities, similar tothose obtained with conventional holographic ap-proaches, one should switch to a page-orientedapproach as predicted in [21]. Besides these antici-pated high capacities, the research on Lippmanndata storage is further motivated by other expectedadvantages, such as
• a simple recording architecture,• a large stability of this recording setup,• required coherence lengths larger than twice
the recording layer thickness only,• an easy replication of the data from one master
disk to a blank one,• a simple implementation of a homodyne detec-
tion for increasing the diffracted signals.
In this paper, we thus investigate such a page-oriented wavelength-multiplexed architecture. First,from simple considerations of grating recordings, wedefine two simple rules for the conception of a datastorage system based on the Lippmann arrange-ment. Second, using these rules as a guide, we designa recording/readout optical setup. Experimentalresults of high-resolution data pages recorded in athick material are then reported.
2. Rules of Conception
A. Fidelity Problem
In case the depth of field of the image is smaller thanthe recording material thickness, Lippmann photo-graphy may suffer from a relatively low fidelity ofthe reconstructed images. This problem can be evi-denced by comparing Lippmann photography to De-nisyuk holography. The recording and readout stepsfor both these techniques are illustrated in Fig. 1. Be-side the image beam (whose electric field complexamplitude is noted A), Denisyuk holography makesuse of a (usually) plane wave reference beam, notedhere O, to record an interference pattern propor-tional toO�A, see Fig. 1(a). It is well known that read-ing out this Denisyuk hologram with the referencebeam, Fig. 1(b), reproduces the image beam with ahigh fidelity: OO�A ∝ A. This situation differs fromLippmann storage in which an incident beam, whosecomplex amplitude is noted B, interferes with its re-flection A onto a mirror, Fig. 1(c). This complex am-plitude B considerably differs from the plane wavereference beam used for reading out, once againnoted O as for Denisyuk holography, see Fig. 1(d).We decompose B in a Fourier series as B ¼OþP
bp, where O is the zero spatial order andPbp the sum over all other spatial orders; the zero
spatial order is indeed identical to the referencebeam used for reading out the data. The informationis thus reconstructed with a plane wave proportionalto O, so that the reconstructed amplitude isOB�A ¼ OðO� þP
b�pÞA. The first term of this sumcorresponds to the desired reconstructed data corre-sponding to the reflected image beam used for record-ing,OO�A ∝ A. The reconstructed wave also containsa second term corresponding to a large amount of
Fig. 1. (Color online) a) Recording scheme for Denisyuk holo-graphy with A and O the image and reference beams; (b) readoutscheme for Denisyuk holography; (c) recording scheme for Lipp-mann storage with image beam A resulting from the reflectionof the incident beam B onto the mirror; (d) readout scheme forLippmann storage, the mirror is replaced by an absorbing layer.e is the recording medium thickness.
b�pA. Thesewaves may correspond to noise and are absent fromDenisyuk holography. Consequently, in general, weexpect the fidelity of the retrieved image to be verylow. This happens if no special care is taken for con-ceiving the optical setup.
Using this plane wave decomposition, in the fol-lowing we derive two simple rules for designing apage-oriented Lippmann system for minimizing thisnoise. These simple rules will also allow us to esti-mate the tolerances of the optical setup.
B. Analysis of Lippmann Architectures for DefiningConstruction Rules
For a thin recording material, i.e., when the materialthickness is smaller than the image depth of field,there is no fidelity problem for recording binary in-tensity images. Indeed for a dark pixel, there is nolight at the corresponding location during recording.There is thus no grating and, therefore, the diffractedintensity vanishes for this location. For a bright pixela grating is recorded, but it does not contain any in-formation about the phase of the recording beam.Indeed, this grating is recorded by a stationary fringepattern resulting from the interference between therecording beam and its reflection onto a mirror. Theposition of the zero order fringe is fixed relatively tothe mirror (typically at a distance of one quarter of awavelength for a metallic mirror): the mirror acts asa phase locker. The phase of the reconstructed beamthus only depends on the phase of the readout beamand on the mirror position but not on the recordingbeam phase. Consequently, for a thin recordingmaterial, we conclude that the intensity of binaryimages is correctly reproduced although the phaseis not.
Conversely, for a thick material, even if a pixel isdark in the Lippmann mirror plane, contributions ofthe diffracted light for this pixel may originate fromparts of the recorded grating that are distant fromthe mirror by more than the depth of field. For sim-plicity, in analyzing the contribution of noise dueto the material thickness, we are going to separatethe Bragg matched diffraction processes from theunmatched processes and show that a Lippmann in-terference pattern can be described as the superim-position of two Denisyuk holograms and a set ofparasitic gratings. Let us once again break the com-plex amplitudes of the incident and reflected beamsunder the form
B ¼ OþX
bp and A ¼ O0 þX
ap: ð1Þ
Amplitudes O and O0 represent the zero spatial or-ders of the incident beam B and of its reflection A.A is the amplitude of the image beam we would liketo reconstruct. The interference pattern betweenthese two beams can be decomposed in three setsof reflection gratings:
B�A ¼ O�ð0:5O0 þX
apÞ þ ð0:5O� þX
b�qÞO0
þX
ap
Xb�q: ð2Þ
This decomposition in Fourier series allows us toidentify three different sets of recorded gratingsproduced by the interfering pattern B�A:
1. Set #1, O�ð0:5O0 þPapÞ, corresponds to the
recording of a first Denisyuk hologram betweenthe zero order of the incident beam and a beam car-rying the information of the original image beam Abut whose zero order is half the one of the originalimage beam A.
2. Similarly, set #2, ð0:5O� þPb�qÞO0, corre-
sponds to the recording of a second Denisyuk holo-gram between the zero order of the reflected imagebeam and a beam carrying the information of thebeam B but whose zero order is half the one of B.
3. Set #3,P
apP
b�q, recorded by interfering allorders different from zero, is a set of gratings thatdoes not exist in conventional Denisyuk holography.These gratings can thus be considered as parasiticgratings.
All these gratings diffract light during the readoutprocess when the material is illuminated by refer-ence beam O. Once again we can identify three setsof diffracted beams:
1. Wave diffracted by set #1 that is the firstDenisyuk hologram. This wave corresponds to theoriginal image beam A, except that its intensitylevels are slightly modified because its zero orderis half the one of the original image. We thus antici-pate gray level distortions in the retrieved data. Thisdiffracted image is at focus in the plane correspond-ing to the image plane of A (referred to as “Image”in Fig. 1).
2. Wave diffracted by set #2, that is the secondDenisyuk hologram. This wave corresponds to thephase conjugated beam of the original beam B with,once again, the intensity levels slightly modifiedbecause its zero order is half the one of the originalimage. This diffracted image is at focus in the planecorresponding to the image plane of B (referred to as“2nd image” in Fig. 1).
3. All waves diffracted by parasitic set #3. Thesewaves are not Bragg matched during the reading outby O. They are responsible for noise and speckle inthe retrieved images.
From this simple analysis we can draw a first im-portant conclusion. In order to retrieve the data fromthe diffracted light, the first Denisyuk hologrammust exist, that is the original image should havea nonvanishing zero order.
In order to get rid of the noise originating form theparasitic gratings, one should select images, i.e., datasets, for which the strength of these parasiticgratings (set #3) is much lower than the set of the
Denisyuk holograms, that is, images for which thezero order is much stronger than any other order.This can be achieved quite simply in a data storagesystem. Typically for binary images whose pixels areselected randomly between two equiprobable inten-sity values “0” and “1”, then half of the energy is con-tained in the zero order. The parasitic gratingsbecome negligible compared to the Denisyuk compo-nent of the Lippmann gratings.
In such a case, the main contribution to the dif-fracted light originates from the two Denisyuk holo-grams. As detailed above, diffraction on these twoholograms replays the two waves A and B�, withslight gray level distortions. These two waves propa-gate in the same direction and carry the same set ofdata. To do so that they are focused in the sameplane, one should do so that, during recording, theimage plane exactly coincides with the Lippmannmirror.
In addition, if the wave fronts of these images areflat in this mirror plane (no phase information), thenA and B are phase conjugated, i.e., A ∝ B�. The mir-ror acts as a phase locker and the two diffractedimages interfere constructively.
All these conclusions are supported by the compu-ter simulations we have conducted with a computersoftware described previously [21,22].
These simple considerations for designing aLippmann storage system are summarized by thesetwo following construction rules:
1. Rule #1: Record pure amplitude images with auniform phase.
2. Rule #2: Place the Lippmann mirror in theimage plane with an accuracy equal or better thanthe depth of field of the image.
It is important to note that the Rule #1 not onlyimposes to use pure amplitude images, but alsonecessitates conceiving an optical system whose exitpupil is set ad infinitum relatively to the Lippmannmirror.
3. Experimental Realization
A. Optical Setup
The scheme of our optical setup is depicted in Fig. 2.Its photograph is shown in Fig. 3. It is fed by a single-mode polarization-preserving fiber. The divergingbeam issued from this fiber freely propagates andgoes through the image mask, on which is displayedthe data page. It is then collected by a first video lens,turned by a polarizing beam splitter cube and thenpasses through a microscope objective. This micro-scope objective images the mask onto the Lippmannmirror. The video lens images the fiber output in theentrance pupil of the microscope objective. This en-sures that the wavefront of the beam in the mirrorplane is totally flat in spite of the curved wavefrontof the beam on the imagemask. If, in addition, we use
pure amplitude masks to input the data pages, thenthe construction Rule #1 is fulfilled.
Typically, when the amplitude mask is removedduring the readout procedure, the reading beam isa plane wave at normal incidence. The diffractedbeam passes through the microscope objective and,thanks to the combination of a quarter-wave plate(not shown in Fig. 2) with the polarizing beam split-ter cube, recorded data are imaged onto the CCDcamera by use of a second video lens.
The two identical video lenses are from Pentax(Model C7528-M, focal length 75mm). We selecteda microscope objective to minimize the field curva-ture and thus to remain within the specifications gi-ven by our Rule #2. It is a Plan Fluor ELWD 40×Cfrom Nikon whose focal length is 5mm. Its correctionring allows correcting for the spherical aberrationsintroduced by the presence of the thick recordinglayer on its glass substrate. We designed the systemso as the numerical aperture of the whole imagingsystem corresponds to the one of the microscope ob-jective that is equal to 0.6. Typically, with this nu-merical aperture at the wavelength of 532nm, thediameter of the Airy pattern at first nulls is 1:08 μm.
The CCD camera is an AVT Marlin F-201B with a1=1:8” CCD sensor from Sony. It is made of 1628 ×1236 square pixels of 4:4 × 4:4 μm2. This sensor sur-face corresponds to an observed image area in theLippmann mirror plane of 478 × 363 μm2.
Fig. 2. (Color online) Optical scheme of the recording/readoutsetup.
Fig. 3. (Color online) Photograph of the optical setup.
The focus of the microscope objective is adjusted bya piezoelectric translation stage. The Lippmannrecording substrate is mounted on a two-axis tiltplatform. With this three-axis adjustment we canmake the image plane coincide with the Lippmannmirror with an accuracy better than 2 μm over thewhole image field.
Ideally, a tunable laser source should be used torecord the gratings. Because we do not possess sucha source, instead, we used three lasers at three fixedwavelengths: (1) a single longitudinal mode, fre-quency doubled Nd:YVO4 laser at 532nm; (2) a mul-timode diode laser at 475nm; (3) a multimode diodelaser at 650nm. The two first ones are used forwavelength multiplexing, the last one in the red isused for the alignment procedure, the recordingmaterial being inactinic in red light. All three wave-lengths are injected in the single-mode fiber and areselected by a set of shutters. The coherence lengths ofthe two first lasers, green and blue, are much largerthan twice the recording layer thickness, which issufficient for Lippmann recording.
B. Recording Substrate
We selected a recording material considering twoantagonist requirements for the Bragg wavelengthselectivity:
• it should be strong enough to readout twosuperimposed gratings recorded at the two wave-lengths of 475 and 532nm without cross talk;
• it should nevertheless be loose enough to allowthe readout at the recording wavelength in spite ofthe unavoidable moderate shrinkage of the recordingmaterial.
We have selected a silver halide holographic gela-tin whose thickness is about 9� 3 μm [23]. Takinginto account the gelatin refractive index, this thick-ness corresponds to a Bragg selectivity (full width athalf-maximum for the amplitude) of about ΔλB ≈
10nm. As required, this value is smaller than therecording wavelength difference, 57nm, but large en-ough to accommodate for the shrinkage of about 1%we have with these silver halide plates.
The structure of the recording substrate is de-picted in Fig. 4. The sensitive gelatin layer is laidon a 1:6mm thick BK7 glass substrate. On top of thisstructure is glued a 100 μm thick, polycarbonate,achromatic quarter-wave plate. Although in our firstrecording tests we used an aluminum mirror set in
contact with the gelatin with an index matching li-quid, all results reported below were obtained byusing the Fresnel reflection on the gelatin–air inter-face as the Lippmannmirror. In spite of lower diffrac-tion efficiency, the use of this Fresnel reflectioninherently ensures that the mirror is in contact withthe gelatin. Because of the bad flatness of our glasssubstrate and gelatin layer we could not guaranteethis contact with the aluminum mirror. About2 cm below the gelatin–air interface, a light-trappingvelvet absorbs the transmitted light.
Material shrinkage occurs during the wet proces-sing. We measured it by recording reflection gratingswith planes waves. By monitoring the diffraction ef-ficiency versus the wavelength of the chemically pro-cessed plates with a Cary 14 spectrophotometer, wemeasured a shrinkage of about 1%. These gratingscan thus be readout at the recording wavelength,as desired for our demonstration.
Before reading out the recorded structure, the ge-latin surface is coated by a black absorbing paint inorder to remove the Fresnel reflection.
C. Data Pages and Masks
Ideally, the data pages should be imprinted on theoptical beam by use of a spatial light modulator. Fortesting our setup, we prefer to use chromium maskson a glass substrate. Because of their relatively lowcost, chromium masks are indeed totally versatile;the pixel pitch can be changed arbitrarily by chan-ging the mask. Furthermore, their ON/OFF ratioof intensity is much larger than the dynamic rangeof the CCD camera. With these chromium maskswe insure that the measured signal-to-noise ratioof the retrieved image is neither limited by the con-trast ratio of the spatial light modulator, nor by itsflicker.
As described in the previous paragraph “Analysisof Lippmann architectures” we anticipate moderatecontrasts in the retrieved images and gray level dis-tortions as well. Therefore, our data pages are en-coded with binary pixels (ON ¼ transparent andOFF ¼ opaque) using bloc coding [24,25]. We use a(3,16) coding, that is, each code (from 0 to 255) of eachASCII character (Byte) of the file to be encoded onthe data page is represented by a bloc of 4 × 4 pixelsamong which three are on the ON state. A typicaldata page is shown in Fig. 5. It is divided into sub-pages in the center of which the square feature isa fiducial pattern used to estimate, during the decod-ing of the diffracted image, the image spatial distor-tions (magnification, rotation, and shifts). Thesepatterns thus allow compensating for these distor-tions. Once these distortions are compensated, foreach bloc, we determine the three pixels bearing thehighest signals and thus attribute an ASCII value forthis bloc. Comparing the decoded ASCII file with theknown original ASCII file, we calculate the bloc errorrate, BLER, defined as the ratio of the number oferroneous blocs to the total number of blocs.
Fig. 4. (Color online) Recording substrate arrangement. Incom-ing light is partly reflected by the interface air/sensitive layer;transmitted light is blocked by the light-trapping velvet.
In order to make its features visible for the reader,the resolution of the page, shown in Fig. 5, is low.It contains 40 × 32 blocs only. However, we testedour setup with data pages of higher resolutions, upto 128 × 96 blocs, whose information capacity is11:5kByte, discarding the fiducial patterns. Forthese high-resolution masks, the pixel pitch is0:88 μm in the Lippmann mirror plane.
Replacing the recording substrate by an aluminummirror, we checked the quality of our setup with thehigh-resolution image detected with the CCD cameraafter being reflected on the mirror, but without anyrecording. We find a bloc error rate of BLER ≈ 0:01,corresponding only to errors located on the edges ofthe 11:5kByte page, thus demonstrating the opticalquality of our setup.
4. Experimental Results and Analyses
In Fig. 6, we show a data page recorded and detectedat 475nm. The pixel pitch of the data mask is2:64 μm. This corresponds to an image depth of fieldof 20 μm that is larger than the recording materialthickness. The retrieved image quality is high. Forsake of comparison, the original data mask is theone previously shown in Fig. 5. We got a bloc errorrate of BLER ≈ 810−4, corresponding to a single error
due to a single speck of dust. We checked that for thisresolution absolutely no cross talk is visible betweentwo images superimposed at the same location andmultiplexed at the two wavelengths 475 and532nm. Thanks to the wavelength Bragg selectivity,they can be retrieved without ambiguity at theirrecording wavelength.
We also tested our setup with the highest availableresolution, that is, with a data mask whose pixelpitch is 0:88 μm. This pixel pitch is smaller than thediameter of the Airy pattern, whose diameter is1:08 μm at the wavelength of 532nm. The depth offield of the image is now about 2 μm. This value isnow much smaller than the holographic layer thick-ness. After recording such a data page at 532nm andprocessing the plate, we obtained during the readoutprocess the raw image shown in Fig. 7. After decod-ing, we found that the error rates are quite nonuni-form among this page. Some subpages are relativelywell decoded, while for other ones the error rate re-mains high. We attribute these inconsistent resultsto variations in the thicknesses of the plastic quarter-wave plate and of the holographic emulsion. Indeed,because of these variations in thickness, it is not pos-sible to exactly make the image plane coincide withthe mirror plane (in practice, the interface “gelatinlayer–air”) over the whole image field. Our concep-tion Rule #2 cannot thus be fulfilled all over thisfield. This explanation for our relatively inconsistentresults is supported by our computer simulations: assoon as the image plane moves away from the mirrorplane by more than or about 2 μm, the detectedimages are blurred and cannot be decoded.
In spite of this problem whose origin is now iden-tified, some of the subpages are correctly decoded.They correspond to locations where the focus is cor-rect. For instance, in Fig. 8(a), we show one of thesubpages of the image detected by the CCD camera.In this image, the image pixel pitch is about 0:88 μmin the mirror plane. Taking into account the ×15magnification of the optical system, this value corre-sponds to 13:3 μm in the CCD plane, that is threetimes the size of a CCD pixel. After compensation
Fig. 5. (Color online) Typical data page with (3,16) bloc coding.This page is made of 5 × 4 subpages made each of 8 × 8 blocs.Fiducial patterns are located in the center of each subpage.
Fig. 6. Image grabbed by the CCD camera corresponding to adata page recorded and retrieved at 475nm. The pixel pitch is2:64 μm.
Fig. 7. Image grabbed by the CCD camera corresponding to adata page recorded and retrieved at 532nmwith the smallest pixelpitch of 0:88 μm.
of the spatial distortions (rotation, magnification,and shift) determined from the positions of the fidu-cial patterns in all subpages, the average value of anensemble of 3 × 3 pixels of the camera is attributed tothe value of one pixel of the data image. The shiftsare compensated within one CCD pixel; we did notemploy any optimized sophisticated resamplingtechnique that could further enhance the perfor-mances of our system [26]. We thus obtain the graylevel image shown in Fig. 8(b). In each bloc made of4 × 4 pixels we then seek for the three pixels bearingthe highest values. These three pixels are then set to“1” and all others to “0”. This binarized image isshown in Fig. 8(c). It has to be compared to the ori-ginal subpage, shown in Fig. 8(d).
The decoded subpage, Fig. 8(c), is thus very similarto the original one, Fig. 8(d). A detailed comparisondemonstrates that they differ by only 3% of the bits ofthe image data, corresponding to a bloc error rate ofBLER ≈ 0:15. This value for a bloc error rate is rela-tively large and should be compensated by the use ofefficient error correction codes. Typically, the use oftwo successive error correction codes, the first on theinitial data stream to be distributed between severalpages and the second at the page level, should beused to reduce the bit error rate on the user file muchbelow 10−12, value compatible with a memory capa-city of a few terabytes [1].
Some of these remaining errors originate fromsmall specks of dust and are thus not inherent fromthe Lippmann technique. This nonperfect bloc errorrate should thus be improved with a cleaner materialand/or by using a material with higher diffractionefficiency. Under our experimental conditions, thediffraction efficiency is indeed about 1%: scattered
light thus greatly contributes to the relatively lowquality of the images.
Nevertheless, these results clearly demonstratethat high-resolution data pages can be recorded inthick holographic media using the Lippmann princi-ple. We are indeed very close to the theoretical reso-lution of the imaging system: the selected pixel pitchis smaller than the Airy pattern diameter. Further-more, the error rate is mainly limited by imperfec-tions and should be easily improved with a betterrecording substrate.
We checked that, together with this high resolu-tion, the Bragg selectivity of the recorded gratingis high enough for wavelength multiplexed images.Typically, reading out at 475nm, the grating corre-sponding to the image shown in Fig. 7 does not giveany observable signal (i.e., signal above the noiselevel).
5. Conclusion
We have defined two simple construction rules fordesigning a page-oriented Lippmann data storagesystem. From these rules we anticipated the correctoperation of such a storage system although distor-tions of the gray levels of the retrieved images areexpected. Using these rules we have built an opti-mized recording setup. With this system we haveindeed demonstrated that the resolution of the datapages is just limited by the numerical aperture of theoptical system: data pages with a pixel pitch of0:88 μm were retrieved in an optical system whoseAiry disk diameter is 1:08 μm and for which the im-age depth of field is smaller than the recordingmaterial thickness. At these high resolutions, theBragg wavelength selectivity is preserved.
The price to pay for the apparent simplicity of theoptical setup (absence of any external referencebeam) is a very severe requirement on its design andadjustment: the image plane should coincide withthe Lippmannmirror over the whole image field withan accuracy better than the image depth of field.
These analyses and experimental results corrobo-rate the previous conclusions of computer simu-lations that predicted that the data capacity ofLippmann storage is comparable with the capacity ofmore conventional holographic approaches providedthat special care is taken during the conception andadjustment of the experimental setup [21]. Thesenew results should thus contribute to the renewalof interest for this elegant technique, which presentsan enormous potential for high-capacity storage.
K. Contreras gratefully acknowledges the supportof the Programme Alban, the European Union Pro-gramme of High Level Scholarship for Latin America(E07D401978PE). The research described here hasbeen supported by Triangle de la Physique contractMIMEA.
Fig. 8. Decoding procedure of a subpage recorded at 532nm, re-trieved at 532nm, and detected with the CCD camera. The pixelpitch is 0:88 μm. (a) Subpage grabbed with the CCD camera with-out any processing; (b) after compensation of distortions, rotation,magnification, and shifts (see text); (c) after decoding; (d) compar-ison with the original subpage before recording.
References1. K. Curtis, L. Dhar, A. Hill, W. Wilson, and M. Ayres, Holo-
graphic Data Storage: From Theory to Practical Systems(Wiley, 2010).
2. H. J. Eichler, P. Kuemmel, S. Orlic, and A. Wappelt, “High-density disk storage by multiplexed microholograms,” IEEEJ. Sel. Top. Quantum Electron. 4, 840–848 (1998).
3. R. R.McLeod, A. J. Daiber, M. E. McDonald, T. L. Robertson, T.Slagle, S. L. Sochava, and L. Hesselink, “Microholographicmultilayer optical disk data storage,” Appl. Opt. 44, 3197–3207 (2005).
4. T. Tanaka and S. Kawata, “Comparison of recording densitiesin three-dimensional optical storage systems: multilayered bitrecording versus angularly multiplexed holographic record-ing,” J. Opt. Soc. Am. A 13, 935–942 (1996).
5. R. R. McLeod, “Impact of phase aberrations caused by multi-layer optical data storage in weakly inhomogeneous media,”J. Opt. Soc. Am. B 26, 308–317 (2009).
6. S. S. Orlov, W. Phillips, E. Bjornson, Y. Takashima, P.Sundaram, L. Hesselink, R. Okas, D. Kwan, and R. Snyder,“High-transfer-rate high-capacity holographic disk data-storage system,” Appl. Opt. 43, 4902–4914 (2004).
7. K. Tanaka, M. Hara, K. Tokuyama, K. Hirooka, K. Ishioka, A.Fukumoto, and K. Watanabe, “Improved performance incoaxial holographic data recording,” Opt. Express 15,16196–16209 (2007).
8. H. Horimai, X. Tan, and J. Li, “Collinear holography,” Appl.Opt. 44, 2575–2579 (2005).
9. Y. Taketomi, J. E. Ford, H. Sasaki, J. Ma, Y. Fainman, and S.H. Lee, “Incremental recording for photorefractive hologrammultiplexing,” Opt. Lett. 16, 1774–1776 (1991).
10. C. Denz, G. Pauliat, and G. Roosen, “Volume hologram multi-plexing using a deterministic phase encoding method,” Opt.Commun. 85, 171–176 (1991).
11. G. A. Rakuljic, V. Leyva, and A. Yariv, “Optical data storageusing orthogonal wavelength multiplexed volume hologram,”Opt. Lett. 17, 1471–1473 (1992).
12. Y. Denisyuk, “The imaging of the optical properties of an ob-ject in a wave field of radiation scattered by it,”Opt. Spectrosc.15, 279–284 (1963).
13. Y. N. Denisyuk, “Imaging properties of light intensity waves:the development of the initial Lippmann ideas,” J. Opt. 22,275–280 (1991).
14. B. Das, J. Joseph, and K. Singh, “Phase modulated gray-scaledata pages for digital holographic data storage,” Opt. Com-mun. 282, 2147–2154 (2009).
15. Z. Göröcs, G. Erdei, T. Sarkadi, F. Ujhelyi, J. Reményi, P.Koppa, and E. Lorincz, “Hybrid multinary modulation usinga phase modulating spatial light modulator and a low-passspatial filter,” Opt. Lett. 32, 2336–2338 (2007).
16. H. Fleisher, P. Pengelly, J. Reynolds, R. Schools, and G.Sincerbox, “An optically accessed memory using theLippmann process for information storage,” in Optical andElectro-Optical Information Processing (MIT, 1965), pp. 1–30.
17. A. S. Hoffman, “Optical information storage in three-dimen-sional media using the Lippmann technique,” Appl. Opt. 7,1949–1954 (1968).
18. G. Maire, G. Pauliat, and G. Roosen, “Homodyne detectionreadout for bit-oriented holographic memories,” Opt. Lett.31, 175–177 (2006).
19. J.-J. Yang and M.-R. Wang, “White light microgratingmultiplexing for high density data storage,” Opt. Lett. 31,1304–1306 (2006).
20. F. Guattari, G. Maire, K. Contreras, C. Arnaud, G. Pauliat, G.Roosen, S. Jradi, and C. Carré, “Balanced homodyne detectionof Bragg microholograms in photopolymer for data storage,”Opt. Express 15, 2234–2243 (2007).
21. K. Contreras, G. Pauliat, C. Arnaud, and G. Roosen, “Applica-tion of Lippmann interference photography to data storage,”J. Europ. Opt. Soc. Rap. Public. 3, 08020 (2008).
22. G. Pauliat and K. Contreras, “Experimental investigation of apage-oriented Lippmann “holographic” data storage system,”Proc. SPIE 7730, 773004 (2010).
24. G. W. Burr, J. Ashley, H. Coufal, R. K. Grygier, J. A. Hoffnagle,C. M. Jefferson, and B. Marcus, “Modulation coding for pixel-matched holographic data storage,” Opt. Lett. 22, 639–641(1997).
25. B. Marcus, “Modulation codes for holographic recording,” inHolographic Data Storage, H. J. Coufal, D. Psaltis, and G. T.Sincerbox, (eds.), Springer Series in Optical Sciences(Springer-Verlag, 2000), pp. 283–292.
26. M. Ayres, A. Hoskins, and K. Curtis, “Image oversampling forpage-oriented optical data storage,” Appl. Opt. 45, 2459–2464(2006).
