Sandwich holography for storing information interferometricallywith a high degree of security
Nils Abramson, Hans Bjelkhagen, and Per Skande
An entirely new method of storing confidential data is presented. The method is based on sandwich holog-raphy, and its-main principle is the following: The basic state of a surface is recorded in one plate of a sand-wich'pair; in the other plate the same surface is recorded slightly deformed. When the two holograms arecombined an interference fringe pattern is formed. By controlling the deformation of the surface it is possi-ble to obtain exactly the interference pattern wainted. This controlled pattern can be used to create infor-mation that is either directly readable digital information or a binary coded information, suitable to auto-matic readout. The data stored in the holograms are not possible to read out from one hologram alone, butthe two holograms must be put together to make the information available. The system is primarily meantto be applied on ID cards but can also be used for other applications where data have to be stored securely.
Introduction
The need for more secure ID cards has increasedrapidly during the last years. Many new ideas usingdifferent optical methods have been suggested forstoring data in a way that makes forgery more difficult.These methods are often based on different ways toscramble an image of the stored information. Using thesame methods in reverse it is possible to unscramble theinformation.
In the suggested holographic solutions'- 6 often acoded reference beam is used to create the scrambledimage. Optical filtering based on Fourier hologramsfor pattern recognition has also been suggested.7-' 0
Sandwich hologram interferometry1-' 4 has mainly beendeveloped for measurements of deformation and vi-bration where this method has some advantages com-pared with conventional hologram interferometry.
The possibility of using sandwich holography as acarrier of confidential information has only been men-tioned briefly by Abramson.'5 In this paper the methodwill be described more in detail, along with an investi-gation that was performed to show how letters can bestored.
Per Skande is with Lasergruppen Holovision AB, S-115 28 Stock-holm, Sweden; the other authors are with Royal Institute of Tech-nology, Division of Production Engineering, S-100 44 Stockholm 70,Sweden.
In conventional sandwich holography used for mea-surements, two holographic plates with the emulsionfacing the object are exposed simultaneously in a specialholder. Then the object is deformed. A new pair ofplates are put in the holder, and an exposure of the de-formed object is performed. After development, theback plate from the first pair is combined with the frontplate from the second pair, and they are glued to-gether.
The interference fringes produced in the sandwichhologram, due to the deformation, can be controlled bytilting the sandwich hologram at reconstruction.Hereby it is possible to compensate for rigid body mo-tions or to determine the sign of a displacement.
Instead of using two pairs of unexposed plates at eachexposure only one holographic plate combined with aclean compensating glass plate of the same thickness asthe holographic plate can be used. At the first exposurethis glass 'plate is positioned in front of the unexposedplate, and at the second exposure the glass plate isplaced behind the holographic plate.
When the sandwich method is used for storing datathe information is produced by interference when thedifference between two images is measured with inter-ferometric precision during reconstruction.
The information is built up by phase differences in-troduced in laser light. As these phase differences aremuch smaller than the random phase variations existingon the surface which is recorded, it is not even theoret-ically possible to find any hidden information using onlyone plate, as no information, coded or uncoded, existson any plate. To create the difference in optical pathlength, resulting in interference fringes, different
EFig. 4. Reconstructed sandwich-hologram information directly
projected on a screen.
Fig. 1. Principle design of the 7 X 7 piezoelectric module: (B) baseplate in steel; (P) piezoelectric crystal; (H) carbide cover on top of thepiezoelectric crystal to eliminate local deformation on the surface of-
the crystal; (E) electrical wire.
HOLOGRAM?
GLASS PLATE
LASER LE OBJECT
ZIJ = ,/EFERENCE MIRROR
GLASSP LATE
~-HOLSRAM 2
LASER SPATIAL
__~~~~~~ __
-t iOBJECT
_ A~FERENCE MIRROR
_ _ i11
Fig. 2. Recording of the information.
LASER EAM
HOLOGRAM 1 HOLOGRAM 2
RECONSTRUCTEDIMAGE
Fig. 3. Reconstruction of the information.
methods can be used: either by translating or de-forming the code surface locally or by putting opticalcomponents, which could change their refractive indexor their optical path length, in the way of the light.
The special object built for our research experimentuses piezoelectric crystals to translate the individualparts of the object surface. The displacement is theorder of one quarter of the wavelength of the used laserlight (0.16 ,m for a He-Ne laser). The test object is amodule consisting of 7 X 7 pixels (piezoelectric ele-ments), as shown in Fig. 1.
When the object is reconstructed from one singlehologram, all the 7 X 7 pixels appear bright. If, how-ever, one hologram (information carrier) is combinedwith another hologram (master) the information is re-vealed because some pixels become dark (pixels movedbetween recordings), while the others remain bright(unmoved pixels). The dark pixels are created bytranslating the surface exactly so that destructive in-terference between the two images is obtained at re-construction of the sandwich hologram. The very finemovement of the individual pixels is caused by applyinga high voltage (dc) from a stable power supply over thepiezoelectric crystals.
If we want to store, for example, the letters i d in thesandwich hologram, we make one hologram, the master,of the module when all the pixels are unactivated anda second hologram, the information carrier, when thepixels forming the letters i d are activated (translatedaround 0.16 um), see Fig. 2. When the holograms aredeveloped the information can be reconstructed byletting an unexpanded laser beam pass through thesandwich hologram so that the letters i d are projectedon a screen, see Figs. 3 and 4. From Fig. 4, showing aphotograph of the projected image, it is seen that theimage is affected by speckles. This can be reduced, andthe contrast can be improved by using a TV system forthe reconstruction process. Then the image from thesandwich hologram is projected on a videocone of a TVcamera so that the information can be seen on a TVscreen.
Hereby the contrast and image quality can be ma-nipulated electronically to obtain a high quality picture
Fig. 5. Sandwich hologram information with improved image quality by means of a TV system.
Table 1. Comparison. Between Different Recording Methods ShowingTheir Influence on the Measured Displacement Between Sandwich Plates
at Reconstruction
Fig. 6. A hologram on film laminated in an ID card combined witha master plate. On a screen behind them is the reconstructed infor-
mation seen.
of the information on the screen (Fig. 5). Instead ofmaking both recordings on glass plates one recordingcan be made on a holographic film. This makes itpossible to laminate the information carrying hologramin an identity card. Combined with the master platethe information can be read out as described earlier.Using suitable plastic material and methods for em-bossing, almost as good results can be obtained as usingtwo glass plates (Fig. 6).
Experimental Results
As our method is based on small phase differences(X/2) between two separate objects images of a codesurface, it is very important for easily readable infor-mation that the positioning of the information carrierhologram on top of the master is made with sufficientaccuracy. A slight positioning error between the twoholograms would result in the interference informationcaused by shifts being affected by a superimposed
7LOi,)G RECORDI:GARRANGEMENT
¶fifc'!lrF1 rlorrectTrrto Oil lil1 s-, AND Z- DIRECTIONS BETWEEN SANDWICHPLATES
r 10 fringes visib- Maximum displacementsin on the 7 7 with still resolvablemodule fringes on the 7 x 7
Z IXYY X Y Z I 0
Off axis hologram 8
1. 40 68 300 330 370 1750o-1400 mm r-2540 -m
Off axis hologram a 200
2. 21 37 230 130 142 1400o = 530 mm r = 1940 no
Focused image hologram -=7_
3. 38 53 260 300 350 -o-0 r 15900
40
4. . 83 153 1517 330 387 -o - 1145 no r =
Parallel reference beam
5. 31 60 272 336 308 1410o = 1110 no r =
Parallel object beamo-130
6. 29 63 100 430 395 1950o = r = 845 no
lensless fourier hologram|
7. 34 64 414 280 310 21009 95 no r 95 n
Fourier hologram a = 120
8. 28 57 117 413 297 777
parallel fringe pattern due to the misalignment. Toestablish how accurately the holograms have to be po-sitioned and to compare different hologram recordingmethods, 1761 measurements have been made on 143holograms using several different recording methods,see Table I. For this investigation special evaluatingequipment was constructed where translations betweenthe two holograms in a sandwich pair can be measuredin x-, y-, and z-directions (Fig. 7).
The best results were obtained by using hologramsof type 6, in which case the largest displacement thatresulted in observable fringes were: x = 430 Arm, y =395 Aim, and z = 1950 jim. At this misalignment it isstill possible to see fringes on the module, and by movingthe hologram in such a direction that the number offringes decreases to zero, the alignment process to ob-tain the information is then completed.
There is, however, no doubt that this method wouldbe complicated and expensive applied to a commercial
Fig. 7. Evaluation equipment for measuring the permitted dis-placement between the holographic plates in the x-, y-, and z-
directions.
system. Thus a new method utilizing the uniqueproperties of the sandwich hologram has been tried.Instead of presenting the information as carried by darkand bright pixels, it is carried by pixels with or withoutfringes.
Data Storing Based on Interference Destruction
In this case some of the pixels are translated "somuch" that no interference fringes can be obtained onthem when reconstructing the sandwich. However,some of the pixels can still produce interference fringesas they are unchanged. Now there is no need for posi-tioning the plates so accurately that the number offringes decreases to zero, but the positioning is satisfyingif the interference fringes can be resolved on the pixels.This system has many advantages compared with thesystem described earlier. The alignment can be lessexact, and there is no need to find the zero fringe posi-tion. Compared with the system described earlierwhere the zero fringe position must be aligned with anaccuracy of about 4 gim, this system can accept around100 times larger misalignment (Fig. 8).
Another advantage of this system is that it is very wellsuited for automated readout. The image of the mod-ule then is projected onto an array of light sensitivedetectors positioned exactly as the pixels. By slightlymoving one plate in relation to the other, the fringepattern starts moving. The detectors will then delivereither an ac or dc signal depending on whether they aredetecting a pixel with or without fringes.
*~~~~~~~~~~~~~- 11
s Ax~~~~itss~l $!! !! ! ! 1!~~~~~~~~~
Fig. 8. Photographs showing the advantage of using the interference destruction method (upper 7 pixels): (a) aligned plates, (letters i d easilyseen); (b) slightly misaligned plates (difficult to read the letters i d but detectable binary code at the upper 7 pixels); (c) misaligned plates
[almost impossible to read the letters i d; easily detectable binary code (1 0 1 00 11)].
When this type of readout is used it is most conve-nient to have the data stored in a binary form insteadof an alphanumerical representation. Using a binarycode the information capacity increases enormously,and thus one single 7 X 7 module can represent around1015 (249) different states. However, one might thinkthat the security level of this system would be lowersince the pixels have to be translated so much that theinterference is destroyed. Would it be possible in asingle plate to observe the difference in positions of thepixels without the interference effect from the masterhologram?
One way to destroy the interference effect could beto rotate the pixels, which of course could be circularinstead of square. If the pixels are uniformly paintedwhite no change could be seen however large the rota-tion. However, let us suppose that there is a mark onthe pixels so that its angle can be seen. Would thenecessary rotation be observable?
Let us assume that if the pixel is covered by morethan 200 fringes it would appear fringe free as comparedwith the other pixels. If the pixel is 1 cm in diameterand illuminated at an angle of 45°, it would have to berotated so that its periphery moves a distance of 100 X2 X A = 0.09 mm.
Can this motion of about 0.1 mm be seen from a dis-tance L of 1 m? If we used a diffraction-limited tele-scope to study the holographic image of the pixel, thediameter d of the lens is needed to resolve the mo-tion:
Discussion
The method discribed for storing confidential datais in our opinion the most secure method availabletoday. It is suited for protecting ID cards, credit cards,passports, and other credentials against forgery. Butit is also extremely useful for sending confidentialmessages (for instance, military messages) and forstoring personal data with good protection for individ-ual integrity.
It is quite possible for one information carrier holo-gram to be used in combination with several differentmasters. A given master then only will be able to accessthe appropriate parts of the information contained inthe information carrier. For example, one master maymake it possible to read entry control data, anotherpurely medical data, a third the credit rating of aperson.
Other advantages of a holographic based securitysystem are that different types of holographic infor-mation can be stored in the same hologram. For ex-ample, our system can easily be combined with theSiemens Holo-Secure-system,1 6 where a picture of theidentity card is holographically stored in the card to becompared with the card itself.
One disadvantage with all holographic systems, ofcourse, is the extra cost of producing the holograms andthen laminating in the card. Also a special readingdevice is needed at every place the information is to becontrolled. However, this reading equipment can becompletely automated and easily connected to a com-puter for data checking.
X x L 0.6328 X 10-3
X 103
0 = - 6 mm.0.1 0.1
Thus the rotation could just be observed if the diameterof the hologram is larger than 6 mm and if a telescopeof absolutely highest possible quality (diffraction-lim-ited resolution) is used.
Let us now study what happens if we instead let thepixel tilt, e.g., rotate around an axis in the plane of itssurface. The needed tilt would result in one point atthe periphery moving 0.063 mm toward the hologram,if the pixel is illuminated from the front. The depthresolution of a telescope is much lower than its trans-versal resolution. In our case the telescope needs amotion of no less than 33 mm before the motion is re-solved. Thus there is a security margin of around 500between the needed displacement of 0.06 mm and thedetectable displacement of 33 mm (with a hologram 6mm in diameter).
Thus we have discovered that the rotation around anin-plane axis does not produce a depth difference largeenough to be resolved by the telescope with a 6-mmaperture. Could perhaps the ellipticity of the tiltedpixel be resolved instead?
Our calculation shows that the ellipticity will be onlyof the order of 2 X 10-6 mm. The smallest transversaldisplacement that would be resolved was 10-1 mm, andthus even for this calculation we have a-high securitymargin, this time of no less than 5 X 105 times.
The method described has been patented. The de-velopment project has been worked out at the labora-tories of Lasergruppen-Holovision AB, sponsored by themain suppliers of ID cards in Sweden: AB ID-Kort,owner of the patent.' 7
The newspaper Dagens Nyheter also sponsored thisproject.
References1. British Patent 37,512 (August 1968).2. U.S. Patent 3,620,590 (November 1971).3. U.S. Patent 3,668,795 (June 1972).4. U.S. Patent 3,647,275 (March 1972).5. U.S. Patent 818,711 (April 1969).6. U.S. Patent 299,294 (October 1972).7. U.S. Patent 48,373 (June 1970).8. S yiss Patent 451,571 (February 1968).9. U.S. Patent 3,483,513 (December 1969).
10. British Patent 1,188,302 (April 1970).11. N. Abramson, Appl. Opt. 13, 2019 (1974).12. N. Abramson, Appl. Opt. 14, 981 (1975).13. N. Abramson, Appl. Opt. 15, 200 (1976).14. N. Abramson, Appl. Opt. 16, 2521 (1977).15. N. Abramson, in Applications of Holography and Optical Data
Processing Proceedings of the International Conference, Jeru-salem 1976 (Pergamon, New York, 1977), p. 269.
