d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1325–1337

avai lab le at www.sc iencedi rec t .com

journa l homepage: www. int l .e lsev ierhea l th .com/ journa ls /dema

Ultrastructural characterization of tooth–biomaterialinterfaces prepared with broad and focused ion beams

E. Coutinhoa, T. Jarmarb, F. Svahnb, A.A. Nevesa, B. Verlindenc,B. Van Meerbeeka,∗, H. Engqvistb

a Leuven BIOMAT Research Cluster, Department of Conservative Dentistry, School of Dentistry, Oral Pathology and Maxillo-FacialSurgery, Catholic University of Leuven, Kapucijnenvoer 7, B-3000 Leuven, Belgiumb Department of Materials Science and Engineering, Uppsala University, Swedenc Department of Materials Science and Engineering, Catholic University of Leuven, Belgium

a r t i c l e i n f o

Article history:

Received 14 January 2009

Accepted 4 June 2009

Keywords:

FIB

BIB

Adhesion

Dental

Dentin

Enamel

TEM

Glass-ionomer

3-Step etch-and-rinse

2-Step self-etch

a b s t r a c t

Current available techniques for transmission electron microscopy (TEM) of tooth–

biomaterial interfaces are mostly ineffective for brittle phases and impair integrated chem-

ical and morphological characterization.

Objectives. The aims of this study were (1) to determine the applicability of new focused ion

beam (FIB) and broad ion beam (BIB) techniques for TEM preparation of tooth–biomaterial

interfaces; (2) to characterize the interfacial interaction with enamel and dentin of a con-

ventional glass-ionomer (Chemfil Superior, DeTrey Dentsply, Germany), a 2-step self-etch

(Clearfil SE, Kuraray, Japan) and a 3-step etch-and-rinse (OptiBond FL, Kerr, USA) adhesives;

and (3) to characterize clinically relevant interfaces obtained from actual Class-I cavities.

Methods. After bonding to freshly extracted human third molars, non-demineralized and

non-stained sections were obtained using the FIB/BIB techniques and examined under TEM.

Results. The main structures generally disclosed in conventional ultramicrotomy samples

were recognized in FIB/BIB-based ones. There were not any major differences between FIB

and BIB concerning the resulting ultrastructural morphology. FIB/BIB-sections enabled to

clearly resolve sub-micron hydroxyapatite crystals on top of hard tissues and the interface

between matrix and filler in all materials, even at nano-scale. Some investigated interfaces

disclosed areas with a distinct “fog” or “melted look”, which is probably an artifact due to

surface damage caused by the high-energy beam. Interfaces with enamel clearly disclosed

the distinct “keyhole” shape of enamel rods sectioned at 90◦, delimited by a thin electron-

lucent layer of inter-rod enamel. At regions where enamel crystals ran parallel with the

interface, we observed a lack of interaction and some de-bonding along with interfacial

void formation.

Significance. The FIB/BIB methods are viable and reliable alternatives to conventional ultrami-

crotomy for preparation of thin sections of brittle and thus difficult to cut biomaterial–hard

tissue interfaces. They disclose additional ultrastructural information about both substrates

and are more suitable for advanced analytic procedures.

© 2009 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.

∗ Corresponding author. Tel.: +32 16 33 75 87; fax: +32 16 33 27 52.E-mail address: bart.vanmeerbeek@med.kuleuven.be (B. Van Meerbeek).

0109-5641/$ – see front matter © 2009 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.doi:10.1016/j.dental.2009.06.002

s 2 5

after extraction. A conventional hand-mixed conventional

1326 d e n t a l m a t e r i a l

1. Introduction

Dentists spend most of their time replacing restorations thatmost often fail prematurely due to interfacial breakdown [1].Hence, a better understanding of the interfacial processesthat cause early degradation of the bond is essential to fur-ther improve adhesive technology [2–4]. Early studies suggestthat the durability of tooth bonding is directly related to thechemical interaction of biomaterial functional groups withinorganic substrate components [5–7]. Nevertheless, mostinvestigations designed to characterize dental interfaces arestill restricted to mainly a combination of mechanical testingand morphological characterization [8–10]. In order to investi-gate the actual intermolecular interactions at the interface, weneed more versatile ultrastructural imaging techniques com-bined with in situ chemical analysis of the biomaterial–hardtissue interface [11,12].

Furthermore, special attention should be devoted tolab-procedures that enable interface samples to be pre-pared that are ideally free from artifacts, or at least onlyentail a low artifact incidence. This means one shouldattempt to avoid using highly chemically reactive andthus potentially sample-damaging routines [13,14]. Unfor-tunately, such specimen-preparation protocols are todayindispensable to prepare ultrathin sections for transmissionelectron microscopy (TEM), which is the current ‘golden stan-dard’ methodology for characterization of dental interfaces[5,15,16]. Moreover, preparation of TEM-sections from hard,brittle and often water-sensitive tissues and/or materials, thatin addition are bonded together (as in case of biomaterial–hard

tissue interfaces), and that are sufficiently thin (70–90 nm)to allow the incident electron beam to pass through, is verychallenging [13,16]. Indeed, there is no easy way to section

Table 1 – Materials, composition and application directions foll

Product Compositiona

Chemfil Tooth Cleanser (DeTreyDentsply, Konstanz, Germany)

10% polyacrylic acid

Chemfil Superior (DeTreyDentsply, Konstanz, Germany)

Polyacrylic acid, water,aluminum-sodium-calcium-flphosphoro-silicate

Clearfil SE (Kuraray, Tokyo,Japan)

Primer: 10-MDP, HEMA, hydrodimethacrylate, photo-initiatBonding: 10-MDP, HEMA, bis-Ghydrophobic dimethacrylate,photo-initiators, silanated co

OptiBond FL (Kerr, Orange, CA,USA)

Etchant: 37.5% phosphoric acithickener. Primer: HEMA, GPDphoto-initiator, ethanol, wateTEGDMA, UDMA, GPDM, HEMfiller, photoinitiator

Gradia Direct (GC, Tokyo, Japan) Silica, prepolymerized filler,fluoro-alumino-silicate glass

a bis-GMA, bisphenol-glycidyl methacrylate; HEMA, 2-hydroxyethyl memonomer; GPDM, glycerol phosphate dimethacrylate; PAMM, phtalic acilate; UDMA, urethane dimethacrylate; 10-MDP, 10-methacryloyloxydecy

( 2 0 0 9 ) 1325–1337

with a diamond knife (ultramicrotomy) through the very hardglass particles of biomaterials such as dental composites orglass-ionomers, or through highly mineralized hard and brittlesubstrates like dentin and in particular enamel.

Recently, novel preparation techniques for cross-sectionTEM-sample preparation have evolved from the developmentof new precise methods to control and manipulate ion beams,i.e. focused-ion-beam (FIB) milling [17], and broad-ion-beam(BIB) milling [18]. As such, these techniques may potentiallyaddress some of the current concerns and limitations of thecurrent ultramicrotomy-based methodology. Therefore, theobjectives of this study were (1) to determine the feasibilityand applicability of the FIB and BIB technique for TEM prepa-ration of tooth–biomaterial interfaces; (2) to characterize, withminimal chemical sample preparation, the interfacial inter-action with enamel and dentin of a particularly challenging(from a sample-preparation perspective) conventional glass-ionomer, having as controls a 2-step self-etch and a 3-stepetch-and-rinse adhesives; and (3) to characterize such inter-faces when all materials were applied onto clinically relevanttooth surfaces (such as walls in actual Class-I cavities) ratherthan onto ‘model’ lab-made flat tooth surfaces.

2. Materials and methods

Human third molars were gathered after informed con-sent as approved by the Commission for Medical Ethicsof the Catholic University of Leuven, stored in 0.5% chlo-ramine/water solution at 4 ◦C and used within 1 month

owed in this study.

Application

Apply to enamel and dentin surfaces andleave undisturbed for 10 s; rinse withwater for 10 s; gently air-dry for 5 s,leaving a moist surface

uoro-Manually mix powder and liquid for 10 s;apply to enamel and dentin surfaces

philicor, water.MA,

lloidal silica

Apply primer and leave for 20 s; gentlyair-dry for 5 s; apply bonding agent; gentleair-blow; light-cure for 20 s

d, silicaM, PAMM,r. Bonding:A, bis-GMA,

Apply etchant for 15 s; rinse for 15 s;gently air-dry for 5 s; scrub surface for 15 swith primer; apply bonding agent; gentleair-blow; light-cure for 30 s

Apply in layers of 1 mm; light-cure eachlayer for 20 s; cure last layer for 40 s

thacrylate; DMA, dimethacrylate; phA-m, phosphoric acid esterd monoethyl methacrylate; TEGDMA, triethylene glycol dimethacry-l dihydrogenphosphate.

glass-ionomer cement (Chemfil Superior, DeTrey Dentsply,Konstanz, Germany), a 2-step self-etch adhesive (Clearfil SE,Kuraray, Tokyo, Japan) and a 3-step etch-and-rinse adhesive

5 ( 2

((wD

tLbitantGwc(

1tLdtmeStdTaei

2

Tslfittancvfcfcfitctp[ttit[

d e n t a l m a t e r i a l s 2

OptiBond FL, Kerr, Orange, CA, USA) were used in this studyTable 1). The glass-ionomer material was applied with and

ithout its respective conditioner (Chemfil Tooth Cleanser,eTrey Dentsply).

Standardized Class-I cavities were manually prepared, ini-ially with a high-speed regular-grit diamond bur (837, Komet,emgo, Germany) and further finished with a pear-shaped fineur (8830L, Komet, Lemgo, Germany), under constant water

rrigation. After sectioning their roots, the remaining dentinhickness was measured with a wax caliper in order to guar-ntee a minimum thickness around 1 mm. The materials wereext applied strictly according to the respective manufac-urer’s instructions (Table 1). A resin composite (Gradia Direct,C, Tokyo, Japan) was used for restoring the cavities treatedith the adhesive resin, in three to four increments up to the

avo-surface angle, with each increment polymerized for 20 sOptilux 500, Demetron/Kerr, Danbury, USA).

All restored teeth were then stored for 1 week at 37 ◦C and00% humidity, after which they were sectioned on their cen-ral portion into mesial and distal pieces (Isomet 1000, Buehler,ake Bluff, USA), fixed in 2.5% glutaraldehyde and step-wiseehydrated in ethanol and HMDS [14]. Both sections werehen metallographically polished with sandpapers and dia-

ond pastes with decreasing grit sizes, dehydrated in 100%thanol and gold-sputtered (Balzers Benchtop Sputter CoaterCD - 030, BAL-TEC, Balzers, Liechtenstein). In order to get tohis basic protocol, several pilot studies were done in order toetermine the most optimal sample-preparation technique.esting conditions involved different embedding proceduresnd materials (epon, epoxy resin, double embedding, and nombedding), different section orientations and different pol-shing strategies.

.1. FIB preparation of TEM samples

he FIB microscope relies on the interaction between sampleurface and heavy ions generated from an ion source (e.g. gal-ium), which are then accelerated by a high voltage electriceld and focused into a fine beam [17,19]. Upon impact withhe sample, the ions generate both secondary and backscat-ered ions and electrons, which can be collected to createn image of the surface, similarly to a conventional scan-ing electron microscope (SEM). The high-energy ion beaman be additionally used to precisely mill the specimen inery complex structures, as well as to cover it with dif-erent coatings. On dual-beam systems, the ion column isomplemented with an electron column mounted 52◦ apartrom each other, thus allowing SEM analysis of ion-milledross-sections. FIB was first used in the microelectronicseld for manipulating circuits, and was then adapted byhe materials science community for sub-micron site-specificross-sectional analysis of sensitive materials and to producehin, electron-transparent foils for TEM [17,20,21]. TEM sam-les can be either produced with ex situ or in situ methods

19]. In the ex situ method, a section is thinned to electronransparency, separated from the main piece, removed from

he vacuum chamber and transferred to a TEM grid. Then situ method requires a micromanipulator to transfer thehinned section to a TEM grid inside the vacuum chamber22].

0 0 9 ) 1325–1337 1327

Thin 100-nm sections between dentin/enamel and thethree different dental biomaterials were prepared in a FEIStrata DB235 dual-beam SEM/FIB (FEI, Eindhoven, The Nether-lands) with an ion accelerating voltage of 30 kV (Fig. 1). Thedevice was also equipped with an in situ lift-out system (Auto-probe 200, Omniprobe, Dallas, USA), which had a tungstenneedle attached to a micromanipulator inside the FIB vac-uum chamber. At first, the whole interface was examinedwith SEM/FIB and specific areas of interest identified. In orderto avoid milling artifacts in the TEM section, a thin layerof platinum was deposited on its surface by an electron-and ion-assisted chemical vapor deposition process (Fig. 1A).Increasingly deep trenches were then roughly milled in eachside of the protective coating, first using a relatively broadand later using a more focused ion beam (Fig. 1B). The result-ing section was extracted with the in situ lift-out technique,by partially separating it from the main body, soldering itwith platinum to the tungsten needle of the micromanipulator(Fig. 1C) and finally transferring it to a special TEM grid insidethe vacuum chamber. The TEM section was subsequently fur-ther thinned with decreasing ion-beam currents (Fig. 1D), andthereafter transferred to TEM (JEOL 2000FX, Tokyo, Japan oper-ating at 200 kV or to a 300 kV FEI Tecnai F30 ST TEM, FEICompany, Eindhoven, Netherlands) for further analysis. Forchemical analysis of the interfaces, energy dispersive X-rayline scans in STEM mode were performed. We took maximumprecautions to avoid beam damage inside the TEM columnby using a cryo-holder, where samples were kept at around−173 ◦C during all analyses (CT3500TR Single Tilt Rotation Liq-uid Nitrogen Cryo Transfer Holder, Gatan, Warrendale, USA).

2.2. BIB preparation of TEM samples

Thin 100-nm interfacial sections of the three investigatedmaterials with dentin/enamel were prepared in a JEOL IonSlicer (EM-09100 IS, JEOL, Tokyo, Japan), at an acceleratingvoltage of 1–8 kV. This device uses a combination of a broadargon-ion beam and a thin plate of shield material (Fig. 1E),where the specimen is irradiated at very low angles (Fig. 1F).In short, a thin-film specimen is prepared by taking advantageof the different sputtering rates of shielding plate and speci-men, which creates a central perforation on the target and athin, electron-transparent area around it (Fig. 1G and H). Therequired prior treatment consists of preparation of a 100-�mthin sample using mechanical polishing [18].

3. Results

Bright-field TEM photomicrographs from dentin and enamelinterfaces with the glass-ionomer cement (conditioned andnon-conditioned) and with the resin adhesives revealed pecu-liar characteristics (Figs. 2–7). In general, the main structuresdisclosed in samples prepared with an ultramicrotome couldalso be recognized on FIB/BIB-based ones, albeit some veryimportant distinctions. We have not observed any major dif-

ferences between the FIB and BIB method concerning theresulting ultrastructural morphology (Figs. 2–5).

When compared to common ultramicrotomy TEM-sections, the FIB method usually produced samples with

1328 d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1325–1337

Fig. 1 – Photomicrographs of the main steps involved in the FIB in situ lift-out (A–D) and the BIB (E-H) TEM preparationprocesses. The FIB requires the deposition of a protective platinum layer over the target area (A), followed by rough millingon both sides of the protective layer at higher beam energy (B). Sample manipulation is done through a tungsten nano-probe(C), after which the section is soldered to a special C-shaped TEM grid and carefully polished with a lower energy beam (D).The BIB method starts with a 0.1-mm thick sample obtained after mechanical polishing and mounted onto the specimenholder (E). A low energy, low angle argon-ion beam irradiates the specimen protected by a masking belt (F), generating thusa perforation on the targeted interface (G) and an electron-transparent thin area around it (H). E, enamel; Ar, adhesive resin.

d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1325–1337 1329

Fig. 2 – TEM photomicrographs of the interface of the conventional glass-ionomer bonded to dentin, prior (A–C) and after (Dand E) application of the polyalkenoic-acid conditioner. General observation of the FIB-prepared section, demonstrating thelimited field of view obtained (A). The interface does not show clear signs of dentin surface demineralization, while theintricate microstructure of dentin with the criss-cross hydroxyapatite organization and micro-pores are revealed by FIB (Band C). However, free hydroxyapatite crystals (black arrows) are clearly distinguishable at the interface after conditioning,and thus available for chemical interactions with the restorative material (D). Note that it is also possible to achievec all shc ed de

c15su

onsiderably higher magnifications, while maintaining overrystals (D and E). Gp, glass particle; M, matrix; Ud, unaffect

onsiderably smaller fields of view, measuring around

0 �m × 10 �m for FIB-TEM (Fig. 2A) versus at least0 �m × 100 �m for ultramicrotomy-TEM. They were alsolightly thicker (≥100 nm for FIB-TEM versus 70–90 nm forltramicrotomy-TEM). Some investigated interfaces also

arpness of all phases and particularly of hydroxyapatitentin.

disclosed areas with a distinct “fog” or “melted look”, which

is probably an artifact due to beam damage related to thepreparation technique with the high-energy ion beam (Fig. 2Band C). Such problem was very localized, particularly observedat FIB-sections of non-conditioned glass-ionomer interfaces

1330 d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1325–1337

Fig. 3 – HAADF STEM photomicrographs and EDS chemical analysis of the interface between dentin and the conventionalglass-ionomer, without prior application of the polyalkenoic-acid conditioner on the dentin surface. General (A) anddetailed (B) views of the interface using the STEM HAADF detector, where contrast is obtained through differences in atomicweight. Note the intact integrity of the siliceous hydrogel layer around the glass particle (Gp). The line indicates wherelinear EDS analysis of the interface was performed (B). EDS analysis demonstrating the relatively sharp transition betweendentin and the glass-ionomer, marked mainly by different concentrations in Ca and Al (C). Gp, glass particle; M, matrix; Ud,

unaffected dentin.

with dentin, and not present in all samples or sectionsprepared.

The main difference between non-conditioned (Fig. 2Band C) and conditioned (Fig. 2D and E) glass-ionomer/dentininterfaces was the presence of a 100- to 200-nm thick layerof hydroxyapatite crystals just on top of the unaffecteddentin (conditioned), clearly resolved even at nano-scale(Fig. 2D and E). Conversely, a sharper transition betweendentin and the glass-ionomer matrix was present on non-conditioned surfaces, both morphologically (Fig. 2B andC) and chemically (Fig. 3A–C). Such transition was alsoobserved with enamel interfaces (Fig. 4A–C). No other majorinterfacial features, such as a hybrid layer, gel-phase, absorp-tion layer, or even remnants of the smear layer, couldbe clearly distinguished. Bubble-shaped features could bedetected all across the interface and glass-ionomer matrix(Fig. 2B and C). Nonetheless, it is not clear if they were

originally part of the matrix, like the so-called ‘multiloc-ular phases’ [10], or product of beam damage. Althoughthere were always a few of them at the start of eachTEM session, both TEM main operators had the subjec-

tive impression that the quantity of bubbles seemed toincrease by the end of examination, despite the careful tech-nique employed. The interface of the glass-ionomer matrixwith the filler particles were though clearly seen, and thedistinct siliceous hydrogel layer was mostly present, evi-dencing the acid–base setting reaction of glass-ionomers(Figs. 2A, 3A, 4B and C). However, attempts to produce thinnersections disrupted the interfaces between filler and matrix,but not as prevalently as the common “broken” filler parti-cles found after ultramicrotomy of glass-ionomer interfaces(Fig. 4A).

Samples produced with the BIB method (Fig. 5A and B) alsohad a more limited field of view than the ultramicrotomy-based ones. Moreover, sample thickness was clearly notuniform and sample location along the interface was moredifficult to control than with the FIB method. Nevertheless,the morphology of the interface was clearly preserved and no

signs of beam damage were found. Therefore, BIB was par-ticularly effective for high-resolution imaging of nano-scalehydroxyapatite crystals at the interface (Fig. 5C), clearly dis-closing their periodic crystalline structure. Similarly to the

d e n t a l m a t e r i a l s 2 5 ( 2

Fig. 4 – TEM high-resolution photomicrographs of theinterface between the conventional glass-ionomer bondedto enamel after polyalkenoic-acid conditioning. Voids areartifacts of preparation and should not be considered foranalysis (A). Glass particles are in close contact with theinterface (B and C), and there are no clear signs ofdemineralization of enamel crystals despite good overalladaptation (A and B). Gp, glass particle; M, matrix; E,enamel.

0 0 9 ) 1325–1337 1331

FIB samples, the dentin structure itself exhibited a some-what peculiar contrast, which allowed a clear focus onpores and, in some cases, double-bands of collagen fibrils(Figs. 2B–E and 5A and B). As with the conventional glass-ionomer, the adhesive material was in intimate contact withthe underlying dental tissue, with no significant signs of de-bonding at the interface (Fig. 5A).

The main phases of the 3-step etch-and-rinse adhesiveinterfaces with dentin were clearly distinguishable (Fig. 6A–C),i.e. a dark ‘unaffected’ dentin and a filler-rich adhesive layer,set apart by an approximately 2-�m thin hybrid layer. Hardfiller particles and their interface with the matrix of the adhe-sive material could also be recognized distinctly (Fig. 6A–C).The hybrid layer had a very uniform appearance, typicallymore homogenous than ultramicrotomy-based, unstainedsections (Fig. 6B). A 100- to 200-nm thick layer of hydroxyap-atite crystals was present just on top of the unaffected dentin(Fig. 6B). Interfaces with enamel clearly disclosed the distinct“keyhole” shape of enamel rods sectioned at 90◦, delimitedby a thin electron-lucent layer of inter-rod enamel (Fig. 7A).Sections were also relatively thicker than the ones commonlyobtained after ultramicrotomy (Fig. 7A–D), but while withthe latter typically only little fragments of enamel remainedattached to the adhesive resin. As a consequence, the etch-pits were rather shallow in comparison, but resulted in clearlydefined micro-tags at rod sheaths, when enamel crystalswere cross-sectioned perpendicular to the bonding interface(Fig. 7C). At regions where enamel crystals ran parallel tothe interface, we observed a lack of interaction and somede-bonding through interfacial void formation (Fig. 7B–E). Dif-ferent crystal orientation within an enamel rod was alsodistinguishable, with most oriented parallel to the central axisof the rod (Fig. 7B). Enamel crystals located most remotely fromthe central rod axis presented a lateral ‘flare’ and intermingledwith crystals from adjacent rods (Fig. 7F).

4. Discussion

The use of high-resolution microscopy to investigate tooth–biomaterial interfaces has contributed much to the greatadvances dental adhesive technology has made in the lastfew decades [3,5]. Among the diverse microscopy techniquesemployed, TEM deserves a special consideration due to therichness in quantitative and qualitative data that can be gath-ered from the interface with a relatively low artifact incidenceand thus high reliability [16]. TEM is therefore the methodof choice of a growing number of research groups in thefield [10,23–25]. Traditionally, as in many other microscopydomains, TEM preparation and analysis procedures haveevolved in parallel with the needs for information withinthe two major research fields, Biology and Materials Science[12,16,19,26]. Considering the field of dentistry, it is not sur-prising that most dental interfaces are prepared and analyzedaccording to routine biological sample-processing procedures[13,16]. Moreover, routine preparation tools available in Mate-

rials Science (such as dimple grinding and ion milling) weretoo much prone to damage the delicate dentin or enamelinterfaces at the time most researchers were developing theirtechniques [15].

1332 d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1325–1337

Fig. 5 – TEM high-resolution photomicrographs of the interface of the 2-step self-etch adhesive bonded to dentin preparedwith BIB (A–C), disclosing similar morphology to those obtained with FIB. As expected, sample thickness is clearly notuniform (A and B), but thinner areas and possibly less beam damage allow higher magnifications to be achieved and clear

cted

imaging of individual hydroxyapatite crystals (C). Ud, unaffehydroxyapatite crystals.

Nevertheless, the current ultramicrotomy-based TEMsample-preparation technique is difficult (sometimes impos-sible) to apply onto hard and/or brittle ‘dental’ substrates,such as those involving the nearly purely inorganic toothenamel, conventional glass-ionomers or other water-baseddental cements, ceramic restorative materials and root posts,metallic and zirconium implant abutments and frameworks.

This is inherently due to the diamond-knife sectioning pro-cedure, which in such cases tends to provoke a controlledfracture, instead of the commonly obtained cleavage in resin-based samples [10,16,24]. Moreover, diamond-knife sectioning

dentin; Hy, hybrid layer; M, matrix; T, tubule; HAp,

also induces significant mechanical deformation of samples(less with a 35◦ knife), besides scratching of TEM specimens,either by small defects on the diamond blade or by looseningof hard filler particles [27].

Compared to the previously used methods, FIB allowssite-specific preparation by first imaging the region to besectioned within an SEM, and then precisely choosing the

actual site for the thin section with micrometer resolution[17,21,28]. Furthermore, interfaces can be investigated with-out any embedding techniques, which may reduce artifacts inboth conventional and analytical electron microscopy [11,22].

d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1325–1337 1333

Fig. 6 – TEM photomicrographs of the interface of the 3-step etch-and-rinse adhesive with dentin. A general view shows thelack of contrast of collagen and polymer-based phases, as well as intactness of filler in the resin matrix (A). Note thep t theh r; M

Ubntsicibsiatolmrdtesw[

l

resence of distinct hydroxyapatite crystals (black arrows) aybrid layer (B and C). Ud, unaffected dentin; Hy, hybrid laye

ltramicrotomy-sectioning without embedding has alreadyeen performed previously, but requires extremely high tech-ical skill and is very time-consuming [29,30]. It is also possibleo further thin a sample in the FIB after initial TEM analy-is, which permits corrections or improvements to be maden highly interesting samples (especially considering furtherhemical analysis). The sample-preparation process can bemaged as it progresses and bad or unsuitable samples cane aborted before going to the TEM, which saves both per-onnel and instrument time. In this regard, the BIB devices considerably simpler to use, working mainly unassisted,fter initial setup, for a few hours, in addition to its ini-ial lower cost [18]. Nevertheless, one important limitationf the BIB technique is the limited control of the precise

ocation for preparation of a TEM sample, since an electronicroscope is not attached to the device and the sample is

elatively unmovable inside the preparation chamber. The BIBevice we employed uses an argon-ion beam that is ‘softer’han the gallium-ion beam used with FIB and therefore isxpected to better preserve ultrastructural detail (as can beeen in Fig. 5). However, FIB devices can also be equipped

ith an argon-ion beam, achieving similar ‘milling’ accuracy

17,18,31].Since material removal is essentially done via gal-

ium/argon ions in both FIB/BIB methodologies, sample

interface just on top of the unaffected dentin within the, matrix; T, tubule; F, filler.

preparation is much less sensitive to the stiff- and brittle-ness of the materials and/or tissues to be cut, such as enamel,glass-ionomers and ceramics [11,22,28,32]. This can be clearlyobserved at the matrix–filler interfaces (Figs. 2–4), where theintimate contact was preserved and no filler particles werechipped out, as can happen with diamond-knife sectioning.Such feature is paramount for characterizing filler distributionand interactions between filler and matrix during materialapplication. Finally, interfaces obtained from actual cavity-preparation procedures can be investigated, which allows theassessment of substrate influence (e.g. dentin tubules, enamelprisms direction, carious dentin) and clinically significantvariables (e.g. C-factor) [33–35].

In our experience, samples obtained via ultramicrotomygenerally give a better morphologic description of the biologi-cal tissue. The presence of an amorphous layer on the surfaceof TEM sections obtained with a FIB has already been docu-mented extensively [17,31], and such surface-smoothing effecthas been reported before for adhesive–dentin interfaces whensectioned using FIB and imaged by SEM [36]. This surface layeris, as a rule of thumb, about 1/10 of the sample thickness, and

can be controlled and/or prevented using a low-energy beamin the final polishing operation. If higher energy beams areused, it is very likely that areas of the section that are moresusceptible to damage, will exhibit a ‘foggy’ or ‘melted’ appear-

1334 d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1325–1337

Fig. 7 – TEM photomicrographs of the unstained and non-demineralized interface between the 3-step etch-and-rinseadhesive and enamel. The distinct “keyhole” shape of enamel rods sectioned at 90◦ is clearly visible (A), delimited by a thinelectron-lucent layer of inter-rod enamel (white arrows). Black arrows (B) show an area of de-bonding at the interface withpresence of voids, contrasting with the etch-pits at enamel crystals perpendicular to the bonding interface at rod sheaths (C,hand-pointer). We can also observe the different crystal orientation within an enamel rod, where most crystals are oriented

eralrom

parallel to the central axis of the rod (B–E). A detail of the latcentral rod axis is shown in close interaction with crystals f

ance, as observed before with enamel and dentin samples[37,38]. The BIB method may, as mentioned above, poten-tially induce less surface amorphization due to the use ofan argon-ion beam instead of the heavier gallium ions com-monly employed in a conventional FIB [18]. Although we have

not observed such pronounced effect on this study, we haveindeed been able to produce very high-resolution photomi-crographs of the hydroxyapatite crystal at dentin interfaces(Fig. 5A–C).

‘flare’ of enamel crystals located most remotely from theadjacent rods (F). F, filler; M, matrix; Er, enamel rods.

The same reasoning applies to the moderate loss in theability of identifying details in some of the phases at theinterface, mainly observed for collagen in the hybrid layer.Although the image contrast may be further improved withpost-staining techniques, such as uranyl-acetate and lead

citrate [13], there are also significant advantages when non-stained samples are examined. As a matter of fact, it allowsvery clear imaging and characterization of hydroxyapatitecrystals in a very important area of the interface, i.e. just

5 ( 2

adybdcghohfitNhtm

ttm(imitsbatca

acf[taSmpiDdcpfm

fbsssfirnc

i

d e n t a l m a t e r i a l s 2

bove the unaffected dentin layer. As previously described,irect evidence of chemical interaction between hydrox-apatite crystals and the main functional constituents ofoth glass-ionomers and self-etch resin adhesives has beenocumented extensively [6,7,39]. FIB-sections disclose verylearly this first layer of hydroxyapatite crystals, thereforeiving the opportunity to properly investigate and analyzeydroxyapatite–monomer interactions with the multitudef techniques today available in a modern TEM, such asigh resolution (HREM), diffraction, high-angle annular darkeld scanning-transmission (HAADF STEM), energy loss spec-roscopy (EELS) and energy-filtered (EFTEM) modes [12,20,28].evertheless, since ion milling involves bombardment witheavy gallium ions, artifacts due to gallium contamina-ion should always be considered in analytical electron

icroscopy.Due to the complex interactions between the ion beam and

he sample, especially regarding different milling rates andhe different phases found in tooth tissues and restorative

aterials, it can be difficult to produce very thin, flat sectionsFigs. 2A, 4A and 7A). Bending of the TEM sections can eas-ly occur as a result of residual internal stresses. This can be

inimized by soldering both edges of the section to the gridn the in situ method [17]. The ion beam can also be slightlyilted in relation to the sample, thus provoking voids in theection around thin areas (Fig. 4A). For these reasons, ion-eam milling is therefore often aborted when the sample isbout 100-nm thick [22,32]. Although samples prepared withhe BIB equipment did not disclose voids at the interface, weould definitely observe their variable thickness during TEMnalysis (Fig. 5A).

Relative to ultramicrotomy, comparative productivity islso not favorable to the FIB technique. It is very time-onsuming to produce multiple samples, i.e. the time gainedor preparing multiple samples of the same interface is low19]. Furthermore, the field of view is considerably smallerhan those obtained with conventional techniques (Fig. 2A),lthough this affects only the number of regions examined.ince the obtained sample size is equivalent to the most usedesh grid size in ultramicrotomy-based samples (Fig. 2A), it is

ossible to conclude that the FIB technique does not necessar-ly affect the overview of the interface at lower magnifications.ual-beam equipments are also extremely expensive andemands skilled personnel for proper operation, particularlyoncerning the interactions of electrons and ions with sam-les. As milling is performed under vacuum, there is a needor proper specimen preparation prior to placement in the FIB

icroscope [17].In accordance with the literature, the in situ lift-out was

ound to be more time consuming than the ex situ methodut with a higher rate of success [17,28]. The higher rate ofuccess is partly due to the eliminated risk of losing the TEMection during the transfer with the glass needle. The rate ofuccess was also higher since it was easier to perform thenal thinning when the TEM section was separated from theest of the sample and mounted on the TEM grid, which sig-

ificantly decreases charging effects and avoids the need forarbon coating [17].

Although glass-ionomers exhibited several problems whennvestigated with the FIB technique, we consider it a very

0 0 9 ) 1325–1337 1335

worthwhile pursuit. Compared to other bonding strategies,glass-ionomers show the most reliable and durable clini-cal retention rates [40,41], which is commonly attributed totheir chemical interaction with hydroxyapatite [7,39]. A betterunderstanding of the chemical interaction at these interfaceswould allow the development of more efficient restorativematerials. However, few studies have been able to characterizeits interface with tooth tissues – particularly with enamel – asconventional glass-ionomers are very difficult to prepare forTEM analysis. Hence, most ultrastructural investigations arelimited to their resin-modified counterparts or to their inter-face with the dentin substrate, which is mainly due to theirhigh brittleness and high water sensitivity [8,10,42,43]. Thisjustifies our particular choice of a conventional glass-ionomerto evaluate the applicability of FIB, as well as indicate thedirection of future research.

Considering the investigated 3-step etch-and-rinse system,the observed hybrid layers were considerably thinner thanreported in the literature [29,30]. It should be noticed thoughthat due to the limitations of the FIB technique, as men-tioned before, a much smaller portion of the tooth–biomaterialinterface has been thoroughly examined. Such drawback isoffset by the much more precise orientation of sections inrelation to samples produced using conventional ultrami-crotomy preparation methods. Section mis-orientation maynaturally strongly influence quantitative measurements at theinterface, such as the hybrid layer thickness [16]. As the FIBtechnique is much more precise to position and control sectioncharacteristics, the commonly accepted values in literaturemay need to be revisited in the near future in studies withlarger sample sizes.

The FIB method was particularly successful for analysis ofenamel interfaces, disclosing features that are generally lostduring ultramicrotomy sectioning (Figs. 4A–C and 7A–F). Bycarefully controlling the position of the sectioning, we havebeen able to prepare TEM samples perpendicular to the intri-cate 3D rod pattern of enamel (Fig. 7A). Enamel tissue has beenwidely recognized as extremely difficult to study, due to itshighly mineralized nature, the abstruse 2D projection at vary-ing section angles of the 3D rod pattern, and the small size ofthe basic constituents [44]. Such problems are obviously aggra-vated when interfaces with biomaterials are studied [15,16].We also observed a dramatic effect of crystal orientation onthe bonding effectiveness to enamel (Fig. 7B–D). Crystals ori-ented parallel to the interface were not etched as easily asperpendicular ones, and tended to be bonded to less optimally.The high-energy ion beam may have then induced voids inthese areas of internal/residual stress concentration [17,19].Rod sheaths also contain more enamel proteins due to theless than optimal packing of enamel crystals that are orientedin different directions [44]. As such, etch-pits should be moreeasily found in these regions, once the organic part is removedby acid conditioning and infiltrated with adhesive resins [5],thus providing a more stable bond substrate [33].

5. Conclusions

Harder materials or phases, such as hydroxyapatite crystalsand filler in restorative materials, do above all benefit from

s 2 5

r

1336 d e n t a l m a t e r i a l

the FIB/BIB methods for TEM preparation, enabling higher-resolution morphologic and chemical analysis. Collagen andpolymeric materials are nevertheless more sensitive to beamdamage and exhibit a lower phase contrast, requiring thusmore care during staining and examination. In addition, theseveral technical demands of the FIB/BIB methods may impairtheir broader application as routine procedures.

Interfacial samples of the adhesive resins with tooth tis-sues clearly disclosed their particular anisotropic morphology.In agreement with previous micro-mechanical studies, wehave been able to show zones of preferential bonding withinthe complex pattern of enamel rods at the interface. Moreover,although more sensitive and prone to artifacts, FIB/BIB showa lot of potential for future research of conventional glass-ionomer interfaces, as more advanced analytical techniquescan be applied to the critical layer of hydroxyapatite crystalsin direct contact with polyalkenoic acids.

The FIB/BIB techniques are thus suitable prepara-tion methodologies of tooth–biomaterial interfaces forTEM and nicely complement information obtained fromultramicrotomy-based techniques. They excel particularly formore advanced TEM analysis, but not without compromises.Fortunately, most of them can be controlled through judiciouspreparation and analysis.

Acknowledgments

This study was supported by the KULeuven (OT/06/55), theResearch Foundation – Flanders (FWO G.0206.07), the ToshioNakao Chair for Adhesive Dentistry, the Göran GustafssonFoundation for Academic Research (Uppsala University), andCAPES Foundation. We would also like to thank DominiqueCrombez and Luk Boon for extensive technical support, as wellas, Dentsply and Kerr for the generous donation of materials.

e f e r e n c e s

[1] Mjor IA, Gordan VV. Failure, repair, refurbishing andlongevity of restorations. Oper Dent 2002;27(5):528–34.

[2] Tay FR, Pashley DH. Dentin bonding—is there a future? JAdhes Dent 2004;6(4):263.

[3] Tay FR, Pashley DH. Dental adhesives of the future. J AdhesDent 2002;4(2):91–103.

[4] Pocius AV. Adhesion and adhesives technology: anintroduction. 1st ed. Cincinnati: Hanser Gardner; 1996.

[5] Van Meerbeek B, De Munck J, Yoshida Y, Inoue S, Vargas M,Vijay P, et al. Buonocore memorial lecture. Adhesion toenamel and dentin: current status and future challenges.Oper Dent 2003;28(3):215–35.

[6] Yoshida Y, Nagakane K, Fukuda R, Nakayama Y, Okazaki M,Shintani H, et al. Comparative study on adhesiveperformance of functional monomers. J Dent Res2004;83(6):454–8.

[7] Yoshida Y, Van Meerbeek B, Nakayama Y, Snauwaert J,Hellemans L, Lambrechts P, et al. Evidence of chemicalbonding at biomaterial–hard tissue interfaces. J Dent Res

2000;79(2):709–14.

[8] Coutinho E, Van Landuyt K, De Munck J, Poitevin A, YoshidaY, Inoue S, et al. Development of a self-etch adhesive forresin-modified glass ionomers. J Dent Res 2006;85(4):349–53.

( 2 0 0 9 ) 1325–1337

[9] Van Landuyt KL, Kanumilli P, De Munck J, Peumans M,Lambrechts P, Van Meerbeek B. Bond strength of a mildself-etch adhesive with and without prior acid-etching. JDent 2006;34(1):77–85.

[10] Tay FR, Smales RJ, Ngo H, Wei SH, Pashley DH. Effect ofdifferent conditioning protocols on adhesion of a GIC todentin. J Adhes Dent 2001;3(2):153–67.

[11] Engqvist H, Schultz-Walz JE, Loof J, Botton GA, Mayer D,Phaneuf MW, et al. Chemical and biological integration of amouldable bioactive ceramic material capable of formingapatite in vivo in teeth. Biomaterials 2004;25(14):2781–7.

[12] Botton GA, Phaneuf MW. Imaging, spectroscopy andspectroscopic imaging with an energy filtered field emissionTEM. Micron 1999;30(2):109–19.

[13] Van Meerbeek B, Yoshida Y, Lambrechts P, Vanherle G, DukeES, Eick JD, et al. A TEM study of two water-based adhesivesystems bonded to dry and wet dentin. J Dent Res1998;77(1):50–9.

[14] Perdigao J, Lambrechts P, Van Meerbeek B, Vanherle G, LopesAL. Field emission SEM comparison of four postfixationdrying techniques for human dentin. J Biomed Mater Res1995;29(9):1111–20.

[15] Nakabayashi N, Pashley DH. Hybridization of dental hardtissues. 1st ed. Tokyo: Quintessence; 1998.

[16] Van Meerbeek B, Vargas M, Inoue S, Yoshida Y, Perdigao J,Lambrechts P, et al. Microscopy investigations. Techniques,results, limitations. Am J Dent 2000;13(Spec No):3D–18D.

[17] Giannuzzi LA, Stevie FA. Introduction to focused ion beams:instrumentation, theory, techniques and practice. 1st ed.New York: Springer; 2004.

[18] Yasuhara A. Development of ion slicer (Thin-Film SpecimenPreparation Equipment). Jeol News 2005;40(1):46–9.

[19] Phaneuf MW. Applications of focused ion beam microscopyto materials science specimens. Micron 1999;30(3):277–88.

[20] Engqvist H, Botton GA, Couillard M, Mohammadi S,Malmstrom J, Emanuelsson L, et al. A novel tool forhigh-resolution transmission electron microscopy of intactinterfaces between bone and metallic implants. J BiomedMater Res A 2006;78(1):20–4.

[21] Engqvist H, Svahn F, Jarmar T, Detsch R, Mayr H, Thomsen P,et al. A novel method for producing electron transparentfilms of interfaces between cells and biomaterials. J MaterSci Mater Med 2008;19(1):467–70.

[22] Giannuzzi LA, Phifer D, Giannuzzi NJ, Capuano MJ.Two-dimensional and 3-dimensional analysis of bone/dentalimplant interfaces with the use of focused ion beam andelectron microscopy. J Oral Maxillofac Surg 2007;65(4):737–47.

[23] Koshiro K, Sidhu SK, Inoue S, Ikeda T, Sano H. New conceptof resin–dentin interfacial adhesion: the nanointeractionzone. J Biomed Mater Res B: Appl Biomater 2006;77(2):401–8.

[24] Van Meerbeek B, Conn Jr LJ, Duke ES, Eick JD, Robinson SJ,Guerrero D. Correlative transmission electron microscopyexamination of nondemineralized and demineralizedresin–dentin interfaces formed by two dentin adhesivesystems. J Dent Res 1996;75(3):879–88.

[25] Reis AF, Giannini M, Pereira PN. Long-term TEM analysis ofthe nanoleakage patterns in resin–dentin interfacesproduced by different bonding strategies. Dent Mater2007;23(9):1164–72.

[26] Williams DB, Carter CB. Transmission electron microscopy: atextbook for materials science. 1st ed. New York: Plenum;1996.

[27] Jesior JC. Use of low-angle diamond knives leads to

improved ultrastructural preservation of ultrathin sections.Scanning Microsc Suppl 1989;3:147–52 [discussion 52–3].

[28] Jarmar T, Palmquist A, Branemark R, Hermansson L,Engqvist H, Thomsen P. Technique for preparation and

5 ( 2

d e n t a l m a t e r i a l s 2

characterization in cross-section of oral titanium implantsurfaces using focused ion beam and transmission electronmicroscopy. J Biomed Mater Res A 2008;87(4):1003–9.

[29] Nakabayashi N, Ashizawa M, Nakamura M. Identification ofa resin–dentin hybrid layer in vital human dentin created invivo: durable bonding to vital dentin. Quintessence Int1992;23(2):135–41.

[30] Van Meerbeek B, Eick JD, Robinson SJ. Epoxy-embeddedversus nonembedded TEM examination of the resin–dentininterface. J Biomed Mater Res 1997;35(2):191–7.

[31] McCaffrey JP, Phaneuf MW, Madsen LD. Surface damageformation during ion-beam thinning of samples fortransmission electron microscopy. Ultramicroscopy2001;87(3):97–104.

[32] Jarmar T, Palmquist A, Branemark R, Hermansson L,Engqvist H, Thomsen P. Characterization of the surfaceproperties of commercially available dental implants usingscanning electron microscopy, focused ion beam, andhigh-resolution transmission electron microscopy. ClinImplant Dent Relat Res 2008;10(1):11–22.

[33] Carvalho RM, Santiago SL, Fernandes CA, Suh BI, PashleyDH. Effects of prism orientation on tensile strength ofenamel. J Adhes Dent 2000;2(4):251–7.

[34] Shirai K, De Munck J, Yoshida Y, Inoue S, Lambrechts P,Suzuki K, et al. Effect of cavity configuration and aging on

the bonding effectiveness of six adhesives to dentin. DentMater 2005;21(2):110–24.

[35] Ogata M, Okuda M, Nakajima M, Pereira PN, Sano H, TagamiJ. Influence of the direction of tubules on bond strength todentin. Oper Dent 2001;26(1):27–35.

0 0 9 ) 1325–1337 1337

[36] Van Meerbeek B, Conn Jr LJ, Duke ES, Schraub D, GhafghaichiF. Demonstration of a focused ion-beam cross-sectioningtechnique for ultrastructural examination of resin–dentininterfaces. Dent Mater 1995;11(2):87–92.

[37] Hayashi Y, Yaguchi T, Ito K, Kamino T. High-resolutionelectron microscopy of human enamel sections preparedwith focused ion beam system. Scanning 1998;20(3):234–5.

[38] Hoshi K, Ejiri S, Probst W, Seybold V, Kamino T, Yaguchi T, etal. Observation of human dentine by focused ion beam andenergy-filtering transmission electron microscopy. J Microsc2001;201(Pt 1):44–9.

[39] Coutinho E, Yoshida Y, Inoue S, Fukuda R, Snauwaert J,Nakayama Y, et al. Gel phase formation at resin-modifiedglass-ionomer/tooth interfaces. J Dent Res 2007;86(7):656–61.

[40] Peumans M, Kanumilli P, De Munck J, Van Landuyt K,Lambrechts P, Van Meerbeek B. Clinical effectiveness ofcontemporary adhesives: a systematic review of currentclinical trials. Dent Mater 2005;21(9):864–81.

[41] van Dijken JW, Pallesen U. Long-term dentin retention ofetch-and-rinse and self-etch adhesives and a resin-modifiedglass ionomer cement in non-carious cervical lesions. DentMater 2008;24(7):915–22.

[42] Yiu CK, Tay FR, King NM, Pashley DH, Carvalho RM, CarrilhoMR. Interaction of resin-modified glass-ionomer cementswith moist dentine. J Dent 2004;32(7):521–30.

[43] Yiu CK, Tay FR, King NM, Pashley DH, Sidhu SK, Neo JC, et al.Interaction of glass-ionomer cements with moist dentin. JDent Res 2004;83(4):283–9.

[44] Ten Cate AR. Oral histology: development, structure andfunction. 5th ed. St. Louis: Mosby; 1998.